filename
stringlengths 17
38
⌀ | gt_das
stringlengths 41
7.52k
⌀ | genshare_das
stringlengths 10
46.3k
⌀ | md_content
stringlengths 0
359k
| healing_error
stringclasses 4
values |
|---|---|---|---|---|
10.1111_sapm.12544.pdf
|
D A T A AVA I L A B I L I T Y S T A T E M E N T
Data sharing not applicable to this article as no datasets were generated or analyzed during the
current study.
| null |
Received: 15 June 2022
Revised: 31 October 2022
Accepted: 2 November 2022
DOI: 10.1111/sapm.12544
O R I G I N A L A R T I C L E
Sobolev-orthogonal systems with tridiagonal
skew-Hermitian differentiation matrices
Arieh Iserles1
Marcus Webb2
1Department of Applied Mathematics and
Theoretical Physics, University of
Cambridge, Wilberforce Road,
Cambridge, UK
2Department of Mathematics, University
of Manchester, Manchester, UK
Correspondence
Marcus Webb, Department of
Mathematics, University of Manchester,
Alan Turing Building, Manchester M13
9PL, UK.
Email: [email protected]
Funding information
Narodowe Centrum Nauki; Simons
Foundation
Abstract
We introduce and develop a theory of orthogonality
with respect to Sobolev inner products on the real line
for sequences of functions with a tridiagonal, skew-
Hermitian differentiation matrix. While a theory of
such L2 -orthogonal systems is well established, Sobolev
orthogonality requires new concepts and their analysis.
We characterize such systems completely as appropri-
ately weighted Fourier transforms of orthogonal poly-
nomials and present a number of illustrative examples,
inclusive of a Sobolev-orthogonal system whose leading
𝑁 coefficients can be computed in (𝑁 log 𝑁) opera-
tions.
K E Y W O R D S
Malmquist–Takenaka functions, orthogonal systems, Sobolev
orthogonality, spectral methods
J E L C L A S S I F I C A T I O N
42C05, 42C10, 42C30, 65M12, 65M70
1
INTRODUCTION
1.1
Orthonormal systems on the real line
The theory of L2-orthonormal systems on the real line with a tridiagonal differentiation matrix has
been developed in Refs. 1–4. In its simplest (real) version, let 𝑤 ≥ 0 be an absolutely continuous,
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and
reproduction in any medium, provided the original work is properly cited.
© 2022 The Authors. Studies in Applied Mathematics published by Wiley Periodicals LLC.
Stud Appl Math. 2022;1–28.
wileyonlinelibrary.com/journal/sapm
1
2
ISERLES and WEBB
nonzero weight function, whose support is symmetric with respect to the origin, and {𝑝𝑛}𝑛∈ℤ+ the
underlying system of orthonormal polynomials, which must satisfy
𝑏𝑛𝑝𝑛+1(𝜉) = 𝜉𝑝𝑛(𝜉) − 𝑏𝑛−1𝑝𝑛−1(𝜉),
𝑛 ∈ ℤ+.
for some real numbers {𝑏𝑛}𝑛∈ℤ+. Setting
𝜑𝑛(𝑥) =
i𝑛
√
2𝜋
∞
∫
−∞
√
𝑝𝑛(𝜉)
𝑤(𝜉)ei𝑥𝜉d𝜉,
𝑥 ∈ ℝ,
𝑛 ∈ ℤ+,
(1)
(2)
we obtain by Parseval’s theorem1 an orthonormal system of functions in L2(ℝ). Moreover, under
the mild assumption that polynomials are dense in L2(ℝ; 𝑤), this system is dense in L2(ℝ) if the
support of 𝑤 is all of ℝ; otherwise, its closure is the Paley–Wiener space
supp 𝑤(ℝ) of all L2(ℝ)
functions whose Fourier transform is supported on supp 𝑤. Moreover, Φ = {𝜑𝑛}𝑛∈ℤ+ obeys
𝜑′
𝑛(𝑥) = −𝑏𝑛−1𝜑𝑛−1(𝑥) + 𝑏𝑛𝜑𝑛+1(𝑥),
𝑛 ∈ ℤ+.
(3)
In vector form, Equation (3) is 𝝋′ = 𝝋, where is the differentiation matrix of the system, which
in this case is tridiagonal and skew-symmetric. Skew symmetry and the tridiagonal form provide
important advantages on the design of spectral methods with the basis Φ.1
In this paper, we generalize the theory to the case of Sobolev-orthogonal systems, where the
Sobolev inner product is of the form
⟨𝜑, 𝜓⟩
𝑣 =
∞∑
𝓁=0
𝑣𝓁 ∫
∞
−∞
𝜑(𝓁)(𝑥)𝜓(𝓁)(𝑥) d𝑥,
(4)
defined by the nonzero, nonnegative sequence {𝑣𝓁}𝓁∈ℤ+ ⊂ [0, ∞). The 𝐇𝑠
corresponds to 𝑣𝓁 = 1 for 𝓁 = 0, 1 … , 𝑠 and 𝑣𝓁 = 0 otherwise.
2(ℝ) norm, where 𝑠 ∈ ℤ+
Besides the resulting theory being of interest in its own right, we can motivate our exploration
in the context of spectral methods for PDEs using the example of the Ornstein–Uhlenbeck process,
𝜕𝑢
𝜕𝑡
=
𝜕2𝑢
𝜕𝑥2
− 𝑎
𝜕
𝜕𝑥
(𝑥𝑢),
𝑥 ∈ ℝ,
𝑡 ≥ 0,
(5)
with coefficient of friction described by the positive constant 𝑎.5,6 Solutions to this PDE satisfy
d
d𝑡 ∫
∞
−∞
[𝑢2
𝑥(𝑥) + 𝑢2(𝑥)] d𝑥 = − ∫
∞
−∞
[2𝑢2
𝑥𝑥(𝑥) + (2 + 3𝑎)𝑢2
𝑥(𝑥) + 𝑎𝑢2(𝑥)]d𝑥,
(6)
which shows that the solution decays monotonically to zero in the 𝐇1
⟨𝑢, 𝑢⟩
drop some terms and show that
exponentially with rate dependent on 𝑎.
𝐻1 ≤ −𝑎⟨𝑢, 𝑢⟩
2(ℝ) norm. In fact, we can
𝐻1 , and hence the norm decreases at least
d𝑡
d
∑𝑁
Now, consider semidiscretizing equation (5) in space by a spectral method 𝑢(𝑥, 𝑡) ≈ 𝑢𝑁(𝑥, 𝑡) ∶=
𝑛=0 𝑎𝑛(𝑡)𝜑𝑛(𝑥), where Φ = {𝜑𝑛}𝑛∈ℤ+ ⊂ L2(ℝ) are orthonormal with respect to the 𝐇1
2(ℝ) inner
1 Also known as Plancharel’s theorem.
ISERLES and WEBB
3
product. If a Galerkin scheme is used with respect to the 𝐇1
2(ℝ) inner product (i.e., the residual
of the PDE at each time 𝑡 is orthogonal to span{𝜑𝑛}𝑁
𝑛=0), then the inequality (6) is also satisfied
by 𝑢𝑁 (cf. Ref. 7(Chapter 8)). It therefore follows that any A-stable discretization in time will be
stable.
The plan of this paper is as follows. In Section 2, basing ourselves upon our earlier theory on L2
inner products, we present a complete framework for the construction of Sobolev-orthogonal sys-
tems on the real line with a tridiagonal differentiation matrix. This leads to two alternatives toward
the construction of 𝐇𝑠
2(ℝ)-orthogonal systems, which are debated in Section 3: the first is the
arguably more obvious approach, yet it leads to formulæ that typically are impossible to express
explicitly, whereas the second, less natural, results in a more constructive approach. Section 4 is
concerned with systems based upon the familiar Hermite weight and Section 5 with bilateral (i.e.,
symmetrized with respect to the origin) Laguerre weights. In Section 6, we discuss Bessel-like
orthogonal systems originating in various ultraspherical weights: in that case, the closure of the
orthogonal system is not 𝐇𝑠
2(ℝ) but a relevant Paley–Wiener space. Section 7 generalizes the dis-
course to nonsymmetric measures. In that instance, our orthogonal systems are complex-valued
but the approach confers some important advantages. In particular, it allows us to generalize the
Malmquist–Takenaka system to Sobolev setting while retaining the most welcome feature of this
system, namely, that the coefficients can be computed rapidly with fast Fourier transform. Finally,
in Section 8, we present brief conclusions.
1.2
Sobolev norms beyond this paper
As an aside, our original interest in orthonormal systems (2) has been motivated in Ref. 1 by the
numerical solution of the linear Schrödinger equation in the semiclassical regime,
i𝜀
𝜕𝑢
𝜕𝑡
= −𝜀2 𝜕2𝑢
𝜕𝑥2
+ 𝑉(𝑥)𝑢,
𝑥 ∈ ℝ,
𝑡 ≥ 0,
(7)
given with an initial condition at 𝑡 = 0, 𝑥 ∈ ℝ. Here 0 < 𝜀 ≪ 1, while the interaction potential 𝑉
is real. The solution of this equation conserves the standard L2 norm (which motivates the use of
L2-orthogonal systems), but it also has another important invariant: its Hamiltonian,
∞
𝐻(𝑢) = ∫
−∞
[𝜀|𝑢𝑥(𝑥)|2 + 𝜀−1𝑉(𝑥)|𝑢(𝑥)|2]d𝑥,
(8)
is conserved. This might be viewed as a conservation of a nonstandard Sobolev norm (if 𝑉 is
positive). While the design of Hamiltonian methods for the Schrödinger equation is still an open
problem, it motivates the work reported in this paper.
We mention in passing another example in which nonstandard Sobolev norms are nonincreas-
ing, the diffusion equation,
𝜕𝑢
𝜕𝑡
=
𝜕
𝜕𝑥
[
𝑎(𝑥)
]
,
𝜕𝑢
𝜕𝑥
𝑥 ∈ ℝ,
𝑡 ≥ 0,
(9)
where 𝑎(𝑥) > 𝑎min > 0 for all 𝑥 ∈ ℝ, given with an initial condition for 𝑡 = 0, 𝑥 ∈ ℝ. It is
readily shown that the norm induced by the following nonstandard Sobolev inner product is
4
ISERLES and WEBB
nonincreasing as a function of time,
∞
⟨𝑢, 𝑢⟩
𝑎 ∶= ∫
−∞
[𝑎(𝑥)𝑢2
𝑥(𝑥) + 𝑢2(𝑥)]d𝑥.
(10)
We do not pursue these general Sobolev inner products in this paper, but anticipate reporting
on such results in the future.
1.3
Related work
Sobolev orthogonality: Polynomials orthogonal with respect to Sobolev inner products associated
with a vector of measures supported on the real line have been considered for a long while, but
the subject received considerable impetus with the introduction of coherent pairs in Ref. 8 and
has been surveyed in Refs. 9, 10. Natural questions, given the constructs (2) and (3) are, first, how
to generate Sobolev-orthogonal systems on the real line and, second, is a Fourier integral of an
orthogonal polynomial system scaled by any reasonable function orthogonal with respect to some
inner product, whether in a classical or Sobolev sense, in line with the L2 theory as briefly reviewed
in Section 2. These related questions are the focus of this paper. Intriguingly, as things stand, the
theory in this paper is heavily based on the theory of classical orthogonal polynomials (as distinct
from Sobolev-orthogonal polynomials). Fourier–Bessel functions11,12: Given a Borel measure d𝜇
and the underlying orthonormal system {𝑝𝑛}𝑛∈ℤ+, we define
𝜑𝑛(𝑥) = ∫
∞
−∞
𝑝𝑛(𝜉)e−i𝑥𝜉d𝜇(𝜉)
(11)
√
as the 𝑛th Fourier–Bessel function: the name is motivated by the Legendre measure d𝜇(𝑥) =
(𝑥). Note the similarity between (2) and (11) (disre-
𝜒(−1,1)(𝑥)d𝑥, whereby 𝜑𝑛(𝑥) =
garding the normalizing factor and the sign in the exponential, neither of which is of much
importance), namely, that both are Fourier transforms of 𝑝𝑛 with added scaling function:
𝑤
in the first instance and 𝑤 in the second.
2𝜋∕𝑥J
√
𝑛+
1
2
Further variation on this theme is the identity
1
∫
−1
T𝑛(𝜉)ei𝑥𝜉
√
d𝜉
1 − 𝜉2
= 𝜋i𝑛J𝑛(𝑥),
𝑛 ∈ ℤ+,
(12)
where T𝑛 is the 𝑛th Chebyshev polynomial of the first kind.11 Note that, unlike (2), Fourier–
Bessel functions need not be orthogonal although, interestingly enough, disregarding signs
and normalizing constants, the two formulæ coincide (and orthogonality is recovered) for the
Legendre measure.
1.4
Brief comments
The name of Charles Hermite is associated with two distinct concepts in this paper: skew-
Hermitian matrices and Hermite polynomials. They are, of course, completely different and
should not be confused.
ISERLES and WEBB
5
Our notation deserves a comment. Thus, we let 𝐇𝑠
2, where 𝑠 ≥ 0, stand for the usual Sobolev
space, equipped with the inner product
⟨𝑓, 𝑔⟩ =
𝑠∑
𝑘=0
𝑣𝑘 ∫ 𝑓(𝑘)(𝜉)𝑔(𝑘)(𝑥)d𝑥,
(13)
where the 𝑣𝑘s are nonnegative and 𝑣0 > 0. With greater generality, it is often helpful to denote
𝐇2,𝑣(ℝ) ∶= {𝜓 ∈ L2(ℝ) ∶ ⟨𝜓, 𝜓⟩
𝑣 < ∞},
(14)
whenever ⟨ ⋅ , ⋅ ⟩
a function 𝑣.
𝑣 is an inner product defined (in a sense that is always clear from the context) by
2
CHARACTERIZATION OF SOBOLEV-ORTHOGONAL SYSTEMS
Let us first state the desiderata. We are interested in functions Φ = {𝜑𝑛}𝑛∈ℤ+ ⊂ L2(ℝ) such that
both of the following properties hold.
A. There exists sequences {𝑏𝑛}𝑛∈ℤ+ ⊂ ℂ ⧵ {0} and {𝑐𝑛}𝑛∈ℤ+ ⊂ ℝ such that
𝜑′
𝑛(𝑥) = −𝑏𝑛−1𝜑𝑛−1(𝑥) + i𝑐𝑛𝜑𝑛(𝑥) + 𝑏𝑛𝜑𝑛+1(𝑥)
for 𝑛 = 0, 1, … (with 𝑏−1 = 0 by convention).
B. Φ is an orthonormal sequence with respect to the Sobolev inner product
⟨𝜑, 𝜓⟩
𝑣 =
∞∑
𝓁=0
∞
𝑣𝓁 ∫
−∞
𝜑(𝓁)(𝑥)𝜓(𝓁)(𝑥) d𝑥,
(15)
(16)
defined by the nonzero, nonnegative sequence {𝑣𝓁}𝓁∈ℤ+ ⊂ [0, ∞) such that
other words, at least one 𝑣𝓁 must be positive.)
∑∞
𝓁=0 𝑣𝓁 > 0. (In
Theorem 1 Refs. 1 and 2. A sequence Φ = {𝜑𝑛}𝑛∈ℤ+ ⊂ L2(ℝ) satisfies criterion (𝐴) if and only if
𝜑𝑛(𝑥) =
ei𝜃𝑛
√
2𝜋
∞
∫
−∞
ei𝑥𝜉𝑝𝑛(𝜉)𝑔(𝜉) d𝜉,
(17)
where
∙ 𝑃 = {𝑝𝑛}𝑛∈ℤ+ is an orthonormal polynomial system on the real line with respect to a probability
measure on the real line with all moments finite and with infinitely many points of increase;
∙ Θ = {𝜃𝑛}𝑛∈ℤ+ ⊂ [0, 2𝜋);
∙ 𝑔 ∈ L2(ℝ) satisfies lim𝜉→±∞
|𝜉𝑘𝑔(𝜉)| = 0 for 𝑘 = 0, 1, 2, …. We call such functions mollifiers.
Remark 1. It is possible to ensure that the parameters {𝑏𝑛}𝑛∈ℤ+ satisfy 𝑏𝑛 > 0 without any genuine
loss of generality. This is achieved by simply setting ei𝜃𝑛 = i𝑛. We henceforth assume that 𝑏𝑛 > 0.
6
ISERLES and WEBB
Remark 2. Under the assumption of Remark 1, the functions Φ are real if and only if 𝑔(𝜉) has even
real part and odd imaginary part (with respect to the origin), and 𝑃 is orthonormal with respect
to an even measure (with respect to the origin). In this case, 𝑏𝑛 > 0 and 𝑐𝑛 = 0 for all 𝑛.
Theorem 1 and Remarks 1 and 2 were proved by the present authors in Refs. 1, 2 along with
results characterizing when such systems are orthogonal with respect to the standard inner prod-
uct on L2(ℝ). The following Theorem generalizes these orthogonality results to the Sobolev inner
products in Equation (16).
Theorem 2. Let 𝜑 satisfy criterion (𝐴), which implies that (17) holds. Then 𝜑 also satisfies criterion
(𝐵) if and only if the mollifier 𝑔 satisfies
𝑤(𝜉) = 𝑣(𝜉)|𝑔(𝜉)|2,
(18)
where 𝑤(𝜉) is the positive weight function with respect to which the polynomials 𝑃 are orthonormal,
𝓁=0 𝑣𝓁𝜉2𝓁. In particular, it is necessary for the nonnegative sequence {𝑣𝓁}𝓁∈ℤ+ to decay
and 𝑣(𝜉) =
sufficiently fast that 𝑣(𝜉) is finite on the support of 𝑤.
∑∞
Proof. By Parseval’s Theorem,
∞
∫
−∞
𝜑𝑛(𝑥)𝜑𝑚(𝑥) d𝑥 = (−i)𝑚−𝑛
∞
∫
−∞
𝑝𝑛(𝜉)𝑝𝑚(𝜉)|𝑔(𝜉)|2 d𝜉.
(19)
Furthermore, since ˆ𝜑(𝓁)(𝜉) = (−i𝜉)𝓁 ̂𝜑(𝜉) (where ̂𝜑 denotes the Fourier transform of 𝜑), we have
∞
∫
−∞
𝑛 (𝑥)𝜑(𝓁)
𝜑(𝓁)
𝑚 (𝑥) d𝑥 = (−i)𝑚−𝑛
∞
∫
−∞
𝑝𝑛(𝜉)𝑝𝑚(𝜉)𝜉2𝓁|𝑔(𝜉)|2 d𝜉.
(20)
Therefore,
⟨𝜑𝑛, 𝜑𝑚
⟩
𝑣 = (−i)𝑚−𝑛
∞
∫
−∞
𝑝𝑛(𝜉)𝑝𝑚(𝜉)𝑣(𝜉)|𝑔(𝜉)|2 d𝜉
(21)
This makes it clear that 𝜑 is orthonormal with respect to the Sobolev inner product if and only if
𝑃 is orthonormal with respect to the measure 𝑣(𝜉)|𝑔(𝜉)|2d𝜉.
■
Remark 3. There are infinitely many choices of 𝑔 which satisfy (18), namely,
√
𝑔(𝜉) =
𝑤(𝜉)
𝑣(𝜉)
ei𝜗(𝜉),
(22)
for any measurable real-valued function 𝜗. Our canonical choice is 𝜗 ≡ 0, although we know of
no good reason, except for simplicity, why this might be superior to other choices.
ISERLES and WEBB
7
It is important to answer what space the resulting orthonormal system is dense in: ideally, this
𝑣, but this need not be the
is the inner product space (14), endowed with the inner product ⟨⋅, ⋅⟩
case.
Theorem 3 (Orthogonal bases of Paley–Wiener spaces). Let Φ = {𝜑𝑛}𝑛∈ℤ+ satisfy the requirements
of Theorem 2 with weight function 𝑤(𝜉) such that polynomials are dense in L2(ℝ; 𝑤(𝜉)d𝜉). Then, Φ
forms a basis for the closure (in 𝐇2,𝑣(ℝ)) of the Paley–Wiener space
Ω(ℝ), where Ω is the support
of 𝑤.
A proof of Theorem 3 can be obtained by modifying Theorem 9 from Ref. 1. The key corollary is
that for a basis Φ satisfying the requirements of Theorem 2 to be complete in L2(ℝ), it is necessary
that the polynomial basis 𝑃 is orthogonal with respect to a measure that is supported on the whole
real line.
3
SOBOLEV CASCADES
In this section, we derive two methods for producing orthonormal systems in the Sobolev space
𝐇𝑠
2(ℝ) where 𝑠 = 0, 1, 2, ….
3.1
Cascades of first and second kinds
For a weight function 𝑤 and 𝑠 ∈ ℤ+, we can define the following two sequences of bases:
⟨𝑠⟩
𝑛 (𝑥) =
𝜑
i𝑛
√
2𝜋
∞
∫
−∞
ei𝑥𝜉 𝑝𝑛(𝜉)
√
∑𝑠
𝑤(𝜉)
𝑘=0 𝜉2𝑘
d𝜉,
where 𝑃 = {𝑝𝑛}𝑛∈ℤ+ are orthonormal polynomials with respect to 𝑤(𝜉), and
𝜑[𝑠]
𝑛 (𝑥) =
i𝑛
√
2𝜋
∞
∫
−∞
ei𝑥𝜉 𝑝[𝑠]
𝑛 (𝜉)
√
𝑤(𝜉) d𝜉,
where 𝑃[𝑠] = {𝑝[𝑠]
𝑛 }𝑛∈ℤ+ are orthonormal polynomials with respect to the weight
𝑤[𝑠](𝜉) =
)
𝜉2𝑘
𝑤(𝜉) =
(
𝑠∑
𝑘=0
1 − 𝜉2(𝑠+1)
1 − 𝜉2
𝑤(𝜉).
(23)
(24)
(25)
By the theory described in Section 2, both systems Φ⟨𝑠⟩ = {𝜑
𝑛 }𝑛∈ℤ+
have skew-Hermitian tridiagonal differentiation matrices and both are orthonormal systems with
respect to the standard 𝐇𝑠
2(ℝ) Sobolev inner product described in the introduction. Furthermore,
all of these systems are bases for (closure of) the Paley–Wiener space
Ω(ℝ), where Ω is the
support of 𝑤.
⟨𝑠⟩
𝑛 }𝑛∈ℤ+ and Φ[𝑠] = {𝜑[𝑠]
8
ISERLES and WEBB
⟨0⟩
𝑛 .
We call the sequence Φ⟨0⟩, Φ⟨1⟩, Φ⟨2⟩ … a Sobolev cascade of the first kind for the weight function
𝑤, and Φ[0], Φ[1], Φ[2] … a Sobolev cascade of the second kind for the weight function 𝑤. Note that
𝜑[0]
𝑛 = 𝜑
While a cascade of the first kind is perhaps a more natural generalization of L2-orthogonality,
it is also more problematic. Typically, the polynomials 𝑝𝑛 might be already known; however, the
⟨𝑠⟩
𝑛 s, is often unknown, even for 𝑠 = 1. The issue
explicit form of the integrals (23), hence of the 𝜑
with cascades of the second kind is different: the polynomials 𝑃[𝑠] are usually unknown for 𝑠 ∈
ℕ even for the most familiar measures such as Legendre or Hermite. On the other hand, once
we know 𝑝[𝑠]
𝑛 is available for all 𝑛, 𝑠 ∈
ℤ+ through the integral (24). Note that to compute (2) in a closed form, we need to be able to
𝑤 for polynomials 𝑝: exactly the same is required for
integrate explicitly Fourier transforms of 𝑝
the computation of (24).
𝑛 explicitly, the closed form of 𝜑[𝑠]
𝑛 and can compute 𝜑[0]
√
3.2
Sobolev cascades of the second kind
Orthogonal systems in a cascade of the second kind have a simple relationship. The following
theorem is a straightforward consequence of the Geronimus transformation.13
Theorem 4. Let 𝑠 ∈ ℤ+. There exists an infinite, lower triangular matrix 𝐶[𝑠] that has bandwidth
2𝑠, such that
𝝋[0] = 𝐶[𝑠]𝝋[𝑠],
(26)
where 𝝋[𝑠] are the elements of Φ[𝑠], arranged into a column vector.
Proof. Since 𝑝[0]
𝑛 and 𝑝[𝑠]
connection coefficient matrix ̃𝐶[𝑠] such that
𝑛 are polynomials of degree 𝑛 (for every 𝑛), there exists a lower triangular
𝑝[0]
𝑛 =
𝑛∑
𝑗=0
𝑛,𝑗𝑝[𝑠]
̃𝐶[𝑠]
𝑗 .
(27)
Since 𝑃[𝑠] is an orthonormal basis with respect to the weight function (
the formula
∑𝑠
𝑘=0 𝜉2𝑘)𝑤(𝜉), we have
̃𝐶[𝑠]
𝑛,𝑗 = ∫
∞
−∞
𝑛 (𝜉)𝑝[𝑠]
𝑝[0]
𝑗 (𝜉)
)
𝜉2𝑘
𝑤(𝜉) d𝜉.
(
𝑠∑
𝑘=0
(28)
∑𝑠
𝑗 (𝜉)(
Since 𝑝[𝑠]
respect to 𝑤, we have that 𝐶[𝑠]
𝑘=0 𝜉2𝑘) is a polynomial of degree at most 𝑗 + 2𝑠, and 𝑃[0] is orthonormal with
𝑛,𝑗 = 0 if 𝑗 ≤ 𝑛 − 2𝑠 − 1, which proves the desired bandwidth of the
𝑤(𝜉) and taking the inverse
matrix. The proof is completed by multiplying Equation (27) by
√
ISERLES and WEBB
Fourier transform:
𝜑[0]
𝑛 =
𝑛∑
𝑗=0
𝑛,𝑗𝜑[𝑠]
𝐶[𝑠]
𝑗 , where 𝐶[𝑠]
𝑛,𝑗 = i𝑛−𝑗 ̃𝐶[𝑠]
𝑛,𝑗.
9
(29)
■
Note further that if the weight function 𝑤 is symmetric, then all the polynomials 𝑝[𝑠]
𝑛 maintain
the parity of 𝑛 and it follows easily that 𝐶[𝑠]
𝑛,𝑗 = 0 for 𝑛 + 𝑗 odd.
Theorem 4 has two consequences. First, if one can calculate {𝜑[0]
𝑁 }, then it is pos-
𝑁 } in (𝑁) operations by applying forward substitution to the
0 , 𝜑[0]
1 , … , 𝜑[0]
0 , 𝜑[𝑠]
sible to calculate {𝜑[𝑠]
banded lower triangular system with matrix 𝐶[𝑠].
1 , … , 𝜑[𝑠]
Second, given 𝑁 + 1 expansion coefficients in the basis Φ[0], we can compute the equivalent
expansion coefficients in the basis Φ[𝑠] in (𝑁) operations. Specifically, if
then
𝑁∑
𝑛=0
𝑛 𝜑[0]
𝑎[0]
𝑛 (𝑥) =
𝑁∑
𝑛=0
𝑛 𝜑[𝑠]
𝑎[𝑠]
𝑛 (𝑥),
𝐶[𝑠]
⊤
𝒂[𝒔] = 𝒂[𝟎],
(30)
(31)
which can be solved in (𝑁) operations by backsubstitution.
A neat idea has been suggested by one of the referees. Let ̃𝐶[𝑠] = 𝐿𝐿⊤ be a Cholesky factorization
of the symmetric matrix ̃𝐶[𝑠] and set
𝒑[𝑠](𝜉) =
⎡
𝑝[𝑠]
0 (𝜉)
⎢
𝑝[𝑠]
⎢
1 (𝜉)
⎢
𝑝[𝑠]
⎢
2 (𝜉)
⎢
⎣
⋮
⎤
⎥
⎥
⎥
.
⎥
⎥
⎦
(32)
Therefore, ̃𝐶[0] = ⟨𝒑[0], 𝒑[0]⟩
∏𝑠
𝑗=1 𝜉2𝑗d𝜇(𝜉).
Orthonormality of the {𝑝[𝑠]
−1
̃𝐶[𝑠]
𝒑[0]. Therefore,
𝑤𝑠 , where ⟨ ⋅ , ⋅ ⟩
𝑤𝑠 is the inner product corresponding to the measure
𝑛 }𝑛∈ℤ+ implies that ⟨𝒑[𝑠], 𝒑[𝑠]⟩
𝑤𝑠 =, while (27) means that 𝒑[𝑠] =
𝐼 = ̃𝐶[𝑠]
−1⟨𝒑[0], 𝒑[0]⟩
𝑤𝑠
̃𝐶[𝑠] − ⊤ = ̃𝐶[𝑠]−1𝐿𝐿⊤ ̃𝐶[𝑠] − ⊤ = ( ̃𝐶[𝑠]−1𝐿)( ̃𝐶[𝑠]−1𝐿)⊤.
(33)
We deduce that ̃𝐶[𝑠]−1𝐿 is an idempotent matrix and both ̃𝐶[𝑠] and 𝐿 being lower triangular deduce
that ̃𝐶[𝑠] = 𝐿. Therefore, practical calculation of connection coefficients involves just Cholesky
factorization of the matrix ̃𝐶[0], of bandwidth 2𝑠.
10
ISERLES and WEBB
In general, it appears that the nonzero entries of 𝐶[𝑠]
𝑛,𝑗 obey no recognizable numerical relations:
for example, the 6 × 6 leading principal submatrix of 𝐶[1] for the Hermite weight is
√
⎡
⎢
⎢
0
⎢
√
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎣
0
0
0
3
2
1
3
0
√
0
√
5
2
3
5
0
0
0
0
√
19
6
0
√
18
19
0
0
0
0
√
39
10
0
√
819
407
0
0
0
0
√
173
38
0
⎤
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎦
.
0
0
0
0
0
√
407
78
(34)
It is difficult to discern a pattern: numerical experiments for large values of 𝑛 indicate that both
𝑛,𝑛 and 𝐶[1]
𝐶[1]
All this does not rule out computing the 𝐶[𝑠]
𝑛), in line with the proof of the Freud conjecture.14
𝑛,𝑗s numerically. Modifying a weight by a quadratic
𝑛+2,𝑛 grow like (
√
factor and computing the connection coefficients is discussed in Refs. 13 and 15.
An alternative approach toward the polynomials 𝑝[𝑠]
𝑛 uses the Christoffel theorem.16(p. 37) Given
a measure d𝜇 and the corresponding set of monic orthogonal polynomials {𝑝𝑛}𝑛∈ℤ+, as well as a
polynomial Ξ(𝜉) =
𝓁=1(𝜉 − 𝜁𝓁), the theorem allows for an explicit construction of polynomials
orthogonal with respect to Ξ(𝜉)d𝜇(𝜉). Specialized to the problem at hand, 𝑟 = 2𝑠 and
∏𝑟
Ξ(𝑥) =
𝑠∑
𝑗=0
𝜉2𝑗 =
1 − 𝜉2(𝑠+1)
1 − 𝜉2
=
(𝜉 − 𝜉𝓁),
𝑠∏
𝓁=−𝑠
𝓁≠0
where 𝜉𝓁 = ei𝜋𝓁∕(𝑠+1), whereby Christoffel’s construction yields
𝑝[𝑠]
𝑛 (𝜉) =
1
ℎ𝑛,𝑠Ξ(𝜉)
where
⎡
𝑝𝑛(𝜉1) 𝑝𝑛+1(𝜉1) ⋯ 𝑝𝑛+2𝑠(𝜉1)
⎢
𝑝𝑛(𝜉2) 𝑝𝑛+1(𝜉2) ⋯ 𝑝𝑛+2𝑠(𝜉2)
⎢
⎢
det
⎢
⎢
⎢
⎣
𝑝𝑛(𝜉𝑠) 𝑝𝑛+1(𝜉𝑠) ⋯ 𝑝𝑛+2𝑠(𝜉𝑠)
𝑝𝑛+1(𝜉) ⋯ 𝑝𝑛+2𝑠(𝜉)
𝑝𝑛(𝜉)
⋮
⋮
⋮
⎤
⎡
𝑝𝑛(𝜉1) 𝑝𝑛+1(𝜉1) ⋯ 𝑝𝑛+2𝑠−1(𝜉1)
⎥
⎢
⎥
⎢
𝑝𝑛(𝜉2) 𝑝𝑛+1(𝜉2) ⋯ 𝑝𝑛+2𝑠−1(𝜉2)
⎥
⎢
ℎ𝑛,𝑠 = det
.
⎥
⎢
⎥
⎢
⎦
⎣
𝑝𝑛(𝜉𝑠) 𝑝𝑛+1(𝜉𝑠) ⋯ 𝑝𝑛+2𝑠−1(𝜉𝑠)
⋮
⋮
⋮
⎤
⎥
⎥
⎥
,
⎥
⎥
⎥
⎦
(35)
(36)
(37)
While the polynomials 𝑝[𝑠]
𝑛 in (36) are monic, they can be easily orthonormalized to fit into
our setting.
ISERLES and WEBB
11
Polynomials of the second cascade display an interesting feature. We recall that an orthogo-
nal polynomial system is semiclassical17 if their weight function 𝑤 obeys the linear differential
equation
𝐴𝑤′ + Bw = 0, A, 𝐵 polynomials, A(𝜉) > 0 for 𝜉 ∈ supp 𝑤.
(38)
The following lemma is valid inter alia for all the examples in the current paper.
Lemma 1.
are semiclassical.
If d𝜇(𝜉) = 𝑤(𝜉)d𝜉 and 𝑤 obeys (38), then all the systems {𝑝[𝑠]
𝑛 }𝑛∈ℤ+ for 𝑠 ∈ ℤ+
Proof. We include a short proof, but note that this result is a special case of Theorem 5.1 of Ref. 18.
𝑗=0 𝜉2𝑗 > 0, 𝜉 ∈ supp 𝑤, and 𝑤𝑠 = Ξ𝑠𝑤. It is trivial to confirm by direct
Let 𝑠 ∈ ℕ. Set Ξ𝑠(𝜉) =
∏𝑠
differentiation that
𝐴Ξ𝑠𝑤′
𝑠 + (𝐵Ξ𝑠 − 𝐴Ξ′
𝑠)𝑤𝑠 = 0
and 𝐴Ξ𝑠 > 0, hence 𝑤𝑠 is consistent with (38) and {𝑝[𝑠]
𝑛 }𝑛∈ℤ+ is semiclassical.
In other words, semiclassicality is preserved throughout a cascade of the second kind.
(39)
■
4
HERMITE-TYPE SYSTEMS
4.1
The Hermite–Sobolev cascade of the first kind
A natural starting point is the Hermite weight 𝑤(𝜉) = e−𝜉2
the definitions in Section 3, is 𝑔(𝜉) = e−𝜉2∕2∕(1 + 𝜉2)1∕2, so
, 𝜉 ∈ ℝ, and 𝑠 = 1. The mollifier, by
√
𝜑𝑛(𝑥) =
i𝑛
√
2𝜋
∞
∫
−∞
̃H𝑛(𝜉)
e−𝜉2
1 + 𝜉2
ei𝑥𝜉d𝜉,
𝑛 ∈ ℤ+,
where
̃H𝑛(𝜉) =
√
1
√
H𝑛(𝜉),
𝑛 ∈ ℤ+,
2𝑛𝑛!
𝜋
(40)
(41)
are the orthonormalised Hermite polynomials. Unfortunately, the integrals (40) are not known
in an explicit form, not even 𝜑0.
In Figure 1, we display the functions 𝜑𝑛, 𝑛 = 0, … , 5, computed by brute-force numerical
quadrature. In the background, in fainter color, we display the familiar Hermite functions that
follow from (2) and are orthonormal in L2(ℝ) (while, by Theorem 2, the 𝜑𝑛s are orthonormal in
𝐇1
2(ℝ)).
12
ISERLES and WEBB
F I G U R E 1
⟨0⟩
𝑛 (𝑥) = (−1)𝑛e−𝑥2∕2 ̃H𝑛(𝑥) in darker shade
𝜑
The first six functions 𝜑
⟨1⟩
𝑛 defined by (40), with corresponding Hermite functions
4.2
The Hermite–Sobolev cascade of the second kind
While the polynomials 𝑝[𝑠]
𝑛 from Subsection 3.1 are unknown for 𝑠 ∈ ℕ, it is possible to generate
them, as explained in Section 3 or directly from the moments: in the simplest nontrivial case,
𝑠 = 1, the moments are
∞
𝜇𝑛 = ∫
−∞
𝜉𝑛(1 + 𝜉2)e−𝜉2
d𝜉 =
⎡
⎢
⎢
⎣
and the first few 𝑝[1]
𝑛 s are
√
2𝑛+1
)
𝜋𝑛!(𝑛+3)
(
𝑛
2
0,
!
𝑝[1]
0 (𝜉) ≡
𝑝[1]
1 (𝜉) =
𝑝[1]
2 (𝜉) =
𝑝[1]
3 (𝜉) =
𝑝[1]
4 (𝜉) =
√
6
3𝜋1∕4
√
5
2
5𝜋1∕4
√
,
𝜉,
(
(
57
2
19𝜋1∕4
√
130
2
39𝜋1∕4
√
9861
2
519𝜋1∕4
)
,
5
6
)
𝜉
,
21
10
𝜉2 −
𝜉3 −
(
𝜉4 −
75
19
𝜉2 +
117
76
, 𝑛 even,
𝑛 odd,
⎤
⎥
⎥
⎦
(42)
)
,
ISERLES and WEBB
13
F I G U R E 2
The polynomials 𝑝[𝑠]
𝑛 for 𝑛 = 2, 3, 4, 5 and 𝑠 = 0, 1, 2, 4 (darker hue corresponds to larger 𝑠)
𝑝[1]
5 (𝜉) =
√
2
52910
2035𝜋1∕4
(
𝜉5 −
245
39
𝜉3 +
335
52
)
𝜉
,
(43)
and so on. Likewise, it is possible to compute recurrence coefficients,
√
√
√
√
√
√
𝑏0 =
5
6
, 𝑏1 =
19
15
, 𝑏2 =
315
190
, 𝑏3 =
1730
741
, 𝑏4 =
38665
13494
, 𝑏5 =
236925
70411
,
(44)
and so on, but difficult to discern any pattern except for the obvious, 𝑏𝑛 = (𝑛1∕2), 𝑛 ≫ 1, a con-
sequence of the proof of the Freud conjecture in Ref. 14. Likewise, we can compute 𝑝[𝑠]
𝑛 for 𝑠 ≥ 2:
Figure 2 displays 𝑝[𝑠]
Computing the 𝜑[1]
𝑛 for 𝑛 = 2, 3, 4, 5 and 𝑠 = 0, 1, 2, 3, 4.
𝑛 s in line with (11) is straightforward:
𝜑[1]
0 (𝑥) =
√
2
3
√
𝜋−1∕4e−𝑥2∕2,
𝜑[1]
1 (𝑥) = −
𝜋−1∕4𝑥e−𝑥2∕2,
4
5
14
ISERLES and WEBB
F I G U R E 3
darker shade
The first six functions 𝜑[1]
𝑛 with corresponding functions 𝜑[0]
𝑛 , which are orthogonal in L2(ℝ), in
𝜑[1]
2 (𝑥) =
1
√
57
√
𝜋−1∕4(6𝑥2 − 1)e−𝑥2∕2,
𝜑[1]
3 (𝑥) = −
2
585
𝜋−1∕4(10𝑥3 − 9𝑥)e−𝑥2∕2,
𝜑[1]
4 (𝑥) =
1
√
𝜋−1∕4(76𝑥4 − 156𝑥2 + 45)e−𝑥2∕2,
𝜑[1]
5 (𝑥) = −
39444
1
√
476190
𝜋−1∕4(156𝑥5 − 580𝑥3 + 405𝑥)e−𝑥2∕2,
(45)
and so on.
Figure 3 displays the above functions 𝜑[1]
same weight 𝑤(𝜉) = (1 + 𝜉2)e−𝜉2
and defined by (2).
𝑛 and, in fainter color, the functions 𝜑
⟨0⟩
𝑛 based on the
Lemma 2. For every 𝑛 ∈ ℤ+, we have 𝜑[1]
𝑛 (𝑥) = 𝜆𝑛(𝑥)e−𝑥2∕2, where 𝜆𝑛 is an 𝑛th-degree polynomial.
Proof. It is enough to prove that
𝜎𝑛(𝑥) =
i𝑛
√
2𝜋
∞
∫
−∞
𝜉𝑛e−𝜉2∕2+i𝑥𝜉d𝜉,
𝑛 ∈ ℤ+,
(46)
is of the asserted form, that is, an 𝑛th degree polynomial times e−𝑥2∕2. This follows readily
𝑛 = 𝜎𝑛+1 because, letting 𝜎𝑛(𝑥) = 𝛼𝑛(𝑥)e−𝑥2∕2, we
by induction on 𝑛 from 𝜎0(𝑥) = e−𝑥2∕2 and 𝜎′
obtain 𝛼𝑛+1(𝑥) = 𝛼′
■
𝑛(𝑥) − 𝑥𝛼𝑛(𝑥).
ISERLES and WEBB
Alternatively, substituting into (15), it is easy to see that
𝜆′
𝑛(𝑥) = −𝑏𝑛−1𝜆𝑛−1(𝑥) + 𝑥𝜆𝑛(𝑥) + 𝑏𝑛𝜆𝑛+1(𝑥),
n ∈ ℤ+.
15
(47)
The proof that 𝜆𝑛 is an 𝑛th degree polynomial follows at once by induction on this differential
recurrence, since 𝑏𝑛 > 0, 𝑛 ∈ ℕ.
The bad news is that the 𝜆𝑛s are not known and, as is trivial to verify, they do not obey a three-
term recurrence relation (hence, by the Favard theorem, cannot be orthogonal with respect to any
Borel measure). However, intriguingly, it follows easily from 𝐇1
𝑛 s that the
𝜆𝑛 are orthogonal with regard to the unconventional inner product
2 orthogonality of the 𝜑[1]
∞
⟨𝑓, 𝑔⟩ = ∫
−∞
{(1 + 𝑥2)𝑓(𝑥)𝑔(𝑥) − 𝑥[𝑓′(𝑥)𝑔(𝑥) + 𝑓(𝑥)𝑔′(𝑥)] + 𝑓′(𝑥)𝑔′(𝑥)]d𝑥.
(48)
It has been proved in Ref. 1 that there exists a unique L2-orthonormal system on the real line that
obeys (3) and where each function is a polynomial multiple of the same L2 function, specifically
Hermite functions (or 𝜑[0]
𝑛 }𝑛∈ℤ+, though, are 𝐇1
2-
orthonormal, they obey (3) and 𝜑[1]
⟨0⟩
𝑛 in present notation). The functions {𝜑[1]
𝑛 (𝑥) = e−𝑥2∕2𝜆𝑛(𝑥).
𝑛 = 𝜑
Lemma 3. The only 𝐇2,𝑣(ℝ)-orthonormal systems (see Equation (14)) with a tridiagonal differenti-
ation matrix that are of the form 𝜑𝑛(𝑥) = 𝐺(𝑥)𝜆𝑛(𝑥), 𝑛 ∈ ℤ+, for some function 𝐺 ∈ L2, (ℝ)𝐺 > 0
(and 𝐺(0) = 1 without loss of generality), where each 𝜆𝑛 is a polynomial of degree 𝑛, correspond to
(
−𝛾𝑥2 + 𝛿𝑥
𝐺(𝑥) = exp
(49)
)
for some constants 𝛾 > 0 and 𝛿 ∈ ℝ. The corresponding weight of orthonormality for 𝑃 in Theorem 1
is
𝑤(𝜉) ∝ 𝑣(𝜉)e−𝜉2∕(2𝛾).
(50)
Proof. We substitute 𝜑𝑛(𝑥) = 𝐺(𝑥)𝜆𝑛(𝑥) into (15), bearing in mind that 𝐺 > 0, to obtain,
𝜆′
𝑛(𝑥) = −𝑏𝑛−1𝜆𝑛−1(𝑥) +
i𝑐𝑛 −
[
𝐺′(𝑥)
𝐺(𝑥)
]
𝜆𝑛(𝑥) + 𝑏𝑛𝜆𝑛+1(𝑥),
𝑛 ∈ ℤ+.
(51)
Since deg 𝜆𝑚 = 𝑚 by assumption, we deduce, comparing degrees, that 𝐺′∕𝐺 is a linear poly-
nomial, and hence, that 𝐺(𝑥) is the exponential of a quadratic polynomial. We can set the
constant term in this quadratic to zero because 𝐺(0) = 1 without loss of generality, so we obtain
Equation (49).
Inverting the representation in Theorem 1, we have
𝑝𝑛(𝜉)𝑔(𝜉) =
(−i)𝑛
√
2𝜋
∞
∫
−∞
𝜑𝑛(𝑥)e−i𝑥𝜉d𝑥 =
(−i)𝑛
√
2𝜋
∞
∫
−∞
𝜆𝑛(𝑥)e−𝛾𝑥2𝜉+𝛿𝑥−i𝑥𝜉d𝑥.
(52)
16
ISERLES and WEBB
The case 𝑛 = 0 tells us that
𝑔(𝜉) ∝ exp(−(𝜉 − i𝛿)2∕(4𝛾)).
Theorem 2 tells us that for 𝐇2,𝑣(ℝ) orthonormality, we require
𝑤(𝜉) = 𝑣(𝜉)|𝑔(𝜉)|2,
(53)
(54)
which completes the proof of necessity of the forms of 𝐺 and 𝑤.
Now we prove sufficiency. Let 𝑤(𝜉) = 𝐶2𝑣(𝜉)e−𝜉2∕2𝛾 where 𝐶 ensures that 𝑤 has unit integral,
𝑔(𝜉) = 𝐶e−(𝜉−i𝛿)2∕(4𝛾), and define Φ as in Theorem 1. By Theorem 2, Φ is an 𝐇2,𝑣(ℝ)-orthonormal
system, so all that remains to prove is that 𝜑𝑛(𝑥) = 𝐺(𝑥)𝜆𝑛(𝑥) where 𝜆𝑛 is a polynomial of degree
𝑛. It is sufficient to show that 𝜌𝑛(𝑥) = ∫ ∞
−∞ 𝜉𝑛𝑔(𝜉)ei𝑥𝜉 d𝜉 is 𝐺(𝑥) times a polynomial of degree 𝑛,
which can be readily shown by induction starting from 𝜌0(𝑥) ∝ 𝐺(𝑥) and leveraging 𝜌𝑛+1(𝑥) =
−i𝜌′
■
𝑛(𝑥).
4.3
An 𝐇∞
𝟐 (ℝ) system based on the Hermite weight
Let 𝜎 ∈ (0, 1), 𝑤(𝜉) = e−𝜉2
, 𝜉 ∈ ℝ. Therefore,
by Theorem 3, the functions 𝜑𝑛, as defined by reqn:phinformula, are orthogonal with respect to
the infinite Sobolev inner product
(i.e., the standard Hermite weight) and 𝑣(𝜉) = e𝜎𝜉2
⟨𝑓, 𝑔⟩
𝑣 =
∞∑
𝓁=0
𝜎𝓁
𝓁! ∫
−∞
∞
𝑓(𝓁)(𝑥)𝑔(𝓁)(𝑥)d𝑥.
(55)
In this case, 𝑝𝑛s are scaled Hermite polynomials and 𝜑𝑛s can be computed explicitly.
Theorem 5. The Hermite weight 𝑤(𝜉) = e−𝜉2
(
(
)𝑛∕2
𝜑[∞]
𝑛
(𝑥) =
√
1
1 + 𝜎
1 − 𝜎
1 + 𝜎
𝜑[0]
𝑛
𝑥
√
1 − 𝜎2
, 𝑥 ∈ ℝ, generates the 𝐇∞
2 (ℝ) system
)
(
exp
)
,
𝜎𝑥2
2(1 − 𝜎2)
𝑛 ∈ ℤ+,
(56)
where 𝜑[0]
𝑛 is the standard 𝑛th Hermite function.
Proof. Let
̃𝜑𝑛(𝑥) =
i𝑛
√
2𝜋
∞
∫
−∞
H𝑛(𝜉)e−
1
2
(1+𝜎)𝜉2+i𝑥𝜉d𝜉,
𝑛 ∈ ℤ+,
(57)
whereby, orthonormalizing Hermite polynomials, (17) yields 𝜑𝑛(𝑥) = ̃𝜑𝑛(𝑥)∕
the standard generating function for Hermite polynomials,
2𝑛𝑛!
𝜋. Using
√
√
∞∑
𝑛=0
̃𝜑𝑛(𝑥)
𝑛!
𝑡𝑛 =
1
√
2𝜋
[
∞
∞∑
∫
−∞
𝑛=0
H𝑛(𝜉)
𝑛!
]
(
(i𝑡)𝑛
exp
−
)
1 + 𝜎
2
𝜉2 + i𝑥𝜉
d𝜉
ISERLES and WEBB
17
1
√
=
2𝜋
∞
∫
−∞
(
exp
2i𝜉𝑡 + 𝑡2 −
(
)
1
=
√
1 + 𝜎
exp
−
𝑥2
2(1 + 𝜎)
)
𝜉2 + i𝑥𝜉
d𝜉
1 + 𝜎
2
(
)
exp
−
2𝑥𝑡
1 + 𝜎
−
1 − 𝜎
1 + 𝜎
𝑡2
1
=
√
1 + 𝜎
(
exp
−
𝑥2
2(1 + 𝜎)
)
⎛
⎜
exp
⎜
⎝
2𝑥
−
√
1 − 𝜎2
(√
)
(√
1−𝜎
1+𝜎
𝑡
−
)2⎞
⎟
⎟
⎠
1−𝜎
1+𝜎
𝑡
(58)
and, using the same generating function in the opposite direction,
∞∑
𝑛=0
̃𝜑𝑛(𝑥)
𝑛!
𝑡𝑛 =
√
1
1 + 𝜎
(
exp
−
𝑥2
2(1 + 𝜎)
) ∞∑
𝑛=0
(
)(√
)𝑛
1
𝑛!
H𝑛
𝑥
−
√
1 − 𝜎2
1 − 𝜎
1 + 𝜎
𝑡
.
(59)
Therefore,
Normalizing,
̃𝜑𝑛(𝑥) =
(
1 − 𝜎
1 + 𝜎
(−1)𝑛
√
1 + 𝜎
)𝑛∕2
(
H𝑛
𝑥
√
1 − 𝜎2
)
(
exp
−
)
.
𝑥2
2(1 + 𝜎)
(60)
𝜑[∞]
𝑛
(𝑥) =
√
(−1)𝑛
√
(
(1 + 𝜎)2𝑛𝑛!
(
1
=
√
1 + 𝜎
1 − 𝜎
1 + 𝜎
𝜋
)𝑛∕2
1 − 𝜎
1 + 𝜎
(
𝜑[0]
𝑛
and (56) is true.
)𝑛∕2
(
H𝑛
)
(
exp
−
)
,
𝑥2
2(1 + 𝜎)
𝑥
√
1 − 𝜎2
)
(
exp
)
,
𝜎𝑥2
2(1 − 𝜎2)
𝑥
√
1 − 𝜎2
𝑛 ∈ ℤ+,
(61)
■
Figure 4 displays the functions 𝜑[∞]
, 𝑛 = 0, … , 5, for three different values of 𝜎 ∈ (0, 1). The
zeros of 𝜑𝑛 are scaled zeros of a Hermite polynomial and, the scaling being monotone, the zeros
are “squeezed” in a uniform manner for increasing 𝜎, as evident in the figure.
𝑛
5
BILATERAL LAGUERRE-TYPE WEIGHTS
Deferring the standard Laguerre weight (which is not symmetric) to Section 7, we let 𝑤(𝜉) = (1 +
𝜉2)e−|𝜉|, 𝜉 ∈ ℝ. Note that the underlying orthogonal polynomials are unknown explicitly, yet can
be computed. The 𝜑[1]
𝑛 s are
𝜑[1]
0 (𝑥) =
2
√
√
3
𝜋
1
1 + 4𝑥2
,
𝜑[1]
1 (𝑥) =
16
√
√
26
𝜋
𝑥
(1 + 4𝑥2)2
,
18
ISERLES and WEBB
F I G U R E 4
The first six 𝐇∞
2 Hermite-type functions 𝜑[∞]
𝑛
for 𝜎 =
1
4
,
1
2
,
3
4
in progressively darker hues
𝜑[1]
2 (𝑥) =
√
2
√
1167
𝜋
1 + 248𝑥2 + 208𝑥4
(1 + 4𝑥2)3
,
𝜑[1]
3 (𝑥) =
√
16
√
23179
𝜋
−21𝑥 + 456𝑥3 + 496𝑥5
(1 + 4𝑥2)4
,
𝜑[1]
4 (𝑥) =
√
2
√
309347971
𝜋
2925 − 128784𝑥2 + 1703264𝑥4 + 3029760𝑥6 + 1369344𝑥8
(1 + 4𝑥2)5
,
𝜑[1]
5 (𝑥) =
√
16
√
22678864934
𝜋
25749𝑥−1017424𝑥3 +5715040𝑥5 +13510400𝑥7 +7744768𝑥9
(1 + 4𝑥2)6
.
(62)
The general formula is a polynomial of degree 2𝑛 − [1 − (−1)𝑛]∕2 in 𝑥 (of the same parity as 𝑛),
divided by (1 + 4𝑥2)𝑛+1. This can be easily verified because by (17) and 𝑔(𝜉) = e−|𝜉|∕2 each 𝜑[1]
𝑛 is
a linear combination of 𝜆𝑛, 𝜆𝑛−2, …, where
𝜆𝑛(𝑥) =
i𝑛
√
2𝜋
∞
∫
−∞
𝜉𝑛e−|𝜉|2+i𝑥𝜉d𝜉,
𝑛 ∈ ℤ+
(63)
and 𝜆′
𝑛(𝑥) = 𝜆𝑛+1(𝑥) implies that
𝜆𝑛(𝑥) = 𝜆(𝑛)
0 (𝑥) =
√
2
2
√
𝜋
d𝑛
d𝑥𝑛
1
1 + 4𝑥2
,
𝑛 ∈ ℤ+.
(64)
ISERLES and WEBB
19
6
BESSEL-LIKE FUNCTIONS
6.1
Transformation of Chebyshev polynomials
We rewrite (12) in the form (17),
𝜑𝑛(𝑥) =
i𝑛
√
2𝜋
1
∫
−1
̃T𝑛(𝜉)ei𝑥𝜉
d𝜉
√
1 − 𝜉2
= (−1)𝑛J𝑛(𝑥),
(65)
√
√
2∕𝜋T𝑛 (except that ̃T0
where ̃T𝑛 =
𝜋) are orthonormal Chebyshev polynomials of the
first kind. It is easy to verify directly that the 𝜑𝑛s cannot be bounded in any Sobolev norm because
the Weber–Schafheitlin formula19(10.22.57) implies that for ℜ𝜆 > 0
≡ T0∕
∞
∫
−∞
𝜑2
𝑛(𝑥)d𝑥
𝑥𝜆
= ∫
∞
−∞
J2
𝑛(𝑥)d𝑥
𝑥𝜆
=
Γ(𝑛 +
1
2
)Γ(𝜆)
2𝜆−1Γ2(
1
2
𝜆 +
1
2
)Γ(
1
2
𝜆 + 𝑛 +
1
2
)
𝜆→0
⟶ ∞.
(66)
If instead of a Chebyshev measure, we use the Legendre measure, 𝑤(𝜉) = 𝜒(−1,1)(𝜉), the state
of affairs is different: 𝑔(𝜉) = 𝜒(−1,1) results in
𝜑𝑛(𝑥) = (−1)𝑛
√
1
2
𝑛 +
𝑥
(𝑥),
J
𝑛+
1
2
𝑥 ∈ ℝ,
(67)
and the 𝜑𝑛s are integrable on ℝ.
6.2
The Legendre weight
The most obvious example of an 𝐇1
2(ℝ) system is based on the Legendre weight 𝑤(𝜉) = 𝜒(−1,1)(𝜉),
√
in which the orthonormal polynomials are 𝑝𝑛 =
𝑛 +
1
2
P𝑛. Thus,
⟨1⟩
𝑛 (𝑥) =
𝜑
√
𝑛 +
i𝑛
√
2𝜋
1
2 ∫
1
−1
√
P𝑛(𝜉)
1 + 𝜉2
ei𝑥𝜉d𝜉,
𝑛 ∈ ℤ+.
(68)
⟨1⟩
𝑛 }𝑛∈ℤ+ is orthonormal in 𝐇1
While {𝜑
2(ℝ), it is not a complete basis because all Fourier spectra
are restricted to [−1, 1], yet it might be of an independent interest. Perhaps, a more vexing issue is
that above integrals are not available in an explicit form. This is not an insurmountable problem in
⟨1⟩
𝑛 s which we can compute for individual values of 𝑥 using a fast Fourier
the computation of the 𝜑
transform,20,21 although it presents an obvious obstacle to their analysis.
In Figure 5, we have computed 𝜑
⟨1⟩
5 numerically. Like other transformed functions
⟨1⟩
(2) or reqn:phinformula, the 𝜑
𝑛 s seem to be endowed with a wealth of structural features
and regularities that have been discussed briefly (for (2)) in Ref. 4 but overall are a subject for
future research.
⟨1⟩
0 , … , 𝜑
20
ISERLES and WEBB
F I G U R E 5
The first six functions 𝜑
⟨1⟩
𝑛
for the Legendre weight
6.3
Sobolev–Legendre cascades
We revisit the essence of Subsections 3.1.1–3.1.2, except that the range of integration is now [−1, 1].
First, we set 𝑤 = 𝜒(−1,1), let 𝑝𝑛 be the (orthonormalized) Legendre polynomials and for every
𝑠 ∈ ℤ+ set,
⟨𝑠⟩
𝑛 (𝑥) =
𝜑
i𝑛
√
2𝜋
1
∫
−1
𝑝𝑛(𝜉)
(
𝑠∑
𝓁=0
)−1∕2
𝜉2𝓁
ei𝑥𝜉d𝜉,
𝑛 ∈ ℤ+.
(69)
Second, we might define 𝑤𝑠(𝜉) = 𝜒(−1,1)(𝜉)
∑𝑠
𝓁=0 𝜉2𝓁, 𝑠 ∈ ℤ+, and set
𝜑[𝑠]
𝑛 (𝑥) =
i𝑛
√
2𝜋
1
∫
−1
𝑝[𝑠]
𝑛 (𝜉)ei𝑥𝜉d𝜉,
𝑛 ∈ ℤ+,
(70)
where {𝑝[𝑠]
at once from Theorem 2 that
𝑛 }𝑛∈ℤ+ is the orthonormal polynomial system corresponding to the weight 𝑤𝑠. It follows
𝑠∑
∞
∫
−∞
𝓁=0
d𝓁𝜑
⟨𝑠⟩
𝑚 (𝑥)
d𝑥𝓁
d𝓁𝜑
⟨𝑠⟩
𝑛 (𝑥)
d𝑥𝓁
d𝑥 =
𝑠∑
∞
∫
−∞
𝓁=0
d𝓁𝜑[𝑠]
𝑚 (𝑥)
d𝑥𝓁
d𝓁𝜑[𝑠]
𝑛 (𝑥)
d𝑥𝓁
d𝑥 = 𝛿𝑚,𝑛
(71)
for all 𝑚, 𝑛 ∈ ℤ+ and both {𝜑
2(ℝ). Of course,
neither is dense in the Sobolev space because their Fourier spectra are restricted to [−1, 1]—
they are dense in an obvious generalization of Paley–Wiener spaces to the realm of Sobolev
𝑛 }𝑛∈ℤ+ are orthonormal sets in 𝐇𝑠
⟨𝑠⟩
𝑛 }𝑛∈ℤ+ and {𝜑[𝑠]
ISERLES and WEBB
21
F I G U R E 6
increasing 𝑠 corresponds to increasing line thickness and darker hue
The Sobolev–Legendre cascade of the second kind: The first six functions 𝜑[𝑠]
𝑛 for 𝑠 = 0, 1, 2:
spaces. The systems (69) and (70) are the Sobolev–Legendre cascades of the first and the second
kinds, respectively.
𝑛 has been given in (67)), 𝜑
We recall a major practical difference between the two cascades: except for the case 𝑠 = 0 (when
⟨0⟩
𝑛 = 𝜑[0]
𝑛 , being an integral
𝜑
in [−1, 1] of a polynomial times ei𝑥𝜉, can be computed at great ease and is a linear combination
of spherical Bessel functions. Consequently, in the sequel, we focus on the Sobolev–Legendre
cascade of the second kind.
⟨𝑠⟩
𝑛 is unknown in an explicit form while 𝜑[𝑠]
Figure 6 displays the beginning (i.e., 𝑠 = 0, 1, 2) of the cascade of the second kind. The obvious
1 and 𝜑[0]
1 , respectively.
is a scalar multiple of 𝜉. Another
0 and the same is true for 𝜑[𝑠]
0 is a scalar multiple of 𝜑[0]
is a constant, whereas 𝑝[𝑠]
1
observation is that 𝜑[𝑠]
This follows from (70) because 𝑝[𝑠]
0
obvious indication is that, as 𝑠 grows, 𝜑[𝑠]
be less interesting than it appears. In particular,
𝑛 converges pointwise to a function 𝜑[∞]
𝑛
Lemma 4. 𝜑[∞]
𝑛
≡ 0.
Proof. Let
𝑢𝑠 = ∫
1
𝑠∑
−1
𝓁=0
𝜉2𝓁d𝜉,
𝑠 ∈ ℤ+.
Then 𝑝[0]
0
√
≡ 1∕
𝑢2 and, by (70),
√
𝜑[𝑠]
0 (𝑥) =
2
𝜋𝑢𝑠
sin 𝑥
𝑥
.
yet this might
(72)
(73)
22
ISERLES and WEBB
F I G U R E 7
corresponds to darker hue
The Sobolev–ultraspherical cascade: The first six functions 𝜑[𝑠]
𝑛 for 𝑠 = 0, 1, 2, 3: increasing 𝑠
However,
and the lemma follows.
𝑢𝑠 =
𝑠∑
1
𝓁=0
𝓁 +
1
2
𝓁→∞
⟶ ∞
(74)
■
Alternatively, lim𝑠→∞ 𝑤[𝑠](𝜉) = (1 − 𝜉2)−1𝜒(−1,1) ∉ L2(ℝ). An obvious remedy, which we do
not pursue in this paper, is to consider the weight 𝑤𝑠(𝜉) =
𝓁=0 𝜎𝓁𝜉2𝓁 for some 𝜎 ∈ (0, 1).
∑𝑠
6.4
The Sobolev-ultraspherical cascade of the second kind
We construct a cascade of the second kind based on the ultraspherical weight 𝑤[0](𝜉) =
(1 − 𝜉2)𝜒(−1,1)(𝜉). Therefore,
𝑤[𝑠](𝜉) = 𝑤[0](𝜉)
𝑠∑
𝓁=0
𝜉2𝓁 = (1 − 𝜉2𝑠+2)𝜒(−1,1),
𝑠 ∈ ℤ+
(75)
and, as 𝑠 → ∞, 𝑤[𝑠] converges weakly to the Legendre weight.
In Figure 7, we display the functions 𝜑[0]
𝑛 , … , 𝜑[3]
𝑛 for 𝑛 = 0, 1, … , 5. While the convergence to
a limit in each figure is quite persuasive, we must beware of “proof by picture:” convergence is
equally pictorially persuasive in Figure 6 where, as we have already seen, it need not take place.
ISERLES and WEBB
23
7
NONSYMMETRIC MEASURES
The most obvious nonsymmetric weight function is the Laguerre weight 𝑤(𝜉) = e−𝜉𝜒[0,∞)(𝜉). In
that case, the 𝜑𝑛s are the Malmquist–Takenaka functions, which have a particularly neat form,2
√
𝜑𝑛(𝑥) =
2
𝜋
i𝑛 (1 + 2i𝑥)𝑛
(1 − 2i𝑥)𝑛+1
,
𝑛 ∈ ℤ+.
(76)
and they are dense in
[0,∞)(ℝ). It is possible to extend them to a system dense in all of L2(ℝ)
by melding them with another system, generated by the mirror image of the Laguerre weight,
e𝜉𝜒(−∞,0](𝜉): together we obtain the same system as (76), except that now 𝑛 ranges over all of ℤ.
It is, of course, perfectly possible for a system with a nonsymmetric measure to be dense in
L2(ℝ), provided that the support of 𝑤 is all of ℝ: an example is the Hermite-type weight 𝑤(𝜉) =
(1 − 𝜉)2e−𝜉2
.
7.1
Shifted Hermite weight
Letting 𝜌 ∈ ℝ, we consider the weight 𝑤(𝜉) = e−(𝜉−𝜌)2
mials are 𝑝𝑛(𝜉) = ̃H𝑛(𝜉 − 𝜌), where ̃H𝑛 is the orthonormalized Hermite polynomial,
H𝑛∕
. The underlying orthonormal polyno-
̃H𝑛 =
𝜋. Therefore, seeking 𝐇1
2(ℝ) orthogonality,
2𝑛𝑛!
√
√
⟨0⟩
𝑛 (𝑥, 𝜌) =
𝜑
⟨1⟩
𝑛 (𝑥, 𝜌) =
𝜑
i𝑛
√
2𝜋
i𝑛
√
2𝜋
∞
∫
−∞
∞
∫
−∞
̃H𝑛(𝜉 − 𝜌)e−(𝜉−𝜌)2∕2+i𝑥𝜉d𝜉 = ei𝜌𝑥𝜑
⟨0⟩
𝑛 (𝑥, 0),
̃H𝑛(𝜉 − 𝜌)
e−(𝜉−𝜌)2∕2+i𝑥𝜉
1 + 𝜉2
√
d𝜉,
𝑛 ∈ ℤ+.
(77)
It is easy to verify that
⟨0⟩
𝑛 (𝑥, 𝜌) = ei𝜌𝑥𝜑
𝜑
⟨0⟩
0 (𝑥, 0),
𝑥, 𝜌 ∈ ℝ.
(78)
⟨0⟩
𝑛 = 𝜑[0]
Thus, 𝜑
situation is more intriguing with regard to 𝜑
𝑛
is merely a complex-valued rotation of the standard Hermite function. The
⟨1⟩
𝑛 . Shifting the variable of integration,
√
⟨1⟩
𝑛 (𝑥, 𝜌) =
𝜑
i𝑛ei𝜌𝑥
√
2𝜋
∞
∫
−∞
̃H𝑛(𝜉)
e−𝜉2
1 + (𝜎 + 𝜉)2
ei𝑥𝜉d𝜉.
(79)
On the face of it, we recover an expression similar to (17), except that 𝑣(𝜉) = 1 + (𝜎 + 𝜉)2 is not
an even function and does not define a Sobolev inner product.
Figure 8 displays the absolute and real values of the complex-valued functions 𝜑
⟨1⟩
𝑛 .
24
ISERLES and WEBB
F I G U R E 8
for 𝑛 = 0, … , 5
Shifted Sobolev–Hermite functions: |𝜑
⟨1⟩
𝑛 (𝑥, 1)| in thicker line and darker color and ℜ𝜑
⟨1⟩
𝑛 (𝑥, 1),
7.2
The Laguerre weight
7.2.1
Sobolev–Laguerre functions of first kind
⟨0⟩
Let 𝑤(𝜉) = e−𝜉𝜒[0,∞)(𝜉), a Laguerre weight. Thus, the 𝜑
𝑛 s are Malmquist–Takenaka functions
(76), which we need to complement with their “reflections” for 𝑛 ∈ −ℕ to form a system dense
⟨1⟩
𝑛 s, 𝑛 ∈ ℤ+, with functions generated with
in L2(ℝ). By similar token, we need to complement 𝜑
𝑤(𝜉) = e𝜉𝜒(−∞,0](𝜉) (and indexed by 𝑛 ∈ −ℕ) to attain completeness in 𝐇1
2(ℝ).
It is possible to compute individual 𝜑
⟨1⟩
𝑛 s explicitly in terms of Bessel functions of the second
kind (a.k.a. Weber functions) Y𝑛
19(10.2.4) and Struve functions 𝐇𝑛
19(11.2.1) . We first let
𝑧 =
1
2
− i𝑥,
g(𝑧) = Y0(𝑧) − 𝐇0(𝑧).
(80)
The functions 𝜑𝑛 can be represented explicitly as linear combinations of derivatives of the
function 𝑔.
Lemma 5. The explicit form of the functions 𝜑
⟨1⟩
𝑛 is
⟨1⟩
𝑛 (𝑥) = −
𝜑
√
2𝜋
4
i𝑛
𝑛∑
𝓁=0
while 𝜑
⟨1⟩
−𝑛−1(𝑥) = (−1)𝑛+1𝜑
⟨1⟩
𝑛 (𝑥), 𝑛 ∈ ℤ+.
(
)
𝑛
𝓁
(
𝑔(𝓁)
1
2
1
𝓁!
)
− i𝑥
,
𝑛 ∈ ℤ+,
(81)
ISERLES and WEBB
Proof. Letting
𝜂𝑛(𝑥) =
1
√
2𝜋
∫
0
∞
𝜉𝑛 e−𝜉∕2+i𝑥𝜉
√
1 + 𝜉2
d𝜉,
𝑛 ∈ ℤ+,
we compute the generating function
∞∑
𝐺(𝑥, 𝑇) =
𝜂𝑛(𝑥)
𝑛!
𝑇𝑛 =
1
√
2𝜋
𝑛=0
√
∞
∫
0
)
1
√
1 + 𝜉2
(
(
exp
−
𝜉
2
)
+ 𝑇𝜉 + i𝑥𝜉
d𝜉
)]
[
(
Y0
1
2
− i𝑥 − 𝑇
− 𝐇0
1
2
− i𝑥 − 𝑇
.
= −
2𝜋
4
Therefore,
25
(82)
(83)
𝜂𝑛(𝑥) =
𝜕𝑛𝐺(𝑥, 𝑇)
𝜕𝑇𝑛
|
|
| 𝑇=0
= (−1)𝑛+1
√
2𝜋
4
(
𝑔(𝑛)
1
2
)
− i𝑥
,
𝑛 ∈ ℤ+.
(84)
Laguerre polynomials L𝑛 are orthonormal and19(18.5.12)
L𝑛(𝑥) =
𝑛∑
𝓁=0
(−1)𝓁
(
)
𝑛
𝓁
𝑥𝓁
𝓁!
,
𝑛 ∈ ℤ+
and it follows from Theorem 1 that
⟨1⟩
𝑛 (𝑥) =
𝜑
∞
i𝑛
√
2𝜋
∫
0
L𝑛(𝜉)
e−𝜉∕2+i𝑥𝜉
√
1 + 𝜉2
d𝜉 = i𝑛
𝑛∑
𝓁=0
(−1)𝓁
(
)
𝑛
𝓁
thereby proving (81) upon the substitution of the explicit form of 𝜂𝓁.
Extending this to 𝑛 ≤ −1 is trivial.
Corollary 1. The functions 𝜑
⟨1⟩
𝑛 for 𝑛 ∈ ℤ+ have the generating function
∞∑
𝑛=0
𝜑
⟨1⟩
𝑛 (𝑥)
𝑛!
𝑡𝑛 = −
√
2𝜋
4
ei𝑡
∞∑
𝓁=0
(i𝑡)𝓁
𝓁!2
(
𝑔(𝓁)
1
2
)
− i𝑥
.
𝜂𝓁(𝑥)
𝓁!
,
(85)
(86)
■
(87)
The proof is elementary, using (81). Moreover, Equation (87) can be easily extended to
)]
√
(
(
)
[
∞∑
𝑛=−∞
𝜑
⟨1⟩
𝑛 (𝑥)
|𝑛|!
𝜁𝑛 =
2𝜋
4
e−i𝜁−1
∞∑
𝓁=0
(−i𝜁−1)𝓁
𝓁!2
𝑔(𝓁)
1
2
+i𝑥
− ei𝜁
∞∑
𝓁=0
(i𝜁)𝓁
𝓁!2
𝑔(𝓁)
1
2
−i𝑥
,
(88)
which makes sense for |𝜁| = 1.
Since19(11.10.5&11.2.7)
𝑧2Y′′
𝑛 (𝑧) + 𝑧Y′
𝑛(𝑧) + (𝑧2 − 𝑛2)Y𝑛(𝑧) = 0,
26
ISERLES and WEBB
𝑧2𝐇′′
𝑛 (𝑧) + 𝑧𝐇′
𝑛(𝑧) + (𝑧2 − 𝑛2)𝐇𝑛(𝑧) =
𝑧𝑛+1
√
2𝑛−1
𝜋Γ(𝑛 +
,
1
2
)
it follows that 𝑔 obeys
𝑧𝑔′′(𝑧) + 𝑔′(𝑧) + 𝑧𝑔(𝑧) = −
√
2
𝜋Γ(
1
2
)
= −
2
𝜋
,
(89)
(90)
and we can express 𝑔(𝓁) as a linear combination of 𝑔 and 𝑔′ with rational coefficients. We do not
⟨𝑠⟩
𝑛 for 𝑠 ≥ 2 (or even the underlying orthogonal
pursue further this course of action. Functions 𝜑
polynomials 𝑝
⟨𝑠⟩
𝑛 ) are no longer available in an explicit form.
7.2.2
Sobolev–Laguerre functions of the second kind
An alternative is to consider the Sobolev–Laguerre cascade of the second kind. While the orthog-
onal polynomials 𝑝[𝑠]
𝓁=0 𝜉2𝓁 are unknown for 𝑠 ∈ ℕ, the
𝑛
underlying moments are trivial to compute and such polynomials can be generated at will. Also,
the computation of the 𝜑[𝑠]
𝑛 s does not present a problem: for example,
for the weight 𝑤𝑠(𝜉) = e−𝜉𝜒[0,∞)
∑𝑠
2
3𝜋
1
1 − 2i𝑥
,
[
i
−
2
87𝜋
4
1 − 2i𝑥
+
3(1 + 2i𝑥)
(1 − 2i𝑥)2
]
,
𝜑[1]
0 (𝑥) =
𝜑[1]
1 (𝑥) =
𝜑[1]
2 (𝑥) =
√
√
√
√
[
2
16211𝜋
i2
34
1 − 2i𝑥
[
−
40(1 + 2i𝑥)
(1 − 2i𝑥)2
+
29(1 + 2i𝑥)2
(1 − 2i𝑥)3
]
,
𝜑[1]
3 (𝑥) =
2
9812127𝜋
i3
−
480
1 − 2i𝑥
+
762(1 + 2i𝑥)
(1 − 2i𝑥)2
−
804(1 + 2i𝑥)2
(1 − 2i𝑥)3
+
559(1 + 2i𝑥)3
(1 − 2i𝑥)4
]
, (91)
and so on: all this seems very similar to (76) and for a good reason: for any 𝑠 ∈ ℤ+, we can expand
the relevant orthonormal polynomials in the Laguerre basis,
𝑝[𝑠]
𝑛 (𝑥) =
𝑛∑
𝑗=0
𝛾[𝑠]
𝑛,𝑗L𝑗(𝑥)
(cf. (28)), whereby it follows from (24) that
𝜑[𝑠]
𝑛 (𝑥) =
∞
𝑛∑
𝑗=0
𝛾[𝑠]
𝑛,𝑗√
2𝜋
∫
0
L𝑗(𝜉)e−𝜉∕2+i𝑥𝜉d𝜉 =
𝑛∑
𝑗=0
𝑛,𝑗𝜑[0]
𝛾[𝑠]
𝑗 (𝑥).
Note that the matrix {𝛾[𝑠]
Theorem 4. A similar construction applies also to 𝜑[𝑠]
𝑛,𝑗}𝑛,𝑗∈ℤ+ is the inverse of the banded connection matrix 𝐶[𝑠]
𝑛 for 𝑛 ≤ −1.
(92)
(93)
⊤
from
ISERLES and WEBB
27
𝑛 = ⟨𝑓, 𝜑[0]
The most remarkable feature of the Malmquist–Takenaka system is that the expansion coeffi-
cients ̂𝑓[0]
⟩ can be computed for −𝑁 + 1 ≤ 𝑛 ≤ 𝑁 in (𝑁 log 𝑁) operations using the
fast Fourier transform. By (31), however, once ̂𝒇[0] is known and assuming that the requisite
derivatives of 𝑓 are available, it costs just (𝑁) operations to compute
𝑛
𝑓[𝑠]
𝑛 =
𝑠∑
∞
∫
−∞
𝓁=0
(𝓁)
𝑓(𝓁)(𝑥)𝜑[𝑠]
𝑛
(𝑥)d𝑥,
−𝑁 + 1 ≤ 𝑛 ≤ 𝑁.
(The derivatives of 𝜑[𝑠]
(𝓁)
𝐶[𝑠]
𝝋[0]
−⊤
.) Altogether, the cost scales as (sN log 𝑁).
𝑛 s can be computed similarly to the functions themselves, 𝝋[𝑠]
𝑛
(94)
(𝓁)
=
8
CONCLUSION
In a sequence of previous papers,1–4 the current authors have sketched different aspects of an
overarching theory of L2-orthonormal systems on the real line with a tridiagonal differentiation
matrix. In this paper, we extend the framework to orthogonality with respect to Sobolev spaces.
Unlike in the case of orthogonal polynomials, where Sobolev orthogonality is of a completely dif-
ferent flavor to orthogonality with respect to a Borel measure,8–10 in our case, we can leverage
many elements of the “L2 theory” to a Sobolev setting: the connection to standard orthogonal
polynomials via a weighted Fourier transform, density in Paley–Wiener spaces, and fast compu-
tation of certain expansions. Other aspects of the theory are new, in particular, the existence of
two cascades, the latter of which can be ascended by banded triangular connection coefficients.
The work of this paper is a stepping stone toward the development of spectral methods on the
real line that respect a wide range of invariants that can be expressed as conservation of a variable-
weight Sobolev norm: a couple of examples have been given in Section 1. We expect to return to
this issue in a forthcoming paper.
A C K N O W L E D G M E N T S
The authors are grateful for very useful and enlightening correspondence with Enno Diekma, Erik
Koelink, and Tom Koornwinder. We gratefully acknowledge the partial support by the Simons
Foundation Award No 663281 granted to the Institute of Mathematics of the Polish Academy
of Sciences for the years 2021–2023. MW acknowledges support by Computational Mathemat-
ics in Quantum Mechanics, Grant of the National Science Centre (SONATA-BIS), project no.
2019/34/E/ST1/00390.
Special thanks are due to the two referees, whose detailed and erudite reports have helped a
great deal in considerably improving the quality of this paper.
D A T A AVA I L A B I L I T Y S T A T E M E N T
Data sharing not applicable to this article as no datasets were generated or analyzed during the
current study.
R E F E R E N C E S
1. Iserles A, Webb M. Orthogonal systems with a skew-symmetric differentiation matrix. Found Comput Math.
2019;19(6):1191-1221.
28
ISERLES and WEBB
2. Iserles A, Webb M. A family of orthogonal rational functions and other orthogonal systems with a
skew-Hermitian differentiation matrix. J Fourier Anal Appl. 2020;26(1):Paper No. 19.
3. Iserles A, Webb M. Fast computation of orthogonal systems with a skew-symmetric differentiation matrix.
Comm Pure Appl Math. 2021;74(3):478-506.
4. Iserles A, Webb M. A differential analogue of Favard’s theorem. In: From Operator Theory to Orthogonal Poly-
nomials, Combinatorics, and Number Theory—A Volume in Honor of Lance Littlejohn’s 70th Birthday, vol. 285
of Oper. Theory Adv. Appl. Birkhäuser/Springer; 2021:239-263.
5. Da Prato G, Zabczyk J. Stochastic Equations in Infinite Dimensions, vol. 152 of Encyclopedia of Mathematics
and its Applications. 2nd ed. Cambridge University Press; 2014.
6. Lawler GF. Introduction to Stochastic Processes. 2nd ed. Chapman & Hall/CRC; 2006.
7. Hesthaven JS, Gottlieb S, Gottlieb D. Spectral Methods for Time-Dependent Problems. vol. 21. Cambridge
University Press, Cambridge; 2007.
8. Iserles A, Koch PE, Nørsett SP, Sanz-Serna JM. On polynomials orthogonal with respect to certain Sobolev
inner products. J Approx Theory. 1991;65(2):151-175.
9. Marcellán F, Alfaro M, Rezola ML. Orthogonal polynomials on Sobolev spaces: old and new directions. J
Comput Appl Math. 1993;48(1–2):113-131.
10. Marcellán F, Xu Y. On Sobolev orthogonal polynomials. Expo Math. 2015;33(3):308-352.
11. Diekema E, Koornwinder TH. Differentiation by integration using orthogonal polynomials, a survey. J Approx
Theory. 2012;164(5):637-667.
12. Mantica G. Fourier-Bessel functions of singular continuous measures and their many asymptotics. Electron
Trans Numer Anal. 2006;25:409-430.
13. Gautschi W. Orthogonal Polynomials: Computation and Approximation. Oxford University Press; 2004.
14. Lubinsky DS, Mhaskar HN, Saff EB. A proof of Freud’s conjecture for exponential weights. Constr Approx.
1988;4(1):65-83.
15. Golub GH, Meurant G. Matrices, Moments and Quadrature with Applications. Princeton University Press; 2009.
16. Ismail MEH. Classical and Quantum Orthogonal Polynomials in One Variable. vol. 98 of Encyclopedia of Math-
ematics and its Applications. Cambridge University Press, Cambridge; 2005. With two chapters by Walter Van
Assche, With a foreword by Richard A. Askey.
17. Hendriksen E, van Rossum H. Semiclassical orthogonal polynomials. In: Orthogonal Polynomials and
Applications (Bar-le-Duc, 1984). vol. 1171 of Lecture Notes in Mathematics. Springer; 1985:354-361.
18. García-Ardila JC, Marcellán F, Marriaga ME. From standard orthogonal polynomials to Sobolev orthogonal
polynomials: the role of semiclassical linear functionals. In: AIMS-Volkswagen Stiftung Workshops. Springer;
2018:245-292.
19. Olver FWJ, Lozier DW, Boisvert RF, Clark CW, eds. NIST Handbook of Mathematical Functions. U.S. Depart-
ment of Commerce, National Institute of Standards and Technology; Cambridge University Press; 2010. With
1 CD-ROM (Windows, Macintosh and UNIX).
20. Townsend A, Webb M, Olver S. Fast polynomial transforms based on Toeplitz and Hankel matrices. Math
Comp. 2018;87(312):1913-1934.
21. Olver S, Slevinsky RM, Townsend A. Fast algorithms using orthogonal polynomials. Acta Numerica.
2020;29:573-699.
How to cite this article: Iserles A, Webb M. Sobolev-orthogonal systems with
tridiagonal skew-Hermitian differentiation matrices. Stud Appl Math. 2022;1-28.
https://doi.org/10.1111/sapm.12544
| null |
10.1073_pnas.2221809120.pdf
|
Data, Materials, and Software Availability. All study data are included in the
article and/or SI Appendix. Sequencing data are available through the National
Center for Biotechnology Information Gene Expression Omnibus, accession num-
ber GSE234805 (
| null |
RESEARCH ARTICLE | MEDICAL SCIENCES
OPEN ACCESS
Proxalutamide reduces SARS- CoV- 2 infection and associated
inflammatory response
Yuanyuan Qiaoa,b,c,1
Abhijit Paroliaa, Tongchen Hea, Caleb Chenga, Xuhong Caoa, Rui Wanga
Qianxiang Zhoug, Liandong Mag, Jonathan Z. Sextond,e,h,i,j, and Arul M. Chinnaiyana,b,c,k,l,2
, Charles J. Zhangd, Yuping Zhanga,b, Xia Jianga, Carla D. Prettoe, Sanjana Eyunnia,
, Fengyun Sua, Stephanie J. Ellisona, Yini Wangf
, Jesse W. Wotringd,1
, Yang Zhenga,1
, Jun Qinf, Honghua Yang,
Contributed by Arul M. Chinnaiyan; received December 23, 2022; accepted June 12, 2023; reviewed by Thirumala- Devi Kanneganti and Amy Moran
Early in the COVID- 19 pandemic, data suggested that males had a higher risk of devel-
oping severe disease and that androgen deprivation therapy might be associated with
protection. Combined with the fact that TMPRSS2 (transmembrane serine protease 2),
a host entry factor for the SARS- CoV- 2 virus, was a well- known androgen- regulated
gene, this led to an upsurge of research investigating androgen receptor (AR)- targeting
drugs. Proxalutamide, an AR antagonist, was shown in initial clinical studies to benefit
COVID- 19 patients; however, further validation is needed as one study was retracted.
Due to continued interest in proxalutamide, which is in phase 3 trials, we examined
its ability to impact SARS- CoV- 2 infection and downstream inflammatory responses.
Proxalutamide exerted similar effects as enzalutamide, an AR antagonist prescribed for
advanced prostate cancer, in decreasing AR signaling and expression of TMPRSS2 and
angiotensin- converting enzyme 2 (ACE2), the SARS- CoV- 2 receptor. However, proxal-
utamide led to degradation of AR protein, which was not observed with enzalutamide.
Proxalutamide inhibited SARS- CoV- 2 infection with an IC50 value of 97 nM, compared
to 281 nM for enzalutamide. Importantly, proxalutamide inhibited infection by mul-
tiple SARS- CoV- 2 variants and synergized with remdesivir. Proxalutamide protected
against cell death in response to tumor necrosis factor alpha and interferon gamma, and
overall survival of mice was increased with proxalutamide treatment prior to cytokine
exposure. Mechanistically, we found that proxalutamide increased levels of NRF2, an
essential transcription factor that mediates antioxidant responses, and decreased lung
inflammation. These data provide compelling evidence that proxalutamide can prevent
SARS- CoV- 2 infection and cytokine- induced lung damage, suggesting that promising
clinical data may emerge from ongoing phase 3 trials.
proxalutamide | SARS- CoV- 2 | COVID- 19 | androgen receptor | cytokines
Over 3 y have passed since the first documented cases of COVID- 19 arose from infec-
tion by the severe acute respiratory syndrome coronavirus 2 (SARS- CoV- 2), yet many
challenges remain worldwide in preventing and treating the disease (1). Robust vac-
cination campaigns led to rapid development, testing, and deployment of several
vaccines effective against infection and serious illness from the initial SARS- CoV- 2
genetic lineages (2–6). However, as the pandemic continued, waning vaccine protection
and emergence of new variants have led to breakthrough infections, as well as many
people now having been infected multiple times (5–9). Booster vaccines, including
bivalent boosters effective against the highly transmissible omicron variant, have been
developed in an effort to overcome these challenges (10). Oral antivirals such as mol-
nupiravir and nirmatrelvir–ritonavir have been developed for high- risk individuals
who contract COVID- 19, but these are also met with obstacles like potential recurrent
infections or contraindications with other commonly prescribed drugs (11–14).
Together, these challenges highlight the ongoing critical need for new therapeutics to
combat SARS- CoV- 2.
As it is the initial step in the viral life cycle, the entry process has been intensely studied
to understand how to potentially block SARS- CoV- 2 infection (15). Early data during
the pandemic showed that the spike (S) protein of SARS- CoV- 2 binds to host
angiotensin- converting enzyme 2 (ACE2) receptors on the cell surface to initiate entry
(16, 17). Cleavage of the spike protein by transmembrane serine protease 2 (TMPRSS2)
facilitates fusion of the viral and cell membranes and cell entry (18, 19). With the presumed
advantage that it will be difficult for the virus to mutate and evade host- directed drugs,
multiple preclinical and clinical research efforts have since followed examining the efficacy
of therapies directly targeting TMPRSS2 and ACE2, albeit with mixed results and several
studies still ongoing (20–25).
Significance
Drugs that target androgen
receptor (AR) signaling, including
those that inhibit production of
androgen ligands (degarelix) and
those that bind to and directly
block AR activity (enzalutamide),
have been investigated in clinical
trials for the treatment of
COVID- 19 but failed to produce
positive results. Another AR
antagonist, proxalutamide, is in
ongoing phase 3 studies for
COVID- 19 after showing initial
positive findings. Data from this
study show that proxalutamide
can inhibit infection of multiple
variants of SARS- CoV- 2 in vitro.
These data suggest that
proxalutamide should continue
to be investigated in clinical
studies as a potential therapy for
COVID- 19.
Author contributions: Y.Q., J.W.W., Y. Zheng, J.Z.S., and
A.M.C. designed research; Y.Q., J.W.W., Y. Zheng, C.J.Z.,
Y. Zhang, X.J., C.D.P., S.E., A.P., T.H., C.C., X.C., R.W., F.S.,
Y.W., J.Q., H.Y., Q.Z., L.M., and J.Z.S. performed research;
Y.Q., J.W.W., Y. Zheng, C.J.Z., Y. Zhang, X.J., C.D.P., S.E.,
A.P., T.H., C.C., X.C., R.W., F.S., Y.W., J.Q., H.Y., Q.Z.,
L.M., J.Z.S., and A.M.C. analyzed data; and Y.Q., J.W.W.,
Y. Zheng, S.J.E., J.Z.S., and A.M.C. wrote the paper.
Reviewers: T.- D.K., St. Jude Children’s Research Hospital;
and A.M., Oregon Health and Science University.
Competing interest statement: H.Y., Q.Z., and L.M. are
part of Kintor Pharmaceutical Limited. The remaining
authors declare no competing interests.
Copyright © 2023 the Author(s). Published by PNAS.
This open access article is distributed under Creative
Commons Attribution License 4.0 (CC BY).
1Y.Q., J.W.W., and Y. Zheng contributed equally to this
work.
2To whom correspondence may be addressed. Email:
[email protected].
This article contains supporting information online at
https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.
2221809120/- /DCSupplemental.
Published July 17, 2023.
PNAS 2023 Vol. 120 No. 30 e2221809120
https://doi.org/10.1073/pnas.2221809120 1 of 10
Since TMPRSS2 is a well- known androgen receptor (AR)- regulated
gene, early hypotheses suggested that inhibition of AR activity could
be a potential treatment strategy for COVID- 19 (26). As demo-
graphic data became available, many reports also observed that males
had higher incidences of severe SARS- CoV- 2 infections that required
intensive care unit (ICU) admission or resulted in death (27–29).
In further support of the initial hypothesis that AR activity may
drive COVID- 19 pathogenesis, a retrospective study during the
first months of the pandemic observed a reduced incidence of
SARS- CoV- 2 infections in prostate cancer patients taking androgen
deprivation therapy (ADT) compared to those not receiving ADT
(30). Other small studies supported this observation and the prem-
ise that anti- androgens could be protective against severe COVID- 19
(31, 32); however, these results quickly became debated as other
studies found no association between ADT and SARS- CoV- 2 infec-
tivity (33–35).
These preliminary observations prompted a burst of basic science
and clinical studies to attempt to elucidate the role of androgens
in SARS- CoV- 2 infection and determine whether AR inhibitors
could be viable treatment options for COVID- 19. Studies with AR
antagonists prescribed for prostate cancer treatment (e.g., enzalu-
tamide, apalutamide, and darolutamide) have since shown that
SARS- CoV- 2 infectivity can be decreased in vitro in certain con-
texts with these drugs (36–38). However, some randomized, con-
trolled clinical trials of AR inhibition in COVID- 19 patients have
not produced encouraging results. For instance, in the Hormonal
Intervention for the Treatment in Veterans with COVID- 19
Requiring Hospitalization (HITCH) trial (NCT04397718) which
tested degarelix, a gonadotropin- releasing hormone (GnRH) antag-
onist that rapidly suppresses testosterone levels, in male veterans
hospitalized with COVID- 19, no improvement in clinical out-
come was observed compared to placebo (39). Similarly, the
COVIDENZA trial (NCT04475601) found no improvement
in outcome of COVID19- positive male or female patients who
were randomized to treatment with enzalutamide vs. standard
of care (40).
In contrast, the AR antagonist proxalutamide was also tested
as a possible treatment for COVID- 19 in randomized, controlled
trials and showed encouraging positive benefits (41–43), but these
findings were met with caution from the scientific community
after a retraction statement was issued for one of the publications,
citing concerns over randomization (44). Proxalutamide is cur-
rently in additional phase 3 trials for COVID- 19 in both outpa-
tient (NCT04870606 and NCT04869228) and hospital
(NCT05009732) settings in different countries, including the
United States. Proxalutamide was originally developed as an AR
antagonist for advanced prostate cancer and is in ongoing phase
2 clinical trials for this indication as well (45–47). Our previous
study found that AR antagonists (enzalutamide, apalutamide, and
darolutamide) and degraders decreased TMPRSS2 and ACE2
expressions and were potent inhibitors of SARS- CoV- 2 infectivity
in vitro (37). Given these data and the continued clinical interest
surrounding proxalutamide in COVID- 19, we sought to test prox-
alutamide for its ability to impact SARS- CoV- 2 infection. We find
that proxalutamide inhibits cellular infection by multiple
SARS- CoV- 2 variants and shows synergistic activity in vitro with
remdesivir, an antiviral demonstrated to have clinical benefit in
COVID- 19 patients (48, 49). Additionally, in vivo studies showed
that prophylactic treatment with proxalutamide can improve over-
all survival in mouse models of the TNFα (tumor necrosis factor
alpha) and IFNγ (interferon gamma)- induced cytokine storm
triggered by SARS- CoV- 2 infection (50), potentially occurring
through increases in the nuclear factor erythroid 2- related factor
2 (NRF2) transcription factor responsible for mediating cellular
antioxidant responses. Altogether, this study provides characteri-
zation of proxalutamide in SARS- CoV- 2 infection models and
provides data to possibly explain positive results that may emerge
from clinical trials of proxalutamide for COVID- 19 treatment.
Results
Proxalutamide is an AR antagonist recently developed for castration-
resistant prostate cancer (CRPC) (47), in comparison to enzalu-
tamide which has been commonly prescribed for CRPC treatment
for several years (51). To first compare the transcriptomic changes
associated with proxalutamide and enzalutamide, RNA- sequencing
(RNA- Seq) analysis was carried out in AR- positive prostate cancer
Lymph Node Carcinoma of the Prostate (LNCaP) cells using
either 20 µM proxalutamide or enzalutamide for 8 h of treatment.
Gene set enrichment analysis was achieved by examining differ-
entially expressed genes in either proxalutamide- or enzalutamide-
treated cells compared to control. The normalized enrichment
score results indicated that androgen responses were the top down-
regulated hallmark in both proxalutamide- and enzalutamide-
treated LNCaP cells (Fig. 1A). Gene set enrichment analysis on
androgen responses further confirmed that proxalutamide signif-
icantly down- regulated androgen- regulated genes that were sup-
pressed by enzalutamide (Fig. 1B), suggesting proxalutamide
suppresses AR signaling. In addition, the effect of proxalutamide
on cell proliferation was examined in LNCaP cells and a castration-
resistant variant of LNCaP called C4- 2B cells. In both LNCaP
and C4- 2B cells, proxalutamide and enzalutamide treatment
resulted in dose- dependent inhibition of cell proliferation in vitro,
but growth inhibition was greater with proxalutamide treatment
compared to enzalutamide at the same concentrations (Fig. 1 C
and D). Importantly, we found that proxalutamide not only sup-
pressed AR signaling but also decreased AR protein levels, which
were not altered by enzalutamide treatment (Fig. 1E), indicating
that proxalutamide possesses stronger inhibition of the AR sign-
aling pathway than enzalutamide. Previously, we reported that
enzalutamide can transcriptionally down- regulate SARS- CoV- 2
entry factors TMPRSS2 and ACE2 (37). Here, we found that
proxalutamide had the same ability to decrease TMPRSS2 and
ACE2 (Fig. 1F). Thus, we postulated that proxalutamide may
block SARS- CoV- 2 infection.
Employing a SARS- CoV- 2 bioassay platform, we have established
an in vitro system with which to examine the various strains of
authentic SARS- CoV- 2 viral infection (37, 52). In this system, cells
were pretreated with the experimental compounds for 24 h prior to
SARS- CoV- 2 infection for an additional 72 h (Fig. 2A). The results
showed that proxalutamide decreased cellular infection by the WA1
strain of SARS- CoV- 2 in a dose- dependent manner with an IC50
value of 97 nM, whereas enzalutamide decreased infectivity with an
IC50 value of 281 nM (Fig. 2B). Representative images of cellular
infectivity by the WA1 strain of SARS- CoV- 2 in control- ,
proxalutamide- , or enzalutamide- treated conditions confirmed that
decreased infection could be achieved by the AR antagonists prox-
alutamide and enzalutamide in LNCaP cells (Fig. 2C). Since several
variants of the SARS- CoV- 2 virus have emerged throughout the
pandemic, we examined the effect of proxalutamide against infec-
tion of multiple strains. The results indicated that proxalutamide
possessed robust inhibitory effects in blocking SARS- CoV- 2 infec-
tion by the most common strains, including WA1, alpha, delta, and
omicron, with IC50 values of 69 nM, 48 nM, 98 nM, and 581 nM,
respectively, in LNCaP cells (Fig. 2D).
Furthermore, remdesivir is a Food and Drug Administration
(FDA)- approved agent for treatment of SARS- CoV- 2 infection
(48, 49). The combinatorial effect of proxalutamide or enzalutamide
2 of 10 https://doi.org/10.1073/pnas.2221809120
pnas.org
A
Enza
*
*
Proxa
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
HALLMARK_INFLAMMATORY_RESPONSE
HALLMARK_APOPTOSIS
HALLMARK_CHOLESTEROL_HOMEOSTASIS
HALLMARK_MTORC1_SIGNALING
HALLMARK_MYOGENESIS
HALLMARK_HEME_METABOLISM
HALLMARK_PI3K_AKT_MTOR_SIGNALING
HALLMARK_UNFOLDED_PROTEIN_RESPONSE
HALLMARK_COMPLEMENT
HALLMARK_TNFA_SIGNALING_VIA_NFKB
NES
2
1
0
−1
−2
HALLMARK_P53_PATHWAY
HALLMARK_HYPOXIA
HALLMARK_INTERFERON_GAMMA_RESPONSE
HALLMARK_ESTROGEN_RESPONSE_EARLY
HALLMARK_G2M_CHECKPOINT
HALLMARK_E2F_TARGETS
HALLMARK_MYC_TARGETS_V1
HALLMARK_MYC_TARGETS_V2
HALLMARK_ANDROGEN_RESPONSE
t
n
u
o
C
l
l
e
C
4×10 4
3×10 4
2×10 4
1×10 4
0
C
E
LNCaP
Enzalutamide
Proxalutamide
1
0
0
0
.
0
<
p
0 24 48 72 96
144 0 24 48 72 96
120
144
120
LNCaP
D
t
n
u
o
C
l
l
e
C
7×10 4
6×10 4
5×10 4
4×10 4
3×10 4
2×10 4
1×10 4
0
30 µM
10 µM
3.33 µM
1.11 µM
Ctrl
1
0
0
0
.
0
<
p
Hours
Proxalutamide
Enzalutamide
0
5
10
20
0
5
10
20
µM
AR
PSA
GAPDH
y
t
i
s
n
e
t
n
i
d
n
a
b
H
D
P
A
G
R
A
/
125
100
75
50
25
0
Proxalutamide
Enzalutamide
3
8
2
0
.
0
=
p
5
0
10 15 20
Concentration (μM)
B
HALLMARK_ANDROGEN_RESPONSE_ Proxalutamide
0.0
−0.2
−0.4
e
r
o
c
s
t
n
e
m
h
c
i
r
n
e
NES = −1.56
p.adj = 0.009
−0.6
0
5000
rank
10000
HALLMARK_ANDROGEN_RESPONSE_Enzalutamide
0.00
e
r
o
c
s
t
n
e
m
h
c
i
r
n
e
−0.25
−0.50
−0.75
NES = −1.99
p.adj = 0.007
0
5000
rank
10000
C4-2B
Enzalutamide
Proxalutamide
3
0
0
0
.
0
=
p
30 µM
10 µM
3.33 µM
1.11 µM
Ctrl
1
0
0
0
.
0
<
p
Hours
0 24 48 72 96
F
120
144 0 24 48 72 96
120
144
ACE2
TMPRSS2
p<0.0001
e
g
n
a
h
c
d
o
f
l
1.25
1.00
0.75
A
N
R
m
e
v
i
t
l
a
e
R
0.50
0.25
0.00
0.5(cid:31)M
Ctrl
3(cid:31)M
1(cid:31)M
3(cid:31)M
Enzalutamide
Proxalutamide
Proxalutamide
ARD61
e
g
n
a
h
c
d
o
l
f
A
N
R
m
e
v
i
t
a
e
R
l
1.25
1.00
0.75
0.50
0.25
0.00
1
0
0
0
.
0
<
p
3
0
0
0
.
0
=
p
1
0
0
0
.
0
<
p
1
0
0
0
.
0
<
p
0.5(cid:31)M
Ctrl
1(cid:31)M
3(cid:31)M
3(cid:31)M
Proxalutamide
Enzalutamide
Proxalutamide
ARD61
Fig. 1. Proxalutamide is a recently developed AR antagonist that also down- regulates AR protein levels. (A) Hallmark of differential expressed gene signatures
in proxalutamide (Proxa) and enzalutamide (Enza) treatment vs. control in LNCaP cells; the asterisk indicates a P value of less than 0.01. (B) Gene set enrichment
of the androgen response pathway in proxalutamide- or enzalutamide- treated LNCaP cells. (C) Cell growth inhibition in enzalutamide- or proxalutamide- treated
LNCaP cells. Ctrl, control. P values were calculated by the two- tailed unpaired t test between ctrl and 30 µM enzalutamide or proxalutamide (not between each
dose). (D) Cell growth inhibition in enzalutamide- or proxalutamide- treated C4- 2B cells. Ctrl, control. P values were calculated by the two- tailed unpaired t test
between ctrl and 30 µM enzalutamide or proxalutamide (not between each dose). (E) Immunoblotting of AR and PSA protein in LNCaP cells after treatment
with various concentrations of proxalutamide and enzalutamide for 24 h. Quantification of band intensity of AR/GAPDH is shown on the right. P values were
calculated by the two- tailed unpaired t test between 20 µM proxalutamide and enzalutamide. (F) Relative mRNA expression of ACE2 and TMPRSS2 in LNCaP cells
after the indicated treatment. P values were calculated by the two- tailed unpaired t test between control and the indicated treatment.
and remdesivir in preventing infection by the SARS- CoV- 2 alpha
strain was examined in induced human alveolar cells (iAEC2)
(Fig. 3A). The results indicated that proxalutamide had a strong
synergistic effect with remdesivir in inhibition of alpha strain
infection and achieved 100% protection against infection (Fig. 3B),
with a synergy score of 14.516 (Fig. 3C). Similarly, the enzalutamide
and remdesivir combination achieved synergy but with a slightly
weaker synergistic effect than the proxalutamide and remdesivir
PNAS 2023 Vol. 120 No. 30 e2221809120
https://doi.org/10.1073/pnas.2221809120 3 of 10
SARS-CoV2 Bioassay
Day 0
Day 1
Day 2
Day 4
Seed LNCaP cells
in 384 well plate
Pre-incubate
compounds
for 24 hours
Infect with
SARS-CoV-2 virus
Fix, Stain, Image
Proxalutamide
Enzalutamide
%
n
o
i
t
c
e
f
n
I
150
100
50
0
IC50: 97 nM
150
100
50
0
V
i
a
b
i
l
i
t
y
%
%
n
o
i
t
c
e
f
n
I
150
100
50
0
IC50: 281 nM
10 -8
10 -7
Concentration (M)
10 -6
10 -5
10 -9
10 -8
10 -7
Concentration (M)
10 -6
150
100
50
0
V
i
a
b
i
l
i
t
y
%
10 -5
3 µM
750 nM
188 nM
23 nM
Viral Control
A
B
C
D
10 -9
i
e
d
m
a
t
u
a
x
o
r
P
l
i
e
d
m
a
t
u
a
z
n
E
l
150
100
50
%
n
o
i
t
c
e
f
n
I
LNCaP
IC50 WA1: 69 nM
IC50 Alpha: 48 nM
IC50 Delta: 98 nM
IC50 Omicron: 581 nM
0
10 -9
10 -8
10 -7
Proxalutamide [M]
10 -6
150
100
50
%
v
i
a
b
i
l
i
t
y
0
10 -5
Viability
WA1 strain
Alpha strain
Delta strain
Omicron strain
Fig. 2. Proxalutamide inhibits multiple strains of SARS- CoV- 2 infection in vitro. (A) Schematic illustration of the SARS- CoV- 2 bioassay. (B) Dose- dependent
inhibition of SARS- CoV- 2 WA1 strain infection by proxalutamide and enzalutamide in LNCaP cells with IC50 values shown for each. Cell viability is also graphed.
(C) Representative images of SARS- CoV- 2 WA1 strain infection after proxalutamide or enzalutamide treatment in LNCaP cells. (D) Dose- dependent inhibition of
infection by multiple strains of SARS- CoV- 2 with proxalutamide treatment in LNCaP cells.
combination (Fig. 3 E and F). Both proxalutamide or enzalutamide
and remdesivir combination treatments had no detrimental effects
on the viability of iAEC2 cells (Fig. 3 D and G). These results suggest
that proxalutamide may have clinical utility in combination with
current SARS- CoV- 2 treatments, such as remdesivir.
SARS- CoV- 2- induced mortality is largely triggered by a cytokine
storm that occurs in the pulmonary system and systemically (53). It
has been reported that TNFα and INFγ can act synergistically to
trigger inflammatory cell death in vitro and in vivo, which mimics
the SARS- CoV- 2- induced cytokine shock syndrome (CSS) that
4 of 10 https://doi.org/10.1073/pnas.2221809120
pnas.org
A
B
)
M
n
(
r
i
v
i
s
e
d
m
e
R
E
)
M
n
(
r
i
v
i
s
e
d
m
e
R
Inhibition of SARS-CoV-2 infection
300
56.99 86.94 82.92 83.59 88.73 97.83 100.00
100
45.09 43.77 57.88 55.17 52.51 88.21 99.06
30
10
0
32.64 37.43 39.69 40.22 64.38 74.19 96.33
11.94 14.53 19.55 25.93 25.47 63.42 95.70
0.00
6.01
6.94
4.58
10.57 56.08 82.77
0
30
100
1000
300
Proxalutamide (nM)
3000 10000
C
F
Inhibition of SARS-CoV-2 infection
300
56.99 66.23 70.50 73.50 79.06 92.28
100
45.09 63.56 71.99 83.04 86.73 94.98
30
10
0
32.64 64.21 69.54 79.51 85.20 97.75
11.94 44.63 55.47 65.48 85.39 95.91
0.00
1.49
30.45 62.57 79.61 97.95
0
30
100
300
1000
3000
Enzalutamide (nM)
D
Viability
300
93.44 124.35 114.51 129.68 104.44 118.83 106.49
100
92.50 105.26 113.44 117.08 91.37 113.52 122.49
30
101.20 102.79 92.72 109.47 97.41 134.10 120.49
10
107.50 121.99 131.75 109.71 109.68 112.65 117.08
0
100.00 132.56 116.73 128.47 148.89 121.12 97.83
0
30
100
1000
300
Proxalutamide (nM)
3000 10000
G
Viability
300
93.44 99.96 106.75 103.63 101.60 98.53
100
92.50 105.09 90.65 91.27 86.04 88.77
30
101.20 106.26 108.09 95.80 95.52 100.91
10
107.50 136.12 108.35 101.32 112.93 103.38
0
100.00 97.32 90.03 93.52 149.88 108.82
0
30
100
300
1000
3000
Enzalutamide (nM)
)
M
n
(
r
i
v
i
s
e
d
m
e
R
)
M
n
(
i
r
i
v
s
e
d
m
e
R
Fig. 3. Proxalutamide and remdesivir combination exerts strong synergistic effect in blocking SARS- CoV- 2 infection in iAEC2. (A) Schematic illustration of the
study design of the SARS- CoV- 2 bioassay on iAEC2 cells. (B) Combination matrix of proxalutamide and remdesivir in inhibition of SARS- CoV- 2 alpha strain infection.
(C) Bliss synergy score of proxalutamide and remdesivir against SARS- CoV- 2 alpha strain infection. (D) Combination matrix of cell viability on proxalutamide and
remdesivir. (E) Combination matrix of enzalutamide and remdesivir in inhibition of SARS- CoV- 2 alpha strain infection. (F) Bliss synergy score of enzalutamide
and remdesivir against SARS- CoV- 2 alpha strain infection. (G) Combination matrix of cell viability on enzalutamide and remdesivir.
occurs in COVID- 19 patients (50). Specifically, TNFα and INFγ
induce a type of inflammatory cell death called PANoptosis, which
is regulated by the PANoptosome and involves molecular compo-
nents of pyroptosis, apoptosis, and necroptosis (50, 54). In an
AR- positive lung cell line, H1437, we demonstrated that the com-
bination of TNFα and INFγ induced maximal cell death compared
to either cytokine alone (Fig. 4A). Interestingly, the cell death
induced by combination treatment with TNFα and INFγ was atten-
uated by proxalutamide and another AR antagonist darolutamide
in a dose- dependent manner (Fig. 4B) but not by enzalutamide or
apalutamide (SI Appendix, Fig. S1A). Additionally, the cell death
triggered by TNFα and INFγ combination treatment was confirmed
by elevated cleaved PARP (c- PARP) levels, which were dose
dependently blocked by proxalutamide and darolutamide (Fig. 4C)
but not enzalutamide or apalutamide (SI Appendix, Fig. S1B).
Similarly, AR protein levels were down- regulated by proxalutamide
and darolutamide (Fig. 4D) but not enzalutamide or apalutamide
(SI Appendix, Fig. S1C). This suggests that AR antagonists such as
proxalutamide or darolutamide may provide additional benefits in
terms of reducing CSS in vivo. In normal mouse prostate organoids,
we confirmed that proxalutamide inhibited murine AR signaling by
decreasing androgen (dihydrotestosterone, DHT)- stimulated induc-
tion of Fkbp5 and Psca target genes; additionally, proxalutamide
decreased Ar mRNA levels (SI Appendix, Fig. S2A). These results
prompted us to examine the in vivo efficacy of proxalutamide in
preventing death in the TNFα and INFγ CSS model (50) in
wild- type C57BL6 male mice. We tested two treatment regimens
of proxalutamide prior to cytokine challenge with the TNFα and
PNAS 2023 Vol. 120 No. 30 e2221809120
https://doi.org/10.1073/pnas.2221809120 5 of 10
A
B
C
H1437
lung adenocarcinoma
Ctrl
TNFα
IFNγ
TNFα+IFNγ
1500
1000
500
e
c
n
e
u
l
f
n
o
c
/
s
l
l
e
c
I
+
P
0
0
24
48
72
Hours of treatment
1500
1000
500
e
c
n
e
u
l
f
n
o
c
/
s
l
l
e
c
I
+
P
0
0
TNFα+IFNγ
DMSO
Proxalutamide 5μM
Proxalutamide 10μM
Proxalutamide 20μM
24
48
72
Hours of treatment
1
0
0
0
0
<
p
.
3
0
0
0
.
0
=
p
Ctrl
TNFα
IFNγ
TNFα+IFNγ
TNFα+IFNγ
DMSO
Darolutamide 10μM
Darolutamide 20μM
7
1
0
0
.
0
=
p
24
48
72
Hours of treatment
1500
1000
500
e
c
n
e
u
l
f
n
o
c
/
s
l
l
e
c
I
+
P
0
0
D
Darolutamide
Proxalutamide
TNFα+IFNγ
+ ++
-
-
-
+ ++
+ ++
+ ++
-
-
+ + + + +
Proxalutamide
Darolutamide
10
20
10
20
c-PARP
Vinculin
µM
µM
AR
Vinculin
Fig. 4. Proxalutamide attenuates CSS–related cell death and mortality. (A) Real- time analysis of cell death in H1437 cells in vitro under control, TNFα, IFNγ, or
combination treatment. Representative images of dead cells under the indicated conditions are shown on the Right. The P value was calculated by the two- tailed
unpaired t test between control and TNFα/IFNγ combination treatment. (B) Real- time analysis of cell death in H1437 cells in vitro under TNFα and IFNγ combination
and various concentrations of proxalutamide or darolutamide. P values were calculated by the two- tailed unpaired t test comparing dimethylsulfoxide (DMSO)
control and 20 µM proxalutamide or darolutamide. (C) Immunoblotting of c- PARP and vinculin (loading control) in H1437 cells after treatment with 10 and 20
µM of proxalutamide or darolutamide with or without TNFα and IFNγ combination for 72 h. (D) Immunoblotting of AR and vinculin in H1437 after treatment with
10 and 20 µM of proxalutamide or darolutamide for 72 h.
INFγ combination. The data showed that both proxalutamide treat-
ment regimens reduced mortality induced by TNFα and INFγ
(SI Appendix, Fig. S2 B and C). Histology evaluation of tissue dam-
age triggered by TNFα and INFγ combination was examined in the
small intestine and lung (SI Appendix, Fig. S2D). Compared with
the PBS treated group, atrophy of the villi and an increase in inflam-
matory cell infiltration in the lamina propria area of the intestine
were observed post- TNFα and IFNγ treatment, which was largely
alleviated with proxalutamide treatment. In addition, TNFα and
IFNγ treatment induced interlobular septal thickening in the lungs
of mice showing focal epithelial hyperplasia, and such effects were
rescued by proxalutamide treatment. Thus, these results suggest that
proxalutamide may reduce TNFα and IFNγ cytokine storm- induced
cell death in vitro and in vivo.
The NRF2 pathway is an important part of cellular defense
through the production of antioxidants, which occurs via binding
of the NRF2 transcription factor to antioxidant response elements
in target genes (55–57). The upregulation of NRF2 has been reported
to control inflammation in several studies (56–60). Here, we found
that proxalutamide increases NRF2 transcriptional activity by
enhancing NRF2 DNA binding in RAW264.7 and THP- 1 cells
(Fig. 5A). In RAW264.7 cells, proxalutamide also up- regulated
NRF2 protein expression in lipopolysaccharide (LPS)- stimulated
conditions (Fig. 5B). In the in vitro CSS model triggered by TNFα
and INFγ combination treatment, proxalutamide augmented NRF2
protein levels and decreased cell death in THP- 1 cells (Fig. 5 C and
D). Apoptotic cell death triggered by TNFα and INFγ combination
treatment was attenuated by proxalutamide (Fig. 5E). Next, we
6 of 10 https://doi.org/10.1073/pnas.2221809120
pnas.org
A
C
TNFα
IFNγ
F
G
B
Proxalutamide
LPS
RAW264.7
- 0.3 1 310
+
+
-
-
+
1.0 1.3 1.3 1.4 1.6 1.6
+
+
µM
NRF2
GAPDH
THP-1
(PMA induced macrophage)
DMSO
Proxalutamide
+
+
+
1.0 2.5 0.4 0.8 3.1 3.5 3.0 3.3
+
+
+
+
+
D
THP-1
1500
TNFα+IFNγ Proxalutamide
E
e
c
n
e
u
l
f
n
o
c
/
s
l
l
e
c
I
+
P
1000
500
0
0
NRF2
Vinculin
TNFα+IFNγ
PBS
Proxalutamide
TNFα+IFNγ
-
-
-
+
+
-
+
+
2
0
0
0
.
0
=
p
1
0
0
0
.
0
<
p
c-PARP
GAPDH
24
48
72
Hours of treatment
p<0.01
4
0
1
x
F
L
A
B
n
i
r
e
b
m
u
n
l
l
e
c
l
a
t
o
T
700
600
500
400
300
200
100
0
1
0
0
0
.
0
<
p
4
0
1
x
r
e
b
m
u
n
s
l
i
h
p
o
r
t
u
e
N
500
400
300
200
100
0
1
0
0
0
.
0
<
p
5
0
.
0
<
p
1
0
.
0
<
p
1
0
0
0
.
0
<
p
Normal-Vehicle
Poly (I:C) Model-Vehicle
Poly (I:C) Model-Dex+Roflumilast
Poly (I:C) Model-Proxalutamide 20mg/kg
Poly (I:C) Model-Proxalutamide 40mg/kg
Fig. 5. Proxalutamide enhances NRF2 transcriptional activity and inhibits acute immune response in the poly (I:C)- induced lung injury animal model. (A)
Proxalutamide increased NRF2 transcriptional activity in RAW264.7 and THP- 1 cells. (B) Immunoblotting of NRF2 protein in RAW264.7 cells with or without LPS
stimulation and indicated concentration of proxalutamide. GAPDH serves as a loading control. (C) Immunoblotting of NRF2 protein in THP- 1 cells with TNFα,
IFNγ, or combination, with or without 20 µM proxalutamide. Vinculin serves as a loading control. (D) Real- time analysis of cell death in THP- 1 cells in vitro treated
with the indicated cytokines. P values were calculated by the two- tailed unpaired t test between the indicated groups. (E) Immunoblotting of c- PARP and GAPDH
in THP- 1 cells after treatment with proxalutamide with or without TNFα and IFNγ combination for 72 h. (F) Schematic illustration of acute immune response in
poly (I:C)- induced acute lung injury animal model. (G) Total cell number and neutrophil cell counts in the bronchoalveolar lavage fluid (BALF) under indicated
treatment. P values were calculated by the two- tailed unpaired t test between the poly (I:C)- vehicle and indicated treatment.
examined proxalutamide in an acute lung injury animal model trig-
gered by poly(I:C), and combination dexamethasone and roflumilast
treatment was used as a positive control (Fig. 5F). In this model,
proxalutamide significantly reduced the total mononuclear cells and
neutrophils in alveolar lavage fluids from poly(I:C)- induced animals
(Fig. 5G). Together, our data show that proxalutamide up- regulates
NRF2 protein levels and decreases inflammation in the lungs induced
by poly(I:C), suggesting a possible benefit of proxalutamide against
SARS- CoV- 2- associated inflammatory responses and mortality in
COVID- 19 patients.
Discussion
Proxalutamide was initially developed as an AR antagonist that
could potentially have efficacy in CRPC patients, including those
that had developed resistance to existing AR- targeted therapies.
PNAS 2023 Vol. 120 No. 30 e2221809120
https://doi.org/10.1073/pnas.2221809120 7 of 10
Results from phase 1 testing in CRPC patients showed that prox-
alutamide was well tolerated, had a favorable pharmacokinetic
profile, and exhibited antitumor activity in select patients (47).
AR- targeting compounds became one of the initial groups of drugs
to be pursued as potential COVID- 19 treatments for the myriad
of reasons discussed in preceding sections. With phase 1 testing
complete, proxalutamide was positioned to be tested in the setting
of COVID- 19, along with other AR- targeted drugs that have been
FDA- approved for prostate cancer for years, such as enzalutamide.
Although positive results were reported for the initial COVID- 19
trials with proxalutamide, clarity is still needed as one of the stud-
ies was retracted last year (41–44). Here, we performed several
in vitro and in vivo assays assessing the activity of proxalutamide
against SARS- CoV- 2 infection and inflammatory responses. We
indeed demonstrate that proxalutamide decreases SARS- CoV- 2
infectivity in vitro, and the compound is active against several
strains of the virus (WA1, alpha, delta, and omicron). Synergy can
be obtained when proxalutamide is combined with remdesivir.
Interestingly, proxalutamide also increases levels of the NRF2
transcription factor.
It is well established that COVID- 19 can be associated with a
cytokine storm, a hyperactivation of the immune system that can
ultimately result in death (53). In this study, we employed two
in vivo lines of experimentation to analyze the effect of proxaluta-
mide on CSS and lung injury. Proxalutamide pretreatment in the
TNFα/IFNγ model of CSS (50) results in a modest increase in
overall survival (SI Appendix, Fig. S2 B and C), mirroring the atten-
uation of in vitro cell death observed with proxalutamide in the
H1437 and THP- 1 cell lines (Figs. 4B and 5D). Using poly(I:C)
that induces inflammatory responses in the lung similar to viral
infections (61), we observe that proxalutamide significantly decreases
total cell and neutrophil levels in BALF (bronchoalveolar lavage
fluid) (Fig. 5G). Altogether, results from these two in vivo models
suggest that proxalutamide can decrease CSS responses and lung
inflammation, but there are associated caveats to note. TNFα and
IFNγ induce PANoptosis in mice that leads to CSS and death, which
has been suggested to mimic severe COVID- 19 in patients (50).
However, TNFα/IFNγ- induced death in mice occurs within hours,
whereas death from acute respiratory distress syndrome (ARDS) in
COVID- 19 patients happens over a much longer time (62).
Additionally, studies have implicated alternative cytokines (e.g., IL- 6
and IL- 1) rather than just TNFα and IFNγ as the primary inducers
of ARDS in COVID- 19 (63). In terms of the poly(I:C) model, it
is prudent to also note that this is a model of lung injury, rather than
lung epithelial cell death. Finally, these in vivo experiments are mod-
els of the possible downstream effects of SARS- CoV- 2 and did not
directly involve animal infection with the virus. It is interesting to
note, however, that proxalutamide increases the DNA binding activ-
ity and expression of Nrf2, and Nrf2 has been shown to be an
essential factor for tempering the immune response and protecting
against sepsis (64, 65). A recent study also shows that SARS- CoV- 2
can inhibit Nrf2 signaling through one of its nonstructural proteins
(66). In line with our findings, Nrf2 agonists consequently inhibited
SARS- CoV- 2 replication (66).
Combined, the data in this study support the notion that proxal-
utamide has antiviral activity against SARS- CoV- 2 and suggest that
it could show positive clinical benefit in cases of COVID- 19, war-
ranting further clinical exploration. In comparison, as mentioned
above, clinical studies with degarelix (HITCH trial, NCT04397718)
and enzalutamide (COVIDENZA trial, NCT04475601) did not find
any improvements in clinical outcome with COVID- 19 (39, 40).
There are a multitude of explanations that could account for these
disparate findings from different AR- targeting drugs. Degarelix is a
GnRH antagonist that prevents release of follicle- stimulating hormone
and luteinizing hormone, thereby leading to suppression of testicular
testosterone release and a decrease in AR activity at the level of
ligand availability (67). In contrast, proxalutamide, like enzaluta-
mide, binds directly to the ligand- binding domain of AR to block
receptor activation (47, 68). As shown in Fig. 1, proxalutamide
and enzalutamide exert similar effects in LNCaP prostate cancer
cells—decreasing or activating similar signaling pathways, decreas-
ing androgen signaling, and decreasing cell proliferation. Relevant
to SARS- CoV- 2, both compounds decrease expression of host
entry receptors ACE2 and TMPRSS2 (Fig. 1F). However, certain
differences exist with these two compounds. For instance, a pre-
clinical report on proxalutamide reported a 3.4- fold higher bind-
ing affinity for AR compared to enzalutamide (47). As shown
here and previously (47), proxalutamide can also decrease AR
protein expression, while enzalutamide does not lead to AR deg-
radation (Fig. 1E). In the SARS- CoV- 2 bioassays, proxalutamide
exhibited increased potency in inhibiting infection compared to
enzalutamide (IC50 of 97 nM for proxalutamide and 281 nM
for enzalutamide, Fig. 2B) and a higher Bliss synergy score with
remdesivir (14.516 and 11.685 for proxalutamide and enzalut-
amide, respectively, Fig. 3). Furthermore, in the cell line models
of cytokine- mediated death with combined TNFα and IFNγ
treatment, addition of proxalutamide prevented cell death
(Fig. 4B), whereas enzalutamide was without effect, even at the
high dose of 20 µM (SI Appendix, Fig. S1A). These data show
that although proxalutamide and enzalutamide are both AR
antagonists, differences in their mechanisms of action exist.
However, since both compounds decrease ACE2 and TMPRSS2
expression and ultimately prevent SARS- CoV- 2 infectivity
in vitro (albeit with different IC50 values), further research is
needed to define the precise mechanisms that could account for
disparate clinical outcomes in COVID- 19 treatment.
Several phase 3 clinical trials of proxalutamide treatment for
COVID- 19, all sponsored by Kintor Pharmaceuticals, are ongoing
in different countries, and these studies should provide more definitive
answers as to its efficacy. One phase 3 randomized, placebo- controlled,
multiregional clinical trial of outpatients with mild or moderate
COVID- 19 (NCT04870606) primarily enrolled patients at centers
across the United States (99%) (69). Efficacy data showed that prox-
alutamide reduced the risk of hospitalization or death compared to
placebo, and proxalutamide continued to show a positive safety
profile (69). An additional outpatient clinical trial of males with
mild to moderate COVID- 19 in Brazil is ongoing (NCT04869228),
with the primary outcome being oxygen requirement at Day 28.
Finally, NCT05009732 is an ongoing phase 3 trial of proxaluta-
mide in hospitalized adults with COVID- 19 that has participating
locations across several countries, including the United States,
China, Philippines, and South Africa. The primary end point for
this study is time to clinical deterioration (need for ICU care,
mechanical ventilation, or mortality). The data presented in our
report suggest that proxalutamide can markedly decrease
SARS- CoV- 2 infectivity and associated inflammatory responses,
which could result in positive clinical benefit, and results from the
clinical studies above are eagerly awaited.
Methods
Cell Culture. LNCaP, RAW264.7, and THP- 1 cells were purchased from the
American Type Culture Collection (ATCC) and cultured in 5% CO2 at 37 °C in
medium as suggested by ATCC. iAEC2 cells [iPSC (SPC2 iPSC line, clone SPC2-
ST- B2, Boston University) derived alveolar epithelial type 2 cells] were maintained
as previously described (52). iAEC2 cells were also subcultured as previously
described (70). Cell lines underwent genotype authentication and were confirmed
to be negative for mycoplasma.
8 of 10 https://doi.org/10.1073/pnas.2221809120
pnas.org
SARS- CoV- 2 Bioassay. SARS- CoV- 2 isolates USA- WA1/2020, hCoV- 19/USA/OR- OHSU-
PHL00037/2021 (Lineage B.1.1.7; Alpha Variant), hCoV- 19/USA/MD- HP05285/2021
(Lineage B.1.617.2; Delta Variant), and hCoV- 19/USA/GA- EHC- 2811C/2021 (Lineage
B.1.1.529; Omicron Variant) were obtained from BEI resources and propagated in
VeroE6 cells (ATCC). Viral titers were established by TCID50 with the Reed and Muench
method. LNCaP or iACE2 cells were plated in 384- well plates and treated with increas-
ing concentrations of proxalutamide or enzalutamide for 24 h prior to SARS- CoV- 2 virus
infection in a Biosafety Level 3 facility. Cells were then incubated for 48 h postinfection
under culture conditions of 5% CO2 and 37°C. Assay plates were fixed, permeabilized,
and labeled with antinucleocapsid SARS- CoV- 2 primary antibody (Antibodies Online,
Cat. #: ABIN6952432) as previously described (52). The remaining of the assay pro-
ceeded as previously described (70).
Fluorescence Imaging and High- Content Analysis. A Thermo- Fisher CX5
high- content microscope with LED excitation (386/23 nm, 650/13 nm) at 10×
magnification was used to image assay plates. Nine fields per well were imaged
at a single Z- plane in these experiments. Imaging, processing, and normalization
were performed as previously described (70, 71).
Gene Expression Analysis. RNA was extracted from LNCaP cells treated with
DMSO, 20 µM proxalutamide, or enzalutamide for 8 h using a Qiagen RNA extrac-
tion kit. RNA quality was determined using a Bioanalyzer RNA Nano Chip. Poly- A
selection was performed with Sera- Mag Oligo(dT)- Coated Magnetic Particles
(38152103010150; GE Healthcare Life Sciences), and libraries were generated
using a KAPA RNA HyperPrep kit (KK8541; Roche Sequencing Solutions). RNA- seq
was performed on an Illumina HiSeq 2500. Reads were aligned with the Spliced
Transcripts Alignment to a Reference mapper to the human reference genome
gh38. Gene differential expression analysis was carried out with edgeR70.
Mouse Prostate Organoid Culture. Whole mouse prostate was dissected from
C57BL6J wild- type mice, and organoid culture was generated according to pre-
vious publication (72). Mouse prostate organoids were treated with 5 µM or 10
µM proxalutamide or enzalutamide for 16 h prior to 10 nM DHT stimulation for
8 h. Total RNA was extracted from organoid culture using the miRNeasy mini kit
(Qiagen), and cDNA was synthesized from 1 µg total RNA using the High- Capacity
cDNA Reverse Transcription Kit (Applied Biosystems). qPCR was performed
using fast SYBR green master mix on the QuantStudio Real- Time PCR Systems
(Applied Biosystems). The SYBR green primer sequences are Fkbp5 forward:
GATTGCCGAGATGTGGTGTTCG, Fkbp5 reverse: GGCTTCTCCAAAACCATAGCGTG; Psca for-
ward: GCACAGTTGCTTTACATCGCGC, Psca reverse: ACAGGTCAGAGTAGCAGCACGT; and
Ar forward: CCTTGGATGGAGAACTACTCCG, Ar reverse: TCCGTAGTGACAGCCAGAAGCT.
Immunoblotting. For western blotting analysis, cells were harvested and lysed
in Pierce RIPA buffer (Thermo Fisher) with added phosphatase (Millipore) and pro-
tease (Roche) inhibitor cocktails. Protein quantification, sodium dodecyl- sulfate
polyacrylamide gel electrophoresis, transfer, blocking, and antibody incubation
were performed as described previously (73), and protein signals were detected
with ECL Primer (Amersham) on a Li- Cor machine. Antibodies were used at dilu-
tions recommended by the manufacturer and consisted of the following: AR
(06- 680, Millipore), PSA (Dako), NRF2 (12721S, Cell Signaling Technology), and
GAPDH (3683S, Cell Signaling Technology).
Real- Time Imaging for Cell Death. The kinetics of cell death were determined
using the IncuCyte ZOOM (Essen BioScience) live- cell automated system. H1437
or THP- 1 cells (1 × 105 cells/well) were seeded in 24- well tissue culture plates. Cells
were treated with 50 ng/mL of human TNFα (Peprotech, AF- 300- 01A) and /or 100
ng/mL of human IFNγ (Peprotech, 300- 02) for the indicated time and stained with
1 µg/mL propidium iodide (PI) (Life Technologies, P3566) following the manufac-
turer’s protocol. The plate was scanned, and fluorescent and phase- contrast images
were acquired in real- time every 4 h. PI- positive dead cells are marked with a red
mask for visualization. The image analysis, masking, and quantification of dead
cells were done using the software package supplied with the IncuCyte imager.
In Vivo TNFα and IFNγ- Induced Inflammatory Shock. C57BL6J mice were pur-
chased from The Jackson Laboratory. Eight- to nine- week- old male C57BL6J mice were
given vehicle or 40 mg/kg proxalutamide by oral gavage either 2 h or once daily for 5 d
prior to cytokine injection. Cytokine combination of 10 μg TNFα (Preprotech, 315- 01A)
and 20 μg IFNγ (Preprotech, 315- 05) was diluted in Dulbecco’s phosphate- buffered
saline (PBS) and injected intraperitoneally. After cytokine injection, animals were under
permanent observation, and survival was assessed every 30 min.
Poly(I:C)- Induced Acute Lung Injury In Vivo Model. Six- to eight- week- old
male BALB/c (Bagg Albino/c) mice were assigned to treatment groups by ran-
domization in BioBook software to achieve similar group mean weight before
treatment; 10 mice were allocated into each group. Group 1 was normal- vehicle;
groups 2 to 5 were challenged with poly(I:C) with vehicle sodium carboxymethly
cellulose (CMC- Na), 10 mg/kg dexamethasone and 20 mg/kg roflumilast com-
bination, 20 mg/kg proxalutamide, or 40 mg/kg proxalutamide, respectively.
Dexamethasone was dissolved in 0.5% CMC- Na to make a suspension at a final
concentration of 1 mg/mL. Roflumilast was dissolved in 0.5% CMC- Na to make a
suspension at a final concentration of 2 mg/mL. Mice were treated with vehicle,
dexamethasone and roflumilast combination, or proxalutamide 16 h and 1 h prior
to poly(I:C) injection and 6 h after poly(I:C) injection. Additional proxalutamide
dose was given 18 h post poly(I:C) injection. Poly(I:C) solution was prepared
to a 0.06% solution in sterile PBS freshly prepared where 1.8 mg poly(I:C) was
dissolved in 3 mL PBS to make a suspension at a final concentration of 0.6 mg/
mL. Twenty- four hours post poly(I:C) injection, all mice were anesthetized with
Zoletil (i.p., 25 to 50 mg/kg, containing 1 mg/mL Xylazine). Lungs were gently
lavaged via the tracheal cannula with 0.5 mL PBS containing 1% fetal bovine
serum (FBS), and the BALF was collected. Then, the lungs were gently lavaged
with another 0.5 mL PBS containing 1% FBS. After lavage, the collected BALF was
stored on ice. The total cell number in BALF was counted using a hemocytometer.
After lavage by PBS, all mice were killed by exsanguination.
Liquid Mass Spectrometry Quantification after TFRE (Transcription
Factors Response Element) Enrichment. Mouse monocyte RAW264.7 cells
(0, 2 h, 4 h, and 8 h) and human monocyte THP- 1 (0, 0.5 h, 2 h, and 6 h) were
treated with 10 μM proxalutamide, respectively. Cells were collected and cocul-
tured with TFRE- binding beads, and the beads were rotated and combined for 1.5
h at 4°C. After the combined TFRE beads were washed 3 times with NETN and 2
times with mass spectrometry (to remove the scale removing agent; if there were
still bubbles, they were washed again with water). Then, 50 μL NH4HCO3 and 1.5
μg tyrosinase were added to the beads. The beads were hydrolyzed overnight,
and the tube wall was lightly spritzed 1 to 2 times in the middle. Two hundred
microliters of 50% acetonitrile + 0.1% formic acid was added to the suspension
for 3 to 5 min, and then, the supernatant was transferred on a magnetic rack to a
new Eppendorf tube; this was then repeated once. The supernatant was vacuum
dried into peptide powder and stored at low temperature. Protein sequences were
identified by liquid chromatography with tandem mass spectrometry.
Statistical Analysis. Statistical analyses were performed by the two- tailed,
unpaired t test, unless otherwise indicated in figure captions. Error bars indicate
mean ± SEM. GraphPad Prism software (version 9) was used for statistical calcu-
lations. No data were excluded from the analyses.
Data, Materials, and Software Availability. All study data are included in the
article and/or SI Appendix. Sequencing data are available through the National
Center for Biotechnology Information Gene Expression Omnibus, accession num-
ber GSE234805 (74).
ACKNOWLEDGMENTS. All strains of SARS- CoV- 2 virus were obtained through
the Biodefense and Emerging Infections Resources Repository of the National
Institute of Allergy and Infectious Diseases that were deposited by the Centers
for Disease Control and Prevention. J.Z.S. is supported by the National Institute
of Diabetes and Kidney Diseases (R01DK120623). J.W.W. is supported by an
American Foundation for Pharmaceutical Education regional award. A.M.C. is a
Howard Hughes Medical Institute Investigator, A. Alfred Taubman Scholar, and
American Cancer Society Professor.
Author affiliations: aMichigan Center for Translational Pathology, University of Michigan, Ann
Arbor, MI 48109; bDepartment of Pathology, University of Michigan, Ann Arbor, MI 48109;
cRogel Cancer Center, University of Michigan, Ann Arbor, MI 48109; dDepartment of Medicinal
Chemistry, College of Pharmacy, University of Michigan, Ann Arbor, MI 48109; eDepartment
of Internal Medicine, University of Michigan, Ann Arbor, MI 48109; fState Key Laboratory
of Proteomics, Beijing Proteome Research Center, National Center for Protein Sciences
(Beijing), Beijing Institute of Lifeomics, Beijing 102206, China; gKintor Pharmaceutical Limited,
Suzhou Industrial Park, Suzhuo 215123, China; hCenter for Drug Repurposing, University
of Michigan, Ann Arbor, MI 48109; iMichigan Institute for Clinical and Health Research,
University of Michigan, Ann Arbor, MI 48109; jDepartment of Pharmacology, University of
Michigan, Ann Arbor, MI 48109; kHHMI, University of Michigan, Ann Arbor, MI 48109; and
lDepartment of Urology, University of Michigan, Ann Arbor, MI 48109
PNAS 2023 Vol. 120 No. 30 e2221809120
https://doi.org/10.1073/pnas.2221809120 9 of 10
1.
2.
3.
4.
5.
6.
7.
K. Koelle, M. A. Martin, R. Antia, B. Lopman, N. E. Dean, The changing epidemiology of SARS- CoV- 2.
Science 375, 1116–1121 (2022).
F. P. Polack et al., Safety and efficacy of the BNT162b2 mRNA Covid- 19 vaccine. N. Engl. J. Med. 383,
2603–2615 (2020).
L. R. Baden et al., Efficacy and safety of the mRNA- 1273 SARS- CoV- 2 vaccine. N. Engl. J. Med. 384,
403–416 (2020).
J. Sadoff et al., Safety and efficacy of single- dose Ad26.COV2.S vaccine against Covid- 19. N. Engl. J.
Med. 384, 2187–2201 (2021).
D. H. Barouch, Covid- 19 vaccines—Immunity, variants, boosters. N. Engl. J. Med. 387, 1011–1020
(2022).
D. M. Altmann, R. J. Boyton, COVID- 19 vaccination: The road ahead. Science 375, 1127–1132
(2022).
E. G. Levin et al., Waning immune humoral response to BNT162b2 Covid- 19 vaccine over 6 months.
N. Engl. J. Med. 385, e84 (2021).
40. K. Welén et al., A phase 2 trial of the effect of antiandrogen therapy on COVID- 19 outcome: No
evidence of benefit, supported by epidemiology and in vitro data. Eur. Urol. 81, 285–293 (2022).
41. F. A. Cadegiani et al., Proxalutamide significantly accelerates viral clearance and reduces time to
clinical remission in patients with mild to moderate COVID- 19: Results from a randomized, double-
blinded, placebo- controlled trial. Cureus 13, e13492 (2021).
42. F. A. Cadegiani et al., Final results of a randomized, placebo- controlled, two- arm, parallel clinical trial
of proxalutamide for hospitalized COVID- 19 patients: A multiregional, joint analysis of the proxa-
rescue androCoV trial. Cureus 13, e20691 (2021).
43. J. McCoy et al., Proxalutamide reduces the rate of hospitalization for COVID- 19 male outpatients:
A randomized double- blinded placebo- controlled trial. Front. Med. (Lausanne) 8, 668698 (2021).
44. Frontiers Editorial Office, Retraction: Proxalutamide reduces the rate of hospitalization for COVID- 19
male outpatients: A randomized double- blinded placebo- controlled trial. Front. Med. (Lausanne) 9,
964099 (2022).
45. Y. Gu et al., Novel strategy of proxalutamide for the treatment of prostate cancer through
8. H. Chemaitelly et al., Waning of BNT162b2 vaccine protection against SARS- CoV- 2 infection in
coordinated blockade of lipogenesis and androgen receptor axis. Int. J. Mol. Sci. 22, 13222 (2021).
Qatar. N. Engl. J. Med. 385, e83 (2021).
9. H. Chemaitelly et al., Immune imprinting and protection against repeat reinfection with SARS-
CoV- 2. N. Engl. J. Med. 387, 1716–1718 (2022), 10.1056/NEJMc2211055.
46. F. Qu et al., Metabolomic profiling to evaluate the efficacy of proxalutamide, a novel androgen
receptor antagonist, in prostate cancer cells. Invest. New Drugs. 38, 1292–1302 (2020).
47. T. Zhou et al., Preclinical profile and phase I clinical trial of a novel androgen receptor antagonist
10. S. Chalkias et al., A bivalent omicron- containing booster vaccine against Covid- 19. N. Engl. J. Med.
GT0918 in castration- resistant prostate cancer. Eur. J. Cancer 134, 29–40 (2020).
387, 1279–1291 (2022).
48. J. H. Beigel et al., Remdesivir for the treatment of Covid- 19—Final report. N. Engl. J. Med. 383,
11. J. M. Coulson, A. Adams, L. A. Gray, A. Evans, COVID- 19 "Rebound" associated with nirmatrelvir/
1813–1826 (2020).
ritonavir pre- hospital therapy. J. Infect. 85, 436–480 (2022).
49. R. L. Gottlieb et al., Early remdesivir to prevent progression to severe Covid- 19 in outpatients. N.
12. J. Hammond et al., Oral nirmatrelvir for high- risk, nonhospitalized adults with Covid- 19. N. Engl. J.
Engl. J. Med. 386, 305–315 (2021).
Med. 386, 1397–1408 (2022).
13. F. Kabinger et al., Mechanism of molnupiravir- induced SARS- CoV- 2 mutagenesis. Nat. Struct. Mol.
Biol. 28, 740–746 (2021).
50. R. Karki et al., Synergism of TNF- α and IFN- γ triggers inflammatory cell death, tissue damage, and
mortality in SARS- CoV- 2 infection and cytokine shock syndromes. Cell 184, 149–168.e17 (2021).
51. H. I. Scher et al., Increased survival with enzalutamide in prostate cancer after chemotherapy. N.
14. H. Ledford, Hundreds of COVID trials could provide a deluge of new drugs. Nature 603, 25–27
Engl. J. Med. 367, 1187–1197 (2012).
(2022).
52. C. Mirabelli et al., Morphological cell profiling of SARS- CoV- 2 infection identifies drug repurposing
15. C. B. Jackson, M. Farzan, B. Chen, H. Choe, Mechanisms of SARS- CoV- 2 entry into cells. Nat. Rev. Mol.
candidates for COVID- 19. Proc. Natl. Acad. Sci. U.S.A. 118, e2105815118 (2021).
Cell Biol. 23, 3–20 (2022).
53. L. Yang et al., The signal pathways and treatment of cytokine storm in COVID- 19. Signal Transduct.
16. J. Lan et al., Structure of the SARS- CoV- 2 spike receptor- binding domain bound to the ACE2 receptor.
Targeted Ther. 6, 255 (2021).
Nature 581, 215–220 (2020).
17. J. Shang et al., Structural basis of receptor recognition by SARS- CoV- 2. Nature 581, 221–224 (2020).
18. M. Hoffmann et al., SARS- CoV- 2 cell entry depends on ACE2 and TMPRSS2 and is blocked by a
54. R. K. S. Malireddi et al., Inflammatory cell death, PANoptosis, mediated by cytokines in diverse
cancer lineages inhibits tumor growth. Immunohorizons 5, 568–580 (2021).
55. Y. Hirotsu et al., Nrf2- MafG heterodimers contribute globally to antioxidant and metabolic networks.
clinically proven protease inhibitor. Cell 181, 271–280 (2020).
Nucleic Acids Res. 40, 10228–10239 (2012).
19. B. Meng et al., Altered TMPRSS2 usage by SARS- CoV- 2 Omicron impacts infectivity and fusogenicity.
Nature 603, 706–714 (2022).
20. J. D. Gunst et al., Efficacy of the TMPRSS2 inhibitor camostat mesilate in patients hospitalized with
Covid- 19- a double- blind randomized controlled trial. EClinicalMedicine 35, 100849 (2021).
21. R. H. Shoemaker et al., Development of an aerosol intervention for COVID- 19 disease: Tolerability of
soluble ACE2 (APN01) administered via nebulizer. PLoS One 17, e0271066 (2022).
22. E. Tobback et al., Efficacy and safety of camostat mesylate in early COVID- 19 disease in an
ambulatory setting: A randomized placebo- controlled phase II trial. Int. J. Infect. Dis. 122, 628–635
(2022).
56. X. L. Chen et al., Activation of Nrf2/ARE pathway protects endothelial cells from oxidant injury and
inhibits inflammatory gene expression. Am. J. Physiol. Heart Circ. Physiol. 290, H1862–H1870
(2006).
57. S. Ruiz, P. E. Pergola, R. A. Zager, N. D. Vaziri, Targeting the transcription factor Nrf2 to ameliorate
oxidative stress and inflammation in chronic kidney disease. Kidney Int. 83, 1029–1041 (2013).
58. E. Singh et al., Management of COVID- 19- induced cytokine storm by Keap1- Nrf2 system: A review.
Inflammopharmacology 29, 1347–1355 (2021).
59. T. O. Khor et al., Nrf2- deficient mice have an increased susceptibility to dextran sulfate sodium-
induced colitis. Cancer Res. 66, 11580–11584 (2006).
23. L. Wettstein, F. Kirchhoff, J. Münch, The transmembrane protease TMPRSS2 as a therapeutic target
60. E. H. Kobayashi et al., Nrf2 suppresses macrophage inflammatory response by blocking
for COVID- 19 treatment. Int. J. Mol. Sci. 23, 1351 (2022).
proinflammatory cytokine transcription. Nat. Commun. 7, 11624 (2016).
24. T. Shapira et al., A TMPRSS2 inhibitor acts as a pan- SARS- CoV- 2 prophylactic and therapeutic. Nature
61. N. C. Stowell et al., Long- term activation of TLR3 by Poly(I:C) induces inflammation and impairs lung
605, 340–348 (2022).
function in mice. Respir. Res. 10, 43 (2009).
25. M. Mahoney et al., A novel class of TMPRSS2 inhibitors potently block SARS- CoV- 2 and MERS- CoV
viral entry and protect human epithelial lung cells. Proc. Natl. Acad. Sci. U.S.A. 118, e2108728118
(2021).
62. C. Wu et al., Risk factors associated with acute respiratory distress syndrome and death in patients
with Coronavirus disease 2019 pneumonia in Wuhan, China. JAMA Intern. Med. 180, 934–943
(2020).
26. K. H. Stopsack, L. A. Mucci, E. S. Antonarakis, P. S. Nelson, P. W. Kantoff, TMPRSS2 and COVID- 19:
63. S. Montazersaheb et al., COVID- 19 infection: An overview on cytokine storm and related
Serendipity or opportunity for intervention? Cancer Discov. 10, 779–782 (2020).
interventions. Virol. J. 19, 92 (2022).
27. J. M. Jin et al., Gender differences in patients with COVID- 19: Focus on severity and mortality. Front.
Public Health 8, 152 (2020).
28. H. Peckham et al., Male sex identified by global COVID- 19 meta- analysis as a risk factor for death
and ITU admission. Nat. Commun. 11, 6317 (2020).
64. Y. Wang et al., A small- molecule inhibitor of Keap1- Nrf2 interaction attenuates sepsis by selectively
augmenting the antibacterial defence of macrophages at infection sites. EBioMedicine 90, 104480
(2023).
65. R. K. Thimmulappa et al., Nrf2 is a critical regulator of the innate immune response and survival
29. D. K. Twitchell et al., Examining male predominance of severe COVID- 19 outcomes: A systematic
during experimental sepsis. J. Clin. Invest. 116, 984–995 (2006).
review. Androg. Clin. Res. Ther. 3, 41–53 (2022).
30. M. Montopoli et al., Androgen- deprivation therapies for prostate cancer and risk of infection by
SARS- CoV- 2: A population- based study (N = 4532). Ann. Oncol. 31, 1040–1045 (2020).
31. A. Goren et al., Anti- androgens may protect against severe COVID- 19 outcomes: Results from a
prospective cohort study of 77 hospitalized men. J. Eur. Acad. Dermatol. Venereol. 35, e13–e15
(2021).
32. V. G. Patel et al., Does androgen deprivation therapy protect against severe complications from
COVID- 19?. Ann. Oncol. 31, 1419–1420 (2020).
33. E. A. Klein et al., Androgen deprivation therapy in men with prostate cancer does not affect risk of
infection with SARS- CoV- 2. J. Urol. 205, 441–443 (2021).
34. A. L. Schmidt et al., Association between androgen deprivation therapy and mortality among
patients with prostate cancer and COVID- 19. JAMA Netw. Open 4, e2134330 (2021).
66. S. Zhang, J. Wang, L. Wang, S. Aliyari, G. Cheng, SARS- CoV- 2 virus NSP14 Impairs NRF2/HMOX1
activation by targeting Sirtuin 1. Cell Mol. Immunol. 19, 872–882 (2022).
67. N. D. Shore, Experience with degarelix in the treatment of prostate cancer. Ther. Adv. Urol. 5, 11–24
(2013).
68. C. Tran et al., Development of a second- generation antiandrogen for treatment of advanced prostate
cancer. Science 324, 787–790 (2009).
69. Cision PR Newswire, Kintor Pharma’s Proxalutamide demonstrated reduction in hospitalization/
mortality for patients with mild to moderate COVID- 19 in phase III MRCT study. https://www.
prnewswire.com/news- releases/kintor- pharmas- proxalutamide- demonstrated- reduction-
in- hospitalizationmortality- for- patients- with- mild- to- moderate- covid- 19- in- phase- iii- mrct-
study- 301518525.html (2022).
70. Y. Qiao et al., Antisense oligonucleotides to therapeutically target SARS- CoV- 2 infection. PLoS One
35. N. J. Shah et al., The impact of androgen deprivation therapy on COVID- 19 illness in men with
18, e0281281 (2023).
prostate cancer. JNCI Cancer Spectr. 6, pkac035 (2022).
71. J. W. Wotring, R. Fursmidt, L. Ward, J. Z. Sexton, Evaluating the in vitro efficacy of bovine lactoferrin
36. D. A. Leach et al., The antiandrogen enzalutamide downregulates TMPRSS2 and reduces cellular
products against SARS- CoV- 2 variants of concern. J. Dairy Sci. 105, 2791–2802 (2022).
entry of SARS- CoV- 2 in human lung cells. Nat. Commun. 12, 4068 (2021).
72. J. Drost et al., Organoid culture systems for prostate epithelial and cancer tissue. Nat. Protoc. 11,
37. Y. Qiao et al., Targeting transcriptional regulation of SARS- CoV- 2 entry factors ACE2 and TMPRSS2.
347–358 (2016).
Proc. Natl. Acad. Sci. U.S.A. 118, e2021450118 (2020).
73. Y. Qiao et al., Autophagy inhibition by targeting PIKfyve potentiates response to immune checkpoint
38. Q. Deng, R. U. Rasool, R. M. Russell, R. Natesan, I. A. Asangani, Targeting androgen regulation of
TMPRSS2 and ACE2 as a therapeutic strategy to combat COVID- 19. iScience 24, 102254 (2021).
39. N. G. Nickols et al., Effect of androgen suppression on clinical outcomes in hospitalized men with
COVID- 19: The HITCH randomized clinical trial. JAMA Netw. Open 5, e227852 (2022).
blockade in prostate cancer. Nat. Cancer 2, 978–993 (2021).
74. Y. Qiao et al., Proxalutamide reduces SARS- CoV- 2 infection and associated inflammatory response.
National Center for Biotechnology Information Gene Expression Omnibus. https://www.ncbi.nlm.
nih.gov/geo/query/acc.cgi?acc=GSE234805. Deposited 13 June 2023.
10 of 10 https://doi.org/10.1073/pnas.2221809120
pnas.org
| null |
10.3390_bs13010066.pdf
|
Data Availability Statement: Data supporting the reported results is kept by the first author.
|
Data Availability Statement: Data supporting the reported results is kept by the first author. Acknowledgments: We would like to thank the visionary, courageous, resourceful research participants, all of whom are engaged in initiatives to transform the lives of individuals and communities. They are the inspiration behind this study, and we want to showcase the great work they are doing.
|
Article
The Experience of Self-Transcendence in Social Activists
Carol Barton 1
and Rona Hart 2,*
1
2
Previously School of Psychology, University of East London, Water Lane, London E15 4LZ, UK
School of Psychology, University of Sussex, Falmer, Brighton BN1 9RH, UK
* Correspondence: [email protected]
Abstract: Every day the wellbeing of disadvantaged individuals and communities is being trans-
formed through the activities of self-transcendent social activists. The positive contagion generated
by their actions is felt globally through influence, replication, leadership training and education.
These people are visionary, brave, and describe their lives as joyful, deeply fulfilled, and impactful.
Seeking no personal recognition or accolade, born from a deep feeling of connectedness and a vision
of how life could be better, participants describe the factors that influenced their decision to dedicate
their lives to serving the greater good. Using Constructivist Grounded Theory, in-depth semi struc-
tured interviews were carried out with eight participants who self-identified as self-transcendent
social activists, who have initiated non-mandated and not-for-profit community action. Data was
analyzed to explore each participant’s personal experiences of self-transcendence and how being self-
transcendent has manifested their life choices. The findings present a definition of ‘self-transcendent
social activism’ and a theoretical model that explains the development of participants’ activism:
trigger, activate, maintain and sustain, resulting in an impact experienced at three levels - individual,
community and global. Theoretical and practical implications are discussed.
Keywords: self-transcendence; social activism; prosocial behavior
1. Introduction
The course of history has been changed by many highly impactful self-transcendent
social activists who committed their lives to bring about social transformation in the
communities and countries in which they lived and served. Nobel Peace Prize winner
(1964), Luther-King Jr., will long be remembered for his non-violent campaign against
racism that resulted in his assassination and racial discrimination being declared illegal in
southern US states. Nobel Peace Prize winner (1984) and former chairperson of the Truth
and Reconciliation Commission, Tutu, was influential in his campaign against apartheid
and for the peace negotiations in South Africa. Whilst the legacy of Gandhi, five times peace
prize nominee, whose non-violent leadership led to his assassination and to independence
for India, is celebrated annually through the award of the international Gandhi Peace Prize.
A review of biographical literature reveals that these courageous, visionary, people of faith
prioritized freedom, equality, and the eradication of poverty above self-interest [1–3]. From
a position of feeling connected to others and a focus that extends beyond their own personal
wellbeing, self-transcendent social activists are people who act to address global problems
such as inequality, poverty, environmental issues and exploitation [4].
Social activism is defined as “instances in which individuals or groups of individuals
who lack full access to institutionalized channels of influence engage in collective action
to remedy a perceived social problem, or to promote or counter changes to the existing
social order” [5] (p. 4). Social activists are therefore individuals or groups who engage in
collective action to bring attention to and resolve social problems. They operate through
groups or social movement organizations that are characterized by varying degrees of
formal and informal structures [5].
Citation: Barton, C.; Hart, R. The
Experience of Self-Transcendence in
Social Activists. Behav. Sci. 2023, 13,
66. https://doi.org/10.3390/
bs13010066
Academic Editor: Andrew Soundy
Received: 5 December 2022
Revised: 9 January 2023
Accepted: 9 January 2023
Published: 11 January 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under
the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Behav. Sci. 2023, 13, 66. https://doi.org/10.3390/bs13010066
https://www.mdpi.com/journal/behavsci
behavioral sciences Behav. Sci. 2023, 13, 66
2 of 22
Self-transcendence is defined as an “increased awareness of dimensions greater
than the self and expansions of personal boundaries within intrapersonal, interpersonal,
transpersonal, and temporal domains” [6] (p. 179). It involves an endeavor to connect to a
larger context with a prosocial intent to serve the greater good. As such, self-transcendence
is a set of values and a state of mind that can prompt the motivation to engage with so-
cial activism. However, to our knowledge there is no early research that examines the
connection between the two concepts qualitatively.
This study endeavors to contribute to the extant literature on the key motivations that
drive social activism through the exploration of self-transcendence. Given the potential
impact that activists have through the work they do in generating positive transformations
in people, groups and entire societies, their goals, life choices and experiences of self-
transcendence within the context of social activism is a worthy scientific undertaking.
1.1. Social Activism
Social activism involves taking positive intentional action and mobilizing resources
to bring about change in society. Activism, both peaceful and aggressive, is expressed in
many forms from writing letters, lobbying, boycotts, protests, strikes, petitions, community
led initiatives, and social media campaigns. Examples of topics that social activists may
engage with include environmental issues, racial equality, gender equality, refugee and
immigration policies, human rights, LGBTQ+ rights, religious freedom, poverty, housing,
anti-war campaigns, welfare policies, and many other topics.
Within the current Western neo-liberal social norms that emphasize individualistic pro-
self goals, and independence rather interdependence, engaging in prosocial activism with
a purpose of benefitting the greater good, might seem exceptional, especially since social
activism is a costly endeavor, and that the chances of successful outcomes are uncertain [7].
The question of what motivates social activism, and whether it is motivated by pro-self or
prosocial intents is particularly intriguing given the contrasting social norm setting.
Given the numerous social causes that social activists are engaged with, motivations
will likely vary in accordance with the goal being pursued, the context, and the ideologies
that underlie the activity. Recent research on the motivations of social activists has therefore
aimed to elicit overarching themes to explore the underlying motives of social activists.
A repeated theme in the literature is that people might engage in activism because of
injustices or deprivation that they suffered or witnessed or because they identify with the
hardship of a particular group whose struggles coincide with their own experiences [8].
This suggests that critical life-events, needs, goals and interests may be key drivers of social
activism [9]. An additional point raised in the literature is that negative emotions (such
as pain, fear, anger or frustration) triggered by one’s sense of deprivation or injustice, or
from a distressing life event, predict the willingness to engage in collective action, as well
as the actual participation [10–12]. The perception that one’s group is negatively evaluated,
disrespected, marginalized or discriminated against, can also induce willingness to engage
in social action, both peaceful and violent [13–17].
Identity features strongly in the social activism literature as a motivating factor. Na-
tional, professional, ethnic, racial, class or sexual orientation identities were found to be
key drivers that motivate people to engage in social activism, and in turn, belonging to a
social movement both intensifies the primacy of these identities, as well as generates a new
identity that forms as a result of affiliating and identifying with the activist group [18–23].
Another motivation to engage in social action is because it renders activists personal,
social or psychological benefits [24]. Gains accrued from activism include a sense of mean-
ing and purpose, positive self-regard, belonging to a group or a community, and improved
wellbeing [24–32]. Social activism was also associated with a feeling of personal signifi-
cance [33], indicating that when people feel that they lack significance, they were more
willing to pursue a political cause, at times involving violent actions, and making personal
sacrifices [25,34]. Klar and Kasser [24] showed that activism was positively related to
self-determination and meeting three basic needs: autonomy, competence, and related-
Behav. Sci. 2023, 13, 66
3 of 22
ness. Another benefit from social activism comes in the form of positive emotions, such as
exhilaration and awe, empowerment, pride, joy and sense of solidarity [11,24,35–37].
Values and ideologies often translate into visions and are also important motivating
factors that can drive people to take social action often by eliciting a sense of social respon-
sibility and urgency [38]. Prosocial values also play a central role in motivating action on
behalf of a social cause [39–42]. Similarly, ideologies, moral convictions and religious beliefs
are positively associated with activism [43–46], and acting on what one sees as core values
engenders a sense of meaning in life, significance, and fulfillment which in turn elevate
self-esteem [47,48]. In the context of political activism, moral convictions were found to be
associated with pride [49]. Interestingly, people may adopt particular values and pursue
social action due to guilt about one’s own privileges or for causing harm, or for not doing
enough [50,51]. Another type of motivation that can drive social action is generativity: the
desire to care about the welfare of future generations [22,52–54]. These are linked with
other prosocial states such as empathy, perspective taking, compassion, accountability, and
sympathy which have been shown as motivational factors that can prompt people into
social action [55–59].
The brief review offered above of the factors that can motivate social activism suggests
that it can be motivated by pro-self or prosocial goals, intents and attitudes, and these
contrasting underlying mindsets, can impact both personal and community results. Pro-self
motives can manifest in the desire to construct and maintain positive self-images of oneself
as worthy and valuable, and might lead to displaying concern for others insomuch as this
serves the need of the ego [60]. This can lead to the ‘white savior’ stereotype: imposing
patronizing models or solutions on those in jeopardy, leading at times to the perpetuation
of their condition [61,62].
In contrast, people who are motivated by prosocial motivations report empathetic
identification with disadvantaged groups, experiencing acute awareness of issues that need
to be changed and a belief that they can make a difference [63]. While they may sacrifice
their personal time and resources, paradoxically their work may result in enhancing their
own personal wellbeing [24], in addition to benefitting the greater good by attracting
attention to social problems, creating solutions, and developing partnerships [64].
The link between motivation, intents and outcomes in social activism raises the ques-
tion of self-transcendence as a driver of social activism. Next, we unpack the concept and
briefly review the literature.
1.2. Self-Transcendence
Frankl [65] (p. 115) maintained that “being human always points to something or
someone greater than self . . . the more one forgets oneself—by giving himself to a cause to
serve . . . the more human he is . . . ” Accordingly, within transpersonal psychology, self-
transcendence involves serving a purpose greater than the self with a selfless intent [66,67].
Reed [68] (p. 397) defines self-transcendence as “the capacity to experience connectedness
and expand self-boundaries in four dimensions: intra-personally by gaining more self-
awareness, inter-personally by relating to others and nature, temporally by integrating
past and future in a meaningful present, and trans-personally by connecting with spiritual
dimensions of indiscernible world”. Other authors argued that the term signifies a devel-
opmental process whereby one’s consciousness expands beyond personal, bounded, and
self-directed ego, to include other people and concerns within that sense of expanded iden-
tity [69]. Maslow’s [70] hierarchy of needs suggests that one of people’s top growth needs is
the desire to reach self-actualization whereby one can realize his or her full potential. How-
ever, it has been argued [71,72] that later in his life Maslow discovered that self-actualizing
individuals were capable of even higher psychological development by transcending their
own self-centered goals, and pursuing higher causes that are other-orientated.
The most coherent and possibly the most cited description of self-transcendence
emerged from Schwartz’s [73,74] values theory. Schwartz [73] conceptualizes values as
beliefs about what is desirable, worthy and important. As such these beliefs shape one’s
Behav. Sci. 2023, 13, 66
4 of 22
perceptions of oneself, others, and situations, guide one’s life goals, priorities, and decision-
making, influencing related behaviors. Schwartz’s [74] values model offers a classification
of two broad bi-polar dimensions, each of which incorporates several values: One axis
ranges between ‘openness to change’ to ‘conservation’ values, and includes self-direction
and stimulation on one side of the axis, and security, conformity and tradition on the other
side. The second axis has ‘self-enhancement’ on one side and ‘self-transcendence’ on the
other. It includes power and achievement on one side, and universalism, benevolence on
the other side. Hedonism is placed across two dimensions: openness to change and self-
enhancement. Self-transcendence values - benevolence and universalism, are characterized
by a reduction in self-centeredness, and the capacity to transcend one’s own selfish needs,
to care for the interest and welfare of others. As such they are considered other-focused,
growth promoting values [73].
An alternative conceptualization of self-transcendence suggests that it is a personality
trait linked to spirituality [75] whereby a person “identifies the self as part of the entire
cosmos” [76] (p. 975), feeling a sense of connection to the universe, interdependence
and responsibility.
It is also seen as a core virtue within the VIA character strengths and virtues clas-
sification, encompassing of character strengths of appreciation of beauty and excellence,
gratitude, hope, spirituality and humor [77]. This trait has been associated with experi-
encing elevation emotions such as awe, ecstasy, amazement, worship, and flow as well as
meaning in life [77].
In terms of its development or emergence, self-transcendence seems to be expressed
more strongly in people who confronted difficult life experiences such as loss or illness,
hence seen as a sign of adversarial growth in terms of the capacity to transcend one’s own
needs and experiences, taxing as they may be, to express universalism and prosociality [78–80].
Further exploration of the concept suggests that it can be active or passive in terms
of its behavioral manifestation. Although self-transcendence is positively associated with
taking action [81], it is indeed possible for self-transcendence to remain passive.
Another finding is that self-transcendence values promote prosocial attitudes and
states (such as empathy, trust, love, affection and compassion) and motivate a variety
of prosocial behaviors (such as offering encouragement, care, or support) [82–85]. Some
gender differences were detected, as women were found to attribute more importance
to self-transcendence values while men attribute more importance to self-enhancement
values [86].
There are indeed some self-benefits that people can gain from holding self-transcendence
values. A positive association was found between self-transcendence and wellbeing,
positive emotions, happiness, quality of life (in severely ill patients), healthy behaviors,
meaning and purpose in life, self-esteem, hope, sense of coherence, mindfulness, flow,
adaptive coping strategies and resilience [87–97].
1.3. Self-Transcendent Social Activism
While theoretically self-transcendence can become a strong motivator for social ac-
tivism, there is little research that explores this point. In two cross-national studies [82], the
authors concluded that people who hold self-transcendence values are more likely to be
involved in political activism. Similarly, Gundelach and Toubøl [98] found that the values
of self-transcendence were associated with activism in the context of refugee solidarity.
Another study on environmental activism examined the relationship between activism and
moral identity and concluded that self-transcendence positively predicts environmental
activism, while self-interest values were associated with apathy leading to low environmen-
tal activism [99]. In another correlational study Hackett [100] found that the association
between self-transcendence values and activist behaviors was stronger when these values
emerged from personal concerns.
Behav. Sci. 2023, 13, 66
5 of 22
To our knowledge no further research has explored the association between self-
transcendence and social activism, and there is no qualitative research which explores how
self-transcendence is manifested in social activism.
1.4. The Current Study
The aim of this study was to qualitatively explore the experiences of a mixed age and
mixed faith group of activists, who self-identify as self-transcendent, in order to answer the
question: ‘In what way does the experience of self-transcendence manifest in the work and
lives of social activists?’
In exploring this topic qualitatively, the paper aims to address a gap in the literature
and contribute to our understanding of the drivers of social activism.
2. Materials and Methods
This study applied a Constructivist Grounded Theory (GT) approach to collect and
analyze qualitative interview data [101], as a means to explore self-transcendence within the
context of social activism. Grounded Theory is particularly useful for exploratory studies
and its key strength is in facilitating the development of theoretical models emerging from
the data. It has several distinctive features [102,103]:
•
Data collection and analysis cycles: Grounded Theory involves an iterative data
collection and analysis process whereby early data collection and initial analyses
inform subsequent decisions on the direction and focus of data to be collected and on
sampling, while the analysis remains open to new emergent topics.
Sampling aimed at theory generation: Sampling in Grounded Theory, is initially
purposive (identifying and selecting participants who are knowledgeable about or
experienced with the phenomenon of interest), and later it becomes theoretically
driven (known as theoretical sampling), since sampling decisions draw on early
analysis and reflect the ongoing theoretical development that occurs as a result of the
data collection and analysis cycles.
Developing a theory from data: Grounded theory is designed in a way that enables
researchers to develop a theory/model from data [102,103]. As such, it requires
the application of inductive reasoning (bottom-up) which enables researchers to
extrapolate a theory from a set of individual cases. This involves moving from the
particular case to the general, and from a detailed description to an abstract level [101].
Data analysis: Analyzing data in Grounded Theory involves applying several tech-
niques. Initial or open coding involves analyzing the text by coding word-by-word
and line-by-line and naming each segment of the data. This is often followed by
focused coding which is aimed at generating conceptual codes [102]. Focused coding
involves the use of some of the following techniques:
•
•
•
(cid:35)
(cid:35)
(cid:35)
(cid:35)
(cid:35)
Axial coding: Relating categories to subcategories and making explicit connec-
tions between them.
Comparative coding: Constant comparisons between data in order to find
similarities and differences and establish analytic distinctions.
In-vivo coding: Preserving participants’ meaning in the coding.
Selective coding: Distinguishing the core categories and connecting them to
other categories.
Core categories: Identifying the components of the model/theory (they are
the ones that most frequently emerge from the data, they have identifiable
properties and are linked to other categories).
Theoretical coding: Specifying the relationships between categories and inte-
grating categories to create a coherent depiction of a model or theory.
(cid:35)
Whilst the nature of the interviewer-imposed questions meant that it was impossible
to totally eliminate researcher bias, interaction between researcher and participants took
the form of clean language open questions and passive listening, enabling participants
to speak openly and spontaneously of their life experiences and for theories to emerge
Behav. Sci. 2023, 13, 66
6 of 22
from participants’ narratives [103]. The research outcomes, therefore, are a co-construction
of a theoretical model based on the data and the interpretation, observations of the first
author, who, for many years, has supported Africa-based social activists through coaching
and consultancy.
2.1. Participants
Criterion sampling (a sub-set of purposive sampling) was used in this study to define
and invite the target participants. It involved searching for participants who meet a certain
criteria. In this study, the key criteria was involvement in social-activism and experiencing
self-transcendence. For the purpose of recruitment and self-selection of participants, the
following definitions were used (see Table 1):
Table 1. Definitions used for purpose of recruitment.
Self-Transcendence
•
•
•
•
A shift in focus from self (ego) to others;
A shift in values and willingness to sacrifice self-interest to serve the greater good;
An increase in moral concern and courage to act and take risks, aligned to moral compass
Social Activism
Non-mandated and not-for-profit practical action carried out by individuals or groups, to
solve societal problems and bring about change for the good of others.
The participants self-identified with the above statements and satisfied the following
inclusion criteria:
•
•
Feeling connected with something greater than oneself
They had initiated a non-mandated not-for-profit community program to reduce
poverty, injustice, homelessness; the program had been operational for at least two years
and positive community impact can be evidenced.
Potential social activists were identified through the first author’s personal networks,
which included former colleagues, coaching and business clients. The researcher also
invited former colleagues to recommend suitable participants. Prospective participants
were initially contacted via an email that informed them that the researcher was seeking
social activists who have experienced self-transcendence; the study aimed to explore their
experience of self-transcendence and how this had manifested in their life choices.
Eight social activists, six females and two males, of mixed nationalities and religions,
aged between 35 and 60 committed to participate in the study. Table 2 details their back-
ground and domain of social activism (pseudonym are used to protect their identities). No
incentives were offered to encourage participation.
Table 2. Participant demographics.
Pseudonym
Gender
Nationality
Country of
Residence
Religion
Context/Projects
Fiona
Jemma
F
F
Swazi
Kenya
Christian
Kenyan
Kenya
Christian
Pastor/Spiritual healer, laying the
foundation for an international network of
home educators
Bringing hope to poor communities affected
by HIV/AIDs by providing education,
medical and social care
Behav. Sci. 2023, 13, 66
7 of 22
Table 2. Cont.
Pseudonym
Gender
Nationality
Country of
Residence
Religion
Context/Projects
Tina
Sam
Todd
Natalie
Tandy
Judith
F
M
M
F
F
F
American
Kenya
Christian
Kenyan
Kenya
Muslim
Filipino
Hawaii
Christian
British (Tobago
origin)
Chinese
American
UK
Hindu
USA/Kenya
Christian
American
Kenya
Christian
Providing education, medical and social
support services for children with disabilities
and employment training and opportunities
for their mothers.
Eisenhower Fellow, developing local leaders,
catalysing positive change, and alleviating
poverty in the largest Kenyan slum
Youth Pastor, building affordable housing
units to support the homeless in Hawaii,
Cambodia and Africa
Teaching Meditation, peace circles and
wellness programmes in US, India, UK and
Virgin Islands
Empowering teachers and transforming
schools in Kenya through leadership training,
instructional coaching and
infrastructure support.
Rescuing and equipping orphans and
destitute children in Kenya and Romania
2.2. Data Collection
Following receipt of ethical approval from the first author’s University, potential
participants were contacted via email. Prior to the interviews, participants were provided
with more detailed information about the purpose of the research including information
about confidentiality and their right to withdraw. Written consent was obtained. A draft
set of questions was provided prior to the interviews.
Semi structured Grounded Theory interviews that lasted between 45 and 80 min
were conducted online by the first author and recorded using Zoom. After reminding
participants about the purpose of the research, interviews commenced by asking “what
does self-transcendence mean to you?” The researcher used clean language [104], open
questions to develop a conversation about their personal experience of becoming self-
transcendent and the role that self-transcendence plays in decision making. Listening
attentively for themes and insights, the researcher asked more probing follow up questions
to stimulate deeper reflection about specific characteristics of self-transcendence and what
factors strengthen or weaken their experience of self-transcendence. Example questions
include: What does the term self-transcendence mean to you? Thinking about your own
experience of becoming self-transcendent, how would you describe that? How has being
self-transcendent influenced your life choices? How does being self-transcendent manifest
itself in your social activism? In other areas of life? What are the benefits and challenges of
being self-transcendent?
Whilst one participant described in some detail her personal experience of becoming
self-transcendent, other interviewees steered the interview in the direction of how being self-
transcendent has motivated and influenced their life choices, and how this manifests in their
pursuit of social activism. The resulting theory, therefore, represents a ‘self-transcendent’
infused model of social activism.
The study followed Grounded Theory guidelines by conducting cycles of data collec-
tion followed by initial analysis which entailed line by line coding [102]. This meant that
between interviews, data were coded, and key themes identified for deeper exploration
were introduced through focused questions in subsequent interviews. For example, in
Behav. Sci. 2023, 13, 66
8 of 22
early interviews ‘courage’ and ‘empathy’, emerged as important themes leading to more
exploratory questions in later interviews.
Whilst time constraints meant that the number of participants and interviews was
limited, the sample size was considered large enough for a robust theory to emerge [105],
and data saturation was achieved within this sample size and interview framework.
2.3. Data Analysis
Interviews were transcribed using a transcription service and manually checked to
ensure verbatim accuracy. This enabled the researcher to gain an in depth understanding
of the data. As noted, data collection and initial analysis (open coding) occurred simul-
taneously. Once open coding was complete for all transcripts, several types of focused
coding techniques were applied to create a more abstract analytical framework [102]. The
first stage included sorting the numerous themes that emerged from the initial coding, to
identify and focus on the most salient ones [102]. Then axial coding was applied as a means
of linking between categories and their subcategories, some of which readily emerged from
the text. Comparative coding followed and involved comparing categories across different
segments of the data in order to find similarities and differences and to establish clearer
distinctions between elements that initially seemed to be entangled together [103]. The
next stage involved selective coding. At this stage it became clear that the focus of the
model would be around the process of becoming self-transcendent social activists. This
stage held the key to reducing the number of categories and focusing the analysis on the
most significant ones which were eventually identified as the core categories [102,103]. The
last stage involved theoretical coding - refining the categories, specifying the relationships
between them, and integrating them into a coherent model [103].
In order to produce a visual representation of the emergent model, the data were
then imported to NVIVO for further analysis. Earlier work by Bazeley [106] and oth-
ers [107,108] demonstrated the usefulness of NVIVO in facilitating a grounded theory
analysis. Hutchison, Johnston and Breckon [107] argued that the benefit of NVIVO is in
providing a transparent account of the analysis process which enhances its rigor. Although
NVIVO can be used to conduct all stages of the Grounded Theory analysis, in this study it
was only used to help generate a clearer account of the model.
The conceptualization of a theoretical model of ‘self-transcendent infused social ac-
tivism’, enabled the researcher to refine, condense, and align the data to the final six themes
which are described below.
3. Results
What started off as an investigation into the experience of self-transcendence in the
lives of social activists became a broader discourse about what motivated participants to
commit their lives to activism, the impact this has had on their personal lives and the com-
munities they serve and more globally. The analysis of data resulted in the emergence of:
A definition of self-transcendence within this context
1.
2. A description of how self-transcendence activism has impacted the lives of partici-
pants and the people they serve
3. A model comprising four continuous stages of activism - trigger, activate, maintain
and sustain.
These are summarized in Table 3.
Behav. Sci. 2023, 13, 66
9 of 22
Table 3. Summary of Results.
Feeling connected to something greater than oneself
Self-awareness
Definition
Increased awareness of social justice issues
Impact
Triggers
Activation
Reduction in self-interest
Desire to be of service
Personal impact
Community impact
Global impact
Early role models and exposure to social injustice
Personal experience of tragedy
Feeling ‘called’ or compelled
Empathy, Compassion and Connection
Courage and faith
Having a vision
Maintain
Personal sacrifice and self-care
A community of like-minded individuals for support
Seeing possibilities and co-production
Having a global perspective
Sustain
Growing leaders
Teaching empathy, awareness and courage
3.1. Definition
The definition domain describes how the participants responded to the question ‘what
does self-transcendence mean to you?’.
3.1.1. Feeling Connected to Something Greater Than Oneself
Without exception, Christian, Hindu and Muslim participants expressed the importance
their faith, combined with a commitment to live a life aligned to their spiritual convictions:
‘It’s definitely, my Christian, commitment and wanting to walk and do things for others’
(Jemma).
‘I’m very strong in my faith, but . . . I don’t want to force that on other people. But I also
make sure that I live my life in the values of my faith and that helps me in terms of how I
walk and interact with the community’ (Sam).
‘When I walk in my calling, directed by God (Fiona)’.
Connection to something greater also included the concept of seeing oneself as part of
a bigger community, connected to all humanity:
‘A small cog in a large wheel’ (Sam),
‘As individuals we are not complete in our separateness’ (Natalie).
‘There’s another expression that says ‘you are because we are’ so you always understand
that your life is connected to others . . . .’. (Fiona)
3.1.2. Self-Awareness
Most respondents noted that self-awareness and self-care are precursors to self-
transcendence and the process of becoming self-transcendent involves self-reflection, self-
knowledge and healing. To help other people in a healthy, safe and benevolent way, first it
is necessary to become a ‘safe person’:
Behav. Sci. 2023, 13, 66
10 of 22
‘ . . . in my process of transcendence, part of my journey was understanding who I am. I
think you cannot transcend yourself if you haven’t taken care of yourself. So, there’s an
element of understanding yourself, growing and knowing who you are’ (Fiona),
‘And then there’s your own growth as a human and your own sort of evolving identity
that sort of interacts with that... it is a process because you have to continually answer
the question of what is actually happening around me, how do I interpret it? How do I
make meaning out of the things I’m seeing?’ (Tandy)
3.1.3. Increased Awareness of Social Justice Issues
The majority of the participants reported a heightened awareness of inequality, poverty
and other prejudices combined with a belief that the situation can be improved. Whereas
other people might not be aware of injustices, participants reported both noticing and
wanting to respond to inequitable access to resources and opportunities:
‘It makes you aware of other people’s lives, other people’s struggles. God put compassion
and empathy in you, and you can’t limit that compassion and empathy to just a small
group of people’ (Judith).
‘It’s how we view the world, how we value things. I cannot sit back and see somebody else
being in total despair’ (Fiona).
3.1.4. Reduction in Self-Interest
Whilst we all need validation, if affirmation, personal gain or enhanced self-esteem is
the motivation; that is not self-transcendence, and this was noted by several participants.
The participants also noted that in self-transcendence, the focus and concern are no longer
on self but on the people being served. When self-gratification desires reduce there is a
much greater sense of freedom:
‘You’re doing things not just for your ego, not to be noticed. You don’t need pub-
lic acclamation. and you’re not doing it for personal gain. Doing it out of love and
compasion—There is something deeper within you’ (Jemma).
‘So basically, it’s about putting others first rather than putting yourself first.’ (Tim)
3.1.5. Desire to Be of Service
The act of serving others was mentioned by several participants who considered it
much more satisfying and rewarding than doing things just for oneself. To serve others
brings great personal blessings, to see the smile on the face of someone you’ve helped, or just
to experience the privilege of serving others, counts for so much more than self-gratification:
‘There can be so much emptiness in just trying to self-gratify. There’s only so much you
can do to self-gratify, but so much joy when you serve others and you see others happy’.
(Jemma)
3.2. Impact
3.2.1. Personal Impact
The work of an activist can be demanding and grueling, but participants overwhelm-
ingly described their lives as joyful, fulfilled, aligned to calling, abundant and meaningful.
Giving joy to others is described as contagious, great fun, extremely rewarding and this
creates a desire to do more:
‘..it can just be exhausting. Honestly just to be so empty, you know . . . .as the social
activist, learning to give your life away, and when you really look at what it definitely
includes, bringing fulfilment, and when you are completely exhausted, exhausted for the
social good.... it gives me a lot of joy. It’s grueling but it gives me joy’ (Jemma)
‘Yes, it does require personal sacrifice. But for me, I don’t see it as personal sacrifice
because I enjoy doing what I do and I see it as an opportunity, I derive a lot of joy. So, for
Behav. Sci. 2023, 13, 66
11 of 22
me I count it as a privilege . . . it makes you want to do it more because you get joy in
other people’s joy. I think joy is contagious, and so, giving joy is just so much fun’ (Tina).
‘Fulfilment, deep fulfilment, challenging, rewarding.’ (Judith)
3.2.2. Community Impact
Eight community programs are represented across four continents. Participants re-
ported working with victims of HIV, disabled children and their families, the homeless
and people living in slums to provide education, medical services, social care, adult skills
and employment training, mediation, infrastructure support, leadership development,
affordable housing and other initiatives to alleviate poverty and empower communities:
‘It began growing organically because when you support a woman she comes with the
entire family. A woman comes with children, youth, adolescents, and she brings the
community. And as a result, she also came with sickness and this affected the education
of the children and became an issue. Socioeconomic empowerment is an issue we’ve been
tackling initially as well. We wanted to see how we can support her to earn. You’re
putting that wholesome completeness in that home. So, we began by offering economic
empowerment, then education for the children, then the technical certificate for their older
children. We were training women to do different skills and assessing their credit to start
little businesses. So that’s how the whole project started . . . . . . .’. (Jemma)
‘We work with children with disabilities and their moms. There is no help in this country
for families that are struggling with that. Every child that comes to our therapy center,
comes with a mama and we provide each mama with employment . . . . . . Our heart is
for people that are struggling with disabilities and their families. We work with a lot of
HIV positive families and they’re just dealing with a lot of problems besides the disability.
There’s so many other problems that come along when you live in poverty. But it’s always
a thrill to be able to help somebody’. (Tina)
3.2.3. Global Impact
Most participants talked about the ripple effect which has been created through
developing international leaders, training others within existing programs, permitting
replication (at no cost) of their community development model, extending their work
internationally. One participant spoke of being invited to speak to UN representatives
about his work to support the homeless:
“God has been good in my life, putting me into this position where I can be influential
to a lot of people as an affordable housing developer. I’m a newbie in this industry, but
I’ve been recognized as the best affordable housing developer in town. And even United
Nations got a hold of my story and my philosophy as a developer... So instead of just
working on developing buildings for people for the money, I follow the need of people. So
my focus is to work with people, find out the need. And that’s one of the reasons why I flew
to Nairobi and I saw even greater need compared to Hawaii, because they’re in need of a
half a million apartments for the 3 million people who live in slum . . . And besides being
a developer, I created a non-profit organization. And we’re reaching out to Cambodia, to
the Philippines. And now I’m thinking about reaching out to Tanzania’. (Todd)
3.3. Triggers
This category refers to the life experiences that set participants on a course of taking action:
3.3.1. Early Role Models and Exposure to Social Injustice
All participants described how the influence of early role models, and the environment
in which they were raised, shaped their outlook and made them more sensitive to injustices
and inequalities:
‘I grew up seeing my parents caring for other people, serving more than to be served and
that’s how I grew to know life’. (Jemma)
Behav. Sci. 2023, 13, 66
12 of 22
One respondent noted how her experience of a difficult childhood led to a sense of
separation, fear and isolation which prompted a spiritual search for reconnection with some-
thing greater, triggering a desire to help others (Natalie). Another recalled his experience
of being raised in an institution:
‘It was shaped with my upbringing growing up in a children’s home which had more than
110 children. It’s not easy growing up in institutions - life was not easy. So that shaped
my thinking about how I wanted to live my life’. (Sam).
Exposure to social justice issues, such as homelessness, apartheid, refugees, triggered
an early response and determination to take action:
‘We had refugees in our home, and you are meant to take care of them. I saw my dad
bring one - he was an Ethiopian refugee when there was war in . . . and then when I was
in the university myself, I brought in a refugee, I’ve always had that desire to reach out to
people who are either homeless or suffering and to serve them’. (Jemma)
3.3.2. Personal Experience of Tragedy
Experiencing personal tragedy, or seeing tragedy close up often triggered negative
emotional and behavioral responses; however, for our participants experiencing trauma it
triggered a motivation to help others:
‘Our son was born at 22 weeks. He survived many heart attacks and we saw him come
back to life many times after having no heartbeat. And, he was a true miracle. and that
was my baby . . . . . . and then God asked me to do a special needs ministry’. (Tina).
‘It was the first time I saw a mother and a child laying on the side of the street and I was
in complete shock. Like I couldn’t believe that traffic wasn’t stopping, and people weren’t
helping her. Like it was such a foreign concept to me. and that definitely was a trigger
too.’. (Judith)
3.3.3. Feeling ‘Called’ or Compelled
Six participants reported a sense of calling, feeling compelled, or hearing from God,
to which, in spite of the personal sacrifices demanded and not knowing where resources
might come from, triggered a conviction to respond. One participant reported seeking
God’s will through prayer and reading the Bible:
‘God has called me to serve the very disadvantage, very poor, in the slum . . . . So out of
obedience to God he called me to go into that community and walk alongside women like
that’. (Jemma)
‘It’s a calling from God truly that he’s asked us to do this. I know that sounds, for some
people kind of weird, but it is definitely what we feel called to do. Now, did I hear a voice
when I say the word calling? No, but I spend a lot of time, praying and reading the Bible
and asking God to keep directing us.’. (Tina).
‘There is the compelling and choosing not to ignore that compelling. God spoke to me and
I know for certain that I heard the call and we responded’. (Judith)
3.4. Activation
These themes moved participants from ‘making a decision’ to take action by the
operationalization of that decision.
3.4.1. Empathy, Compassion and Connection
Common themes running through all interviews were how compassion and empathy
led to taking action. Empathy enables one to identify and connect with the community, as
opposed to sympathy which can be seen as adopting a superior position and imposing
solutions. Feeling compassionate often draws one into becoming deeply empathetic.
‘When you talk about transcendence, transcendence is not about sympathy. It must have
empathy. If empathy is not in you then you’re totally missing the point. So, empathy
Behav. Sci. 2023, 13, 66
13 of 22
enables you to identify and connect with the community. Whereas sympathy puts you on
a higher position and you’ve got power. Sympathy is all about listening with your head.
But empathy is about listening with your heart’. (Sam)
‘So, my job I believe is to inspire all these people that there is a choice that we can make
to have compassion and empathy for other people who are less fortunate than them/me’.
(Todd)
3.4.2. Courage and Faith
Without courage, self-transcendence can remain passive. All participants spoke of the
need to exercise courage, an internal quality that you carry on the inside–courage to admit
one does not know all the answers, to be unpopular, to travel across the world and live in
dangerous places, and courage to take risks. The notion of faith includes believing that
resources will be provided, and things will work out whilst the path remains unclear:
‘For sure you can’t do what we do without courage. You need both self-courage and you
just need overall courage. . . . .I want to learn the courage to say I’m not here to help. I’m
here to walk with you and everything. and even the courage to have a brave face to go
into hard places.’ (Sam).
So you have to sacrifice something in order to be courageous and to step up and do,
especially when you’re trying to help other people. You gotta have courage’. (Todd)
3.4.3. Having a Vision
Participants reported observing patterns and seeing life through a lens of possibilities.
Rather than looking at problems and what does not work, starting from the position of
appreciating what works, seeing potential in others—what they are capable of becoming
and having a visualization of what might be:
‘And I always say, because it is God’s work, he provides the resource, it’s his vision’.
(Jemma).
‘I was primed to see things in a way that would make me want to do something about it.
. . . . . . .. for several months prior to the vision trip that I took’ (Tandy).
‘So you have a vision. It’s challenging, but it’s also extremely rewarding, because I’ve
been doing it for some time, like for instance our rescue centre in Romania, those kids are
now grown’. (Judith)
3.5. Maintain
The life of an activist can be demanding and exhausting. The resolve to remain
committed is strengthened through several factors:
3.5.1. A Community of Like-Minded Individuals for Support
Surrounding oneself with a supportive circle of encouraging, like-minded people who
act as co-mentors increases motivation and provides opportunities to work collectively:
‘If you have healthy intimate relationships and strong connections with other people,
there’s an exchange - you’re learning with other people, you’re serving with other people.
I think that increases self-transcendence because you get the opportunity to watch other
people being courageous’ (Fiona).
For family members, the support of a partner and family to cheer you on is vital:
‘I don’t think that God’s going to call me one way and my husband another way because
we are in this together as a married couple. and so, we make decisions together’. (Tina)
3.5.2. Personal Sacrifice and Self-Care
The importance of exercising self-care, taking breaks and time out, spending time with
family, spiritual connection and devotion were reported as being important to maintain
good emotional, spiritual and physical health:
Behav. Sci. 2023, 13, 66
14 of 22
‘I made sacrifices thinking that I could withstand it, thinking that my marriage could
withstand it. I’ve made a lot of sacrifices. I think first of all, money, it took me five years,
before I launched xxx . . . ..And so that’s a very concrete data point around the financial
cost’ (Tandy)
‘I want to have more time with my daughter. I kinda need to start being selfish myself.
That’s called self-care and boundaries.’ (Sam).
‘ self-care is obviously very important. Having healthy boundaries is really important
. . . . So, I have to go to the source, which is God, he has an abundance. So, if I’m not
going to the source, it’s like not plugging my computer battery in. It’s not going to last
very long’. (Judith)
‘And of course, in this kind of work, you really have to know how to take care of yourself.
I’m here trying to recover. Cause the last week I was working so much, but I am happy.’
(Jemma)
3.5.3. Seeing Possibilities and Co-Production
Co-production is when a community comes together to influence and design policies
and services that benefit all, rather than becoming consumers of solutions supplied by non-
community members. This approach creates a sense of interdependence and connectedness
whereby people develop confidence to care for each other and co-create solutions. Co-
production is seen as an essential factor in maintaining programs and accomplishing
community empowerment. For many participants, this has involved exchanging western
comforts to live in a Nairobi slum, to truly understand what this feels like on a day-to-
day basis:
‘Once you start putting that community in a box and you’re not within that box you’re
outside, then you’re not in the community, then that’s a problem. You’re not actually
working with the community - you are working against the community. Or, you’re
actually looking in terms of “how do I bring a fix” with me?’ (Sam)
‘I think that just being at the same level with everybody here is an important piece. Living
with them, working side by side, shoulder to shoulder, trying to understand what they’re
going through, even though ultimately I can never fully understand’ (Tina)
3.6. Sustain
Participants expressed a desire to see the life of a self-transcendent activist become
more commonplace, describing the possibility in terms of ‘heaven on earth’ or utopia, a
world filled with more justice, equitable opportunities and resources, joy, compassion,
gratitude and kindness. Poverty, oppression and greed would be reduced. Important
factors that lead to sustaining impact and growth are identified below:
3.6.1. Having a Global Perspective
Technology and media support a sense of connection with people all over the globe.
Problems experienced by individuals, communities and countries are no longer viewed in
isolation and participants reported how recognizing the interconnection of all things leads
to the development of global solutions and co-operation that grows organically, often from
something small to something that has global impact:
‘We seem to have embodied this ethos on a global scale because we have kids from all over
the world’ (Fiona)
‘What I do - I offer up and create and hold space for entrepreneurs to also discover their
own purpose and their own capacities and their own power’ (Sam)
‘Because the more people that are connected and understand this and are able to move
outside of themselves, the better society is because then everybody, everybody becomes a
resource but in a positive way, not in an exploitative way, but in a synergistic way, like
in a way that that brings beauty to society’ (Fiona)
Behav. Sci. 2023, 13, 66
15 of 22
3.6.2. Growing Leaders
Leaving a legacy means training the next generation of leaders. Where this is ne-
glected, the potential impact of initiatives is not sustainable. An example offered by one
participant was of a situation where an influential community leader who had initiated
many community programs, unexpectedly died before training successors. His death
resulted in a fight for leadership and political chaos:
‘He was able to develop so many other things, but he failed in one thing. He failed in
grooming leaders to take over from where he was. So indirectly you can say he was
self-centered in his leadership because if he had intentionally groomed other leaders, we
would not be having the chaos we are having with the political parties’ (Sam).
‘What I’ve done mostly I’ve chosen to mentor others then meet other people who are
committed and have the same heart and the same calling. Increasingly, I’m investing
my time doing that mentoring, coaching so that more people can develop that attitude.’.
(Jemma)
3.6.3. Teaching Empathy, Awareness and Courage
Without courage, self-transcendence can remain passive. According to Sam, ‘with-
out empathy, you’re missing the point’. Self-reflection, self-knowledge and healing are
necessary precursors to helping others. Awareness of social justice issues is a trigger for
many activists. Embedding the concepts of empathy, self-awareness, awareness of social
injustice and courage into the educational, mentoring and coaching methods deployed by
participants and their organizations was reported to be a high priority:
‘There are others that are coming behind me that I need to teach and I need to teach them
to be courageous’. (Fiona)
‘There needs to be a way in terms of how we start breaking those walls and start having
conversations in terms of me and you, this is where I come from and where you come
from. Not based on tribe ethnicity or your race or your religion - then we start developing
empathy in a different way. So my priority now is I’m doing more in terms of one-to-one
where people just want to have a conversation’. (Sam)
The resultant model brings these themes together in a continuous process of self-
transcendent infused social activation which results in individual, community and (in the
case of participants) global impact.
4. Discussion
In the midst of alarming news about escalating and urgent global problems, where
every day millions live without adequate food, water and sanitation; children die from
malnutrition, HIV kills thousands of people, increased carbon dioxide and other human-
made emissions injure the planet and human activities create a wave of extinction of plants
and animals, the lives of individuals and the well-being of disadvantaged communities is
being transformed through the activities of impactful self-transcendent social activists. The
positive contagion of their actions is felt globally through influence, replication, leadership
training and education.
Experiencing notable levels of eudemonic wellbeing [109] participants describe their
lives as joyful, deeply fulfilled, privileged, spiritual and meaningful. Leading meaningful
lives sensed as a calling, seeking no personal recognition or accolades, born from a deep
feeling of connectedness and a vision of how life could be better, participants described
what motivated them to ‘focus on what really matters’ (Jemma) by committing their lives
to a self-transcendent purpose directed towards serving others [65].
What started off as an exploration into the experience of self-transcendence within the
context of social activism, led to the emergence of a ‘self-transcendence infused’ values
driven model (see Figure 1) of social activism that describes four key processes—trigger,
activate, maintain and sustain. The model presents a continuous process of activism that
Behav. Sci. 2023, 13, 66
16 of 22
generates personal joy fulfilment and meaning whilst creating a ripple effect of positive
contagion that can be leveraged to address community and global issues.
Figure 1. Self-transcendent social activism.
4.1. Self-Transcendent Social Activism
A combination of early role models, exposure to social injustice, personal experience
of tragedy and feeling ‘called’ triggered a resolve to help others; findings that are align to
research carried out by Dutt and Grabe [110]. Empathy, compassion, a sense of connection,
courage and faith moved participants from simply having a vison of how life could be
better, to take action. Maintaining social activism requires sacrifice and is challenging;
participants listed a number of factors that enhanced their commitment and motivation
including being surrounded by a community of like-minded individuals for support [111],
willingness to make personal sacrifices, self-care, seeing possibilities rather than problems
and adopting an empathetic approach that empowers communities. Sustaining momentum,
so that the ripple effect of their activism reaches new communities and future generations
and becomes more universally contagious, involves having a global perspective, growing
leaders, and embedding the concepts of empathy, awareness and courage into coaching,
mentoring and educational organisation systems.
4.2. Context
The study results have broader implications as shown in matrix below (Figure 2),
which depicts comparative levels of activism and self-transcendence.
Initiated in Hawaii, the approach taken to develop housing projects for the homeless
has been extended to Cambodia and Kenya and is recognized by the UN. The approach that
led to the creation and organic growth of a center of educational, medical and social support
facilities located in a Kenya slum emaciated by HIV, is being multiplied through a ‘franchise’
type methodology and mentoring like-minded activists. A program which started many
years ago in Romania, to rescue orphans, has led to a similar program being brought to
Kenya. These are examples of how the influence of participant’s activism extends well
beyond local communities. Participants self-identified as self-transcendent social activists
thereby occupying quadrant B on the matrix above. High self-transcendence combined
with high social activism has led to the development of sustainable co-produced community
enterprises [60,64]. Here, a number of factors have coalesced, resulting in significant com-
Behav. Sci. 2023, 13, 66 16 of 22 The resultant model brings these themes together in a continuous process of self-transcendent infused social activation which results in individual, community and (in the case of participants) global impact. 4. Discussion In the midst of alarming news about escalating and urgent global problems, where every day millions live without adequate food, water and sanitation; children die from malnutrition, HIV kills thousands of people, increased carbon dioxide and other human-made emissions injure the planet and human activities create a wave of extinction of plants and animals, the lives of individuals and the well-being of disadvantaged commu-nities is being transformed through the activities of impactful self-transcendent social ac-tivists. The positive contagion of their actions is felt globally through influence, replica-tion, leadership training and education. Experiencing notable levels of eudemonic wellbeing [109] participants describe their lives as joyful, deeply fulfilled, privileged, spiritual and meaningful. Leading meaningful lives sensed as a calling, seeking no personal recognition or accolades, born from a deep feeling of connectedness and a vision of how life could be better, participants described what motivated them to ‘focus on what really matters’ (Jemma) by committing their lives to a self-transcendent purpose directed towards serving others [65]. What started off as an exploration into the experience of self-transcendence within the context of social activism, led to the emergence of a ‘self-transcendence infused’ values driven model (see Figure 1) of social activism that describes four key processes—trigger, activate, maintain and sustain. The model presents a continuous process of activism that generates personal joy fulfilment and meaning whilst creating a ripple effect of positive contagion that can be leveraged to address community and global issues. Figure 1. Self-transcendent social activism. 4.1. Self-Transcendent Social Activism A combination of early role models, exposure to social injustice, personal experience of tragedy and feeling ‘called’ triggered a resolve to help others; findings that are align to research carried out by Dutt and Grabe [110]. Empathy, compassion, a sense of connec-tion, courage and faith moved participants from simply having a vison of how life could be better, to take action. Maintaining social activism requires sacrifice and is challenging; Behav. Sci. 2023, 13, 66
17 of 22
munity and global impact. By fully identifying with disadvantaged communities, working
alongside them, contributing much needed resources and skills, empowering, training and
co-producing sustainable initiatives, participants have delivered significant results.
Figure 2. Self-transcendence + social-activism = impact.
Quadrant A represents non-activated self-transcendence where the impact of a self-
transcendent lifestyle remains individualistic. Feelings of connection to something greater
than oneself and the motivation to do something meaningful are incubated before being
activated. Life for research participants commenced in this space as self-awareness, aware-
ness of injustice, and a desire to be of service increased. Feeling empathetic, compassionate
and connected, believing they had a role to play in helping to improve the lives of others,
exercising faith and bravery, overcoming challenges to pursue a goal or conviction [112];
participants moved from quadrant A to quadrant B by demonstrating commitment and a
willingness to step out of comfort zones and confront challenge [113].
Quadrant C represents a form of activism that is not infused with self-transcendence
values. Often more ego than eco driven, and sometimes driven by entrepreneurism and
a desire to generate profit, frequently less impactful ‘solutions’ are imposed rather than
co-created and are short-lived [60,114].
The research did not involve collecting Quadrant D data, which represents low self-
transcendence and low activism; however, from spiritual literature [115], we may speculate
that, for some, this is a lonely position, possibly with high levels of neuroticism and
alienation [116] representing potential ground for further social activism.
Self-transcendent social activism, which involves the integration of ego and eco goals
is highly impactful. This form of activism leads to the development of co-produced
sustainable initiatives and solutions that empower local communities and create positive
contagion. In comparison, non-self-transcendent activism, often motivated by personal
agendas, and the need for personal recognition leads to ‘outsider’ imposed, less sustainable,
models and often causes resentment. Self-transcendent activism operates from a position
of ‘empathy’. According to Sam, ‘empathy involves listening with the heart, whereas
sympathy involves listening to the head.’ ‘If empathy is not in you then you’re totally
missing the point’.
Passive self-transcendence may benefit an individual; however, increasing societal
impact involves transitioning from passive to active self-transcendence. Amongst other
Behav. Sci. 2023, 13, 66 17 of 22 participants listed a number of factors that enhanced their commitment and motivation including being surrounded by a community of like-minded individuals for support [111], willingness to make personal sacrifices, self-care, seeing possibilities rather than problems and adopting an empathetic approach that empowers communities. Sustaining momen-tum, so that the ripple effect of their activism reaches new communities and future gen-erations and becomes more universally contagious, involves having a global perspective, growing leaders, and embedding the concepts of empathy, awareness and courage into coaching, mentoring and educational organisation systems. 4.2. Context The study results have broader implications as shown in matrix below (Figure 2), which depicts comparative levels of activism and self-transcendence. Figure 2. Self-transcendence + social-activism = impact. Initiated in Hawaii, the approach taken to develop housing projects for the homeless has been extended to Cambodia and Kenya and is recognized by the UN. The approach that led to the creation and organic growth of a center of educational, medical and social support facilities located in a Kenya slum emaciated by HIV, is being multiplied through a ‘franchise’ type methodology and mentoring like-minded activists. A program which started many years ago in Romania, to rescue orphans, has led to a similar program being brought to Kenya. These are examples of how the influence of participant’s activism ex-tends well beyond local communities. Participants self-identified as self-transcendent so-cial activists thereby occupying quadrant B on the matrix above. High self-transcendence combined with high social activism has led to the development of sustainable co-pro-duced community enterprises [60,64]. Here, a number of factors have coalesced, resulting in significant community and global impact. By fully identifying with disadvantaged communities, working alongside them, contributing much needed resources and skills, empowering, training and co-producing sustainable initiatives, participants have deliv-ered significant results. Quadrant A represents non-activated self-transcendence where the impact of a self-transcendent lifestyle remains individualistic. Feelings of connection to something greater than oneself and the motivation to do something meaningful are incubated before being activated. Life for research participants commenced in this space as self-awareness, Behav. Sci. 2023, 13, 66
18 of 22
things, moving from passive to active requires developing a vision of how life can be
better [63], believing one can make a difference, and having courage. Courage can be
taught [112]. The implications and application of this study are far reaching. The study
suggests that teaching and modelling empathy, compassion and courage and embedding
each stage of the ‘self-transcendent social activism model’, into coaching, mentoring and
educational interventions will result in increased positive individual and community
impact, generating a ripple effect of positive contagion which can be leveraged to address
global challenges.
4.3. Limitations and Future Research
A number of limitations of the current study should be considered when examining
the results and conclusions. Findings were based on eight interviews with participants
who self-identified as self-transcendent social activists. A limitation of the study was the
predominance of female (6), and Christian (6), participants.
Within the scope of the interviews, arguably, data saturation was achieved, and no
new information emerged from latter interviews. However, given more time, it would be
possible to increase the sample size and to extend the range of interview questions. The
researcher has attempted to eliminate personal bias; however, a number of participants
were known to her. Future research, deploying a quantitative methodology to evidence
impact and the use of scales to measure the relationship between transcendence, activism
and wellbeing would strengthen findings.
Researching activism within the context of quadrant C—to include volunteerism,
entrepreneurialism, and career activism would prove insightful.
Furthermore, testing the model in terms of training, taking before and after mea-
surements to evidence the effectiveness of interventions designed to develop empathy,
compassion and courage, is suggested by the researcher.
5. Conclusions
This study contributes to the extant of the literature by expanding our understanding
of self-transcendence as a driver of social activism. It has resulted in the development of
a new model of ‘self-transcendent social activism’ containing four key processes: trigger,
activate, maintain and sustain engagement with social activism.
Author Contributions: Conceptualization: C.B. and R.H.; methodology: C.B. and R.H.; software:
Not relevant; validation: C.B.; formal analysis: C.B.; investigation: C.B.; resources: C.B. and R.H.;
data curation: C.B.; writing—original draft preparation: C.B.; writing—review and editing: R.H.
and C.B.; visualization: C.B. and R.H.; supervision: R.H.; project administration: C.B. and R.H.;
funding acquisition: not relevant. All authors have read and agreed to the published version of
the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: The study was conducted in accordance with the Decla-
ration of Helsinki, and approved by the University of East London Ethics Committee for studies
involving humans.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: Data supporting the reported results is kept by the first author.
Acknowledgments: We would like to thank the visionary, courageous, resourceful research partici-
pants, all of whom are engaged in initiatives to transform the lives of individuals and communities.
They are the inspiration behind this study, and we want to showcase the great work they are doing.
Conflicts of Interest: The authors declare no conflict of interest. However, we do note that the first
author is a coach and consultant who has had some professional involvement with the organizations
represented by several of the research participants.
Behav. Sci. 2023, 13, 66
References
19 of 22
Luther King, M. The Autobiography Of Martin Luther King, Jr.; Warner Books: New York, NY, USA, 1998.
1.
2. Mandella, N. Long Walk to Freedom; Abacus: London, UK, 1995.
3.
4.
5.
6. McCarthy, V.L.; Ling, R.N.J.; Carini, R.M. The Role of Self-Transcendence: A Missing Variable in the Pursuit of Successful Aging?
Tutu, D. No Future Without Forgiveness; Doubleday: New York, NY, USA, 2000.
Barker, C.; Martin, B.; Zournazi, M. Emotional self-management for activists. Reflective Pract. 2008, 9, 423–435. [CrossRef]
Briscoe, F.; Gupta, A. Social activism in and around organizations. Acad. Manag. Ann. 2016, 10, 671–727. [CrossRef]
7.
8.
9.
Res. Gerontol. Nurs. 2013, 6, 178–186. [CrossRef] [PubMed]
Jasko, K.; Szastok, M.; Grzymala-Moszczynska, J.; Maj, M.; Kruglanski, A.W. Rebel with a cause: Personal significance from
political activism predicts willingness to self-sacrifice. J. Soc. Issues 2019, 75, 314–349. [CrossRef]
Smith, H.J.; Pettigrew, T.F.; Pippin, G.M.; Bialosiewicz, S. Relative deprivation: A theoretical and meta-analytic review. Personal.
Soc. Psychol. Rev. 2012, 16, 203–232. [CrossRef]
Fielding, K.S.; McDonald, R.; Louis, W.R. Theory of planned behaviour, identity and intentions to engage in environmental
activism. J. Environ. Psychol. 2008, 28, 318–326. [CrossRef]
10. Miller, D.A.; Cronin, T.; Garcia, A.L.; Branscombe, N.R. The relative impact of anger and efficacy on collective action is affected by
feelings of fear. Group Process. Intergroup Relat. 2009, 12, 445–462. [CrossRef]
11. Tausch, N.; Becker, J.C. Emotional reactions to success and failure of collective action as predictors of future action intentions:
A longitudinal investigation in the context of student protests in Germany. Br. J. Soc. Psychol. 2013, 52, 525–542. [CrossRef]
[PubMed]
12. Van Zomeren, M. Four core social-psychological motivations to undertake collective action. Soc. Personal. Psychol. Compass 2013,
7, 378–388. [CrossRef]
13. Kteily, N.; Bruneau, E. Backlash: The politics and real-world consequences of minority group dehumanization. Personal. Soc.
Psychol. Bull. 2017, 43, 87–104. [CrossRef]
14. De Zavala, A.G.; Cichocka, A.; Eidelson, R.; Jayawickreme, N. Collective narcissism and its social consequences. J. Personal. Soc.
Psychol. 2009, 97, 1074–1081. [CrossRef] [PubMed]
15. Lyons-Padilla, S.; Gelfand, M.J.; Mirahmadi, H.; Farooq, M.; Van Egmond, M. Belonging nowhere: Marginalization & radicaliza-
tion risk among Muslim immigrants. Behav. Sci. Policy 2015, 1, 1–12.
16. Piazza, J.A. Rooted in poverty? Terrorism, poor economic development, and social cleavages. Terror. Political Violence 2006,
18, 159–177. [CrossRef]
17. Piazza, J.A. Poverty, minority economic discrimination, and domestic terrorism. J. Peace Res. 2011, 48, 339–353. [CrossRef]
18. McAdam, D.; Friedman, D. Collective identity and activism. In Frontiers in Social Movement Theory; Yale University Press: New
Haven, CT, USA, 1992; pp. 156–174.
19. Dono, J.; Webb, J.; Richardson, B. The relationship between environmental activism, pro-environmental behaviour and social
20.
21.
identity. J. Environ. Psychol. 2010, 30, 178–186. [CrossRef]
Schulte, M.; Bamberg, S.; Rees, J.; Rollin, P. Social identity as a key concept for connecting transformative societal change with
individual environmental activism. J. Environ. Psychol. 2020, 72, 101525. [CrossRef]
Foster, M.D.; Tassone, A.; Matheson, K. Tweeting about sexism motivates further activism: A social identity perspective. Br. J. Soc.
Psychol. 2021, 60, 741–764. [CrossRef] [PubMed]
22. Matsuba, M.K.; Pratt, M.W.; Norris, J.E.; Mohle, E.; Alisat, S.; McAdams, D.P. Environmentalism as a context for expressing
identity and generativity: Patterns among activists and uninvolved youth and midlife adults. J. Personal. 2012, 80, 1091–1115.
[CrossRef]
23. Valocchi, S. Activism as a Career, Calling, and Way of Life. J. Contemp. Ethnogr. 2013, 42, 169–200. [CrossRef]
24. Klar, M.; Kasser, T. Some benefits of being an activist: Measuring activism and its role in psychological well-being. Political Psychol.
2009, 30, 755–777. [CrossRef]
25. Dugas, M.; Bélanger, J.J.; Moyano, M.; Schumpe, B.M.; Kruglanski, A.W.; Gelfand, M.J.; Touchton-Leonard, K.; Nociti, N. The
quest for significance motivates self-sacrifice. Motiv. Sci. 2016, 2, 15–32. [CrossRef]
26. Venhoeven, L.A.; Bolderdijk, J.; Steg, L. Explaining the paradox: How proenvironmental behaviour can both thwart and foster
well-being. Sustainability 2013, 5, 1372–1386. [CrossRef]
27. Becker, J.C.; Tausch, N.; Wagner, U. Emotional consequences of collective action participation: Differentiating self-directed and
outgroup-directed emotions. Personal. Soc. Psychol. Bull. 2011, 37, 1587–1598. [CrossRef]
28. Tropp, L.R.; Brown, A.C. What benefits the group can also benefit the individual: Group-enhancing and individual-enhancing
motives for collective action. Group Process. Intergroup Relat. 2004, 7, 267–282. [CrossRef]
29. Back, E.A.; Back, H.; Knapton, H.M. Group belongingness and collective action: Effects of need to belong and rejection sensitivity
on willingness to participate in protest activities. Scand. J. Psychol. 2015, 56, 537–544. [CrossRef]
30. Wong, P.T.P. Self-Transcendence: A Paradoxical Way to Become Your Best. Int. J. Existent. Psychol. Psychother. 2016, 6, 9.
31.
Sohr, S. Eco-activism and well-being: Between flow and burnout. In Life-Goals and Well-Being: Towards a Positive Psychology of
Human Striving; Schmuck, P., Sheldon, K.M., Eds.; Hogrefe & Huber: Cambridge, WA, USA, 2001; pp. 202–215.
Behav. Sci. 2023, 13, 66
20 of 22
32. Eigner, S. The relationship between “protecting the environment” as a dominant life goal and subjective well-being. In Life-Goals
and Well-Being: Towards a Positive Psychology of Human Striving; Schmuck, P., Sheldon, K.M., Eds.; Hogrefe & Huber: Cambridge,
WA, USA, 2001; pp. 182–201.
33. Kruglanski, A.W.; Jasko, K.; Chernikova, M.; Dugas, M.; Webber, D. To the fringe and back: Violent extremism and the psychology
34.
of deviance. Am. Psychol. 2017, 72, 217–230. [CrossRef]
Jasko, K.; LaFree, G.; Kruglanski, A. Quest for significance and violent extremism: The case of domestic radicalization. Political
Psychol. 2017, 38, 815–831. [CrossRef]
35. Drury, J.; Reicher, S. Explaining enduring empowerment: A comparative study of collective action and psychological outcomes.
Eur. J. Soc. Psychol. 2005, 35, 35–58. [CrossRef]
36. Evripidou, A.; Drury, J. This is the time of tension: Collective action and subjective power in the Greek anti-austerity movement.
Contention: Multidiscip. J. Soc. Protest 2013, 1, 31–51. [CrossRef]
37. Apoifis, N. Anarchy in Athens: An Ethnography of Militancy, Emotions and Violence; Manchester University Press: Manchester, UK, 2016.
38. Lee, Y.; Kim, S.; Kim, M.; Choi, J. Antecedents and interrelationships of three types of pro-environmental behavior. J. Bus. Res.
2014, 67, 2097–2105. [CrossRef]
39. Liss, M.; Crawford, M.; Popp, D. Predictors and correlates of collective action. Sex Roles 2004, 50, 771–779. [CrossRef]
40. Miller, J.M.; Krosnick, J.A. Threat as a motivator of political activism: A field experiment. Political Psychol. 2004, 25, 507–523.
[CrossRef]
41. Unsworth, K.L.; McNeill, I.M. Increasing pro-environmental behaviors by increasing self-concordance: Testing an intervention.
J. Appl. Psychol. 2017, 102, 88–103. [CrossRef]
42. Van Zomeren, M.; Postmes, T.; Spears, R.; Bettache, K. Can moral convictions motivate the advantaged to challenge social
inequality? Extending the social identity model of collective action. Group Process. Intergroup Relat. 2011, 14, 735–753. [CrossRef]
43. Den Hond, F.; De Bakker, F.G. Ideologically motivated activism: How activist groups influence corporate social change activities.
44.
Acad. Manag. Rev. 2007, 32, 901–924. [CrossRef]
Skitka, L.J.; Bauman, C.W.; Sargis, E.G. Moral conviction: Another contributor to attitude strength or something more? J. Personal.
Soc. Psychol. 2005, 88, 895–917. [CrossRef]
Skitka, L.J.; Bauman, C.W. Moral conviction and political engagement. Political Psychol. 2008, 29, 29–54. [CrossRef]
45.
46. Atran, S.; Ginges, J. Religious and sacred imperatives in human conflict. Science 2012, 336, 855–857. [CrossRef]
47. Molinario, E.; Kruglanski, A.W.; Bonaiuto, F.; Bonnes, M.; Cicero, L.; Fornara, F.; Scopelliti, M.; Admiraal, J.; Beringer, A.;
Dedeurwaerdere, T.; et al. Motivations to Act for the Protection of Nature Biodiversity and the Environment: A Matter of
“Significance”. Environ. Behav. 2020, 52, 1133–1163. [CrossRef]
Sherman, D.K.; Cohen, G.L. The psychology of self-defense: Self-affirmation theory. Adv. Exp. Soc. Psychol. 2006, 38, 183–242.
Skitka, L.J.; Hanson, B.E.; Wisneski, D.C. Utopian hopes or dystopian fears? Exploring the motivational underpinnings of
moralized political engagement. Personal. Soc. Psychol. Bull. 2017, 43, 177–190. [CrossRef] [PubMed]
48.
49.
50. Kleres, J.; Wettergren, Å. Fear, hope, anger, and guilt in climate activism. Soc. Mov. Stud. 2017, 16, 507–519. [CrossRef]
51.
Jacobsson, K.; Lindblom, J. Emotion work in animal rights activism: A moral-sociological perspective. Acta Sociol. 2013, 56, 55–68.
[CrossRef]
52. Huo, H.; Qi, H.; Newton, N.J. Examining Identity and Generativity among Middle-aged Female Activists in Two Cultural
53.
Contexts. J. Adult Dev. 2022, 29, 318–327. [CrossRef]
Singer, J.A.; King, L.A.; Green, M.C.; Barr, S.C. Personal identity and civic responsibility: “Rising to the occasion” narratives and
generativity in community action student interns. J. Soc. Issues 2002, 58, 535–556. [CrossRef]
54. Alisat, S.; Norris, J.E.; Pratt, M.W.; Matsuba, M.K.; McAdams, D.P. Caring for the earth: Generativity as a mediator for the
prediction of environmenta narratives from identity among activists and nonactivists. Identity: Int. J. Theory Res. 2014, 14, 177–194.
[CrossRef]
Suk, J.; Abhishek, A.; Zhang, Y.; Ahn, S.Y.; Correa, T.; Garlough, C.; Shah, D.V. # MeToo, networked acknowledgment, and
connective action: How “empowerment through empathy” launched a social movement. Soc. Sci. Comput. Rev. 2021, 39, 276–294.
56. Todorova, M. Interpreting for refugees: Empathy and activism. In Intercultural Crisis Communication; Bloomsbury Publishing:
55.
London, UK, 2019; pp. 153–171.
57. Edrington, C.L. Social Movements and Identification: An Examination of How Black Lives Matter and March for Our Lives Use
Identification Strategies on Twitter to Build Relationships. J. Mass Commun. Q. 2022, 99, 643–659. [CrossRef]
58. Elisha, O. Moral ambitions of grace: The paradox of compassion and accountability in Evangelical faith-based activism. Cult. Anthropol.
2008, 23, 154–189. [CrossRef]
59. Yong, A. All Mixed up: Multi/Racial Liberation and Compassion-Based Activism. Religions 2020, 11, 402. [CrossRef]
60. Crocker, J. From Egosystem to Ecosystem: Implications for Relationships, Learning and Well-Being. In Transcending Self-Interest;
Wament, H., Bauer, J., Eds.; American Psychological Association: Washington, DC, USA, 2008; pp. 171–194.
61. McVeigh, K. Comic Relief to ditch White savior stereotype appeals. Guard 2018, 23, 29–37.
62. Kherbaoui, J.; Aronson, B. Bleeding through the band-aid: The white savior industrial complex. In Routledge Handbook of Critical
Studies in Whiteness; Hunter, S., Van der Westhuizen, C., Eds.; Routledge: Oxon, UK, 2021; pp. 269–279.
63. Korsgaard, M.; Meglino, B. Beyond the Individualistic Self—A framework for pro-social motives and judgements. In Transcending
Self-Interest; Wayment, H., Bauer, J., Eds.; APA: Washington, DC, USA, 2008; pp. 183–198.
Behav. Sci. 2023, 13, 66
21 of 22
64. Cahn, E. No More Throw-Away People; Essential: Washington, DC, USA, 2000.
65.
66. Aron, A.; Aron, E.A.; Tudor, M.; Nelson, G. Close Relationships as Including Other in the Self. J. Personal. Soc. Psychol. 1991,
Frankl, V.E. Self-Transcendence as a Human Phenomenon. J. Humanist. Psychol. 1966, 6, 97–106. [CrossRef]
60, 241–253. [CrossRef]
67. Hartman, D.; Zimberoff, D. Self-transcendence and Ego Surrender, a Quiet-enough Ego or an Ever-quieter enough Ego; Wellness Press:
Washington, DC, USA, 2005.
68. Reed, P.G. Demystifying Self-Transcendence for Mental Health Nursing Practice and Research. Arch. Psychiatr. Nurs. 2009,
23, 397–400. [CrossRef] [PubMed]
69. McCrathy, V.L.; Bockweg, A. The role of transcendence in a holistic view of successful aging: A concept analysis and model of
transcendence. J. Holist. Nurs. 2013, 31, 83–91. [CrossRef] [PubMed]
70. Maslow, A.H. Lessons from the peak experiences. J. Humanist. Psychol. 1962, 2, 9–18. [CrossRef]
71. Kaufman, S.B. Transcend: The New Science of Self-Actualization; Penguin: London, UK, 2021.
72. Koltko-Rivera, M.E. Rediscovering the later version of Maslow’s hierarchy of needs: Self-transcendence and opportunities for
73.
74.
theory, research, and unification. Rev. Gen. Psychol. 2006, 10, 302–317. [CrossRef]
Schwartz, S.H. An overview of the Schwartz theory of basic values. Online Read. Psychol. Cult. 2012, 2, 11–23. [CrossRef]
Schwartz, S.H. Universals in the content and structure of values: Theoretical advances and empirical tests in 20 countries.
Adv. Exp. Soc. Psychol. 1992, 25, 1–65.
75. Piedmont, R.L. Does spirituality represent the sixth factor of personality? Spiritual transcendence and the five-factor model.
J. Personal. 1999, 67, 985–1013. [CrossRef]
76. Cloninger, C.R.; Svrakic, D.M.; Przybeck, T.R. A psychobiological model of temperament and character. Arch. Gen. Psychiatry
1993, 50, 975–990. [CrossRef]
77. Peterson, C.; Seligman, M. Character Strengths and Virtues; Oxford University Press: New York, NY, USA, 2004.
78. Ellermann, C.R.; Reed, P.G. Self-transcendence and depression in middle-age adults. West. J. Nurs. Res. 2001, 23, 698–713.
[CrossRef]
79. Klaas, D. Testing two elements of spirituality in depressed and non-depressed elders. Int. J. Psychiatr. Nurs. Res. 1998, 4, 452–462.
Stinson, C.K.; Kirk, E. Structured reminiscence: An intervention to decrease depression and increase self-transcendence in older
80.
women. J. Clin. Nurs. 2006, 15, 208–218. [CrossRef]
81. Vecchione, M.; Schwartz, S.H.; Caprara, G.V.; Schoen, H.; Cieciuch, J.; Silvester, J.; Bain, P.; Bianchi, G.; Baslevent, C.;
Mamali, C.; et al. Personal values and political activism: A cross-national study. Br. J. Psychol. 2015, 106, 84–106. [Cross-
Ref]
82. Caprara, G.V.; Alessandri, G.; Eisenberg, N. Prosociality: The contribution of traits, values, and self-efficacy beliefs. J. Personal.
Soc. Psychol. 2012, 102, 1289–1303. [CrossRef]
83. Ardenghi, S.; Rampoldi, G.; Bani, M.; Strepparava, M.G. Personal values as early predictors of emotional and cognitive empathy
among medical students. Curr. Psychol. 2021, 1–9. [CrossRef]
84. Danioni, F.; Barni, D. The relations between adolescents’ personal values and prosocial and antisocial behaviours in team sports.
Int. J. Sport Exerc. Psychol. 2019, 17, 459–476. [CrossRef]
85. Tamir, M.; Schwartz, S.H.; Cieciuch, J.; Riediger, M.; Torres, C.; Scollon, C.; Vishkin, A. Desired emotions across cultures:
86.
87.
A value-based account. J. Personal. Soc. Psychol. 2016, 111, 67–82. [CrossRef]
Schwartz, S.H.; Rubel, T. Sex differences in value priorities: Cross-cultural and multi-method studies. J. Personal. Soc. Psychol.
2005, 89, 1010–1028. [CrossRef]
Sortheix, F.M.; Lönnqvist, J.E. Personal value priorities and life satisfaction in Europe: The moderating role of socioeconomic
development. J. Cross-Cult. Psychol. 2014, 45, 282–299. [CrossRef]
88. Russo, C.; Barni, D.; Zagrean, I.; Lulli, M.A.; Vecchi, G.; Danioni, F. The resilient recovery from substance addiction: The role of
89.
self-transcendence values and hope. Mediterr. J. Clin. Psychol. 2021, 9, 1–20.
Fegg, M.J.; Wasner, M.; Neudert, C.; Borasio, G.D. Personal values and individual quality of life in palliative care patients. J. Pain
Symptom Manag. 2005, 30, 154–159. [CrossRef]
90. Danioni, F.; Ranieri, S.; Villani, D. The role of personal values in gambling: A preliminary study with Italian adolescents. J. Genet.
Psychol. 2020, 181, 413–426. [CrossRef]
91. Neill, J. Transcendence and transformation in the life patterns of women living with rheumatoid arthritis. Adv. Nurs. Sci. 2002,
24, 27–47. [CrossRef]
92. Coward, D.D. Self-transcendence and correlates in a healthy population. Nurs. Res. 1996, 45, 116–121. [CrossRef]
93. Coward, D.D. Facilitation of self-transcendence in a breast cancer support group II. Oncol. Nurs. Forum 2003, 30, 291–300.
[CrossRef]
94. Nygren, B.; Aléx, L.; Jonsén, E.; Gustafson, Y.; Norberg, A.; Lundman, B. Resilience, sense of coherence, purpose in life and
self-transcendence in relation to perceived physical and mental health among the oldest old. Aging Ment. Health 2005, 9, 354–362.
[CrossRef]
95. Csikszentmihalyi, M. Flow: The Psychology of Optimal Experience; Harper & Row: New York, NY, USA, 1990.
96. Yaden, D.B.; Haidt, J.; Hood Jr, R.W.; Vago, D.R.; Newberg, A.B. The Varieties of Self-Transcendent Experience. Rev. Gen. Psychol.
2017, 17, 1089–2680. [CrossRef]
Behav. Sci. 2023, 13, 66
22 of 22
97. Batthyany, A.; Russo-Netzer, P. Psychologies of meaning.
In Meaning in Positive and Existential Psychology; Batthyany, A.,
Russo-Netzer, P., Eds.; Springer: Berlin/Heidelberg, Germany, 2014; pp. 3–22.
98. Gundelach, P.; Toubøl, J. High-and low-risk activism: Differential participation in a refugee solidarity movement. Mobilization Int.
99.
Q. 2019, 24, 199–220. [CrossRef]
Jia, F.; Soucie, K.; Alisat, S.; Curtin, D.; Pratt, M. Are environmental issues moral issues? Moral identity in relation to protecting
the natural world. J. Environ. Psychol. 2017, 52, 104–113. [CrossRef]
100. Hackett, J.D. Values anchoring: Strengthening the link between values and activist behaviors. Soc. Influ. 2014, 9, 99–115.
[CrossRef]
101. Charmaz, K. Grounded Theory Methods in Social Justice Research. In The Sage Handbook of Qualitative Research; Denzin, N.K.,
Lincoln, Y.S., Eds.; Sage: London, UK, 2011; pp. 359–380.
102. Charmaz, K. Constructing Grounded Theory; Sage: London, UK, 2006.
103. Glaser, B.G. Constructivist Grounded Theory? Glaser, Barney G. Constructivist Grounded Theory? [47 paragraphs]. Forum Qual.
Soz./Forum Qual. Soc. Res. 2002, 3, 12.
104. Sullivan, W.; Rees, J. Clean Language: Revealing Metaphors and Opening Minds; Crown House Publishing: Carmarthen, Wales, 2008.
105. Padgett, D. Does the Glove Really Fit? Qualitative Research and Clinical Social Work Practice. Soc. Work. 1998, 43, 373–381.
[CrossRef]
106. Bazeley, P. Qualitative Data Analysis with NVivo; Sage: London, UK, 2007.
107. Hutchison, A.J.; Johnston, L.H.; Breckon, J.D. Using QSR-NVivo to facilitate the development of a grounded theory project:
An account of a worked example. Int. J. Soc. Res. Methodol. 2010, 13, 283–302. [CrossRef]
108. Bringer, J.D.; Johnston, L.H.; Brackenridge, C.H. Using computer assisted qualitative data analysis software to develop a grounded
theory project. Field Methods 2006, 18, 245–266. [CrossRef]
109. Hart, R. Positive Psychology—The Basics; Routledge: London, UK, 2020.
110. Dutt, A.; Grabe, S. Lifetime activism, marginality, and psychology: Narratives of lifelong feminist activists committed to social
change. Qual. Psychol. 2014, 1, 107–122. [CrossRef]
111. Bettencourt, B.A.; Dillmann, G.; Wollman, N. The intragroup dynamics of maintaining a successful grassroots organization: A
case study. J. Soc. Issues 1996, 52, 169–186. [CrossRef]
112. Biswas-Diener, R. The Courage Quotient: How Science Can Make You Braver; Willey & Sons: Hoboken, NJ, USA, 2012.
113. Maddi, S.R. Hardiness: An Operationalization of Existential Courage. J. Humanist. Psychol. 2004, 44, 279–298. [CrossRef]
114. Bell, K.M. Raising Africa? Celebrity and the rhetoric of the white savior. Portal: J. Multidiscip. Int. Stud. 2013, 10, 1–24.
115. Rohr, R. The Divine Dance; Whitaker House: Pennsylvania, PA, USA, 2016.
116. Levenson, M.R.; Jennings, P.A.; Aldwin, C.M.; Shiraishi, R.W. Self-transcendence: Conceptualization and measurement. Int. J.
Aging Hum. Dev. 2005, 60, 127–143. [CrossRef] [PubMed]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.
| null |
10.1371_journal.pone.0265477.pdf
|
Data Availability Statement: The raw data was
collected from Taiwan Centers for Disease Control
(CDC) and it is available at www.cdc.gov.tw/En.
The tool and dataset are publicly available at github.
com/mahsaashouri/Taiwan-COVID-19-Interactive-
tool.
|
The raw data was collected from Taiwan Centers for Disease Control (CDC) and it is available at www.cdc.gov.tw/En . The tool and dataset are publicly available at github. com/mahsaashouri/Taiwan
|
RESEARCH ARTICLE
Interactive tool for clustering and forecasting
patterns of Taiwan COVID-19 spread
Mahsa Ashouri, Frederick Kin Hing PhoaID*
Institute of Statistical Science, Academia Sinica, Taipei, Taiwan
* [email protected]
Abstract
The COVID-19 data analysis is essential for policymakers to analyze the outbreak and man-
age the containment. Many approaches based on traditional time series clustering and fore-
casting methods, such as hierarchical clustering and exponential smoothing, have been
proposed to cluster and forecast the COVID-19 data. However, most of these methods do
not scale up with the high volume of cases. Moreover, the interactive nature of the applica-
tion demands further critically complex yet compelling clustering and forecasting tech-
niques. In this paper, we propose a web-based interactive tool to cluster and forecast the
available data of Taiwan COVID-19 confirmed infection cases. We apply the Model-based
(MOB) tree and domain-relevant attributes to cluster the dataset and display forecasting
results using the Ordinary Least Square (OLS) method. In this OLS model, we apply a
model produced by the MOB tree to forecast all series in each cluster. Our user-friendly
parametric forecasting method is computationally cheap. A web app based on R’s Shiny
App makes it easier for practitioners to find clustering and forecasting results while choosing
different parameters such as domain-relevant attributes. These results could help in deter-
mining the spread pattern and be utilized by medical researchers.
Introduction
The Coronavirus Disease 2019 (COVID-19) from Wuhan (Hubei, China), which started
spreading quickly in late December 2019, was announced as an outbreak by the public health
emergency of international in January 2020 and a pandemic by the World Health Organization
(WHO) on March 11, 2020. It transmits from person to person and causes symptoms like high
fever, cough, and shortness of breath after a 2-to-14-day infection period [1]. On December 15,
2020, more than 72.8 million people were confirmed by COVID-19, with 742 cases confirmed
in Taiwan. Confirmed cases grew exponentially across all continents [2]. The world has
changed dramatically ever since the first case broke out, and many countries have encountered
multiple crises, such as health crises, financial crises, and economic collapses [3]. At that time,
Taiwan had successfully curbed the spread for more than a year since the outbreak started.
Taiwan center of disease control reported the first confirmed infection case on January 21,
2020, a 50-year-old woman who was a teacher in Wuhan. Due to early responses and active
contact tracing policies, Taiwan managed to contain the spread successfully with a record of
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Ashouri M, Phoa FKH (2022) Interactive
tool for clustering and forecasting patterns of
Taiwan COVID-19 spread. PLoS ONE 17(6):
e0265477. https://doi.org/10.1371/journal.
pone.0265477
Editor: Chun-Hsi Huang, Southern Illinois
University, UNITED STATES
Received: August 23, 2021
Accepted: March 2, 2022
Published: June 30, 2022
Copyright: © 2022 Ashouri, Kin Hing Phoa. This is
an open access article distributed under the terms
of the Creative Commons Attribution License,
which permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The raw data was
collected from Taiwan Centers for Disease Control
(CDC) and it is available at www.cdc.gov.tw/En.
The tool and dataset are publicly available at github.
com/mahsaashouri/Taiwan-COVID-19-Interactive-
tool.
Funding: FKHP, AS-TP-109-M07, Academia Sinica,
https://www.sinica.edu.tw/ FKHP, 107-2118-M-
001-011-MY3 and 109-2321-B-001-013, Ministry
of Science and Technology (Taiwan), https://www.
most.gov.tw/.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
1 / 11
PLOS ONE Competing interests: The authors have declared
that no competing interests exist.
Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
250 consecutive days without any locally transmitted cases. However, Taiwan started to face a
sharp surge of confirmed cases in late April 2021 [4]. The policy-making and spread patterns
of the disease depend on many factors (such as environmental factors [5]), which may not fol-
low the previously available models. Therefore, creating a more efficient and accurate interac-
tive analytical tool is essential in identifying the spread pattern and providing helpful
information to enact effective policies.
Time series clustering is essential to determine similarities and/or differences in the behav-
ior of COVID-19 across cities, states, or countries, and it is advantageous in selecting forecast-
ing models. [6] measured the similarity of the COVID-19 time series between states using the
dynamic time warping distance (DTW) as the similarity matrix and applied a hierarchical clus-
tering approach to analyze the behavioral relationships in the United States (US) pandemic. As
a result, they found different pandemic behaviors in eastern and western zones. [7] suggested a
non-negative matrix factorization (NMF) followed by a k-means clustering procedure on the
coefficients of the NMF basis to cluster the US states into different communities. Their method
not only has the advantage of capturing patterns, but it has also reflected the spread and con-
trol of the pandemic by July 25, 2020.
[8] used an unsupervised machine learning technique to identify COVID-19 cases. They
applied a lung radiography dataset to the Robust Continuous Clustering algorithm (RCC) to
identify confirmed patients.
Forecasting the pattern of the COVID-19 pandemic is critical to health services, health
policymakers, healthcare providers, and epidemiologists. Various time series approaches
aim to forecast the COVID-19 pandemic using statistical modeling. For example, [9] pro-
posed a time series statistical approach to predict the short-term behavior of COVID-19.
They applied multiplicative trend to forecast the number of confirmed cases and deaths
globally and presented a 10-day-ahead competitive forecast over four months. [10] intro-
duced an objective approach to predict the continuation of COVID-19. They produced fore-
casts using models from the exponential smoothing family suitable for the short-term time
series. [2] presented a simple interactive non-linear method to forecast the number of con-
firmed cases. Their method took the expected recoveries and deaths into account to deter-
mine the maximum daily growth rate. Finally, [11] suggested a simplified and accurate
method using fast linear regressions with only a few parameters to forecast deaths, which
can consider the effect of many complexities of the epidemic process.
R’s Shiny app
R’s Shiny app [12] is a package from RStudio [13] developed for an easier and more efficient
result visualization. This web-based application allows users to change the model parameters
and interact with results. In the COVID-19 subject, many researchers published medical and
epidemiological research regarding interactive data analysis and visualization with the R Shiny
framework.
For instance, the COVID-19 tracker [14] in R’s Shiny package provides more context for
daily headlines and a fresh perspective of historical turning points.
[15] developed a COVID-19 worldwide web-based application using R’s Shiny package.
They design the tool for the country-specific analysis to visualize epidemiological pandemic
indicators. [16] suggested a COVID-19 watcher of the updated information for medical and
public use. Their tool aggregated the data from different resources and visualized them using
an online dashboard.
This research proposes an interactive web-based R’s Shiny app to cluster and forecast Tai-
wan COVID-19 time series while benefiting from domain-relevant attributes. Our tool helps
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
2 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
users choose from various parameters to interact with results. For example, users can identify
possible domain-relevant splitting variables of interest.
The dataset contains the number of confirmed cases in the cities, townships, and districts of
Taiwan. The data was collected from Taiwan Centers for Disease Control (CDC)and contains
183 daily series with a length of 155 from January 1 to June 4, 2021. We assume that this gov-
ernmental data is legitimate and trustworthy.
Our app can also be used to analyze COVID-19 time series data of any places in the rest of
the world, only if the dataset follows the same structure below. There should be eight columns
in the dataset corresponding to administrative types, city name, a YES/NO on whether the city
has an airport, a YES/NO on whether the case is imported or local otherwise, the number of
cases, the region category, the number of population, and the date. Among them, all except the
number of cases, the number of population, and the dates are categorical entries. In addition,
all rows are arranged in the ascending order of the dates for each city. Finally, the first row
should be the name of the titles of these eight columns.
Methodology
To cluster and forecast COVID-19 time series, we applied the method suggested by [17], and
we will briefly explain it in this section. This clustering approach applies domain-relevant attri-
butes and time series temporal patterns (trend, seasonality, and autocorrelation). Domain-rel-
evant attributes are cross-sectional attributes that link time series into sub-groups. For
example, the sales volume of items in a supermarket can be divided into different sub-groups.
Similarly, the COVID-19 cases can also be grouped based on geographical features.
This method based on the model-based partitioning tree (MOB) [18] is automated for clus-
tering large collections of time series. It consists of fitting local parametric models into differ-
ent subsets based on a recursive partitioning algorithm. The parameters and split points are
estimated using an objective function and a greedy forward search. To determine which vari-
able should be used for partitioning, we test each model score for parameter instability in each
node. Each node of the resulting tree is associated with a parametric statistical model. When
using the MOB algorithm, we need to specify the outcome, the predictors, the splitting vari-
ables, and the ‘fit’ function. The next part will discuss how the ordinary least squares (OLS)
model is used as the ‘fit’ function within the MOB framework.
To capture time series temporal patterns, [17] suggested an OLS model with predictors to
model their trend, seasonality, and autocorrelation. This model is parametric and flexible in
trend shapes (e.g., linear, quadratic) and seasonal patterns (e.g., seasonal dummies or a smooth
function for slowly changing seasonality). These predictors allow incorporating external attri-
butes valuable for clustering or forecasting time series. For instance, we can include the ‘Easter’
dummy variable indicating the timing of Easter.
Y ¼ Trend þ Season þ ARðpÞ þ External data þ error;
ð1Þ
Where AR(p) is a weighted average of lags in order p, and p can be equal to seasonality order or
specified based on the data type and domain knowledge. As an example, Eq 1 can be written as:
yt ¼ a0 þ a1f ðtÞ
þb1Season1t þ b2Season2t þ � � � þ bm(cid:0) 1Seasonðm(cid:0) 1Þt
þg1yt(cid:0) 1 þ g2yt(cid:0) 2 þ � � � þ gpyt(cid:0) p
þdzt þ �t;
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
ð2Þ
3 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
where yt (t = 1, 2, . . ., T) is the value of series at time t, f(t) is a function of the time index that
captures trend (e.g., linear, quadratic), Seasonjt is a dummy variable taking value 1 if time t is in
season j, m is the number of seasons (e.g., for a daily time series with day-of-week seasonality,
m = 7), and zt is the external data at time t. Furthermore, yt−j is the jth lagged value. One advan-
tage of OLS models is the interpretability of coefficients. The contribution of each feature to the
output will be equal to its coefficient. For example, if there is a linear trend, α1 measures the
changes in yt from one period to the next due to the passage of time while holding other vari-
ables in the model constant. As another example, with quadratic trend, α1 f(t) would be a0
1t þ
1t2 means when a0
a00
1 are positive, the trend is increasing while holding other variables in
the model constant.
1 and a00
Using the MOB partitioning tree and pseudo-R notations with partitioning variables
[Z1, . . ., Zq], Eq 2 can be written as:
yt ¼ a0 þ a1f ðtÞ
þb1Season1t þ b2Season2t þ � � � þ bm(cid:0) 1Seasonðm(cid:0) 1Þt
ð3Þ
þg1yt(cid:0) 1 þ g2yt(cid:0) 2 þ � � � þ gpyt(cid:0) p
þdztjZ1 þ � � � þ Zq:
This approach creates clusters with the same domain-relevant attribute profile and the simi-
lar trend, seasonality, and autocorrelation pattern. Based on this approach, we can cluster the
time series using Algorithm 1.
Algorithm 1: MOB time series clustering algorithm
• Zero time series: separate ‘all zero’ time series
• Normalize the series: subtract the mean and divide the standard
deviation
• MOB tree: run the tree on the series using Eq 3
• Prune the MOB tree: stop the tree when reaching the best improvement
on Mean Square Error (MSE), tree simplicity, and AIC [19] or BIC [20]
• Coefficient plot: compare OLS models in non-neighboring clusters and
check their differences/similarities
Finally, we computed forecasts by one linear model in each cluster produced by the MOB
partitioning tree. We apply the same linear model for series in the same cluster to produce
forecasts. We generate forecasts at fixed time t with h steps ahead (the lagged values of y are
replaced by their forecasted values if they occur in periods after the forecast origin). We also
compare our OLS forecast results with the Exponential Smoothing (ETS) approach. For run-
ning ETS, we applied functions ‘est’ forecast package [21] in R. We run this function indepen-
dently on each series. Then, we use the average of Root Mean Square Errors (RMSEs) and
Mean Absolute Error (MAE) across all series and display box and density plots for forecast
errors. We define the forecast error as the difference between the observed value and its fore-
cast. For better visualization, we do not plot the outliers.
Clustering and forecasting Taiwan COVID-19 confirmed cases
The collected dataset includes 183 daily series (cities, townships, and districts) with a length of
161 from January 1 to June 10, 2021, and the number excluding zero time series is one. Before
running the clustering method, we scaled the data by subtracting the mean and dividing the
standard deviation. Additionally, we partition the data into training and test sets, with the last
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
4 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
Table 1. Domain-relevant attribute categories used in Taiwan COVID-19 confirmed infection cases.
Domain-relevant attributes
Categories
Region
Administrative
Population
Imported
Airport
north, east, west, south, null (imported cases)
township/city, district, null (imported cases)
numeric—no categories
yes, no (local cases)
yes, no (the city has an international airport or not)
https://doi.org/10.1371/journal.pone.0265477.t001
7 days as our test set and the rest as the training set. Then we combine the training and test sets
and update the model and forecast one-week-ahead of the confirmed cases. Note that we
update the model in each cluster while keeping the clustering results unchanged.
For Taiwan COVID-19 daily dataset, we included the following predictors in the MOB-
based clustering and forecasting OLS model (‘fit’ function): a linear trend, six seasonal dum-
mies, and lags 1 to 7. Also domain-relevant splitting variables includes geographical division,
including ‘region’ (6 categories), ‘administrative’ (3 categories), ‘population’ (numeric),
‘imported’ (2 categories), and ‘airport’ (2 categories) (Table 1).
Interactive tool
Table 2 demonstrates the interactive panel inside our tool with three options to choose from,
the MOB depth (number of splits +1), prune option, and domain-relevant attributes (splitting
variables). Additionally, the ‘choose file to upload’ button lets users upload the desired dataset.
Our web-based interactive tool consists of eight parts (displays in Figs 1 to 6). For better
visualization, we divide the results into six figures.
The number on the top shows the MSE for all splits in the MOB partitioning tree. The first
MOB-heatmap includes two parts (Fig 1). The right part displays the MOB tree, which helps
users see domain-relevant attributes and split order accessioned with each cluster based on the
specified depth, prune options, and domain-relevant attributes. The left part is the time series
heatmap of all clusters, displaying time series patterns. Each row represents one series, and
darker color means higher values (color pallet: white, green, and red). Vertical stripes specify
similarities among the series in each cluster.
The second MOB-heatmap (Fig 2) is similar to the first plot, except it combines periods
into seasonal aggregations to highlight the seasonal effects. In both heatmaps, we order series
based on their values. For example, series with higher values in a similar period gather in the
same area. Also, based on the number of series in each cluster, the size of the cluster box would
be different.
Table 2. Taiwan COVID-19 interactive tool panel.
Categories
Application
Choose file to upload
let users upload the Taiwan COVID-19 dataset
MOB depth (number of
splits + 1)
Prune option
Splitting variables
changes from ‘no split’ to ‘full tree’, which controls the tree simplicity
AIC or BIC
include all available options for domain-relevant attributes (splitting variables).
Options are ‘region’, ‘administrative’, ‘population’, ‘imported’, and ‘airport’
Screenshot
let users screenshot the result
https://doi.org/10.1371/journal.pone.0265477.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
5 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
Fig 1. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 1.
https://doi.org/10.1371/journal.pone.0265477.g001
Fig 2. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 2.
https://doi.org/10.1371/journal.pone.0265477.g002
Fig 3. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 3.
https://doi.org/10.1371/journal.pone.0265477.g003
The following plot shows the time series line chart in gray and the average line in red (Fig
3). The coefficient plot displays OLS coefficients for predictors in all clusters (Fig 4). In other
words, each line represents one model connecting the coefficients for each predictor. This plot
is useful for users to choose the number of clusters. Also, by clicking on the coefficient points,
its value will appear in the box below. The final plots, the forecast error box, and density plots,
display forecast errors for the OLS and ETS methods on a one-week test set (Fig 5). In the
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
6 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
Fig 4. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 4.
https://doi.org/10.1371/journal.pone.0265477.g004
following tables, we first examine the linear models in each cluster and OLS results by comput-
ing Pearson [22] and concordance correlation coefficient [23] (between forecasted and
observed values). Then we compare OLS and ETS approaches using RMSE and MAE across
all series. Lastly, we presented the one-week-ahead forecast results (by updated model on com-
bined training and test sets) of all cities, townships, and districts in Taiwan computed by OLS
and ETS models (Fig 6). Users can download the forecasting results in an excel file by clicking
on the ‘Excel’ button next to tables.
Fig 5. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 5.
https://doi.org/10.1371/journal.pone.0265477.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
7 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
Fig 6. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 6.
https://doi.org/10.1371/journal.pone.0265477.g006
Figs 1 to 6 demonstrate the screenshots of our interactive tool results of Taiwan COVID-19
confirmed cases.
In Fig 1, in the top left side panel, we chose three as the MOB tree depth (two splits), AIC as
the prune option, and all splitting variables as domain-relevant attributes, which resulted in
three clusters differing in terms of ‘population’ and ‘region’. Changing options in the panel
update results shown in Figs 1 to 6. In Fig 1, the first split divides the series into population
more than 198795 and population less than 198795, and for more populated areas, there is no
further splits while in the less populated area there is one further split on the ‘region’, shows
series in central, east, south, islands (up) behave differently from north, null (imported cases)
(down).
Table 3 represents the final clusters of confirmed cases with the number of series.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
8 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
Table 3. Cluster categories of Taiwan COVID-19 confirmed infection cases by choosing three as the MOB depth,
AIC as pruning option, and region, population, imported, administrative, and airport as domain-relevant
attributes.
Cluster 1
Cluster 2
Cluster 3
Cluster categories
Population: more than 198795
Population: less than 198795
Region: central, east, south and islands
Population: less than 198795
Region: north and null (imported cases)
https://doi.org/10.1371/journal.pone.0265477.t003
Number of series
26
103
54
The heatmap in this figure shows in early June—when pick started—the number of con-
firmed cases is higher and more frequent (frequent dark green and red points—‘spiky’ series)
in the more populated areas (cluster 1), while the number of confirmed cases is fewer and less
frequent (frequent light green points) in the less populated areas (clusters 2 and 3). Also, the
diverse distribution of cases (time series temporal patterns), based on populations and regions,
is visible between the final clusters.
The heatmap in Fig 2 shows changes in the number of confirmed cases on different days of
the week. Based on this plot, the number of reported cases in all clusters is lower on Mondays
and Tuesdays and slightly higher on Sundays. Fig 3 shows the line chart of all series with their
average (red line) in each cluster. The comparison of the series and the average line in different
clusters shows the visible between-cluster variability. Clusters 1 (more populated areas) show
more confirmed cases. In cluster 3, the imported case series demonstrates a continuous report
of confirmed cases from the beginning of the year. Another series (Wanhua District) shows a
high jump in early June when the breakdown started in Taiwan.
Based on the coefficient plot in Fig 4, coefficients in all clusters differ mainly in terms of
lags (daily autocorrelation coefficients). The trend and seasonal dummies do not seem to vary
across clusters.
The final part of our web-based interactive tool is the forecast results displayed in Figs 5
and 6. We presented the forecasting performance on a one-week test set, using one OLS model
in each cluster, and compared it with forecasts generated by ETS, a more complex method.
Error box and density plots in Fig 5 show the one-week-ahead forecast errors of three clus-
ters using OLS and ETS models. Based on the error distribution of these two approaches, we
can see that for the Taiwan COVID-19 dataset, OLS performs significantly better. We compute
Pearson and concordance correlation coefficients between observed and forecasted values to
evaluate the OLS performance and forecast precision on each cluster. Based on these coeffi-
cients, the forecasting result, computed by three OLS models, is precise. We also compared
their performances using RMSE and MAE, and the results are the same as in plots. Lastly, we
present two tables in Fig 6 that indicate the one-week-ahead forecast for all cities, townships,
and districts in Taiwan using updated OLS and ETS models on combined training and test
sets in each cluster.
Conclusion
This research proposes an interactive web-based Shiny app for clustering and forecasting Tai-
wan COVID-19 confirmed infection cases. This tool is designed based on the MOB partition-
ing tree, cross-sectional attributes called domain-relevant attributes, and time series temporal
patterns (trend, seasonality, and autocorrelation). Our tool helps users analyze Taiwan
COVID-19 data via changing factors, including MOB depth, model complexity parameter
(AIC or BIC), and domain-relevant attributes.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
9 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
One advantage of our tool is grouping the series into interpretable clusters in which we can
label a certain cluster by its corresponding domain-relevant attributes. This MOB-based clus-
tering approach results in a single parametric OLS model in each cluster used to forecast all
series in that cluster. Clustering series into groups with similar temporal patterns led us to
enough accurate forecasts of Taiwan COVID-19 confirmed cases. This OLS forecasting
approach has low computational complexity in forecasting these cases.
Our clustering results determine the different spread patterns of confirmed infection cases
in the least populated in different regions and most populated areas. For example, the number
of confirmed cases in populated areas is higher than in other places. Also, the COVID-19 time
series shows different seasonality patterns on certain days of the week, higher on Sundays and
lower on Mondays and Tuesdays.
Another advantage of our tool is its usefulness in handling the existence of missing values
(missing completely at random (MCAR) or missing at random (MAR) variables)—displayed
in gray in heatmaps. In addition, users can have the most updated results of the COVID-19
transmission in Taiwan by simply updating the dataset in the tool. Although this tool is specifi-
cally designed for Taiwan COVID-19 confirmed cases, it can be easily applied to other regions
and/or countries with few changes and updates.
The OLS and ETS forecast results show an increase in infected cases in different cities. Note
that these results are before vaccine rollout, and we need to adjust the model to consider the
vaccination effect on the forecasting results. In addition, the concordance of our forecast is not
studied in this work, and we expect that the forecast has a no-more-than moderate concor-
dance. It is a future task to improve the concordance of our forecast.
Supporting information
S1 File.
(PDF)
Acknowledgments
The authors would like to thank Ms. Ula Tzu-Ning Kung for providing English editing service
in this paper.
Author Contributions
Conceptualization: Mahsa Ashouri, Frederick Kin Hing Phoa.
Formal analysis: Mahsa Ashouri.
Funding acquisition: Frederick Kin Hing Phoa.
Investigation: Mahsa Ashouri.
Methodology: Mahsa Ashouri, Frederick Kin Hing Phoa.
Software: Mahsa Ashouri.
Supervision: Frederick Kin Hing Phoa.
Validation: Mahsa Ashouri, Frederick Kin Hing Phoa.
Visualization: Mahsa Ashouri.
Writing – original draft: Mahsa Ashouri, Frederick Kin Hing Phoa.
Writing – review & editing: Frederick Kin Hing Phoa.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
10 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
References
1. Anastassopoulou C, Russo L, Tsakris A, Siettos C. Data-based analysis, modelling and forecasting of
the COVID-19 outbreak. PloS one. 2020; 15(3):e0230405. https://doi.org/10.1371/journal.pone.
0230405 PMID: 32231374
2. Perc M, Gorisˇek Miksić N, Slavinec M, Stozˇer A. Forecasting Covid-19. Frontiers in Physics. 2020;
8:127. https://doi.org/10.3389/fphy.2020.00127
3. Gopinath G. The great lockdown: Worst economic downturn since the Great Depression. IMFBlog–
Insights & analysis on economics & finance. 2020;.
4.
Is Taiwan’s COVID success story in jeopardy?;.
5. Aabed K, Lashin MM. An analytical study of the factors that influence COVID-19 spread. Saudi journal
of biological sciences. 2021; 28(2):1177–1195. https://doi.org/10.1016/j.sjbs.2020.11.067 PMID:
33262677
6. Rojas I, Rojas F, Valenzuela O. Estimation of COVID-19 dynamics in the different states of the United
States using Time-Series Clustering. medRxiv. 2020;.
7. Chen J, Yan J, Zhang P. Clustering US States by Time Series of COVID-19 New Case Counts with
Non-negative Matrix Factorization. arXiv preprint arXiv:201114412. 2020;.
8. Householder J, Householder A, Gomez-Reed JP, Park F, Zhang S. Clustering COVID-19 Lung Scans.
arXiv preprint arXiv:200909899. 2020;.
9. Petropoulos F, Makridakis S, Stylianou N. COVID-19: Forecasting confirmed cases and deaths with a
simple time-series model. International Journal of Forecasting. 2020;.
10. Petropoulos F, Makridakis S. Forecasting the novel coronavirus COVID-19. PloS one. 2020; 15(3):
e0231236. https://doi.org/10.1371/journal.pone.0231236 PMID: 32231392
11. Srivastava A, Xu T, Prasanna VK. Fast and Accurate Forecasting of COVID-19 Deaths Using the SIkJα
Model. arXiv preprint arXiv:200705180. 2020;.
12. RStudio, Inc. Shiny: web application framework for R.; 2013. Available from: http://www.rstudio.com/
shiny/.
13. RStudio Team. RStudio: Integrated Development Environment for R; 2020. Available from: http://www.
rstudio.com/.
14. Parker E. COVID-19 tracker.; 2019. Available from: https://shiny.rstudio.com/gallery/covid19-tracker.
html.
15.
Tebe´ C, Valls J, Satorra P, Tobı´as A. COVID19-world: a shiny application to perform comprehensive
country-specific data visualization for SARS-CoV-2 epidemic. BMC Medical Research Methodology.
2020; 20(1):1–7. https://doi.org/10.1186/s12874-020-01121-9 PMID: 32958001
16. Wissel BD, Van Camp P, Kouril M, Weis C, Glauser TA, White PS, et al. An interactive online dashboard
for tracking covid-19 in us counties, cities, and states in real time. Journal of the American Medical Infor-
matics Association. 2020;. https://doi.org/10.1093/jamia/ocaa071 PMID: 32333753
17. Ashouri M, Shmueli G, Sin CY. Tree-based methods for clustering time series using domain-relevant
attributes. Journal of Business Analytics. 2019; p. 1–23. https://doi.org/10.1080/2573234X.2019.
1645574
18.
Zeileis A, Hothorn T, Hornik K. Model-based recursive partitioning. Journal of Computational and
Graphical Statistics. 2008; 17(2):492–514. https://doi.org/10.1198/106186008X319331
19. Akaike H. Information theory and an extension of the maximum likelihood principle. In: Selected papers
of hirotugu akaike. Springer; 1998. p. 199–213.
20. Schwarz G, et al. Estimating the dimension of a model. The annals of statistics. 1978; 6(2):461–464.
https://doi.org/10.1214/aos/1176344136
21. Hyndman RJ, Athanasopoulos G. Forecasting: principles and practice. OTexts; 2018.
22.
23.
Freedman D, Pisani R, Purves R. Statistics (international student edition). Pisani, R Purves, 4th edn
WW Norton & Company, New York. 2007;.
Lawrence I, Lin K. A concordance correlation coefficient to evaluate reproducibility. Biometrics. 1989; p.
255–268.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
11 / 11
PLOS ONE
| null |
10.2196_41005.pdf
|
Data Availability
The deidentified data analyzed in this study are available from the corresponding author upon reasonable request.
|
Data Availability The deidentified data analyzed in this study are available from the corresponding author upon reasonable request.
|
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
Original Paper
An Unguided, Computerized Cognitive Behavioral Therapy
Intervention (TreadWill) in a Lower Middle-Income Country:
Pragmatic Randomized Controlled Trial
Arka Ghosh1, PhD; Rithwik J Cherian1,2, MSc; Surbhit Wagle1,3, MTech; Parth Sharma4, MTech; Karthikeyan R
Kannan1, BTECH; Alok Bajpai5, MBBS, MD, DPM; Nitin Gupta1,6, PhD
1Department of Biological Sciences and Bioengineering, Indian Institute of Technology Kanpur, Kanpur, India
2Department of Cognitive Science, Indian Institute of Technology Kanpur, Kanpur, India
3Institute of Physiological Chemistry, University Medical Center Mainz, Mainz, Germany
4Department of Computer Science and Engineering, Indian Institute of Technology Kanpur, Kanpur, India
5Counseling Service, Indian Institute of Technology Kanpur, Kanpur, India
6Mehta Family Center for Engineering in Medicine, Indian Institute of Technology Kanpur, Kanpur, India
Corresponding Author:
Nitin Gupta, PhD
Department of Biological Sciences and Bioengineering
Indian Institute of Technology Kanpur
IIT Campus
Kanpur, 208016
India
Phone: 91 5122594384
Email: [email protected]
Abstract
Background: Globally, most individuals who are susceptible to depression do not receive adequate or timely treatment. Unguided
computerized cognitive behavioral therapy (cCBT) has the potential to bridge this treatment gap. However, the real-world
effectiveness of unguided cCBT interventions, particularly in low- and middle-income countries (LMICs), remains inconclusive.
Objective:
In this study, we aimed to report the design and development of a new unguided cCBT–based multicomponent
intervention, TreadWill, and its pragmatic evaluation. TreadWill was designed to be fully automated, engaging, easy to use, and
accessible to LMICs.
Methods: To evaluate the effectiveness of TreadWill and the engagement level, we performed a double-blind, fully remote,
and randomized controlled trial with 598 participants in India and analyzed the data using a completer’s analysis.
Results: The users who completed at least half of the modules in TreadWill showed significant reduction in depression-related
(P=.04) and anxiety-related (P=.02) symptoms compared with the waitlist control. Compared with a plain-text version with the
same therapeutic content, the full-featured version of TreadWill showed significantly higher engagement (P=.01).
Conclusions: Our study provides a new resource and evidence for the use of unguided cCBT as a scalable intervention in
LMICs.
Trial Registration: ClinicalTrials.gov NCT03445598; https://clinicaltrials.gov/ct2/show/NCT03445598
(J Med Internet Res 2023;25:e41005) doi: 10.2196/41005
KEYWORDS
computerized cognitive behavioral therapy; cCBT; depression; digital intervention; mobile phone
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 1
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
Introduction
Background
Globally, >264 million individuals experience depressive
disorders [1]. Despite the availability of evidence-based
pharmacological and psychological treatment approaches, 76%
to 85% of the individuals experiencing mental health disorders
do not receive any treatment in low- and middle-income
countries (LMICs) [2]. Barriers to accessing treatment for mental
health disorders include the lack of access to treatment options,
high cost, the fear of social stigma, and an inclination to
self-manage the problem [3-5]. In India, there is a treatment
gap of 85.2% for major depressive disorders [6].
One approach to bridging this treatment gap is to deliver
computerized psychotherapy. The first therapeutic chatbot,
ELIZA, was developed in 1966 [7]; it was a rudimentary
program based on text rephrasing rather than evidence-based
methods. The first computer-assisted cognitive behavioral
therapy (CBT)–based program for depression was delivered in
1982 [8]. However, in the past 2 decades, the advent of stable
internet connection and the pervasiveness of smartphones and
computers have made it feasible to deploy technological
interventions at scale.
Computerized CBT (cCBT) has gained traction as a viable
treatment modality, with >200 trials conducted to date [9]. cCBT
for depressive disorders, both guided and unguided, has been
evaluated in several clinical trials worldwide. In both guided
and unguided cCBT interventions, the intervention is provided
by a software; in guided interventions, a guide or a coach is
additionally involved who provides encouragement, technical
assistance, and explanations of the intervention, whereas in a
strictly unguided intervention, the user should not have any
interaction with a human guide. Recent studies and
meta-analyses have indicated that for depressive symptoms,
guided cCBT interventions are more beneficial than unguided
cCBT interventions [10-14]. Carlbring et al [15] showed
equivalent effects between guided cCBT interventions and
face-to-face CBT. Including guided cCBT intervention with
treatment as usual does not add any extra benefits [16].
Moreover, although guided cCBT intervention can be a feasible
option in high-income countries [17], it is not feasible in LMICs
because of the acute shortage of mental health professionals
who can act as qualified guides [18].
Unguided cCBT interventions have the potential to bridge the
treatment gap in LMICs. The evidence for unguided cCBT
interventions is mixed, with some meta-analyses showing that
they are effective with a small or medium effect size [14,19,20]
and some showing that they are not effective [21-23]. The
effectiveness of the unguided interventions is reduced by the
high dropout rates. Note that the unguided studies included in
the meta-analyses often involved initial contact with humans
for diagnostic interviews [24-29], weekly telephone contact
support [30,31], or treatment as usual [16,32]. Even minimal
human contact can increase adherence to the interventions
compared with a study without any such contact [33,34]. Indeed,
Fleming et al [35] found that adherence rates observed in trial
settings failed to translate into the real world. Recent
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
meta-analyses have reported a positive correlation between
treatment adherence and treatment effects [14,19]. In addition,
a recent meta-analysis found that existing guided or unguided
cCBT interventions had low acceptability among patients, which
was even less than that of waitlist [10].
interventions have been conducted
Studies on cCBT
predominantly
in high-income countries [36]; however,
systematic reviews on depression and mental health disorders
in LMICs have been done by Martínez et al [37] and Fu et al
[38], respectively. A recent meta-analysis reported that 92% of
the studies on diagnosed depression had been conducted in
Western Europe, North America, and Australia [39]. The
interventions have been developed, evaluated, and made
available for free only in these high-income regions. There is a
need for unguided interventions that are more effective, have
higher adherence, and are available free of cost for wide
accessibility in LMICs.
Objective
In this study, we developed and evaluated such a cCBT-based
multicomponent intervention, TreadWill. We included several
features in TreadWill that could increase adherence to and
improve the effectiveness of a completely unguided intervention.
We also developed an active control version of TreadWill that
presented the same therapeutic content without these features.
We designed a fully remote 3-armed randomized controlled
trial (RCT)—an experimental version of TreadWill, a plain-text
version of TreadWill with the same therapeutic CBT content
(active control), and a waitlist control. We hypothesized that
the participants in the experimental group would show
significantly greater improvement in depressive and anxiety
symptom severity. We also hypothesized that the participants
in the experimental group would show significantly more
engagement in terms of modules completed and absolute time
used compared with the active control group participants.
Methods
Study Design
We designed a fully remote RCT to test the effectiveness of the
experimental version of TreadWill compared with an active
control version and a waitlist control version. We planned to
recruit 600 participants with a 1:1:1 distribution across the 3
groups. We implemented simple randomization using an
automated randomization function (developed in Python; version
3.4.3; Python Software Foundation). This trial was registered
at ClinicalTrials.gov before commencement (NCT03445598).
Participant Recruitment and Screening
We recruited the participants using both offline and web-based
publicity. We displayed flyers in residential hostels, research
buildings, and lecture halls at the Indian Institute of Technology,
Kanpur. A press release helped with coverage in newspapers
and social media.
The publicity material included a website link to join this study.
The link opened a web page that provided information regarding
the study and accepted the email ID of the interested
participants. Over the next 3 steps, the web page collected the
J Med Internet Res 2023 | vol. 25 | e41005 | p. 2
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
it (the
demographic data, baseline Patient Health Questionnaire-9
(PHQ-9) score [40], and informed consent from the potential
participants. The entire participant recruitment process was
automated
(including self-reports and self-administered
questionnaires) to eliminate human contact and maintain
scalability. To be eligible to participate in the study, an
individual must be an Indian resident aged between 16 and 35
years. They must be fluent in English and have had access to
an internet-enabled computer or tablet device. They must have
had scored between 5 and 19 (both inclusive) in the PHQ-9 with
a score of 0 on the ninth question. We decided to include
participants with mild symptoms of depression (a score of 5-9
in the PHQ-9) and exclude those with severe symptoms (a score
>19) because our program was targeted not at the clinical
population but at a wider population with susceptibility for
depression. We excluded individuals who were unemployed,
had a diagnosis of bipolar disorder or psychosis, or reported
that they just wanted to check out the site and did not plan to
complete
trial
commencement to exclude casual visitors to the website).
Because of the pragmatic nature of our study, we included
participants regardless of whether they were receiving treatment
for depression. Once a potential participant met the inclusion
and exclusion criteria and provided informed consent (for
individuals aged between 16 and 17 years, informed consent
was also required from a parent or guardian), the individual was
scheduled to be recruited in the study. After 18 hours, the
individual was randomized to 1 of the 3 groups and received a
unique link via email. The delay of 18 hours was included to
prevent individuals from signing up using disposable temporary
email IDs. They were counted as participants in the study only
after clicking on the unique link and were led to a sign-up page
(for participants assigned to the experimental or active control
groups). The participants assigned to the waitlist control group
were led to a page to collect their baseline Generalized Anxiety
Disorder-7 (GAD-7) scores [41]; GAD-7 scores of the
experimental and active control groups were taken just before
the start of the first module in the intervention. The participants
did not receive any monetary compensation.
last condition was added after
Ethics Approval
The Institutional Ethics Committee of the Indian Institute of
Technology Kanpur provided ethical clearance to conduct this
study (IITK/IEC/2017-18 II/1).
Safety Check
At any stage in the intervention, if we detected severe depressive
symptoms or suicidal ideation, we blocked access to TreadWill.
Severe depressive symptoms were determined as a total score
of >19 on the PHQ-9. Suicidal ideation was detected by a score
of >0 on the ninth question of the PHQ-9 and a total score of
>4 on the Suicidal Intent Questionnaire [42]. In such cases,
email and SMS text messaging alerts were sent to the
participants (and their buddy, if they had one in the program),
requesting them to seek professional help. For participants aged
between 16 and 17 years, an email notification was sent to the
parent or guardian as well. The participants had not been
informed of this exclusion criterion; therefore, they did not
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
intentionally suppress their scores for the sake of continuing
the intervention.
Automated Notifications
Participant contact was minimal and automated. Participants
who initiated the recruitment process but did not complete it
were sent automated email reminders encouraging them to
complete the process. Participants also received periodic
automated email and SMS text messaging reminders nudging
them to use TreadWill (Table S1 in Multimedia Appendix 1
provides details). The program asked the participants about their
preferred time to log in; using this information, email and SMS
text messaging alerts were sent 10 minutes in advance to remind
the participants. The research team did not initiate any direct
contact with the participants. Technical support via email was
provided in case the participants sent an email requesting for
it.
Active Control Version
The active control version presented the same CBT content as
the experimental version in the same 6 modules, but used plain
text instead of slides, videos, and conversations. Each module
had Introduction, Learn, and Discuss sections, but the Practice
section was excluded. The content was not tailored according
to the participant. The active control version included the CBT
forms, but excluded games, such as SupportGroup, PeerGroup,
and the option to involve a buddy. The participants received
only essential email notifications (Table S1 in Multimedia
Appendix 1 presents the details of notifications). The active
control version was introduced to test whether the additional
interactive elements introduced in the experimental version
increased user engagement.
Development of the Intervention
We used the Django framework (Django Software Foundation)
for developing the TreadWill website. We used Google Slides
(Google LLC) to embed the slides and YouTube (Google LLC)
to embed the videos on the website. We used images with a
Creative Commons license for use in slides and videos. We
used images from the internet for the Identify the friendly face
game [43]. The content and the website underwent multiple
rounds of checking by the development team and other
volunteers to fix errors before launching the trial.
Assessments
We used the PHQ-9 [40] and GAD-7 [41] questionnaires to
measure depressive and anxiety symptom severity, respectively.
For the experimental and active control group participants, the
PHQ-9 and GAD-7 were administered before the beginning of
each module, after completing all the modules, and at the 90-day
follow-up time point. The first PHQ-9 (administered before
randomization, as it was an inclusion-exclusion criterion) and
the first GAD-7 (administered after randomization but before
the first module of the intervention) served as the baseline
scores. For the waitlist control group, the PHQ-9 and GAD-7
were administered at baseline and after a 42-day interval (this
interval was chosen to be at par with the expected intervention
duration of approximately 6 weeks for completing the 6 modules
in the experimental group). After submitting the 42-day
J Med Internet Res 2023 | vol. 25 | e41005 | p. 3
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
assessments, the waitlisted participants were also given access
to the intervention.
Blinding
All the participants followed the same recruitment procedure.
Consequently, the participants were unaware of which version
of TreadWill they were assigned to (they did not even know
that 2 different versions existed). Therefore, we expected
placebo effects in the 2 groups to be similar. In addition, the
PHQ-9 and GAD-7 data were self-reported on the website;
therefore, there was no scope for evaluator bias.
Data Security and Privacy
All the participants agreed to allow their data to be used for
research purposes and to be reported in a deidentified format.
All participant data were transferred over Secure Sockets Layer.
The only personal identifiers provided by the participants were
their email IDs and phone numbers. Before analyzing the data,
all the participants’ email IDs and phone numbers were removed
from the data set.
Primary, Secondary, and Exploratory End Points
We performed a completer’s analysis (Discussion section). The
primary end point was the final PHQ-9 score in participants
who completed at least half of the intervention (3 out of 6
modules). The primary end point was decided after the trial
completion but before any data analysis. We decided the cutoff
point at 3 modules to ensure that all the participants were
exposed to the cognitive aspect of CBT, which we introduced
in the third module. We also analyzed the data of participants
who had completed all the modules. A similar analysis approach
based on module completion in web-based studies has been
used by Christensen et al [44], Keefe et al [45], and Rollman
et al [46] (the Discussion section elaborates on the rationale for
using this analysis approach).
TreadWill was primarily designed to help individuals with
depressive symptoms. Therefore, PHQ-9 was our primary
outcome measure. However, as anxiety and depression are
highly comorbid, we wanted to check whether the techniques
presented in TreadWill also helped in the reduction of anxiety
symptoms. Thus, for the experimental and the active control
groups, the secondary end point was the GAD-7 score in
participants who completed at least half of the intervention (3
out of 6 modules). Other secondary end points included PHQ-9
and GAD-7 scores at the 90-day follow-up.
The intermediate PHQ-9 and GAD-7 scores (after every module)
and 2 surveys conducted after the module 3 and the module 6
were used as exploratory end points.
Statistical Analyses
Owing to the high dropout rate, we did not assume the PHQ-9
and GAD-7 scores to be normally distributed; therefore, we
used nonparametric statistical
the
effectiveness of the intervention. We used the Kruskal-Wallis
test for comparing the reduction in depression or anxiety
symptom severity from baseline to the primary end point among
the 3 groups. All the tests were 2 tailed unless otherwise
mentioned. For post hoc analysis between the groups, we used
tests for analyzing
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
the Mann-Whitney U test. The tests were conducted using
MATLAB (MathWorks) and Python (Python Software
Foundation).
Because this was the first trial of TreadWill, we did not have a
prior estimate of the dropout rate and could not perform power
calculations. We chose the sample size of 600 participants based
on the previous studies of similar nature [16,47].
Results
Approach Taken for Developing the Intervention
We aimed to develop and evaluate a fully automated
intervention, TreadWill, that would be engaging and effective
without any expert guidance or contact. We reviewed the
existing cCBT interventions before starting the development
process and considered factors that may be responsible for the
high dropout rates. The common shortcomings that we identified
included the lack of interactive content, lack of tailoring of the
content to different users, lack of peer support for users, and
lack of engaging games. Different interventions addressed some
of these shortcomings by including the corresponding features;
however, none of them included all the features. We developed
TreadWill
the
development process, we used the inputs on initial prototypes
from the institute counselors and psychiatrists and from 13 pilot
users (not included in the eventual trial), before finalizing the
content and user experience in TreadWill. We hypothesized
that TreadWill would lead to a high adherence rate and a
significant reduction in depressive and anxiety symptom
severity. As we did not plan to charge the users, we also
expected TreadWill to be more accessible, especially in LMICs,
compared with paid interventions.
these shortcomings. During
to address
Design of TreadWill
We designed the therapeutic content of TreadWill based on
CBT, using the book by Beck [48] as the primary reference.
TreadWill delivered the core concepts of CBT in a structured
format with 6 modules (Table S2 in Multimedia Appendix 1
shows the details) in an easily understandable language. Each
module consisted of 4 sections: Introduction, Learn, Discuss,
and Practice. In the Introduction section, an automated virtual
therapist explained the importance of the module through
interactive text-based dialogue. The Learn section included
psychoeducation in the form of slides and videos. Slides
consisted of multiple infographics that were presented
sequentially (Figure S1 in Multimedia Appendix 1). Videos
consisted of animated content with a voiceover explaining the
concepts that were visible on the screen. In the Discuss section,
the participants learned to apply the psychoeducation to real-life
situations through conversations. These conversations were
text-based dialogues with an automated virtual patient (Figure
S2 in Multimedia Appendix 1), presented in an interactive
format designed to simulate human chat. Although the
conversations were preprogrammed, in many instances, the
participants could choose from >1 response, thus providing
some control to the user in steering the dialogue. The Practice
section included interactive quizzes on the material covered in
each module.
J Med Internet Res 2023 | vol. 25 | e41005 | p. 4
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
To ensure sequential progression through the intervention, only
the first module was initially accessible to the user and the later
modules were locked. After the completion of all sections in a
module and a 4-day gap since its unlocking (to prevent rushing
through the modules), the next module was unlocked. Steps
within a module were also unlocked gradually upon the
completion of the preceding steps. Each participant had a
maximum of 90 days to complete the 6 modules starting from
their first log-in. After 90 days, they could continue to use the
modules that were already unlocked until then but could not
unlock new modules. We did this to restrict participants’
exposure to new therapeutic content after 90 days and provide
a clear deadline, as recommended by several studies [49-51].
Interactive Games and CBT Forms in TreadWill
TreadWill included 2 interactive games. The Identifying thinking
errors game was aimed at training the participants in spotting
thinking errors in their negative automatic thoughts. The
gameplay involved the presentation of a situation, a related
negative automatic thought, and a list of 10 thinking errors from
which the participant had to select one or more thinking errors
present in the thought. Selecting the correct option allowed the
participant to move to the next level. When an incorrect option
was selected, feedback was provided along with an opportunity
to try again. The Identify the friendly face game is based on the
training paradigm developed by Dandeneau and Baldwin [52]
to train participants to overcome the negative attention bias and
improve their self-esteem, thereby reducing the risk of
depression [53,54]. The game presented 4 images in a 2×2 grid
with 3 faces showing a negative emotion and 1 face showing a
positive emotion. The participant was allowed 5 seconds to find
the positive image and thus increase their score. If the participant
responded or if 5 seconds elapsed, a new set of images was
displayed. The gameplay incentivized quick attention to positive
emotions. The difficulty of the game continuously adapted to
the participants’ competence: incorrect responses increased the
frequency of faces with obvious emotions, and correct responses
increased the frequency of faces with subtle emotions.
TreadWill provided an interactive interface to fill in the forms
commonly used in CBT: Thought record worksheet, Core belief
worksheet, Behavioral experiment worksheet, Problem-solving
worksheet, Prepare for setback worksheet, and Schedule activity
worksheet (Table S3 in Multimedia Appendix 1). The forms
allowed participants to apply CBT techniques to their situations
and save the information for future reference.
Peer and Family Support in TreadWill
Individuals looking for support on the internet might have low
social support in real life [55]. In such cases, web-based
peer-based support has been shown to be effective in reducing
depressive symptoms [56,57]. Keeping this in mind, we designed
the SupportGroup and PeerGroup features in TreadWill to
provide a social space where participants could connect with
other TreadWill users and potentially help each other in solving
their problems. Posts in the SupportGroup were visible to all
the TreadWill participants. The participants could upvote or
downvote posts, add comments, and send thank you messages
to each other. PeerGroups were smaller groups of 10 members
each, designed in such a way that the posts in a PeerGroup were
visible only to the members of that PeerGroup.
We provided the participants with the option to invite a family
member or a friend as their buddy who would receive weekly
updates about the participant’s activities in TreadWill. We
hypothesized that the involvement of the buddy would motivate
the participants to complete the program. We sent an email to
this buddy if the participant failed to use TreadWill regularly
and requested them to nudge the participant.
Content Tailoring in TreadWill
Content tailoring has the potential to increase adherence to
cCBT interventions, as participants are more likely to stick with
a program if they find the content relatable [50,58,59]. In
TreadWill, we implemented tailoring by selecting examples in
the conversations based on the participant’s occupation (high
school students, college students, or working professionals). In
addition, we tailored the conversations based on participants’
thoughts, beliefs, and situations in the following manner. First,
we asked the participants to select relatable intermediate and
core beliefs from the Dysfunctional Attitude Scale [60], negative
automatic thoughts from the Automatic Thoughts Questionnaire
[61], and stressful situations from a curated list. Then, we made
the simulated virtual patients in the subsequent conversations
identify with similar beliefs, thoughts, and situations, and the
participant’s goal was to help the simulated patient by using the
CBT techniques learned in that module.
The automated email and SMS text messaging notifications
received by the users were also tailored according to their
preferences (Methods section).
Participants
Recruitment commenced on February 14, 2018, with a planned
enrollment of 600 participants. The primary completion date
was March 2, 2019, after full enrollment, and the secondary
completion date was May 31, 2019. Of the 5188 individuals
who started the registration process for the study, 598 (11.53%)
participants completed all the steps and met the study inclusion
criteria (2 other participants who did not meet the inclusion
criteria were initially included owing to a software bug but were
excluded when we cross-checked the data during data analysis).
The 598 participants were randomly assigned to the 3 study
arms with equal probability (Methods section), resulting in 204
experimental, 189 active control, and 205 waitlist control
participants (Figure 1). The participants in the 3 groups were
found to be balanced in terms of age, sex, the severity of
depressive symptoms, occupation, the use of other interventions,
motivation for joining, and the occurrence of recent traumatic
events (Table 1). The baseline PHQ-9 scores in the 3 groups
were not significantly different: Kruskal-Wallis H(2)=2.04
(P=.36). However, a sex bias (478/598, 79.9% male) was
observed because participants in our study were recruited mainly
from Indian engineering colleges where the students were
predominantly male [62]. In addition, in India, there is a 56%
gender gap in mobile internet use [63].
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 5
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
Figure 1. The flow of participants in the trial. In the experimental and the active control groups, the follow-up scores of only those participants who
had completed at least 3 modules were analyzed. In the waitlist group, the 42-day interval scores of only those participants who had also submitted the
baseline scores were analyzed. GAD-7: Generalized Anxiety Disorder-7; PHQ-9: Patient Health Questionnaire-9.
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 6
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
Table 1. Baseline and demographic characteristics of the participants recruited in the study.a
Groups
Experimental (n=204)
Active control
(n=189)
Waitlist con-
trol (n=205)
All (n=598)
Group comparison—test result
H(2)
Chi-square (df; n=598)
P value
Age (years), mean (SE)
23.76 (0.30)
23.42 (0.28)
23.48 (0.29)
23.56 (0.17)
0.205
N/Ab
Sex, n (%)
Male
Female
160 (78.4)
44 (21.6)
151 (79.9)
167 (81.5)
478 (79.9)
38 (20.1)
38 (18.5)
120 (20.1)
N/A
0.586 (2)
Traumatic event or death of a loved one, n (%)
N/A
0.199 (2)
Yes
No
Joining for help, n (%)
Yes
No
22 (10.8)
182 (89.2)
180 (88.2)
24 (11.8)
18 (9.5)
20 (9.8)
60 (10)
171 (90.5)
185 (90.2)
538 (90)
157 (83.1)
180 (87.8)
517 (86.5)
32 (16.9)
25 (12.2)
81 (13.5)
N/A
2.722 (2)
Secondary help, n (%)
N/A
4.968 (6)
.90
.75
.91
.26
.55
185 (90.7)
178 (94.2)
193 (94.1)
556 (93)
None
Counseling
Medication
Both
5 (2.5)
12 (5.9)
2 (1)
Occupation, n (%)
High school Student
1 (0.5)
Between school and
college
8 (3.9)
4 (2.1)
4 (2.1)
3 (1.6)
2 (1.1)
4 (2.1)
4 (2)
6 (2.9)
2 (1)
4 (2)
6 (2.9)
13 (2.2)
22 (3.7)
7 (1.2)
7 (1.2)
18 (3)
N/A
7.620 (14)
.91
College student
113 (55.4)
103 (54.5)
120 (58.5)
336 (56.2)
Coaching after college
33 (16.2)
33 (17.5)
30 (14.6)
96 (16.1)
Working professionals
39 (19.1)
40 (21.2)
37 (18)
116 (19.4)
Self-employed
Freelancers
Volunteers
PHQ-9c, mean (SE)
7 (3.4)
2 (1)
1 (0.5)
4 (2.1)
3 (1.6)
0 (0)
4 (2)
2 (1)
2 (1)
15 (2.5)
7 (1.2)
3 (0.5)
10.76 (0.26)
10.81 (0.25)
10.42 (0.27)
10.66 (0.15)
2.04
N/A
.36
aThe 3 groups were not statistically different in these characteristics, as indicated by the statistical tests reported in the last column.
bN/A: not applicable.
cPHQ-9: Patient Health Questionnaire-9.
Effectiveness of TreadWill
In the primary analysis, we included the participants who
completed at least 3 modules in the experimental group or the
active control group. For this analysis, we used the last PHQ-9
scores submitted by these participants, excluding the follow-up
questionnaire. Henceforth, we refer to the time of these last
scores as the primary end point. In the waitlist control group,
all users who submitted the questionnaires after the waiting
period were included in the analysis.
We compared the reductions in the PHQ-9 scores from the
baseline to the primary end point between the 3 groups (Figures
2A and 2B; Table 2). The 3 groups showed significant
differences in the reductions in the PHQ-9 score (Kruskal-Wallis
test H(2)=8.93; P=.01); a post hoc test with Bonferroni
correction revealed that the experimental group showed a larger
reduction than the waitlist control group (2.73 vs 1.12;
Mann-Whitney U=1027; experimental group: n=22; waitlist
control group: n=139; P=.04). The differences in PHQ-9
reductions between the experimental and the active control
groups were not significant (U=96; experimental group: n=22;
active control group: n=7; P=.34).
in
the
reductions
In secondary analysis, the 3 groups showed significant
differences
score
(Kruskal-Wallis test H(2)=8.02; P=.02); a post hoc test with
Bonferroni correction showed a larger reduction in the
experimental group than in the waitlist control group (3.27 vs
0.89; Mann-Whitney U=637.50; experimental group: n=22;
the GAD-7
in
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 7
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
waitlist control group: n=94; P=.02). The differences in the
GAD-7 reductions between the experimental and the active
control groups were not significant (U=52.5; experimental
group: n=22; active control group: n=7; P=.22).
We also checked the reduction in the PHQ-9 and GAD-7 scores
for the smaller set of the experimental group participants who
completed all 6 modules (Figures 2C and 2D; Table 2); this
analysis could not be performed for the active control group
because only 1 participant from that group completed all 6
modules. This analysis also showed that the experimental group
had a significantly larger reduction in PHQ-9 scores compared
with the waitlist control group (4.20 vs 1.12; Mann-Whitney
U=368.5; experimental group: n=10; waitlist control group:
n=139; P=.01) and GAD-7 scores (3.40 vs 0.89; Mann-Whitney
U=260.5; experimental group: n=10; waitlist control group:
n=94; P=.02). The participants who completed all modules in
the experimental and the active control groups did not differ
demographically or in their baseline PHQ-9 scores from the
rest of the participants (Table S4 in Multimedia Appendix 1).
The reductions observed in the PHQ-9 and GAD-7 scores in
the experimental and the active control groups at the primary
end point were maintained at the 90-day follow-up period
(Figures 2A and 2B). Thus, both the full-featured version of
TreadWill (experimental) and the plain-text version of TreadWill
(active control) were effective in reducing depression- and
anxiety-related symptoms in participants who completed all or
at least 3 modules.
We checked whether the novel features of the experimental
version of TreadWill were able to increase engagement
compared with the active control version. Every module was
completed by more participants in the experimental version
than in the active control version (Figure 3A). The odds of
completing at least 3 modules were 3 times higher for a
participant in the experimental group compared with a
participant in the active control group (odds ratio 3.004, 95%
CI 1.247-7.237; P=.01). The experimental group participants
used TreadWill for an average of 79.8 minutes and the active
control group participants for 26.1 minutes; the difference was
statistically significant (Mann-Whitney U=10,290; experimental
group, n=181; active control group, n=159; P<.001; Figure 3B).
Thus, the full-featured version of TreadWill had higher
engagement and less attrition than the plain-text version.
Furthermore, we checked whether the level of engagement with
TreadWill was related to the reductions in depressive and
anxiety symptoms. We found that the reduction in the PHQ-9
score was positively correlated with the number of modules
completed within each group (experimental group: Spearman
ρ=0.38; P=.003; n=61; Figure 3C; active control group: ρ=0.51;
P<.001; n=41; Figure 3D) and with the total use time
(experimental group: ρ=0.39; P=.002; n=61; Figure 3E; active
control group: ρ=0.47; P=.002; n=41; Figure 3F). The reduction
in the GAD-7 score was also moderately correlated with the
number of modules completed (experimental group: ρ=0.27;
P=.04; n=57; Figure S3A in Multimedia Appendix 1; active
control group: ρ=0.43; P=.009; n=37; Figure S3B in Multimedia
Appendix 1) and total use time (experimental group: ρ=0.25;
P=.07; n=57; Figure S3C in Multimedia Appendix 1; active
control group: ρ=0.35; P=.04; n=37; Figure S3D in Multimedia
Appendix 1).
Figure 2. Changes in Patient Health Questionnaire-9 (PHQ-9) and Generalized Anxiety Disorder-7 (GAD-7) scores after using TreadWill. (A) and
(B) Violin plots show PHQ-9 (A) or GAD-7 (B) scores at baseline, primary end point, and follow-up for the experimental group, the active control
group, and the waitlist group participants. Primary end point is defined as the latest PHQ-9 or GAD-7 score submitted after completing at least 3 modules.
For PHQ-9, experimental group: n=22, active control group: n=7, waitlist group: n=139; for GAD-7, experimental group: n=22, active control group:
n=7, waitlist group: n=94. (C) and (D) Violin plots show the change from baseline to program completion in PHQ-9 (C) or GAD-7 (D) score for the
experimental group participants who completed all 6 modules (blue violin). For waitlist group participants (orange violin), the plots show the change
from the score at the baseline to the score after the 42-day waiting interval (considered as the primary end point for the waitlist group). Red horizontal
lines: median; black: mean. Error bars represent SE.
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 8
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
Table 2. Average changes (SE) in Patient Health Questionnaire-9 (PHQ-9) and Generalized Anxiety Disorder-7 (GAD-7) scores for the 3 groups from
the baseline to the primary end point or to the completion of all modules.a
Groups
Experimental (n=204)
Active control (n=189)
Waitlist control (n=205)
Average change
(SE)
Values, n (%)
Average change
(SE)
Values, n (%)
Average change
(SE)
Values, n (%)
PHQ-9
Primary end point
−2.73 (1.27)
22 (10.8)
−5.14 (2.28)
7 (3.7)
−1.12 (0.37)
139 (67.8)
Completion of all modules
−4.20 (0.83)
10 (4.9)
—b
—
—
—
GAD-7
Primary end point
−3.27 (0.97)
22 (10.8)
−1.43 (0.92)
7 (3.7)
−0.89 (0.42)
94 (45.9)
Completion of all modules
−3.40 (0.82)
10 (4.9)
—
—
—
—
aAs only 1 participant in the active control group completed all modules, the corresponding values were not analyzed.
bNot available.
Figure 3. Adherence with TreadWill and the relationship between intervention use and symptom reductions. (A) The graph shows the number of
participants in the experimental (blue) and the active control (red) groups who completed the indicated number of modules. (B) Violin plots show the
total use times of the experimental and the active control group participants. Red horizontal lines: median, black: mean. Error bars represent SE. (C)
and (D) The reduction in Patient Health Questionnaire-9 (PHQ-9) scores versus the number of modules completed by the experimental group participants
(C) and the active control group participants (D). (E) and (F) The reduction in PHQ-9 scores versus the total use time in hours for the experimental
group participants (E) and the active control group participants (F). In all cases, the reduction in PHQ-9 scores was calculated by subtracting the last
PHQ-9 score (excluding follow-up) from the baseline score; a positive value indicates improvement. Some points in the graphs are overlapping.
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 9
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
Evaluating the Role of Possible Confounding Factors
Differences in time and motivation can act as confounding
factors in the performance of an intervention. To test whether
the observed differences between the experimental and the
waitlist groups were affected by these factors, we performed
additional analyses.
We had planned to take the PHQ-9 and GAD-7 scores in the
waitlist group at 42 days, as the experimental group participants
were also expected to take 42 days to submit the final
questionnaire (6 modules at the rate of 1 module per week).
However, variability in the actual timing of score submission
was inevitable in a fully unguided and remote study. In our data
set, we found that the actual timing of the final questionnaire
was 63.7 (SD 26.8) days for the experimental group and 47.4
(SD 10.5) days for the waitlist group participants. To check if
this difference in timing can explain the difference in the
performance, we split the waitlist group participants into 2
subgroups depending on when they submitted the PHQ-9
questionnaires: the first subgroup included participants who
had submitted before 46 days (mean 43.56, SD 1.10 days;
nWL1=99), and the second subgroup included participants who
submitted after 46 days (mean 57.05, SD 15.91 days; nWL2=40);
by design, the mean number of days for the 2 subgroups were
significantly different (Mann-Whitney U=3960; P<.001).
However, the mean reductions in PHQ-9 scores for these 2
subgroups were not significantly different (U=1987.5; P=.97).
This indicates that for the waitlist group, the difference in the
number of days in the observed range did not affect the PHQ-9
scores significantly.
To check whether the higher reduction in the PHQ-9 scores in
the experimental group than in the waitlist group can be
explained by motivation, we performed the following analysis.
In our study, the waitlist group participants were given the
option to sign up for the experimental version of the intervention
once the waitlist period was over (ie, when their formal
participation in the study had ended, they were not considered
as experimental group participants). It is reasonable to expect
that the waitlist group participants who actually signed up for
this option, despite the long gap of at least 42 days, were more
motivated than the rest. We created a subgroup of these more
motivated waitlist group participants and compared their
performances with that of the remaining participants. These 2
subgroups did not show a significant difference in the reductions
in the PHQ-9 scores (U=2160; motivated: n=64; unmotivated:
n=75; P=.31).
Another potential concern is that the users who happened to
improve spontaneously may be likely to complete more
modules; by performing a completer’s analysis, we may be
selecting for such spontaneous improvers. We performed an
additional analysis to check whether this was the case in our
data. On the basis of this argument, the participants who went
on to complete module 3 after completing module 2 would have
seen more improvement in their PHQ-9 scores at the end of
module 2 compared with the participants who dropped out just
after completing module 2. We compared the reductions in
PHQ-9 scores (from baseline to the end of module 2) of these
2 subgroups and found no significant difference (U=93; dropout:
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
n=9; continued: n=22; P=.81). Similarly, we compared the
reductions in PHQ-9 scores (from baseline to the end of module
1) of participants who dropped out after completing module 1
and those who went on to complete the next module, and we
did not find any significant difference (U=488; dropout: n=30;
continued: n=31; P=.74). Thus, the idea that (spontaneous)
improvement in performance encourages the participants to
complete more modules is not supported by our data.
On the basis of these analyses, we conclude that the higher
reduction in PHQ-9 scores observed in the experimental group
can be attributed to the effect of completion of the modules,
rather than differences in the timing of questionnaires or in
motivation.
Feedback on the Features of TreadWill
We programmed TreadWill to present surveys containing 15
questions using a 5-point Likert scale to quantify the
participants’ feedback on various aspects of TreadWill. For
example, one of the questions stated I found the email reminders
helpful, to which the participant responded by selecting one of
the following options: strongly agree, somewhat agree, neither
agree nor disagree, somewhat disagree, and strongly disagree,
which were mapped to a score of 2, 1, 0, −1, and −2, respectively
(Table S5 in Multimedia Appendix 1 lists all questions). The
surveys were conducted at 2 time points: after completing 3
modules and at the end of the intervention.
In the experimental group, of the 22 participants who completed
at least 3 modules, the first survey was submitted by 22 (100%)
participants and the second survey was submitted by 18 (82%)
participants. The participants reported positive feedback on
most aspects of TreadWill (Figure 4A): mean feedback scores
over all questions were significantly >0 for both the first survey
(mean 1.16, SE 0.12; n=15 questions; t21=9.20; P<.001; 2-tailed
t test) and the second survey (mean 1.29, SE 0.10; n=15
questions; t17=12.56; P<.001; 2-tailed t
test). The scores
remained largely consistent between the 2 surveys (Pearson
r=0.87; P<.001; n=15). The strongest positive feedback was
received for questions related to the ease of English used (mean
1.86, SE 0.10 in the first survey and mean 1.83, SE 0.12 in the
second survey), the relatability of the examples (mean 1.23, SE
0.25 and mean 1.72, SE 0.13, respectively), the ease of using
the CBT forms (mean 1.45, SE 0.18 and mean 1.22, SE 0.17),
the engaging nature of the conversations (mean 1.36, SE 0.21
and mean 1.50, SE 0.20), the helpfulness of the Learning slides
(mean 1.73, SE 0.10 and mean 1.67, SE 0.14), and the
helpfulness of the Learning videos (mean 1.55, SE 0.13 and
mean 1.67, SE 0.14). The features with the lowest ratings
included the PeerGroup, which received weak positive feedback
(mean 0.73, SE 0.23 and mean 0.61, SE 0.28) and the buddy
feature, which received neutral feedback in both surveys (mean
0.09, SE 0.22 and mean 0.39, SE 0.20).
In the active control group, of the 7 participants who completed
at least 3 modules, 7 (100%) and 5 (71%) participants submitted
their first and second surveys, respectively. The survey questions
were slightly different in the active control group (Table S5 in
Multimedia Appendix 1); the first 6 questions judged their
opinion on aspects they experienced directly, and the next 9
J Med Internet Res 2023 | vol. 25 | e41005 | p. 10
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
questions were asked in a prospective manner, for example, I
would prefer to have email reminders. In the first 6 questions,
the participants reported overall positive feedback (Figure 4B).
In the 9 prospective questions, they showed interest in having
only some of the proposed features, including conversations
and game elements. Curiously, many features that the active
they would not
control group participants
thought
prefer—including SMS text messaging reminders, videos, and
slides—were actually found to be useful by the experimental
group participants who experienced the features (Figures 4A
and 4B).
No participant reported any adverse events through the contact
form on the website.
Figure 4. Feedback on the features of TreadWill. Violin plots show survey responses by the experimental group (A) and the active control group
participants (B). Y-axis labels: 2=“strongly agree,” 1=“somewhat agree,” 0=“neither agree nor disagree,” −1=“somewhat disagree,” and −2=“strongly
disagree.” Black lines indicate mean (SE).
Exploratory Analysis
To check if the content provided was engaging, we provided
the experimental group participants with the option to provide
feedback on the slides, videos, and conversations using like and
dislike buttons. The slides, videos, and conversations were
viewed 467, 205, and 1479 times, respectively, over all modules,
of which nearly 17.1% (80/467), 19.5% (40/205), and 20.14%
(298/1479) instances resulted in likes or dislikes feedback
(Figures S4A, S4B, and S4C in Multimedia Appendix 1). We
found that the feedback included more likes than dislikes for
slides (mean 8.0 SE 2.30 likes vs mean 0, SE 0 dislikes;
Wilcoxon W=45; n=10 slides; P<.001; Figure S4D in
Multimedia Appendix 1); videos (mean 7.4, SE 1.51 likes vs
mean 0.60, SE 0.54 dislikes; W=15; n=5 videos; P=.06; Figure
S4E in Multimedia Appendix 1); and conversations (mean 1.88,
SE 0.24 likes vs mean 0.12, SE 0.033 dislikes; W=6015; n=149
conversations; P<.001; Figure S4F in Multimedia Appendix
1). The participants also had the option of providing descriptive
feedback on these elements. The subjective feedback was mostly
positive, with participants frequently mentioning that they liked
the given examples. One participant mentioned that they would
have preferred to type their own answers in conversations
(instead of choosing from prewritten text options). A word cloud
created from the collated subjective feedback showed that the
most frequently used words in feedback included given,
example, liked, and idea (Figure S4G in Multimedia Appendix
1). TreadWill allowed participants to revisit previously
completed conversations to refresh their memory; this option
was used 17 times by the participants.
TreadWill allowed participants to attach one or more word tags
from a list of 44 tags to posts in the SupportGroup. A word
cloud of the tags used during the study revealed the topics that
were most commonly discussed by the participants: wasting
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
time, loneliness, guilt, self-esteem, and trust (Figure S5A in
Multimedia Appendix 1). We also analyzed the entries made
by the participants in the CBT forms (worksheets) to identify
the common themes in their activities and concerns (Figures
S5B and S5C in Multimedia Appendix 1). We checked the most
commonly selected situations, thoughts, and beliefs from the
lists presented to the experimental group participants. The most
selected situation, thought, and belief were I am concerned
about my career,I should be doing something better, and If I
don’t work very hard, I’ll fail, respectively. (Figure S6 in
Multimedia Appendix 1 presents the 10 most frequently selected
situations, thoughts, and beliefs.)
All waitlist group participants had the option to use the
experimental intervention once their participation in the waitlist
group was complete. Of the 205 waitlist group participants, 70
(34.1%) signed up to use the experimental group (of which
64/70, 91.4% submitted the follow-up). Of these 70 participants,
7 (10%) completed at least 3 modules and 5 (7.14%) completed
all 6 modules. These values were comparable with the
completion rates in the experimental group participants. In
addition, we calculated the reduction in PHQ-9 and GAD-7
scores from waitlist posttreatment time point to the primary end
point for the participants who completed at least 3 modules.
The reductions in PHQ-9 (mean 3.14, SE 1.14; n=7) and GAD-7
(mean 4.28, SE 0.75; n=7) scores were statistically similar to
those of the experimental group participants.
Discussion
Principal Findings
We have presented the design of an unguided cCBT–based
multicomponent intervention, TreadWill, aimed at high user
engagement and universal accessibility. A fully remote RCT
with 598 participants was performed to test the effectiveness
J Med Internet Res 2023 | vol. 25 | e41005 | p. 11
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
of TreadWill in reducing depression- and anxiety-related
symptoms. The results of the trial show that the full-featured
(experimental) and the plain-text (active control) versions of
TreadWill effectively reduced both PHQ-9 and GAD-7 scores
for the participants who completed at least 3 modules compared
with the waitlist control group. The number of participants who
completed at least 3 modules in the experimental group was
nearly 3 times more than in the active control group. The extra
features included in the experimental version increased
adherence compared with the active control version in terms of
both the time of engagement and the number of modules
completed. The results also showed that the number of modules
completed correlated with the reduction in the symptom severity
of a participant.
Two automated surveys presented during the intervention for
taking participant feedback showed that the participants
perceived TreadWill as useful and easy to use and found most
of the interactive features helpful. In addition, the feedback
provided by the participants using like and dislike buttons on
different elements of the modules indicated that the participants
found the content relatable and useful. Our target population
was tech savvy and educated individuals (high school students,
college students, and working young adults). We expected this
target demographic to be comfortable with English to understand
the material. We kept the language used in TreadWill simple
enough for nonnative speakers to understand. The survey results
confirmed that Easy English was one of the highest-rated
features of TreadWill (Figure 4).
Completer’s Analysis
An intention-to-treat analysis allows one to assess whether
assigning a particular intervention helps the participant. In an
intention-to-treat analysis, all participants assigned to the
interventions are analyzed, regardless of their completion status;
the missing data are imputed or carried forward from earlier
observations. The missing data problem is manageable in most
studies, in which participants are recruited and monitored by
experimenters, and the participants generally have high intrinsic
motivation or perceived psychological pressure (owing to the
involvement of others) or receive compensation for participating
in the study. However, in a web-based, remote intervention, the
intention-to-treat analysis might not be suitable, as previously
noted by Christensen et al [44]. The problem becomes even
more severe when a study, such as ours, is completely unguided;
there is no compensation for the participants, and there are no
psychological barriers to joining or leaving the study at any
time, just by using a smartphone. Although such open designs
pose a problem for the intention-to-treat analysis, they are
desirable in other aspects, as they mimic the real-life use patterns
of smartphone-based self-help interventions.
An alternative analysis approach is to perform a completer’s
analysis, in which the data of only those participants are
analyzed who actually use the intervention. A completer’s
analysis allows one to assess whether completing a particular
intervention helps a participant. This is a more restricted claim
compared with what can be made with an intention-to-treat
analysis, especially from the perspective of a public health
agency that has to decide which interventions to recommend to
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
people. However, in emerging cases of smartphone-based
self-help interventions for which an intention-to-treat analysis
is not ideal, a completer’s analysis can be a reasonable
alternative. This approach has also been used in previous studies
either in isolation or in combination with an intention-to-treat
analysis [30,44-46,64-68].
Another rationale for using an intention-to-treat analysis is that,
in the presence of dropouts, including all participants in the
analysis maintains the equivalence established among the
different groups at the baseline. Although we performed a
completer’s analysis, we found that the baseline equivalence
was also maintained in our data. The baseline PHQ-9 scores for
all the participants who were included in our primary analysis
after removing dropouts remained similar (Kruskal-Wallis
H(2)=1.11; P=.57; Figure 2A).
Limitations
We did not require a clinical diagnosis of depression for
including participants in the study because our goal was to create
an accessible tool catering to both clinical and subclinical
populations. Given that the prevalence of subclinical depression,
defined as a score in the range of 5 to 9 on the PHQ-9, is fairly
high at 15% to 20% [40,69,70], an unguided intervention can
be
immensely beneficial. We used only self-reported
assessments for measuring symptom severity. Although
self-reported assessments have their drawbacks [71], it was
essential given the pragmatic nature of the study with an
unguided intervention. For the same reason, we also included
participants undergoing other treatments (42/598, 7% of our
participants; Table 1). We used only 1 questionnaire each for
assessing severity of depression and anxiety symptom. This
decision was made keeping in mind that filling long
questionnaires on the web is not a pleasant experience for users
and might increase dropout rates [34]. In addition, while
including multiple questionnaires for assessing the same disorder
might improve validity, it also increases the risk of obtaining
false-positive results by chance. Owing to the high dropout rate,
we were unable to perform an intention-to-treat analysis.
Although a completer’s analysis might be justified for a fully
remote RCT, future work can evaluate TreadWill in a more
traditional trial setting to assess intention-to-treat effects. Finally,
our participants were young, mostly male, and tech-savvy
college students, which reduces the generalizability of our results
to the wider and much diverse population of India.
Adherence Rates in cCBT Interventions
Deprexis, a well-evaluated intervention, reported a full
adherence rate of 7.5% in a fully unguided evaluation [50]. The
high adherence rates observed in trial settings often fail to
translate into the real world [35]. In real-world studies,
adherence can be very low: 5.6% in the study by Lara et al [72],
13.11% in the study by Morgan et al [73], and 5% in the study
by March et al [74]. In a fully remote trial of an app-based
intervention, Arean et al [47] reported that 57.9% of the
participants did not even download their assigned apps.
Similarly, in another study involving no human contact, Morriss
et al [34] reported that only 57.3% of participants randomized
to the experimental group signed up for the intervention and
only 42.5% accessed it more than once. Morriss et al [34] further
J Med Internet Res 2023 | vol. 25 | e41005 | p. 12
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
reported an attrition rate of 84.9% at the 3-week follow-up, with
the attrition rate increasing at later follow-up points. In another
recent study, Oehler et al [75] observed that the minimal dose
was received by only 2.10% of the participants for the unguided
version of iFightDepression. Guarino et al [76] also reported in
a recent study that out of 2484 participants who signed up, only
562 started one of the modules, and the module completion
rates ranged from 1% to 13% even by liberal definitions of
module completion. The completion rates of web-based,
self-help, and unguided cCBT interventions are comparable
with the completion rates of massive open web-based courses,
which have been reported to range from 3% to 6% [77,78]. The
adherence rate observed for TreadWill, 12.1% (22/181) for
moderate use and 5.5% (10/181) for full completion, is
comparable with that reported in previous real-world studies.
At this adherence level, TreadWill can benefit a significant
number of people from the general population as a fully
automated and scalable intervention. A web-based self-help
intervention has a low opportunity cost; joining and dropping
out of one intervention usually does not prevent a user from
using another intervention; and trying out multiple apps before
settling on one is a common behavioral pattern observed with
smartphone apps. In addition, it is possible that the existing
cCBT interventions and other digital mental health interventions
do not provide help in the format that users expect on the web.
The 12.1% adherence rate for moderate use was observed in
our study despite additional challenges compared with other
studies. Every step in our study, from participant recruitment
to assessment, was fully automated; the lack of human contact
is known to affect the commitment of participants [34].
Christensen et al [64] evaluated the cCBT intervention
MoodGYM in 2 different settings: in a trial setting in which
participants were called every week by human guides and
provided instructions on completing the intervention, the
completion rate for all 5 modules was 22.5%, but in an open
setting (with no human contact), only 0.49% (97/19,607) of the
participants completed the intervention. The intervention used
in both cases was identical; the only difference was the
interaction between participants and experimenters in the trial
setting. This study shows that although it is possible to obtain
higher completion rates in standard trial settings, these rates do
not translate into real-world settings. Therefore, we used a
pragmatic trial with no human contact, and even though the
adherence rates are low, they are expected to be a more faithful
representation of the real-world completion rates. Furthermore,
it has been reported that male sex and young age significantly
increase the chance of dropout [79]. The average age of our
participant group was 23.6 (SE 0.17) years, and 79.9% (478/598)
were male, which could have contributed to the dropout rate.
Contrary to the practice of giving money or gift cards to
participants [47,80-82], we did not reward participants for
submitting assessments or for participation. In several studies
[47,80-82], participants were paid even after submitting the
baseline assessments. The practice of paying participants is
likely to influence adherence to the intervention owing to the
rule of reciprocity [83] and influence the assessment responses.
Participants getting paid might feel that they owe it to the
researchers to use the program and try to give answers in the
assessments that they think the researchers expect. Not giving
a reward also supports our pragmatic trial design; as in the real
world, paying participants to use the intervention will be
unsustainable.
The generally positive comments that we received from the
participants on the content (Figure S4 in Multimedia appendix
1) and various features of the intervention in the 2 surveys
(Figure 4) suggested that the user dropout was not because of
the unacceptability of the intervention. To check this further,
we compared the survey responses of the experimental group
participants who had completed all 6 modules with those who
dropped out before completing 3 modules but completed a
survey. The average feedback score was not lower in the dropout
group than that in the completer group (Figure S7 in Multimedia
Appendix 1).
Implications
Our study shows that even in low-resource settings, a
cCBT-based intervention without expert support can help users
who at least partially complete the intervention. This implication
is immensely encouraging, as the number of mental health
professionals is extremely low in India [84,85]. The reduction
in PHQ-9 scores in our study was 2.73 for users who completed
at least 3 modules and 4.20 for users who completed all modules.
This level of reduction in a low-threshold intervention, with the
potential to have a population-level impact, can be considered
clinically significant [32]. Our study established TreadWill as
a potential population-level intervention. This is among the
largest studies conducted in India on digital mental health
[86-89]. Our study is also the first fully web-based trial
conducted in India and provides a template for conducting
web-based trials for other mental health conditions in the
country. Future work should focus on strategies, such as using
gamification, serious games, or chatbots to build therapeutic
alliance, to improve adherence to self-help interventions. Future
work can also focus on making the intervention more similar
to general web-based apps, so that users receive help in the
formats with which they are familiar. In this study, we created
tailored content for high school and college students and
working professionals. Future work should also target the
unemployed population and other susceptible groups.
Acknowledgments
The authors thank Romit Chaudhary, Divya Chauhan, Vinay Agarwal, Sahars Kumar, Pearl Sikka, Rahul Gupta, Nikhil Vanjani,
and Sandarsh Pandey for their help in development of TreadWill. The authors thank Pranjul Singh, Aditya Patil, and Pearl Sikka
for their help in developing the videos of TreadWill. The authors thank Prof. Braj Bhushan, Shoukkathali K, Rita Singh, Akanksha
Awasthy, and Dr. Gitanjali Narayanan for helpful discussions, and they thank Dr Shikha Jain, Mrityunjay Bhargava, Pratibha
Mishra, Jagriti Agnihotri, Swastika Tandon, Akash A, Harsh Agarwal, and the Indian Institute of Technology Kanpur media cell
for promoting the visibility of the trial. The authors thank Silky Gupta and Aarush Mohit Mittal for helping with data analysis
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 13
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
and feedback on the manuscript. The authors thank many members of the Indian Institute of Technology Kanpur community for
helping as pilot users of the intervention during its development, members of the counseling and psychiatry team for their feedback,
and members of the Lab of Neural Systems for testing the beta version of TreadWill and giving feedback on the content and user
experience. The authors thank Arun Shankar and Ranjeet Kumar for labeling images in “Identify the friendly face” game. This
work was supported by the Cognitive Science Research Initiative of the Department of Science & Technology (grant
DST/CSRI/2018/102). The funding agency had no role in the design or implementation of the study and in the interpretation of
the results.
Data Availability
The deidentified data analyzed in this study are available from the corresponding author upon reasonable request.
Authors' Contributions
AG and NG conceptualized the project; AG, RJC, SW, AB, and NG designed the research; AG, RJC, SW, PS, and KRK developed
the intervention; AG, AB, and NG recruited participants; AG and NG analyzed data; AG and NG wrote the paper with inputs
from all coauthors.
Conflicts of Interest
None declared.
Multimedia Appendix 1
Supplementary figures and tables.
[DOCX File , 885 KB-Multimedia Appendix 1]
Multimedia Appendix 2
CONSORT-EHEALTH (V 1.6.1) Checklist.
[PDF File (Adobe PDF File), 728 KB-Multimedia Appendix 2]
References
1.
GBD 2017 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence,
and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990-2017: a systematic
analysis for the Global Burden of Disease Study 2017. Lancet 2018 Nov 10;392(10159):1789-1858 [FREE Full text] [doi:
10.1016/S0140-6736(18)32279-7] [Medline: 30496104]
3.
2. Wang PS, Aguilar-Gaxiola S, Alonso J, Angermeyer MC, Borges G, Bromet EJ, et al. Use of mental health services for
anxiety, mood, and substance disorders in 17 countries in the WHO world mental health surveys. Lancet 2007 Sep
08;370(9590):841-850 [FREE Full text] [doi: 10.1016/S0140-6736(07)61414-7] [Medline: 17826169]
Andrade LH, Alonso J, Mneimneh Z, Wells JE, Al-Hamzawi A, Borges G, et al. Barriers to mental health treatment: results
from the WHO world mental health surveys. Psychol Med 2014 Apr;44(6):1303-1317 [FREE Full text] [doi:
10.1017/S0033291713001943] [Medline: 23931656]
Patel V, Saxena S, Lund C, Thornicroft G, Baingana F, Bolton P, et al. The Lancet Commission on global mental health
and sustainable development. Lancet 2018 Oct 27;392(10157):1553-1598. [doi: 10.1016/S0140-6736(18)31612-X] [Medline:
30314863]
Saxena S, Thornicroft G, Knapp M, Whiteford H. Resources for mental health: scarcity, inequity, and inefficiency. Lancet
2007 Sep 08;370(9590):878-889. [doi: 10.1016/S0140-6736(07)61239-2] [Medline: 17804062]
4.
5.
6. Murthy RS. National mental health survey of India 2015-2016. Indian J Psychiatry 2017;59(1):21-26 [FREE Full text] [doi:
10.4103/psychiatry.IndianJPsychiatry_102_17] [Medline: 28529357]
7. Weizenbaum J. ELIZA—a computer program for the study of natural language communication between man and machine.
8.
9.
Commun ACM 1966 Jan;9(1):36-45 [FREE Full text] [doi: 10.1145/357980.357991]
Selmi PM, Klein MH, Greist JH, Johnson JH, Harris WG. An investigation of computer-assisted cognitive-behavior therapy
in the treatment of depression. Behav Res Methods Instrum 1982 Mar;14(2):181-185 [FREE Full text] [doi:
10.3758/bf03202150]
Andersson G, Carlbring P. Cognitive behavioral therapy delivered using the internet. In: Wenzel A, editor. Handbook of
Cognitive Behavioral Therapy: Applications. Volume 2. Washington, DC, USA: American Psychological Association;
2021:607-631.
10. Cuijpers P, Noma H, Karyotaki E, Cipriani A, Furukawa TA. Effectiveness and acceptability of cognitive behavior therapy
delivery formats in adults with depression: a network meta-analysis. JAMA Psychiatry 2019 Jul 01;76(7):700-707 [FREE
Full text] [doi: 10.1001/jamapsychiatry.2019.0268] [Medline: 30994877]
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 14
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
11. Gilbody S, Brabyn S, Lovell K, Kessler D, Devlin T, Smith L, REEACT collaborative. Telephone-supported computerised
cognitive-behavioural therapy: REEACT-2 large-scale pragmatic randomised controlled trial. Br J Psychiatry 2017
May;210(5):362-367 [FREE Full text] [doi: 10.1192/bjp.bp.116.192435] [Medline: 28254959]
Johansson R, Andersson G. Internet-based psychological treatments for depression. Expert Rev Neurother 2012
Jul;12(7):861-870. [doi: 10.1586/ern.12.63] [Medline: 22853793]
12.
13. Karyotaki E, Efthimiou O, Miguel C, Bermpohl FM, Furukawa TA, Cuijpers P, Individual Patient Data Meta-Analyses for
Depression (IPDMA-DE) Collaboration, et al. Internet-based cognitive behavioral therapy for depression: a systematic
review and individual patient data network meta-analysis. JAMA Psychiatry 2021 Apr 01;78(4):361-371 [FREE Full text]
[doi: 10.1001/jamapsychiatry.2020.4364] [Medline: 33471111]
14. Wright JH, Owen JJ, Richards D, Eells TD, Richardson T, Brown GK, et al. Computer-assisted cognitive-behavior therapy
for depression: a systematic review and meta-analysis. J Clin Psychiatry 2019 Mar 19;80(2):18r12188 [FREE Full text]
[doi: 10.4088/JCP.18r12188] [Medline: 30900849]
15. Carlbring P, Andersson G, Cuijpers P, Riper H, Hedman-Lagerlöf E. Internet-based vs. face-to-face cognitive behavior
therapy for psychiatric and somatic disorders: an updated systematic review and meta-analysis. Cogn Behav Ther 2018
Jan;47(1):1-18. [doi: 10.1080/16506073.2017.1401115] [Medline: 29215315]
16. Gilbody S, Littlewood E, Hewitt C, Brierley G, Tharmanathan P, Araya R, REEACT Team. Computerised cognitive
behaviour therapy (cCBT) as treatment for depression in primary care (REEACT trial): large scale pragmatic randomised
controlled trial. BMJ 2015 Nov 11;351:h5627 [FREE Full text] [doi: 10.1136/bmj.h5627] [Medline: 26559241]
17. Andrews G, Basu A, Cuijpers P, Craske MG, McEvoy P, English CL, et al. Computer therapy for the anxiety and depression
disorders is effective, acceptable and practical health care: an updated meta-analysis. J Anxiety Disord 2018 Apr;55:70-78
[FREE Full text] [doi: 10.1016/j.janxdis.2018.01.001] [Medline: 29422409]
18. Bruckner TA, Scheffler RM, Shen G, Yoon J, Chisholm D, Morris J, et al. The mental health workforce gap in low- and
middle-income countries: a needs-based approach. Bull World Health Organ 2011 Mar 01;89(3):184-194 [FREE Full text]
[doi: 10.2471/BLT.10.082784] [Medline: 21379414]
19. Karyotaki E, Riper H, Twisk J, Hoogendoorn A, Kleiboer A, Mira A, et al. Efficacy of self-guided internet-based cognitive
behavioral therapy in the treatment of depressive symptoms: a meta-analysis of individual participant data. JAMA Psychiatry
2017 Apr 01;74(4):351-359 [FREE Full text] [doi: 10.1001/jamapsychiatry.2017.0044] [Medline: 28241179]
20. Twomey C, O'Reilly G, Bültmann O, Meyer B. Effectiveness of a tailored, integrative internet intervention (Deprexis) for
depression: updated meta-analysis. PLoS One 2020 Jan 30;15(1):e0228100 [FREE Full text] [doi:
10.1371/journal.pone.0228100] [Medline: 31999743]
21. Garrido S, Millington C, Cheers D, Boydell K, Schubert E, Meade T, et al. What works and what doesn't work? A systematic
review of digital mental health interventions for depression and anxiety in young people. Front Psychiatry 2019 Nov
13;10:759 [FREE Full text] [doi: 10.3389/fpsyt.2019.00759] [Medline: 31798468]
22. Harrer M, Adam SH, Baumeister H, Cuijpers P, Karyotaki E, Auerbach RP, et al. Internet interventions for mental health
in university students: a systematic review and meta-analysis. Int J Methods Psychiatr Res 2019 Jun;28(2):e1759 [FREE
Full text] [doi: 10.1002/mpr.1759] [Medline: 30585363]
23. Välimäki M, Anttila K, Anttila M, Lahti M. Web-based interventions supporting adolescents and young people with
depressive symptoms: systematic review and meta-analysis. JMIR Mhealth Uhealth 2017 Dec 08;5(12):e180 [FREE Full
text] [doi: 10.2196/mhealth.8624] [Medline: 29222079]
25.
24. Berger T, Hämmerli K, Gubser N, Andersson G, Caspar F. Internet-based treatment of depression: a randomized controlled
trial comparing guided with unguided self-help. Cogn Behav Ther 2011;40(4):251-266. [doi: 10.1080/16506073.2011.616531]
[Medline: 22060248]
de Graaf LE, Gerhards SA, Arntz A, Riper H, Metsemakers JF, Evers SM, et al. Clinical effectiveness of online computerised
cognitive-behavioural therapy without support for depression in primary care: randomised trial. Br J Psychiatry 2009
Jul;195(1):73-80 [FREE Full text] [doi: 10.1192/bjp.bp.108.054429] [Medline: 19567900]
Farrer L, Christensen H, Griffiths KM, Mackinnon A. Internet-based CBT for depression with and without telephone
tracking in a national helpline: randomised controlled trial. PLoS One 2011;6(11):e28099 [FREE Full text] [doi:
10.1371/journal.pone.0028099] [Medline: 22140514]
26.
27. Meyer B, Bierbrodt J, Schröder J, Berger T, Beevers CG, Weiss M, et al. Effects of an internet intervention (Deprexis) on
severe depression symptoms: randomized controlled trial. Internet Interv 2015 Mar;2(1):48-59 [FREE Full text] [doi:
10.1016/j.invent.2014.12.003]
28. Mira A, Bretón-López J, García-Palacios A, Quero S, Baños RM, Botella C. An internet-based program for depressive
symptoms using human and automated support: a randomized controlled trial. Neuropsychiatr Dis Treat 2017 Mar
31;13:987-1006 [FREE Full text] [doi: 10.2147/NDT.S130994] [Medline: 28408833]
Spek V, Nyklícek I, Smits N, Cuijpers P, Riper H, Keyzer J, et al. Internet-based cognitive behavioural therapy for
subthreshold depression in people over 50 years old: a randomized controlled clinical trial. Psychol Med 2007
Dec;37(12):1797-1806. [doi: 10.1017/S0033291707000542] [Medline: 17466110]
29.
30. Christensen H, Griffiths KM, Jorm AF. Delivering interventions for depression by using the internet: randomised controlled
trial. BMJ 2004 Jan 31;328(7434):265 [FREE Full text] [doi: 10.1136/bmj.37945.566632.EE] [Medline: 14742346]
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 15
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
31.
Phillips R, Schneider J, Molosankwe I, Leese M, Foroushani PS, Grime P, et al. Randomized controlled trial of computerized
cognitive behavioural therapy for depressive symptoms: effectiveness and costs of a workplace intervention. Psychol Med
2014 Mar;44(4):741-752 [FREE Full text] [doi: 10.1017/S0033291713001323] [Medline: 23795621]
32. Klein JP, Berger T, Schröder J, Späth C, Meyer B, Caspar F, et al. Effects of a psychological internet intervention in the
treatment of mild to moderate depressive symptoms: results of the EVIDENT study, a randomized controlled trial. Psychother
Psychosom 2016;85(4):218-228 [FREE Full text] [doi: 10.1159/000445355] [Medline: 27230863]
33. Moritz S, Schilling L, Hauschildt M, Schröder J, Treszl A. A randomized controlled trial of internet-based therapy in
depression. Behav Res Ther 2012 Aug;50(7-8):513-521. [doi: 10.1016/j.brat.2012.04.006] [Medline: 22677231]
34. Morriss R, Kaylor-Hughes C, Rawsthorne M, Coulson N, Simpson S, Guo B, et al. A Direct-to-public peer support program
(big white wall) versus web-based information to aid the self-management of depression and anxiety: results and challenges
of an automated randomized controlled trial. J Med Internet Res 2021 Apr 23;23(4):e23487 [FREE Full text] [doi:
10.2196/23487] [Medline: 33890858]
Fleming T, Bavin L, Lucassen M, Stasiak K, Hopkins S, Merry S. Beyond the trial: systematic review of real-world uptake
and engagement with digital self-help interventions for depression, low mood, or anxiety. J Med Internet Res 2018 Jun
06;20(6):e199 [FREE Full text] [doi: 10.2196/jmir.9275] [Medline: 29875089]
35.
36. Lehtimaki S, Martic J, Wahl B, Foster KT, Schwalbe N. Evidence on digital mental health interventions for adolescents
and young people: systematic overview. JMIR Ment Health 2021 Apr 29;8(4):e25847 [FREE Full text] [doi: 10.2196/25847]
[Medline: 33913817]
37. Martínez P, Rojas G, Martínez V, Lara MA, Pérez JC. Internet-based interventions for the prevention and treatment of
depression in people living in developing countries: a systematic review. J Affect Disord 2018 Jul;234:193-200. [doi:
10.1016/j.jad.2018.02.079] [Medline: 29529553]
Fu Z, Burger H, Arjadi R, Bockting CL. Effectiveness of digital psychological interventions for mental health problems in
low-income and middle-income countries: a systematic review and meta-analysis. Lancet Psychiatry 2020 Oct;7(10):851-864
[FREE Full text] [doi: 10.1016/S2215-0366(20)30256-X] [Medline: 32866459]
38.
39. Köhnen M, Dreier M, Seeralan T, Kriston L, Härter M, Baumeister H, et al. Evidence on technology-based psychological
interventions in diagnosed depression: systematic review. JMIR Ment Health 2021 Feb 10;8(2):e21700 [FREE Full text]
[doi: 10.2196/21700] [Medline: 33565981]
40. Kroenke K, Spitzer RL, Williams JB. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med 2001
41.
Sep;16(9):606-613 [FREE Full text] [doi: 10.1046/j.1525-1497.2001.016009606.x] [Medline: 11556941]
Spitzer RL, Kroenke K, Williams JB, Löwe B. A brief measure for assessing generalized anxiety disorder: the GAD-7.
Arch Intern Med 2006 May 22;166(10):1092-1097. [doi: 10.1001/archinte.166.10.1092] [Medline: 16717171]
42. Gupta SC, Anand R, Trivedi JK. Development of a suicidal intent questionnaire. Indian J Psychiatry 1983 Jan;25(1):57-62
[FREE Full text] [Medline: 21847253]
Psychological Image Collection at Stirling (PICS). URL: http://pics.stir.ac.uk/ [accessed 2023-04-10]
43.
44. Christensen H, Griffiths KM, Korten A. Web-based cognitive behavior therapy: analysis of site usage and changes in
depression and anxiety scores. J Med Internet Res 2002 Jan;4(1):e3 [FREE Full text] [doi: 10.2196/jmir.4.1.e3] [Medline:
11956035]
45. Keefe RS, Cañadas E, Farlow D, Etkin A. Digital intervention for cognitive deficits in major depression: a randomized
controlled trial to assess efficacy and safety in adults. Am J Psychiatry 2022 Jul;179(7):482-489. [doi:
10.1176/appi.ajp.21020125] [Medline: 35410496]
46. Rollman BL, Herbeck Belnap B, Abebe KZ, Spring MB, Rotondi AJ, Rothenberger SD, et al. Effectiveness of online
collaborative care for treating mood and anxiety disorders in primary care: a randomized clinical trial. JAMA Psychiatry
2018 Jan 01;75(1):56-64 [FREE Full text] [doi: 10.1001/jamapsychiatry.2017.3379] [Medline: 29117275]
47. Arean PA, Hallgren KA, Jordan JT, Gazzaley A, Atkins DC, Heagerty PJ, et al. The use and effectiveness of mobile apps
for depression: results from a fully remote clinical trial. J Med Internet Res 2016 Dec 20;18(12):e330 [FREE Full text]
[doi: 10.2196/jmir.6482] [Medline: 27998876]
48. Beck JS. Cognitive Behavior Therapy: Basics and Beyond. 2nd edition. New York, NY, USA: Guilford Press; Aug 17,
2011.
49. Andersson G, Carlbring P, Berger T, Almlöv J, Cuijpers P. What makes internet therapy work? Cogn Behav Ther 2009;38
Suppl 1:55-60. [doi: 10.1080/16506070902916400] [Medline: 19675956]
50. Meyer B, Berger T, Caspar F, Beevers CG, Andersson G, Weiss M. Effectiveness of a novel integrative online treatment
for depression (Deprexis): randomized controlled trial. J Med Internet Res 2009 May 11;11(2):e15 [FREE Full text] [doi:
10.2196/jmir.1151] [Medline: 19632969]
51. Nordin S, Carlbring P, Cuijpers P, Andersson G. Expanding the limits of bibliotherapy for panic disorder: randomized trial
of self-help without support but with a clear deadline. Behav Ther 2010 Sep;41(3):267-276. [doi: 10.1016/j.beth.2009.06.001]
[Medline: 20569776]
52. Dandeneau SD, Baldwin MW. The buffering effects of rejection-inhibiting attentional training on social and performance
threat among adult students. Contemp Educ Psychol 2009 Jan;34(1):42-50 [FREE Full text] [doi:
10.1016/j.cedpsych.2008.05.004]
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 16
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
53. Orth U, Robins RW, Meier LL, Conger RD. Refining the vulnerability model of low self-esteem and depression: disentangling
the effects of genuine self-esteem and narcissism. J Pers Soc Psychol 2016 Jan;110(1):133-149 [FREE Full text] [doi:
10.1037/pspp0000038] [Medline: 25915133]
Sowislo JF, Orth U. Does low self-esteem predict depression and anxiety? A meta-analysis of longitudinal studies. Psychol
Bull 2013 Jan;139(1):213-240. [doi: 10.1037/a0028931] [Medline: 22730921]
54.
55. Houston TK, Cooper LA, Ford DE. Internet support groups for depression: a 1-year prospective cohort study. Am J Psychiatry
2002 Dec;159(12):2062-2068. [doi: 10.1176/appi.ajp.159.12.2062] [Medline: 12450957]
56. Griffiths KM, Mackinnon AJ, Crisp DA, Christensen H, Bennett K, Farrer L. The effectiveness of an online support group
for members of the community with depression: a randomised controlled trial. PLoS One 2012;7(12):e53244 [FREE Full
text] [doi: 10.1371/journal.pone.0053244] [Medline: 23285271]
57. Tomasino KN, Lattie EG, Ho J, Palac HL, Kaiser SM, Mohr DC. Harnessing peer support in an online intervention for
older adults with depression. Am J Geriatr Psychiatry 2017 Oct;25(10):1109-1119 [FREE Full text] [doi:
10.1016/j.jagp.2017.04.015] [Medline: 28571785]
58. Carlbring P, Maurin L, Törngren C, Linna E, Eriksson T, Sparthan E, et al. Individually-tailored, internet-based treatment
for anxiety disorders: a randomized controlled trial. Behav Res Ther 2011 Jan;49(1):18-24. [doi: 10.1016/j.brat.2010.10.002]
[Medline: 21047620]
Johansson R, Sjöberg E, Sjögren M, Johnsson E, Carlbring P, Andersson T, et al. Tailored vs. standardized internet-based
cognitive behavior therapy for depression and comorbid symptoms: a randomized controlled trial. PLoS One 2012;7(5):e36905
[FREE Full text] [doi: 10.1371/journal.pone.0036905] [Medline: 22615841]
59.
60. Weissman AN, Beck AT. Development and validation of the dysfunctional attitude scale: a preliminary investigation. In:
Proceedings of the 62nd Annual Meeting of the American Educational Research Association. 1978 Mar Presented at: AERA
'78; March 27-31, 1978; Toronto, Canada p. 1-33 URL: https://files.eric.ed.gov/fulltext/ED167619.pdf
61. Hollon SD, Kendall PC. Cognitive self-statements in depression: development of an automatic thoughts questionnaire.
Cognit Ther Res 1980 Dec;4(4):383-395 [FREE Full text] [doi: 10.1007/bf01178214]
62. All India survey on higher education 2018-19. Department of Higher Education, India. 2021. URL: https://aishe.gov.in/
aishe/gotoAisheReports [accessed 2022-02-17]
63. The mobile gender gap report 2019. GSMA. 2021. URL: https://www.gsma.com/mobilefordevelopment/resources/
mobile-gender-gap-report-2019/ [accessed 2022-02-17]
64. Christensen H, Griffiths KM, Korten AE, Brittliffe K, Groves C. A comparison of changes in anxiety and depression
symptoms of spontaneous users and trial participants of a cognitive behavior therapy website. J Med Internet Res 2004 Dec
22;6(4):e46 [FREE Full text] [doi: 10.2196/jmir.6.4.e46] [Medline: 15631970]
65. Christensen H, Griffiths KM, Mackinnon AJ, Brittliffe K. Online randomized controlled trial of brief and full cognitive
behaviour therapy for depression. Psychol Med 2006 Dec;36(12):1737-1746. [doi: 10.1017/S0033291706008695] [Medline:
16938144]
66. Harrer M, Apolinário-Hagen J, Fritsche L, Salewski C, Zarski AC, Lehr D, et al. Effect of an internet- and app-based stress
intervention compared to online psychoeducation in university students with depressive symptoms: results of a randomized
controlled trial. Internet Interv 2021 Apr;24:100374 [FREE Full text] [doi: 10.1016/j.invent.2021.100374] [Medline:
33718001]
67. Twomey C, O'Reilly G, Byrne M, Bury M, White A, Kissane S, et al. A randomized controlled trial of the computerized
CBT programme, MoodGYM, for public mental health service users waiting for interventions. Br J Clin Psychol 2014
Nov;53(4):433-450. [doi: 10.1111/bjc.12055] [Medline: 24831119]
68. Wright B, Tindall L, Littlewood E, Allgar V, Abeles P, Trépel D, et al. Computerised cognitive-behavioural therapy for
depression in adolescents: feasibility results and 4-month outcomes of a UK randomised controlled trial. BMJ Open 2017
Jan 27;7(1):e012834 [FREE Full text] [doi: 10.1136/bmjopen-2016-012834] [Medline: 28132000]
69. Khaled SM. Prevalence and potential determinants of subthreshold and major depression in the general population of Qatar.
J Affect Disord 2019 Jun 01;252:382-393. [doi: 10.1016/j.jad.2019.04.056] [Medline: 31003107]
70. Kroenke K, Strine TW, Spitzer RL, Williams JB, Berry JT, Mokdad AH. The PHQ-8 as a measure of current depression
71.
in the general population. J Affect Disord 2009 Apr;114(1-3):163-173. [doi: 10.1016/j.jad.2008.06.026] [Medline: 18752852]
Sato H, Kawahara J. Selective bias in retrospective self-reports of negative mood states. Anxiety Stress Coping 2011
Jul;24(4):359-367. [doi: 10.1080/10615806.2010.543132] [Medline: 21253957]
72. Lara MA, Tiburcio M, Aguilar Abrego A, Sánchez-Solís A. A four-year experience with a web-based self-help intervention
for depressive symptoms in Mexico. Rev Panam Salud Publica 2014 May;35(5-6):399-406. [Medline: 25211568]
73. Morgan C, Mason E, Newby JM, Mahoney AE, Hobbs MJ, McAloon J, et al. The effectiveness of unguided internet
cognitive behavioural therapy for mixed anxiety and depression. Internet Interv 2017 Oct 24;10:47-53 [FREE Full text]
[doi: 10.1016/j.invent.2017.10.003] [Medline: 30135752]
74. March S, Batterham PJ, Rowe A, Donovan C, Calear AL, Spence SH. Trajectories of change in an open-access internet-based
cognitive behavior program for childhood and adolescent anxiety: open trial. JMIR Ment Health 2021 Jun 18;8(6):e27981
[FREE Full text] [doi: 10.2196/27981] [Medline: 34142971]
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 17
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
75. Oehler C, Scholze K, Reich H, Sander C, Hegerl U. Intervention use and symptom change with unguided internet-based
cognitive behavioral therapy for depression during the COVID-19 pandemic: log data analysis of a convenience sample.
JMIR Ment Health 2021 Jul 16;8(7):e28321 [FREE Full text] [doi: 10.2196/28321] [Medline: 34115604]
76. Detweiler Guarino I, Cowan DR, Fellows AM, Buckey JC. Use of a self-guided computerized cognitive behavioral tool
during COVID-19: evaluation study. JMIR Form Res 2021 May 31;5(5):e26989 [FREE Full text] [doi: 10.2196/26989]
[Medline: 33973856]
77. Breslow L, Pritchard DE, DeBoer J, Stump GS, Ho AD, Seaton DT. Studying learning in the worldwide classroom research
into edX's first MOOC. Res Pract Assess 2013;8:13-25 [FREE Full text]
78. Reich J, Ruipérez-Valiente JA. The MOOC pivot. Sci Educ 2019 Jan 11;363(6423):130-131. [doi: 10.1126/science.aav7958]
[Medline: 30630920]
79. Karyotaki E, Kleiboer A, Smit F, Turner DT, Pastor AM, Andersson G, et al. Predictors of treatment dropout in self-guided
web-based interventions for depression: an 'individual patient data' meta-analysis. Psychol Med 2015 Oct;45(13):2717-2726
[FREE Full text] [doi: 10.1017/S0033291715000665] [Medline: 25881626]
81.
80. Clarke G, Kelleher C, Hornbrook M, Debar L, Dickerson J, Gullion C. Randomized effectiveness trial of an internet, pure
self-help, cognitive behavioral intervention for depressive symptoms in young adults. Cogn Behav Ther 2009;38(4):222-234
[FREE Full text] [doi: 10.1080/16506070802675353] [Medline: 19440896]
Pratap A, Renn BN, Volponi J, Mooney SD, Gazzaley A, Arean PA, et al. Using mobile apps to assess and treat depression
in Hispanic and Latino populations: fully remote randomized clinical trial. J Med Internet Res 2018 Aug 09;20(8):e10130
[FREE Full text] [doi: 10.2196/10130] [Medline: 30093372]
Schure MB, Lindow JC, Greist JH, Nakonezny PA, Bailey SJ, Bryan WL, et al. Use of a fully automated internet-based
cognitive behavior therapy intervention in a community population of adults with depression symptoms: randomized
controlled trial. J Med Internet Res 2019 Nov 18;21(11):e14754 [FREE Full text] [doi: 10.2196/14754] [Medline: 31738173]
82.
83. Cialdini RB. Influence: The Psychology of Persuasion. New York, NY, USA: William Morrow; Dec 26, 2006.
84. Garg K, Kumar CN, Chandra PS. Number of psychiatrists in India: baby steps forward, but a long way to go. Indian J
85.
Psychiatry 2019 Jan;61(1):104-105 [FREE Full text] [doi: 10.4103/psychiatry.IndianJPsychiatry_7_18] [Medline: 30745666]
Patel V, Xiao S, Chen H, Hanna F, Jotheeswaran AT, Luo D, et al. The magnitude of and health system responses to the
mental health treatment gap in adults in India and China. Lancet 2016 Dec 17;388(10063):3074-3084. [doi:
10.1016/S0140-6736(16)00160-4] [Medline: 27209149]
86. Kanuri N, Arora P, Talluru S, Colaco B, Dutta R, Rawat A, et al. Examining the initial usability, acceptability and feasibility
of a digital mental health intervention for college students in India. Int J Psychol 2020 Aug;55(4):657-673 [FREE Full text]
[doi: 10.1002/ijop.12640] [Medline: 31867730]
87. Mehrotra S, Sudhir P, Rao G, Thirthalli J, Srikanth TK. Development and pilot testing of an internet-based self-help
intervention for depression for Indian users. Behav Sci (Basel) 2018 Mar 22;8(4):36 [FREE Full text] [doi:
10.3390/bs8040036] [Medline: 29565278]
88. Rodriguez-Villa E, Rozatkar AR, Kumar M, Patel V, Bondre A, Naik SS, et al. Cross cultural and global uses of a digital
mental health app: results of focus groups with clinicians, patients and family members in India and the United States. Glob
Ment Health (Camb) 2021 Aug 24;8:e30 [FREE Full text] [doi: 10.1017/gmh.2021.28] [Medline: 34512999]
Srivastava P, Mehta M, Sagar R, Ambekar A. Smartteen- a computer assisted cognitive behavior therapy for Indian
adolescents with depression- a pilot study. Asian J Psychiatr 2020 Apr;50:101970. [doi: 10.1016/j.ajp.2020.101970]
[Medline: 32114331]
89.
Abbreviations
CBT: cognitive behavioral therapy
cCBT: computerized cognitive behavioral therapy
GAD-7: Generalized Anxiety Disorder-7
LMICs: low- and middle-income countries
PHQ-9: Patient Health Questionnaire-9
RCT: randomized controlled trial
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 18
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Ghosh et al
Edited by A Mavragani; submitted 12.07.22; peer-reviewed by D Ramanathan, Q Liao, JC Buckey; comments to author 01.11.22;
revised version received 23.02.23; accepted 08.03.23; published 26.04.23
Please cite as:
Ghosh A, Cherian RJ, Wagle S, Sharma P, Kannan KR, Bajpai A, Gupta N
An Unguided, Computerized Cognitive Behavioral Therapy Intervention (TreadWill) in a Lower Middle-Income Country: Pragmatic
Randomized Controlled Trial
J Med Internet Res 2023;25:e41005
URL: https://www.jmir.org/2023/1/e41005
doi: 10.2196/41005
PMID: 37099376
©Arka Ghosh, Rithwik J Cherian, Surbhit Wagle, Parth Sharma, Karthikeyan R Kannan, Alok Bajpai, Nitin Gupta. Originally
published in the Journal of Medical Internet Research (https://www.jmir.org), 26.04.2023. This is an open-access article distributed
under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits
unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of
Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on
https://www.jmir.org/, as well as this copyright and license information must be included.
https://www.jmir.org/2023/1/e41005
XSL•FO
RenderX
J Med Internet Res 2023 | vol. 25 | e41005 | p. 19
(page number not for citation purposes)
| null |
10.3389/fnbeh.2020.584731
|
DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be
made available by the authors, without undue reservation.
|
DATA AVAILABILITY STATEMENT The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
|
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 1
ORIGINAL RESEARCH
published: 16 November 2020
doi: 10.3389/fnbeh.2020.584731
Developmental Fluoxetine Exposure
Alters Behavior and Neuropeptide
Receptors in the Prairie Vole
Rebecca H. Lawrence1,2, Michelle C. Palumbo2,3, Sara M. Freeman1,2,4,
Caleigh D. Guoynes1,5 and Karen L. Bales1,2,6*
1 Department of Psychology, University of California, Davis, Davis, CA, United States, 2 California National Primate Research
Center, University of California, Davis, Davis, CA, United States, 3 Department of Behavioral Neuroscience, Oregon Health &
Science University, Portland, OR, United States, 4 Department of Biology, Utah State University, Logan, UT, United States,
5 Department of Psychology, University of Wisconsin, Madison, WI, United States, 6 Department of Neurobiology, Physiology
and Behavior, University of California, Davis, Davis, CA, United States
Developmental exposure to selective serotonin reuptake inhibitor (SSRI) increases the
risk of Autism Spectrum Disorder (ASD), however, the underlying neurobiology of this
effect is not fully understood. Here we used the socially monogamous prairie vole as a
translational model of developmental SSRI exposure. Paired female prairie voles (n = 20)
were treated with 5 mg/kg subcutaneous fluoxetine (FLX) or saline (SAL) daily from birth
of the second litter until the day of birth of the 4th litter. This design created three cohorts
of FLX exposure: postnatal exposure in litter 2, both prenatal and postnatal exposure in
litter 3, and prenatal exposure in litter 4. Post-weaning, subjects underwent behavioral
testing to detect changes in sociality, repetitive behavior, pair-bond formation, and
anxiety-like behavior. Quantitative receptor autoradiography was performed for oxytocin,
vasopressin 1a, and serotonin 1a receptor density in a subset of brains. We observed
increased anxiety-like behavior and reduced sociality in developmentally FLX exposed
adults. FLX exposure decreased oxytocin receptor binding in the nucleus accumbens
core and central amygdala, and vasopressin 1a receptor binding in the medial amygdala.
FLX exposure did not affect serotonin 1A receptor binding in any areas examined.
Changes to oxytocin and vasopressin receptors may underlie the behavioral changes
observed and have translational implications for the mechanism of the increased risk of
ASD subsequent to prenatal SSRI exposure.
Keywords: oxytocin receptor, vasopressin receptor, serotonin receptor, 5-HT, autism, antidepressant, SSRI,
autoradiography
INTRODUCTION
In humans, antidepressant medication, most frequently a selective serotonin reuptake inhibitor
(SSRI), is commonly prescribed to pregnant and lactating women with major depression (Boukhris
et al., 2016). Use of SSRIs during pregnancy has increased dramatically over the last several
decades, with estimates ranging from 6 to 13% of pregnancies in the United States (Cooper et al.,
2007; Andrade et al., 2008; Alwan et al., 2011). Pharmacological treatment of maternal depression
is typically recommended during the prenatal period, primarily because of the well-established
negative effects of maternal depression (Davalos et al., 2012; Jarde et al., 2016). However, there
Edited by:
Tamas Kozicz,
Mayo Clinic, United States
Reviewed by:
William J. Giardino,
Stanford University, United States
Joseph Lonstein,
Michigan State University,
United States
Caroline Hostetler,
Oregon Health & Science University,
United States
*Correspondence:
Karen L. Bales
[email protected]
Specialty section:
This article was submitted to
Behavioral Endocrinology,
a section of the journal
Frontiers in Behavioral Neuroscience
Received: 18 July 2020
Accepted: 23 October 2020
Published: 16 November 2020
Citation:
Lawrence RH, Palumbo MC,
Freeman SM, Guoynes CD and
Bales KL (2020) Developmental
Fluoxetine Exposure Alters Behavior
and Neuropeptide Receptors
in the Prairie Vole.
Front. Behav. Neurosci. 14:584731.
doi: 10.3389/fnbeh.2020.584731
Frontiers in Behavioral Neuroscience | www.frontiersin.org
1
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 2
Lawrence et al.
Developmental SSRIs in Voles
may be side effects of SSRIs leading to preterm labor, altered
gestational
length and early delivery (Hayes et al., 2012),
congenital heart malformations (Knudsen et al., 2014; Gentile,
2015a), persistent pulmonary hypertension (Grigoriadis et al.,
2014), and adverse neurodevelopmental outcomes (El Marroun
et al., 2014; Glover and Clinton, 2016). There is reason for
concern about
the effects of early exposure to SSRIs on
the developing brain. SSRIs can cross the placental barrier
(Hendrick et al., 2003; Rampono et al., 2009) and enter into
breast milk (Kristensen et al., 1999; Rampono et al., 2000).
Exposed infants show altered brain activity measured via EEG
(Videman et al., 2017).
A growing body of research indicates increased rates of
Autism Spectrum Disorder (ASD) in prenatally SSRI-exposed
children (Croen et al., 2011; El Marroun et al., 2014; Gidaya
et al., 2014; Gentile, 2015b; Boukhris et al., 2016; Andalib
et al., 2017). While others have found no relationship when
controlling for maternal factors (Hviid et al., 2013; Kobayashi
et al., 2016) recent meta-analyses indicate that SSRI-exposure
does increase autism diagnosis when pooling across studies (Man
et al., 2015; Kaplan et al., 2017). Disentangling the effects of the
underlying psychiatric condition of the mother from the effects
of SSRIs on fetal development is difficult, and causality remains
to be established.
Decades of research have indicated a link between ASD and
serotonin, starting with the finding of hyperserotonemia in a
subset of individuals shortly after the disorder was first described
(Schain and Freedman, 1961). Hyperserotonemia has remained
a consistent finding in a large subgroup of individuals diagnosed
with ASD, with roughly one third of individuals presenting with
high whole blood serotonin levels (Schain and Freedman, 1961;
Anderson et al., 1987; Hranilovic et al., 2007; Gabriele et al., 2014;
Muller et al., 2016). This finding has led researchers to suggest
that hyperserotonemia underlies differences in the brain which
are responsible for the appearance of autistic behavior (Whitaker-
Azmitia, 2005; Yang et al., 2014). Animal models corroborate
that hyperserotonemia leads to behavioral and neuroendocrine
changes consistent with those seen in autism (Whitaker-Azmitia,
2005; McNamara et al., 2008; Veenstra-VanderWeele et al., 2012;
Madden and Zup, 2014; Tanaka et al., 2018). Developmental
hyperserotonemia decreases the number of oxytocinergic cells
in the paraventricular nucleus of the hypothalamus in both rats
(McNamara et al., 2008) and prairie voles (Martin et al., 2012),
while decreasing affiliative behavior and increasing anxiety.
The effects of hyperserotonemia on the brain are rooted in
serotonin’s critical role during early development as a trophic
factor, long before it begins to function as a neurotransmitter. As
a growth factor, it regulates development of its own and related
systems and guides cell division, differentiation, migration,
myelination, synaptogenesis, and dendritic pruning (Lauder,
1993; Azmitia, 2001; Wirth et al., 2017). Because serotonin
exposure at this time also functions to autoregulate its own
innervation throughout
the brain via a negative feedback
can cause
mechanism,
organizational change which may enduringly alter serotonergic
neurotransmission (Whitaker-Azmitia, 2001). Despite
the
relative paucity of serotonin neurons, they innervate almost all
hyperserotonemia
developmental
parts of the brain, making this system a powerful mediator of
brain activity in many regions. Thus, alterations in serotonin
during development may be particularly influential.
Significant overlap exists in psychiatric conditions associated
with serotonin dysfunction and ASD. For instance, heightened
rates of anxiety and depression may be seen in ASD populations
(Lugnegård et al., 2011) and serotonin-based treatments,
including SSRIs, show efficacy in treating some symptoms of
ASD (Kolevzon et al., 2006; Hollander et al., 2012). Furthermore,
the serotonin precursor, worsens
depletion of
repetitive behavior symptoms in ASD (McDougle et al., 1993,
1996). In addition, gastrointestinal problems are prevalent in
ASD (Adams et al., 2011; Chaidez et al., 2014; McElhanon et al.,
2014), and serotonin is highly involved in gut motility (Sikander
et al., 2009). These comorbidities suggest that disrupted serotonin
signaling may underlie the neurobiology of autism.
tryptophan,
The serotonin system has important interactions with other
systems in the brain. One such example is the interaction
seen in the serotonin and oxytocin (OT) systems, both during
development and in adulthood. Animal models indicate these
systems are anatomically interconnected. Fibers from the dorsal
and median raphe project to the paraventricular (PVN) and
supraoptic (SON) nuclei of the hypothalamus, where oxytocin
receptors (OTR) are distributed around them (Emiliano et al.,
2007). Serotonin acts on OT neurons via serotonin receptors
located in the PVN and SON, where OT is produced (Osei-
Owusu et al., 2005). Likewise, OT acts via OTR on serotonin
neurons in the raphe nuclei, where serotonin is produced,
which may mediate the release of serotonin and have a role
in the anxiolytic effects of OT (Yoshida et al., 2009). While
evidence suggests that these two neurochemical systems may be
working in tandem, it is not yet clear how early SSRI use may
affect neural OT.
Vasopressin (AVP) is structurally and genetically similar to
OT, and both play a central role in modulating the development
of normal social behavior (Carter, 2014). Direct approaches to
target the oxytocinergic and vasopressinergic systems are aimed
at treating social dysfunction in disorders such as ASD. Although
clinical results remain contradictory regarding whether effects
are prosocial or antisocial (De Dreu et al., 2010; Guastella
et al., 2010), recent advances in our understanding of the
complex neurobiology of OT and AVP signaling, release, and
degradation present promising avenues for understanding social
function in ASD.
Animal models are useful in establishing causal links to long-
term effects of perinatal SSRI exposure on social behavior in
offspring (Zucker, 2017). Results are complicated by age, sex, and
context-specific effects. Pre- and postnatal FLX exposure resulted
in loss of a preference for a social partner vs. an empty cage, and
a deficit in social recognition, in mice (Bond et al., 2020). When
rats were tested as pre-adolescents, prior exposure to perinatal
FLX prevented effects of maternal stress on play behavior in
both sexes, but also resulted in an increase in aggressive play in
males only (Gemmel et al., 2017). When tested as adults, perinatal
exposure resulted in sex-specific increases in social behaviors
(Gemmel et al., 2019). Another study of perinatal exposure found
decreases in social interaction in male rats when tested as adults
Frontiers in Behavioral Neuroscience | www.frontiersin.org
2
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 3
Lawrence et al.
Developmental SSRIs in Voles
(Silva et al., 2018). In addition, some types of social behavior (i.e.,
pair bonding) are not present in rats and mice, necessitating a
different animal model.
p = 0.106). We therefore determined that 5 mg/kg was a more
appropriate dose for the current study (data are available in
Supplementary Figure S1).
In the present study, we used the prairie vole as a translational
model of developmental SSRI exposure. Prairie voles are
socially monogamous microtine rodents that form lasting adult
heterosexual pair bonds characterized by the formation of
a partner preference,
intrasexual aggression, and bi-parental
care. Prairie voles are highly social and have a well described
neurohypophyseal nonapeptide system (for review see Young
et al., 2011) and can be tested in standardized assays of social
behavior and anxiety-like behavior (e.g., partner preference,
elevated plus maze). Here we use the prairie vole to examine
how developmental exposure to a SSRI affects adult behavior and
neural OTR, vasopressin 1a (V1aR), and serotonin 1A (5-HT1a)
receptors and to determine if these changes replicate aspects of
the symptomology of ASD.
MATERIALS AND METHODS
voles
Subjects
(Microtus
laboratory-housed prairie
Subjects were
ochrogaster) from the breeding colony at the University of
California, Davis. This colony was derived from a lineage of
stock which was wild-caught near Champaign, IL. Animals
were housed on a 14:10 light dark cycle with lights on at 0600.
Food (Purina high-fiber rabbit chow) and water were available
ad libitum in the home cage. Breeding pairs and offspring
prior to weaning were housed in large polycarbonate cages
(44 cm × 22 cm × 16 cm) and were given compressed
for bedding. Offspring were weaned on
cotton nestlets
postnatal
(PND) 20 and housed in small polycarbonate
cages (27 cm × 16 cm × 16 cm) throughout testing with a
same-sex sibling when available and a similarly aged non-sibling
when not. All procedures were reviewed and approved by the
Institutional Animal Care and Use Committee of the University
of California, Davis.
Drugs
Fluoxetine hydrochloride (Sigma-Aldrich, St. Louis, MO,
United States) was dissolved in isotonic saline in a concentration
of 1 mg/ml. It was then filtered into sterile solution and injected
subcutaneously at the nape of the neck in a dose of 5 mg/kg.
This dose was chosen based on the literature and the results of
our own prior dose finding study. Both 5 and 10 mg/kg doses
of FLX are commonly used in other rodent studies for perinatal
administration (Gemmel et al., 2017, 2019; Grieb and Ragan,
2019). In the prairie vole dose-finding study, we examined the
effect of 5 mg/kg FLX, 10 mg/kg FLX, or saline (SAL) vehicle
on forced swim behavior and sucrose preference in socially
isolated adult female prairie voles. At 5 mg/kg, females struggled
significantly less (when compared to SAL, t36 = −2.92, p = 0.005),
and spent approximately 40% less time immobile (although this
was not statistically significant). In contrast, at 10 mg/kg struggle
behavior did not differ from SAL, and time spent immobile
trended toward an increase (when compared to saline, t37 = 1.64,
Design and Procedures
Virgin prairie voles (20 male, 20 female) were paired and allowed
to raise a litter of pups together undisturbed. On the day of birth
of the second litter, females were hand caught and pups were
briefly removed. Litters were culled to two male and two female
pups when possible. Females were given a subcutaneous injection
of 5 mg/kg FLX or SAL at the nape of the neck and returned
to the home cage along with her pups. On subsequent days, the
female was hand caught and FLX or SAL was injected without
removing the pups from the nipples. Females were dosed daily
in this way with either FLX or SAL until the day of birth of the
fourth litter. This design created three cohorts of FLX exposure:
postnatal exposure in litter 2 (POST), both prenatal and postnatal
exposure in litter 3 (PRE + POST), and prenatal exposure in
litter 4 (PRE) (Figure 1). The average interbirth interval for litter
2–3 was 22.7 ±0.34 days (range 21–26), and for litter 3–4 was
22.9 ±0.19 days (range 21–24).
Parental Care of Prenatally Exposed
Offspring
Parental care is minimally altered following treatment with FLX
(Villalba et al., 1997), however the effects of withdrawal prior to
weaning has not been examined in prairie voles. Parental care
of prenatally FLX-exposed subjects (litter 4) was quantified in
the home cage to determine whether FLX withdrawal would
significantly alter parental behavior. Undisturbed parental care
was observed in the home cage for 20 min once during the
morning and once in the afternoon on 2 days between PND 1-
3. Behaviors were quantified in real-time using Behavior Tracker
1.5 (behaviortracker.com) using methods previously validated
to measure the type and amount of parental care (Perkeybile
et al., 2013). Both maternal and paternal behavior was measured,
including huddling, non-huddling contact,
licking/grooming,
pup retrieval, nest building, and maternal nursing postures.
Behavioral Tests
After weaning, subjects underwent behavioral testing. Half of
each litter, one male and one female when possible, underwent
behavioral testing during periadolescence, between PND21 and
PND39. Periadolescent subjects underwent alloparental care,
elevated plus maze, and open field testing in that order. The
other half of each litter, one male and one female when
possible, underwent behavioral testing as adults, between PND45
and PND120. Adult subjects were tested for alloparental care,
elevated plus maze, and open field;
in addition, they also
underwent intrasexual adult affiliation and partner preference
testing. All behaviors were quantified using Behavior Tracker
1.5 (behaviortracker.com). Behavioral tests occurred from 1
to 5 days apart.
Frontiers in Behavioral Neuroscience | www.frontiersin.org
3
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 4
Lawrence et al.
Developmental SSRIs in Voles
FIGURE 1 | Timeline of maternal daily dosing and subject exposure. GD, gestational day; PND, postnatal day.
Alloparental Care
A minimum of 24 h after weaning, subjects were tested with a
novel pup to measure alloparental care behavior as previously
described (Bales et al., 2004a). Subjects were placed into an arena
consisting of two polycarbonate cages (27 cm × 16 cm × 16 cm)
connected by a short clear tube for a 45-minute acclimation
period. This period was followed by a 10 min test in which a
novel pup (PND 0-4) was placed into the arena. The subject
was free to move about the arena and interact with the pup.
Tests were video-recorded and later scored by a trained observer
blind to condition. Behaviors quantified included frequency and
latency of approach, sniffing, licking and grooming the pup,
autogrooming, physical contact with the pup, huddling, pup
retrievals, non-injurious biting, attacks, digging, and location
in the arena relative to the pup. Digging and autogrooming
were considered potential stereotypical behaviors. When attacks
occurred, the test was immediately stopped and the subject
removed from the arena. If possible, injuries were treated and
the pup returned to the home cage. If necessary, the pup was
euthanized. Each pup was used for no more than two test sessions.
Following testing, animals were returned to their home cage.
Sex differences in prairie voles in this test are well-established,
with males responding with higher levels of alloparental care
than females. This sex difference, although already present in
peri-adolescents, becomes more marked as animals become adult
(Roberts et al., 1998).
considered a potential stereotypical behavior. Following testing
animals were returned to their home cage.
It is worth noting that at baseline, prairie voles spend a
higher amount of time in the open arms of the elevated plus-
maze than mice typically do (Komada et al., 2008). While
across 90 genetically engineered strains, mice spent an average
of 9.19 ± 0.36% time in the open arms of the maze, prairie
voles often spend 35–75% of their time in the open arms
(Bales et al., 2004b; Greenberg et al., 2012). Male prairie voles
tend to spend more time in the open arms, or exhibit higher
frequencies of open arm entries, than females (Bales et al., 2004b;
Greenberg et al., 2012).
Open Field
The open field test was used as a second measure of anxiety
and exploration (Ramos and Mormède, 1997). The open field
consisted of a 40 cm × 40 cm × 40 cm plexiglass arena with a
grid marked on the floor. The subject was placed in the center of
the arena and behavior was digitally recorded for 10 min. Time
spent in the center and the periphery was measured, as well as
the frequency of rearing. Tests were video recorded and later
scored using Behavior Tracker by trained observers with an inter-
rater reliability greater than 90%. Following testing animals were
returned to their home cage. Sex differences for prairie voles are
not well established and are absent in some studies (Greenberg
et al., 2012); we did not therefore predict any sex differences at
baseline for this test.
Elevated Plus Maze
The elevated plus-maze was used as a measure of anxiety and
exploration (Insel et al., 1995) based on the rodent predisposition
to prefer dark enclosed spaces (Campos et al., 2013). The maze
consisted of two open and two enclosed opaque arms, each 67 cm
long and 5.5 cm wide. The arms were elevated 1 m above the
floor. Each vole was placed into the center of the maze and its
behavior was scored for 5 min. Any animals that jumped off
the open arms of the maze were captured and placed back into
the center of the maze. If a subject jumped off the maze three
times, the test was stopped. Throughout the course of the study,
only four animals jumped off the maze, and data from only two
animals had to be removed due to jumping. Trained observers
blind to conditions scored behavior live for duration of time
in the open and closed arms, freezing, and autogrooming with
an inter-rater reliability greater than 90%. Autogrooming was
Intrasexual Adult Affiliation
Subjects were placed into a novel arena (27 cm × 16 cm × 16 cm)
with a stimulus animal of the same sex and body size for
5 min as a low-threat, low-aggression social interaction task
(Perkeybile and Bales, 2015). Behavior was video recorded and
later scored by an observer blind to the treatment condition. The
ethogram used to score behavior included affiliative behaviors
(sniffing, physical contact, allogrooming, and play), anxiety
related behaviors (rearing, digging, abrupt withdrawal), and
aggressive behaviors (lunging, wrestling, chasing). Digging and
autogrooming were considered potential stereotypical behaviors.
Prior to testing, stimulus animals were screened for aggressive
behavior with a novel animal, and were not used if they displayed
high levels of aggression. Stimulus animals were collared prior
to the start of testing to allow for identification during later
behavioral scoring. Stimulus animals were used for a maximum
Frontiers in Behavioral Neuroscience | www.frontiersin.org
4
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 5
Lawrence et al.
Developmental SSRIs in Voles
of 2 tests, and were not reused if they experienced an aggressive
interaction. Tests were continuously monitored for high levels of
aggression and were stopped if necessary. Intense aggression was
rarely seen. Following testing, animals were returned to the home
cage. At baseline, we expected males to be more aggressive and
less affiliative than females (Bales and Carter, 2003b).
(American Radiolabeled Chemicals, St. Louis, MO, United States)
were exposed to Biomax MR film (Kodak, Rochester, NY,
United States) for 72 h and then developed. We have previously
reported a sex difference in the nucleus accumbens shell, with
males displaying higher OTR binding than females at baseline
(Guoynes et al., 2018).
Partner Preference
This test is commonly used as an operational index of the
formation of a pair-bond in the prairie vole (Williams et al.,
1992; Bales and Carter, 2003a; Bales et al., 2013). Male subjects
were housed with a female “partner” for 24 h prior to testing
and female subjects were housed with a male partner for 6 h
prior to testing. These durations have been previously shown to
be sufficient time for the formation of a partner preference and
account for the sex difference in time to pair bond formation
(Williams et al., 1992; DeVries and Carter, 1999). Following this
cohabitation, the opposite-sex mate of the subject (partner) and
a non-related opposite-sex animal matched on age and weight to
the mate (stranger) were tethered in opposing ends of a three-
chamber testing apparatus. The subject was placed untethered in
the empty middle chamber and was free to move about all three
chambers and interact with either the partner or stranger for 3 h.
The test was digitally recorded, and the duration of time in each
of the three locations was quantified, as was the duration of side
by side contact with the stranger and partner.
Brain Extraction and Tissue Sectioning
Brains were taken from behaviorally tested animals of both ages
(juvenile and adult), but only brains from the PRE + POST
exposure cohort were analyzed for receptor binding (see below).
Twenty-four hours after completion of all behavioral testing,
subjects were euthanized via cervical dislocation and rapid
decapitation under deep anesthesia. Brains were removed quickly
and placed in powdered dry ice and then stored at −80◦C
until sectioning. Brain tissue was sectioned coronally in 20 µm
slices at 20◦C on a cryostat (Leica) and thaw mounted on
Fisher Superfrost Plus slides. Slides were stored at −80◦C until
the time of assay.
OTR and V1aR Autoradiography
Because they showed the largest effects on behavior, quantitative
receptor autoradiography for OTR, V1aR, and 5-HT1aR was
performed for the PRE + POST exposure cohort. Analyses
were carried out on the right side of the brain only, as tissue
punches were taken from the left side for additional analyses.
Tissue was allowed to thaw in slide boxes containing desiccant
packets. OTR and V1aR autoradiography was performed as
previously reported (Perkeybile and Bales, 2015) with minor
adjustments. For OTR binding, the ligand 125I-OVTA [125I-
ornithine vasotocin [d(CH2)5[Tyr(Me)2, Thr4, Orn8, (125I)Tyr9-
NH2] analog], 2200Ci/mmol (Perkin Elmer, Waltham, MA,
United States) was used. For V1aR binding, the ligand 125I-
LVA [125I-lin-vasopressin [125I-phenylacetyl-D-Tyr(ME)-Phe-
Gln-Asn-Arg-Pro-Arg-Tyr-NH2] analog], 2200Ci/mmol (Perkin
Elmer, Waltham, MA, United States) was used. After assay
completion, slides along with 125I-autoradiographic standards
5-HT1A Autoradiography
For 5-HT1A binding, 3.0 nM [3H]WAY-100635, 74Ci/mmol
(Perkin Elmer, Waltham, MA, United States) was used. Tissue
was rinsed in 50 mM Tris–HCl buffer (pH 7.5) followed by a
120 min incubation in the tracer buffer at room temperature.
10 nM of L-485,870, a dopamine antagonist, was included to
prevent binding of WAY-100635 to Dopamine D4 receptors.
Following the incubation period, tissue was rinsed twice in
50 mM Tris buffer at 4◦C and then dipped in dH2O and air dried.
Tissue was exposed to Carestream BioMax MR Film (Kodak,
Rochester, NY, United States) for 6 weeks with 3H microscale
standards (American Radiolabeled Chemicals, St. Louis, MO,
United States). We had no a priori predictions as far as 5-HT1A
binding sex differences at baseline for this species.
Quantification
Experimenters were blind to conditions during autoradiogram
quantification. ImageJ software (National Institutes of Health,
Bethesda, MD, United States) was used to quantify OTR optical
binding density (OBD) in previously reported (Insel and Shapiro,
1992) regions of interest (ROI) including the nucleus accumbens
core and shell, anterior central amygdala, and the lateral septum,
and for V1aR in the medial amygdala,
lateral septum, and
ventral pallidum. 5-HT1aR OBD were quantified in the anterior
and posterior lateral septum, dorsal hippocampus, dorsal raphe,
and frontal cortex using MCID Core Digital Densitometry
system (Cambridge, United Kingdom). The ten standard OBD
values were used to generate a standard curve. Three separate
measurements for ROIs and background OBD were averaged to
yield normalized values and account for individual variation in
background across samples.
Data Analysis
Statistical analyses were conducted using SAS 9.4 (SAS Institute,
Cary, NC, United States). All analyses were carried out using
generalized linear mixed models (GLMM) utilizing backward
selection to eliminate non-significant variables from the model.
Significance level was set at p < 0.05 for all analyses and all
tests were two-tailed. Data were checked for normality, and if
not normally distributed, square root, quad root, or reciprocal
transformation was used. If data was not transformable to
normality, a GLMM was still used as recommended by Feir-
Walsh and Toothaker (1974). Post hoc analyses utilized least
squares means when the omnibus test was significant. The
random factor used in all analyses was a pair ID (for the subject’s
parents) to account for differences due to parenting or genetic
background for subjects within the same litter or across litters.
Drug condition was nested within this term, as each female
maintained a consistent drug condition throughout the study and
thus all offspring of a given pair had the same drug condition.
Frontiers in Behavioral Neuroscience | www.frontiersin.org
5
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 6
Lawrence et al.
Developmental SSRIs in Voles
When a three-way interaction was statistically significant, all two-
way interactions which included the variables in the three-way
interaction were left in the model even if not significant.
Parental Care
A multivariate mixed model was used for analysis of parental
care behavior. All three types of nursing were included in one
model, as were behaviors that were examined concomitantly in
both mothers and fathers that were not independent, such as
huddling. Factors included in the model were pair ID and drug
condition of the mother prior to cessation of treatment, as well as
age of pups at observation and time of day as covariates.
Alloparental Care Test
For the alloparental care analyses, variables were summed for
duration of time in the same location (with the pup) or different
location (without the pup) in the testing arena. A ratio was
created to examine relative proportion of time spent in the
same location as the pup relative to duration in a different
location than the pup using the equation: ratio = with the
pup/(with the pup + without the pup). Factors included in
the model were pair ID, drug condition, sex, exposure cohort,
age group, and interactions of these factors. Also analyzed were
time spent in contact to the pup, time spent retrieving the
pup, time spent in proximity to the pup, latency to approach,
duration of social investigation, duration of licking, and duration
of huddling over the pup.
Elevated Plus Maze
For the elevated plus maze analysis, a ratio was created to
examine the proportion of time spent on the open arms relative
to total time on the maze using the equation: ratio = time
on open arms/(time on open arms + time on closed arms).
Factors included in the model were pair ID, drug condition, sex,
exposure cohort, age group, and interactions of these factors.
Autogrooming, entries onto the arms of the maze, and duration
of freezing, were also analyzed.
Open Field Test
For the open field test analyses, a ratio was created to examine
proportion of time spent in the center of the arena relative to
total time using the equation: ratio = time in center/(time in
center + time in periphery). Factors included in the model were
pair ID, drug condition, sex, exposure cohort, age group, and
interactions of these factors. Rearing was also analyzed.
Intrasexual Adult Affiliation
For the intrasexual adult affiliation analyses, the frequency of
aggressive behavior was calculated by summing the frequencies
of lunging and wrestling. Factors included in the model for
each behavior (including affiliative, anxiety-like, and aggressive
behaviors, as described above) were pair ID, drug condition, sex,
exposure cohort, and interactions of these factors.
Partner Preference Test
For between-group partner preference test analyses, a difference
score was created to examine duration of time spent in the same
FIGURE 2 | Parental care of prenatal exposure subjects. (A) Mean (±SEM)
total, neutral, lateral, and active nursing duration comparing mothers
previously exposed to saline to mothers previously exposed to fluoxetine.
(B) Mean (±SEM) duration of nest building in mothers previously exposed to
saline and their male pair-mates (fathers) compared to mothers previously
exposed to fluoxetine and their pair-mates. *p < 0.05.
cage as the partner relative to time spent with the stranger using
the equation: difference = time with partner - time with stranger.
The same procedure was used to examine physical contact
with the partner relative to contact with the stranger using the
equation: difference = time in contact with the partner - time in
contact with the stranger. Duration of time spent in the empty
chamber was analyzed separately, and square root transformed
for analyses to make the residuals for this model normally
distributed. Factors included in the model were pair ID, drug
condition, sex, exposure cohort, and interactions of these factors.
Within-group partner preference analyses for the SAL and
FLX groups were performed using matched t-tests for time
spent in contact with the partner vs. time spent in contact
with the stranger.
Oxytocin, Vasopressin 1a, and Serotonin
1a Receptor Binding
For all binding analyses, density of binding in three sequential
areas of each ROI were averaged for each individual. The model
Frontiers in Behavioral Neuroscience | www.frontiersin.org
6
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 7
Lawrence et al.
Developmental SSRIs in Voles
included pair ID, drug condition, sex, age group, and interactions
of these factors.
Pearson correlations were calculated for the 4 ROIs quantified
for OTR and the 3 ROIs quantified for V1aR with difference
in time in physical contact and duration of
time in the
empty chamber in the partner preference test. Correlation
of OTRs in the central amygdala and proportion of
time
on the open arms of
the elevated plus maze was also
examined. When multiple comparisons were made within a
single behavioral or neuroanatomical test, a Benjamini-Hochberg
false discovery rate adjustment for multiple comparisons was
used (Benjamini and Hochberg, 1995).
RESULTS
Parental Care
Parental care of the PRE cohort was minimally altered by the drug
condition of the mother, either FLX withdrawal or no withdrawal
from SAL at the time of parenting. Drug condition did not alter
total duration of nursing, nor did it alter duration of neutral
nursing postures or lateral nursing postures. However, duration
of active nursing was altered by drug condition (F1,51 = 5.11,
p < 0.05), with FLX-withdrawing dams spending more time
in active nursing than those who had been treated with SAL
(Figure 2A). Nest building duration was also greater in FLX-
withdrawing mothers (F1,51 = 4.06, p < 0.05) as well as their
untreated male pair-mates (F1,51 = 4.79, p < 0.05) compared
to pairs in which mothers were previously treated with SAL
(Figure 2B). Because of the high amount of variability in this
behavior, we also analyzed nest-building with a non-parametric
Kruskal-Wallis test. The duration of nest-building in FLX-
withdrawing mothers, compared to SAL mothers, remained
1 = 4.62, p < 0.05), however, the effect was non-
significant (χ2
significant in their male mates (χ2
1 = 1.14, p > 0.05). All
other behaviors observed were not affected by drug condition
including maternal huddling, paternal huddling, maternal non-
huddling contact, paternal non-huddling contact, maternal
licking and grooming, paternal licking and grooming, maternal
pup retrieval, paternal pup retrieval, maternal autogrooming, or
paternal autogrooming.
Behavior of Developmentally Exposed
Offspring
Alloparental Care Test
Duration of overall pup physical contact was greater in males
than in females (F1,167 = 8.28, p < 0.01). A three-way interaction
of condition, sex, and age group (F1,167 = 3.77, p < 0.05) indicated
that among FLX subjects, adult females were in contact with the
pup less than periadolescent females (t41 = 2.88, p < 0.05) and
that among SAL subjects, periadolescent females were in contact
with the pup less than periadolescent males (t49 = 2.06, p < 0.05).
Adult females spent less time in contact with the pup compared
to adult males exposed to either SAL (t52 = 1.97, p < 0.05) or
FLX (t44 = 2.83, p < 0.01) (Figure 3A). Put another way, females
were in contact with the pup less than males under matching
conditions, with the exception of FLX periadolescent females,
FIGURE 3 | Alloparental care behavior. (A) Mean (± SEM) duration of physical
contact with the pup comparing saline and fluoxetine exposure by age and
sex. (B) Mean (±SEM) duration of pup retrieval comparing saline and
fluoxetine exposure by exposure cohort. (C) Mean (± SEM) latency to
approach the pup, sniffing, and huddling comparing saline and fluoxetine
exposure. *p < 0.05, **p < 0.01.
which spent more time in contact with the pup than did FLX
periadolescent males.
Duration of time spent retrieving the pup tended to be
greater in males than in females (F1,163 = 3.69, p = 0.057).
A drug condition by cohort interaction (F2,163 = 3.44, p < 0.05)
(Figure 3B) indicated that in the PRE + POST cohort, FLX
subjects spent more time retrieving the pup than SAL subjects
(t63 = 2.34, p < 0.05), and that in FLX subjects, PRE + POST
Frontiers in Behavioral Neuroscience | www.frontiersin.org
7
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 8
Lawrence et al.
Developmental SSRIs in Voles
FIGURE 4 | Elevated plus maze. Mean (±SEM) proportion of time spent in the
open arms relative to total time comparing saline and fluoxetine exposure by
age. *p < 0.05.
FIGURE 5 | Open field test. Mean (±SEM) proportion of time spent in the
center relative to total time comparing saline and fluoxetine exposure by age
and sex. Different letters indicate a significant difference at p < 0.05.
subjects spent more time retrieving than PRE (t60 = 2.40,
p < 0.05) and POST (t58 = 2.47, p < 0.05) subjects.
Fluoxetine exposure had no effect on proximity to the pup,
licking the pup, latency of approach, social investigation, or
huddling (Figure 3C). Ratio of time spent in the same chamber
of the testing arena as the pup relative to total time was not
altered by drug condition, nor was latency to approach the pup,
duration of sniffing, huddling, licking, or grooming of the pup.
There was no indication of heightened repetitive behavior with
FLX exposure, and duration of autogrooming and digging were
not altered by drug condition.
Elevated Plus Maze
Proportion of time spent in the open arms relative to total
time on the maze showed an interaction of drug condition
and age group (F1,141 = 4.02, p < 0.05) such that FLX-
exposed adults spent a lower proportion of time in the open
arms compared to SAL-exposed adults (t64 = 2.21, p < 0.05),
while there was no such difference in periadolescent subjects
(Figure 4). Drug condition did not alter the number of entries
onto the arms of the maze, duration of freezing, or duration
of autogrooming.
Open Field Test
Proportion of time spent in the center of the open field relative
to total time showed a three-way interaction of drug condition,
sex, and age group (F4,119 = 4.66, p < 0.01) (Figure 5). In
SAL-exposed females, periadolescents spent more time in the
center than adults (t39 = 2.48, p = 0.01), while this was not true
for FLX-exposed subjects (t30 = 1.29, p = 0.20). Among SAL
exposed subjects, time in the center was greater in adult males
than adult females (t31 = 3.42, p < 0.001), in periadolescent
females than periadolescent males (t44 = 1.94, p = 0.05), and in
adult males than periadolescent males (t36 = 3.00, p < 0.01).
There was also a trend level difference between SAL males and
SAL females (t76 = 1.91, p = 0.06). There were no sex or age
group differences within the FLX-exposed subjects. Duration
of autogrooming and frequency of rearing were not affected
by drug condition.
the stimulus animal,
Intrasexual Adult Affiliation Test
Duration of sniffing of
the primary
form of social investigation, did not differ by drug condition.
Duration of allogrooming of the stimulus animal showed a
trend level interaction of drug condition and sex (F1,91 = 3.73,
p = 0.057). FLX exposed males spent more time allogrooming
than SAL exposed males (t49 = 1.77, p = 0.07), and SAL females
spent more time allogrooming than SAL males (t48 = 1.91,
p = 0.059). Duration of time in physical contact with the stimulus
animal, autogrooming, or frequency of rearing were not altered
by drug condition.
Frequency of aggressive behavior was not altered by drug
condition. In contrast, duration of digging showed an interaction
of treatment and sex (F1,73 = 4.62, p < 0.05) (Figure 6A). SAL
males dug more than SAL females (t48 = 2.53, p < 0.05), but there
was no sex difference in FLX exposed subjects.
Duration of play with the stimulus animal showed an
interaction of drug condition and sex (F1,91 = 5.75, p < 0.05)
(Figure 6B). FLX males played more than FLX females (t45 = 2.23,
p < 0.05) and SAL males (t49 = 2.36, p < 0.05).
greater
Partner Preference Test
Difference in duration of time in the partner and stranger
chambers was
compared to males
in females
(F1,74 = 12.95, p < 0.001) but did not differ by cohort or
drug condition (Figure 7A). Difference in duration of time in
side-by-side contact with the partner and the stranger was not
altered by cohort but did show an interaction of sex and drug
condition (F1,73 = 4.01, p < 0.05) (Figure 7B). SAL females spent
more time in physical contact with the partner than SAL males
(t40 = 2.62, p < 0.01), but there was no sex difference in the FLX
condition. Within the SAL group, females formed a significant
preference for the partner (t24 = 3.44, p = 0.002), while males
did not (t16 = −0.14, p = 0.891). Within the FLX group, neither
females (t18 = 1.672, p = 0.121) nor males (t16 = 1.816, p = 0.07)
formed a significant preference.
Duration of time spent in the empty chamber in the partner
preference test showed an interaction of drug condition and
exposure cohort (F2,70 = 4.17, p < 0.05) (Figure 7C). Subjects
Frontiers in Behavioral Neuroscience | www.frontiersin.org
8
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 9
Lawrence et al.
Developmental SSRIs in Voles
Oxytocin receptors binding in the anterior central amygdala
was decreased with FLX exposure compared to SAL exposure
(F1,46 = 8.42, p < 0.01). There was no effect of sex on
OTR binding in the central amygdala. A condition by age
(Figures 8D,
group interaction (F1,46 = 3.98, p = 0.05)
9B)
indicated that FLX adults had lower OTR binding
compared to SAL adults (t66 = 3.26, p < 0.01), and that
SAL adults had higher OTR binding than SAL periadolescents
(t34 = 2.01, p = 0.05), but
this age difference was not
found with FLX exposure. OTR binding in the lateral
septum was not altered by drug condition (Figure 8E),
sex, or age group.
Oxytocin receptors binding did not correlate with difference
in contact between the partner and stranger or duration in the
empty chamber in the partner preference test. There was also no
correlation between OTR binding in the central amygdala and
proportion of time on the open arms of the elevated plus maze.
Vasopressin 1a Receptors
Vasopressin 1a binding in the medial amygdala was reduced
by FLX exposure compared to SAL exposure (F1,47 = 4.20,
p < 0.05) (Figures 10A, 9C). V1aR binding in the medial
amygdala was not altered by sex or age group. V1aR binding
in the lateral septum was not altered by drug condition,
sex, or age group (Figure 10B). V1aR binding in the ventral
pallidum was not altered by drug condition, sex, or age
group (Figure 10C).
Vasopressin 1a binding density in the three ROIs quantified
did not correlate with difference in contact between the
partner and stranger or duration in the empty chamber in
the partner preference test once adjusted to account
for
multiple comparisons.
Serotonin 5-HT1a Receptors
Unexpectedly, there was no effect of FLX exposure on 5-HT1A
receptor binding density in any ROI examined (anterior and
posterior lateral septum, dorsal hippocampus, dorsal raphe,
frontal cortex) nor were there any significant interactions of age
group, sex, and ROI (Figures 11A–E).
DISCUSSION
Understanding the etiology of
the increased risk of ASD
associated with developmental SSRI exposure is an area of
research which can greatly benefit from animal models. Here, we
used the prairie vole as a translational model in which to examine
how exposure to an SSRI, FLX, affects behavior, neuropeptide
receptors, and serotonin receptors in the brain.
We examined three primary behavioral domains which are
repetitive behavior,
associated with ASD:
social behavior,
and anxiety-like behavior. The first
the
two represent
impaired social
two primary diagnostic criteria for ASD,
repetitive behavior;
communication and stereotyped or
the
frequently
comorbid in ASD (White et al., 2009; van Steensel et al.,
communication domain of
2011). Modeling
the heightened anxiety
third represents
social
the
FIGURE 6 | Intrasexual adult affiliation. (A) Mean (± SEM) duration of digging
comparing saline and fluoxetine exposure by sex. (B) Mean (±SEM) duration
of play comparing saline and fluoxetine exposure by sex. *p < 0.05.
in the PRE cohort that were exposed to FLX spent more time
in the empty chamber than those exposed to SAL (t26 = 2.06,
p < 0.05). Time in the empty chamber was not altered by sex,
nor were there differences by drug condition in the PRE + POST
or POST conditions.
Quantitative Receptor Autoradiography
Oxytocin Receptors
Oxytocin receptors binding in the nucleus accumbens core was
lower in FLX subjects compared to SAL subjects (F1,43 = 3.96,
p = 0.05) and was greater in adult compared to periadolescent
subjects (F1,43 = 7.18, p < 0.01). A drug condition by sex
interaction (F1,43 = 4.89, p < 0.05) (Figures 8A, 9A) indicated
that FLX females had less OTR binding than SAL females
(t31 = 2.84, p < 0.01) and FLX males (t30 = 2.20, p < 0.05).
A drug condition by age group interaction (F1,43 = 5.02,
p < 0.05) (Figure 8B) indicated that FLX adults had less OTR
binding than SAL adults (t28 = 2.73, p < 0.01). Adults also
had greater OTR binding compared to periadolescents with SAL
exposure (t34 = 3.50, p = 0.001), but this was not the case
with FLX exposure (t30 = 0.31, p = 0.76). OTR binding in the
nucleus accumbens shell did not differ by drug condition or
sex. Adult subjects had greater OTR binding in the nucleus
accumbens shell than periadolescents (F1,45 = 3.92, p = 0.05;
Figure 8C).
Frontiers in Behavioral Neuroscience | www.frontiersin.org
9
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 10
Lawrence et al.
Developmental SSRIs in Voles
FIGURE 7 | Partner preference test. (A) Mean (±SEM) difference in duration between time spent in the partner chamber and the stranger chamber comparing saline
and fluoxetine exposure by sex. (B) Mean (±SEM) difference in duration between time spent in side-by-side contact with the pair-mate and the stranger comparing
saline and fluoxetine exposure by sex. (C) Mean (±SEM) duration of time in the empty chamber comparing saline and fluoxetine exposure by exposure cohort.
*p < 0.05, **p < 0.01, ***p < 0.001.
ASD is particularly difficult
in animal models. Verbal
language is uniquely human, and thus the precise deficits
found in individuals with ASD cannot be modeled in
any animal species.
We examined sociality by measuring species-typical behaviors
involved in social interaction and looking for deficits in FLX
exposed subjects. Social investigation (sniffing) was not altered
by FLX with a novel social partner, be it a pup or an adult
conspecific. Affiliative behavior, which is ubiquitous in prairie
voles, was altered by FLX exposure (Table 1). We observed
changes in alloparental care (Figures 3A,B), in play behavior with
a same-sex adult (Figure 6B), and in time spent in the empty
Frontiers in Behavioral Neuroscience | www.frontiersin.org
10
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 11
Lawrence et al.
Developmental SSRIs in Voles
FIGURE 8 | Oxytocin receptor binding. (A) Mean (± SEM) optical binding density in the nucleus accumbens core comparing saline and fluoxetine exposure by sex.
(B) Mean (±SEM) optical binding density in the nucleus accumbens core comparing saline and fluoxetine exposure by age. (C) Mean (±SEM) optical binding density
in the nucleus accumbens shell comparing saline and fluoxetine exposure by age. (D) Mean (±SEM) optical binding density in the central amygdala comparing saline
and fluoxetine exposure by age. (E) Mean (±SEM) optical binding density in the lateral septum comparing saline and fluoxetine exposure. *p < 0.05, **p < 0.01,
***p < 0.001.
chamber of the partner preference test (Figure 7C). The changes
in alloparental care were primarily in retrieval behavior, with
males that had been treated with both prenatal and postnatal FLX
spending significantly more time retrieving (Figure 3B). These
males were picking up the pup in their mouths and running
excitedly around the test arena, in an apparently less organized
manner of providing care for the pup.
During the partner preference test, prenatal FLX exposure
also led subjects of both sexes to opt out of social interaction in
favor of time alone in the empty cage (Figure 7C), indicating
that FLX led to a rejection of social interaction very atypical
of prairie voles. However, FLX males also spent more time
in play behavior with stimulus males during the intrasexual
affiliation test. Much as the research in humans suggests, prenatal
SSRI exposure may increase the likelihood of asociality, or the
alteration or disorganization of sociality; but it does so in subtle,
non-deterministic ways.
The
neurohypophyseal
and
vasopressin, are likely candidates to be involved in such shifts in
sociality due to their developmental interaction with serotonin
nonapeptides,
oxytocin
Frontiers in Behavioral Neuroscience | www.frontiersin.org
11
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 12
Lawrence et al.
Developmental SSRIs in Voles
FIGURE 9 | Representative autoradiograms of oxytocin and vasopressin 1a receptor binding. Please note that tissue punches were taken from the left side of each
brain to assess additional outcome measures not reported here. (A) Oxytocin receptor binding in the nucleus accumbens core shows a sex by drug condition
interaction (see also Figure 8A). (B) Oxytocin receptor binding in the central amygdala shows an age by drug condition interaction (see also Figure 8C).
(C) Vasopressin 1a receptor binding in the medial amygdala shows a drug condition effect (see also Figure 9A).
Frontiers in Behavioral Neuroscience | www.frontiersin.org
12
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 13
Lawrence et al.
Developmental SSRIs in Voles
FIGURE 10 | Vasopressin 1a receptor binding. (A) Mean (±SEM) optical binding density in the medial amygdala comparing saline and fluoxetine exposure. (B) Mean
(±SEM) optical binding density in the lateral septum comparing saline and fluoxetine exposure. (C) Mean (±SEM) optical binding density in the ventral pallidum
comparing saline and fluoxetine exposure. *p < 0.05.
as well as their important roles in social behavior across species
(Carter and Perkeybile, 2018). We found that FLX exposure
reduced the binding density of oxytocin receptors in the nucleus
accumbens core and the central amygdala (Figures 8A,B,D),
and the binding density of vasopressin 1a receptors in the
medial amygdala (Figure 9A). While the nucleus accumbens
shell has been strongly implicated in studies of prairie vole pair
bonding, oxytocin receptors in the core are under-studied in the
neurobiology of social behavior in voles, and may represent a
new avenue of investigation.
It is likely that changes in OTR and AVPR1a underlie the
differences found not only in social behavior, as described above,
but also in anxiety-like behavior. Anxiety-like behavior was
altered in the elevated plus maze (Figure 4), where adults spent
less time on the open arms if developmentally exposed to FLX,
regardless of the timing of exposure. This result is in line with
previous research which has reported an increase in anxiety-like
behavior in adults exposed to an SSRI developmentally (Ansorge
et al., 2004; Boulle et al., 2016). We also found that FLX
exposed subjects had lower OTR in the central amygdala during
adulthood but not during periadolescence (Figure 8D). The
amygdala is an area of the brain that is highly involved in
anxiety and emotion regulation (Babaev et al., 2018). OTRs in
the central amygdala are known to be involved in anxiety, as
well as regulation of the hypothalamic-pituitary-adrenal axis,
and can play a role in mediating the stress response (Neumann
et al., 2000). Likewise, V1aR in the amygdala mediate stress and
anxiety, with binding at V1aRs linked to heightened anxiety,
reducing time spent in the open arms of the elevated plus
maze (Hernández et al., 2016). Taken together, one potential
mechanism by which developmental exposure to FLX increases
anxiety in adulthood may be the reduction of OTRs and V1aRs
in the amygdala.
there was no indication of
While developmental FLX altered social and anxiety related
behaviors,
increased repetitive
behaviors in FLX exposed subjects. We found no increase
in stereotypies
tests examined.
Autogrooming and digging were not increased by FLX exposure
in any of the behavioral tests in which they were measured.
the behavioral
in any of
Changes in offspring behavior may have been mediated by
changes in the behavior of the mothers treated with FLX,
although these were relatively subtle. In particular, mothers
that were withdrawing from FLX spent extra time in active
nursing (Figure 2A) and in nest-building (Figure 2B). The
male pair mates of the FLX-withdrawing mothers also spent
higher amounts of time in nest-building (although this effect was
eliminated when the data were examined non-parametrically).
Unfortunately, we missed the opportunity to assess the quality
of the nests being produced (Figure 2B). Nest quality is an often-
used measure of parental behavior in rodents and other species
(Mann, 1993; Deacon, 2012). In three-spined sticklebacks, FLX
reduced measures of male nest quality (Sebire et al., 2015); while
in mice, females prenatally treated with FLX displayed lower nest
quality during early days postpartum (Svirsky et al., 2016). The
quality of the nest could affect various measures for the offspring
including survival (Hamilton et al., 1997), thermoregulation
(Gaskill et al., 2013), and even sleep (Harding et al., 2019). It is
possible that the FLX-withdrawing parents put in extra time nest-
building, while still producing low quality nests. A disorganized
approach to nest-building would be consistent with the active
nursing behavior of the mothers, which is when they locomote
around the cage with the pups still attached to the nipples (prairie
vole pups have milk teeth). Given that the pups are being bounced
against substrate as they are dragged around, we have generally
regarded this as a lower quality form of maternal behavior.
Active nursing is also higher in prairie vole mothers that are
broadly characterized as “low contact” mothers (Perkeybile et al.,
2013). Future research on this topic should include nest quality
as a variable in aiding understanding of the effects of FLX on
parental behavior.
A major limitation of this study is that we did not find
a partner preference in the SAL-treated males (Figure 7B).
A possible explanation for this is that the daily injections
inadvertently created a prenatal stress paradigm to which all
subjects were exposed. Daily saline injections in pregnant rats
have been shown to be sufficient to change several aspects of
stress reactivity and the serotonin system in offspring (Peters,
1982). Prenatal stress has been shown to alter the social behavior
Frontiers in Behavioral Neuroscience | www.frontiersin.org
13
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 14
Lawrence et al.
Developmental SSRIs in Voles
FIGURE 11 | Serotonin receptor 1a binding. (A) Mean (±SEM) optical binding density in the dorsal hippocampus comparing saline and fluoxetine exposure.
(B) Mean (±SEM) optical binding density in the dorsal raphe comparing saline and fluoxetine exposure. (C) Mean (±SEM) optical binding density in the frontal cortex
comparing saline and fluoxetine exposure. (D) Mean (±SEM) optical binding density in the anterior lateral septum comparing saline and fluoxetine exposure.
(E) Mean (±SEM) optical binding density in the posterior lateral septum comparing saline and fluoxetine exposure.
of offspring (Weinstock, 2001; Schulz et al., 2011; Wilson and
Terry, 2013) and likely prevented any of our animals from
forming a preference. However, the finding that prenatally FLX
exposed subjects spent more of their time alone compared
interest
to SAL treated animals suggests a change in social
above and beyond that involved in the formation of a partner
preference. Furthermore, maternal stress adds ecological validity
given that in human prenatal SSRI use there is an underlying
psychiatric condition for which pharmacological treatment with
SSRIs has been prescribed. Chronic stress is frequently used in
Frontiers in Behavioral Neuroscience | www.frontiersin.org
14
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 15
Lawrence et al.
Developmental SSRIs in Voles
TABLE 1 | Summary of behavioral effects of fluoxetine exposure.
Behavioral test
Measure
Effect of fluoxetine
Interacts with
Results
Alloparental Care
Physical contact
Pup retrieval
Same chamber as pup
Latency to approach
Sniff
Huddle
Lick and groom
Autogroom
Dig
Elevated plus maze
Ratio of time on open arms
Arm entries
Freeze
Autogroom
Open field test
Ratio of time in center
Intrasexual adult affiliation
Autogroom
Rear
Sniff
Allogroom
Physical contact
Autogroom
Rear
Aggression
Dig
Play
Partner preference test
Difference in partner and
stranger chamber time
Difference in side-by-side
contact
Empty chamber time
Y, significant effect; N, no effect; peri, periadolescent.
Y
Y
N
N
N
N
N
N
N
Y
N
N
N
Y
N
N
N
Y
N
N
N
N
Y
Y
N
Y
Y
Sex, age group
Exposure cohort
FLX adult female < FLX peri female
SAL peri female < SAL peri male
FLX PRE + POST > SAL
PRE + POST FLX
PRE + POST > FLX PRE, FLX
POST
–
–
–
–
–
–
–
Age
–
–
–
–
–
–
–
–
–
–
FLX adult < SAL adults
–
–
–
Sex, age group
Eliminated sex and age differences
seen in SAL
–
–
–
Sex
–
–
–
–
Sex
Sex
–
Sex
Exposure cohort
–
–
–
FLX male > SAL male (trend)
Eliminated sex difference seen in
SAL
–
–
–
–
Eliminated sex difference seen in
SAL
FLX male > FLX female FLX
male > SAL male
–
Eliminated sex difference seen in
SAL
FLX PRE > PRE SAL
the laboratory to induce a learned helplessness phenotype of
depressive-like behavior to model depression (Pollak et al., 2010).
An interesting and unexpected finding was that FLX exposure
eliminated sex differences across multiple behavioral tests. One
example is the change in physical contact with the pup seen
in the alloparental care test (Figure 3A). Male prairie voles
are typically more alloparental than females, and here we saw
that with FLX exposure, male periadolescents were not more
alloparental than females, as was the case with SAL exposure.
Male alloparental care is directly impacted by estrogen receptor
expression, and sex-dependent changes in alloparental care
with increasing age are based on changes in estrogen receptor
expression (Perry et al., 2015). FLX exposure also eliminated
the sex difference in partner and stranger contact in the partner
preference test (Figure 7B). Both alloparental care and partner
preference are examples of behaviors that show well-established
sex differences in prairie voles. Estrogen receptor α expression
has been implicated in reducing heterosexual adult contact in
the partner preference test as well as male alloparental care
behavior (Lei et al., 2010). FLX has estrogenic effects both in vivo
and in vitro (Jacobsen et al., 2015; Pop et al., 2015; Muller
et al., 2016), as does its bioactive metabolite norfluoxetine (Lupu
et al., 2015). There is evidence in the literature for sex-specific
effects of FLX on estrogen receptor expression (Adzic et al.,
2017). FLX may have altered estrogen receptor expression, which
in turn reduced affiliative behavior specifically in males, thus
abolishing the sex differences seen in the SAL exposure groups.
Future work should more thoroughly characterize the effects of
developmental FLX on steroid receptors to further understand its
behavioral effects.
Frontiers in Behavioral Neuroscience | www.frontiersin.org
15
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 16
Lawrence et al.
Developmental SSRIs in Voles
Developmental timing is likely to be important in SSRI
exposure. While some work has suggested that in humans, any
chronic exposure in the year prior to birth results in heightened
risk (Croen et al., 2011), others have found that either the first
or third trimester are the periods of greatest risk (Oberlander
et al., 2008; Croen et al., 2011; Harrington et al., 2014). In order
to address the effects of exposure timing, we evaluated behavior
in three different gross exposure cohorts spanning prenatal
and postnatal development. We found few effects of FLX that
were specific to an exposure cohort with the notable exception
of increased duration in the empty chamber of the partner
preference test in the PRE cohort. It is likely that creating shorter
dosing periods which translate to specific trimesters in human
pregnancy would be beneficial to more accurately determining
how to best limit risk to offspring based on timing of exposure.
It is also worth pointing out that due to study design,
offspring with different exposure timing were born to mothers of
different parity and were potentially subject to different maternal
hormone exposures. For example, pups that were part of the
PRE + POST cohort were being nursed by mothers which
were becoming pregnant again. To the extent that variation in
maternal hormones due to parity or pregnancy may have affected
hormones during the postpartum estrus or lactation (Bridges and
Byrnes, 2006; Bridges, 2016), altering pup hormonal exposure in
utero or through milk, these exposures may have varied in this
study. In addition, all subjects in that cohort were litter 3 for their
parents, whereas subjects in the POST cohort were all litter 2, and
subjects in the PRE cohort were all litter 4; which could have also
had effects on hormone exposure.
We have shown here that developmental SSRI exposure
alters OTR and AVPR1a, but not 5-HT1A, binding. Because
FLX’s mechanism works to increase serotonin neurotransmission
by blocking reuptake of serotonin, it was surprising to find
that 5-HT1A receptor binding was unchanged by FLX in all
regions examined. Studies in mice have shown that perinatal
FLX can regularize 5-HT1A levels that have been altered by
other developmental factors (Nagano et al., 2012; Stagni et al.,
2015). For the current study, it appears that the behavioral
effects were mediated by OTR and V1aR without concomitant
changes in the 5HT system. However, while there was no
change in serotonin receptor density, actions on OTR and V1aR
subsequent to FLX exposure may have been precipitated by
changes in the peptides themselves, the function or location of
the receptor, or other downstream cellular mechanistic pathways.
Serotonin developmentally autoregulates its own innervation
throughout the brain (Herlenius and Lagercrantz, 2004) and is
plastic throughout development. Fetal exposure to FLX is poorly
understood, yet it is clear that it leads to changes that last well into
adulthood (Kiryanova et al., 2013). While SSRIs are presumed to
increase extracellular serotonin in the long term, short term SSRI
exposure can reduce raphe cell firing by acting on autoreceptors
leading to a reduction in extracellular serotonin (Tao et al., 2000).
Such activity may have neurodevelopmental consequences for
offspring that have yet to be elucidated fully, but which warrant
further investigation.
The serotonin system is also an extensive system with 15
different types of receptors (Carr and Lucki, 2011). We chose to
examine the 1A receptor because of its autoreceptor function, but
it may be the case that other exclusively post-synaptic serotonin
receptors were altered while 1A was not. Further work examining
other serotonin receptor populations will be important to clarify
how serotonergic neurotransmission is altered by SSRI use
prenatally. It is also possible that species differences between
mice and voles may have altered the effects of FLX on 5-HT1A
receptor binding.
Another area that should be considered is how exposure
interacts with the maternal and early postnatal environment, as
environmental moderation of SSRI effects may underlie their
effects (Alboni et al., 2017). Since the prevalent and incident use
of SSRI-exposed pregnancies has increased in the last two decades
(Alwan et al., 2011), it is of the utmost importance that we more
clearly understand the causes and consequences that prenatal
SSRI exposure may have on the developing brain.
DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be
made available by the authors, without undue reservation.
ETHICS STATEMENT
The animal study was reviewed and approved by Institutional
the University of
Animal Care and Use Committee of
California, Davis.
AUTHOR CONTRIBUTIONS
RL and KB designed the research. RL, MP, CG, and SF conducted
the experiments. RL, SF, and KB analyzed the data. RL wrote the
first draft of the manuscript. All authors edited the manuscript.
FUNDING
This work was supported by an Autism Science Foundation
predoctoral fellowship to RL and HD071998 to KB.
ACKNOWLEDGMENTS
Special thanks to Kenny Nguyen, Tiffany Chen, Gabriel Larke,
Jennifer Nicosia, Elizabeth Sahagun-Parez, Erin Mast, J’aime
Gass, Henry Yang, and Amira Shweyk for their indispensable help
in carrying out data collection, and to Cindy Clayton for her
excellent veterinary care. Many thanks to Forrest Rogers for the
preparation of Figure 9.
SUPPLEMENTARY MATERIAL
for this article can be found
at: https://www.frontiersin.org/articles/10.3389/fnbeh.
The Supplementary Material
online
2020.584731/full#supplementary-material
Frontiers in Behavioral Neuroscience | www.frontiersin.org
16
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 17
Lawrence et al.
REFERENCES
Adams, J. B., Johansen, L. J., Powell, L. D., Quig, D., and Rubin, R. A. (2011).
Gastrointestinal flora and gastrointestinal status in children with autism –
comparisons to typical children and correlation with autism severity. BMC
Gastroenterol. 11:22. doi: 10.1186/1471-230X-11-22
Adzic, M., Mitic, M., and Radojcic, M. (2017). Mitochondrial estrogen receptors
as a vulnerability factor of chronic stress and mediator of fluoxetine treatment
in female and male rat hippocampus. Brain Res. 1671, 77–84. doi: 10.1016/j.
brainres.2017.07.007
Developmental SSRIs in Voles
Bridges, R. S. (2016). Long-term alterations in neural and endocrine processes
induced by motherhood in mammals. Horm. Behav. 77, 193–203. doi: 10.1016/
j.yhbeh.2015.09.001
Bridges, R. S., and Byrnes, E. M. (2006). Reproductive experience reduces
circulating 17β-estradiol and prolactin levels during proestrus and alters
estrogen sensitivity in female rats. Endocrinology 147, 2575–2582. doi: 10.1210/
en.2005-0917
Campos, A. C., Fogaca, M. V., Aguiar, D. C., and Guimaraes, F. S. (2013). Animal
models of anxiety disorders and stress. Rev. Bras. Psiquiatr. 35, S101–S111.
doi: 10.1590/1516-4446-2013-1139
Alboni, S., van Dijk, R. M., Poggini, S., Milior, G., Perrotta, M., Drenth, T., et al.
(2017). Fluoxetine effects on molecular, cellular and behavioral endophenotypes
of depression are driven by the living environment. Mol. Psychiatry 22, 552–561.
doi: 10.1038/mp.2015.142
Carr, G. V., and Lucki, I. (2011). The role of serotonin receptor subtypes in treating
depression: a review of animal studies. Psychopharmacology 213, 265–287. doi:
10.1007/s00213-010-2097-z
Carter, C. S. (2014). Oxytocin pathways and the evolution of human behavior. Ann.
Alwan, S., Reefhuis, J., Rasmussen, S. A., and Friedman, J. M. (2011). Patterns
of antidepressant medication use among pregnant women in a United States
population. J. Clin. Pharmacol. 51, 264–270. doi: 10.1177/0091270010373928
Andalib, S., Emamhadi, M. R., Yousefzadeh-Chabok, S., Shakouri, S. K., Høilund-
Carlsen, P. F., Vafaee, M. S., et al. (2017). Maternal SSRI exposure increases the
risk of autistic offspring: a meta-analysis and systematic review. Eur. Psychiat.
45, 161–166. doi: 10.1016/j.eurpsy.2017.06.001
Anderson, G. M., Freedman, D. X., Cohen, D. J., Volkmar, F. R., Hoder, E. L.,
McPhedran, P., et al. (1987). Whole blood serotonin in autistic and normal
Subjects. J. Child Psychol. Psychiatry 28, 885–900. doi: 10.1111/j.1469-7610.
1987.tb00677.x
Andrade, S. E., Raebel, M. A., Brown, J., Lane, K., Livingston, J., Boudreau, D., et al.
(2008). Use of antidepressant medications during pregnancy: a multisite study.
Am. J. Obstet. Gynecol. 198:194.e1–15. doi: 10.1016/j.ajog.2007.07.036
Ansorge, M. S., Zhou, M., Lira, A., Hen, R., and Gingrich, J. A. (2004). Early-
life blockade of the 5-HT transporter alters emotional behavior in adult mice.
Science 306, 879–881. doi: 10.1126/science.1101678
Azmitia, E. C. (2001). Modern views on an ancient chemical: serotonin effects
on cell proliferation, maturation, and apoptosis. Brain Res. Bull. 56, 413–424.
doi: 10.1016/S0361-9230(01)00614-1
Rev. Psychol. 65, 17–39. doi: 10.1146/annurev-psych-010213-115110
Carter, C. S., and Perkeybile, A. M. (2018). The monogamy paradox: what do love
and sex have to do with it? Front. Ecol. Evol. 6:202. doi: 10.3389/fevo.2018.00202
Chaidez, V., Hansen, R. L., and Hertz-Picciotto, I. (2014). Gastrointestinal
problems in children with autism, developmental delays or typical cevelopment.
J. Autism Dev. Disord. 44, 1117–1127. doi: 10.1007/s10803-013-1973-x
Cooper, W. O., Willy, M. E., Pont, S. J., and Ray, W. A. (2007). Increasing use
of antidepressants in pregnancy. Am. J. Obstet. Gynecol. 196:544.e1–5. doi:
10.1016/j.ajog.2007.01.033
Croen, L. A., Grether, J. K., Yoshida, C. K., Odouli, R., and Hendrick, V.
(2011). Antidepressant use during pregnancy and childhood autism spectrum
disorders. Arch. Gen. Psychiatry 68, 1104–1112. doi: 10.1001/archgenpsychiatry.
2011.73
Davalos, D. B., Yadon, C. A., and Tregellas, H. C. (2012). Untreated prenatal
maternal depression and the potential risks to offspring: a review. Arch.
Womens Ment. Health 15, 1–14. doi: 10.1007/s00737-011-0251-1
De Dreu, C. K. W., Greer, L. L., Handgraaf, M. J. J., Shalvi, S., Van Kleef, G. A.,
Baas, M., et al. (2010). The neuropeptide oxytocin regulates parochial altruism
in intergroup conflict among humans. Science 328, 1408–1411. doi: 10.1126/
science.1189047
Babaev, O., Chatain, C. P., and Krueger-Burg, D. (2018). Inhibition in the amygdala
Deacon, R. (2012). Assessing burrowing, nest construction, and hoarding in mice.
anxiety circuitry. Exp. Mol. Med. 50:18. doi: 10.1038/s12276-018-0063-8
J. Vis. Exp. 59:e2607. doi: 10.3791/2607
Bales, K. L.,
and Carter, C. S.
to
oxytocin facilitates partner preferences in male prairie voles (Microtus
ochrogaster). Behav. Neurosci. 117, 854–859. doi: 10.1037/0735-7044.1
17.4.854
(2003a). Developmental
exposure
Bales, K. L., and Carter, C. S. (2003b). Sex differences and developmental effects
of oxytocin on aggression and social behavior in prairie voles (Microtus
ochrogaster). Horm. Behav. 44, 178–184. doi: 10.1016/s0018-506x(03)00154-5
Bales, K. L., Kim, A. J., Lewis-Reese, A. D., and Carter, C. S. (2004a). Both oxytocin
and vasopressin may influence alloparental behavior in male prairie voles.
Horm. Behav. 45, 354–361. doi: 10.1016/j.yhbeh.2004.01.004
Bales, K. L., Pfeifer, L. A., and Carter, C. S. (2004b). Sex differences and
developmental effects of manipulations of oxytocin on alloparenting and
anxiety in prairie voles. Dev. Psychobiol. 44, 123–131. doi: 10.1002/dev.10165
Bales, K. L., Perkeybile, A. M., Conley, O. G., Lee, M. H., Guoynes, C. D., Downing,
G. M., et al. (2013). Chronic intranasal oxytocin causes long-term impairments
in partner preference formation in male prairie voles. Biol. Psychiatry 74,
180–188. doi: 10.1016/j.biopsych.2012.08.025
Benjamini, Y., and Hochberg, Y. (1995). Controlling the false discovery rate: a
practical and powerful approach to multiple testing. J. R. Stat. Soc. 57, 289–300.
doi: 10.1111/j.2517-6161.1995.tb02031.x
Bond, C. M., Johnson, J. C., Chaudhary, V., McCarthy, E. M., McWhorter, M. L.,
and Woehrle, N. S. (2020). Perinatal fluoxetine exposure results in social
deficits and reduced monoamine oxidase gene expression in mice. Brain Res.
1727:146282. doi: 10.1016/j.brainres.2019.06.001
Boukhris, T., Sheehy, O., Mottron, L., and Bérard, A. (2016). Antidepressant use
during pregnancy and the risk of autism spectrum disorder in children. JAMA
Pediatr. 170, 117–124. doi: 10.1001/jamapediatrics.2015.3356
Boulle, F., Pawluski, J. L., Homberg, J. R., Machiels, B., Kroeze, Y., Kumar, N.,
et al. (2016). Prenatal stress and early-life exposure to fluoxetine have enduring
effects on anxiety and hippocampal BDNF gene expression in adult male
offspring. Dev. Psychobiol. 58, 427–438. doi: 10.1002/dev.21385
DeVries, A. C., and Carter, C. S. (1999). Sex differences in temporal parameters
of partner preference in prairie voles (Microtus ochrogaster). Can. J. Zool. 77,
885–889. doi: 10.1139/z99-054
El Marroun, H., White, T., Verhulst, F. C., and Tiemeier, H. (2014). Maternal use
of antidepressant or anxiolytic medication during pregnancy and childhood
neurodevelopmental outcomes: a systematic review. Eur. Child Adoles. Psych.
23, 973–992. doi: 10.1007/s00787-014-0558-3
Emiliano, A. B. F., Cruz, T., Pannoni, V., and Fudge, J. L. (2007). The
interface of oxytocin-labeled cells and serotonin transporter-containing fibers
in the primate hypothalamus: a substrate for SSRIs therapeutic effects?
Neuropsychopharmacology 32, 977–988. doi: 10.1038/sj.npp.1301206
Feir-Walsh, B. J., and Toothaker, L. E. (1974). An empirical comparison
of the Anova F-test, normal scores test, and Kruskal-Wallis test under
violation of assumptions. Edu. Psychol. Meas. 34, 789–799. doi: 10.1177/
001316447403400406
Gabriele, S., Sacco, R., and Persico, A. M. (2014). Blood serotonin levels
in autism spectrum disorder: a systematic review and meta-analysis.
Eur. Neuropsychopharmacol. 24, 919–929. doi: 10.1016/j.euroneuro.2014.
02.004
Gaskill, B. N., Gordon, C. J., Pajor, E. A., Lucas, J. R., Davis, J. K., and Garner, J. P.
(2013). Impact of nesting material on mouse body temperature and physiology.
Phys. Behav. 11, 87–95. doi: 10.1016/j.physbeh.2012.12.018
Gemmel, M., De Lacalle, S., Mort, S. C., Hill, L. A., Charlier, T. D., and Pawluski,
J. L. (2019). Perinatal fluoxetine has enduring sexually differentiated effects on
neurobehavioral outcomes related to social behaviors. Neuropharmacology 144,
70–81. doi: 10.1016/j.neuropharm.2018.10.009
Gemmel, M., Hazlett, M., Bögi, E., De Lacalle, S., Hill, L. A., Kokras, N.,
et al. (2017). Perinatal fluoxetine effects on social play, the HPA system, and
hippocampal plasticity in pre-adolescent male and female rats: interactions
with pre-gestational maternal stress. Psychoneuroendocrinology 84, 159–171.
doi: 10.1016/j.psyneuen.2017.07.480
Frontiers in Behavioral Neuroscience | www.frontiersin.org
17
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 18
Lawrence et al.
Developmental SSRIs in Voles
Gentile, S. (2015a). Early pregnancy exposure to selective serotonin reuptake
inhibitors,
structural malformations, and hypothesized
teratogenic mechanisms. Expert Opin. Drug Metab. Toxicol. 11, 1585–1597.
doi: 10.1517/17425255.2015.1063614
risks of major
Gentile, S.
(2015b). Prenatal antidepressant
exposure and the
in children. Are we looking at
132–137. doi:
risk of
the fall
10.1016/j.jad.2015.
182,
autism spectrum disorders
of Gods?
04.048
J. Affect. Disord.
Gidaya, N. B., Lee, B. K., Burstyn, I., Yudell, M., Mortensen, E. L., and Newschaffer,
C. J. (2014). In utero exposure to selective serotonin reuptake inhibitors and
risk for autism spectrum disorder. J. Autism Dev. Disord. 44, 2558–2567. doi:
10.1007/s10803-014-2128-4
Glover, M. E., and Clinton, S. M. (2016). Of rodents and humans: a comparative
review of the neurobehavioral effects of early life SSRI exposure in preclinical
and clinical research. Int. J. Dev. Neurosci. 51, 50–72. doi: 10.1016/j.ijdevneu.
2016.04.008
Greenberg, G. D., van Westerhuyzen, J. A., Bales, K. L., and Trainor, B. C. (2012).
Is it all in the family? The effects of early social structure on neural-behavioral
systems of prairie voles (Microtus ochrogaster). Neuroscience 216, 46–56. doi:
10.1016/j.neuroscience.2012.04.063
Grieb, Z. A., and Ragan, C. M. (2019). The effects of perinatal SSRI exposure on
anxious behavior and neurobiology in rodent and human offspring. Eur. J.
Neuropsychopharm. 29, 1169–1184. doi: 10.1016/j.euroneuro.2019.07.239
Grigoriadis, S., VonderPorten, E. H., Mamisashvili, L., Tomlinson, G., Dennis, C.-
L., Koren, G., et al. (2014). Prenatal exposure to antidepressants and persistent
pulmonary hypertension of the newborn: systematic review and meta-analysis.
BMJ 348:f6932. doi: 10.1136/bmj.f6932
Insel, T. R., Preston, S., and Winslow, J. T. (1995). Mating in the monogamous
male: behavioral consequences. Physiol. Behav. 57, 615–627. doi: 10.1016/0031-
9384(94)00362-9
Insel, T. R., and Shapiro, L. E. (1992). Oxytocin receptor distribution reflects social
organization in monogamous and polygamous voles. Proc. Natl. Acad. Sci.
U.S.A 89, 5981–5985.
Jacobsen, N. W., Hansen, C. H., Nellemann, C., Styrishave, B., and Halling-
Sørensen, B. (2015). Effects of selective serotonin reuptake inhibitors on three
sex steroids in two versions of the aromatase enzyme inhibition assay and in the
H295R cell assay. Toxicol. in Vitro 29, 1729–1735. doi: 10.1016/j.tiv.2015.07.005
Jarde, A., Morais, M., Kingston, D., Giallo, R., MacQueen, G. M., Giglia, L., et al.
(2016). Neonatal outcomes in women with untreated antenatal depression
compared with women without depression: a systematic review and meta-
analysis. JAMA Psychiatry 73, 826–837. doi: 10.1001/jamapsychiatry.2016.0934
Kaplan, Y. C., Keskin-Arslan, E., Acar, S., and Sozmen, K. (2017). Maternal SSRI
discontinuation, use, psychiatric disorder and the risk of autism in children:
a meta-analysis of cohort studies. Br. J. Clin. Pharmacol. 83, 2798–2806. doi:
10.1111/bcp.13382
Kiryanova, V., McAllister, B. B., and Dyck, R. H.
of developmental
(2013). Long-term
review of
literature. Dev. Neurosci. 35, 437–449. doi: 10.1159/000
to fluoxetine:
exposure
a
outcomes
the animal
355709
Knudsen, T. M., Hansen, A. V., Garne, E., and Andersen, A.-M. N. (2014).
Increased risk of severe congenital heart defects in offspring exposed to selective
serotonin-reuptake inhibitors in early pregnancy – an epidemiological study
using validated EUROCAT data. BMC Pregnancy Childbirth 14:333. doi: 10.
1186/1471-2393-14-333
Guastella, A. J., Einfeld, S. L., Gray, K. M., Rinehart, N. J., Tonge, B. J., Lambert,
T. J., et al. (2010). Intranasal oxytocin improves emotion recognition for youth
with autism spectrum disorders. Biol. Psychiatry 67, 692–694. doi: 10.1016/j.
biopsych.2009.09.020
Kobayashi, T., Matsuyama, T., Takeuchi, M., and Ito, S. (2016). Autism spectrum
disorder and prenatal exposure to selective serotonin reuptake inhibitors: a
systematic review and meta-analysis. Reprod. Toxicol. 65, 170–178. doi: 10.
1016/j.reprotox.2016.07.016
Guoynes, C. D., Simmons, T. C., Downing, G. M., Jacob, S., Solomon, M., and
Bales, K. L. (2018). Chronic intranasal oxytocin has dose-dependent effects on
central oxytocin and vasopressin systems in prairie voles (Microtus ochrogaster).
Neuroscience 369, 292–302. doi: 10.1016/j.neuroscience.2017.11.037
Hamilton, H. H., Lukefahr, S. D., and McNitt, J. L. (1997). Maternal nest quality and
its influence on litter survival and weaning performance in commercial rabbits.
J. Anim. Sci. 75, 926–933. doi: 10.2527/1997.754926x
Harding, E. C., Franks, N. P., and Wisden, W. (2019). The temperature dependence
of sleep. Front. Neurosci. 13:336. doi: 10.3389/fnins.2019.00336
Harrington, R. A., Lee, L.-C., Crum, R. M., Zimmerman, A. W., and Hertz-
Picciotto, I. (2014). Prenatal SSRI use and offspring with autism spectrum
disorder or developmental delay. Pediatrics 133, e1241–e1248. doi: 10.1542/
peds.2013-3406
Hayes, R. M., Wu, P., Shelton, R. C., Cooper, W. O., Dupont, W. D., Mitchel,
E., et al. (2012). Maternal antidepressant use and adverse outcomes: a cohort
study of 228,876 pregnancies. Am. J. Obstet. Gynecol. 207, 49.e1–49.e9. doi:
10.1016/j.ajog.2012.04.028
Hendrick, V., Stowe, Z. N., Altshuler, L. L., Hwang, S., Lee, E., and Haynes, D.
(2003). Placental passage of antidepressant medications. Am. J. Psychiatry 160,
993–996. doi: 10.1176/appi.ajp.160.5.993
Herlenius, E., and Lagercrantz, H. (2004). Development of neurotransmitter
systems during critical periods. Exp. Neurol. 190, 8–21. doi: 10.1016/j.
expneurol.2004.03.027
Hernández, V. S., Hernández, O. R., Perez de la Mora, M., Gómora, M. J.,
Fuxe, K., Eiden, L. E., et al. (2016). Hypothalamic vasopressinergic projections
innervate central amygdala GABAergic neurons: implications for anxiety and
stress coping. Front. Neural Circuit 10:92. doi: 10.3389/fncir.2016.00092
Hollander, E., Soorya, L., Chaplin, W., Anagnostou, E., Taylor, B. P., Ferretti,
C. J., et al. (2012). A double-blind placebo-controlled trial of fluoxetine for
repetitive behaviors and global severity in adult autism spectrum disorders. Am.
J. Psychiatry 169, 292–299. doi: 10.1176/appi.ajp.2011.10050764
Hranilovic, D., Bujas-Petkovic, Z., Vragovic, R., Vuk, T., Hock, K., and Jernej,
B. (2007). Hyperserotonemia in adults with autistic disorder. J. Autism Devel.
Disord. 37, 1934–1940. doi: 10.1007/s10803-006-0324-6
Hviid, A., Melbye, M., and Pasternak, B. (2013). Use of selective serotonin reuptake
inhibitors during pregnancy and risk of autism. N. Engl. J. Med. 369, 2406–2415.
doi: 10.1056/NEJMoa1301449
Kolevzon, A., Mathewson, K. A., and Hollander, E. (2006). Selective serotonin
reuptake inhibitors in autism: a review of efficacy and tolerability. J. Clin.
Psychiat. 67, 407–414. doi: 10.4088/jcp.v67n0311
Komada, M., Takao, K., and Miyakawa, T. (2008). Elevated plus maze for mice.
J. Vis. Exp. 22:1088. doi: 10.3791/1088
Kristensen, J. H., Ilett, K. F., Hackett, L. P., Yapp, P., Paech, M., and Begg, E. J.
(1999). Distribution and excretion of fluoxetine and norfluoxetine in human
milk. Br. J. Clin. Pharmacol. 48, 521–527. doi: 10.1046/j.1365-2125.1999.00040.
x
Lauder, J. M. (1993). Neurotransmitters as growth regulatory signals: role of
receptors and second messengers. Trends Neurosci. 16, 233–240. doi: 10.1016/
0166-2236(93)90162-F
Lei, K., Cushing, B. S., Musatov, S., Ogawa, S., and Kramer, K. M. (2010). Estrogen
receptor-α in the bed nucleus of the stria terminalis regulates social affiliation
in male prairie voles (Microtus ochrogaster). PLoS One 5:e8931. doi: 10.1371/
journal.pone.0008931
Lugnegård, T., Hallerbäck, M. U., and Gillberg, C. (2011). Psychiatric comorbidity
in young adults with a clinical diagnosis of Asperger syndrome. Res. Dev.
Disabil. 32, 1910–1917. doi: 10.1016/j.ridd.2011.03.025
Lupu, D., Pop, A., Cherfan, J., Kiss, B., and Loghin, F. (2015). In vitro modulation
of estrogen receptor activity by norfluoxetine. Med. Pharm. Rep. 88, 386–390.
doi: 10.15386/cjmed-476
Madden, A. M. K., and Zup, S. L.
(2014). Effects of developmental
hyperserotonemia on juvenile play behavior, oxytocin and serotonin receptor
expression in the hypothalamus are age and sex dependent. Physiol. Behav. 128,
260–269. doi: 10.1016/j.physbeh.2014.01.036
Man, K. K. C., Tong, H. H. Y., Wong, L. Y. L., Chan, E. W., Simonoff, E., and
Wong, I. C. K. (2015). Exposure to selective serotonin reuptake inhibitors
during pregnancy and risk of autism spectrum disorder in children: a systematic
review and meta-analysis of observational studies. Neurosci. Biobehav. Rev. 49,
82–89. doi: 10.1016/j.neubiorev.2014.11.020
Mann, P. E. (1993). 22- measurement of maternal behavior. Methods Neurosci. 14,
343–358. doi: 10.1016/B978-0-12-185277-1.50027-7
Martin, M. M., Liu, Y., and Wang, Z. (2012). Developmental exposure to a
serotonin agonist produces subsequent behavioral and neurochemical changes
in the adult male prairie vole. Physiol. Behav. 105, 529–535. doi: 10.1016/j.
physbeh.2011.09.015
Frontiers in Behavioral Neuroscience | www.frontiersin.org
18
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 19
Lawrence et al.
Developmental SSRIs in Voles
McDougle, C., Naylor, S. T., Cohen, D. J., Aghajanian, G. K., Heninger, G. R., and
Price, L. H. (1996). Effects of tryptophan depletion in drug-free adults with
autistic disorder. Arch. Gen. Psychiatry 53, 993–1000. doi: 10.1001/archpsyc.
1996.01830110029004
McDougle, C. J., Naylor, S. T., Goodman, W. K., Volkmar, F. R., Cohen, D. J., and
Price, L. H. (1993). Acute tryptophan depletion in autistic disorder: a controlled
case study. Biol. Psychiatry 33, 547–550. doi: 10.1016/0006-3223(93)90011-2
McElhanon, B. O., McCracken, C., Karpen, S., and Sharp, W. G. (2014).
Gastrointestinal symptoms in autism spectrum disorder: a meta-analysis.
Pediatrics 133, 872–883. doi: 10.1542/peds.2013-3995
McNamara, I. M., Borella, A. W., Bialowas, L. A., and Whitaker-Azmitia, P. M.
(2008). Further studies in the developmental hyperserotonemia model (DHS)
of autism: social, behavioral and peptide changes. Brain Res. 1189, 203–214.
doi: 10.1016/j.brainres.2007.10.063
Muller, C. L., Anacker, A. M. J., and Veenstra-VanderWeele, J. (2016). The
serotonin system in autism spectrum disorder: from biomarker to animal
models. Neuroscience 321, 24–41. doi: 10.1016/j.neuroscience.2015.11.010
Nagano, M., Liu, M., Inagaki, H., Kawada, T., and Suzuki, H. (2012). Early
intervention with fluoxetine reverses abnormalities in the serotonergic system
and behavior of rats exposed prenatally to dexamethasone. Neuropharmacology
63, 292–300. doi: 10.1016/j.neuropharm.2012.03.027
Neumann, I. D., Krömer, S. A., Toschi, N., and Ebner, K. (2000). Brain oxytocin
inhibits the (re)activity of the hypothalamo-pituitary-adrenal axis in male rats:
involvement of hypothalamic and limbic brain regions. Regul. Pept. 96, 31–38.
doi: 10.1016/s0167-0115(00)00197-x
Oberlander, T. F., Warburton, W., Misri, S., Aghajanian, J., and Hertzman, C.
(2008). Effects of timing and duration of gestational exposure to serotonin
reuptake inhibitor antidepressants: population-based study. Br. J. Psychiat. 192,
338–343. doi: 10.1192/bjp.bp.107.037101
Osei-Owusu, P.,
James, A., Crane,
(2005). 5-
Hydroxytryptamine 1A receptors in the paraventricular nucleus of
the
hypothalamus mediate oxytocin and adrenocorticotropin hormone release and
some behavioral components of the serotonin syndrome. J. Pharmacol. Exp.
Ther. 313, 1324–1330. doi: 10.1124/jpet.104.082073
J., and Scrogin, K. E.
Perkeybile, A. M., and Bales, K. L. (2015). Early rearing experience is related to
altered aggression and vasopressin production following chronic social isolation
in the prairie vole. Behav. Brain Res. 283, 37–46. doi: 10.1016/j.bbr.2015.01.025
Perkeybile, A. M., Griffin, L., and Bales, K. L. (2013). Natural variation in
early parental care correlates with social behaviors in adolescent prairie voles
(Microtus ochrogaster. Front. Behav. Neurosci 7:21. doi: 10.3389/fnbeh.2013.
00021
Perry, A. N., Carter, C. S., and Cushing, B. S. (2015). Effects of postnatal estrogen
manipulations on juvenile alloparental behavior. Horm. Behav. 75, 11–17. doi:
10.1016/j.yhbeh.2015.07.019
Peters, D. A. V. (1982). Prenatal stress: effects on brain biogenic amine and plasma
corticosterone levels. Pharmacol. Biochem. Behav. 17, 721–725. doi: 10.1016/
0091-3057(82)90353-7
Pollak, D. D., Rey, C. E., and Monje, F. J. (2010). Rodent models in depression
research: classical strategies and new directions. Ann. Med. 42, 252–264. doi:
10.3109/07853891003769957
Schain, R. J., and Freedman, D. X. (1961). Studies on 5-hydroxyindole metabolism
in autistic and other mentally retarded children. J. Pediatr. 58, 315–320. doi:
10.1016/S0022-3476(61)80261-8
Schulz, K. M., Pearson, J. N., Neeley, E. W., Berger, R., Leonard, S., Adams, C. E.,
et al. (2011). Maternal stress during pregnancy causes sex-specific alterations in
offspring memory performance, social interactions, indices of anxiety, and body
mass. Physiol. Behav. 104, 340–347. doi: 10.1016/j.physbeh.2011.02.021
Sebire, M., Elphinstone Davis, J., Hatfield, R., Winberg, S., and Katsiadaki, I. (2015).
Prozac affects stickleback nest quality without altering androgen, spiggin, or
aggression levels during a 21-day breeding test. Aqua. Toxicol. 168, 78–89.
doi: 10.1016/j.aquatox.2015.09.009
Sikander, A., Rana, S. V., and Prasad, K. K. (2009). Role of serotonin in
gastrointestinal motility and irritable bowel syndrome. Clin. Chim. Acta 403,
47–55. doi: 10.1016/j.cca.2009.01.028
Silva, A. S., Toffoli, L. V., Estrada, V. B., Veríssimo, L. F., Francis-Oliveira, J.,
Moreira, E. G., et al. (2018). Maternal exposure to fluoxetine during gestation
and lactation induces long lasting changes in the DNA methylation profile of
offspring’s brain and affects the social interaction of rat. Brain Res. Bull. 142,
409–413. doi: 10.1016/j.brainresbull.2018.09.007
Stagni, F., Giacomini, A., Guidi, S., Ciani, E., Ragazzi, E., Filonzi, M., et al.
(2015). Long-term effects of neonatal treatment with fluoxetine on cognitive
performance in Ts65Dn mice. Neurobiol. Dis. 74, 204–218. doi: 10.1016/j.nbd.
2014.12.005
Svirsky, N., Levy, S., and Avitsur, R. (2016). Prenatal exposure to selective
serotonin reuptake inhibitors (SSRI) increases aggression and modulates
maternal behavior in offspring mice. Dev. Psychobiol. 58, 71–82. doi: 10.1002/
dev.21356
Tanaka, M., Sato, A., Kasai, S., Hagino, Y., Kotajima-Murakami, H., Kashii, H., et al.
(2018). Brain hyperserotonemia causes autism-relevant social deficits in mice.
Mol. Autism 9:60. doi: 10.1186/s13229-018-0243-3
Tao, R., Ma, Z., and Auerbach, S. B. (2000). Differential effect of local infusion
of serotonin reuptake inhibitors in the raphe versus forebrain and the role of
depolarization-induced release in increased extracellular serotonin. J. Pharm.
Exp. Ther. 294, 571–579.
van Steensel, F. J. A., Bögels, S. M., and Perrin, S. (2011). Anxiety disorders in
children and adolescents with autistic spectrum disorders: a meta-analysis. Clin.
Child Fam. Psychol. Rev. 14, 302–317. doi: 10.1007/s10567-011-0097-0
Veenstra-VanderWeele, J., Muller, C. L., Iwamoto, H., Sauer, J. E., Owens, W. A.,
Shah, C. R., et al. (2012). Autism gene variant causes hyperserotonemia,
serotonin receptor hypersensitivity, social impairment and repetitive behavior.
Proc. Natl. Acad. Sci. U.S.A. 109, 5469–5474. doi: 10.1073/pnas.1112345109
Videman, M., Tokariev, A., Saikkonen, H., Stjerna, S., Heiskala, H., Mantere,
O., et al. (2017). Newborn brain function is affected by fetal exposure to
maternal serotonin reuptake inhibitors. Cereb. Cortex 27, 3208–3216. doi: 10.
1093/cercor/bhw153
Villalba, C., Boyle, P. A., Caliguri, E.
selective
serotonin
in male
Behav.
32,
J., and De Vries, G.
inhibitor
voles
reuptake
and female prairie
J.
(1997).
fluoxetine
(Microtus
10.1006/hbeh.1997.
184–191.
doi:
the
Effects
of
on social behaviors
ochrogaster). Horm.
1420
Pop, A., Lupu, D.
(2015).
Estrogenic/antiestrogenic activity of selected selective serotonin reuptake
inhibitors. Med. Pharm. Rep. 88, 381–385. doi: 10.15386/cjmed-474
J., Kiss, B., and Loghin, F.
I., Cherfan,
Weinstock, M.
(2001). Alterations induced by gestational stress in brain
morphology and behaviour of the offspring. Prog. Neurobiol. 65, 427–451. doi:
10.1016/S0301-0082(01)00018-1
Ramos, A., and Mormède, P. (1997). Stress and emotionality: a multidimensional
and genetic approach. Neurosci. Biobehav. Rev. 22, 33–57. doi: 10.1016/S0149-
7634(97)00001-8
Whitaker-Azmitia, P. M. (2001). Serotonin and brain development: role in human
developmental diseases. Brain Res. Bull. 56, 479–485. doi: 10.1016/S0361-
9230(01)00615-3
Rampono, J., Kristensen, J. H., Hackett, L. P., Paech, M., Kohan, R., and Ilett,
K. F. (2000). Citalopram and demethylcitalopram in human milk; distribution,
excretion and effects in breast fed infants. Br. J. Clin. Pharmacol. 50, 263–268.
doi: 10.1046/j.1365-2125.2000.00253.x
Whitaker-Azmitia, P. M.
(2005). Behavioral and cellular consequences of
increasing serotonergic activity during brain development: a role in
J. Dev. Neurosci. 23, 75–83. doi: 10.1016/j.ijdevneu.2004.
autism? Int.
07.022
Rampono, J., Simmer, K., Ilett, K. F., Hackett, L. P., Doherty, D. A., Elliot, R., et al.
(2009). Placental transfer of SSRI and SNRI antidepressants and effects on the
neonate. Pharmacopsychiatry 42, 95–100. doi: 10.1055/s-0028-1103296
White, S. W., Oswald, D., Ollendick, T., and Scahill, L. (2009). Anxiety in children
and adolescents with autism spectrum disorders. Clin. Psychol. Rev. 29, 216–
229. doi: 10.1016/j.cpr.2009.01.003
Roberts, R. L., Miller, A. K., Taymans, S. E., and Carter, C. S. (1998).
Role of social and endocrine factors in alloparental behavior of prairie
voles (Microtus ochrogaster). Can. J. Zool. 76, 1862–1868. doi: 10.1139/z9
8-156
Williams, J. R., Catania, K. C., and Carter, C. S. (1992). Development of partner
preferences in female prairie voles (Microtus ochrogaster): the role of social
and sexual experience. Horm. Behav. 26, 339–349. doi: 10.1016/0018-506X(92)
90004-F
Frontiers in Behavioral Neuroscience | www.frontiersin.org
19
November 2020 | Volume 14 | Article 584731
fnbeh-14-584731
November 10, 2020
Time: 15:58
# 20
Lawrence et al.
Developmental SSRIs in Voles
Wilson, C. A., and Terry, A. V. (2013). Variable maternal stress in rats alters
locomotor activity, social behavior, and recognition memory in the adult
offspring. Pharmacol. Biochem. Behav. 104, 47–61. doi: 10.1016/j.pbb.2012.
12.015
Zucker, I. (2017). Risk mitigation for children exposed to drugs during gestation:
a critical role for animal preclinical behavioral testing. Neurosci. Biobehav. Rev.
77, 107–121. doi: 10.1016/j.neubiorev.2017.03.005
Wirth, A., Holst, K., and Ponimaskin, E. (2017). How serotonin receptors regulate
morphogenic signalling in neurons. Prog. Neurobiol. 151, 35–56. doi: 10.1016/j.
pneurobio.2016.03.007
Conflict of Interest: The reviewer CH declared a shared affiliation, with no
collaboration, with one of the authors, MP, to the handling editor at the time of
review.
Yang, C.-J., Tan, H.-P., and Du, Y.-J. (2014). The developmental disruptions
of
serotonin signaling may involved in autism during early brain
development. Neuroscience 267, 1–10. doi: 10.1016/j.neuroscience.2014.
02.021
Yoshida, M., Takayanagi, Y., Inoue, K., Kimura, T., Young, L. J., Onaka, T., et al.
(2009). Evidence that oxytocin exerts anxiolytic effects via oxytocin receptor
expressed in serotonergic neurons in mice. J. Neurosci. 29, 2259–2271. doi:
10.1523/JNEUROSCI.5593-08.2009
Young, K. A., Gobrogge, K. L., Liu, Y., and Wang, Z.
insights
32:53–69.
(2011). The
from a socially monogamous
10.1016/j.yfrne.2010.
doi:
neurobiology of pair bonding:
rodent.
07.006
Front. Neuroendocrinol.
The remaining authors declare that the research was conducted in the absence of
any commercial or financial relationships that could be construed as a potential
conflict of interest.
Copyright © 2020 Lawrence, Palumbo, Freeman, Guoynes and Bales. This is an
open-access article distributed under the terms of the Creative Commons Attribution
License (CC BY). The use, distribution or reproduction in other forums is permitted,
provided the original author(s) and the copyright owner(s) are credited and that the
original publication in this journal is cited, in accordance with accepted academic
practice. No use, distribution or reproduction is permitted which does not comply
with these terms.
Frontiers in Behavioral Neuroscience | www.frontiersin.org
20
November 2020 | Volume 14 | Article 584731
| null |
10.1073_pnas.2301852120.pdf
|
Data, Materials, and Software Availability. Cryo-EM images data have
been deposited in Electron Microscopy Public Image Archive (EMPIAR) data-
base with accession number EMPIAR-11544 (https://www.ebi.ac.uk/empiar/
EMPIAR-11544/) (41).
| null |
INAUGURAL ARTICLE |
BIOPHYSICS AND COMPUTATIONAL BIOLOGY
OPEN ACCESS
Quantification of gallium cryo-FIB milling damage in biological
lamellae
Bronwyn A. Lucasa,1,2,3
and Nikolaus Grigorieff a,b,1
This contribution is part of the special series of Inaugural Articles by members of the National Academy of Sciences elected in 2021.
Contributed by Nikolaus Grigorieff; received February 1, 2023; accepted April 20, 2023; reviewed by Jürgen M. Plitzko and Elizabeth Villa
Cryogenic electron microscopy (cryo-EM) can reveal the molecular details of biologi-
cal processes in their native, cellular environment at atomic resolution. However, few
cells are sufficiently thin to permit imaging with cryo-EM. Thinning of frozen cells
to <500 nm lamellae by focused-ion-beam (FIB) milling has enabled visualization of
cellular structures with cryo-EM. FIB milling represents a significant advance over
prior approaches because of its ease of use, scalability, and lack of large-scale sam-
ple distortions. However, the amount of damage it causes to a thinned cell section
has not yet been determined. We recently described an approach for detecting and
identifying single molecules in cryo-EM images of cells using 2D template matching
(2DTM). 2DTM is sensitive to small differences between a molecular model (tem-
plate) and the detected structure (target). Here, we use 2DTM to demonstrate that
under the standard conditions used for machining lamellae of biological samples,
FIB milling introduces a layer of variable damage that extends to a depth of 60 nm
from each lamella surface. This layer of damage limits the recovery of information
for in situ structural biology. We find that the mechanism of FIB milling damage is
distinct from radiation damage during cryo-EM imaging. By accounting for both
electron scattering and FIB milling damage, we estimate that FIB milling damage
with current protocols will negate the potential improvements from lamella thinning
beyond 90 nm.
electron cryomicroscopy | template matching | ribosome | focused-ion-beam milling
Cryogenic electron microscopy (cryo-EM) has enabled visualization of purified macro-
molecular complexes at atomic resolution (1, 2). A more complete understanding of
molecular function requires visualizing their location, structure, and interactions in the
native cellular environment. The internal architecture of cells can be preserved with high
fidelity by vitrification allowing for the visualization of molecules at high resolution directly
in the cell (in situ) with cryo-EM (3). However, with few exceptions, cells are too thick
to be electron transparent and therefore require thinning.
Cryo-EM of vitreous sections (CEMOVIS) is one solution to generating thin slices of
high-pressure frozen cells using a cryo-ultramicrotome (4). However, the process requires
a skilled user, is difficult to automate, and introduces compression artifacts, which together
have limited the widespread utility of this approach (5).
Focused-ion-beam (FIB) milling is a technique in common use in materials science
that has been adapted to produce thin cell sections for in situ cryo-EM under cryogenic
conditions (6–8). In place of a physical ultramicrotome, a focused beam of ions, typically
produced from a gallium liquid metal ion source (LMIS) or plasma, is used to sputter
material above and below a thin section of the cell known as a lamella (8). FIB milling
has higher throughput relative to CEMOVIS because of its ease of use, commercial avail-
ability, and computational control allowing for automation of lamella production (9–11).
As a result, cryo-FIB milling for lamella preparation of cells has recently seen widespread
adoption and is now the predominant method for preparing cells for in situ cryo-EM
(12).
It has been demonstrated recently that it is possible to generate near-atomic resolution
reconstructions by averaging subtomograms from vitreously frozen cells (13, 14). These
successes highlight the need for a more quantitative understanding of potential sample
damage introduced during FIB milling that could limit both the resolution of in situ
reconstructions and the ability to accurately localize molecules in cells.
Organic materials are particularly sensitive to radiation damage upon interaction with
high-energy particles. Simulations of the stopping range in matter (SRIM) of ions in a
glancing incidence beam at 30 keV, the typical conditions for cryo-lamella preparation for
transmission electron microscopy (TEM), will implant Ga+ ions in frozen cells to a depth
Significance
The molecular mechanisms of
biological macromolecules and
their assemblies are often
studied using purified material.
However, the composition,
conformation, and function of
most macromolecules depend on
their cellular context, which must
be studied inside cells. Focused-
ion-beam (FIB) milling enables
cryogenic electron microscopy to
visualize macromolecules in cells
at near atomic resolution by
generating thin sections of frozen
cells. However, the extent of FIB
milling damage to frozen cells is
unknown. Here, we show that
Ga+ FIB milling introduces
damage to a depth of ~60 nm
from each lamella surface,
leading to a loss of recoverable
information of up to 20% in
100 nm samples. FIB milling with
Ga+ therefore presents both an
opportunity and an obstacle for
structural cell biology.
Author contributions: B.A.L. and N.G. designed research;
B.A.L. performed research; B.A.L. contributed new
reagents/analytic tools; B.A.L. analyzed data; B.A.L. and
N.G. interpretated results; and B.A.L. and N.G. wrote the
paper.
Reviewers: J.M.P., Max-Planck-Institut fur Biochemie; and
E.V., University of California San Diego.
The authors declare no competing interest.
Copyright © 2023 the Author(s). Published by PNAS.
This open access article is distributed under Creative
Commons Attribution License 4.0 (CC BY).
1To whom correspondence may be addressed. Email:
[email protected] or [email protected].
2Present address: Division of Biochemistry, Biophysics
and Structural Biology, Department of Molecular and
Cell Biology, University of California Berkeley, Berkeley,
CA 94720.
3Present address: Center for Computational Biology,
University of California Berkeley, Berkeley, CA 94720.
This article contains supporting information online at
https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.
2301852120/-/DCSupplemental.
Published May 22, 2023.
PNAS 2023 Vol. 120 No. 23 e2301852120
https://doi.org/10.1073/pnas.2301852120 1 of 9
of 20 to 30 nm (7, 15). After accounting for removal of ~10 nm of
material due to the concurrent milling action, the implantation
zone is anticipated to be ~5 to 20 nm from the lamella surface (7).
Cascading atomic collisions between Ga+ ions and sample atoms as
the Ga+ ions imbed in the sample will introduce additional damage
to an unknown depth from each lamella surface (16). Such damage
introduced during FIB milling would decrease the usable volume
of a lamella and could limit the resolution of in situ–determined
structures. In a recent study (17), subtomogram averaging was used
to generate high-resolution reconstructions of ribosomes taken from
varying distances from the argon plasma FIB–milled lamella surface.
To assess the argon plasma FIB damage, the B-factors affecting these
reconstructions were analyzed, showing five-fold higher B-factors
near the surface compared to 60 nm into the lamella. However, the
B-factor analysis did not separate the contribution of the subtomo-
gram alignment errors to the overall B-factors, thereby likely over-
estimating the extent of FIB damage. In our study, we set out to
quantify the degree and depth of FIB damage caused by the more
commonly used Ga+ LMIS.
We have recently described an approach, 2D template matching
(2DTM) (18), to locate molecular assemblies in three dimensions
with high precision in 2D cryo-EM images of unmilled cells (19,
20) and FIB-milled lamellae (21). Cross-correlation of a
high-resolution template generated from a molecular model with
a cryo-EM image produces a 2DTM signal-to-noise ratio (SNR)
that reflects the similarity between the template and the individual
target molecules in the image (18–21).
In the present study, we apply 2DTM to quantify target integ-
rity within FIB-milled lamellae at single-molecule resolution. We
find that Ga+ FIB milling appreciably reduces target integrity to
a depth of ~60 nm from the lamella surface. We find that the
nature of FIB milling damage is distinct from electron radiation
damage, consistent with interatomic collisions, rather than elec-
tronic interactions, being primarily responsible for the damage.
By comparing signal loss due to FIB milling damage to signal loss
in thick samples due to inelastic electron scattering and molecular
overlap, we show that recovery of structural information in
100 nm lamellae is reduced by ~20%.
Results
FIB Milling Introduces a Layer of Reduced Structural Integrity.
A 2DTM template represents an ideal, undamaged model of the
molecule to be detected. Any damage introduced during FIB
milling will therefore decrease the correlation with the undamaged
template, leading to a lower 2DTM SNR. Ribosomes are present
at high density and relatively evenly distributed in the cytoplasm
of the yeast Saccharomyces cerevisiae (21) and therefore present
an ideal 2DTM target to quantify differences in target integrity.
We prepared FIB-milled lamellae of S. cerevisiae cells of thickness
varying from 120 nm to 260 nm (Fig. 1 A and B). In 30 images
of the yeast cytoplasm from four lamellae, we located 11,030 large
ribosomal subunits (LSUs) using 2DTM (Fig. 1 C and D and
SI Appendix, Fig. S1 A and B).
We estimated the z-coordinate of each LSU relative to the image
defocus plane with 2 nm precision (Fig. 1 E and F, Materials and
Methods). We found that the LSUs were located in a slab oriented
at an angle of ~6 to 11° relative to the defocus plane, consistent
with the milling angle relative to the grid surface (Fig. 1 C–F).
The 2DTM SNRs of LSUs were noticeably lower at the edge of
the lamellae than at the center and did not correlate with defocus
(Fig. 1 E and F), indicating that this is unlikely to be the result of
defocus estimation error. We used the tilt axis and angle estimated
from the contrast transfer function (CTF) fit (21, 22), which
indicates the pretilt of the lamella introduced during milling to
adjust the coordinate frame to reflect the position of each LSU
relative to the lamella center (Fig. 1 G and H).
On average, the 2DTM SNRs were higher in the center and
lower toward the surface in all lamellae examined (Fig. 1 G and
H and SI Appendix, Fig. S2). The maximum 2DTM SNR
decreased with increasing lamella thickness (SI Appendix, Fig. S1B)
as observed previously (18–21). However, we observed a different
2DTM SNR profile as a function of z-coordinate in regions of
different thicknesses. The 2DTM SNRs in ≤ ∼ 150-nm-thick
lamellae increased toward the center of the lamella (e.g.: Fig. 1G),
while in ≥ ∼ 150-nm-thick lamellae, they reached a plateau (e.g.:
Fig. 1H). This is consistent with decreased structural integrity of
LSUs close to each lamella surface.
Quantification of the Damage Profile Reveals Damage up to
~60 nm from Each Lamella Surface. To assess the depth of the
damage, we focused on images of 200 nm lamellae because we
were able to detect targets throughout most of the volume, and
both the number and 2DTM SNRs of the detected targets reached
a plateau in the center, indicating that there is a zone of minimal
damage. In seven images of 200 nm lamellae, we calculated the
mean 2DTM SNR in bins of 10 nm from the lamella surface
and divided this by the undamaged SNR ( SNRu ), defined as the
mean 2DTM SNR of the targets between 90 and 100 nm from
the lamella surface. Both the relative 2DTM SNR (Fig. 2A) and
the number of LSUs detected (Fig. 2B) increased as a function of
distance from the lamella surface. The lower number of detected
LSUs at the lamella surface is likely a consequence of targets having
a 2DTM SNR that falls below the chosen 2DTM SNR threshold
of 7.85 at which we expect a single false positive per image (18).
The low number of targets in the 10 nm bin prevented an accurate
Gaussian fit (R2 = 0.8), and thus, this population was excluded
from further analysis. In each of the bins >60 nm from the lamella
surface (Fig. 2C), the distribution of 2DTM SNRs was Gaussian
and not significantly different from the undamaged bin (t test
P > 0.05, SI Appendix, Table S1). However, for each of the bins ≤
60 nm from the lamella surface, the distribution shifts significantly
(t test P < 0.0001, SI Appendix, Table S1) to the left, i.e., lower
SNR values (Fig. 2C). This indicates that the structural similarity
between target and template decreases closer to the lamella surface.
We interpret this as a loss of target integrity due to FIB milling
damage up to ~60 nm from the lamella surface.
We found that the change in the mean 2DTM SNR at a par-
ticular depth from the lamella surface ( d ), relative to, SNRu can
be described by an exponential decay function:
SN Rd
SN Ru
= 1 − Y 0 ⋅ e−
[1]
d
k ,
where Y 0 and k are the fit and decay constants of our model.
A least-squares fit gave values of Y 0 = 0.31 and k = 37.03 nm
(R2 = 0.99) (Fig. 2D). Since SN Rd ∕SN Ru represents the remain-
ing signal, the exponential model indicates a steep decline in dam-
age in the first ~10 to 20 nm from the lamella surface, possibly
explaining why few LSUs were detected in this range.
The observed damage profile was absent in images of unmilled
Mycoplasma pneumoniae cells, confirming that the observed pat-
tern results from FIB milling and is not a result of error in the
z-estimation in 2DTM (SI Appendix, Figs. S4 and S5).
Mechanism of FIB Milling Damage. To characterize the mechanism
of FIB milling damage, we compared its profile to the damage
introduced by exposure to electrons during cryo-EM imaging.
2 of 9 https://doi.org/10.1073/pnas.2301852120
pnas.org
A
120 nm lamella
B
200 nm lamella
C
E
1500
1000
500
0
)
Å
(
Z
-500
-1000
-1500
G
R
N
S
M
T
D
2
20
18
16
14
12
10
8
369 LSU
1.4
1.2
1.0
0.8
2222222 0000000000000000000000000000
2000
2222000000000000
0020
22
00
0022
00
00
44444400004040040400000000000000000000
4000
000000000000
4400000000000
0000000000
440000000
00000000
040
00
6
6000
8000
corrected Y (Å)
R
e
a
l
t
i
v
e
2
D
T
M
S
N
R
D
F
2000
1000
732 LSU
)
Å
(
Z
0
00000000000 00000000000000000000060000000660000066000006666666000000000000000000000000000000000000000000000000000000000000000000000000000066666666666666666666
6000
6
00000
8000
corrected Y (Å)
-1000
-2000
H
R
N
S
M
T
D
2
20
18
16
14
12
10
8
1.4
1.2
1.0
0.8
R
e
a
l
t
i
v
e
2
D
T
M
S
N
R
-1250
-1000
-750
-500
-250 0
250
500
750
1000
1250
-1250
-1000
-750
-500
-250 0
250
500
750
1000
1250
Lamella Z (Å)
Lamella Z (Å)
Fig. 1. Visualization of yeast cytoplasmic ribosomes in 3D with 2DTM. (A) An electron micrograph of the yeast cytoplasm in a 120-nm region of a lamella. Scale
bars in (A and B) represent 50 nm. (B) As in (A), showing a 200-nm lamella. (C) Significant LSUs located in 3D in the image in (A) with 2DTM. (D) As in (C), showing
the results for the image in (B). (E) Scatterplot showing a side view of the LSUs in (A). The color coding indicates the 2DTM SNR of each significant detection relative
to the mean 2DTM SNR in each image. The z-coordinate represents the position of each target relative to the microscope defocus plane. (F) As in (E), showing the
results from the image in (B). (G) Scatterplot showing the 2DTM SNR of each detected LSU in the image in (A), as a function of z-coordinate relative to the center
of the lamella. (H) As in (G), showing the z-coordinate relative to the center of the lamella of each LSU detected in the image shown in (B).
Cryo-EM imaging causes radiation damage, introducing differences
between the template and the target structure that are more
pronounced at high spatial frequencies (23). To measure radiation
damage, we generated a series of images with different exposures by
including different numbers of frames from the original movie in
the summed image. Using the locations and orientations identified
with 2DTM using a high-resolution template as above, we sought
to calculate the contribution of different spatial frequencies to the
2DTM SNR. To achieve this, we generated a series of low-pass
filtered templates with a sharp cutoff at different spatial frequencies
and calculated the change in the 2DTM SNR of each identified
LSU relative to the original high-resolution template as a function
of electron exposure relative to a 20 electrons/Å2 exposure (Fig. 3 A
and B). We find that the 2DTM SNR of templates low-pass filtered
PNAS 2023 Vol. 120 No. 23 e2301852120
https://doi.org/10.1073/pnas.2301852120 3 of 9
A
u
R
N
S
/
R
N
S
1.4
1.2
1.0
0.8
0.6
C
100
y
c
n
e
u
q
e
r
F
80
60
40
20
0
500
400
300
200
100
0
0
20
40
60
Depth (nm)
80
100
B
s
t
e
g
r
a
t
d
e
t
c
e
t
e
d
f
o
r
e
b
m
u
N
D
10 20 30 40 50 60 70 80 90 100
Depth (nm)
60
50
40
30
20
10
u
R
N
S
/
R
N
S
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0.6
0.8
1.0
SNR/SNRu
1.2
1.4
0
20
40
60
Depth (nm)
80
100
Fig. 2. The number and 2DTM SNR values of detected LSUs increase as a function of distance from the lamella surface. (A) Boxplot showing the 2DTM SNR of
LSUs at the indicated lamella depths, relative to the undamaged SNR ( SNRu ) in each image from 200 nm lamellae. Boxes represent the interquartile range (IQR),
middle lines indicate the median, whiskers represent 1.5× IQR, and dots represent values outside of this range. (B) Scatterplot showing the number of detected
targets in the indicated z-coordinate bins. (C) Gaussian fits to the distribution of 2DTM SNRs for LSUs identified in z-coordinate bins of 10 nm. Red indicates
populations with means significantly different from the mean in the center of the lamella. Blue indicates that the mean in a bin is not significantly different from
the mean in the lamella center. Fitting statistics are indicated in SI Appendix, Table S1. (D) Scatterplot showing the mean change in 2DTM SNR relative to SNRu
at the indicated depths relative to the lamella surface estimated from the Gaussian fits in (C). The line shows the exponential fit (R2 = 0.99). Error bars indicate
the SD from the Gaussian fit.
to between 1/10 and 1/7 Å−1 increases with increasing exposure.
The 2DTM SNRs of templates low-pass filtered with a cutoff at
higher resolutions begin to decrease with increasing exposure
(Fig. 3 A and B). Templates filtered to 1/5 Å−1 have a maximum
2DTM SNR at 32 electrons/Å2, while templates filtered to 1/3 Å−1
have a maximum 2DTM SNR at 28 electrons/Å2 (Fig. 3B).
To estimate the extent of FIB milling damage on different spatial
frequencies, we binned detected targets by lamella depth and calcu-
lated SNRd ∕SN Ru . We found that for templates filtered to < 1/5 Å−1,
SNRd ∕SN Ru fluctuated for targets detected further from the lamella
center. This is likely due to differences in the defocus position that
result in some of the targets having weak contrast (contrast transfer
function close to zero) and therefore not contributing meaningful
signal at different spatial frequencies relative to targets in the center
of the lamella. For templates filtered to > 1/5 Å−1, the profile was
similar between the different bins and approximately constant across
spatial frequencies (Fig. 3 C and D). This is consistent with a model
in which the FIB-damaged targets have effectively lost a fraction of
their structure, compared to undamaged targets, possibly due to
displacement of a subset of atoms by colliding ions.
Radiation damage of nucleic acids has been well documented
with one of the most labile bonds being the phosphodiester bond
in the nucleic acid backbone (24) (Fig. 3E). We observed an accel-
erated loss of signal from phosphorous atoms relative to the aver-
age loss of signal for the whole template as a function of electron
exposure (Fig. 3F). This is consistent with the phosphorous atoms
being more mobile due to breakage of phosphodiester linkages in
response to electron exposure. We did not observe a consistent
difference in the accelerated loss of signal from phosphorous in
the lamella z-coordinate groups (Fig. 3F). This indicates that the
mechanism for FIB milling damage is distinct from the radiation
damage observed during cryo-EM imaging.
Sample Thickness Limits 2DTM SNR More Than FIB Milling
Damage. Above we report that using the most common protocol
for cryo-lamella generation by LMIS Ga+ FIB milling introduces
a variable layer of damage up to 60 nm from each lamella surface.
Lamellae for cryo-EM and electron cryotomography (cryo-ET)
are typically milled to 100 to 300 nm, meaning that the damaged
layer comprises 50 to 100% of the volume. Thicker lamellae will
have a lower proportion of damaged particles. However, thicker
lamellae will also suffer from signal loss due to the increased loss
of electrons due to inelastic scattering and scattering outside the
aperture, as well as the increased number of other molecules in the
sample contributing to the background in the images. For a target
inside a cell, the loss of 2DTM SNR with increasing thickness has
been estimated as (19):
SN Rt
SN R0
= e−t ∕𝜆SNR ,
[2]
where t denotes the sample thickness, SN R0 is the 2DTM SNR in
the limit of an infinitesimally thin sample, and the decay constant
𝜆SNR = 426 nm. Optimal milling conditions for high-resolution
imaging of FIB-milled lamellae will therefore need to strike a
balance between lamella thickness and FIB damage.
To assess the relative impact of these two factors on target detec-
tion with 2DTM, we plotted the proportional loss in signal due to
4 of 9 https://doi.org/10.1073/pnas.2301852120
pnas.org
A
1.08
Exposure (e/Å2)
20
B
e
0
2
R
N
S
/
R
N
S
M
T
D
2
22
24
26
28
30
32
34
36
0.0
0.1
0.2
0.3
0.4
0.5
Low-pass Filter (1/Å)
Depth (nm)
D
1.06
e
0
2
1.04
1.02
1.00
0.98
1.05
1.00
0.95
0.90
0.85
0.80
R
N
S
/
R
N
S
M
T
D
2
C
u
R
N
S
R
N
S
/
E
0.0
0.1
0.2
0.3
0.4
0.5
Low-pass Filter 1/Å
F
R
N
S
f
o
n
o
i
t
r
o
p
o
r
P
s
u
o
r
o
h
p
s
o
h
P
m
o
r
f
0.04
0.03
0.02
0.01
0.00
15
1.08
1.06
1.04
1.02
1.00
0.98
1.05
1.00
0.95
0.90
0.85
0.80
2.12 Å
3 Å
5 Å
6.25 Å
10 Å
25
35
20
Exposure (electrons/Å2)
30
40
2.12 Å
3 Å
5 Å
6.25 Å
10 Å
0
20
40
60
Depth (nm)
80
100
Depth (nm)
u
R
N
S
R
N
S
/
20
30
40
50
60
70
80
90
100
20
30
40
50
60
70
80
90
100
0
10
20
30
40
Exposure (electrons/Å2)
Fig. 3. The mechanism of FIB milling damage is distinct from radiation damage during cryo-EM imaging. (A) Plot showing the change in 2DTM SNR with the
template low-pass filtered to the indicated spatial frequency in images collected with the indicated number of electrons/Å2 relative to images collected with
20 electrons/Å2. (B) Plot showing the change in 2DTM SNR as a function of electron exposure of templates low-pass filtered to the indicated spatial frequency.
(C) As in (A), showing the change in the 2DTM SNR in the indicated lamella z-coordinate bins relative to the SNR in the undamaged bin ( SNRu ). (D) Plot showing
the change in 2DTM SNR for templates low-pass filtered to the indicated spatial frequencies as a function of lamella z-coordinate bins. (E) Diagram showing a
segment of an RNA strand of two nucleotides. The blue circle designates the phosphate; the two red arrows indicate the location of the backbone phosphodiester
bonds. (F) Plot showing the relative contribution of template phosphorous atoms to the 2DTM SNR relative to the full-length template at the indicated exposure
without dose weighting, calculated using Eq. 6.
electrons lost in the image and background (Fig. 4, red curve). We
can estimate the average loss of 2DTM SNR, SN Rd ∕SN Ru , due
to FIB milling damage from the product of the loss (Eq. 1) from
both surfaces:
SN Rd
SN Ru
1 − ed −t ∕k
(
1 − e−d ∕k
𝛿d .
)
1
t ∫
[3]
=
)
0 (
⋅
t
Combining these two sources of signal loss gives the expected
overall 2DTM SNR as a function of sample thickness (Fig. 4,
black curve):
SN Rd
SN Ru
=
(∫
t
0 (
1 − e−d ∕k
⋅
)
1 − ed −t ∕k
t
(
𝛿d
)
e−t ∕𝜆SNR
)
[4]
.
This model predicts that in samples thicker than ~90 nm, the
relative loss in the signal due to the loss of electrons contributing
to the image, as well as molecular overlap, is greater than the
relative change due to FIB milling damage (Fig. 4A). In lamellae
thinner than 90 nm, however, FIB milling damage will dominate
and negate any benefit from further thinning. The difference
between the expected signal loss given by Eq. 4) and signal loss
solely from lost electrons and molecular overlap represents the
potential gain if FIB milling damage could be avoided. Without
FIB damage, the potential improvement in 2DTM SNR would
be between ~10% in 200 nm lamellae and ~20% in 100 nm
lamellae (Fig. 4). The model in Eq. 4) ignores the variable degree
of damage expected to occur across LSUs that we used as probes
to measure damage and that have a radius of ~15 nm. However,
the resulting error in the measured damage constant k (Eq. 1) is
PNAS 2023 Vol. 120 No. 23 e2301852120
https://doi.org/10.1073/pnas.2301852120 5 of 9
R
N
S
M
T
D
2
e
v
i
t
a
l
e
R
1.0
0.8
0.6
0.4
0
50
FIB damage
Electron scattering
Expected relative signal
100
150
Lamella thickness (nm)
200
250
300
Fig. 4. Signal loss due to increased inelastic and multiple electron scattering
in thicker samples outweighs the effect of FIB damage on LSU 2DTM SNRs.
Plot showing the expected signal recovery in lamellae of indicated thickness
as a function of signal loss due to electron scattering (red curve), FIB damage
(blue curve), and their product (black curve).
expected to be small since k (~37 nm) significantly exceeds the
LSU radius, and hence, the variable damage can be approximated
by an average damage uniformly distributed across the target.
We also expect that the number of detected targets will be
reduced by FIB milling damage. The number of detected LSUs
was variable across lamellae, likely due to biological differences in
local ribosome concentration. In undamaged parts of a subset of
200-nm-thick lamellae, we identified ~425 LSU in z-coordinate
intervals of 10 nm. If this density were maintained throughout
the lamella, we would expect to detect ~40% more targets in these
lamellae.
We conclude that FIB damage reduces the number and integrity
of detected targets but that signal loss due to electrons lost to the
image, as well as background from overlapping molecules, is a
greater limiting factor for target detection and characterization
with 2DTM than FIB milling damage in lamellae thicker than
~90 nm. These data agree with other empirical observations that
thinner lamellae are optimal for recovery of structural information
and generation of high-resolution reconstructions.
It may be possible to restore signal in images otherwise lost to
inelastic scattering using Cc-correctors (25). This would be par-
ticularly impactful for thick samples such as FIB-milled cellular
lamellae. With the use of a Cc-corrector, the signal loss in thick
samples would be reduced, and FIB milling damage may become
the main limiting factor for in situ structural biology.
Discussion
Ga+ LMIS FIB milling is currently the preferred method for gen-
erating thin, electron-transparent cell sections for in situ cryo-EM.
We use 2DTM to evaluate the structural integrity of macromol-
ecules in FIB-milled lamellae and provide evidence that FIB-
milled lamellae have a region of structural damage to a depth of
up to 60 nm from the lamella surface. By evaluating the relative
similarity of a target molecule to a template model, 2DTM pro-
vides a sensitive, highly position-specific, single-particle evaluation
of sample integrity.
2DTM SNRs Provide a Readout of Sample Integrity and Image
Quality. Changes to the 2DTM SNR provide a readout of the
relative similarity of a target molecule to a given template. We
have previously shown that relative 2DTM SNRs discriminate
between molecular states and can reveal target identity (20, 21).
In this study, we show that changes in 2DTM SNRs can also
reflect damage introduced during FIB milling and radiation
damage introduced during cryo-EM imaging. A previous
attempt to measure FIB damage has relied on visual changes
in the sample near the surface. These changes are difficult
to quantify in terms of damage, and they could in part be
caused by other mechanisms such as ice accumulation after
milling (26). Argon plasma FIB damage has been assessed by
comparing subtomogram averages of particles from different
distances from the lamella surface and estimating their B-factors
(17), which may overestimate the amount of damage due to
unrelated contributions to the measured B-factors. The 2DTM
SNR represents an alternative, more quantitative metric to assess
sample integrity.
2DTM SNRs have also been used as a metric to assess image
quality (27) and the fidelity of simulations (28). 2DTM, therefore,
represents a sensitive, quantitative, and versatile method to meas-
ure the dependence of data quality on sample preparation and
data collection strategies, as well as new hardware technologies
and image processing pipelines. Tool and method developers could
use standard datasets and 2DTM to rapidly and quantitatively
assess how any changes to a pipeline affect data quality.
Estimating Errors in z-Coordinates and Thickness. The z-coordinates
of each LSU were determined by modulating the template with a
CTF corresponding to a range of defoci and identifying the defocus
at which the cross-correlation with the 2D projection image was
maximized (18). This quantification relies on an accurate estimate
of defocus. The error in the z-coordinates determined this way was
estimated to be about 60 Å (20). However, it is unlikely that these
errors explain the observed decrease in 2DTM SNRs of LSUs near
the edge of the lamellae because 1) the reduction in 2DTM SNRs
correlates strongly with the z-coordinate within the lamella, and 2),
we did not observe a consistent decrease in the number of detected
LSUs (SI Appendix, Fig. S4A) or their 2DTM SNRs (SI Appendix,
Fig. S4B) as a function of z-coordinate in images of unmilled M.
pneumoniae cells (20). It remains possible that differences in cell
density, residual motion (20) or differences in the size and resolution
of the LSUs could contribute to the differences in the profile of
2DTM SNRs as a function of z-coordinate. In the future, it may
be informative to examine the 2DTM SNRs of ribosomes and
other complexes in other thin samples such as the extensions of
mammalian cells.
Undulations at the lamella surface caused by curtaining or other
milling artifacts could contribute to the reduced number of ribo-
somes detected near the lamella surface. We aimed to minimize
the effect of curtaining in our analysis by calculating the lamella
thickness in 120 × 120 pixel (127.2 × 127.2 Å) patches across an
image and limiting our analysis to images with a thickness stand-
ard deviation (SD) of less than 20 nm. The curtaining on the
remaining lamellae cannot account for the reduced particle integ-
rity toward the lamella surface.
Possible Mechanisms of FIB Milling Damage. We find evidence
for FIB milling damage consistent with an exponential decay
of the amount of damage as a function of distance from the
lamella surface, as measured by the 2DTM SNR. Unlike electron
radiation damage, FIB damage 1) causes a reduction in the total
signal and does not preferentially affect higher spatial frequencies
contributing to the 2DTM SNR calculation, and 2), unlike
electron beam radiation damage, it does not preferentially affect
the phosphodiester bond in the RNA backbone. This suggests
that different mechanisms are responsible for the damage caused
by high-energy electrons and ions.
6 of 9 https://doi.org/10.1073/pnas.2301852120
pnas.org
At the energy ranges used for FIB milling, the interactions
between the bombarding ion and sample atoms can be modeled
as a cascade of atom displacements resulting from the transfer of
momentum from the incident Ga+ ions to the sample atoms (16).
Atoms involved in the cascade will be displaced, while the position
of other atoms will not change. This is consistent with our obser-
vation that FIB damage decreases the LSU target signal overall
without changing the relative contribution from different spatial
frequencies. Further study is required to test this hypothesis and
investigate the mechanism of FIB milling damage in more detail.
SRIM simulations predict implantation of Ga+ up to ~25 nm
into the sample (7, 15). This implies that the damage deeper in
the sample is caused by secondary effects, possibly reflecting dis-
placed sample atoms that were part of the collision cascade. We
observe a different pattern of particles within 20 nm of the lamella
surface (Fig. 3 C and D). One possible explanation is that
implanted Ga+ ions cause additional damage. However, SRIM
simulations cannot account for the full intensity profile of a Ga+
beam, and poorly match with experiment especially at low beam
currents (29). Moreover, the use of a protective organoplatinum
layer during FIB milling, as done in our experiments, will further
change the effective profile of the beam acting on the sample (30).
Further work is required to connect the quantification of particle
integrity with the implantation of Ga+ ions during biological
lamella preparation.
Implications for Generating High-Resolution Reconstructions
from FIB-Milled Samples. We have shown that particles on the
edge of a lamella have reduced structural integrity relative to
particles near the center of the lamella (Figs. 1 and 2). We
found that FIB milling damage reduces the total 2DTM
SNR. At distances >20 nm from the lamella surface the rate of
signal loss is similar at different spatial frequencies, in contrast
to radiation damage during cryo-EM imaging (Fig. 3). The
practical implication of this finding is that particles >30 nm
from the lamella surface can be included during subtomogram
averaging without negatively affecting the resolution of the
reconstruction, provided that they can be accurately aligned.
We also predict that more particles will be required relative to
unmilled samples. This is consistent with the observation that
more particles <30 nm from the lamella surface are required to
achieve the same resolution relative to >30 nm from the lamella
surface from argon plasma FIB–milled lamellae (17). The depth
at which particle quality is noticeably poorer is consistent
between gallium and argon FIB–milled samples. This suggests
that argon plasma FIB milling is not a solution to mitigate the
damage introduced during gallium FIB milling.
Due to the small number of particles detected within 10 nm
of the lamella surface, these particles were not examined in more
detail. Since ribosomes are ~25 nm in diameter, it is likely that
these particles are more severely damaged compared to particles
further away from the surface. 2DTM relies on high-resolution
signal and therefore excludes more severely damaged particles that
may be included using a low-resolution template matching
approach, such as 3D template matching used typically to identify
particles for subtomogram averaging. We therefore advise exclud-
ing particles detected within 10 nm of the lamella surface.
Alternate Methods for the Preparation of Thin Cell Sections.
FIB damage reduces both the number of detected targets and the
available signal per target. However, the damaged volume still
contributes to the sample thickness, reducing the usable signal by
10 to 20% in lamellae of typical thicknesses (Fig. 4). Therefore, it
would be advantageous to explore other strategies for cell thinning.
Plasma FIBs allow different ions to be used for milling, and this
may change the damage profile (31). Larger atoms such as xenon
will have a higher sputtering yield and may result in reduced lamella
damage, as has been demonstrated for milling of silicon samples
(32, 33). The 2DTM-based approach described here provides a
straightforward way to quantify the relative damaging effects of dif-
ferent ion species by generating curves as shown in Fig. 4.
CEMOVIS generates thin sections using a diamond knife
rather than high-energy ions and would therefore not introduce
radiation damage (4). It is unclear how the large-scale compression
artifacts introduced by this method affect particle integrity (5).
CEMOVIS has the additional benefit of being able to generate
multiple sections per cell and thereby enable serial imaging of
larger cell volumes. If the compression artifacts are unevenly dis-
tributed throughout a section, leaving some regions undistorted,
automation could make CEMOVIS a viable strategy for structural
cell biology in the future.
To retain the benefits of fast, reliable, high-throughput lamella
generation with cryo-FIB milling, strategies to remove the damaged
layer should be explored. In the Ga+ FIB, there are two properties
that are easily tunable, the beam current, which affects the rate of
ions to which the sample is exposed, and the energy of the ion beam.
Lowering the current and the total exposure is unlikely to decrease
the damage layer when milling thick samples because 1) there will
be a minimum number of collisions required to sputter a sufficiently
large volume to generate a lamella and 2) because the total exposure
will greatly exceed the steady-state dose at which implantation of
ions into the sample and sputtering are at equilibrium, such that any
additional exposure will not cause additional damage. Consistently,
we observe damage throughout the lamella and do not observe dra-
matic increases in the damage layer close to the milling edge or when
the organo-Pt layer is compromised relative to images collected fur-
ther from the milling edge, which have been exposed to a lower dose
(SI Appendix, Fig. S6). Alternately polishing the final ~50 nm from
each lamella surface with a low energy (~5 kV) beam, which has the
advantage of being easily implementable using the current configu-
ration of most cryo-FIB-SEMs, would be expected to decrease the
damage layer.
Materials and Methods
Yeast Culture and Freezing. S. cerevisiae strain BY4741 (ATCC) colonies were
inoculated in 20 mL of yeast extract–peptone–dextrose (YPD) media, diluted 1/5,
and grown overnight at 30 °C with shaking to mid-log phase. The cells were then
diluted to 10,000 cells/mL, treated with 10 µg/mL cycloheximide (Sigma) for
10 min at 30 °C with shaking. 3 µL were applied to a 2/1 or 2/2 Quantifoil 200
mesh SiO2 Cu grid, allowed to rest for 15 s, back side blotted for 8 s at 27 °C, 95%
humidity, and plunge-frozen in liquid ethane at –184 °C using a Leica EM GP2
plunger. Frozen grids were stored in liquid nitrogen until FIB milled.
FIB Milling. Grids were transferred to an Aquilos 2 cryo-FIB/SEM, sputter coated
with metallic Pt for 10 s and then coated with organo-Pt for 30 s and milled in a
series of sequential milling steps using a 30 kV Ga+ LMIS beam using the follow-
ing protocol: rough milling 1: 0.1 nA rough milling 2: 50 pA lamella polishing:
10 pA at a stage tilt of 15° (milling angle of 8°) or 18° (milling angle of 11°).
Over and under tilt of 1° was used to generate lamellae of relatively consistent
thickness during the 50 pA milling steps. No SEM imaging was performed after
the milling started to prevent introducing additional damage.
Cryo-EM Data Collection and Image Processing. Cryo-EM data were collected
following the protocol described in ref. 21 using a Thermo Fisher Krios 300 kV
electron microscope equipped with a Gatan K3 direct detector and Gatan energy
filter with a slit width of 20 eV at a nominal magnification of 81,000× (pixel size of
1.06 Å2) and a 100-µm objective aperture. Movies were collected at an exposure
rate of 1 e−/Å2/frame to a total dose of 50 e−/Å2 (dataset 1) with correlated double
sampling using the microscope control software SerialEM (34).
PNAS 2023 Vol. 120 No. 23 e2301852120
https://doi.org/10.1073/pnas.2301852120 7 of 9
Images were processed as described previously (21). Briefly, movie frames
were aligned using the program unblur (35) in the cisTEM graphical user
interface (GUI) (36) with or without dose weighting using the default param-
eters where indicated in the text. Defocus, astigmatism, and sample pretilt
were estimated using a modified version of CTFFIND4 (20, 22) in the cisTEM
GUI (36). Images of the cytoplasm were identified visually for further analysis.
Images visually containing organelles were excluded. Images of 3D densities
and 2DTM results were prepared in ChimeraX (37).
2DTM. The atomic coordinates corresponding to the yeast LSU from the Protein Data
Bank (PDB), code 6Q8Y (38) were used to generate a 3D volume using the cisTEM
program simulate (28) and custom scripts as in ref. (21). 2DTM was performed using
the program match_template (20) in the cisTEM GUI (36) using an in-plane search
step of 1.5° and an out-of-plane search step of 2.5°. Significant targets were defined
as described in ref. (20) and based on the significance criterion described in ref. (18).
The coordinates were refined using the program refine_template (20) in rotational
steps of 0.1° and a defocus range of 200 Å with a 20 Å step (2 nm z-precision). The
template volume was placed in the identified locations and orientations using the
program make_template_result (20) and visualized with ChimeraX (37).
To generate the results in Fig. 3 A–D, we applied a series of sharp low-pass fil-
ters in steps of 0.01 Å−1 to the template using the e2proc3d.py function in EMAN2
(39). We used the locations and orientations from the refined 2DTM search with
the full-length template to recalculate the 2DTM SNR with each modified template
using the program refine_template (20) by keeping the positions and orien-
tations fixed. The normalized cross-correlation was determined by dividing the
SNR calculated with each low-pass filtered template to the SNR of the full-length
template for each target.
Calculation of Pretilt and Coordinate Transform. We used Python scripts
to extract the rotation angle and pretilt from the cisTEM (36) database gener-
ated using the tilt-enabled version of the program CTFFIND4 (21, 22), perform a
coordinate transform to convert the 2DTM coordinates to the lamella coordinate
frame, and plot the 2DTM SNR as a function of lamella z-coordinate.
Calculation of Sample Thickness and Depth. We estimated the lamella
thickness per image by first summing the movie frames without dose weighting
using the EMAN2 program, alignframes (39), and then calculating the average
intensity of a sliding box of 120 × 120 pixels ( I ) relative to the same area of an
image collected over vacuum ( Io ). We then used the mean free path for electron
scattering ( 𝜆 ) of 283 nm (19) to estimate the local sample thickness ti using the
Beer-Lambert law (40):
The sample thickness was determined by taking the mean across the image. Only
images with a SD of <20 nm across the image were included for estimation of
the damage profile (Fig. 2B). The depth of each LSU relative to the lamella surface
was calculated by assuming that the LSUs are evenly distributed in z and defining
the median lamella z-coordinate as the lamella center (e.g.: Fig. 1 G and H and
SI Appendix, Fig. S2).
Measuring Change in Signal with Electron Exposure. We compared the
change in the 2DTM SNR of each individual LSU as a function of electron exposure
at different positions relative to the edge of the lamella in bins of 10 nm. We used
the locations and orientations of LSUs identified in dose-filtered images exposed
to 50 e−/Å2 to assess the correlation at the same locations and orientations in
different numbers of unweighted frames corresponding to total exposures of
8-36 e−/Å2.
To calculate the relative contribution of phosphorous to the 2DTM SNR, all
phosphorous atoms in the PDB file were deleted, and a template was generated
as described above without recentering so that it aligned with the full-length
template. We used the locations and orientations from the refined 2DTM search
with the full-length template for each exposure to calculate the 2DTM SNR with
the template lacking phosphorous ( SNRΔP ) using the program refine_template
(20) and keeping the positions and orientations fixed. The relative contribution
of phosphorous atoms to the 2DTM SNR ( SNRP ) at each exposure was calculated
using the following equation:
SNRP = 1 −
SNR
ΔP
SNRFL
.
[6]
Data, Materials, and Software Availability. Cryo-EM images data have
been deposited in Electron Microscopy Public Image Archive (EMPIAR) data-
base with accession number EMPIAR-11544 (https://www.ebi.ac.uk/empiar/
EMPIAR-11544/) (41).
ACKNOWLEDGMENTS. We thank Johannes Elferich, Ximena Zottig, and other
members of the Grigorieff lab (University of Massachusetts Chan Medical School),
Russo lab (Medical Research Council Laboratory of Molecular Biology), and de
Marco lab (Monash University) for helpful discussions. We are also grateful for
the use of and support from the cryo-EM facilities at Janelia Research Campus
and UMass Chan Medical School. B.A.L. and N.G. gratefully acknowledge funding
from the Chan Zuckerberg Initiative, grant #2021-234617 (5022).
ti = − ln (
I
Io
)𝜆.
[5]
Author affiliations: aRNA Therapeutics Institute, University of Massachusetts Chan
Medical School, Worcester, MA 01605; and bHHMI, University of Massachusetts Chan
Medical School, Worcester, MA 01605
1.
K. M. Yip, N. Fischer, E. Paknia, A. Chari, H. Stark, Atomic-resolution protein structure determination
by cryo-EM. Nature 587, 157–161 (2020).
2.
T. Nakane et al., Single-particle cryo-EM at atomic resolution. Nature 587, 152–156 (2020).
3. W. Baumeister, R. Grimm, J. Walz, Electron tomography of molecules and cells. Trends Cell Biol. 9,
4.
5.
81–85 (1999).
A. Al-Amoudi, L. P. O. Norlen, J. Dubochet, Cryo-electron microscopy of vitreous sections of native
biological cells and tissues. J. Struct. Biol. 148, 131–135 (2004).
A. Al-Amoudi, D. Studer, J. Dubochet, Cutting artefacts and cutting process in vitreous sections for
cryo-electron microscopy. J. Struct. Biol. 150, 109–121 (2005).
6. M. Marko, C. Hsieh, W. Moberlychan, C. A. Mannella, J. Frank, Focused ion beam milling of vitreous
water: Prospects for an alternative to cryo-ultramicrotomy of frozen-hydrated biological samples. J.
Microsc. 222, 42–47 (2006).
7. M. Marko, C. Hsieh, R. Schalek, J. Frank, C. Mannella, Focused-ion-beam thinning of frozen-hydrated
8.
9.
biological specimens for cryo-electron microscopy. Nat. Methods 4, 215–217 (2007).
A. Rigort et al., Focused ion beam micromachining of eukaryotic cells for cryoelectron tomography.
Proc. Natl. Acad. Sci. U.S.A. 109, 4449–4454 (2012).
G. Buckley et al., Automated cryo-lamella preparation for high-throughput in-situ structural biology.
J. Struct. Biol. 210, 107488 (2020).
14. B. A. Himes, P. Zhang, emClarity: Software for high-resolution cryo-electron tomography and
subtomogram averaging. Nat. Methods 15, 955–961 (2018).
15. Y. Fukuda, A. Leis, A. Rigort, “Preparation of vitrified cells for TEM by cryo-FIB microscopy” in
Biological Field Emission Scanning Electron Microscopy (John Wiley & Sons, 2019), pp. 415–438
16. L. A. Giannuzzi, B. I. Prenitzer, B. W. Kempshall, “Ion - solid interactions” in Introduction to Focused
Ion Beams (2005), pp. 13–52.
17. C. Berger et al., Plasma FIB milling for the determination of structures in situ. Nat. Commun. 114,
1–12 (2023).
18. J. P. Rickgauer, N. Grigorieff, W. Denk, Single-protein detection in crowded molecular environments
in cryo-EM images. Elife 6, e25648 (2017).
19. J. P. Rickgauer, H. Choi, J. Lippincott-Schwartz, W. Denk, Label-free single-instance protein detection
in vitrified cells. bioRxiv [Preprint] (2020). https://doi.org/10.1101/2020.04.22.053868 (Accessed
24 April 2020).
20. B. A. Lucas et al., Locating macromolecular assemblies in cells by 2D template matching with
cisTEM. Elife 10, e68946 (2021).
21. B. A. Lucas, K. Zhang, S. Loerch, N. Grigorieff, In situ single particle classification reveals distinct 60S
maturation intermediates in cells. Elife 11, e68946 (2022).
22. A. Rohou, N. Grigorieff, CTFFIND4: Fast and accurate defocus estimation from electron micrographs.
10. T. Zachs et al., Fully automated, sequential focused ion beam milling for cryo-electron tomography.
J. Struct. Biol. 192, 216–221 (2015).
Elife 9, e52286 (2020).
23. L. A. Baker, J. L. Rubinstein, Radiation damage in electron cryomicroscopy. Methods Enzymol. 481,
11. S. Klumpe et al., A modular platform for automated cryo-FIB workflows. Elife 10, 70506 (2021).
12. F. R. Wagner et al., Preparing samples from whole cells using focused-ion-beam milling for cryo-
371–388 (2010).
24. J. F. Ward, Molecular mechanisms of radiation-induced damage to nucleic acids. Adv. Radiat. Biol. 5,
electron tomography. Nat. Protocols 615, 2041–2070 (2020).
181–239 (1975).
13. D. Tegunov, L. Xue, C. Dienemann, P. Cramer, J. Mahamid, Multi-particle cryo-EM refinement with M
25. J. L. Dickerson, C. J. Russo, Phase contrast imaging with inelastically scattered electrons from any
visualizes ribosome-antibiotic complex at 3.5 Å in cells. Nat. Methods 18, 186–193 (2021).
layer of a thick specimen. Ultramicroscopy 237, 113511 (2022).
8 of 9 https://doi.org/10.1073/pnas.2301852120
pnas.org
26. C. J. Russo, J. L. Dickerson, K. Naydenova, Cryomicroscopy in situ: What is the smallest
34. D. N. Mastronarde, Automated electron microscope tomography using robust prediction of
molecule that can be directly identified without labels in a cell? Faraday Discuss 240, 277–302
(2022).
specimen movements. J. Struct. Biol. 152, 36–51 (2005).
35. T. Grant, N. Grigorieff, Measuring the optimal exposure for single particle cryo-EM using a 2.6 Å
27. J. Elferich, G. Schiroli, D. T. Scadden, N. Grigorieff, Defocus corrected large area cryo-EM (DeCo-LACE)
reconstruction of rotavirus VP6. Elife 4, e06980 (2015).
for label-free detection of molecules across entire cell sections. Elife 11, e80980 (2022).
28. B. Himes, N. Grigorieff, Cryo-TEM simulations of amorphous radiation-sensitive samples using
36. T. Grant, A. Rohou, N. Grigorieff, CisTEM, user-friendly software for single-particle image processing.
Elife 7, e35383 (2018).
multislice wave propagation. IUCrJ 8, 943–953 (2021).
37. E. F. Pettersen et al., UCSF ChimeraX: Structure visualization for researchers, educators, and
29. Y. Greenzweig, Y. Drezner, S. Tan, R. H. Livengood, A. Raveh, Current density profile characterization
developers. Protein Sci. 30, 70 (2021).
and analysis method for focused ion beam. Microelectron. Eng. 155, 19–24 (2016).
38. P. Tesina et al., Structure of the 80S ribosome–Xrn1 nuclease complex. Nat. Struct. Mol. Biol. 26,
30. M. Schaffer et al., Optimized cryo-focused ion beam sample preparation aimed at in situ structural
275–280 (2019).
studies of membrane proteins. J. Struct. Biol 197, 73–82 (2017).
39. G. Tang et al., EMAN2: An extensible image processing suite for electron microscopy. J. Struct. Biol.
31. G. Sergey et al., Oxygen plasma focused ion beam scanning electron microscopy for biological
157, 38–46 (2007).
samples. bioRxiv [Preprint] (2018). https://doi.org/10.1101/457820 (Accessed 31 October 2018).
40. W. J. Rice et al., Routine determination of ice thickness for cryo-EM grids. J. Struct. Biol. 204, 38
32. L. A. Giannuzzi, N. S. Smith, TEM specimen preparation with plasma FIB Xe + Ions Microsc.
(2018).
Microanal. 17, 2011 (2011).
33. R. Kelley et al., Xe + FIB milling and measurement of amorphous silicon damage. Microsc.
Microanal. 19, 862–863 (2013).
41. A. L. Bronwyn, N. Grigorieff, Quantification of gallium cryo-FIB milling damage in biological
lamella. Electron Microscopy Public Image Archive. https://www.ebi.ac.uk/empiar/EMPIAR-11544/.
Deposited 7 May 2023.
PNAS 2023 Vol. 120 No. 23 e2301852120
https://doi.org/10.1073/pnas.2301852120 9 of 9
| null |
10.7554_elife.76554.pdf
|
Data availability
Sequencing data have been deposited in GEO under accession code GSE157681.
The following dataset was generated:
Author(s)
Brown JM, Helsley R,
Kadam A, Neumann
C
Year
2021
Dataset title
Dataset URL
Database and Identifier
The Gut Microbe- Derived
Metabolite Trimethylamine
is a Biomarker of and
Therapeutic Target in
Alcohol- Associated Liver
Disease
http://www. ncbi.
nlm. nih. gov/ geo/
query/ acc. cgi? acc=
GSE157681
NCBI Gene Expression
Omnibus, G
|
Data availability Sequencing data have been deposited in GEO under accession code GSE157681. The following dataset was generated:
|
University of Kentucky
University of Kentucky
UKnowledge
UKnowledge
Pediatrics Faculty Publications
Pediatrics
1-27-2022
Gut Microbial Trimethylamine Is Elevated in Alcohol-Associated
Gut Microbial Trimethylamine Is Elevated in Alcohol-Associated
Hepatitis and Contributes to Ethanol-Induced Liver Injury in Mice
Hepatitis and Contributes to Ethanol-Induced Liver Injury in Mice
Robert N. Helsley
University of Kentucky, [email protected]
Tatsunori Miyata
Cleveland Clinic
Anagha Kadam
Cleveland Clinic
Venkateshwari Varadharajan
Cleveland Clinic
Naseer Sangwan
Cleveland Clinic
Follow this and additional works at: https://uknowledge.uky.edu/pediatrics_facpub
See next page for additional authors
Part of the Gastroenterology Commons, Hepatology Commons, and the Internal Medicine Commons
Right click to open a feedback form in a new tab to let us know how this document benefits you.
Right click to open a feedback form in a new tab to let us know how this document benefits you.
Repository Citation
Repository Citation
Helsley, Robert N.; Miyata, Tatsunori; Kadam, Anagha; Varadharajan, Venkateshwari; Sangwan, Naseer;
Huang, Emily C.; Banerjee, Rakhee; Brown, Amanda L.; Fung, Kevin K.; Massey, William J.; Neumann,
Chase; Orabi, Danny; Osborn, Lucas J.; Schugar, Rebecca C.; McMullen, Megan R.; Bellar, Annette; Poulsen,
Kyle L.; Kim, Adam; Pathak, Vai; and Mrdjen, Marko, "Gut Microbial Trimethylamine Is Elevated in Alcohol-
Associated Hepatitis and Contributes to Ethanol-Induced Liver Injury in Mice" (2022). Pediatrics Faculty
Publications. 321.
https://uknowledge.uky.edu/pediatrics_facpub/321
This Article is brought to you for free and open access by the Pediatrics at UKnowledge. It has been accepted for
inclusion in Pediatrics Faculty Publications by an authorized administrator of UKnowledge. For more information,
please contact [email protected].
Gut Microbial Trimethylamine Is Elevated in Alcohol-Associated Hepatitis and
Gut Microbial Trimethylamine Is Elevated in Alcohol-Associated Hepatitis and
Contributes to Ethanol-Induced Liver Injury in Mice
Contributes to Ethanol-Induced Liver Injury in Mice
Digital Object Identifier (DOI)
https://doi.org/10.7554/elife.76554
Notes/Citation Information
Notes/Citation Information
Published in eLife, v. 11, e76554.
© 2022, Helsley et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use and redistribution provided that the original author and source are credited.
The first 20 authors (including the one from the University of Kentucky) are shown on the author list
above. Please refer to the downloaded document for the complete author list.
Authors
Authors
Robert N. Helsley, Tatsunori Miyata, Anagha Kadam, Venkateshwari Varadharajan, Naseer Sangwan, Emily
C. Huang, Rakhee Banerjee, Amanda L. Brown, Kevin K. Fung, William J. Massey, Chase Neumann, Danny
Orabi, Lucas J. Osborn, Rebecca C. Schugar, Megan R. McMullen, Annette Bellar, Kyle L. Poulsen, Adam
Kim, Vai Pathak, and Marko Mrdjen
This article is available at UKnowledge: https://uknowledge.uky.edu/pediatrics_facpub/321
RESEARCH ARTICLE
Gut microbial trimethylamine is elevated
in alcohol- associated hepatitis and
contributes to ethanol- induced liver injury
in mice
Robert N Helsley1,2,3†, Tatsunori Miyata4†, Anagha Kadam1,2†,
Venkateshwari Varadharajan1,2, Naseer Sangwan1,2, Emily C Huang4,
Rakhee Banerjee1,2, Amanda L Brown1,2, Kevin K Fung1,2, William J Massey1,2,
Chase Neumann1,2, Danny Orabi1,2, Lucas J Osborn1,2, Rebecca C Schugar1,2,
Megan R McMullen4, Annette Bellar4, Kyle L Poulsen4, Adam Kim4, Vai Pathak5,
Marko Mrdjen1,2,4, James T Anderson1,2, Belinda Willard1,2, Craig J McClain6,
Mack Mitchell7, Arthur J McCullough2,4, Svetlana Radaeva8, Bruce Barton9,
Gyongyi Szabo10, Srinivasan Dasarathy2,4, Jose Carlos Garcia- Garcia11,
Daniel M Rotroff5, Daniela S Allende12, Zeneng Wang1,2, Stanley L Hazen1,2,13,
Laura E Nagy2,4, Jonathan Mark Brown1,2*
1Department of Cardiovascular and Metabolic Sciences, Lerner Research Institute
of the Cleveland Clinic, Cleveland, United States; 2Center for Microbiome and
Human Health, Lerner Research Institute, Cleveland Clinic, Cleveland, United States;
3Department of Pediatrics, Division of Pediatric Gastroenterology, Hepatology, and
Nutrition, College of Medicine, University of Kentucky, Lexington, United States;
4Department of Inflammation and Immunity, Lerner Research Institute, Cleveland
Clinic, Cleveland, United States; 5Department of Quantitative Health Sciences,
Lerner Research Institute, Cleveland Clinic, Cleveland, United States; 6Department
of Medicine, University of Louisville, Louisville, United States; 7Department of
Internal Medicine, University of Texas Southwestern Medical Center, Dallas, United
States; 8National Institute on Alcohol Abuse and Alcoholism, Bethesda, United
States; 9Department of Population and Quantitative Health Sciences, University of
Massachusetts Medical School, Worcester, United States; 10Department of Medicine,
Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, United
States; 11Life Sciences Transformative Platform Technologies, Procter & Gamble,
Cincinnati, United States; 12Department of Anatomical Pathology, Cleveland Clinic,
Cleveland, United States; 13Department of Cardiovascular Medicine, Heart and
Vascular and Thoracic Institute, Cleveland Clinic, Cleveland, United States
Abstract There is mounting evidence that microbes residing in the human intestine contribute
to diverse alcohol- associated liver diseases (ALD) including the most deadly form known as alcohol-
associated hepatitis (AH). However, mechanisms by which gut microbes synergize with excessive
alcohol intake to promote liver injury are poorly understood. Furthermore, whether drugs that
selectively target gut microbial metabolism can improve ALD has never been tested. We used liquid
chromatography tandem mass spectrometry to quantify the levels of microbe and host choline
co- metabolites in healthy controls and AH patients, finding elevated levels of the microbial metab-
olite trimethylamine (TMA) in AH. In subsequent studies, we treated mice with non- lethal bacterial
*For correspondence:
[email protected]
†These authors contributed
equally to this work
Competing interest: See page
22
Funding: See page 22
Received: 21 December 2021
Preprinted: 01 January 2022
Accepted: 31 December 2021
Published: 27 January 2022
Reviewing Editor: Hossein
Ardehali, Northwestern
University, United States
Copyright Helsley et al. This
article is distributed under the
terms of the Creative Commons
Attribution License, which
permits unrestricted use and
redistribution provided that the
original author and source are
credited.
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
1 of 28
Research article
choline TMA lyase (CutC/D) inhibitors to blunt gut microbe- dependent production of TMA in the
context of chronic ethanol administration. Indices of liver injury were quantified by complemen-
tary RNA sequencing, biochemical, and histological approaches. In addition, we examined the
impact of ethanol consumption and TMA lyase inhibition on gut microbiome structure via 16S rRNA
sequencing. We show the gut microbial choline metabolite TMA is elevated in AH patients and
correlates with reduced hepatic expression of the TMA oxygenase flavin- containing monooxygenase
3 (FMO3). Provocatively, we find that small molecule inhibition of gut microbial CutC/D activity
protects mice from ethanol- induced liver injury. CutC/D inhibitor- driven improvement in ethanol-
induced liver injury is associated with distinct reorganization of the gut microbiome and host liver
transcriptome. The microbial metabolite TMA is elevated in patients with AH, and inhibition of TMA
production from gut microbes can protect mice from ethanol- induced liver injury.
Editor's evaluation
This paper aims to understand the mechanisms by which gut microbes synergize with excessive
alcohol intake to cause liver injury, and whether drugs that selectively target gut microbial metab-
olism can improve alcohol- associated liver disease (ALD). The authors used liquid chromatography
tandem mass spectrometry to quantify the levels of microbe and host choline co- metabolites in
controls and patients with alcohol- associated hepatitis (AH). They also treated mice with bacterial
choline trimethylamine (TMA) lyase inhibitors to reduce gut microbe- dependent TMA produc-
tion, followed by measurement of Indices of liver injury. They showed that gut microbial choline
metabolite TMA is increased in AH patients, which correlates with reduced liver expression of the
TMA oxygenase Flavin- containing monooxygenase 3 (FMO3). They also show that inhibition of gut
microbial CutC/D activity protects from ethanol- induced liver injury in mouse models, which was
associated with reorganization of the gut microbiome and host liver transcriptome. The authors
conclude that microbial TMA is elevated in patients with AH, and inhibition of TMA production by
gut microbes protects against ethanol- induced liver injury.
Introduction
Alcohol- associated liver disease (ALD) includes a spectrum of liver pathologies including steatosis,
fibrosis, cirrhosis, and the most severe manifestation known as alcohol- associated hepatitis (AH).
Shortly after diagnosis AH patients die at a staggering rate of 40–50% (Masarone et al., 2016;
Kochanek et al., 2017). Despite many attempts, an effective therapy for this deadly disease has
been elusive. Similar to other components of the spectrum of ALD, AH has consistently been linked
to reorganization of the gut microbiome and dysregulation of microbe- host interactions (Chen et al.,
2011; Yan et al., 2011; Mutlu et al., 2009; Mutlu et al., 2012; Tripathi et al., 2018; Ciocan et al.,
2018; Llopis et al., 2016; Duan et al., 2019; Smirnova et al., 2020; Gao et al., 2019; Puri et al.,
2018; Lang and Schnabl, 2020). It is well appreciated that chronic alcohol use can elicit structural
alterations in the gut barrier, allowing either live bacteria themselves or microbe- associated molecule
patterns (MAMPs), such as lipopolysaccharide (LPS), to enter the portal circulation where they can
directly engage pattern recognition receptors (PRRs) such as Toll- like receptors (TLRs) or NOD- like
receptors (NLRP3, NLRP6, etc.) to promote hepatic inflammation and tissue injury (Wilkinson et al.,
1974; Tarao et al., 1979; Uesugi et al., 2001; Paik et al., 2003; DeSantis et al., 2013; Knorr et al.,
2020). In addition to MAMP- PRR interactions, gut microbes can act as a collective endocrine organ,
producing a vast array of small molecules, proteins, and lipid metabolites that can engage dedicated
host receptor systems to also impact liver disease progression (Brown and Hazen, 2015). Collectively,
these MAMP- PRR and microbial metabolite- host receptor interactions converge to promote ALD and
many other diseases of uncontrolled inflammation (Brown and Hazen, 2015; Gilbert et al., 2018).
Although there is now clear evidence that microbe- host interactions play a key role in liver disease
progression (Chen et al., 2011; Yan et al., 2011; Mutlu et al., 2009; Mutlu et al., 2012; Tripathi et al.,
2018; Ciocan et al., 2018; Llopis et al., 2016; Duan et al., 2019; Smirnova et al., 2020; Gao et al.,
2019; Puri et al., 2018; Lang and Schnabl, 2020; Wilkinson et al., 1974; Tarao et al., 1979; Uesugi
et al., 2001; Paik et al., 2003; DeSantis et al., 2013; Knorr et al., 2020; Brown and Hazen, 2015;
Gilbert et al., 2018), ALD drug discovery to this point has focused primarily on targets encoded by the
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
2 of 28
Medicine
Research article
human genome. Our knowledge is rapidly expanding as to how microbes intersect with ALD progres-
sion, including cataloging microbial genomes. We also now understand the repertoire of MAMPs gut
microbes harbor as well as the vast array of metabolites that they produce in both patients with ALD
and animal models of ethanol- induced liver injury (Chen et al., 2011; Yan et al., 2011; Mutlu et al.,
2009; Mutlu et al., 2012; Tripathi et al., 2018; Ciocan et al., 2018; Llopis et al., 2016; Duan et al.,
2019; Smirnova et al., 2020; Gao et al., 2019; Puri et al., 2018; Lang and Schnabl, 2020; Wilkinson
et al., 1974; Tarao et al., 1979; Uesugi et al., 2001; Paik et al., 2003; DeSantis et al., 2013; Knorr
et al., 2020; Brown and Hazen, 2015; Gilbert et al., 2018). However, there are very few examples of
where this information has been leveraged into safe and effective therapeutic strategies. In general,
the microbiome- targeted therapeutic field has primarily focused on either anti-, pre-, or pro- biotic
approaches, yet these microbial community- restructuring approaches have resulted in very modest or
non- significant effects in clinical studies of liver disease (Kwak et al., 2014; Asgharian et al., 2020;
Reijnders et al., 2016; Madjd et al., 2016). As an alternative microbiome- targeted approach, we and
others have begun developing non- lethal selective small molecule inhibitors of bacterial enzymes with
the intention of reducing levels of disease- associated microbial metabolites with mechanistic rationale
for contribution to disease pathogenesis (Roberts et al., 2018; Wang et al., 2015; Gupta et al.,
2020; Organ et al., 2020; Orman et al., 2019). In fact, we have recently shown that small molecule
inhibition of the gut microbial transformation of choline into trimethylamine (TMA), the initial and
rate- limiting step in the generation of the cardiovascular disease (CVD)- associated metabolite trime-
thylamine N- oxide (TMAO), can significantly reduce disease burden in animal models of atheroscle-
rosis, thrombosis, heart failure, and chronic kidney disease (Roberts et al., 2018; Wang et al., 2015;
Gupta et al., 2020; Organ et al., 2020). Although the gut microbial TMAO pathway has been studied
mostly in the context of CVD (Wang et al., 2011; Koeth et al., 2013; Zhu et al., 2016; Zhu et al.,
2017; Tang et al., 2013; Wang et al., 2014b; Trøseid et al., 2015; Tang and Hazen, 2014), recent
studies found that breath levels of the primary metabolite TMA and other related co- metabolites
are elevated in patients with ALD (Hanouneh et al., 2014; Ascha et al., 2016). These data showed
promise, but whether the gut microbial TMAO pathway is causally related to ALD has never been
explored. Hence, here we set out to understand how the gut microbial TMA/TMAO pathway may
play a contributory role in ALD susceptibility and progression, and to test whether selective drugs that
lower gut microbial production of TMA can be an effective therapeutic strategy. In an era when host
genetics/genomics approaches dominate, this work reminds us that genes and metabolic products
produced by gut bacteria play equally important roles in modulating disease susceptibility. Whereas
pathways encoded by the host genome have long been pursued as drug targets, this work provides
proof of concept that rationally designed drugs that target bacterial metabolism likely have untapped
therapeutic potential in ALD and beyond.
Results
Circulating levels of the gut microbial metabolite TMA are elevated in
AH
In a previous collaborative study, we reported that the highly volatile microbial metabolite TMA is
elevated in exhaled breath of patients with AH (Hanouneh et al., 2014), and related co- metabolites,
such as trimethyllysine and carnitine, can serve as prognostic indicators of mortality in AH (Ascha et al.,
2016). Given the extremely volatile nature of TMA, it is readily detectable in breath, but is challenging
to accurately quantitate levels in the circulation because TMA rapidly dissipates during collection
and storage. To reduce the volatility of TMA and enable its analysis in the circulation, we coordinated
patient blood collection utilizing rapid acidification of separated plasma (protonated TMA has a lower
vapor pressure) across a large multi- center AH consortium (Defeat Alcoholic Steatohepatitis [DASH]
consortium) (Crabb et al., 2016; Vatsalya et al., 2020; Saha et al., 2019). This provided us the
unique opportunity to accurately quantify circulating TMA levels in human subjects, including those
with moderate or severe AH for the first time. Patient demographics and clinical characteristics for the
cohort examined are summarized in Figure 1—source data 1; Figure 1—source data 2. Importantly,
MELD score, Maddrey’s discriminant function score, Child- Pugh score, aspartate aminotransferase
(AST), total bilirubin, creatinine, and international normalized ratio were higher in patients with severe
AH compared to moderate AH patients, while serum albumin was lower in severe AH compared to
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
3 of 28
Medicine
Research article
moderate AH patients. In agreement with previous breath metabolomics studies (Hanouneh et al.,
2014; Ascha et al., 2016), plasma TMA levels were significantly elevated in moderate and severe AH
patients compared to healthy controls (Figure 1A). However, the CVD- related co- metabolite TMAO
was reciprocally decreased in AH patients (Figure 1B). Given the reciprocal alterations in plasma TMA
and TMAO levels, we next examined the expression of the host liver enzyme flavin- containing mono-
oxygenase 3 (FMO3) which is the predominant TMA to TMAO converting enzyme in the adult liver
(Cashman, 2002). Interestingly, mRNA levels for FMO3 are uniquely repressed in patients with more
severe AH (AH with liver failure [MELD 22–28] and AH with emergency liver transplant [MELD 18–21]),
but not in other liver disease etiologies such as non- alcoholic fatty liver disease (NAFLD) or viral
hepatitis (Figure 1C). In agreement with reduced mRNA levels (Figure 1C), patients with severe AH
undergoing emergency liver transplant have marked reduction in FMO3 protein (Figure 1D), which
likely contributes to elevations in plasma TMA (Figure 1A). Although ethanol feeding in mice does
not consistently result in reduced hepatic Fmo3 expression (data not shown), a single injection of lipo-
polysaccharide (LPS) to induce acute hepatic inflammation is associated with both a reduction in the
expression of Fmo3 and a significant increase in the TMA receptor trace amine- associated receptor 5
(Taar5) (Figure 1E). It is important to note that circulating choline levels was not significantly altered
in patients with AH compared to healthy controls (Figure 1—figure supplement 1). However, plasma
levels of one of the gut microbial substrates for TMA production (carnitine) and other TMA pathway
co- metabolites (e.g. betaine and γ-butyrobetaine) were elevated in patients with AH compared to
healthy controls (Figure 1—figure supplement 1). These findings, in addition to previous breath
metabolomic studies (Hanouneh et al., 2014; Ascha et al., 2016), provide evidence that TMA and
related co- metabolites may allow for discrimination of AH from other liver diseases.
Microbial choline TMA lyase inhibition protects mice from ethanol-
induced liver injury
We next sought to establish whether a causal relationship between gut microbial TMA production and
ALD progression exists, and to test the hypothesis that selectively drugging microbial choline trans-
formation can serve as a mechanism for improving host liver disease and attenuating ethanol- induced
liver injury in mice. Mice were individually treated with two recently reported non- lethal bacterial
choline TMA lyase inhibitors, iodomethylcholine (IMC) and fluoromethylcholine (FMC) (Roberts et al.,
2018). These small molecule inhibitors exhibit potent in vivo inhibition of the gut microbial choline
TMA lyase enzyme CutC (Craciun and Balskus, 2012), and have been shown to effectively block
bacterial choline to TMA conversion in vivo (Roberts et al., 2018). Designed as suicide substrate
mechanism- based inhibitors, past studies reveal that the vast majority of IMC and FMC is retained in
the gut within luminal bacteria and excreted in the feces with limited systemic exposure of the polar
drug in the host (Roberts et al., 2018; Gupta et al., 2020; Organ et al., 2020).
IMC treatment effectively blunted ethanol- induced increases in plasma TMA and TMAO (Figure 2A
and B). IMC also produced modest increases in plasma choline and betaine, while reducing plasma
carnitine, particularly in pair- fed mice (Figure 2C–E). IMC also prevented ethanol- induced increases
in alanine aminotransferase (ALT) and hepatic steatosis (Figure 2F, G, and K). Interestingly, IMC
treatment prevented ethanol- induced increases in hepatic triglycerides (Figure 2G), and reduced
hepatic total and cholesterol esters, but not free cholesterol, in both pair- and ethanol- fed conditions
(Figure 2H–I). IMC treatment also reduced the expression levels of the pro- inflammatory cytokine
tumor necrosis factor α (Tnfα) (Figure 2L). Although IMC was well tolerated in several previous mouse
studies in the setting of standard rodent chow- feeding (Roberts et al., 2018; Gupta et al., 2020;
Organ et al., 2020), here we found an unexpected reduction in food intake and body weights in mice
receiving both IMC and ethanol (Figure 2—figure supplement 1A, B). Although IMC was clearly
protective against ethanol- induced liver injury, this potential drug- ethanol interaction prompted us
to test another structurally distinct gut microbe- targeted choline TMA lyase inhibitor FMC (Roberts
et al., 2018; Figure 3 and Figure 3—figure supplement 1).
Importantly, FMC was well tolerated and did not significantly alter liquid diet intake or body
weights throughout the 25- day chronic ethanol feeding study (Figure 2—figure supplement 1C, D).
FMC treatment trended toward reducing plasma TMA (Figure 3A), and more dramatically suppressed
plasma TMAO levels (Figure 3B). Unlike IMC, which also altered other co- metabolites such as choline,
betaine, and carnitine (Figure 2B- E), FMC did not significantly alter these TMA co- metabolites
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
4 of 28
Medicine
Research article
Figure 1. The gut microbial volatile metabolite trimethylamine (TMA) is elevated in alcohol- associated hepatitis
(AH). Plasma TMA (A) and trimethylamine N- oxide (TMAO) (B) levels in patients considered healthy (n = 13 for TMA
and 20 for TMAO), or who have moderate (MELD < 20) (n = 52 for TMA and 111 for TMAO) or severe (MELD >
20) (n = 83 for TMA and 152 for TMAO) AH. (C) RNA sequencing results from liver tissues of patients with different
Figure 1 continued on next page
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
5 of 28
Medicine
Research article
Figure 1 continued
pathologies, including: healthy controls (HC, n = 10), early AH (EAH, n = 12; MELD 7–8), AH with liver failure (AHL,
n = 18; MELD 22–28), explant tissue from patients with severe AH with emergency liver transplants (ExAH, n = 10;
MELD 18–21), non- alcohol- associated fatty liver disease (NAFLD; n = 8), hepatitis C virus (HCV; n = 9), and hepatitis
C virus with cirrhosis (HCV_Cirr, n = 9). Gene expression was measured by transcripts per million (TPM). Boxplots
of average expression for Fmo3 in different disease groups; error bars indicate SD (q < 0.05 in comparison to
healthy controls). (D) Liver FMO3 protein expression measured by Western blot from healthy patients and patients
with severe AH undergoing emergency liver transplant (Maddrey’s discriminant function 45–187). (E) Liver Tnfa,
Il1b, Fmo3, and Taar5 transcript levels were measured by qPCR from female WT mice injected with either saline or
lipopolysaccharide (LPS) for 6 hr. N = 6; unpaired Student’s t- test. *p ≤ 0.05; ***p ≤ 0.001.
The online version of this article includes the following source data and figure supplement(s) for figure 1:
Source data 1. Demographic and clinical parameters for entire cohort of healthy controls and patients with AH.
Source data 2. Demographic and clinical parameters for subset of healthy controls and patients with AH included
in TMA assay.
Source data 3. Liver flavin- containing monooxygenase 3 (FMO3) protein expression measured by Western blot
from healthy patients (HC) and patients with severe alcohol- associated hepatitis (AH) undergoing emergency liver
transplant (Maddrey’s discriminant function 45–187).
Source data 4. Liver flavin- containing monooxygenase 3 (FMO3) protein expression measured by Western blot
from healthy patients (HC) and patients with severe alcohol- associated hepatitis (AH) undergoing emergency liver
transplant (Maddrey’s discriminant function 45–187).
Source data 5. Liver HSC70 protein expression measured by Western blot from healthy patients (HC) and patients
with severe alcohol- associated hepatitis (AH) undergoing emergency liver transplant (Maddrey’s discriminant
function 45–187).
Source data 6. Liver HSC70 protein expression measured by Western Blot from healthy patients (HC) and patients
with severe alcohol- associated hepatitis (AH) undergoing emergency liver transplant (Maddrey’s discriminant
function 45–187).
Figure supplement 1. Levels of trimethylamine (TMA)- related metabolites in alcohol- associated hepatitis (AH).
(Figure 3C- E). More importantly, as with IMC (Figure 2F–K), FMC treatment significantly protected
against ethanol- induced ALT elevations (Figure 3F), hepatic steatosis (Figure 3G and K), and reduced
total and esterified cholesterol levels without altering free cholesterol (Figure 3H–J). However, FMC
trended to reduce but did not significantly alter Tnfα expression (Figure 3L). To determine whether
these effects were generalizable in other models of ethanol- induced liver injury, we exposed control
and FMC- treated mice to a 10- day chronic model in which mice were allowed free access to a 5% vol/
vol (27% kcal) for 10 days (Figure 3—figure supplement 1). In this 5%–10- day ethanol feeding model
FMC treatment did not significantly alter food intake, body weight, or blood ethanol levels, but was
able to selectively suppress TMA and TMAO levels (Figure 3—figure supplement 1). FMC treat-
ment in the 5%–10- day model significantly reduced plasma AST and ALT levels, and trended toward
lowering liver triglycerides (Figure 3—figure supplement 1). However, in this short- term model there
were no apparent differences in hepatic cytokine/chemokine gene expression with either ethanol
exposure or FMC treatment (Figure 3—figure supplement 1 and data not shown). Collectively, these
data demonstrate that gut microbe- targeted choline TMA lyase inhibition with two structurally distinct
inhibitors (IMC or FMC) can generally protect mice against ethanol- induced liver injury.
Microbial choline TMA lyase inhibitors promote remodeling of the
gut microbiome and host liver transcriptome in an ethanol-dependent
manner
One theoretical advantage of the selective microbe- targeted choline TMA lyase inhibitors, compared
to antibiotic or MAMP- PRR- targeted therapies, is that they are anticipated to exert less selective
pressure for development of drug resistance given their non- lethal nature. However, microbes that
preferentially utilize choline as a carbon or nitrogen source might be anticipated to have reduced
competitive advantage in the presence of the inhibitor. We therefore next examined whether IMC or
FMC treatment was associated with alterations in choline utilizers and other members of the murine
gut microbiome community that are known to be correlated with ethanol- induced liver injury (Chen
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
6 of 28
Medicine
Research article
Figure 2. Small molecule choline trimethylamine (TMA) lyase inhibition with iodomethylcholine (IMC) protects mice against ethanol- induced liver
injury. Nine- to eleven- week- old female C57BL6/J mice were fed either ethanol- fed or pair- fed in the presence and absence of IMC as described in the
methods. Plasma levels of TMA (A), trimethylamine N- oxide (TMAO) (B), choline (C), carnitine (D), and betaine (E) were measured by mass spectrometry
(n = 4–5). Plasma alanine aminotransferase (ALT) (F) was measured enzymatically (n = 4–5). Liver triglycerides (G), total cholesterol (H), cholesterol esters
(I), and free cholesterol (J) were measured enzymatically (n = 4–5). (K) Representative H&E staining of livers from pair and EtOH- fed mice in the presence
and absence of IMC. (L) Hepatic messenger RNA levels of tumor necrosis factor alpha (Tnfα). Statistics were completed by a two- way analysis of variance
(ANOVA) followed by a Tukey’s multiple comparison test. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001. All data are presented as mean ± SEM,
unless otherwise noted.
The online version of this article includes the following figure supplement(s) for figure 2:
Figure supplement 1. Small molecule inhibition with iodomethylcholine (IMC), but not fluoromethylcholine (FMC), reduces food intake in ethanol- fed
mice.
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
7 of 28
Medicine
Research article
Figure 3. Small molecule choline trimethylamine (TMA) lyase inhibition with fluoromethylcholine (FMC) protects mice against ethanol- induced liver
injury. Nine- to eleven- week- old female C57BL6/J mice were fed either ethanol- fed or pair- fed in the presence and absence of FMC as described in the
methods. Plasma levels of TMA (A), trimethylamine N- oxide (TMAO) (B), choline (C), carnitine (D), and betaine (E) were measured by mass spectrometry
(n = 3–5). Plasma alanine aminotransferase (ALT) (F) were measured at necropsy (n = 4–5). Liver triglycerides (G), total cholesterol (H), cholesterol esters
(I), and free cholesterol (J) were measured enzymatically (n = 4–5). (K) Representative H&E staining of livers from pair and EtOH- fed mice in the presence
and absence of FMC. (L) Hepatic messenger RNA levels of tumor necrosis factor alpha (Tnfα). Statistics were completed by a two- way analysis of
variance (ANOVA) followed by a Tukey’s multiple comparison test. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001. All data are presented as mean ±
SEM, unless otherwise noted.
The online version of this article includes the following figure supplement(s) for figure 3:
Figure supplement 1. Small molecule inhibition of gut microbial trimethylamine (TMA) lyase activity with fluoromethylcholine (FMC) in a second model
of ethanol- induced liver injury.
Figure supplement 2. A single bolus of ethanol does not significantly alter trimethylamine (TMA) or trimethylamine N- oxide (TMAO) levels in mice.
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
8 of 28
Medicine
Research article
et al., 2011; Yan et al., 2011; Mutlu et al., 2009; Mutlu et al., 2012; Tripathi et al., 2018; Ciocan
et al., 2018; Llopis et al., 2016; Duan et al., 2019; Smirnova et al., 2020; Gao et al., 2019; Puri
et al., 2018; Lang and Schnabl, 2020). It is important to note that both IMC (Figure 4A–E) and FMC
(Figure 4F–J) altered the gut microbiome, with some consistent, yet several distinct differences. Non-
metric multidimensional scaling (NMDS) of microbial taxa revealed distinct clusters, indicating that
both IMC and FMC promoted clear restructuring of the cecal microbiome in an ethanol- dependent
manner (Figure 4A and F). Under pair- feeding conditions, both IMC and FMC caused a reciprocal
decrease in the relative abundance of Bacteroidetes and increase in Firmicutes (Figure 4B and G).
However, under ethanol- fed conditions IMC resulted in increased Bacteroidetes and reduced Firmic-
utes, and FMC treatment resulted in more modest reductions in Bacteroidetes and increased Firmic-
utes (Figure 4B and G). When examining drug- specific alterations at the genus level, we found that
under both pair- and ethanol- fed conditions, IMC treatment promoted significant increases in Faeca-
libaculum and Escherichica/Shigella, and reductions in Bacteroidales_S24- 7 (Figure 4C–E , and H–I).
FMC, however, most significantly altered Turicibacter, Oscillibacter, and Lachnospiraceae, and it is
important to note that these FMC- induced alterations were different between pair- and ethanol- fed
groups (Figure 4C–E , and H–I). Collectively, these data demonstrate that inhibition of gut microbial
choline to TMA transformation with a selective non- lethal small molecule inhibitor promotes restruc-
turing of the gut microbiome in an ethanol- dependent manner.
To more globally understand the effects of choline TMA lyase inhibitors on the host liver, we
performed unbiased RNA sequencing in mice undergoing pair or ethanol feeding treated either with
or without IMC (Figure 5). NMDS and hierarchical clustering analysis showed clear separation between
all four groups (Figure 5A and B). In pair- fed mice, IMC treatment caused significant decreases in
several genes encoding major urinary proteins (Mup2, Mup10, Mup11, and Mup18) and cytochrome
p450 enzymes (Cyp3a16, Cyp3a44), while increasing other genes involved in xenobiotic metabolism
(Ephx1, Cyp4a31) and hormone/cytokine signaling (Lepr, Fgf21, Il22ra1) (Figure 5C). Under ethanol-
feeding conditions, IMC treatment most significantly altered genes involved in hepatocyte metab-
olism (Cyp8b1, Ugt1a5, Pnpla5, Sult2a8, Ces3a, and Cmah), RNA processing (Ddx21, Ftsj3, Dus1l,
and Cmah), and again major urinary proteins (Mup2, Mup10, Mup11, and Mup20) (Figure 5D and E).
These unbiased RNASeq data demonstrate that gut microbe- targeted choline TMA lyase inhibitors
can alter the host liver transcriptome in an ethanol feeding- dependent manner.
The microbe-derived metabolite TMA elicits rapid hormone-like
signaling effects in mouse liver
The gut microbe- derived co- metabolites TMA and TMAO are generated postprandially in both
rodents and humans after a substrate- rich meal is ingested (Schugar et al., 2018; Boutagy et al.,
2015). Given the acute meal- related production and recent identification of candidate host receptors
for TMA (Li et al., 2013; Wallrabenstein et al., 2013) and TMAO (Chen et al., 2019), we hypothe-
sized that TMA may be acting as a gut microbe- derived hormone to promote liver injury. However,
currently nothing is known regarding the acute hormone- like signaling effects stimulated by TMA in
the liver. To address this gap, we infused TMA directly into the portal circulation draining the gut (i.e.
portal vein) of fasted mice and examined global phosphorylation events stimulated in the liver 10 min
later using a phosphoproteomics approach (Figure 6A). It is important to note that this experiment
provided high levels of exogenous TMA via direct injection, and future studies should focus on more
physiologically relevant modes of TMA production like provision of gut bacteria that can naturally or
be genetically engineered to produce high levels. A total of 36 liver proteins exhibited site- specific
hypo- or hyper- phosphorylation 10 min after administration of TMA relative to vehicle- injected mice
(Figure 6B). Several of the TMA- driven phosphorylation events represented proteins that are enriched
in key hormonal signaling pathways known to impact hepatic metabolism. For example, portal vein
infusion of TMA resulted in altered phosphorylation of proteins implicated in protein kinase A (PKA)
signaling, including A kinase anchor protein 1 (AKAP1) (Huang et al., 1999) and FK506- binding protein
15 (FKBP15) (Nooh and Bahouth, 2017), and insulin signaling including insulin receptor substrate 2
(IRS2) (Araki et al., 1994; Figure 6B–D). TMA infusion was also associated with altering the phosphor-
ylation of several guanine nucleotide exchange factors (GEF), including Rac/Cdc42 guanine nucleo-
tide exchange factor 6 (Arhgef6) and Rho GTPase activating protein 17 (ARHGAP17) (Zhou et al.,
2016; Aslan, 2019), and proteins involved in RNA processing/splicing including signal recognition
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
9 of 28
Medicine
Research article
Figure 4. Small molecule choline trimethylamine (TMA) lyase inhibition promotes remodeling of the gut microbiome in an ethanol- dependent manner.
Nine- to eleven- week- old female C57BL6/J mice were fed either ethanol- fed or pair- fed in the presence and absence of iodomethylcholine (IMC)
or fluoromethylcholine (FMC) as described in the methods. (A) Non- metric multidimensional scaling (NMDS) plots based on the Bray- Curtis index
between the pair, EtOH, pair + 0.06% IMC, and EtOH + 0.06% IMC groups, Statistical analysis was performed with permutational multivariate analysis
of variance (PERMANOVA), and p- values are labeled in plots. R2 values are noted for comparisons with significant p- values and stand for percentage
variance explained by the variable of interest. (B) Boxplots of relative abundance patterns for Firmicutes and Bacteroidetes distinguishing pair, EtOH,
pair + 0.06% IMC and EtOH + 0.06% IMC groups. Statistical analysis was performed with Mann- Whitney U test (also called the Wilcoxon rank- sum
test, p- values are labeled in plots). Plotted are interquartile ranges (boxes), and dark lines in boxes are medians. (C) Stacked bar charts of relative
abundance (left y- axis) of the top 20 genera assembled across all four groups (pair, EtOH, pair + 0.06% IMC, and EtOH + 0.06% IMC groups). Pairwise
Figure 4 continued on next page
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
10 of 28
Medicine
Research article
Figure 4 continued
differential abundance analyses between (D) pair- fed and pair- fed + 0.06% IMC and (E) EtOH- fed and EtOH- fed + 0.06% IMC group. Statistical analysis
was performed with White’s non- parametric t- test (p- values are labeled in plots). (F) NMDS plots based on the Bray- Curtis index between the pair,
EtOH, pair + 0.006% FMC, and EtOH + 0.006% FMC groups, Statistical analysis was performed with permutational multivariate analysis of variance
(PERMANOVA), and p- values are labeled in plots. R2 values are noted for comparisons with significant p- values and stand for percentage variance
explained by the variable of interest. (G) Boxplots of relative abundance patterns for Firmicutes and Bacteroidetes distinguishing pair, EtOH, pair +
0.006% FMC, and EtOH + 0.006% FMC groups. Statistical analysis was performed with Mann- Whitney U test (also called the Wilcoxon rank- sum test,
p- values are labeled in plots). Plotted are interquartile ranges (boxes), and dark lines in boxes are medians. (H) Stacked bar charts of relative abundance
(left y- axis) of the top 20 genera assembled across all four groups (pair, EtOH, pair + 0.06% FMC, and EtOH + 0.006% FMC groups). Pairwise differential
abundance analyses between (I) pair- fed and pair- fed + 0.06% FMC, and (J) EtOH- fed and EtOH- fed + 0.006% FMC group. Statistical analysis was
performed with White’s non- parametric t- test (p- values are labeled in plots).
particle 14 (SRP14) (Strub and Walter, 1990) and serine- and arginine- rich splicing factor 1 (SRSF1)
(Cho et al., 2011; Figure 6B and C). These data have identified acute TMA- driven signaling events
in the liver in vivo, and potentially link TMA to acute alterations in PKA-, insulin-, and GEF- driven
signaling cascades that deserve further exploration.
Discussion
Although drug discovery has historically targeted pathways in the human host, there is untapped
potential in therapeutically targeting the gut microbial endocrine organ to treat advanced liver
disease. This paradigm shift is needed in light of the clear and reproducible associations between the
gut microbiome in viral, alcohol- associated, and non- alcohol- associated liver diseases (Chen et al.,
2011; Yan et al., 2011; Mutlu et al., 2009; Mutlu et al., 2012; Tripathi et al., 2018; Ciocan et al.,
2018; Llopis et al., 2016; Duan et al., 2019; Smirnova et al., 2020; Gao et al., 2019; Puri et al.,
2018; Lang and Schnabl, 2020; Wilkinson et al., 1974; Tarao et al., 1979; Uesugi et al., 2001; Paik
et al., 2003; DeSantis et al., 2013; Knorr et al., 2020; Brown and Hazen, 2015). Now we are faced
with both the challenge and opportunity to test whether microbe- targeted therapeutic strategies
can improve health in the human metaorganism without negatively impacting the symbiotic relation-
ships between microbes and host. Although traditional microbiome manipulating approaches such as
antibiotics, prebiotics, probiotics, and fecal microbial transplantation have shown their own unique
strengths and weaknesses, each of these presents unique challenges particularly for use in chronic
diseases such as end stage liver disease. As we move toward selective non- lethal small molecule ther-
apeutics, the goal is to have exquisite target selectivity and limited systemic drug exposure given that
the targets are microbial in nature. This natural progression parallels the paradigm shifts in oncology
which have transitioned from broadly cytotoxic chemotherapies to target- selective small molecule
and biologics- based therapeutics. Here, we provide the first evidence that the gut microbial choline
metabolite TMA is elevated in the plasma of patients with AH, which corroborates previous reports
showing that TMA is also prominent in the breath of patients with AH (Hanouneh et al., 2014). Hence,
further studies are warranted to determine whether combined measures of breath and blood TMA
can serve as a prognostic biomarker to accurately predict AH- related mortality. Here, we also show
for the first time that a selective non- lethal small molecule drug that reduces bacterial production of
TMA can prevent ethanol- induced liver injury in mice. We also demonstrate that direct administration
of TMA can elicit rapid signaling effects in the liver, supporting the notion that gut microbial metabo-
lites produced postprandially can act in an endocrine- like manner to alter host signal transduction and
associated disease pathogenesis. Collectively, these studies suggest that selective drugs targeting the
gut microbial TMA pathway may hold promise for treating AH.
As drug discovery advances in the area of small molecule non- lethal bacterial enzyme inhibitors, it is
key to understand how these drugs impact microbial ecology in the gut and other microenvironments.
As we have previously reported (Roberts et al., 2018; Gupta et al., 2020; Organ et al., 2020), gut
microbe- targeted choline TMA lyase inhibitors (IMC and FMC) induced a significant remodeling of the
cecal microbiome in mice. In the current studies there were some consistent, but many different cecal
microbiome alterations when comparing IMC and FMC (Figure 4), yet both drugs similarly improved
ethanol- induced liver injury. As small molecule bacterial enzymes inhibitors are developed, it will be
extremely important to understand their effects on microbial ecology, and it is expected that some
of the beneficial effects of these drugs will indeed originate from the restructuring of gut microbiome
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
11 of 28
Medicine
Research article
Figure 5. Small molecule choline trimethylamine (TMA) lyase inhibition with iodomethylcholine (IMC) alters the hepatic transcriptome in response to
ethanol. Nine- to eleven- week- old female C57BL6/J mice were fed either ethanol- fed or pair- fed in the presence and absence of IMC as described in
the methods. RNA was isolated from the livers and subjected to next- generation sequencing. (A) Non- metric multidimensional scaling (NMDS) plots;
each point represents a single sample from a single mouse. Positions of points in space display dissimilarities in the transcriptome, with points further
from one another being more dissimilar. (B–C) Row- normalized expression for the top 25 DEGs shown by heat map (B) while the volcano plot (C)
summarizes log2 fold changes vs. significance in response to IMC treatment in pair (left) and ethanol (right) feeding (n = 4). (D) Summary of significantly
differentially regulated pathways in mice treated with IMC in the ethanol- fed mice (n = 4).
communities. In fact, this is not an uncommon mechanism by which host targeted drugs impact human
health. A recent study showed that nearly a quarter of commonly used host- targeted drugs have
microbiome- altering properties (Maier et al., 2018), and in the context of diabetes therapeutics it is
important to note metformin’s anti- diabetic effects are partially mediated by the drug’s microbiome
altering properties (Wu et al., 2017). Given the strong association between gut microbiome and liver
disease (Chen et al., 2011; Yan et al., 2011; Mutlu et al., 2009; Mutlu et al., 2012; Tripathi et al.,
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
12 of 28
Medicine
Research article
Figure 6. Trimethylamine (TMA) rapidly reorganizes liver signal transduction in vivo. (A) Schematic of experiment; female C57BL/6 mice were fasted
overnight (12 hr fast), and then injected directly into the portal vein with vehicle (saline), or TMA, and only 10 min later liver tissue was harvested
for phosphoproteomic analysis to identify TMA- responsive phosphorylation events in mouse liver (n = 4 per group). (B) List of proteins that were
differentially phosphorylated (p < 0.05) upon TMA administration in vivo. (C) A doubly charged ion was present in the phospho- enriched sample
Figure 6 continued on next page
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
13 of 28
Medicine
Research article
Figure 6 continued
that was identified as the KSpSVEGLEPAENK from signal recognition particle 14 kDa protein (Srp14). The CID spectra of this ion is dominated by
H3PO4 loss from the precursor ion consistent with the presence of a pS or pT residue. The mass difference between the y11 and y10 ions is consistent
with modification at S45. The observed chromatograms for this peptide from the saline and TMA samples are shown and the TMA/saline ratio
was determined to be 3.6 (p- value 0.0114). (D) A doubly charged ion was present in the phospho- enriched sample that was identified as the
RLpSEEACPGVLSVAPTVTQPPGR from A- kinase anchor protein 1. The CID spectra of this ion is dominated by fragmentation C- terminal to the proline
residues. The mass of the b7 ion is consistent with modification at S55. The observed chromatograms for this peptide from the saline and TMA samples
are shown and the TMA/saline ratio was determined to be 0.4.
2018; Ciocan et al., 2018; Llopis et al., 2016; Duan et al., 2019; Smirnova et al., 2020; Gao et al.,
2019; Puri et al., 2018; Lang and Schnabl, 2020; Wilkinson et al., 1974; Tarao et al., 1979; Uesugi
et al., 2001; Paik et al., 2003; DeSantis et al., 2013; Knorr et al., 2020; Brown and Hazen, 2015;
Gilbert et al., 2018), it may prove advantageous to find therapeutics that beneficially remodel the gut
microbiome as well as engage either their microbe or host target of interest.
The metaorganismal TMA/TMAO pathway represents only one of many microbial metabolic circuits
that have been associated with human disease (Figure 7). In fact, many microbe- associated metabo-
lites such as short chain fatty acids, secondary bile acids, phenolic acids, polyamines, and others have
more recently been associated with many human diseases (Brown and Hazen, 2015; Gilbert et al.,
2018). In an ethanol- and meal- related manner, gut microbes produce a diverse array of metabolites
that reach micromolar to millimolar concentrations in the blood, making the collective gut microbiome
an active endocrine organ (Brown and Hazen, 2015). Small molecule metabolites are well known to
be mediators of signaling interactions in the host, and this work provides evidence that diet/ethanol-
microbe- host metabolic interplay can be causally linked to ethanol- induced liver injury. Our work, and
that of many others, demonstrates that there is clear evidence of bi- directional crosstalk between
the gut microbial endocrine organ and host liver metabolism. As drug discovery advances, it will be
important to move beyond targets based solely in the human host. This work highlights that non-
lethal gut microbe- targeted enzyme inhibitors may serve as effective therapeutics in AH and provides
proof of concept that this may be a generalizable approach to target metaorganismal crosstalk in
other disease contexts. In fact, selective inhibition of bacterial enzymes has the advantage over host
targeting given that small molecules can be designed to avoid systemic absorption and exposure,
thereby minimizing potential host off target effects. As shown here with the gut microbial TMA/TMAO
pathway, it is easy to envision that other microbe- host interactions are mechanistically linked to host
disease pathogenesis, serving as the basis for the rational design of microbe- targeted therapeutics
that improve human health.
Key resources table
Methods
Reagent type
(species) or
resource
Strain, strain
background Mice
(Females)
Designation
Source or reference
Identifiers
Additional information
9–11 Weeks
Jackson Laboratories
C57BL6/J, RRID:IMSR_
JAX:000664
5–8 per study
Biological
sample (Humans)
Plasma samples from
285 patients
Cleveland Clinic Foundation;
University of Louisville;
University of Massachusetts
Medical School; University of
Texas Southwestern Medical
Center
Not provided
Biological
sample (Humans)
Liver samples from five
healthy donors
Clinical Resource for Alcoholic
Hepatitis Investigations at
Johns Hopkins University
Not provided
Biological
sample (Humans)
Liver samples from
five patients with severe
AH
Clinical Resource for Alcoholic
Hepatitis Investigations at
Johns Hopkins University
Not provided
Continued on next page
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
14 of 28
Medicine
Research article
Continued
Reagent type
(species) or
resource
Antibody
Antibody
Designation
Source or reference
Identifiers
Additional information
Anti- FMO3 (Rabbit
monoclonal)
Abcam
Anti- HSC70 (Mouse
monoclonal)
Santa Cruz Biotechnology
Cat# ab126790, RRID:
AB_11128907
Cat# sc- 7298, RRID:
AB_627761
1:1000 (WB)
1:1000 (WB)
Cat#: NA934- 100UL, RRID:
AB_772206
1:5000 (WB)
Antibody
Anti- rabbit IgG HRP
GE- Healthcare
Antibody
Anti- mouse IgG HRP
GE- Healthcare
NA931V, RRID: AB_772210
1:5000 (WB)
F: CCAC CACG CTCT TCTG TCTAC
R:AGGGTCTGGGCCATAGAACT
F:AGTTGACGGACCCCAAAAG
R:AGCTGGATGCTCTCATCAGG
F: CCCACATGCTTTGAGAGGAG
R:GGAAGAGTTGGTGAAGACCG
F:AAAGAAAAGCTGCCAAGA
R:AAGGGAAGCCAACACACA
F:GCGGCAGGTCCATCTACG
R:GCCATCCAGCCATTCAGTC
F:TGCACCCAAACCGAAGTC
R:GTCAGAAGCCAGCGTTCACC
F: ACTT GGGG ACCA CCTA TTCCT
R:ATCGCCAATCAGACGCTCC
Sequence- based
reagent
Mouse Tnfα
Sequence- based
reagent
Mouse Il1β
Sigma
Sigma
PCR primers
PCR primers
Sequence- based
reagent
Sequence- based
reagent
Sequence- based
reagent
Mouse Fmo3
Sigma
PCR primers
Mouse Taar5
Sigma
PCR primers
Mouse CyclophilinA
Sigma
PCR primers
Sequence- based
reagent
Mouse Cxcl1
Sequence- based
reagent
Mouse Grp78
IDT
IDT
PCR primers
PCR primers
Commercial
assay or kit
Commercial
assay or kit
AST Commercial Kit
Sekisui Diagnostics
319–30
ALT Commercial Kit
Sekisui Diagnostics
318–30
Commercial
assay or kit
Triglyceride Commercial
Kit
Wako
994–02891
Commercial
assay or kit
Total Cholesterol
Commercial Kit
Fisher Scientific
TR134321
Commercial
assay or kit
Free Cholesterol
Commercial Kit
Wako
Commercial
assay or kit
RNAeasy Lipid Tissue
Mini Kit
Qiagen
993–02501
74804
Commercial
assay or kit
Commercial
assay or kit
Thermo Scientific Pierce
TiO2 Phosphopeptide
Enrichment and Clean-
up Kit
Fisher Scientific
PI88301
RNAeasy Purification Kit Qiagen
74004
Chemical
compound, drug Iodomethylcholine (IMC)
Synthesized at the Cleveland
Clinic
Not provided
Chemical
compound, drug
Fluoromethylcholine
(FMC)
Synthesized at the Cleveland
Clinic
Not provided
Chemical
compound, drug
Trimethylamine
Hydrochloride
Chemical
compound, drug Lipopolysaccharide
Sigma
Sigma
Continued on next page
T72761
L4391
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
15 of 28
Medicine
Research article
Continued
Reagent type
(species) or
resource
Software,
algorithm
Software,
algorithm
Software,
algorithm
Software,
algorithm
Software,
algorithm
Software,
algorithm
Other
Other
Designation
Source or reference
Identifiers
Additional information
GraphPad Prism
GraphPad Software, Inc
8.4
DADA2
Phyloseq
microbiomeSeq
Ggplot2
vegan
https://benjjneb.
github.io/dada2/
dada-installation.html;
Callahan et al., 2016
https://www.bioconductor.
org/packages/release/
bioc/html/phyloseq.html
https://github.com/
umerijaz/microbiomeSeq
https://cran.r-project.
org/web/
packages/
ggplot2/index.html
1.16
4.1, RRID:SCR_013080
1: RRID:SCR_002630
3.3.5, RRID:SCR_014601
https://cran.r-project.org/web/
packages/
vegan/index.html
2.5–7
Supersignal West Pico
Plus Substrate
Thermo Fisher
Diet
Dyets
34577
710260
Overview of human study populations
We made use of three different human study populations, detailed below, that included patients with
severities of AH/ALD. It must be noted that one limitation of our study is that each of these cohorts
used slightly different diagnostic criteria for defining the severity/stage of AH/ALD.
Human study populations and sample collection for TMA measurement
A total of 285 subjects were included in this study. De- identified plasma samples, along with clinical
and demographic data, were obtained from (1) the Northern Ohio Alcohol Center (NOAC) at the
Cleveland Clinic biorepository including 21 healthy individuals and 15 patients diagnosed and (2) the
Defeat Alcoholic Steatohepatitis (DASH) consortium (Cleveland Clinic, University of Louisville School of
Medicine, University of Massachusetts Medical School, and University of Texas Southwestern Medical
Center) including 249 patients with AH. Diagnosis with AH was performed using clinical and labora-
tory criteria, with MELD score utilized for distinguishing moderate (MELD < 20) and severe (MELD >
20) AH, as recommended by the NIAAA Alcoholic Hepatitis consortia (Crabb et al., 2016). A detailed
description of patient recruitment, inclusion and exclusion criteria for the DASH consortium has been
reported in previous studies (Vatsalya et al., 2020). Patients with AH were classified as moderate
(MELD < 20, n = 112) and severe (MELD ≥ 20, n = 152) according to the MELD score at admission
as part of either of two independent clinical trials (ClincalTrials.gov identifier # NCT01809132 and
NCT03224949) or the NOAC biorepository. These studies were approved by the Institutional Review
Boards of all four participating institutions and all study participants consented prior to collection of
data and blood samples. Clinical and demographic data for the entire cohort is presented in Figure
1—source data 1 and for the sub- set of subjects used for TMA analysis is presented in Figure 1—
source data 2.
In order to be able to measure volatile compounds such as TMA, blood was collected in EDTA-
coated tubes and immediately placed on ice. Plasma was separated by centrifugation at 1200× g for
15 min at 4°C. Plasma was rapidly acidified by adding 25 mL of 1 M hydrochloric acid (HCL) to 500 mL
of aliquoted plasma, followed by vigorous vortexing. Acidified plasma samples were stored at –80°C
in air- tight O- ring cryovials (Fisher Scientific, product # 02- 681- 373) until being processed for quantifi-
cation of TMA and other volatile compounds. A non- acidified sample was also collected for standard
plasma biochemistries.
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
16 of 28
Medicine
Research article
Figure 7. Graphical summary depicting the proposed role of trimethylamine (TMA) in the progression of alcohol- associated liver disease (ALD). Gut
microbiota can elicit both metabolism- dependent and metabolism- independent effects in ALD. Relevant to this manuscript, intestinal microbes
metabolize dietary L- carnitine, choline, or phosphatidylcholine (PC) to form TMA, which is a volatile compound that originates exclusively from
gut bacterial metabolism and is elevated in ALD. Importantly, TMA can also be converted to trimethylamine N- oxide (TMAO) by hepatic flavin
monooxygenase 3 (FMO3), and TMAO has recently been linked to cardiovascular disease (CVD) promotion in humans. Metabolism- independent effects
are the result of gut hyperpermeability (leaky gut), allowing bacterial cell wall products such as lipopolysaccharide (LPS) and peptidoglycans to enter
into the blood stream and engage with host pattern recognition receptors (PRR) to promote hepatic inflammation. Collectively, metabolism- dependent
pathways such as TMA production as well as metabolism- independent pathways provide multiple bacterially derived ‘hits’ to promote ALD progression.
The small molecule bacterially targeted CutC/D inhibitors iodomethylcholine (IMC) and fluoromethylcholine (FMC) can effectively blunt ethanol- induced
liver injury in mice.
Analysis of hepatic FMO3 expression across different liver disease
etiologies
For data shown in main Figure 1 panel C, we leveraged access to publicly available bulk liver RNA
sequencing data from patients with different liver disease etiologies (Argemi et al., 2019). For this
cohort, early AH (EAH) was defined as MELD 7–8, severe AH with liver failure (AHL) with MELD 22–28,
and AHL with emergency liver transplant (ExAH) with MELD 18–21. All raw fastq files were down-
loaded from SRA (PRJNA531223) and dbGAP (phs001807.v1.p1) (Argemi et al., 2019). Fastq files
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
17 of 28
Medicine
Research article
were aligned to the human genome (GRCh38, indices downloaded from https://github.com/pach-
terlab/kallisto-transcriptome-indices/releases/download/ensembl-96/homo_sapiens.tar.gz; Pachter,
2018) using Kallisto version 0.44.0 with 100 bootstraps calculated (Bray et al., 2016). Data were then
merged with clinical data and analyzed with Sleuth in gene_mode with aggregation_column set to
Ensemble Gene ID; in addition, extra_bootstrap_summary and read_bootstrap_tpm were set to true
(Pimentel et al., 2017). Differential expression was measured with Sleuth using a cutoff of q < 0.05.
Human study populations and sample collection for liver Western
blotting
De- identified samples from five livers explanted from severe AH patients during liver transplantation
or five wedge biopsies from healthy donor livers were snap- frozen in liquid nitrogen and stored at
–80°C. Samples were provided by the Clinical Resource for Alcoholic Hepatitis Investigations at Johns
Hopkins University (R24 AA0025107, Z. Sun PI). Written informed consent was obtained from each
patient included in the study and the study protocol conforms to the ethical guidelines of the 1975
Declaration of Helsinki as reflected in a priori approval by the Institutional Review Boards at Johns
Hopkins Medical Institutions. This cohort utilized Maddrey’s Discriminant Function as the primary
indicator of disease severity, with an average score of 102.5 ± 27.7. MELD scores are not available for
this cohort. Descriptive biochemical and clinical data for this cohort have been reported previously
(Tripathi et al., 2018).
Immunoblotting
Whole tissue homogenates were made from tissues in a modified RIPA buffer as previously described
(Warrier et al., 2015; Helsley et al., 2019; Schugar et al., 2017; Lord et al., 2016), and protein was
quantified using the bicinchoninic assay (Pierce). Proteins were separated by 4–12% SDS- PAGE, trans-
ferred to polyvinylidene difluoride membranes, and then proteins were detected after incubation with
specific antibodies as previously described (Warrier et al., 2015; Helsley et al., 2019; Schugar et al.,
2017; Lord et al., 2016) and listed in the Key resources table.
Real-time PCR analysis of gene expression
Tissue RNA extraction and qPCR analysis was performed as previously described (Helsley et al.,
2019). The mRNA expression levels were calculated based on the ΔΔ-CT method using cyclophilin A
as the housekeeping gene. qPCR was conducted using the Applied Biosystems 7500 Real- Time PCR
system. All primer sequences are listed in the Key resources table.
Chemical synthesis of gut microbe-targeted choline TMA lyase
inhibitors
The small molecule choline TMA lyase inhibitors IMC and FMC have been previously described as
potent and selective mechanism- based inhibitors targeted microbial CutC (Roberts et al., 2018).
Here, IMC and FMC were synthesized and structurally characterized as outlined below using both
multinuclear NMR analysis and high- resolution mass spectrometry. 1H- and 13C- NMR spectra for
IMC and FMC were recorded on a Bruker Ascend spectrometer operating at 400 MHz. Chemical
shifts are reported as parts per million (ppm). IMC iodide was prepared using a previously reported
method using 2- dimethylethanolamine and diiodomethane as reactants in acetonitrile followed by
recrystallization from dry ethanol. 1H- and 13C- NMRs of IMC were both consistent with that in the
reported literature (Mistry et al., 2002), as well as consistent based on proton and carbon chem-
ical shift assignments indicated below. High- resolution MS corroborated the expected cation mass
and provided further evidence of structural identity. 1H- NMR (400 MHz, D2O): 5.24 (s, 2H, -N- CH2- I),
4.06–3.99 (m, 2H, -CH2- CH2- OH), 3.68–3.62 (m, 2H, -N- CH2- CH2-), 3.29 (s, 6H, -N(CH3)2); 13C- NMR (100
MHz, D2O): 66.1 (- CH2- CH2- OH), 55.8 (- N- CH2- CH2-), 52.9 (- N(CH3)2), 33.0 (- N- CH2- I); HRMS (ESI/TOF):
m/z (M+) calculated for C5H13INO, 230.0036; found, 230.0033. The synthesis of fluoromethylcholine
chloride was performed using the procedure below. 1H- and 13C- NMRs of FMC were consistent with
that in the reported literature (Gao et al., 2019). High- resolution MS was also consistent with the
expected cation mass. Chloro(fluoro)methane (2.05 kg, 29.9 mol, 6 eq) was bubbled into a solution
of 2- dimethylaminoethanol (444.0 g, 4.98 mol, 500 mL, 1 eq) in THF (1000 mL) at –70°C for 4 hr. The
mixture was then transferred to an autoclave and heated to 80°C and stirred for 18 hr (pressure:
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
18 of 28
Medicine
Research article
~15–50 psi). During this period, a white precipitate formed. The solid was isolated by filtration, washed
with cold THF (600 mL), and dried under vacuum to give fluoromethylcholine chloride as a white solid
(1.14 kg, 70.7% yield, 98.0% purity). 1H- NMR (400 MHz, D2O): 5.44 (s, 1H, -N- CH2- F), 5.32 (s, 1H,
-N- CH2- F), 4.04–3.98 (m, 2H, -CH2- CH2- OH), 3.60–3.54 (m, 2H, -N- CH2- CH2-), 3.19 (s, 6H, -N(CH3)2);
13C- NMR (100 MHz, D2O): 97.8 and 95.6 (- N- CH2- F), 62.9 (- CH2- CH2- OH), 55.1 (- N- CH2- CH2-), 48.0
(- N(CH3)2); HRMS (ESI/TOF): m/z (M+) calculated for C5H13FNO (M+) 122.0976, found 122.0975.
Ethanol feeding trials in mice
All mice were maintained in an Association for the Assessment and Accreditation of Laboratory
Animal Care, International- approved animal facility. All experimental protocols were approved by
the Institutional Animal Care and Use Committee (IACUC) at the Cleveland Clinic. Age- and weight-
matched female C57BL6/J mice were randomized into pair- and ethanol- fed groups and adapted to
control liquid diet for 2 days. Two models of chronic ethanol feeding were used. (1) A 25- day chronic
model in which mice were allowed free access to increasing concentrations of ethanol for 25 days
(i.e. chronic feeding model) as previously described (McCullough et al., 2018). In this model, the
ethanol- fed mice were acclimated to ethanol as follows: 1% vol/vol for 2 days, 2% vol/vol for 2 days,
4% vol/vol (22% kcal) for 1 week, 5% vol/vol (27% kcal) for 1 week, and last 6% vol/vol (32% kcal) for
1 week and is denoted as 32%, day 25. (2) A 10- day chronic model in which mice were allowed free
access to a 5% vol/vol (27% kcal) for 10 days (Bertola et al., 2013). Ethanol- fed mice were allowed ad
libitum access to liquid diet. Control mice were pair- fed a diet that received isocalorically substituted
maltose dextrin for ethanol. Some cohorts received choline TMA lyase inhibitors IMC (0.06% wt/wt) or
FMC (0.006% wt/wt) in these liquid diets throughout the entire 10- to 25- day feeding period. Lieber-
DeCarli high- fat ethanol and control diets were purchased from Dyets (catalog number 710260; Beth-
lehem, PA).
LPS injections
Female C57BL6/J mice at 10.5 weeks of age were injected intraperitoneally with either 15 mg/kg LPS
(500 µg/mL, Sigma L4391) or a matched volume (30 mL/kg) of sterile saline. After 6 hr, mice were euth-
anized with ketamine/xylazine and the liver was immediately collected and homogenized in TRIzol.
RNA was extracted using chloroform phase separation and purified using Qiagen RNeasy kit.
Liver histology and immunohistochemistry
For histological analysis, formalin- fixed tissues were paraffin embedded, sectioned, and stained with
hematoxylin and eosin. Formalin- fixed samples are coded at the time of collection for blinded analysis.
Measurement of plasma aminotransferase levels
To determine the level of hepatic injury in mice, plasma was used to quantify ALT and AST levels using
a commercially available enzymatic assay (Sekisui Diagnostics, Lexington, MA) according to manufac-
turer’s instruction.
Measurement of hepatic lipid levels
Extraction of liver lipids and quantification of total plasma and hepatic triglycerides, cholesterol, and
cholesterol esters was conducted using enzymatic assays as described previously (Warrier et al.,
2015; Helsley et al., 2019; Schugar et al., 2017; Lord et al., 2016).
Quantification of TMA-related metabolites in acidified plasma
Stable isotope dilution high- performance liquid chromatography with on- line tandem mass spectrom-
etry (LC- MS/MS) was used for quantification of levels of TMAO, TMA, choline, carnitine, betaine, and
γ-butyrobetaine in plasma, as previously described (Wang et al., 2014a). Their d9(methyl) isotopo-
logues were used as internal standards. LC- MS/MS analyses were performed on a Shimadzu 8050
triple quadrupole mass spectrometer. IMC and d2- IMC, along with other metabolites, were moni-
tored using multiple reaction monitoring of precursor and characteristic product ions as follows: m/z
230.0 → 58.0 for IMC; m/z 232.0 → 60.1 for d2- IMC; m/z 76.0 → 58.1 for TMAO; m/z 85.0 → 66.2 for
d9- TMAO; m/z 60.2 → 44.2 for TMA; m/z 69.0 → 49.1 for d9- TMA; m/z 104.0 → 60.1 for choline; m/z
113.1 → 69.2 for d9- choline; m/z 118.0 → 58.1 for betaine; m/z 127.0 → 66.2 for d9- betaine.
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
19 of 28
Medicine
Research article
Cecal microbiome analyses
16S rRNA amplicon sequencing were done for V4 region using via miSEQ from mouse cecal contents.
Raw 16S amplicon sequence and metadata were demultiplexed using split_ libraries_ fastq. py script
implemented in QIIME1.9.1 (Caporaso et al., 2010). Demultiplexed fastq file was split into sample
specific fastq files using split_ sequence_ file_ on_ sample_ ids. py script from Qiime1.9.1 (Caporaso
et al., 2010). Individual fastq files without non- biological nucleotides were processed using Divisive
Amplicon Denoising Algorithm (DADA) pipeline (Callahan et al., 2016). The output of the dada2
pipeline (feature table of amplicon sequence variants [an ASV table]) was processed for alpha and beta
diversity analysis using phyloseq (McMurdie and Holmes, 2013), and microbiomeSeq (http://www.
github.com/umerijaz/microbiomeSeq) packages in R. Alpha diversity estimates were measured within
group categories using estimate_richness function of the phyloseq package (McMurdie and Holmes,
2013). Multidimensional scaling (also known as principal coordinate analysis [PCoA]) was performed
using Bray- Curtis dissimilarity matrix (Knorr et al., 2020) between groups and visualized by using
ggplot2 package (Wickham, 2009). We assessed the statistical significance (p < 0.05) throughout and
whenever necessary, we adjusted p- values for multiple comparisons according to the Benjamini and
Hochberg method to control false discovery rate (Benjamini, 2010) while performing multiple testing
on taxa abundance according to sample categories. We performed an analysis of variance (ANOVA)
among sample categories while measuring the of alpha diversity measures using plot_anova_diver-
sity function in microbiomeSeq package (http://www.github.com/umerijaz/microbiomeSeq). Permu-
tational multivariate analysis of variance (PERMANOVA) with 999 permutations was performed on
all principal coordinates obtained during PCoA with the ordination function of the microbiomeSeq
package. Wilcoxon (non- parametric) test was performed on ASV’s abundances against metadata vari-
ables levels using their base functions in R (Tilt, 1999).
RNA sequencing in mouse tissues
RNA sequencing libraries were generated from mouse liver using the Illumina mRNA TruSeq Direc-
tional library kit and sequenced using an Illumina HiSeq4000 (both according to the manufacturer’s
instructions). RNA sequencing was performed by the University of Chicago Genomics Facility, and data
analysis and data availability are described in detail in the online supplement. Briefly, RNA samples
were checked for quality and quantity using the Bio- analyzer (Agilent). RNA sequencing libraries were
generated using the Illumina mRNA TruSEQ Directional library kit and sequenced using an Illumina
HiSEQ4000 (both according to the manufacturer’s instructions). RNA sequencing was performed by
the University of Chicago Genomics Facility. Raw sequence files will be deposited in the Sequence
Read Archive before publication (SRA). Single- end 100 bp reads were trimmed with Trim Galore
(v.0.3.3, https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/. ) and controlled for quality
with FastQC (v0.11.3, http://www.bioinformatics.bbsrc.ac.uk/projects/fastqc) before alignment to the
Mus musculus genome (Mm10 using UCSC transcript annotations downloaded July 2016). Reads
were aligned using the STAR alignerSTAR in single pass mode (v.2.5.2a_modified, RRID:SCR_004463,
https://github.com/alexdobin/STAR) with standard parameters. Raw counts were loaded into R
(http://www.R-project.org/) and edgeR was used to perform upper quantile, between- lane normaliza-
tion, and DE analysis. Values generated with the cpm function of edgeR, including library size normal-
ization and log2 conversion, were used in figures. Heat maps were generated of top 50 differentially
expressed transcripts using pheatmap. Reactome- based pathway analysis was performed using an
open- sourced R package: ReactomePA. RNA sequencing data have been deposited into the National
Institutes of Health (NIH)- sponsored GEO repository (accession number GSE157681).
Phosphoproteomics analyses to examine TMA-induced signaling events
in mouse liver
The goal of this experiment was to unbiasedly identify TMA- responsive signaling events in mouse liver
after an acute exposure (10 min) of TMA. To closely mimic physiological route of delivery, we delivered
saline or TMA directly into the portal vein in fasted mice. Briefly, C57BL/6 mice were fasted overnight
(12 hr fast), and between the hours of 9:00–10:00 am (2–3 hr into light cycle), mice were anesthetized
using isoflurane (4% for induction and 2% for maintenance). Once fully anesthetized, a midline lapa-
rotomy was performed, and the portal vein was visualized under a Leica M650 surgical microscope.
Briefly, a fresh 10 mM stock of trimethylamine hydrochloride (TMA- HCL) made in sterile saline, and the
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
20 of 28
Medicine
Research article
pH of stock solution was adjusted to 7.4. Mice then received 20 μL of either saline vehicle or TMA- HCL
via direct syringe infusion (Becton- Dickson product #309306); 9.75 min later a small aliquot (50 μL)
of portal blood was collected by pulling back on injection syringe left in place following injection. In
saline vehicle injected mice, portal blood levels of TMA ranged from 0.49 to 2.22 μM and TMAO levels
ranged from 2.53 to 7.14 μM. In mice injected with TMA- HCL, portal blood levels of TMA ranged
from 125.36 to 319.55 μM and TMAO levels ranged from 9.68 to 17.48 μM. Exactly 10 min after initial
injection, the liver was rapidly snap- frozen by immersion in liquid nitrogen. Liver samples were homog-
enized, the protein was precipitated with acetone, and the protein concentration was measured. A
total of 1 mg of protein from each sample was digested with trypsin and the resulting tryptic peptides
were subjected to phosphoserine and phosphothreonine enrichment using the Thermo Scientific
Pierce TiO2 Phosphopeptide Enrichment and Clean- up Kit (Fisher # PI88301). The enrichment was
performed based on the manufacturer’s instructions. The enriched peptide samples were subjected
to C18 clean- up prior to LC- MS analysis. The LC- MS system was a Finnigan LTQ- Obitrap Elite hybrid
mass spectrometer system. The HPLC column was a Dionex 15 cm × 75 µm id Acclaim Pepmap C18,
2 μm, 100 Å reversed- phase capillary chromatography column. Five μL volumes of the extract were
injected and the peptides eluted from the column by an acetonitrile/0.1% formic acid gradient at
a flow rate of 0.25 μL/min were introduced into the source of the mass spectrometer on- line. The
microelectrospray ion source is operated at 1.9 kV. The digest was analyzed using the data- dependent
multitask capability of the instrument acquiring full scan mass spectra to determine peptide molecular
weights and product ion spectra to determine amino acid sequence in successive instrument scans.
The LC- MS/MS data files were searched against the mouse UnitProtKB database (downloaded in
December 2019 contains 17,017 sequences) using Sequest bundled into Proteome Discoverer 2.4.
Cysteine carbamidomethylation was set as a fixed modification and oxidized methionine, protein
N- terminal acetylation, and phosphorylation of serine, threonine, and tyrosine were considered as
dynamic modification. A maximum of two missed cleavages were permitted. The peptide and protein
false discovery rates were set to 0.01 using a target- decoy strategy. Phosphorylation sites were identi-
fied using ptmRS node in PD2.4. The relative abundance of the positively identified phosphopeptides
was determined using the extracted ion intensities (Minora Feature Detection node) with Retention
time alignment. All peptides were included in the quantitation, the peptide intensities were normal-
ized to total peptide amount. Missing values were imputed in Perseus using a normal distribution. A
total of 789 phosphopeptides were identified with 36 phosphopeptides determined to be two- fold
different in the TMA and saline samples with a p- value < 0.05 (t- test).
Statistical analysis
All statistical analyses were performed using GraphPad Prism and p < 0.05 was considered statistically
significant. All data are presented as mean ± SEM, unless otherwise noted in the figure legends. All
data were tested for equal variance and normality. For two- group comparison of parametric data,
a two- tailed Student’s t- test was performed, while non- parametric data were analyzed with Mann-
Whitney U test (also called the Wilcoxon rank- sum test). For studies comparing vehicle and TMA lyase
inhibitors in pair- and ethanol- fed mice, a two- way ANOVA was performed, followed by Tukey’s tests
for post hoc analysis. For human studies in AH patients, statistical significance was determined by
ANOVA and a Tukey’s honest significant difference post hoc test (p < 0.05).
Acknowledgements
This work was supported in part by National Institutes of Health grants P50 AA024333 (AJM, SD,
DSA, LEN, JMB), R01 DK120679 (JMB), P01 HL147823 (JMB, SLH), U01 AA026938 (LEN, JMB), P50
CA150964 (JMB), U01 AA021890 (LEN, SD), U01 AA021893 (SD, BB, CJM, MM, GS, and AJM), R01
HL103866 (SLH), R01 HL144651 (ZW), R01 HL130819 (ZW), U01 AA026980 (CJM), P50 AA 024337
(CJM), R21 AR 071046 (SD), R01 GM119174 (SD), R01 DK113196 (SD), R56 HL141744 (SD), U01
DK061732 (SD), U01 AA026977 (GS), UH3 AA026970 (GS), K99 AA028048 (AK), a Leducq Transatlantic
Networks of Excellence Award (SLH), a JSPS Overseas Research Fellowship 201960331 (TM), and the
American Heart Association (Postdoctoral Fellowships 17POST3285000 to RNH and 15POST2535000
to RCS). The Orbitrap Elite instrument used for proteomics was purchased via an NIH shared instru-
ment grant 1S10RR031537 (BW).
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
21 of 28
Medicine
Research article
Additional information
Competing interests
Zeneng Wang: Kaiser Permanente (CME lecture sessions) Advisory Board for Incyte (on treatment
of cholangiocarcinoma). Stanley L Hazen: Z.W. report being named as co- inventor on pending and
issued patents held by the Cleveland Clinic relating to cardiovascular diagnostics and therapeutics.
Z.W. reports being eligible to receive royalty payments for inventions or discoveries related to cardio-
vascular diagnostics or therapeutics from Zehna Therapeutics, Cleveland Heart Lab, a wholly owned
subsidiary of Quest Diagnostics, and Procter & Gamble. The other authors declare that no competing
interests exist.
Funding
Funder
Grant reference number Author
National Institutes of
Health
P50 AA024333
Arthur J McCullough
Srinivasan Dasarathy
Daniela S Allende
Laura E Nagy
Jonathan Mark Brown
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
R01 DK120679
Jonathan Mark Brown
P01 HL147823
U01 AA026938
Jonathan Mark Brown
Stanley L Hazen
Laura E Nagy
Jonathan Mark Brown
P50 CA150964
Jonathan Mark Brown
U01 AA021890
U01 AA021893
Laura E Nagy
Srinivasan Dasarathy
Srinivasan Dasarathy
Bruce Barton
Craig J McClain
Marko Mrdjen
Gyongyi Szabo
Arthur J McCullough
R01 HL103866
Stanley L Hazen
R01 HL144651
Zeneng Wang
R01 HL130819
Zeneng Wang
U01 AA026980
Craig J McClain
P50 AA 024337
Craig J McClain
R21 AR 071046
Srinivasan Dasarathy
R01 GM119174
Srinivasan Dasarathy
R01 DK113196
Srinivasan Dasarathy
R56 HL141744
Srinivasan Dasarathy
U01 DK061732
Srinivasan Dasarathy
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
22 of 28
Medicine
Research article
Funder
Grant reference number Author
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
National Institutes of
Health
JSPS Overseas Research
Fellowship
American Heart
Association
American Heart
Association
U01 AA026977
Gyongyi Szabo
UH3 AA026970
Gyongyi Szabo
K99 AA028048
Anagha Kadam
1S10RR031537
Belinda Willard
Leducq Transatlantic
Networks of Excellence
Award
Stanley L Hazen
201960331
Tatsunori Miyata
17POST3285000
Robert N Helsley
15POST2535000
Rebecca C Schugar
The funders had no role in study design, data collection and interpretation, or the
decision to submit the work for publication.
Author contributions
Robert N Helsley, Chase Neumann, Lucas J Osborn, Rebecca C Schugar, Megan R McMullen, Annette
Bellar, Kyle L Poulsen, Adam Kim, Vai Pathak, Marko Mrdjen, James T Anderson, Belinda Willard,
Craig J McClain, Mack Mitchell, Arthur J McCullough, Svetlana Radaeva, Bruce Barton, Gyongyi
Szabo, Srinivasan Dasarathy, Jose Carlos Garcia- Garcia, Daniel M Rotroff, Zeneng Wang, Stanley
L Hazen, Laura E Nagy, Jonathan Mark Brown, Conceptualization, Data curation, Formal analysis,
Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Super-
vision, Validation, Visualization, Writing – original draft, Writing – review and editing; Tatsunori Miyata,
Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualiza-
tion, Writing – original draft, Writing – review and editing; Anagha Kadam, Data curation, Formal
analysis, Investigation, Methodology, Writing – original draft, Writing – review and editing; Venkatesh-
wari Varadharajan, Naseer Sangwan, Emily C Huang, Rakhee Banerjee, Amanda L Brown, Conceptu-
alization, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing
– review and editing; Kevin K Fung, Conceptualization, Data curation, Formal analysis, Investigation,
Methodology, Project administration, Supervision, Validation, Visualization, Writing – original draft,
Writing – review and editing; William J Massey, Conceptualization, Data curation, Formal analysis,
Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Vali-
dation, Visualization, Writing – original draft, Writing – review and editing; Danny Orabi, Conceptual-
ization, Data curation, Formal analysis, Writing – original draft, Writing – review and editing; Daniela
S Allende, Data curation, Formal analysis, Methodology, Visualization, Writing – original draft, Writing
– review and editing
Author ORCIDs
Robert N Helsley
William J Massey
Bruce Barton
Srinivasan Dasarathy
Jonathan Mark Brown
http://orcid.org/0000-0001-5000-3187
http://orcid.org/0000-0002-2087-6048
http://orcid.org/0000-0001-7878-8895
http://orcid.org/0000-0003-1774-0104
http://orcid.org/0000-0003-2708-7487
Ethics
Clinical trial registration NCT01809132; NCT03224949.
Human subjects: Patients with AH were classified as moderate (MELD < 20, n=112) and severe (MELD
≥20, n=152) according to the MELD score at admission as part of either of two independent clinical
trials ( ClincalTrials. gov identifier # NCT01809132 and NCT03224949) or the NOAC biorepository.
These studies were approved by the Institutional Review Boards of all 4 participating institutions
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
23 of 28
Medicine
Research article
and all study participants consented prior to collection of data and blood samples. Written informed
consent was obtained from each patient included in the study and the study protocol conforms to the
ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the Institu-
tional Review Boards at Johns Hopkins Medical Institutions.
All mice were maintained in an Association for the Assessment and Accreditation of Laboratory Animal
Care, International- approved animal facility. All experimental protocols were approved by the institu-
tional animal care and use committee (IACUC) at the Cleveland Clinic.
Decision letter and Author response
Author response https://doi.org/10.7554/eLife.76554.sa2
Additional files
Supplementary files
• Transparent reporting form
Data availability
Sequencing data have been deposited in GEO under accession code GSE157681.
The following dataset was generated:
Author(s)
Brown JM, Helsley R,
Kadam A, Neumann
C
Year
2021
Dataset title
Dataset URL
Database and Identifier
The Gut Microbe- Derived
Metabolite Trimethylamine
is a Biomarker of and
Therapeutic Target in
Alcohol- Associated Liver
Disease
http://www. ncbi.
nlm. nih. gov/ geo/
query/ acc. cgi? acc=
GSE157681
NCBI Gene Expression
Omnibus, GSE157681
References
Araki E, Lipes MA, Patti ME, Brüning JC, Haag B, Johnson RS, Kahn CR. 1994. Alternative pathway of insulin
signalling in mice with targeted disruption of the IRS- 1 gene. Nature 372:186–190. DOI: https://doi.org/10.
1038/372186a0, PMID: 7526222
Argemi J, Latasa MU, Atkinson SR, Blokhin IO, Massey V, Gue JP, Cabezas J, Lozano JJ, Van Booven D, Bell A,
Cao S, Vernetti LA, Arab JP, Ventura- Cots M, Edmunds LR, Fondevilla C, Stärkel P, Dubuquoy L, Louvet A,
Odena G, et al. 2019. Defective HNF4alpha- dependent gene expression as a driver of hepatocellular failure in
alcoholic hepatitis. Nature Communications 10:3126. DOI: https://doi.org/10.1038/s41467-019-11004-3, PMID:
31311938
Ascha M, Wang Z, Ascha MS, Dweik R, Zein NN, Grove D, Brown JM, Marshall S, Lopez R, Hanouneh IA.
2016. Metabolomics studies identify novel diagnostic and prognostic indicators in patients with alcoholic
hepatitis. World Journal of Hepatology 8:499–508. DOI: https://doi.org/10.4254/wjh.v8.i10.499, PMID:
27057307
Asgharian H, Homayouni- Rad A, Mirghafourvand M, Mohammad- Alizadeh- Charandabi S. 2020. Effect of
probiotic yoghurt on plasma glucose in overweight and obese pregnant women: a randomized controlled
clinical trial. European Journal of Nutrition 59:205–215. DOI: https://doi.org/10.1007/s00394-019-01900-1,
PMID: 31069459
Aslan JE. 2019. Platelet Rho GTPase regulation in physiology and disease. Platelets 30:17–22. DOI: https://doi.
org/10.1080/09537104.2018.1475632, PMID: 29799302
Benjamini Y. 2010. Discovering the false discovery rate. Journal of the Royal Statistical Society 72:405–416. DOI:
https://doi.org/10.1111/j.1467-9868.2010.00746.x
Bertola A, Mathews S, Ki SH, Wang H, Gao B. 2013. Mouse model of chronic and binge ethanol feeding (the
NIAAA model). Nature Protocols 8:627–637. DOI: https://doi.org/10.1038/nprot.2013.032, PMID: 23449255
Boutagy NE, Neilson AP, Osterberg KL, Smithson AT, Englund TR, Davy BM, Hulver MW, Davy KP. 2015.
Short- term high- fat diet increases postprandial trimethylamine- N- oxide in humans. Nutrition Research (New
York, N.Y.) 35:858–864. DOI: https://doi.org/10.1016/j.nutres.2015.07.002, PMID: 26265295
Bray NL, Pimentel H, Melsted P, Pachter L. 2016. Near- optimal probabilistic RNA- seq quantification. Nature
Biotechnology 34:525–527. DOI: https://doi.org/10.1038/nbt.3519, PMID: 27043002
Brown JM, Hazen SL. 2015. The gut microbial endocrine organ: bacterially derived signals driving
cardiometabolic diseases. Annual Review of Medicine 66:343–359. DOI: https://doi.org/10.1146/annurev-med-
060513-093205, PMID: 25587655
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
24 of 28
Medicine
Research article
Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJA, Holmes SP. 2016. DADA2: High- resolution sample
inference from Illumina amplicon data. Nature Methods 13:581–583. DOI: https://doi.org/10.1038/nmeth.3869,
PMID: 27214047
Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello EK, Fierer N, Peña AG, Goodrich JK,
Gordon JI, Huttley GA, Kelley ST, Knights D, Koenig JE, Ley RE, Lozupone CA, McDonald D, Muegge BD,
Pirrung M, Reeder J, et al. 2010. QIIME allows analysis of high- throughput community sequencing data. Nature
Methods 7:335–336. DOI: https://doi.org/10.1038/nmeth.f.303, PMID: 20383131
Cashman JR. 2002. Human flavin- containing monooxygenase (form 3): polymorphisms and variations in chemical
metabolism. Pharmacogenomics 3:325–339. DOI: https://doi.org/10.1517/14622416.3.3.325, PMID: 12052141
Chen Y, Yang F, Lu H, Wang B, Chen Y, Lei D, Wang Y, Zhu B, Li L. 2011. Characterization of fecal microbial
communities in patients with liver cirrhosis. Hepatology (Baltimore, Md.) 54:562–572. DOI: https://doi.org/10.
1002/hep.24423, PMID: 21574172
Chen S, Henderson A, Petriello MC, Romano KA, Gearing M, Miao J, Schell M, Sandoval- Espinola WJ, Tao J,
Sha B, Graham M, Crooke R, Kleinridders A, Balskus EP, Rey FE, Morris AJ, Biddinger SB. 2019. Trimethylamine
N- Oxide Binds and Activates PERK to Promote Metabolic Dysfunction. Cell Metabolism 30:1141-1151.. DOI:
https://doi.org/10.1016/j.cmet.2019.08.021, PMID: 31543404
Cho S, Hoang A, Sinha R, Zhong XY, Fu XD, Krainer AR, Ghosh G. 2011. Interaction between the RNA binding
domains of Ser- Arg splicing factor 1 and U1- 70K snRNP protein determines early spliceosome assembly. PNAS
108:8233–8238. DOI: https://doi.org/10.1073/pnas.1017700108, PMID: 21536904
Ciocan D, Rebours V, Voican CS, Wrzosek L, Puchois V, Cassard AM, Perlemuter G. 2018. Characterization of
intestinal microbiota in alcoholic patients with and without alcoholic hepatitis or chronic alcoholic pancreatitis.
Scientific Reports 8:4822. DOI: https://doi.org/10.1038/s41598-018-23146-3, PMID: 29555983
Crabb DW, Bataller R, Chalasani NP, Kamath PS, Lucey M, Mathurin P, McClain C, McCullough A, Mitchell MC,
Morgan TR, Nagy L, Radaeva S, Sanyal A, Shah V, Szabo G, NIAAA Alcoholic Hepatitis Consortia. 2016.
Standard Definitions and Common Data Elements for Clinical Trials in Patients With Alcoholic Hepatitis:
Recommendation From the NIAAA Alcoholic Hepatitis Consortia. Gastroenterology 150:785–790. DOI: https://
doi.org/10.1053/j.gastro.2016.02.042, PMID: 26921783
Craciun S, Balskus EP. 2012. Microbial conversion of choline to trimethylamine requires a glycyl radical enzyme.
PNAS 109:21307–21312. DOI: https://doi.org/10.1073/pnas.1215689109, PMID: 23151509
DeSantis DA, Ko CW, Liu Y, Liu X, Hise AG, Nunez G, Croniger CM. 2013. Alcohol- induced liver injury is
modulated by Nlrp3 and Nlrc4 inflammasomes in mice. Mediators of Inflammation 2013:751374. DOI: https://
doi.org/10.1155/2013/751374, PMID: 24453428
Duan Y, Llorente C, Lang S, Brandl K, Chu H, Jiang L, White RC, Clarke TH, Nguyen K, Torralba M, Shao Y, Liu J,
Hernandez- Morales A, Lessor L, Rahman IR, Miyamoto Y, Ly M, Gao B, Sun W, Kiesel R, et al. 2019.
Bacteriophage targeting of gut bacterium attenuates alcoholic liver disease. Nature 575:505–511. DOI: https://
doi.org/10.1038/s41586-019-1742-x, PMID: 31723265
Gao B, Lang S, Duan Y, Wang Y, Shawcross DL, Louvet A, Mathurin P, Ho SB, Stärkel P, Schnabl B. 2019. Serum
and Fecal Oxylipins in Patients with Alcohol- Related Liver Disease. Digestive Diseases and Sciences 64:1878–
1892. DOI: https://doi.org/10.1007/s10620-019-05638-y, PMID: 31076986
Gilbert JA, Blaser MJ, Caporaso JG, Jansson JK, Lynch SV, Knight R. 2018. Current understanding of the human
microbiome. Nature Medicine 24:392–400. DOI: https://doi.org/10.1038/nm.4517, PMID: 29634682
Gupta N, Buffa JA, Roberts AB, Sangwan N, Skye SM, Li L, Ho KJ, Varga J, DiDonato JA, Tang WHW, Hazen SL.
2020. Targeted Inhibition of Gut Microbial Trimethylamine N- Oxide Production Reduces Renal Tubulointerstitial
Fibrosis and Functional Impairment in a Murine Model of Chronic Kidney Disease. Arteriosclerosis, Thrombosis,
and Vascular Biology 40:1239–1255. DOI: https://doi.org/10.1161/ATVBAHA.120.314139, PMID: 32212854
Hanouneh IA, Zein NN, Cikach F, Dababneh L, Grove D, Alkhouri N, Lopez R, Dweik RA. 2014. The breathprints
in patients with liver disease identify novel breath biomarkers in alcoholic hepatitis. Clinical Gastroenterology
and Hepatology 12:516–523. DOI: https://doi.org/10.1016/j.cgh.2013.08.048, PMID: 24036050
Helsley RN, Varadharajan V, Brown AL, Gromovsky AD, Schugar RC, Ramachandiran I, Fung K, Kabbany MN,
Banerjee R, Neumann CK, Finney C, Pathak P, Orabi D, Osborn LJ, Massey W, Zhang R, Kadam A, Sansbury BE,
Pan C, Sacks J, et al. 2019. Obesity- linked suppression of membrane- bound O- acyltransferase 7 (MBOAT7)
drives non- alcoholic fatty liver disease. eLife 8:e49882. DOI: https://doi.org/10.7554/eLife.49882, PMID:
31621579
Huang LJ, Wang L, Ma Y, Durick K, Perkins G, Deerinck TJ, Ellisman MH, Taylor SS. 1999. NH2- Terminal targeting
motifs direct dual specificity A- kinase- anchoring protein 1 (D- AKAP1) to either mitochondria or endoplasmic
reticulum. The Journal of Cell Biology 145:951–959. DOI: https://doi.org/10.1083/jcb.145.5.951, PMID:
10352013
Knorr J, Wree A, Tacke F, Feldstein AE. 2020. The NLRP3 Inflammasome in Alcoholic and Nonalcoholic
Steatohepatitis. Seminars in Liver Disease 40:298–306. DOI: https://doi.org/10.1055/s-0040-1708540, PMID:
32526788
Kochanek K, Xu JQ, Arias E. 2017. Deaths: Final Data for 2017. National Vital Statistics Reports. National Center
for Health Statistics.
Koeth RA, Wang Z, Levison BS, Buffa JA, Org E, Sheehy BT, Britt EB, Fu X, Wu Y, Li L, Smith JD, DiDonato JA,
Chen J, Li H, Wu GD, Lewis JD, Warrier M, Brown JM, Krauss RM, Tang WHW, et al. 2013. Intestinal microbiota
metabolism of L- carnitine, a nutrient in red meat, promotes atherosclerosis. Nature Medicine 19:576–585. DOI:
https://doi.org/10.1038/nm.3145, PMID: 23563705
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
25 of 28
Medicine
Research article
Kwak DS, Jun DW, Seo JG, Chung WS, Park S- E, Lee KN, Khalid- Saeed W, Lee HL, Lee OY, Yoon BC, Choi HS.
2014. Short- term probiotic therapy alleviates small intestinal bacterial overgrowth, but does not improve
intestinal permeability in chronic liver disease. European Journal of Gastroenterology & Hepatology 26:1353–
1359. DOI: https://doi.org/10.1097/MEG.0000000000000214, PMID: 25244414
Lang S, Schnabl B. 2020. Microbiota and Fatty Liver Disease- the Known, the Unknown, and the Future. Cell Host
& Microbe 28:233–244. DOI: https://doi.org/10.1016/j.chom.2020.07.007, PMID: 32791115
Li Q, Korzan WJ, Ferrero DM, Chang RB, Roy DS, Buchi M, Lemon JK, Kaur AW, Stowers L, Fendt M, Liberles SD.
2013. Synchronous evolution of an odor biosynthesis pathway and behavioral response. Current Biology
23:11–20. DOI: https://doi.org/10.1016/j.cub.2012.10.047, PMID: 23177478
Llopis M, Cassard AM, Wrzosek L, Boschat L, Bruneau A, Ferrere G, Puchois V, Martin JC, Lepage P, Le Roy T,
Lefèvre L, Langelier B, Cailleux F, González- Castro AM, Rabot S, Gaudin F, Agostini H, Prévot S, Berrebi D,
Ciocan D, et al. 2016. Intestinal microbiota contributes to individual susceptibility to alcoholic liver disease. Gut
65:830–839. DOI: https://doi.org/10.1136/gutjnl-2015-310585, PMID: 26642859
Lord CC, Ferguson D, Thomas G, Brown AL, Schugar RC, Burrows A, Gromovsky AD, Betters J, Neumann C,
Sacks J, Marshall S, Watts R, Schweiger M, Lee RG, Crooke RM, Graham MJ, Lathia JD, Sakaguchi TF, Lehner R,
Haemmerle G, et al. 2016. Regulation of Hepatic Triacylglycerol Metabolism by CGI- 58 Does Not Require
ATGL Co- activation. Cell Reports 16:939–949. DOI: https://doi.org/10.1016/j.celrep.2016.06.049, PMID:
27396333
Madjd A, Taylor MA, Mousavi N, Delavari A, Malekzadeh R, Macdonald IA, Farshchi HR. 2016. Comparison of
the effect of daily consumption of probiotic compared with low- fat conventional yogurt on weight loss in
healthy obese women following an energy- restricted diet: a randomized controlled trial. The American Journal
of Clinical Nutrition 103:323–329. DOI: https://doi.org/10.3945/ajcn.115.120170, PMID: 26702123
Maier L, Pruteanu M, Kuhn M, Zeller G, Telzerow A, Anderson EE, Brochado AR, Fernandez KC, Dose H, Mori H,
Patil KR, Bork P, Typas A. 2018. Extensive impact of non- antibiotic drugs on human gut bacteria. Nature
555:623–628. DOI: https://doi.org/10.1038/nature25979, PMID: 29555994
Masarone M, Rosato V, Dallio M, Abenavoli L, Federico A, Loguercio C, Persico M. 2016. Epidemiology and
Natural History of Alcoholic Liver Disease. Reviews on Recent Clinical Trials 11:167–174. DOI: https://doi.org/
10.2174/1574887111666160810101202, PMID: 27515957
McCullough RL, McMullen MR, Sheehan MM, Poulsen KL, Roychowdhury S, Chiang DJ, Pritchard MT,
Caballeria J, Nagy LE. 2018. Complement Factor D protects mice from ethanol- induced inflammation and liver
injury. American Journal of Physiology. Gastrointestinal and Liver Physiology 315:G66–G79. DOI: https://doi.
org/10.1152/ajpgi.00334.2017, PMID: 29597356
McMurdie PJ, Holmes S. 2013. phyloseq: an R package for reproducible interactive analysis and graphics of
microbiome census data. PLOS ONE 8:e61217. DOI: https://doi.org/10.1371/journal.pone.0061217, PMID:
23630581
Mistry JS, Abraham DJ, Kozikowski AP, Hanin I. 2002. Neurochemistry of aging. 2. Design, synthesis and
biological evaluation of halomethyl analogs of choline with high affinity choline transport inhibitory activity.
Journal of Medicinal Chemistry 34:2031–2036. DOI: https://doi.org/10.1021/jm00111a016, PMID: 2066974
Mutlu E, Keshavarzian A, Engen P, Forsyth CB, Sikaroodi M, Gillevet P. 2009. Intestinal dysbiosis: a possible
mechanism of alcohol- induced endotoxemia and alcoholic steatohepatitis in rats. Alcoholism, Clinical and
Experimental Research 33:1836–1846. DOI: https://doi.org/10.1111/j.1530-0277.2009.01022.x, PMID:
19645728
Mutlu EA, Gillevet PM, Rangwala H, Sikaroodi M, Naqvi A, Engen PA, Kwasny M, Lau CK, Keshavarzian A. 2012.
Colonic microbiome is altered in alcoholism. American Journal of Physiology. Gastrointestinal and Liver
Physiology 302:G966–G978. DOI: https://doi.org/10.1152/ajpgi.00380.2011, PMID: 22241860
Nooh MM, Bahouth SW. 2017. Two barcodes encoded by the type- 1 PDZ and by phospho- Ser312 regulate
retromer/WASH- mediated sorting of the ß1- adrenergic receptor from endosomes to the plasma membrane.
Cellular Signalling 29:192–208. DOI: https://doi.org/10.1016/j.cellsig.2016.10.014, PMID: 27816670
Organ CL, Li Z, Sharp TE, Polhemus DJ, Gupta N, Goodchild TT, Tang WHW, Hazen SL, Lefer DJ. 2020.
Nonlethal Inhibition of Gut Microbial Trimethylamine N- oxide Production Improves Cardiac Function and
Remodeling in a Murine Model of Heart Failure. Journal of the American Heart Association 9:e016223. DOI:
https://doi.org/10.1161/JAHA.119.016223, PMID: 32390485
Orman M, Bodea S, Funk MA, Campo AM- D, Bollenbach M, Drennan CL, Balskus EP. 2019. Structure- Guided
Identification of a Small Molecule That Inhibits Anaerobic Choline Metabolism by Human Gut Bacteria. Journal
of the American Chemical Society 141:33–37. DOI: https://doi.org/10.1021/jacs.8b04883, PMID: 30557011
Pachter LS. 2018. kallisto- transcriptome- indices. 944fbcc. Github. https://github.com/pachterlab/kallisto-
transcriptome-indices/releases
Paik YH, Schwabe RF, Bataller R, Russo MP, Jobin C, Brenner DA. 2003. Toll- like receptor 4 mediates
inflammatory signaling by bacterial lipopolysaccharide in human hepatic stellate cells. Hepatology (Baltimore,
Md.) 37:1043–1055. DOI: https://doi.org/10.1053/jhep.2003.50182, PMID: 12717385
Pimentel H, Bray NL, Puente S, Melsted P, Pachter L. 2017. Differential analysis of RNA- seq incorporating
quantification uncertainty. Nature Methods 14:687–690. DOI: https://doi.org/10.1038/nmeth.4324, PMID:
28581496
Puri P, Liangpunsakul S, Christensen JE, Shah VH, Kamath PS, Gores GJ, Walker S, Comerford M, Katz B,
Borst A, Yu Q, Kumar DP, Mirshahi F, Radaeva S, Chalasani NP, Crabb DW, Sanyal AJ, TREAT Consortium. 2018.
The circulating microbiome signature and inferred functional metagenomics in alcoholic hepatitis. Hepatology
(Baltimore, Md.) 67:1284–1302. DOI: https://doi.org/10.1002/hep.29623, PMID: 29083504
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
26 of 28
Medicine
Research article
Reijnders D, Goossens GH, Hermes GDA, Neis EPJG, van der Beek CM, Most J, Holst JJ, Lenaerts K, Kootte RS,
Nieuwdorp M, Groen AK, Olde Damink SWM, Boekschoten MV, Smidt H, Zoetendal EG, Dejong CHC,
Blaak EE. 2016. Effects of Gut Microbiota Manipulation by Antibiotics on Host Metabolism in Obese Humans:
A Randomized Double- Blind Placebo- Controlled Trial. Cell Metabolism 24:63–74. DOI: https://doi.org/10.
1016/j.cmet.2016.06.016, PMID: 27411009
Roberts AB, Gu X, Buffa JA, Hurd AG, Wang Z, Zhu W, Gupta N, Skye SM, Cody DB, Levison BS, Barrington WT,
Russell MW, Reed JM, Duzan A, Lang JM, Fu X, Li L, Myers AJ, Rachakonda S, DiDonato JA, et al. 2018.
Development of a gut microbe- targeted nonlethal therapeutic to inhibit thrombosis potential. Nature Medicine
24:1407–1417. DOI: https://doi.org/10.1038/s41591-018-0128-1, PMID: 30082863
Saha B, Tornai D, Kodys K, Adejumo A, Lowe P, McClain C, Mitchell M, McCullough A, Dasarathy S,
Kroll- Desrosiers A, Barton B, Radaeva S, Szabo G. 2019. Biomarkers of Macrophage Activation and Immune
Danger Signals Predict Clinical Outcomes in Alcoholic Hepatitis. Hepatology (Baltimore, Md.) 70:1134–1149.
DOI: https://doi.org/10.1002/hep.30617, PMID: 30891779
Schugar RC, Shih DM, Warrier M, Helsley RN, Burrows A, Ferguson D, Brown AL, Gromovsky AD, Heine M,
Chatterjee A, Li L, Li XS, Wang Z, Willard B, Meng Y, Kim H, Che N, Pan C, Lee RG, Crooke RM, et al. 2017.
The TMAO- Producing Enzyme Flavin- Containing Monooxygenase 3 Regulates Obesity and the Beiging of
White Adipose Tissue. Cell Reports 19:2451–2461. DOI: https://doi.org/10.1016/j.celrep.2017.05.077, PMID:
28636934
Schugar RC, Willard B, Wang Z, Brown JM. 2018. Postprandial gut microbiota- driven choline metabolism links
dietary cues to adipose tissue dysfunction. Adipocyte 7:49–56. DOI: https://doi.org/10.1080/21623945.2017.
1398295, PMID: 29172946
Smirnova E, Puri P, Muthiah MD, Daitya K, Brown R, Chalasani N, Liangpunsakul S, Shah VH, Gelow K,
Siddiqui MS, Boyett S, Mirshahi F, Sikaroodi M, Gillevet P, Sanyal AJ. 2020. Fecal Microbiome Distinguishes
Alcohol Consumption From Alcoholic Hepatitis But Does Not Discriminate Disease Severity. Hepatology
(Baltimore, Md.) 72:271–286. DOI: https://doi.org/10.1002/hep.31178, PMID: 32056227
Strub K, Walter P. 1990. Assembly of the Alu domain of the signal recognition particle (SRP): dimerization of the
two protein components is required for efficient binding to SRP RNA. Molecular and Cellular Biology 10:777–
784. DOI: https://doi.org/10.1128/mcb.10.2.777-784.1990, PMID: 2153922
Tang WHW, Wang Z, Levison BS, Koeth RA, Britt EB, Fu X, Wu Y, Hazen SL. 2013. Intestinal microbial metabolism
of phosphatidylcholine and cardiovascular risk. The New England Journal of Medicine 368:1575–1584. DOI:
https://doi.org/10.1056/NEJMoa1109400, PMID: 23614584
Tang WHW, Hazen SL. 2014. The contributory role of gut microbiota in cardiovascular disease. The Journal of
Clinical Investigation 124:4204–4211. DOI: https://doi.org/10.1172/JCI72331, PMID: 25271725
Tarao K, Moroi T, Nagakura Y, Ikeuchi T, Suyama T, Endo O, Fukushima K. 1979. Relationship between
endotoxaemia and protein concentration of ascites in cirrhotic patients. Gut 20:205–210. DOI: https://doi.org/
10.1136/gut.20.3.205, PMID: 437553
Tilt EM. 1999. Nonparametric statistical methods. Myles Hollander and Douglas A. Wolfe, Wiley, Chichester.
Statistics in Medicine 19:2- X. DOI: https://doi.org/10.1002/(SICI)1097-0258(20000530)19:103.0.CO;2-X
Tripathi A, Debelius J, Brenner DA, Karin M, Loomba R, Schnabl B, Knight R. 2018. The gut- liver axis and the
intersection with the microbiome. Nature Reviews. Gastroenterology & Hepatology 15:397–411. DOI: https://
doi.org/10.1038/s41575-018-0011-z, PMID: 29748586
Trøseid M, Ueland T, Hov JR, Svardal A, Gregersen I, Dahl CP, Aakhus S, Gude E, Bjørndal B, Halvorsen B,
Karlsen TH, Aukrust P, Gullestad L, Berge RK, Yndestad A. 2015. Microbiota- dependent metabolite
trimethylamine- N- oxide is associated with disease severity and survival of patients with chronic heart failure.
Journal of Internal Medicine 277:717–726. DOI: https://doi.org/10.1111/joim.12328, PMID: 25382824
Uesugi T, Froh M, Arteel GE, Bradford BU, Thurman RG. 2001. Toll- like receptor 4 is involved in the mechanism
of early alcohol- induced liver injury in mice. Hepatology (Baltimore, Md.) 34:101–108. DOI: https://doi.org/10.
1053/jhep.2001.25350, PMID: 11431739
Vatsalya V, Cave MC, Kong M, Gobejishvili L, Falkner KC, Craycroft J, Mitchell M, Szabo G, McCullough A,
Dasarathy S, Radaeva S, Barton B, McClain CJ. 2020. Keratin 18 Is a Diagnostic and Prognostic Factor for
Acute Alcoholic Hepatitis. Clinical Gastroenterology and Hepatology 18:2046–2054. DOI: https://doi.org/10.
1016/j.cgh.2019.11.050, PMID: 31811953
Wallrabenstein I, Kuklan J, Weber L, Zborala S, Werner M, Altmüller J, Becker C, Schmidt A, Hatt H, Hummel T,
Gisselmann G. 2013. Human trace amine- associated receptor TAAR5 can be activated by trimethylamine. PLOS
ONE 8:e54950. DOI: https://doi.org/10.1371/journal.pone.0054950, PMID: 23393561
Wang Z, Klipfell E, Bennett BJ, Koeth R, Levison BS, Dugar B, Feldstein AE, Britt EB, Fu X, Chung Y- M, Wu Y,
Schauer P, Smith JD, Allayee H, Tang WHW, DiDonato JA, Lusis AJ, Hazen SL. 2011. Gut flora metabolism of
phosphatidylcholine promotes cardiovascular disease. Nature 472:57–63. DOI: https://doi.org/10.1038/
nature09922, PMID: 21475195
Wang Z, Levison BS, Hazen JE, Donahue L, Li XM, Hazen SL. 2014a. Measurement of trimethylamine- N- oxide by
stable isotope dilution liquid chromatography tandem mass spectrometry. Analytical Biochemistry 455:35–40.
DOI: https://doi.org/10.1016/j.ab.2014.03.016, PMID: 24704102
Wang Z, Tang WHW, Buffa JA, Fu X, Britt EB, Koeth RA, Levison BS, Fan Y, Wu Y, Hazen SL. 2014b. Prognostic
value of choline and betaine depends on intestinal microbiota- generated metabolite trimethylamine- N- oxide.
European Heart Journal 35:904–910. DOI: https://doi.org/10.1093/eurheartj/ehu002, PMID: 24497336
Wang Z, Roberts AB, Buffa JA, Levison BS, Zhu W, Org E, Gu X, Huang Y, Zamanian- Daryoush M, Culley MK,
DiDonato AJ, Fu X, Hazen JE, Krajcik D, DiDonato JA, Lusis AJ, Hazen SL. 2015. Non- lethal Inhibition of Gut
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
27 of 28
Medicine
Research article
Microbial Trimethylamine Production for the Treatment of Atherosclerosis. Cell 163:1585–1595. DOI: https://
doi.org/10.1016/j.cell.2015.11.055, PMID: 26687352
Warrier M, Shih DM, Burrows AC, Ferguson D, Gromovsky AD, Brown AL, Marshall S, McDaniel A, Schugar RC,
Wang Z, Sacks J, Rong X, Vallim T, Chou J, Ivanova PT, Myers DS, Brown HA, Lee RG, Crooke RM, Graham MJ,
et al. 2015. The TMAO- Generating Enzyme Flavin Monooxygenase 3 Is a Central Regulator of Cholesterol
Balance. Cell Reports 10:326–338. DOI: https://doi.org/10.1016/j.celrep.2014.12.036, PMID: 25600868
Wickham H. 2009. Ggplot2: Elegant Graphics for Data Analysis. New York, NY: Springer Publishing Company,
Incorporated. DOI: https://doi.org/10.1007/978-0-387-98141-3
Wilkinson SP, Arroyo V, Gazzard BG, Moodie H, Williams R. 1974. Relation of renal impairment and
haemorrhagic diathesis to endotoxaemia in fulminant hepatic failure. Lancet (London, England) 1:521–524.
DOI: https://doi.org/10.1016/s0140-6736(74)92711-1, PMID: 4131970
Wu H, Esteve E, Tremaroli V, Khan MT, Caesar R, Mannerås- Holm L, Ståhlman M, Olsson LM, Serino M,
Planas- Fèlix M, Xifra G, Mercader JM, Torrents D, Burcelin R, Ricart W, Perkins R, Fernàndez- Real JM,
Bäckhed F. 2017. Metformin alters the gut microbiome of individuals with treatment- naive type 2 diabetes,
contributing to the therapeutic effects of the drug. Nature Medicine 23:850–858. DOI: https://doi.org/10.
1038/nm.4345, PMID: 28530702
Yan AW, Fouts DE, Brandl J, Stärkel P, Torralba M, Schott E, Tsukamoto H, Nelson KE, Brenner DA, Schnabl B.
2011. Enteric dysbiosis associated with a mouse model of alcoholic liver disease. Hepatology (Baltimore, Md.)
53:96–105. DOI: https://doi.org/10.1002/hep.24018, PMID: 21254165
Zhou W, Li X, Premont RT. 2016. Expanding functions of GIT Arf GTPase- activating proteins, PIX Rho guanine
nucleotide exchange factors and GIT- PIX complexes. Journal of Cell Science 129:1963–1974. DOI: https://doi.
org/10.1242/jcs.179465, PMID: 27182061
Zhu W, Gregory JC, Org E, Buffa JA, Gupta N, Wang Z, Li L, Fu X, Wu Y, Mehrabian M, Sartor RB, McIntyre TM,
Silverstein RL, Tang WHW, DiDonato JA, Brown JM, Lusis AJ, Hazen SL. 2016. Gut Microbial Metabolite TMAO
Enhances Platelet Hyperreactivity and Thrombosis Risk. Cell 165:111–124. DOI: https://doi.org/10.1016/j.cell.
2016.02.011, PMID: 26972052
Zhu W, Wang Z, Tang WHW, Hazen SL. 2017. Gut Microbe- Generated Trimethylamine N- Oxide From Dietary
Choline Is Prothrombotic in Subjects. Circulation 135:1671–1673. DOI: https://doi.org/10.1161/
CIRCULATIONAHA.116.025338, PMID: 28438808
Helsley, Miyata, Kadam, et al. eLife 2022;11:e76554. DOI: https://doi.org/10.7554/eLife.76554
28 of 28
Medicine
| null |
10.1088_1758-5090_acfb3c.pdf
|
Data availability statement
The data that support the findings of this study
are openly available at the following URL/DOI:
www.ncbi.nlm.nih.gov/bioproject/PRJNA953644.
The other data that supports the findings of this study
are available upon reasonable request.
|
Data availability statement The data that support the findings of this study are openly available at the following URL/DOI: www.ncbi.nlm.nih.gov/bioproject/PRJNA953644 . The other data that supports the findings of this study are available upon reasonable request.
|
Biofabrication 15 (2023) 045025
https://doi.org/10.1088/1758-5090/acfb3c
Biofabrication
PAPER
RECEIVED
18 April 2023
REVISED
21 August 2023
ACCEPTED FOR PUBLICATION
19 September 2023
PUBLISHED
27 September 2023
3D bioprinted endothelial cell-microglia coculture for diabetic
retinopathy modeling
Haixiang Wu1,∗, Fangcheng Xu1, Yunfang Luo2, Yibao Zhang2 and Min Tang3,∗
1 Department of Ophthalmology, Eye & ENT Hospital, Fudan University, Shanghai 200031, People’s Republic of China
2 Department of Biomedical Research, Cyberiad Biotechnology Ltd, Shanghai 201112, People’s Republic of China
3 Institute of Interdisciplinary Integrative Medical Research, Shanghai University of Traditional Chinese Medicine, Shanghai 201203,
∗
People’s Republic of China
Authors to whom any correspondence should be addressed.
E-mail: [email protected] and [email protected]
Keywords: diabetic retinopathy, bioprinting, microglia, glucose level, coculture
Supplementary material for this article is available online
Abstract
Diabetic retinopathy (DR) is a common diabetes complication leading to vision impairment or
blindness due to retinal vasculature alterations. Hyperglycemia induces structural alterations,
inflammation, and angiogenic factor upregulation. Current treatments targeting vascular
endothelial growth factor are insufficient for approximately 20% of DR patients, necessitating
alternative approaches. Microglia (MG), essential for retinal homeostasis, remains underexplored
in DR. This study used digital light processing bioprinting to construct a 3D coculture model of
endothelial cells (ECs) and MG under varying glucose conditions, with a hydrogel stiffness of
4.6–7.1 kPa to mimic the extracellular matrix property of retina plexiform. Our results showed that
high glucose levels influenced both EC and microglial phenotypes, gene expression, and angiogenic
potential. Increasing glucose from 5 mM to 25 mM reduces drug efficacy by 17% for Aflibercept in
EC monoculture, and 25% and 30% for Aflibercept and Conbercept in EC-MG coculture,
respectively, suggesting that diabetic condition and MG presence could interfere with drug
responses. In conclusion, our findings emphasize the importance of cellular interactions and
microenvironmental factors in DR therapy, aiming to identify novel strategies and improve
understanding of MG’s role in disease pathogenesis.
1. Introduction
Diabetic retinopathy (DR) is a prevalent and debil-
itating ocular complication of diabetes, character-
ized by pathological changes in the retinal vascu-
lature, affecting roughly one-third of individuals with
diabetes, and is the most common disease affect-
ing retina vasculature [1–4]. The likelihood of DR
rises with the duration of diabetes and levels of
glycosylated hemoglobin. Advanced stages of DR
are characterized by vascular leakage and prolif-
eration. In more advanced cases, proliferative DR
(PDR) and diabetic macular edema (DME) may
develop. Elevated blood glucose levels, characteristic
of diabetes, exert deleterious effects on retinal blood
vessels,
initiating a cascade of pathophysiological
changes that culminate in DR [5]. Hyperglycemia
instigates the production of advanced glycation end
products, which accumulate in the retinal microvas-
culature, inducing structural changes and promot-
ing the release of pro-inflammatory cytokines [5].
These cytokines contribute to the breakdown of the
blood-retinal barrier, increasing vascular permeab-
ility and leading to fluid accumulation in the ret-
ina, which impairs vision. Elevated glucose levels
also drive the overproduction of reactive oxygen spe-
cies, resulting in oxidative stress within the retina
[6]. Oxidative stress activates signaling pathways that
promote retinal cell apoptosis, inflammation, and
the upregulation of angiogenic factors. Excessive vas-
cular endothelial growth factor (VEGF) promotes
abnormal, fragile blood vessels, a process known as
neovascularization, which could eventually lead to
vision loss [7]. DR treatments predominantly focus
© 2023 IOP Publishing Ltd
Biofabrication 15 (2023) 045025
H Wu et al
on targeting VEGF [4, 8]. However, roughly 20% of
DR patients do not respond adequately to anti-VEGF
therapy, and around 40% of patients experienced
persistent DME chronically under treatment [9, 10].
Studies have revealed the potential role of microglia
(MG), which is distributed throughout the retina,
such as the plexiform layers, ganglion cell layer, and
nerve fiber layer, in various retinopathies and neur-
ological disorders with ocular manifestations [11].
The importance of MG in retinal angiogenesis is sup-
ported by evidence of their participation in vasculo-
genesis within the human fetal retina before retinal
vasculature formation and their association with the
growth of vessels [12–15]. MGs and ECs also exhibit
close proximity, implying potential interactions dur-
ing vascular development [14]. A more thorough
understanding of the dynamic interplay between ret-
inal ECs and MGs is essential for advancing our
knowledge of DR.
Available modeling systems for retinal diseases
including DR have been extensively reviewed [16].
Retinal disease investigations have predominantly
used animal models [17]. However, concerns about
species-specific variations and ethical considerations
limit their universal applicability. While 2D cell
cultures have elucidated pivotal signaling pathways
associated with various retinal cell types and their
responses to diabetes-related factors, they do not
adequately represent the intricate multicellular inter-
actions or crosstalk with ECM found within the native
retina [18–20]. Tissue explants are more complex and
biomimetic models, providing an improved repres-
entation but fall short of allowing precise control over
individual factors [21]. This limitation hampers the
study of different DR variants. Harnessing advance-
ments in biofabrication and biomaterials, scient-
ists have developed biomimetic in vitro 3D mod-
els. Techniques such as manual-casting 3D culture,
3D bioprinting, organoid culture, and organ-on-chip
models have been employed to emulate either the
intricate cellular interactions or the distinct proper-
ties of the tissue ECM [22–24]. While there has been
a focus on creating 3D models for the normal ret-
ina, such as the outer-blood-retina-barrier, and dis-
eases like age-related macular degeneration, models
specific to DR are less commonly explored [25–28].
Furthermore, the impact of ECM properties on cellu-
lar dynamics, as well as the effects of various retina-
specific signals including electrical signals or glucose
levels on the ECM, have yet to be investigated [29].
Among the diverse biofabrication technologies,
3D bioprinting has shown particular promise for the
generation of complex and controllable biological
constructs. Utilizing natural material-derived bioma-
terials, this technology offers exceptional flexibility
and capacity in mimicking multicellular microen-
vironments, elucidating cellular interactions, and
reproducing extracellular matrix (ECM) properties
2
in three-dimensional constructs [30, 31]. Bioprinting
techniques can be categorized into three main types
[32]. Extrusion-based bioprinting is the most pre-
valent form of bioprinting, where a continuous fila-
ment of bioink is extruded to build 3D structures in
a controlled manner [33, 34]. Inkjet-based bioprint-
ing involves depositing droplets of biomaterials and
cells on a substrate. The precise control of droplet size
and position enables the creation of complex, high-
resolution 3D constructs [35, 36]. The nozzle-based
methods, including extrusion-based and inkjet-based
bioprinting can handle multi-material constructs eas-
ily by switching of nozzles [37]. Vat photopoly-
merization or light-based bioprinting, such as ste-
reolithography, digital light processing (DLP), and
two-photon polymerization leverage light to poly-
merize photosensitive materials, avoiding the poten-
tial risks of shear stress in nozzle-based bioprinting
methods [38, 39]. DLP, notably effective in gener-
ating tissue-mimetic microenvironments, relies on a
digital micromirror device consisting of millions of
micromirrors that can be switched ‘on’ or ‘off ’ to
precisely control the shape of projected light [40].
This bioprinting technology delivers high cell viabil-
ity, superior printing speed, and single-cell level res-
olution, making it especially suited for the generation
of cell-encapsulating constructs [41–43].
In this study, we employed a DLP bioprinter to
generate a 3D bioprinted coculture model of EC and
MG, using relevant biomaterials, including gelatin
and hyaluronic acid-derived photosensitive materi-
als, to achieve tunable stiffness to more accurately
mimic retinal tissue. Gelatin has been used to gen-
erate a retinal capillary bed, and hyaluronic acid has
been used to print the retina [28, 44, 45]. We used a
combination of the two biomaterials to generate the
hydrogel constructs in this study. We observed alter-
ations in cellular growth and phenotype within the
3D hydrogels due to varying glucose concentrations
in the culture medium. We observed changes in cel-
lular growth and phenotype within the 3D hydrogels
in response to varying glucose concentrations in the
culture medium. Our results indicated that high gluc-
ose levels had an inhibitory effect on the growth of
both ECs and MGs; however, this effect was atten-
uated in coculture conditions. When subjected to
anti-VEGF treatments, the coculture model demon-
strated increased resistance compared to ECs alone,
further validating a protective role for MG under dia-
betic conditions. Gene expression profiling revealed
that the coculture conditions and various glucose
conditions could all pose significant alterations on
both cell types. By manipulating glucose concentra-
tions in the culture medium, we aim to reveal novel
insights into the intricate interactions between ECs
and MG under diabetic conditions, ultimately seek-
ing to identify potential therapeutic strategies and
enhance clinical outcomes for patients.
Biofabrication 15 (2023) 045025
H Wu et al
2. Materials and methods
2.1. Cell culture
The human MG cell line HMC3 (ATCC) was cul-
tured in minimum essential medium (Gibco), with
10% fetal bovine serum (FBS, Gibco), 1% penicillin-
streptomycin (p s−1, Gibco), and 1% non-essential
amino acids (Gibco). The human endothelial cell
(EC)
line PUMC-HUVEC-T1 (HUVEC, NSTI-
BMCR) was cultured in low-glucose Dulbecco’s
Modified Eagle Medium (DMEM, Gibco), with 10%
FBS, 1% p s−1, and 25 ng ml−1 VEGF. Cells were pas-
saged using 0.05% trypsin-EDTA (Gibco). Cells were
authenticated using short tandem repeat analysis.
2.2. Bioprinting of cellular constructs
A DLP bioprinter
550 A, Cyberiad
(Azure
Biotechnology) was employed to fabricate cell-
encapsulated hydrogels. The printing parameters
were set to 50% light intensity and 15–20 s expos-
ure time based on the rheological measurements
(supplementary figure 1(a)). A bioink composed of
4% gelatin methacrylate (GelMA, Sinobioprint), 1%
hyaluronic acid methacrylate (HAMA), and 0.3%
lithium phenyl-2,4,6-trimethylbenzoylphosphinate
(LAP, TCI chemicals) was prepared as a prepoly-
mer solution and temporarily stored at 37 ◦C in the
dark. HMC3 and HUVEC cells between passages
3-7 were used for the experiments. Both cell types
were enzymatically dissociated and resuspended to
achieve a concentration of 20 million cells ml−1. For
single cell type printing, the cells are ready for use. For
coculture printing, cells were mixed at a 1:1 ratio. The
cell suspension was temporarily placed on ice prior
to use. Immediately before printing, the cell mixture
and bioink were thoroughly combined at a 1:1 ratio
and loaded for printing.
48 h. After drying, the samples were treated with an
iridium coating through a sputter coater (Emitech).
The coated samples were then inspected using a scan-
ning electron microscope (Zeiss) and the SEM images
were analyzed using ImageJ to quantify the pore sizes.
2.4. Cell proliferation evaluation
Cell-encapsulated hydrogels were generated and cul-
tured in testing media. For HMC3,
the testing
media included the original culture medium (with
5.5 mM glucose, G5.5), and medium supplemen-
ted with additional glucose to achieve concentrations
of 10 mM (G10), 15 mM (G15), 25 mM (G25),
50 mM (G50), and 100 mM (G100). For HUVEC, the
testing media included the original culture medium
without VEGF (with 5.5 mM glucose, G5.5), and
medium supplemented with additional glucose to
achieve concentrations of G10, G15, G25, G50, and
G100. For coculture models, a 1:1 mixture of the
single-cell type testing media at corresponding gluc-
ose concentrations was prepared. CellTiter-Glo 3D
(CTG, Promega) was used to evaluate cell prolifera-
tion under different culture conditions, with at least
three replicates measured for each condition.
2.5. Cell viability evaluation
Cell-encapsulated hydrogels were generated and cul-
tured in testing media as described in the previous
section. Three glucose concentrations were studied:
G5.5, G10, and G25. On day 5 post-printing, samples
were stained using a Live-Dead Viability/Cytotoxicity
Kit (Invitrogen) with 1:2000 calcein-AM and 1:500
ethidium homodimer-1 diluted in PBS. The samples
were stained at 37 ◦C for 15 min and immediately
imaged using a fluorescence microscope. At least
three replicates were stained and measured for each
condition.
2.3. Mechanical characterization
For stiffness assessment, hydrogel samples were
incubated overnight at 37 ◦C in phosphate-buffered
saline (PBS) or different glucose-supplemented
media, and measurements were obtained the fol-
lowing day using a nanoindenter (Piuma, Optics11)
in matrix scan mode. At least three replicates were
measured for each sample.
The rheological properties of the bioinks were
assessed using a Discovery Hybrid Rheometer HR-2
(TA Instruments) with a parallel plate. Storage and
loss moduli were recorded at a constant strain of 0.1%
and a frequency of 5 rad s−1 at 37 ◦C. The bioink was
exposed to a light source with a wavelength of 405 nm,
at a power density of 30 mW cm−2, for a duration of
30 s.
For scanning electron microscopy (SEM) meas-
urement, the bioprinted samples were snap-frozen
using liquid nitrogen and then immediately subjected
to a freeze-drying (Labconco) process for no less than
2.6. Vascular formation assay
HUVEC and HMC3 cells were seeded in six-well
plates. Seeding density was 0.3 million cells ml−1.
When cell confluency reached 30%–50%, zsGreen
and RFP lentivirus (Hanbio Biotechnology) were
used to transfect the cells. The multiplicity of infec-
tion of 30 was used for both cells. A 2x lentivirus solu-
tion with 5 µg ml−1 polybrene was prepared in the
cells’ original medium. Cells were treated with the 2x
lentivirus solution for 4 h at 37 ◦C in the incubator,
and then the same volume of fresh medium was added
to the well to dilute the solution to 1x. The incubation
continued for 24 h, and the lentivirus solution was
removed. The transfected cells were cultured in a fresh
medium and monitored for fluorescence expression.
HMC3-RFP cells were bioprinted to form cell-
encapsulated hydrogels, and HUVEC-zsGreen cells
were immediately seeded on top. Three glucose
concentrations were studied,
including G5, G10,
and G25. Images were taken with a fluorescence
3
Biofabrication 15 (2023) 045025
H Wu et al
microscope (Olympus) at 24 h post-printing and
seeding. Images were analyzed using the ImageJ plu-
gin angiogenesis analyzer [46].
2.7. Immunofluorescence staining
Bioprinted samples were fixed with 4% paraformal-
dehyde (Beyotime) for 1 h, rinsed with Dulbecco’s
PBS (DPBS), permeabilized with 0.2% Triton X-100
(Merck) for 15 min, and blocked with 5% bovine
serum albumin (Merck) for 1 h. All procedures
were performed at room temperature. Primary anti-
bodies, including anti-IBA1 (1:100, Cell Signaling
Technology) and VE-cadherin (1:100, Invitrogen),
were diluted in staining buffer (BioLegend), and the
samples were incubated in primary antibody solu-
tion at 4 C overnight. The next day, the samples were
rinsed with DPBS and incubated with secondary anti-
bodies (1:200, Cell Signaling Technology) and DAPI
(1:1000) diluted in staining buffer at room temperat-
ure for 2 h. The samples were then rinsed with DPBS
and immediately imaged on a confocal microscope
(Leica).
2.8. RNAseq and data analysis
For single-cell-type bioprinted samples containing
either HUVEC or HMC3 cells, hydrogels were dis-
sociated using 2 mg ml−1 Collagenase I (Yeasen)
at 37 ◦C for 1 h. Subsequently, cells were harves-
ted and centrifuged at 300 g for 5 min. Cell pel-
lets containing a minimum of 1 million cells were
treated with 1 ml of TRIzol (Invitrogen) and stored
at −80 ◦C until all samples were collected. In cocul-
ture samples, zsGreen-HMC3 and HUVEC cells were
employed for printing. Hydrogels were dissociated
using the same methodology as for single-cell-type
samples. Following hydrogel digestion, cells were
washed with DPBS and centrifuged at 300 g for 5 min.
Cells were resuspended in DPBS containing 1% FBS
and 40 ng ml−1 DNase I for flow cytometry sort-
ing (Fusion, BD). HMC3 and HUVEC cells were
separated and immediately lysed with TRIzol. Total
RNA was extracted from TRIzol samples using an
RNA extraction kit (Zymo). The library was gener-
ated through messenger RNA (mRNA) purification
using poly-T oligo-attached magnetic beads, cDNA
synthesis, end repair, A-tailing, adapter ligation, size
selection, and PCR amplification. Paired-end sequen-
cing was conducted using the Illumina NovaSeq 6000
platform by Novogene.
The raw RNAseq FASTQ data were processed
by first trimming low-quality reads using TrimGalore
v0.6.7. The trimmed data was analyzed for transcript-
level quantification using Salmon v0.13.1 in
quasi-mapping mode, with Gencode Release 33
(GRCh38.p13) used for annotation. Transcript-level
quantification was subsequently converted to gene-
level quantification using Tximport. Differential
expression analysis was then conducted using DESeq2
v1.31.15 to identify pair-wise differentially expressed
4
genes between single-culture and coculture at G5.5,
as well as between coculture at G5.5, G10, and G25.
The criteria to select differentially expressed genes
were padj < 0.05 and |log2FC| > 1.
2.9. Drug testing
HUVEC and HMC3 cells were enzymatically disso-
ciated to create single-cell suspensions. The cells were
then counted, and three distinct cell suspensions were
prepared: (1) only HUVEC cells, (2) only HMC3
cells, and (3) a 1:1 mixture of HUVEC and HMC3
cells. These suspensions were centrifuged at 200 g for
5 min, with the supernatant removed and the cells
resuspended in base medium to achieve a concentra-
tion of 5 × 106 cells ml−1.
The printing material was prepared as previously
described, and the cell suspensions were combined
with the printing material at a 1:1 ratio before being
printed onto 96-well plates. Following bioprinting,
100 µl of the drug testing medium was added. The
G5.5 drug testing medium was composed of a 1:1
mixture of the MEM and low glucose DMEM as
base medium and supplemented with 10% FBS, 1%
P/S, and 1% NEAA. The G10 and G25 drug test-
ing medium were formulated by adding glucose to
the G5.5 testing medium. Nine experimental groups
were established, with each of the three drug testing
media evaluated on the following cell conditions: (1)
HUVEC monoculture, (2) HMC3 monoculture, and
(3) HUVEC/HMC3 coculture.
and
Ranibizumab
Drugs tested in this study are drugs com-
monly used in clinical practice, including Aflibercept
(EYLEA, Bayer), Conbercept (Lumitin, Kanghong
(Lucentis,
Biotechnology),
Novartis). After 24 h of incubation, the supernatant
was removed and replaced with 100 µl of drug-
containing medium, prepared to achieve final drug
concentrations of 1 µg or 10 µg, in each well. For
control wells, the media were replaced with fresh
corresponding testing medium without drugs. The
plates were returned to the CO2 incubator for contin-
ued culture. Following 72 h of treatment, the super-
natant was aspirated and replaced with 100 µl of a
1:1 mixture of CTG solution and PBS in each well.
The plates were incubated for 30 min while shaking at
350 rpm min−1 on a shaker, shielded from light. After
incubation, the plates were removed from the shaker,
and luminescence was measured using a preheated
plate reader (Tecan) for 30 min. Cell viability was
assessed using GraphPad Prism 9, with at least three
replicates measured and analyzed for each condition.
2.10. Statistics
Statistical analyses were performed with GraphPad
Prism 9. One-way or two-way ANOVA was used
to analyze differences in multiple groups, followed
by Tukey’s multiple comparisons test for post-hoc
analysis. T-tests were used to compare between two
groups. At least triplicates were tested and analyzed
Biofabrication 15 (2023) 045025
H Wu et al
in all experiments. The results were considered stat-
istically significant if the p-value was less than 0.05.
3. Results
3.1. Development of a 3D bioprinted
retina-mimetic coculture model
We employed a DLP bioprinter operating at 405 nm
to generate HMC3, HUVEC, and HMC3/HUVEC
coculture models (figure 1(a), supplementary figure
1(b)). To establish a reliable coculture model of EC
and MG that accurately replicates the native retina’s
ECM properties, we optimized both biomaterial and
bioprinting parameters. The bioink was composed
of GelMA and HAMA due to their proven com-
patibility with retinal cell culture and the flexibility
they provide in manipulating stiffness characteristics
[28, 41, 42, 44, 47]. The compressive modulus of a
healthy native retina has been reported to be approx-
imately 10–20 kPa, with the plexiform layers exhib-
iting a softer range of 1.3–7.7 kPa [23, 48]. As MG
primarily inhabits the plexiform layers while extend-
ing processes to surveil the retina, we tested different
GelMA concentrations to achieve a hydrogel stiffness
of 4.6 kPa, corresponding to the average plexiform
layer stiffness (figure 1(b)). The stiffness remained
consistent whether the hydrogels were bioprinted
with or without cells (figure 1(c)), and increased
when incubated in a medium with a higher gluc-
ose level (i.e. G25) (supplementary figure 1(c)). The
bioprinted samples exhibited adequate pore sizes for
cell migration (supplementary figure 1(d)).
We evaluated cell proliferation within the hydro-
gel at different glucose concentrations, ranging from
G5.5 to G100 guided by both physiological relev-
ance and findings from previous studies [49, 50].
A glucose level of 5.5 mM corresponds to normal
physiological blood glucose levels, while all other con-
ditions, starting from 10 mM, fall within the diabetic
range [51]. EC and MG, printed and cultured alone,
exhibited reduced proliferation at elevated glucose
levels when assessed on day 3 (figures 1(d) and (e)),
indicating that increased glucose concentration neg-
atively impacted cell growth. In contrast, the cocul-
ture model displayed a markedly different behavior.
We monitored cell proliferation at additional time
points due to the emphasis on the coculture model.
On days 1 and 3, all high glucose concentration con-
ditions, except for G100, promoted cell proliferation
in the coculture model (figure 1(f)). By day 5, G10 and
G15 maintained enhanced cell proliferation, while
higher glucose levels appeared to reduce proliferation.
The absolute cell numbers were highest in the G10
group on day 1 and day 3 and in the G15 group on
day 5 (figure 1(g)). These observations suggest that
coculture conditions may provide a protective effect
on cells under diabetic conditions. In the severe form
of DR, known as PDR, abnormal neovascularization
occurs. Our findings indicate that while EC or MG
alone may experience growth inhibition due to high
glucose levels, coculture could be protective, promot-
ing cellular proliferation.
3.2. Impact of glucose levels on EC and MG
phenotypes
Given the observed decrease in cell numbers in single-
cell-type models, we conducted a cell viability assess-
ment to determine whether cells were experiencing
reduced proliferation or cell death due to the sup-
plemented glucose (figure 2(a), supplementary figure
2(a)). Building upon our previous findings, we selec-
ted two glucose concentrations for further invest-
igation: G10, representing a typical diabetic gluc-
ose threshold, and G25, which demonstrated inhib-
itory effects on cell proliferation in the coculture
model during extended culture periods. These were
compared to the control group G5.5. Cell viability
remained high, with all culture conditions, including
monocultures and the coculture, maintaining levels
above 90% for both cell types (figure 2(b), supple-
mentary figure 2(b)).
In the G5.5 condition, ECs formed the most
closed lumen-like structures. However, in the G10
condition,
the number of closed-lumen struc-
tures decreased, and more open-lumen-like shapes
were observed. In the G25 condition, most lumens
remained open. We further examined the reorganiz-
ational potential of ECs under the influence of MG
coculture and varying glucose levels. To emulate tra-
ditional tube formation assays, we first bioprinted
HMC3-RFP cells within the hydrogel and imme-
diately seeded HUVEC-zsGreen cells on top of the
bioprinted hydrogel. Tube formation occurred rap-
idly within one day, and fluorescent images were
obtained (figures 3(a) and (b)). Similar to single-
cultured ECs, more complete mesh structures formed
in the G5.5 condition, and a decreasing trend was
observed with increasing glucose levels (figure 3(c)).
The total area of meshes displayed a similar pattern.
Nevertheless, ECs in the G25 condition still exhib-
ited aberrant angiogenic potential, as evidenced by
the significantly higher number of extremities and
total length of isolated branches, indicative of new
sprouting behavior (figure 3(d)).
IBA1, a calcium-binding protein, plays a critical
role in regulating microglial function, particularly in
activated MG [52]. In the context of DR, hyperre-
flective retinal spots (HRS) are clinically observed
by physicians using optical coherence tomography.
Studies have reported a positive correlation between
IBA1-positive MG activation and retinal HRS in
patients with non-PDR (NPDR) and PDR [53]. Our
results revealed that in the diabetic coculture con-
ditions, G10 and G25, a considerably higher IBA1
5
Biofabrication 15 (2023) 045025
H Wu et al
Figure 1. (a) Schematic representation of the bioprinted co-culture model, comprising endothelial cells (EC) and microglia
(MG), fabricated using a digital light processing (DLP) bioprinter and formulated bioinks. (b) Quantification of hydrogel
stiffness for constructs printed with bioink formulations containing 2%, 3%, and 4% GelMA, following overnight incubation in
PBS. (c) Stiffness of hydrogel without (acellular) and with cells, measured after overnight incubation in G5.5. (d) Relative cell
numbers in 3D bioprinted EC models on day 3 cultured in different glucose media. (e) Relative cell numbers in 3D bioprinted
MG models on day 3 cultured in different glucose media. (f) Heatmap illustrating relative cell counts in 3D co-cultured EC-MG
hydrogels for various glucose level media on days 1, 3, and 5 post-printing. The control group was the G5.5 on each day. (g)
Absolute luminescence measurements of coculture models in different media on days 1, 3, and 5 post-printing. A minimum of
three replicate measurements were obtained for each condition. Data are presented as mean ± standard deviation (SD).
∗p < 0.05, ∗∗∗p < 0.001.
expression was detected in MG, suggesting MG activ-
ation (figure 4(a)). Additionally, under increased
glucose concentrations, MG cells displayed morpho-
logical alterations. In G25, an amoeboid morphology
of MG was observed, further signifying an activated
state (figure 4(b)).
3.3. Influence of coculture on gene expression of
EC and MG
induced significant changes
The coculture model
in cell proliferation at various glucose levels.
Consequently, we performed RNA sequencing (RNA-
seq) with EC and MG isolated from both single-cell-
type bioprinted samples and cocultured bioprinted
samples for further analysis. Notable changes in gene
6
expression were observed in both cell types following
coculture.
Compared to single-cultured MG, 1038 genes
were upregulated in cocultured MG, and 574 genes
were downregulated (supplementary figure 3(a)).
Gene ontology (GO) analysis of the biological pro-
cesses of the cocultured MG revealed significant
enrichment in pathways related to vasculature devel-
opment, organogenesis, and cellular responses to
various stimuli (figure 5(a)). The top enriched path-
ways included regulation of vasculature development,
angiogenesis, endothelium development, and cellular
migration. These findings suggested that MGs might
play critical roles in modulating the development
and maintenance of vascular structures and organ
systems. Enrichment in pathways related to cellular
Biofabrication 15 (2023) 045025
H Wu et al
Figure 2. (a) Live-dead staining of bioprinted HUVEC and HMC3 hydrogels cultured in selected glucose levels: 5.5 mM, 10 mM,
and 25 mM. Scale bar = 500 µm. The white arrow indicates closed-lumen structures. White arrowhead indicates open-lumen
structures. (b) Quantification of cell viability. At least triplicate measurements were obtained. Data are presented as mean ± SD.
Figure 3. (a) Tube formation testing in selected glucose levels, G5.5, G10, and G25. HUVECs were labeled with zsGreen and
HMC3 labeled with RFP. Scale bar = 500 µm. (b) Analysis of tube formation images using ImageJ angiogenesis analyzer plugin.
Blue: meshes. Green: branches to extremities. Magenta: segments. Dark blue: isolated elements. (c) Quantification of the number
of meshes and total mesh area (µm2). (d) Quantification of total isolated branch length and number of extremities.
migration, such as ameboid-type cell migration, EC
migration, and epithelial cell migration, implied their
role in tissue remodeling and repair processes.
Additionally, RNA-seq revealed the involvement
of cocultured MG in responding to changes in oxy-
gen levels, evidenced by the enrichment of pathways
7
Biofabrication 15 (2023) 045025
H Wu et al
Figure 4. (a) Immunofluorescent staining of IBA-1 in bioprinted co-culture hydrogels, with the nucleus counterstained using
DAPI. (b) Fluorescent imaging displaying the morphologies of microglia labeled with RFP, where G25 microglia exhibit an
amoeboid morphology. Scale bar = 50 µm.
related to hypoxia and oxygen level response. This
suggested a potential role of these cells in adapt-
ing to various physiological and pathological condi-
tions affecting oxygen availability. Furthermore, the
enriched pathways related to cell–cell adhesion, EC
proliferation and differentiation, and regulation of
apoptotic cell clearance indicated that cocultured MG
could contribute to tissue integrity and homeostasis.
GO analysis of molecular function in cocultured MG
further revealed these cells were also involved in cel-
lular communication, signaling, response to environ-
mental stimuli, protein binding, enzymatic activity,
and regulatory functions (figure 5(b)).
In cocultured EC compared to single-cultured
EC, 493 genes were upregulated, and 395 were down-
regulated (supplementary figure 3(b)). The GO ana-
lysis of the biological process in cocultured ECs
revealed significant enrichment in immune-related
pathways (figure 6(a)). The top enriched pathways
encompassed a range of processes, including defense
responses to viruses and other pathogens, regula-
tion of viral processes, cytokine-mediated signaling,
interferon production and response, and regulation
of innate immune response. These findings sugges-
ted that the coculture system might promote coordin-
ating immune responses against infections and fine-
tuning the innate immune response to prevent excess-
ive inflammation and tissue damage. GO analysis
of molecular function in cocultured ECs further
revealed significant enrichment in pathways includ-
ing cytokine activity, receptor ligand activity, and sig-
naling receptor activator activity (figure 6(b)). The
coculture system enhanced cellular communication
and signaling, with the cells involved in modulating
immune responses and other cellular processes. The
cocultured cells were also enriched in ECM inter-
actions and structural integrity, evidenced by the
enrichment of pathways related to integrin binding,
ECM structural constituent, and collagen binding.
The GO analysis revealed distinct patterns of gene
expression changes, validating the hypothesis that the
coculture system would provide a more biomimetic
and complex model for EC and MG.
3.4. Diabetic culture conditions alter gene
expression in cocultured EC and MG
In the subsequent section, we examined the effects of
high glucose conditions, G10 and G25, which sim-
ulated diabetic environments, in comparison to the
control group, G5.5, on cocultured ECs and MG.
Our analysis revealed more significant differences in
gene expression profiles between normal and diabetic
conditions, highlighting the substantial impact of the
diabetic milieu on cellular processes and pathways.
However, fewer differences were observed between
the two diabetic conditions with distinct glucose
levels, suggesting that the cellular responses might
have reached a saturation point or adapted to the dia-
betic environment (supplementary figures 4(a) and
(b)).
Analysis of
the differentially expressed genes
in MG under high glucose conditions, specific-
ally G10 compared to G5,
identified several key
upregulated genes and their associated pathways
8
Biofabrication 15 (2023) 045025
H Wu et al
Figure 5. (a) GO analysis of biological processes, showcasing the top 10 enriched pathways in co-cultured MG compared to
monoculture MG in G5.5 medium. (b) GO analysis of molecular functions, illustrating the top 10 enriched pathways in
co-cultured MG versus monoculture MG in G5.5 medium.
(figure 7, supplementary figure 5(a)). For instance, we
observed the upregulation of genes such as IL1RL1,
PTGS2, and NOS3, which were involved in regu-
lating inflammation and immune response, suggest-
ing that MG might exhibit increased inflammatory
activity in response to elevated glucose levels. Genes
like MUC5B, DIO2, and CDH11 were upregulated,
implicating changes in mucus production, thyroid
hormone metabolism, and cell adhesion. The upreg-
ulation of MUC5B has been associated with inflam-
mation and fibrosis, and studies have reported MG’s
role in secreting fibronectin and its binding pro-
teins to form ECM bridges [54, 55]. These altera-
tions might have influenced the interaction between
MG and their surrounding ECM environment under
high glucose conditions. We also noted the upreg-
ulation of genes involved in neural communication
and synaptic function, such as NLGN1 and GIPC2,
indicating potential changes in the communication
between MG and other cell types in the diabetic
milieu. A few genes and pathways were altered in
G25 vs. G10 conditions (supplementary figure 5(b)).
The upregulation of ST8SIA5, which was involved
in synthesizing glycoproteins and glycolipids, sug-
gested potential alterations in cell surface proper-
ties and interactions under the higher glucose con-
centration of G25. TG was involved in the form-
ation of the ECM, indicating that higher glucose
levels might have impacted the extracellular envir-
onment and, consequently, cellular communication
and signaling. The upregulation of genes such as
FAM156A and TLN2, which were involved in regulat-
ing cell adhesion and cytoskeletal organization, sug-
gested that the higher glucose concentration in G25
could have influenced MG adhesion and motility.
Genes associated with the TGF-beta signaling path-
way and chaperone-mediated protein folding, such
as ACVR2A and DNAJB14, indicated that MG might
9
Biofabrication 15 (2023) 045025
H Wu et al
Figure 6. (a) GO analysis of biological processes, presenting the top 10 enriched pathways in co-cultured EC relative to
monoculture EC in G5.5 medium. (b) GO analysis of molecular functions, revealing the top 10 enriched pathways in co-cultured
EC compared to monoculture EC in G5.5 medium.
have changed signaling and stress response mechan-
isms in the G25 condition.
Upon analysis of the differentially expressed genes
in EC under G10 compared to G5.5, we identi-
fied several key upregulated genes and their associ-
ated pathways (figure 8, supplementary figure 6(a)).
Notably, genes such as THBS1-AS1, EPHA1, SPON1,
and POSTN are involved in ECM remodeling and
cell adhesion. Additionally, upregulated genes like
ALOX12, PTGS2, and CYP1B1 regulate inflammation
and oxidative stress, while IGFBP3 and NR4A1 reg-
ulate cell growth and metabolism. Furthermore, we
observed the upregulation of genes like ATF3, EGR3,
and ZFP36, which are involved in stress response
and transcriptional regulation. These findings sug-
gest that high glucose conditions may affect pathways
related to ECM remodeling, inflammation, oxidative
stress, cell growth, and stress response in cocultured
ECs. Notably, some downregulated genes, such as
RPS28P7, HNRNPA1P48, RPL13AP5, and RPL7P9,
are involved in ribosomal function, and RNA pro-
cessing could impact protein synthesis in the cells
under high glucose conditions. The downregulation
of genes related to translation and RNA processing
may lead to reduced cellular activity and protein
expression.
Additionally, we observed the downregulation
of genes such as ISCA1P1 and COX7B, which are
associated with mitochondrial function and energy
metabolism. This could indicate decreased metabolic
activity or potential alterations in energy produc-
tion pathways under high glucose conditions. Other
genes, such as FZD3, are involved in the Wnt signaling
pathway, which plays a crucial role in cell differenti-
ation, proliferation, and migration. Downregulation
of FZD3 suggests potential changes in these pro-
cesses, which may affect the behavior of cocultured
EC in a high glucose environment. Upon analysis of
10
Biofabrication 15 (2023) 045025
H Wu et al
Figure 7. Heatmap representing the top differentially expressed genes (DEGs) of co-cultured MG under normal conditions (G5.5)
and diabetic conditions (G10, G25).
the genes upregulated in G25 compared to G10, we
identified a few genes and their associated pathways,
which may provide insights into the differential cellu-
lar responses between these two high glucose condi-
tions (supplementary figure 6(b)). Some upregulated
genes in G25, such as LINC00475 and CCDC171,
are long non-coding RNAs (lncRNAs) and coiled-
coil domain-containing proteins. The upregulation of
these genes might suggest alterations in gene regulat-
ory mechanisms under the higher glucose concentra-
tion of G25. UNC13A, another upregulated gene in
G25, is involved in synaptic vesicle priming and neur-
otransmitter release. The upregulation of UNC13A
may indicate potential changes in cellular commu-
nication and signaling pathways in response to the
higher glucose concentration. ALG14, BOLA2B, and
YPEL1 are genes involved in protein glycosylation,
iron-sulfur cluster assembly, and cell proliferation.
The upregulation of these genes in G25 compared
to G10 might indicate changes in protein modifica-
tion, cellular metabolism, and proliferation rates in
response to the higher glucose concentration.
3.5. Enhanced drug resistance in cocultured
models under diabetic conditions
Lastly, we employed the bioprinted models to evalu-
ate commonly used anti-VEGF DR drugs, including
aflibercept, conbercept, and ranibizumab. We con-
structed 3D models of EC monocultures, MG mono-
culture, and EC-MG cocultures, and tested the drugs
in three different glucose concentrations, G5.5, G10,
and G25.
We first
focused on the glucose impact on
drug efficacies. Upon treatment with a 10 µg ml−1
dosage, we observed that in EC monoculture samples,
both aflibercept and conbercept exhibited inhibit-
ory effects to varying degrees at G5.5 (figure 9(a)).
Aflibercept’s efficacy was reduced by 17% at 25 mM,
11
Biofabrication 15 (2023) 045025
H Wu et al
Figure 8. Heatmap representing the top differentially expressed genes (DEGs) of co-cultured EC under normal conditions (G5.5)
and diabetic conditions (G10, G25).
while Conbercept’s efficacy slightly decreased with
increasing glucose levels. HMC3 was unrespons-
ive to either drug, as anticipated (figure 9(b)).
In the coculture model, all three drugs displayed
greater efficacy reduction under high glucose condi-
tions (figure 9(c)). Both Conbercept and Aflibercept
were only effective at G5.5, with efficacy decreas-
ing by 18% and 25% for Aflibercept, and 31% and
30% for Conbercept at G10 and G25, respectively.
Ranibizumab did not exhibit significant inhibitory
effects in either monoculture or coculture models
under our testing conditions. These trends were con-
sistent with the observations made when using a
lower drug dosage of 1 µg ml−1 (supplementary
figures 7(a)–(c)).
We then analyzed the influence of cellular com-
position on drug responses at each glucose concen-
tration. Upon treatment with a 10 µg ml−1 dosage,
at the normoglycemic level G5.5, both the EC mono-
culture and the coculture exhibited sensitivity to
Aflibercept and Conbercept (supplementary figure
8(a)). Although MG did not respond to either drug,
its addition did not appear to alter drug efficacy in
coculture models. At the diabetic level G10, coculture
significantly reduced the efficacy by 32% and 31% for
Aflibercept and Conbercept, respectively, compared
to EC monoculture (supplementary figure 8(b)), and
by 6% for Ranibizumab, though not statistically signi-
ficant. At G25, Conbercept’s efficacy declined by 26%
in coculture compared to EC monoculture (supple-
mentary figure 8(c)). These observations were con-
sistent at a lower dosage of 1 µg ml−1 (supplementary
figures 9(a)–(c)).
In conclusion, our results suggest
that EC’s
response to anti-VEGF treatment may be diminished
in diabetic conditions, and MG’s presence may fur-
ther exert a protective effect on EC under high glucose
conditions. Our study highlights the importance of
considering glucose concentrations and cellular inter-
actions in evaluating drug resistance, emphasizing the
12
Biofabrication 15 (2023) 045025
H Wu et al
Figure 9. (a) Drug responses to aflibercept, conbercept, and ranibizumab at a drug concentration of 10 µg ml−1 in EC
monoculture under various glucose levels. (b) Drug responses to aflibercept, conbercept, and ranibizumab at a drug concentration
of 10 µg ml−1 in MG monoculture under various glucose levels. (c) Drug responses to aflibercept, conbercept, and ranibizumab
at a drug concentration of 10 µg ml−1 in EC-MG co-culture under different glucose levels. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.
value of cocultured models to enhance understand-
ing and potentially overcome drug resistance in DR
treatments.
4. Discussion
In conclusion, our study successfully established a
3D bioprinted coculture model with retina-mimetic
mechanical properties, offering a valuable platform
to investigate the interplay between EC and MG
under varying glucose conditions representative of
DR patients. The coculture displayed a protective
effect on cell proliferation in diabetic conditions,
emphasizing the potential roles of MG in promot-
ing cellular proliferation and modulating EC beha-
vior. RNA-seq analysis revealed that the coculture
system induced substantial changes in gene expres-
sion profiles of both EC and MG, with GO ana-
lysis uncovering distinct patterns. Cocultured MG
displayed enrichment in pathways related to vascu-
lature development, cellular migration, and response
to oxygen levels, while cocultured EC exhibited signi-
ficant enrichment in immune-related pathways. Our
study revealed a marked activation of MG under dia-
betic co-culture conditions, with pronounced IBA1
expression and morphological shifts, particularly in
the higher glucose concentrations (G10 and G25).
This finding built upon a complex landscape of MG’s
role in diabetes and offers a nuanced understanding of
their activation patterns. Traditionally, MG activation
has been classified into the M1 phenotype, associated
with pro-inflammatory functions, and the M2 phen-
otype, linked to anti-inflammatory responses and tis-
sue repair. These categories are stimulated by spe-
cific cytokines and environmental factors, reflecting
a dichotomy that is overly simplistic for the highly
plastic and heterogeneous (phenotypical, regional,
and functional) state of MG in DR revealed by single-
cell sequencing [56, 57]. This emerging knowledge
regarding the extensive variety of microglial pheno-
types underscores the imperative for employing more
sophisticated techniques and models for MG invest-
igation. The coculture system devised in this invest-
igation stands as a robust and versatile tool, facilit-
ating the precise study of MG and EC phenotypes.
By providing a controlled environment, our model
13
Biofabrication 15 (2023) 045025
H Wu et al
contributed new insights into their dynamic roles
in tissue remodeling, repair processes, and immune
responses, all under varying glucose scenarios.
Our findings have elucidated that glucose levels
exert a significant influence on EC and MG phen-
impacting
otypes within the coculture models,
angiogenic potential, morphological alterations, and
activation states. These observations emphasize the
central role of cellular interactions and environmental
factors in the progression of DR, while also highlight-
ing the multifaceted nature of cellular responses to
elevated glucose conditions. Changes in gene expres-
sion linked to inflammation, immune response, cell
adhesion, metabolism, and signaling pathways fur-
ther underscore this complexity. Evidence from prior
studies indicates that diabetic conditions can result
in the stiffening of the vascular basement membrane
[58]. Our bioprinted hydrogel seems to have partially
recapitulated these changes in ECM properties under
diabetic conditions. Specifically, at the glucose level
of 25 mM, there was a marked increase in hydrogel
stiffness relative to hydrogels incubated with lower
glucose concentrations. This result points to a more
comprehensive role for glucose in not only shap-
ing cellular behaviors but also in modulating ECM
properties.
The mechanisms driving the Influence of glucose
on ECM stiffness remain to be elucidated and warrant
further investigation, as they may yield insights into
novel therapeutic strategies. Additionally, our exam-
ination of drug resistance in cocultured models has
unveiled limitations in current DR treatments. The
diminished effectiveness of commonly used drugs,
such as aflibercept and conbercept in coculture, par-
ticularly at elevated glucose concentrations, under-
scores the urgent need for innovative approaches
that account for cellular interactions and microen-
vironmental factors in DR. In summary, our study
introduces a physiologically relevant 3D bioprinted
coculture model of MG and EC that simulates the nat-
ive retina vasculature environment for probing gluc-
ose’s effects on cells and ECM. Future research in this
direction promises to unveil the underlying mech-
anisms and identify potential targets to devise more
potent treatments. This would not only overcome
anti-VEGF resistance but also enhance our under-
standing of DR, paving the way for improved patient
outcomes.
Data availability statement
The data that support the findings of this study
are openly available at the following URL/DOI:
www.ncbi.nlm.nih.gov/bioproject/PRJNA953644.
The other data that supports the findings of this study
are available upon reasonable request.
Acknowledgments
This work was sponsored by Grants from Shanghai
Natural Science Foundation (20ZR1409700). Authors
Thank Kaiwen Tao and Xiaotong He for assistance
in the bioink preparation and bioprinter set up.
Figure 1(a) was generated using the licensed version
of BioRender.
Conflict of interest
The authors declare no conflict of interest.
ORCID iD
Min Tang https://orcid.org/0000-0002-6084-1827
References
[1] Ting D S W, Cheung G C M and Wong T Y 2016 Diabetic
retinopathy: global prevalence, major risk factors, screening
practices and public health challenges: a review Clin. Exp.
Ophthalmol. 44 260–77
[2] Teo Z L et al 2021 Global prevalence of diabetic retinopathy
and projection of burden through 2045 systematic review
and meta-analysis Ophthalmology 128 1580–91
[3] Antonetti D A, Klein R and Gardner T W 2012 Diabetic
retinopathy New Engl. J. Med. 366 1227–39
[4] Uemura A et al 2021 VEGFR1 signaling in retinal
angiogenesis and microinflammation Prog. Retin. Eye Res.
84 100954
[5] Klaassen I, Van Noorden C J F and Schlingemann R O 2013
Molecular basis of the inner blood-retinal barrier and its
breakdown in diabetic macular edema and other
pathological conditions Prog. Retin. Eye Res. 34 19–48
[6] Kang Q and Yang C 2020 Oxidative stress and diabetic
retinopathy: molecular mechanisms, pathogenetic role and
therapeutic implications Redox Biol. 37 101799
[7] Aiello L P et al 1994 Vascular endothelial growth factor in
ocular fluid of patients with diabetic retinopathy and other
retinal disorders New Engl. J. Med. 331 1480–7
[8] Rosenfeld P J, Brown D M, Heier J S, Boyer D S, Kaiser P K,
Chung C Y and Kim R Y 2006 Ranibizumab for neovascular
age-related macular degeneration New Engl. J. Med.
355 1419–31
[9] Wallsh J O and Gallemore R P 2021 Anti-VEGF-resistant
retinal diseases: a review of the latest treatment options Cells
10 1049
[10] Bressler N M, Beaulieu W T, Glassman A R, Blinder K J,
Bressler S B, Jampol L M, Melia M and Wells J A III for the
Diabetic Retinopathy Clinical Research Network 2018
Persistent macular thickening following intravitreous
aflibercept, bevacizumab, or ranibizumab for
central-involved diabetic macular edema with vision
impairment: a secondary analysis of a randomized clinical
trial JAMA Ophthalmol. 136 257–69
[11] Salter M W and Beggs S 2014 Sublime microglia: expanding
roles for the guardians of the CNS Cell 158 15–24
[12] Chen S et al 2017 CNS macrophages control neurovascular
development via CD95L Cell Rep. 19 1378–93
[13] Foulquier S et al 2019 The role of receptor MAS in
microglia-driven retinal vascular development Angiogenesis
22 481–9
[14] Checchin D, Sennlaub F, Levavasseur E, Leduc M and
Chemtob S 2006 Potential role of microglia in retinal blood
vessel formation Invest. Ophthalmol. Vis. Sci. 47 3595–602
14
Biofabrication 15 (2023) 045025
H Wu et al
[15] Ding X, Gu R, Zhang M, Ren H, Shu Q, Xu G and Wu H
[35] Li X, Liu B, Pei B, Chen J, Zhou D, Peng J, Zhang X, Jia W
2018 Microglia enhanced the angiogenesis, migration and
proliferation of co-cultured RMECs BMC Ophthalmol.
18 249
[16] Schnichels S, Paquet-Durand F, Löscher M, Tsai T, Hurst J,
Joachim S C and Klettner A 2021 Retina in a dish: cell
cultures, retinal explants and animal models for
common diseases of the retina Prog. Retin. Eye Res.
81 100880
[17] Olivares A M, Althoff K, Chen G F, Wu S, Morrisson M A,
DeAngelis M M and Haider N 2017 Animal models of
diabetic retinopathy Curr. Diabetes Rep. 17 93
[18] Eastlake K, Banerjee P J, Angbohang A, Charteris D G,
Khaw P T and Limb G A 2016 Müller glia as an important
source of cytokines and inflammatory factors present in the
gliotic retina during proliferative vitreoretinopathy Glia
64 495–506
[19] Arroba A I, Mazzeo A, Cazzoni D, Beltramo E, Hernández C,
Porta M, Simó R and ´A M V 2016 Somatostatin protects
photoreceptor cells against high glucose–induced apoptosis
Mol. Vis. 22 1522–31
[20] Zhang Y, Lv X, Hu Z, Ye X, Zheng X, Ding Y, Xie P and Liu Q
2017 Protection of Mcc950 against high-glucose-induced
human retinal endothelial cell dysfunction Cell Death Dis.
8 e2941
[21] Valdés J, Trachsel-Moncho L, Sahaboglu A, Trifunovi´c D,
Miranda M, Ueffing M, Paquet-Durand F and
Schmachtenberg O 2016 Organotypic retinal explant
cultures as in vitro alternative for diabetic retinopathy
studies Altex 33 459–64
[22] Chen K, Rowley A P, Weiland J D and Humayun M S 2014
Elastic properties of human posterior eye J. Biomed. Mater.
Res. A 102 2001–7
[23] Ferrara M, Lugano G, Sandinha M T, Kearns V R,
Geraghty B and Steel D H W 2021 Biomechanical properties
of retina and choroid: a comprehensive review of techniques
and translational relevance Eye 35 1818–32
[24] Aizawa Y and Shoichet M S 2012 The role of endothelial cells
in the retinal stem and progenitor cell niche within a 3D
engineered hydrogel matrix Biomaterials 33 5198–205
[25] Paek J et al 2019 Microphysiological engineering of
self-assembled and perfusable microvascular beds for the
production of vascularized three-dimensional human
microtissues ACS Nano 13 7627–43
[26] Chung M, Lee S, Lee B J, Son K, Jeon N L and Kim J H 2018
Wet-AMD on a chip: modeling outer blood-retinal barrier
in vitro Adv. Healthcare Mater. 7 1700028
[27] Achberger K et al 2019 Merging organoid and
organ-on-a-chip technology to generate complex multi-layer
tissue models in a human retina-on-a-chip platform eLife
8 e46188
[28] Song M J et al 2023 Bioprinted 3D outer retina barrier
uncovers RPE-dependent choroidal phenotype in advanced
macular degeneration Nat. Methods 20 149–61
[29] Khodamoradi M, Eskandari M, Keshvari H and Zarei R 2021
An electro-conductive hybrid scaffold as an artificial Bruch’s
membrane Mater. Sci. Eng. C 126 112180
[30] Akbari M and Khademhosseini A 2022 Tissue bioprinting
for biology and medicine Cell 185 2644–8
and Xu T 2020 Inkjet bioprinting of biomaterials Chem. Rev.
120 10596–636
[36] Gudapati H, Dey M and Ozbolat I 2016 A comprehensive
review on droplet-based bioprinting: past, present and
future Biomaterials 102 20–42
[37] Xu T, Zhao W, Zhu J-M, Albanna M Z, Yoo J J and Atala A
2013 Complex heterogeneous tissue constructs containing
multiple cell types prepared by inkjet printing technology
Biomaterials 34 130–9
[38] You S et al 2023 High cell density and high-resolution 3D
bioprinting for fabricating vascularized tissues Sci. Adv.
9 eade7923
[39] Shah M, Ullah A, Azher K, Rehman A U, Juan W, Aktürk N,
Tüfekci C S and Salamci M U 2023 Vat
photopolymerization-based 3D printing of polymer
nanocomposites: current trends and applications RSC Adv.
13 1456–96
[40] Yu C, Schimelman J, Wang P, Miller K L, Ma X, You S,
Guan J, Sun B, Zhu W and Chen S 2020 Photopolymerizable
biomaterials and light-based 3D printing strategies
for biomedical applications Chem. Rev.
120 10695–743
[41] Tang M et al 2020 Three-dimensional bioprinted
glioblastoma microenvironments model cellular
dependencies and immune interactions Cell Res.
30 833–53
[42] Tang M et al 2021 Rapid 3D bioprinting of glioblastoma
model mimicking native biophysical heterogeneity Small
17 2006050
[43] Yu C et al 2020 A sequential 3D bioprinting and orthogonal
bioconjugation approach for precision tissue engineering
Biomaterials 258 120294
[44] Wang P, Li X, Zhu W, Zhong Z, Moran A, Wang W, Zhang K
and Chen S 2018 3D bioprinting of hydrogels for retina cell
culturing Bioprinting 11 e00029
[45] Liu Y, Wang R, Zarembinski T I, Doty N, Jiang C,
Regatieri C, Zhang X and Young M J 2013 The application of
hyaluronic acid hydrogels to retinal progenitor cell
transplantation Tissue Eng. A 19 135–42
[46] Carpentier G, Berndt S, Ferratge S, Rasband W, Cuendet M,
Uzan G and Albanese P 2020 Angiogenesis analyzer for
imageJ—a comparative morphometric analysis of
“endothelial tube formation assay” and “fibrin bead assay
Sci. Rep. 10 11568
[47] Ma X, Yu C, Wang P, Xu W, Wan X, Lai C S E, Liu J,
Koroleva-Maharajh A and Chen S 2018 Rapid 3D
bioprinting of decellularized extracellular matrix with
regionally varied mechanical properties and biomimetic
microarchitecture Biomaterials 185 310–21
[48] Qu Y, He Y, Zhang Y, Ma T, Zhu J, Miao Y, Dai C,
Humayun M, Zhou Q and Chen Z 2018 Quantified elasticity
mapping of retinal layers using synchronized acoustic
radiation force optical coherence elastography Biomed. Opt.
Express 9 4054–63
[49] Wang Q, Zhang X, Wang K, Zhu L, Qiu B, Chen X, Lin X
and Nie Y 2021 An in vitro model of diabetic retinal vascular
endothelial dysfunction and neuroretinal degeneration J.
Diabetes Res. 2021 9765119
[31] Daly A C, Prendergast M E, Hughes A J and Burdick J A 2021
[50] Fiorello M L, Treweeke A T, Macfarlane D P and Megson I L
Bioprinting for the biologist Cell 184 18–32
[32] Murphy S V, De Coppi P and Atala A 2020 Opportunities
and challenges of translational 3D bioprinting Nat. Biomed.
Eng. 4 370–80
2020 The impact of glucose exposure on bioenergetics and
function in a cultured endothelial cell model and the
implications for cardiovascular health in diabetes Sci. Rep.
10 19547
[33] Ozbolat I T and Hospodiuk M 2016 Current advances and
[51] Mouri M and Badireddy M 2023 Hyperglycemia StatPearls
future perspectives in extrusion-based bioprinting
Biomaterials 76 321–43
[34] Ouyang L, Yao R, Zhao Y and Sun W 2016 Effect of bioink
properties on printability and cell viability for 3D
bioplotting of embryonic stem cells Biofabrication 8 035020
(StatPearls Publishing)
[52] Ito D, Imai Y, Ohsawa K, Nakajima K, Fukuuchi Y and
Kohsaka S 1998 Microglia-specific localisation of a novel
calcium binding protein, Iba1 Brain Res. Mol. Brain Res.
57 1–9
15
Biofabrication 15 (2023) 045025
H Wu et al
[53] Trotta M C, Gesualdo C, Petrillo F, Cavasso G, Corte A D,
D’Amico G, Hermenean A, Simonelli F and Rossi S 2022
Serum Iba-1, GLUT5, and TSPO in patients with diabetic
retinopathy: new biomarkers for early retinal neurovascular
alterations? A pilot study Transl. Vis. Sci. Technol. 11 16
[54] Seibold M A et al 2011 A common MUC5B promoter
polymorphism and pulmonary fibrosis New Engl. J. Med.
364 1503–12
[55] Khorasani A M, Mohammadi B, Saghafi M R,
Mohammadi S, Ghaffari S, Mirsadraee M and Khakzad M R
2023 The association between MUC5AC and MUC5B genes
expression and remodeling progression in severe
neutrophilic asthma: a direct relationship Respir. Med.
213 107260
[56] Hu Z et al 2022 Single-cell transcriptomics reveals novel role
of microglia in fibrovascular membrane of proliferative
diabetic retinopathy Diabetes 71 762–73
[57] Uddin Md I, Kilburn T C, Duvall C L and Penn J S 2020
Visualizing HIF-1α mRNA in a subpopulation of bone
marrow-derived cells to predict retinal neovascularization
ACS Chem. Biol. 15 3004–12
[58] Yang X et al 2016 Basement membrane stiffening promotes
retinal endothelial activation associated with diabetes FASEB
J. 30 601–11
16
| null |
10.1371_journal.pclm.0000285.pdf
| null |
R code and aggregated data used in climate risk calculations are available at https://github.com/groundfish- climatechange/fish-footprints . Confidential data may be acquired
|
RESEARCH ARTICLE
Stay or go? Geographic variation in risks due
to climate change for fishing fleets that adapt
in-place or adapt on-the-move
Jameal F. SamhouriID
Kate RichersonID
7, Lyall BellquistID
H. BeaudreauID
Abigail Harley11, Chris J. HarveyID
Amanda Phillips1,5, Leif K. RasmusonID
L. SeldenID
14
1*, Blake E. Feist1, Michael Jacox2, Owen R. LiuID
3,4,
5, Erin Steiner5, John Wallace5, Kelly Andrews1, Lewis Barnett6, Anne
8,9, Mer Pozo BuilID
1, Isaac C. Kaplan1, Karma NormanID
2, Melissa A. HaltuchID
1,
5,10,
12,13, Eric J. Ward1, Curt WhitmireID
5, Rebecca
1 Conservation Biology Division, Northwest Fisheries Science Center, National Marine Fisheries Service,
National Oceanic and Atmospheric Administration, Seattle, Washington, United States of America,
2 Environmental Research Division, Southwest Fisheries Science Center, National Marine Fisheries Service,
National Oceanic and Atmospheric Administration, Monterey, California, United States of America, 3 Under
Contract to the Northwest Fisheries Science Center, National Marine Fisheries Service, National Oceanic and
Atmospheric Administration, Ocean Associates, Inc., Seattle, Washington, United States of America, 4 NRC
Research Associateship Program, Northwest Fisheries Science Center, National Marine Fisheries Service,
National Oceanic and Atmospheric Administration, Seattle, Washington, United States of America, 5 Fishery
Resource Analysis and Monitoring Division, Northwest Fisheries Science Center, National Marine Fisheries
Service, National Oceanic and Atmospheric Administration, Seattle, Washington, United States of America,
6 Resource Assessment and Conservation Engineering Division, Alaska Fisheries Science Center, National
Marine Fisheries Service, National Oceanic and Atmospheric Administration, Seattle, Washington, United
States of America, 7 School of Marine and Environmental Affairs, University of Washington, Seattle,
Washington, United States of America, 8 The Nature Conservancy, Sacramento, California, United States of
America, 9 Scripps Institution of Oceanography, La Jolla, California, United States of America, 10 Resource
Ecology and Fisheries Management Division, Alaska Fisheries Science Center, National Marine Fisheries
Service, National Oceanic and Atmospheric Administration, Seattle, Washington, United States of America,
11 Sustainable Fisheries Division, West Coast Region, National Marine Fisheries Service, National Oceanic
and Atmospheric Administration, Seattle, Washington, United States of America, 12 Marine Fisheries
Research Project, Marine Resources Program, Oregon Department of Fish and Wildlife, Newport, Oregon,
United States of America, 13 Department of Fisheries, Wildlife, and Conservation Sciences, Oregon State
University, Corvallis, Oregon, United States of America, 14 Department of Biological Sciences, Wellesley
College, Wellesley, Massachusetts, United States of America
* [email protected]
Abstract
From fishers to farmers, people across the planet who rely directly upon natural resources
for their livelihoods and well-being face extensive impacts from climate change. However,
local- and regional-scale impacts and associated risks can vary geographically, and the
implications for development of adaptation pathways that will be most effective for specific
communities are underexplored. To improve this understanding at relevant local scales, we
developed a coupled social-ecological approach to assess the risk posed to fishing fleets by
climate change, applying it to a case study of groundfish fleets that are a cornerstone of fish-
eries along the U.S. West Coast. Based on the mean of three high-resolution climate projec-
tions, we found that more poleward fleets may experience twice as much local temperature
change as equatorward fleets, and 3–4 times as much depth displacement of historical
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Samhouri JF, Feist BE, Jacox M, Liu OR,
Richerson K, Steiner E, et al. (2024) Stay or go?
Geographic variation in risks due to climate change
for fishing fleets that adapt in-place or adapt on-
the-move. PLOS Clim 3(2): e0000285. https://doi.
org/10.1371/journal.pclm.0000285
Editor: Athanassios C. Tsikliras, Aristotle University
of Thessaloniki, GREECE
Received: August 11, 2023
Accepted: December 28, 2023
Published: February 9, 2024
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pclm.0000285
Copyright: This is an open access article, free of all
copyright, and may be freely reproduced,
distributed, transmitted, modified, built upon, or
otherwise used by anyone for any lawful purpose.
The work is made available under the Creative
Commons CC0 public domain dedication.
Data Availability Statement: R code and
aggregated data used in climate risk calculations
are available at https://github.com/groundfish-
climatechange/fish-footprints. Confidential vessel-
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
1 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
level logbook, landings, and registration data may
be acquired by direct request from the California,
Oregon, and Washington Departments of Fish and
Wildlife, subject to a non-disclosure agreement.
Funding: JFS received funding for this work from
the the David and Lucille Packard Foundation
2019-69817. The work of ORL and RLS was
supported by that funding. The funders had no role
in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
environmental conditions in their fishing grounds. Not only are they more highly exposed to
climate change, but some poleward fleets are >10x more economically-dependent on
groundfish. While we show clear regional differences in fleets’ flexibility to shift to new fisher-
ies via fisheries diversification (‘adapt in-place’) or shift their fishing grounds in response to
future change through greater mobility (‘adapt on-the-move’), these differences do not
completely mitigate the greater exposure and economic dependence of more poleward
fleets. Therefore, on the U.S. West Coast more poleward fishing fleets may be at greater
overall risk due to climate change, in contrast to expectations for greater equatorward risk in
other parts of the world. Through integration of climatic, ecological, and socio-economic
data, this case study illustrates the potential for widespread implementation of risk assess-
ment at scales relevant to fishers, communities, and decision makers. Such applications will
help identify the greatest opportunities to mitigate climate risks through pathways that
enhance flexibility and other dimensions of adaptive capacity.
Introduction
Climate change is shaping the availability of nature’s benefits to people and will continue to do
so for generations [1,2]. While global-scale projections provide coarse, qualitative expectations
for how climate impacts will manifest in different regions and sectors, there is much more lim-
ited understanding of risks due to climate change at local scales. Yet regionally-specific infor-
mation about the effects of biophysical changes on natural resource-dependent industries and
communities is critical for adaptation planning and strategic responses from resource manage-
ment agencies [3–5]. For communities that rely upon harvest of natural resources for their
lives and livelihoods, the scale and intensity of expected environmental change in customary
use areas for agriculture, fisheries, forestry, and other industries is especially important [6,7].
A clear challenge lies in determining how adaptation within or outside of these areas can
enhance climate resilience, using tractable, resonant, and scalable approaches.
Environmental change is spatially heterogeneous and will intersect with dynamic social fac-
tors to determine risk due to climate change [3,8,9]. For instance, it is already apparent that
rates of warming at the poles exceed those toward the equator [10], patterns of historical vari-
ability in local physical forcing will interact with anthropogenic climate change to determine
future conditions [11–14], and short-term extreme events fueled by climate change, as well as
long-term gradual change, can create localized hotspots of impact [15,16]. In the ocean, warm-
ing waters can cause shifts in species’ ranges or alterations in target species productivity that
lead to changes in local abundance that vary over space [17–19]. This heterogeneity will fuel
divergent ecological responses of species to create spatial variability in the exposure of human
communities to these impacts [20].
Social vulnerability of human communities, based on their sensitivity and adaptive capacity
to respond to biophysical changes, also varies geographically. For fisheries and fishing commu-
nities, the potential to adapt to change–whether driven by climate, markets, regulations, or
other factors–differs enormously based on a variety of historical contingencies as well as con-
temporary circumstances [21–26]. For example, the diversity of species a fishing community
has access to or other potential sources of non-fishing revenue can act as buffers during times
of ecological or financial volatility [27]. The ability to cope, adapt, and transform fishing prac-
tices in response to climate change [28] is influenced strongly by variation across domains of
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
2 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
adaptive capacity, which include assets, flexibility, organization, learning, and agency [20,29–
31]. A recurrent challenge lies in determining how to measure and manage these different
domains of adaptive capacity in tangible ways. Coupled social-ecological analyses of a fishing
community’s risk due to climate change integrate the magnitude of environmental change it
will experience, the sensitivity to such change, and adaptive capacity.
The flexibility domain of adaptive capacity (e.g., occupational multiplicity, technological
diversity; [30]) is especially pertinent to fishing communities. The potential for spatial redistri-
bution of target species due to changing ocean conditions encourages particular focus on two
of the more tangible, and non mutually-exclusive, attributes of flexibility: fisher or fleet mobil-
ity and species diversification. More mobile fishers and fleets can ‘adapt on-the-move’,
responding to changes in the availability of target species by changing where they fish [32],
while more diversified fishers may ‘adapt in-place’, continuing to operate in historical fishing
grounds while switching species [33]. Scientific advice that captures variability in mobility and
diversification provides effective support for decision makers managing fisheries in the face of
climate change [29].
In much of Europe and North America, groundfish fishing fleets that use bottom trawl gear
to target demersal species have formed the backbone of fishing communities for decades to
centuries. Many of the most well-developed future projections of the impacts of climate change
for fisheries are rooted in predictions of declining abundance of groundfish species (e.g.,
[17,34–36]), which tend to be characterized by high-quality, fishery-independent data, strongly
influenced by environmental forcing, and prone to overfishing due to their life-history charac-
teristics. Surprisingly, however, there are relatively few studies that explicitly connect climate
change to coupled social-ecological risk for groundfish fishing fleets. On the U.S. West Coast,
this gap in understanding is a crucial one, as the groundfish fishery in this region is a corner-
stone of the commercial fishing industry and economies of entire fishing communities [37–
39]. Groundfish are caught by bottom trawl off of the coasts of California, Oregon, and Wash-
ington, including catch by some vessels participating in state-managed bottom trawl fisheries
that capture federally-managed groundfish incidentally. Most catch is managed under the
Pacific Coast Groundfish Fishery Management Plan by the Pacific Fishery Management Coun-
cil (PFMC). This federally-managed fishery consists of nearly 100 species that include rock-
fishes (Sebastes spp.), roundfishes (e.g., sablefish), and flatfishes (e.g., Dover sole). The bottom
trawl groundfish fishery once generated >$100M USD (2021 USD) and engaged >400 vessels
across all three US West Coast states (Fig 1A and 1B). As of 2019, these values have fallen by a
factor of five or more, with annual revenues at just over $20M USD and fewer than 75 vessels
remaining in the fleet despite consistency in the number of port groups buying bottom trawl
groundfish over the same time period (Fig 1C and 1D).
While several West Coast groundfish stocks were rebuilt during the last two decades [40]
and total allowable catches have been increasing [41], utilization of many species remains low
[42], and much of the revenue generated from this fishery is now concentrated within fewer
ports, primarily in Oregon (Fig 1E). These patterns coincide with declines in the number of
fish buyers, reduced processing capacity, and increased spatial consolidation of processing,
which in turn may impact the magnitude and distribution of fishing effort [37,43,44].
Together, these trends suggest that port-level bottom trawl groundfish fishing fleets (hereafter,
groundfish fleets) are a useful set of fleets on which to focus because each is subject to the same
regulations and market forces, operates within a similar geographic area, experiences environ-
mentally-driven change in species’ availability, and therefore shares common opportunities
and challenges.
The confluence of long-term declines in revenue and participation along with increased
geographic consolidation (Fig 1E) suggests that the risk due to climate change for U.S. West
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
3 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
Fig 1. Historical changes in the groundfish fishery. (a) Ex-vessel revenue coastwide, (b) mean (±SD) annual ex-vessel revenue by state for 2011–2019, (c)
number of port groups, (d) number of vessels, and (e) revenue consolidation (estimated with the absolute Theil Index, calculated for each port group; [45]). A
port group represents a collection of individual ports; these groups were developed by the Pacific Fisheries Management Council (S1 Table). All revenue data
were adjusted for inflation to 2021 USD. See S1 Text for methodological details.
https://doi.org/10.1371/journal.pclm.0000285.g001
Coast groundfish fleets may be high and heterogeneous, yet neither these risks nor regional
variability in the potential for these fleets to mitigate risk has been rigorously explored. To
close this knowledge gap, we assessed the coupled social-ecological risk of groundfish fleets
along the U.S. West Coast to climate change. We focused this assessment on projected envi-
ronmental change within present-day fishing grounds, in combination with quantitative anal-
yses surrounding the economic dependence of the fleets on groundfish and the fleets’ relative
mobility and capacity to diversify into other fisheries, based on past fishing behaviors. We
hypothesize that regional variation in the magnitude of future ocean change will create geo-
graphically variable exposure. In addition, we predict that consolidation of the groundfish fleet
over time has concentrated economic dependence on bottom trawl-caught groundfish in
fewer places, altering sensitivity to future changes in groundfish fisheries. Finally, we expect
that fleet composition and fisheries portfolios vary from place to place, causing inconsistency
in the capacity for fleets to cope with risk posed by climate change across the coast.
Methods
Overview
We approached the question of what climate change portends for groundfish fleets on the U.S.
West Coast using a coupled social-ecological approach. We define coupled social-ecological
risk due to climate change as the combination of exposure to projected environmental or
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
4 / 28
204060801001990200020102020Revenue (millions)ACaliforniaOregonWashington05101520Revenue(mean ± 1SD, millions)2011−2019B141618201990200020102020Number of Port GroupsC1002003004001990200020102020Number of VesselsD0.40.60.81.01990200020102020Absolute Theil IndexEPLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
Fig 2. Conceptual framework to consider coupled social-ecological risk due to climate change. (a) Assuming fleets change target species while remaining in
current fishing grounds (adapt in-place); (b) assuming fleets shift fishing grounds while targeting current species (adapt on-the-move). We define coupled
social-ecological risk due to climate change as the combination of exposure to projected environmental or ecological change and the sensitivity and adaptive
capacity (i.e., social vulnerability) of the affected community, or more formally, Risk = (Exposure2+Vulnerability2)1/2 (Eq 7) where Vulnerability =
(Sensitivity2+(Lack of Adaptive Capacity) 2)1/2 (Eq 6). This approach is adapted from frameworks in [3,20]. In both panels, redder colors indicate higher
exposure due to warming.
https://doi.org/10.1371/journal.pclm.0000285.g002
ecological change and the social vulnerability of the affected community. Social vulnerability is
defined in terms of sensitivity and adaptive capacity. We assessed fleet-specific risk in two
ways (Fig 2). First, we evaluated risk if fleets change target species while continuing to fish in
current fishing grounds (the adapt in-place assessment). Second, we assessed risk if fleets shift
fishing grounds while targeting current species (the adapt on-the-move assessment). This eval-
uation builds on the general framework of the Intergovernmental Panel on Climate Change
(IPCC) [3], and more recent reviews and developments introduced by [9,26,46–48]. We define
each groundfish fleet as the collection of vessels landing groundfish caught using bottom trawl
gear and delivered to buyers in the same port group (S1 Table). We note that this definition of
a groundfish fleet is inclusive of vessels with federal permits for the fishery and vessels partici-
pating in state-managed bottom trawl fisheries that capture federally-managed groundfish
incidentally.
For the adapt in-place assessment, we estimated exposure as the amount of thermal change
expected between the periods 1990–2020 and 2065–2095 within the present-day fishing
grounds used by each fleet. We estimated the flexibility dimension of adaptive capacity based
on an index of diversification, defined as realized opportunities to participate in multiple fish-
eries in each port group from 2011–2019, and encompassing a recent period of consistent
management regulations [37].
For the adapt on-the-move assessment, we estimated exposure as the projected extent of
horizontal (change in latitude and/or longitude) and vertical (change in depth) displacement
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
5 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
of near-bottom isotherms representative of present-day fishing grounds for each fleet between
the periods 1990–2020 and 2065–2095 (S1 and S2 Figs; [49]). We estimated the flexibility
dimension of adaptive capacity based on an index of mobility, defined based on documented
distances of fishing grounds from landing ports during 2011–2019.
For both the adapt in-place and adapt on-the-move assessments, we defined sensitivity as
the economic dependence of each groundfish fleet on bottom trawl groundfish relative to total
commercial fishing revenue, including pink shrimp, Dungeness crab, and Pacific whiting, gen-
erated by those fleets within the U.S. Exclusive Economic Zone and state waters during the
period of 2011–2019. This approach assumes that more economically-dependent fleets are
more susceptible to harm if climate change negatively affects bottom trawl groundfish. To esti-
mate overall risk due to climate change for groundfish fleets, we calculated a social vulnerabil-
ity index based on the sensitivity and adaptive capacity estimates, and combined it with
estimates of exposure. We describe these calculations in detail below.
Defining fishing footprints
The foundation of this risk assessment is the location of fishing grounds for each groundfish
fleet. We defined the spatial footprints of each of 14 fleets based on fishery-dependent catch
data available from logbooks from 2011–2019 in Washington, Oregon, and California. We
retrieved these data from the Pacific Fisheries Information Network (PacFIN; http://pacfin.
psmfc.org). To connect these data with specific fishing communities, we associated footprints
with port groups of landing for each bottom trawl tow in the database (following [50,51]; S1
Table). There are nearly 300 ports where groundfish are landed and the distinction between
ports can often be as small as two different sides of a small bay. The port groupings were devel-
oped by the PFMC for biennial groundfish harvest specifications. In addition, aggregating
individual ports into port groups is necessary to provide a feasible set of geographic areas for a
coastwide climate risk analysis. Finally, analysis at the individual port-level would violate con-
fidentiality requirements, because there are often fewer than three buyers in any one port.
We pre-processed the logbook data to remove problematic hauls prior to development of
footprints (https://zenodo.org/record/7916821). Specifically, we included hauls lasting at least
0.2 hours but not more than 24 hours, and removed hauls with coordinates outside of the U.S.
EEZ, and those on land or outside of a customary catch depth (>2,000 m) or area (defined
based on locations of bottom trawl tows during the period 2010–2015). We evaluated the
depth reported for each haul using the Imap R package (https://github.com/John-R-Wallace-
NOAA/Imap), which overlays hauls with the National Geophysical Data Center (NGDC)
bathymetry (at a resolution of 3 arc-seconds, or ~90m at the Equator) [52–54]. We retained
hauls reporting a depth within 250 m of the NGDC depth, assuming accurate reported haul
locations. However, we assumed that if reported depths were inaccurate by >250 m, the haul
locations were likely to be similarly erroneous. Finally, we assumed that failure to report depth
was not indicative of positional error, but a simple misstep on the skipper’s part, so we
acquired the missing depth from NGDC based on the geocoordinates of the set (start) point
for each haul. Combined, these filters reduced the size of the logbook dataset by ~4% across all
years (S2 Table).
For each fleet, we extracted all tows from the period 2011–2019 from the logbook data,
excluding fleets with fewer than 3 vessels reporting logbook data during that time period. We
used the summed weight of landed catch of all groundfish species actively managed or listed as
ecosystem component species in the groundfish fishery management plan used by the PFMC
(Tables 3–1, 3–2 in https://www.pcouncil.org/documents/2016/08/pacific-coast-groundfish-
fishery-management-plan.pdf/), along with the geocoordinates of trawl set points, to create a
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
6 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
kernel density surface [33]. We calculated kernel density with a 10 km bandwidth, using the
density.ppp function in the sp package in R [55]. The kernel density allowed us to define the
footprint of each fleet, using a percent volume contour that represents the boundary of the
area that contains 75% of the volume of the kernel density distribution. The percent volume
contour was determined using the getvolumeUD function in the adehabitat package in R [56].
The position of each fleet’s footprint on the coast was relatively unchanged by the choice of the
50, 75, 90, or 95 percent volume contour (S4 Fig), and would not influence the rank order
exposure of fleets, or the relationships between exposure and latitude, described below given
the large-scale patterns of projected bottom temperature change, horizontal displacement, and
vertical displacement (S1 and S2 Figs).
Exposure
Poor ocean bottom conditions are the most relevant hazard for the life stages of groundfish
species caught with bottom trawl gear, and temperature is an established predictor of ground-
fish species’ range shifts [57]. We obtained projected bottom temperatures–the basis for a
regional assessment of hazard–from an ensemble of regional downscaled ocean projections
[11] produced using the Regional Ocean Modeling System (ROMS; S1 Fig). The ROMS
domain spans the California Current ecosystem from 30˚-48˚N latitude and from the coast to
134˚W longitude at 0.1˚ degree (~7–11 km) horizontal resolution with 42 terrain-following
vertical layers. The regional projections were forced with output from three Earth System
Models (ESMs) contributing to phase 5 of the Coupled Model Intercomparison Project
(CMIP5): Geophysical Fluid Dynamics Laboratory (GFDL) ESM2M, Hadley Center Had-
GEM2-ES (HADL), and Institut Pierre Simon Laplace (IPSL) CM5A-MR. While we only used
the high-emissions Representative Concentration Pathway (RCP) 8.5 scenario, which is the
highest-emission scenario and one which appears to be increasingly unlikely [58], the ESMs
were chosen to bracket the spread of potential future change. Specifically, GFDL and HADL
represent low and high ends of the spectrum, respectively, for the projected magnitude of
warming in the CMIP5 ensemble [11,59]. The relatively weak warming in GFDL under
RCP8.5 is comparable to the CMIP5 ensemble mean warming under RCP4.5. We focused on
30-year historic (1990–2020) and future (2065–2095) periods to best capture interdecadal vari-
ability [59] in ocean conditions characteristic of the California Current ecosystem.
We estimated exposure based on analysis of projected bottom temperatures within each
fleet’s fishing footprint. For the adapt in-place assessment, we calculated exposure eadaptin−place,p
for each fleet operating out of port group p as the thermal state change normalized by historic
thermal variability within each fishing footprint, addressing the question: if the footprint of fish-
ing effort for a fleet remains stationary, how much will the environment change within it rela-
tive to the scale of variability it normally experiences?
To obtain estimates of eadaptin−place, ESM,p for each ESM we spatially joined bottom tempera-
ture projections to the fleet footprints (using the sf library in R; [60]), and calculated the mean
and standard deviation in bottom temperature during the historic period, thistoric,ESM,p,c and
σhistoric,ESM,p,c, respectively, and the mean bottom temperature during the future period, tfuture,
ESM,p,c, for each ROMS cell c within each footprint. We estimated exposure as the difference in
the average future and historic temperatures across all cells within each footprint, t future;ESM;p
and t historic;ESM;p, divided by the average standard deviation in historic bottom temperature
across all cells within each footprint, shistoric;ESM;p, or
eadaptin(cid:0) place;ESM;p ¼
t future;ESM;p (cid:0)
t historic;ESM;p
shistoric;ESM;p
:
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
ð1Þ
7 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
Therefore, the units for this exposure metric are essentially standard deviations of tempera-
ture change relative to the historic baseline.
For the adapt on-the-move assessment, we calculated exposure for each fleet based on hori-
zontal (change in latitude and/or longitude) and vertical (change in depth) displacement of
isotherms representative of present-day fishing grounds (S2 Fig). Displacement is a metric that
characterizes environmental change in terms of the minimum distance that must be traveled
to track constant temperature contours [49], addressing the question: if the footprint of fishing
effort for a fleet moves to find a future environment that matches the historical one, how far
will it have to go? In the case of bottom temperature, we calculated both horizontal and vertical
displacement for each ROMS cell. We excluded ROMS cells in which >10% of their area was
inaccessible to the trawl fishery due to presence of untrawlable habitat or the most recent spa-
tial fishery regulations (2020-present; S2 Text, S3 Fig). Sensitivity analysis revealed that the
choice of the 10% threshold for inaccessible habitat did not qualitatively change conclusions.
To capture movement on finer spatial scales than the 0.1˚ degree resolution of the ROMS out-
put, displacements were interpolated to capture the minimum distance required (i.e., it is not
necessary to move a full 0.1˚ degree to the next grid cell if a partial movement would account
for the temperature change). As with eadaptin−place,ESM,p, we joined the summaries of displace-
ment to the fleet footprints, and calculated the average value of horizontal and vertical dis-
placement for each fleet and ESM, or eadapton−the−move,ESM,Hd,p and eadapton−the−move,ESM,VD,p,
respectively. The units for the horizontal and vertical displacement metrics are in kilometers
that would have to be shifted to maintain an isotherm.
Sensitivity
We calculated sensitivity in the same way for both the adapt in-place and adapt on-the-move
assessments, focusing on the economic dependence of fleets on bottom trawl groundfish. To
obtain information on fisheries landings by port group, on 3 October 2022 we downloaded
data for all bottom trawl groundfish vessels for the period 2011–2019 from PacFIN’s compre-
hensive fish tickets table. We calculated sensitivity s of vessel v in year y to changes in revenue
r (adjusted for inflation to 2021 USD) from the bottom trawl-caught groundfish gbt in port
group p in relation to all fisheries f and port groups in which it participates as
sf ¼gbt;p;y;v ¼
PP
rf ¼gbt;p;y;v
PF
p¼1
f ¼1 rf ;p;y;v
:
ð2Þ
We calculated annual sensitivity of each fleet Sf = gbt,p,y based on the median value of sf = gbt,
p,y across vessels for each year and port group as
Sf ¼gbt;p;y ¼ medianðsf ¼gbt;p;y;vÞ:
ð3Þ
Adaptive capacity
Adaptive capacity is a complex and multifaceted concept, defined by the Intergovernmental
Panel on Climate Change as “[t]he ability of a system to adjust to climate change (including cli-
mate variability and extremes), to moderate potential damages, to take advantage of opportu-
nities, or to cope with the consequences” ([61], p. 9). Evaluating adaptive capacity
comprehensively requires assessment of multiple domains, including assets, flexibility, organi-
zation, learning, and agency [29–31]. Here we focused on the flexibility domain as it pertains
to coping capacity, the “ability to react to and reduce the adverse effects of experienced haz-
ards” ([62], p. 72). Specifically, we quantified diversification and mobility within the
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
8 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
groundfish fleets, equating reduced diversification and mobility with reduced capacity to cope
and adapt.
Adapt in-place: Diversification. For the adapt in-place assessment, we quantified pres-
ent-day fisheries diversification within each of the port groups associated with each groundfish
fleet in terms of opportunities to participate in other fisheries from 2011–2019. For this analy-
sis, we selected a measure that invites consideration of the full cross-section of a port group
(e.g., processors, deckhands, owners, captains, etc.) that may offer resilience to a groundfish
fleet should it experience negative impacts of climate change. We did not subset to only those
vessels that participated in the bottom trawl groundfish fishery, as we wanted to reflect the
potential for future adaptation within a port group given current fishing opportunities defined
as broadly as possible.
Specifically, we generated an annual fisheries participation network [25,38] for each port
group to derive an edge density metric. In these networks, different fisheries are depicted as
nodes, while pairs of nodes are connected by lines, called edges, that integrate information
about vessels participating in both fisheries (S5 Fig; further methodological details provided in
[63]). Edge density of a network is defined as the ratio of the number of edges present to the
total possible edges in the network [64]. Higher edge density implies that fishers in these ports
have, on average, access to a greater range of alternative fishing opportunities if one node (fish-
ery) is compromised because of poor stock availability, a fishery closure, or other regulatory
actions [25,38]. Edge density scales with network size (it is easier to achieve a high density in a
low complexity network), so comparisons across networks of different sizes should be made
with the knowledge that port groups with fewer fisheries will necessarily have more diversifica-
tion potential than those with more fisheries.
We created annual fisheries participation networks using species landings data retrieved
from PacFIN’s comprehensive fish tickets table on 29 December 2021. These networks repre-
sent the most recent available data for the period 2011–2019 [63], and are summarized annu-
ally from week 46 in one year through week 45 in the following year (e.g., November 2018 to
November 2019) to capture the beginning of the Dungeness crab (Metacarcinus magister) fish-
ing season, a fishery in which many bottom trawl groundfish vessels also participate. We classi-
fied nodes based on the species groupings described by [65]. We report diversification as the
annual edge density value of each port group’s fisheries participation network.
Adapt on-the-move: Mobility. For the adapt on-the-move assessment, we characterized
each fleet’s mobility based on documented changes in the distance of fishing grounds to port
from 2011–2019. This approach assumed that fleets from port groups fishing farther from port
were more mobile, while acknowledging that many factors influence this metric (e.g., bathym-
etry, stock availability, vessel size and gear, spatial closures, substrate, etc.). We calculated
mobility mp,y,v of vessel v in year y based on its landings-weighted distance from port. For each
vessel v in year y, we calculated the straight-line distance d from the set location l of each haul
to the port of landing p, then weighted each distance calculation by the groundfish landings
associated with that haul before selecting the median value for each vessel in each year:
mp;y;v ¼ medianðdp;y;v;lÞ
We calculated annual mobility of each fleet Mp;y based on the median value of mp,y,v for
each year and port
Mp;y ¼ medianðmp;y;vÞ;
ð4Þ
ð5Þ
and report the 95th percentile of Mp;y as our annual index of mobility. This approach assumes
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
9 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
each vessel contributes equally to fleet mobility, rather than weighting mobility by each vessel’s
landings, and captures the upper limit of mobility for each fleet.
Assessment of risk due to climate change
We integrated measures of exposure, sensitivity, and adaptive capacity of the groundfish fleets
on the U.S. West Coast to evaluate coupled social-ecological risk to climate change. Our defini-
tions follow those of the IPCC [62], such that high exposure to climate change, given the haz-
ard of projected warming bottom temperatures [11], and high vulnerability, together imply
high risk. Vulnerability is defined broadly as “the propensity or predisposition to be adversely
affected” ([3], p. 5), and here we calculate it by integrating our measure of sensitivity (eco-
nomic dependence) with our measures of adaptive capacity (diversification or mobility).
p, thermal change, E*
Specifically, we calculated median exposure values based on thermal change relative to his-
toric variability, horizontal displacement, and vertical displacement across the 3 ESMs for each
fleet, and rescaled the median exposure values to index values of E*
p, horizontal
displacement, and E*
p, vertical displacement such that their minimum values were 0 and their maxima
were 1 (the maximum thermal change relative to historic variability, horizontal displacement,
and vertical displacement expected across all fleets). We calculated the average value of Sf ¼gbt;p;y
across 2011–2019 and rescaled it to create a sensitivity index S*
and a maximum value of 1, with 1 reflecting the maximum observed across all fleets. For each
of the measures of adaptive capacity, we calculated their average annual values across 2011–
2019, and rescaled the resultant quantities such that their minimum values were 0 and their
maxima were 1, with 1 reflecting the minimum diversification or mobility observed across all
fleets. This reversal of scale converted these indices into measures of a relative lack of capacity
p and relative lack of mobility M*
to cope and adapt, due to a relative lack of diversification D*
p.
We calculated vulnerability of each fleet under the adapt in-place assessment Vp, adapt n-place
and under the adapt on-the-move assessment Vp, adapt on-the-move, as the Euclidean distance to
the origin of the location represented by sensitivity S*
p values, such that
p with a minimum value of 0
p and either D*
p or M*
and
Vp;adapt in(cid:0) place ¼ ðS∗
p
2 þ D∗
p
2Þ1=2
Vp;adapt on(cid:0) the(cid:0) move ¼ ðS∗
p
2 þ M∗
p
2Þ1=2:
ð6AÞ
ð6BÞ
With this calculation, we assume vulnerability to be equally affected by sensitivity and adap-
tive capacity. Following [46] (their Fig 2, right), we represented this vulnerability to climate
change visually, and used it to distinguish between fleets of greater or lesser concern and those
that are potential adapters or have high latent risk.
Our ultimate interest was in the combined risk due to climate change of each fleet under
the adapt in-place assessment Rp, adapt in-place and under the adapt on-the-move assessment Rp,
adapt on-the-move. Specifically, we defined this integrated measure of exposure and vulnerability
as the Euclidean distance to the origin of the location associated with each value of E*p,i and
vulnerability Vp,j,
Rp;adapt in(cid:0) place ¼ ðE∗
p;thermal change
2 þ Vp;adapt in(cid:0) place
2Þ1=2:
Rp;adapt on(cid:0) the(cid:0) move ¼ ðE∗
p;vertical displacement
2 þ Vp;adapt on(cid:0) the(cid:0) move
2Þ1=2:
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
ð7AÞ
ð7BÞ
10 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
With these calculations, we assume risk to be equally affected by exposure and vulnerability,
and interpret fleet risk relative to other fleets in this analysis, rather than capturing an absolute
measure of risk.
Geographical patterns
To evaluate whether there were geographical patterns in the exposure, sensitivity, adaptive
capacity, and risk metrics, we conducted regressions of these variables against latitude. Specifi-
cally, we used the glmmTMB package to evaluate (i) the fixed effects of latitude on thermal
change relative to historic variability, horizontal displacement, or vertical displacement for
each ESM separately; (ii) the fixed effect of latitude and the random effect of year on sensitivity,
diversification, and mobility; and, (iii) the fixed effect of latitude on each of the risk metrics. In
all of the models, we weighted the regressions by the number of vessels composing each fleet.
For the sensitivity and diversification models, we used a logit link and the ordered beta family
because the data represent proportions. For the mobility model, we used a log link and the
Gaussian family to adequately capture the long tail in the distribution of landings-weighted
distance from port across fleets, and included splines (number of knots = 3). All other models
used an identity link and the Gaussian family. While the convention when plotting regressions
is to have the explanatory variable on the x-axis, we decided to plot latitude on the y-axis
because it provides a more intuitive representation of poleward and equatorward shifts in fish-
ing fleets operating off the U.S. West Coast. To evaluate the leverage of individual fleets in
these analyses, we re-ran the regressions described above using leave one out cross validation
(LOOCV; see S3 Text for details).
Results
We found that the sensitivity of groundfish fleets along the U.S. West Coast, based on their
share of earnings from the groundfish fishery, varied substantially from close to zero to near
complete dependence (Fig 3). The more equatorward San Francisco, Santa Barbara, and Los
Angeles fleets derived <10% of their revenue from the bottom trawl groundfish fishery during
2011–2019, while the more poleward fleets landing in Puget Sound, Astoria, and Fort Bragg
captured �80% of their revenue from the bottom trawl groundfish fishery (Fig 3). Overall,
though there was a fair amount of interannual variability in the relationship, sensitivity
increased significantly with latitude (p <0.001; Fig 3, S3D Table). The Santa Barbara fleet had
high leverage, but did not modify the positive relationship observed in the full data set (S15
Fig). These estimates of sensitivity based on economic dependence of groundfish fleets on bot-
tom trawl groundfish were used in both the adapt in-place and adapt on-the-move risk
assessments.
We centered our analysis of exposure to climate change within present-day fishing foot-
prints (Fig 4A) of U.S. West Coast groundfish fishing fleets. These footprints indicate extensive
fishing along the coast, particularly off Washington and Oregon (Fig 4B) where fishing
grounds overlapped considerably more and generally occupied larger areas, compared with
the fishing footprints of fleets landing catch in California-based port groups (Fig 4C and 4D).
The landings-weighted depth of the catch, while highly variable for some port groups, was gen-
erally shallower for fleets landing catch in ports south of Point Conception, California, than
those farther north (S6 Fig). In addition, these equatorward fleets tended to be composed of
smaller-size vessels (S7 Fig).
On average across the three ESMs, we estimated that between the historic (1990–2020) and
projected (2065–2095) periods, there would be one standard deviation or more of near-bottom
ocean warming within present-day fishing footprints, ~5km of horizontal displacement of
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
11 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
Fig 3. Economic dependence, as a measure of sensitivity, of U.S. West Coast groundfish fleets to changes in the
fishery, in relation to latitude. The black line represents the relationship between mean economic dependence (2011–
2019; proportion of groundfish revenue relative to revenue from all commercial fisheries) and latitude, while grey
shading indicates the SE of this relationship, which was statistically significant (p < 0.001, S3 Table). Colors
correspond to the state in which each port occurs (blue: California, yellow: Oregon, red: Washington).
https://doi.org/10.1371/journal.pclm.0000285.g003
bottom isotherms, and 10s to 100s of meters displacement of bottom isotherms into deeper
waters (vertical displacement). We also found that exposure under adapt in-place and adapt
on-the-move strategies increased significantly with latitude (S3A–S3C Table). Compared to
more equatorward fleets, we found that poleward fleets will experience twice as much local
temperature change within present-day fishing footprints (Fig 4E), relative to historic variabil-
ity, and 3–4 times as much vertical thermal displacement if they move to follow thermal pro-
files of present-day fishing footprints (Fig 4F). The Puget Sound, Astoria, Santa Barbara, and
Los Angeles fleets had high leverage in the regressions with both measures of exposure (S9–
S14 Figs), but did not modify the positive relationship observed in the full data set (except for
the IPSL-based regression of local temperature change within present-day fishing footprints,
which was highly uncertain; S11 Fig). Horizontal displacement of bottom isotherms in pres-
ent-day fishing footprints is more uncertain across the ESMs and its association with latitude
varied in sign depending on the ESM (S8 Fig). Because the sign of the association between hor-
izontal displacement and latitude varied between ESMs, we did not calculate an average hori-
zontal displacement across ESMs to include in the overall risk estimates reported below.
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
12 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
Fig 4. Fishing footprints and geographic exposure to climate change within fishing footprints. (a) Fishing footprint from 2011–2019 (dark gray regions) for
U.S. West Coast groundfish fleets. Alternating light/dark green regions on land delineate the 14 port groups, which are numbered with corresponding names
listed in inset legend. Three enlargement maps to the right show the 14 port groups landing bottom trawl-caught groundfish on land (numbered), but with
distinct, individually delineated fishing footprints (corresponding circled numbers) associated with fleets fishing off Oregon and Washington (b) and
California (c, d). Estimates of exposure of these fleets to climate change based on comparison of 30-year historic (1990–2020) and future (2065–2095) periods
for (e) bottom temperature change relative to historic variability, and (f) vertical displacement of bottom isotherms. In (e) and (f), point size scales with the
number of vessels in each fleet and these relationships were statistically significant (p < 0.001, S3 Table). GFDL, HADL and IPSL correspond to the three Earth
system models used to develop dynamically downscaled projections of bottom temperature. GEBCO 2023 (NOAA NCEI Visualization) base map (https://noaa.
maps.arcgis.com/home/item.html?id=8050bfc4eb4444758f194db95f817184). Credit: General Bathymetric Chart of the Oceans (GEBCO); NOAA National
Centers for Environmental Information (NCEI).
https://doi.org/10.1371/journal.pclm.0000285.g004
Our two measures of the adaptive capacity of the groundfish fishing fleets showed contrast-
ing changes with latitude (Fig 5). Diversification, which we used as a proxy for the potential to
adapt if fleets continue to fish where they are now (adapt in-place), declined significantly with
increasing latitude (Fig 5A, S3E Table; p<0.001). While statistically significant, the differences
in diversification between poleward and equatorward fleets due strictly to latitudinal position
were small and uncertain in absolute magnitude (S16 Fig) and unlikely to be especially impact-
ful to fleet-specific vulnerability (65–75% of potential edges were realized in most networks).
In addition, the Puget Sound and Santa Barbara fleets had high leverage (S16 Fig). In contrast,
fleets in poleward ports generally caught groundfish farther from ports of landing (~80km-
250km) compared to ports in more equatorward California (in most cases <50km). Therefore
fleet mobility (interquartile range of mobility: 40–90 km), which we use as a proxy for the
potential for fleets to adapt by moving to new fishing grounds (adapt on-the-move), increased
significantly with increasing latitude (Fig 5B, S3F Table; p<0.001). The Puget Sound fleet had
high leverage in the regression of mobility against latitude, but did not modify the positive
relationship observed in the full data set (S17 Fig).
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
13 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
Fig 5. Geographic variation in fleet fisheries diversification and fleet mobility. Relationships between the latitude of ports of landings for U.S. West Coast
groundfish fleets and two elements of the flexibility dimension of adaptive capacity: (a) diversification based on edge density of fisheries participation networks;
and (b) mobility based on landings-weighted distance from port to fishing grounds. Points indicate averages across 2011–2019, point size scales with the
number of vessels in each fleet, and these relationships were statistically significant (diversification: p = 0.015, mobility: p < 0.001, S3 Table). Colors correspond
to the state in which each port occurs (blue: California, yellow: Oregon, red: Washington).
https://doi.org/10.1371/journal.pclm.0000285.g005
Collectively, we found that the coupled social-ecological risk of poleward groundfish fishing
fleets was elevated compared to more equatorward fleets (Fig 6, S19 Fig). Sensitivity created
the greatest variation in vulnerability (y-axes in S18 Fig), which tended to be highest for fleets
landing at ports in northern California, Oregon, and Washington. Under an adapt in-place
strategy, risk was greatest for more poleward fleets because of their greater exposure and
higher sensitivity (Fig 6A). Under an adapt on-the-move strategy, the greater exposure and
sensitivity of more poleward fleets to climate change was dampened by their greater mobility,
and fleets had similar risk scores from either being more vulnerable or more exposed, but not
necessarily both more vulnerable and more exposed simultaneously (S19 Fig). Overall, latitude
had a greater effect on risk of groundfish fleets to climate change under an adapt in-place strat-
egy (compare slopes in S3G and S3H Table, risk scores in S19 Fig).
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
14 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
Fig 6. Coupled social-ecological risk due to climate change for groundfish fleets on the U.S. West Coast. (a) Assuming fleets change target species while
remaining in current fishing grounds (adapt in-place); (b) assuming fleets shift fishing grounds while targeting current species (adapt on-the-move). Larger
points and font sizes indicate fleets composed of a greater number of vessels, and these relationships were statistically significant (p < 0.001, S3 Table). Colors
correspond to the state in which each port occurs (blue: California, yellow: Oregon, red: Washington).
https://doi.org/10.1371/journal.pclm.0000285.g006
Discussion
The translation of global-to-local projected impacts of climate change can facilitate strategic
planning that helps resource-dependent communities and industries take a proactive role in
their futures. One form this translation can take is climate risk assessments that are performed
at scales relevant to individuals, communities, and decision makers [4]. Such steps increase the
reliability and relevance of information by representing important social and biophysical pro-
cesses more accurately and providing user-specific context. Focusing on the bottom trawl
groundfish fishery along the U.S. West Coast, we found that more poleward fleets face greater
risk due to climate change because of higher exposure and greater sensitivity in the form of
economic dependence on groundfish. Specifically, we showed that poleward risk was greater if
fleets rely on existing groundfish fishing grounds, which necessitates diversifying to other spe-
cies and can come at a cost (e.g., investment in additional permit and gear types), rather than
shifting fishing grounds and maintaining current catch composition. This result suggests that
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
15 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
an adapt on-the-move strategy will better mitigate risk than an adapt in-place strategy for
high-latitude fleets, assuming that the variable costs of fishing (e.g., due to changes in fuel
prices and labor wages) relative to ex-vessel revenues remain similar to the present. These gen-
eral inferences emerge from application of one indicator for each dimension of risk, which is
an oversimplification, but also offers transparency and the potential for replicability for other
fleets and regions. Our findings contrast with similar work in other parts of the world, such as
Europe, where lower-latitude fleets and fisheries are expected to face greater climate risk
[35,36,66]. While existing within-fishery flexibility on the U.S. West Coast provides some
promise for coping with, reacting to, and adapting to projected impacts of climate change [67],
our analysis highlights how further development of this and other dimensions of adaptive
capacity could enhance resilience of these fishing fleets.
Building climate resilience for fishing fleets
Parsing risk into its constituents (exposure, sensitivity, and adaptive capacity, under two con-
trasting adaptation strategies) suggests different types of interventions that can be imple-
mented to reduce risk. Communities may have similar risk scores, but contrasting sources of
risk, and therefore may respond favorably to customized interventions. Mitigating risk may
require more proactive efforts to improve adaptive capacity, such as fisheries portfolio diversi-
fication or enhancing fleet mobility, or to reduce sensitivity through expansion of revenue
streams, among other solutions [29,46,68]. For example, in California, there are existing prece-
dents for enhancing adaptive capacity for fleets with latent risk (low sensitivity and low adaptive
capacity). For instance, following the implementation of individual fishing quotas in 2011,
members of the Fort Bragg, Morro Bay, Monterey, and Santa Barbara fleets organized quota
risk pools with the support of local government and non-government organizations to navigate
bycatch constraints, thereby enhancing resilience within the new regulatory environment [69].
In contrast, the suite of interventions for fleets that are potential adapters (because they
have higher adaptive capacity and sensitivity, e.g., Fort Bragg or Astoria) are more likely to
focus on a reduction in sensitivity. Livelihood diversification (e.g., through mariculture or
tourism activities) can dampen sensitivity while also improving adaptive capacity in-place,
whereas improving access to fish for other target species and in new (or previously closed) fish-
ing grounds are more exclusively directed at reducing sensitivity [46,68]. Finally, there are
interventions that could rescale the risk landscape across all fleets, such as recent efforts to cre-
ate increased market share for groundfish [70]. Increased consumer demand for a diversity of
groundfish could increase profit margins, augment financial safety nets for fishers, and provide
an opportunity to take advantage of currently underutilized and abundant stocks. However,
creation of market demand in specific areas requires resolution of mismatches between loca-
tions of fishery landings, seafood processing, and seafood markets (e.g., through accurate map-
ping of seafood supply chains and rescuing of stranded capital; [71]). In addition, market
demand interventions may exacerbate ecological risk if they incentivize localized depletion of
stocks to meet growing local demand [68,72].
Historical contingencies in management, market, and ecological forces provide important
context for evaluating the most useful interventions, regardless of whether risk due to climate
change is higher or lower for these fleets. These forces create a geography of pre-existing vul-
nerability, akin to that documented in other regions where shrinkage and disappearance of
fishing communities has occurred [73] or where implementation of new management mea-
sures has set the stage for responses to subsequent shocks [25,74]. For the bottom trawl
groundfish fishery on the U.S. West Coast, revenue has become more concentrated within
fewer fleets over the last several decades, a trend that continued throughout the 2011–2019
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
16 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
period we focused on in this study. Furthermore, the narrower continental shelf available to
California fleets has led to smaller fishing footprints (areal extent) and a lower projected expo-
sure to expected ocean warming for equatorward groundfish fleets (Fig 4), which also tend to
be composed of smaller, less mobile vessels (Fig 5 and S7 Fig, [73]). These trends are a result of
the biogeographic context in which each fleet operates, a changed regulatory environment, his-
torical impacts to more equatorward groundfish stocks [75], and various other factors (e.g.,
geographic locations of buyers, processors, and associated infrastructure; [37,45]). As in other
fisheries (e.g., Dungeness crab; [76]), practices that level the playing field for the many smaller
vessels composing equatorward groundfish fleets may help to reduce their climate risk. In con-
trast, for more poleward groundfish fleets that have high sensitivity, it may be more effective to
employ approaches that bolster other dimensions of adaptive capacity such as organization,
e.g., via social capital building to create cooperatives [46]. Each fleet’s history complicates the
many possible paths forward, but potential futures are made less opaque with the information
we have provided here on climate risk.
Future directions for assessing climate risk in fisheries
Our approach to understanding spatial heterogeneity in climate risk for fishing fleets in gen-
eral, and on the U.S. West Coast in particular, highlights opportunities for future research. The
data and methods we used to estimate exposure, sensitivity, and adaptive capacity, and to com-
bine them into a risk index, deserve further examination. For instance, we found that estimates
of exposure based on horizontal displacement of bottom isotherms are highly uncertain (S8
Fig). This result underscores the challenge of generating expectations about future ocean con-
ditions and use, and brings into question how other environmental factors that affect species
distributions, such as dissolved oxygen [77,78] may change and interact with the behavior of
fishing fleets [79–82]. Another avenue of future research is integrating expectations for other
fisheries in the participation networks (S5 Fig, [38,63]) that are likely to experience climate
effects, which will add complexity to estimates of adaptive capacity. For example, Dungeness
crab fisheries at higher latitudes may be negatively impacted by ocean acidification effects by
the late 21st Century [51], and numerous Pacific salmon (Oncorhynchus spp.) populations
along the U.S. West Coast are highly vulnerable to climate impacts at multiple life history
stages [83]. An extension of this work could connect species distributions projected using
dynamically downscaled ESM outputs (e.g., [84–86]) to fishing footprints directly, using
expected changes in the resources themselves within customary use areas to derive estimates
of exposure. Such an approach could capture the potential for more equatorward species mov-
ing into footprints while others move out ([87–89]; but see [90]), and would also need to
address the potential for fleets to capitalize on these changes under existing regulations. There
is also the question of how best to identify fishing areas, or footprints, for estimating exposure.
Here we identified the primary fishing grounds where the majority of harvested biomass is
extracted based on vessel landings by port. Alternative approaches could use metrics such as
revenue [91], fisher days [33], or could define fishing areas specific to vessel home ports [23].
There are also alternative approaches for describing sensitivity and adaptive capacity. For
example, rather than focus solely on economic dependence on a target species relative to all
other commercial fisheries, it would be informative to quantify the economic dependence of
fleets on target species relative to all other income streams including those outside of commer-
cial fisheries. Such data are not necessarily widely available, though household survey research
in small-scale fisheries provides a template for pursuing this line of inquiry [92–94]. Addition-
ally, the sensitivity and adaptive capacity of crew on fishing vessels may be quite different than
for captains or owners. Strong social identity related to participation in particular fisheries
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
17 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
could affect fishers’ willingness or ability to adapt by shifting to new fisheries or livelihood
activities [95,96]. Ideally, future work to understand risk of fishing communities will embrace
a participatory approach in which notions of community, vulnerability, and adaptive capacity
are co-developed [97] and considered alongside perceptions of other risks beyond climate
change [98]. Approaches such as fisheries learning exchanges may have the added benefit of
building trust amongst stakeholders to allow for increases in flexibility in response to climate
change, without jeopardizing ecological sustainability [99].
While we chose to analyze fleets defined by common fishing grounds and ports of landing
as one type of community, there are other units of community analysis that are equally or
more compelling (e.g., communities-of-place defined shoreside, [100,101]; and fisher net-
works emergent as communities-of-practice [102,103]). Different rubrics for describing com-
munities may lead to greater or lesser emphasis on mobility and diversification as primary
metrics to index adaptive capacity. Being able to fish a larger portfolio of species can buffer
fishers’ revenues against change and high variability [65]–but doing so often requires owning
multiple permits, which may be cost prohibitive for many participants or difficult to manage
given current jurisdictional boundaries [104]. This insight could lead to deeper exploration of
geographic gradients in the assets dimension of adaptive capacity.
We do not know whether current levels of diversification and mobility are at an upper
bound or if there is room for further adjustment given current costs (fuel consumption, insur-
ance, etc.; [105]). Fishing new species may be constrained by fisheries regulations that are slow
to adapt to shifting species distributions [21]. Specifically, for the bottom trawl groundfish fish-
ery, some quota categories are restricted to certain geographic regions, which would be prob-
lematic if stocks move out of the designated areas [104]. Similarly, mobility may be limited for
smaller-vessel fleets and larger-vessel fleets with more diversified catch, as has been demon-
strated on the U.S. East Coast [73]. Diversification and mobility aspects of flexibility are under-
pinned by enabling conditions that intersect with other domains of adaptive capacity such as
assets (e.g., financial resources), learning (e.g., access to knowledge, adaptable skill sets), and
organization (e.g., community cohesion), all of which may vary across different community
typologies [29,30,46]. Future work to explore these issues, for example through retrospective
evaluation of community changes associated with adaptive capacity measures existing prior to
a disruptive event [25,74], would be illuminating.
Assessments of risk due to climate change can be used to communicate potential impacts to
people, regions, or sectors at local scales [5], and in so doing can provide rationale for
medium- to long-term policy decisions intended to improve resilience. This case study pro-
vides a practical implementation of the widely-used IPCC risk assessment framework at a geo-
graphic scale that is relevant to fishers, communities, and U.S. federal fisheries managers. It
achieves this appropriately-scaled outcome by integrating climatic, ecological, and socio-eco-
nomic data from a regionally large-volume, relatively profitable, lynchpin fishery. These kinds
of data are commonly available from many of the largest-volume, greatest-value fisheries glob-
ally. However, given that these data were also available for the relatively small fleets we assessed
here, this framework may be viable for smaller-scale fisheries as well, especially with creative
approaches to generating information streams (e.g., improving understanding of fishing
grounds, economic dependence on target species, and mobility via structured surveys and par-
ticipatory workshops; [97]). Similar analyses for fleets in other regions, coupled with scenario
planning efforts [106,107], can provide more comprehensive insight into the risks of climate
change for fisheries. This insight can be used to identify regions with the greatest potential to
improve resilience to climate change through government-based regional action plans, self-
determined actions, and via new legislation for fishery disaster responses (e.g., in the U.S. via
the Fishery Resource Disasters Improvement Act) [26,29].
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
18 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
The contrasts observed here among U.S. West Coast groundfish fleets have explanations
ranging from physics to market forces, and contingencies fueled by historical and present-day
regulations. They add to evidence from the U.S. that more poleward fishing fleets may be at
greater risk due to climate change [51,86], in contrast to expectations for greater equatorward
risk in other parts of the world, such as Europe [35,36,66]. While the potential for the adapt
on-the-move strategy to mitigate greater poleward risk exceeded that for the adapt in-place
strategy, our results imply that neither of these within-fisheries flexibility measures are suffi-
cient to disrupt fundamental geographic patterning of risk. Rather, alternative adaptation
approaches that build out other attributes of flexibility, including those external to commercial
fisheries, and alternative dimensions of adaptive capacity not addressed here, may prove most
fruitful for ameliorating latitudinal patterns of climate risk. For example, increased agency for
fishers to access new target species entering their fishing grounds, introduction of greater flexi-
bility to shift fishing permits quickly, and organizational support to develop new markets are
all aspects of adaptive capacity that can reduce climate risk. Evaluations of climate risk and
adaptation approaches that capture these other types of issues need not be more complex, but
instead can strive for transparency, replicability, and comparability with this one. While the
insights presented here are specific to the U.S. West Coast, they suggest that coupled social-
ecological risk assessments like this one offer a promising path forward for evaluating climate
adaptation options in other regions around the world.
Supporting information
S1 Fig. Bottom temperature change, horizontal displacement of bottom temperature, and
vertical displacement of bottom temperature projected by three dynamically downscaled
Earth System Models (GFDL, HAD, IPSL) for the period 2025–2055 and 2065–2095.
(TIFF)
S2 Fig. Schematic of thermal displacement calculation. (a) Historical (1990–2020) bottom
temperature, (b) bottom temperature change between historical and future (2065–2095) bot-
tom temperatures, and (c) future bottom temperature and thermal displacement. The thermal
displacement calculation is illustrated for an example location at 124.2˚W, 43.9˚N. At that
location the historical mean temperature was 10.1˚C and the projected bottom temperature
increase is 2.2˚C. In the future period, moving from the future temperature (12.3˚C) to the his-
torical temperature (10.1˚C) requires an offshore horizontal displacement of 25 km, with an
associated 98 m increase in bottom depth (vertical displacement). This example uses projec-
tions forced by the IPSL Earth Systems Model, assuming Amendment 28 bottom trawl fishery
closures.
(TIFF)
S3 Fig. Contextual map, indicating the landing ports and port groups for groundfish fleets
on the U.S. West Coast, as well as fishery closure areas and untrawlable habitat. Landing
ports are represented by white squares, while hatched regions show areas closed to bottom
trawl fishing and red regions show untrawlable habitat. Green shading reflects 20km inland
buffer for each of the 14 IO-PAC port groups. Left map shows fishery closures under Amend-
ment 19, from ~2003–2019, and right map shows fishery closures from 2020 to present under
Amendment 28 which were used for thermal displacement calculations. GEBCO 2023 (NOAA
NCEI Visualization) base map (https://noaa.maps.arcgis.com/home/item.html?id=
8050bfc4eb4444758f194db95f817184). Credit: General Bathymetric Chart of the Oceans
(GEBCO); NOAA National Centers for Environmental Information (NCEI).
(TIFF)
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
19 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
S4 Fig. Fishing footprints from 2011–2019 for U.S. West Coast groundfish fleets, using the
50, 75, 90, and 95 percent volume contour.
(TIFF)
S5 Fig. Example fisheries participation networks for 3 port groups on the U.S. West Coast.
Example fisheries participation networks for the Puget Sound (left), Coos Bay (middle), and
Morro Bay (right) port groups on the U.S. West Coast (2019). Each fishery is depicted as a
node, while pairs of nodes are connected by lines, called edges, that integrate information
about vessels participating in both fisheries. In these examples, Coos Bay and Morro Bay have
higher edge densities than Puget Sound, implying that fishers in these port groups have access
to a greater range of alternative fishing opportunities if one node (fishery) is compromised
because of poor stock availability, a fishery closure, or other regulatory actions.
(EPS)
S6 Fig. Groundfish fleet depths. Landings-weighted depth of fishing grounds for U.S. West
Coast groundfish fleets from 2011–2019 (median with 95% confidence interval).
(TIFF)
S7 Fig. Groundfish fleet vessel lengths. Vessel lengths for U.S. West Coast groundfish fleets
from 2011–2019 (median with 95% confidence interval).
(TIFF)
S8 Fig. Horizontal displacement of fishing footprints. Estimates of exposure of U.S. West
Coast groundfish fleets to climate change based on comparison of 30-year historic (1990–
2020) and future (2065–2095) periods for horizontal displacement of bottom isotherms. Note
that the direction of the association between horizontal displacement and latitude varied
between the three Earth System Models (GFDL, HADL, IPSL).
(TIFF)
S9 Fig. Leave one out cross validation for regression of exposure based on bottom tempera-
ture change relative to historical variability using the GFDL Earth System Model against
latitude. Points and error bars represent estimates of the coefficient of this regression (±2 SE)
with the corresponding fleet removed from the data, red line indicates the mean estimate of
the coefficient with all fleets included in the analysis. Changes in sign of the coefficient indicate
a difference in the qualitative directional relationship between exposure based on bottom tem-
perature change relative to historical variability and latitude.
(TIFF)
S10 Fig. Leave one out cross validation for regression of exposure based on bottom temper-
ature change relative to historical variability using the HADL Earth System Model against
latitude. Points and error bars represent estimates of the coefficient of this regression (±2 SE)
with the corresponding fleet removed from the data, red line indicates the mean estimate of
the coefficient with all fleets included in the analysis. Changes in sign of the coefficient indicate
a difference in the qualitative directional relationship between exposure based on bottom tem-
perature change relative to historical variability and latitude.
(TIFF)
S11 Fig. Leave one out cross validation for regression of exposure based on bottom temper-
ature change relative to historical variability using the IPSL Earth System Model against
latitude. Points and error bars represent estimates of the coefficient of this regression (±2 SE)
with the corresponding fleet removed from the data, red line indicates the mean estimate of
the coefficient with all fleets included in the analysis. Changes in sign of the coefficient indicate
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
20 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
a difference in the qualitative directional relationship between exposure based on bottom tem-
perature change relative to historical variability and latitude.
(TIFF)
S12 Fig. Leave one out cross validation for regression of exposure based on vertical dis-
placement of bottom temperature using the GFDL Earth System Model against latitude.
Points and error bars represent estimates of the coefficient of this regression (±2 SE) with the
corresponding fleet removed from the data, red line indicates the mean estimate of the coeffi-
cient with all fleets included in the analysis. Changes in sign of the coefficient indicate a differ-
ence in the qualitative directional relationship between exposure based on vertical
displacement of bottom temperature and latitude.
(TIFF)
S13 Fig. Leave one out cross validation for regression of exposure based on vertical dis-
placement of bottom temperature using the HADL Earth System Model against latitude.
Points and error bars represent estimates of the coefficient of this regression (±2 SE) with the
corresponding fleet removed from the data, red line indicates the mean estimate of the coeffi-
cient with all fleets included in the analysis. Changes in sign of the coefficient indicate a differ-
ence in the qualitative directional relationship between exposure based on vertical
displacement of bottom temperature and latitude.
(TIFF)
S14 Fig. Leave one out cross validation for regression of exposure based on vertical dis-
placement of bottom temperature using the IPSL Earth System Model against latitude.
Points and error bars represent estimates of the coefficient of this regression (±2 SE) with the
corresponding fleet removed from the data, red line indicates the mean estimate of the coeffi-
cient with all fleets included in the analysis. Changes in sign of the coefficient indicate a differ-
ence in the qualitative directional relationship between exposure based on vertical
displacement of bottom temperature and latitude.
(TIFF)
S15 Fig. Leave one out cross validation for regression of economic dependence, as a mea-
sure of sensitivity, against latitude. Points and error bars represent estimates of the coeffi-
cient of this regression (±2 SE) with the corresponding fleet removed from the data, red line
indicates the mean estimate of the coefficient with all fleets included in the analysis. Changes
in sign of the coefficient indicate a difference in the qualitative directional relationship
between economic dependence and latitude.
(TIFF)
S16 Fig. Leave one out cross validation for regression of diversification against latitude.
Points and error bars represent estimates of the coefficient of this regression (±2 SE) with the
corresponding fleet removed from the data, red line indicates the mean estimate of the coeffi-
cient with all fleets included in the analysis. Changes in sign of the coefficient indicate a differ-
ence in the qualitative directional relationship between diversification and latitude.
(TIFF)
S17 Fig. Leave one out cross validation for regression of mobility against latitude. Points
and error bars represent estimates of the coefficient of this regression (±2 SE) with the corre-
sponding fleet removed from the data, red line indicates the mean estimate of the coefficient
with all fleets included in the analysis. Changes in sign of the coefficient indicate a difference
in the qualitative directional relationship between mobility and latitude.
(TIFF)
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
21 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
S18 Fig. Social vulnerability of groundfish fleets on the U.S. West Coast. We assume that
fleets either (a) adapt in-place by changing target species while remaining in current fishing
grounds, or (b) adapt on-the-move by shifting fishing grounds while targeting current species.
Font size and color scales with projected exposure to climate change. Vertical and horizontal
lines represent median values across fleets.
(EPS)
S19 Fig. Social vulnerability of groundfish fleets on the U.S. West Coast relative to expo-
sure to climate change. Social vulnerability, defined as sensitivity relative to adaptive capacity,
in relation to exposure to climate change for U.S. West Coast groundfish fleets, under the
assumption that fleets (a) adapt in-place by changing target species while remaining in current
fishing grounds, or (b) adapt on-the-move by shifting fishing grounds while targeting current
species. Font size, point size, and Euclidean distance from the origin scales with risk, while
color corresponds to latitude.
(EPS)
S1 Table. Linkage between individual ports and IO-PAC port groups. The port groupings
were developed by the PFMC for biennial groundfish harvest specifications. Aggregating indi-
vidual ports into port groups is necessary to provide a feasible set of geographic areas for a
coastwide climate risk analysis. Analysis at the individual port-level would violate confidential-
ity requirements, because there are often fewer than three buyers in any one port.
(DOCX)
S2 Table. Percent reduction in hauls to achieve a clean dataset. Percent reduction in hauls
to achieve a clean dataset by reason for years 2011–2019, based on processing steps detailed
here: https://zenodo.org/record/7916821.
(DOCX)
S3 Table. Statistical results. Summary of statistical results of regressions of (a-c) exposure, (d)
sensitivity, (e-f) adaptive capacity, and (g-h) risk indices relative to latitude of each fleet.
(DOCX)
S1 Text. Methods related to Fig 1. Methods Related to Fig 1.
(DOCX)
S2 Text. Exposure: Spatial considerations for thermal displacement. Description of fishery
closure areas and untrawlable habitat that influenced calculations of horizontal and vertical
thermal displacement.
(DOCX)
S3 Text. Leave one out cross validation analyses for regressions.
(DOCX)
Acknowledgments
This study was supported by the David and Lucille Packard Foundation 2019–69817 and the
NOAA Integrated Ecosystem Assessment (IEA) and Climate and Fisheries Adaptation
(CAFA) Programs. The authors appreciate the data sharing and discussions with the Califor-
nia, Oregon, and Washington Departments of Fish and Wildlife and the Pacific States Marine
Fisheries Commission. This manuscript benefited from reviews by Mary Hunsicker, Kristin
Marshall, and Kayleigh Somers, as well as from inspiring discussions and presentations at the
Effects of Climate Change on the World’s Oceans Conference held in Bergen, Norway in April
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
22 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
2023. We thank Su Kim and Vicky Krikelas for designing Fig 1, all of the groundfish that
hopped into trawl nets to make this work possible, and The Clash for their entire catalog.
Author Contributions
Conceptualization: Jameal F. Samhouri, Michael Jacox, Owen R. Liu, Lyall Bellquist, Melissa
A. Haltuch, Abigail Harley, Chris J. Harvey, Isaac C. Kaplan, Karma Norman, Leif K. Ras-
muson, Rebecca L. Selden.
Data curation: Jameal F. Samhouri, Michael Jacox, Owen R. Liu, Kate Richerson, Erin Steiner,
John Wallace, Mer Pozo Buil, Amanda Phillips, Curt Whitmire, Rebecca L. Selden.
Formal analysis: Jameal F. Samhouri, Blake E. Feist, Michael Jacox, Owen R. Liu, Kate Richer-
son, Erin Steiner, John Wallace, Mer Pozo Buil, Amanda Phillips, Eric J. Ward, Curt Whit-
mire, Rebecca L. Selden.
Funding acquisition: Jameal F. Samhouri.
Investigation: Jameal F. Samhouri, Owen R. Liu, Kate Richerson, Erin Steiner, Kelly Andrews,
Lewis Barnett, Anne H. Beaudreau, Lyall Bellquist, Melissa A. Haltuch, Abigail Harley,
Chris J. Harvey, Isaac C. Kaplan, Karma Norman, Leif K. Rasmuson, Rebecca L. Selden.
Methodology: Jameal F. Samhouri, Blake E. Feist, Michael Jacox, Owen R. Liu, John Wallace,
Melissa A. Haltuch, Abigail Harley, Chris J. Harvey, Curt Whitmire, Rebecca L. Selden.
Project administration: Jameal F. Samhouri.
Resources: Jameal F. Samhouri.
Software: Jameal F. Samhouri, Owen R. Liu, Kate Richerson, Erin Steiner, John Wallace,
Amanda Phillips, Eric J. Ward, Rebecca L. Selden.
Supervision: Jameal F. Samhouri.
Validation: Jameal F. Samhouri.
Visualization: Jameal F. Samhouri, Blake E. Feist, Michael Jacox, Rebecca L. Selden.
Writing – original draft: Jameal F. Samhouri, Erin Steiner, Rebecca L. Selden.
Writing – review & editing: Jameal F. Samhouri, Blake E. Feist, Michael Jacox, Owen R. Liu,
Kate Richerson, Erin Steiner, John Wallace, Kelly Andrews, Lewis Barnett, Anne H. Beau-
dreau, Lyall Bellquist, Mer Pozo Buil, Melissa A. Haltuch, Abigail Harley, Chris J. Harvey,
Isaac C. Kaplan, Karma Norman, Amanda Phillips, Leif K. Rasmuson, Eric J. Ward, Curt
Whitmire, Rebecca L. Selden.
References
1. Chaplin-Kramer R, Sharp RP, Weil C, Bennett EM, Pascual U, Arkema KK, et al. Global modeling of
nature’s contributions to people. Science. 2019; 366: 255. https://doi.org/10.1126/science.aaw3372
PMID: 31601772
2.
3.
Johnson JA, Baldos UL, Corong E, Hertel T, Polasky S, Cervigni R, et al. Investing in nature can
improve equity and economic returns. Proc Natl Acad Sci. 2023; 120: e2220401120. https://doi.org/
10.1073/pnas.2220401120 PMID: 37364118
IPCC. Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral
Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental
Panel on Climate Change. Field CB, Barros VR, Dokken DJ, Mach KJ, Mastrandrea MD, Bilir TE,
et al., editors. Cambridge, UK and New York, NY, USA: Cambridge University Press; 2014.
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
23 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
4. Doblas-Reyes F, So¨ rensson A, Almazroui M, Dosio A, Gutowski W, Haarsma R, et al. IPCC AR6 WGI
Chapter 10: Linking global to regional climate change. 2021. pp. 1363–1512. https://doi.org/10.1017/
9781009157896.012
5. Hinkel J. “Indicators of vulnerability and adaptive capacity”: Towards a clarification of the science–pol-
icy interface. Glob Environ Change. 2011; 21: 198–208. https://doi.org/10.1016/j.gloenvcha.2010.08.
002
6. Howden SM, Soussana J-F, Tubiello FN, Chhetri N, Dunlop M, Meinke H. Adapting agriculture to cli-
mate change. Proc Natl Acad Sci. 2007; 104: 19691–19696. https://doi.org/10.1073/pnas.
0701890104 PMID: 18077402
7.
8.
9.
Thiault L, Mora C, Cinner JE, Cheung WWL, Graham NAJ, Januchowski-Hartley FA, et al. Escaping
the perfect storm of simultaneous climate change impacts on agriculture and marine fisheries. Sci
Adv. 2019; 5: eaaw9976. https://doi.org/10.1126/sciadv.aaw9976 PMID: 31807697
Turner BL, Kasperson RE, Matson PA, McCarthy JJ, Corell RW, Christensen L, et al. A framework for
vulnerability analysis in sustainability science. Proc Natl Acad Sci U S A. 2003; 100: 8074. https://doi.
org/10.1073/pnas.1231335100 PMID: 12792023
Thiault L, Jupiter S, Johnson J, Cinner J, Jarvis R, Heron S, et al. Harnessing the potential of vulnera-
bility assessments for managing social-ecological systems. Ecol Soc. 2021;26. https://doi.org/10.
5751/ES-12167-260201
10. Rantanen M, Karpechko AY, Lipponen A, Nordling K, Hyva¨rinen O, Ruosteenoja K, et al. The Arctic
has warmed nearly four times faster than the globe since 1979. Commun Earth Environ. 2022; 3: 1–
10. https://doi.org/10.1038/s43247-022-00498-3
11. Pozo Buil M, Jacox MG, Fiechter J, Alexander MA, Bograd SJ, Curchitser EN, et al. A Dynamically
Downscaled Ensemble of Future Projections for the California Current System. Front Mar Sci. 2021;8.
https://doi.org/10.3389/fmars.2021.612874
12. Holt J, Polton J, Huthnance J, Wakelin S, O’Dea E, Harle J, et al. Climate-Driven Change in the North
Atlantic and Arctic Oceans Can Greatly Reduce the Circulation of the North Sea. Geophys Res Lett.
2018; 45: 11,827–11,836. https://doi.org/10.1029/2018GL078878
13. Alexander MA, Shin S, Scott JD, Curchitser E, Stock C. The Response of the Northwest Atlantic
Ocean to Climate Change. J Clim. 2020; 33: 405–428. https://doi.org/10.1175/JCLI-D-19-0117.1
14. Echevin V, Ge´ vaudan M, Espinoza-Morribero´ n D, Tam J, Aumont O, Gutierrez D, et al. Physical and
biogeochemical impacts of RCP8.5 scenario in the Peru upwelling system. Biogeosciences. 2020; 17:
3317–3341. https://doi.org/10.5194/bg-17-3317-2020
15. Smith KE, Burrows MT, Hobday AJ, King NG, Moore PJ, Sen Gupta A, et al. Biological Impacts of
Marine Heatwaves. Annu Rev Mar Sci. 2023;15: null. https://doi.org/10.1146/annurev-marine-
032122-121437 PMID: 35977411
16.
Intergovernmental Panel on Climate Change (IPCC), editor. Impacts of 1.5˚C Global Warming on Nat-
ural and Human Systems. Global Warming of 15˚C: IPCC Special Report on Impacts of Global Warm-
ing of 15˚C above Pre-industrial Levels in Context of Strengthening Response to Climate Change,
Sustainable Development, and Efforts to Eradicate Poverty. Cambridge: Cambridge University
Press; 2022. pp. 175–312. https://doi.org/10.1017/9781009157940.005
17. Morley JW, Selden RL, Latour RJ, Fro¨ licher TL, Seagraves RJ, Pinsky ML. Projecting shifts in thermal
habitat for 686 species on the North American continental shelf. PLOS ONE. 2018; 13: e0196127.
https://doi.org/10.1371/journal.pone.0196127 PMID: 29768423
18.
Free CM, Thorson JT, Pinsky ML, Oken KL, Wiedenmann J, Jensen OP. Impacts of historical warming
on marine fisheries production. Science. 2019; 363: 979–983. https://doi.org/10.1126/science.
aau1758 PMID: 30819962
19. Selden RL, Thorson JT, Samhouri JF, Bograd SJ, Brodie S, Carroll G, et al. Coupled changes in bio-
mass and distribution drive trends in availability of fish stocks to US West Coast ports. ICES J Mar Sci.
2020; 77: 188–199. https://doi.org/10.1093/icesjms/fsz211
20.
Thiault L, Jupiter S, Johnson J, Cinner J, Jarvis R, Heron S, et al. Harnessing the potential of vulnera-
bility assessments for managing social-ecological systems. Ecol Soc. 2021;26. https://doi.org/10.
5751/ES-12167-260201
21. Pinsky ML, Fogarty M. Lagged social-ecological responses to climate and range shifts in fisheries.
Clim Change. 2012; 115: 883–891. https://doi.org/10.1007/s10584-012-0599-x
22. Barange M, Merino G, Blanchard JL, Scholtens J, Harle J, Allison EH, et al. Impacts of climate change
on marine ecosystem production in societies dependent on fisheries. Nat Clim Change. 2014; 4: 211–
216. https://doi.org/10.1038/nclimate2119
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
24 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
23. Colburn LL, Jepson M, Weng C, Seara T, Weiss J, Hare JA. Indicators of climate change and social
vulnerability in fishing dependent communities along the Eastern and Gulf Coasts of the United States.
Mar Policy. 2016; 74: 323–333. https://doi.org/10.1016/j.marpol.2016.04.030
24. Beaudreau AH, Ward EJ, Brenner RE, Shelton AO, Watson JT, Womack JC, et al. Thirty years of
change and the future of Alaskan fisheries: Shifts in fishing participation and diversification in response
to environmental, regulatory and economic pressures. Fish Fish. 2019; 20: 601–619. https://doi.org/
10.1111/faf.12364
25.
Fisher MC, Moore SK, Jardine SL, Watson JR, Samhouri JF. Climate shock effects and mediation in
fisheries. Proc Natl Acad Sci. 2021;118. https://doi.org/10.1073/pnas.2014379117 PMID: 33397723
26. Koehn LE, Nelson LK, Samhouri JF, Norman KC, Jacox MG, Cullen AC, et al. Social-ecological vul-
nerability of fishing communities to climate change: A U.S. West Coast case study. PLOS ONE. 2022;
17: e0272120. https://doi.org/10.1371/journal.pone.0272120 PMID: 35976855
27. Cline TJ, Schindler DE, Hilborn R. Fisheries portfolio diversification and turnover buffer Alaskan fishing
communities from abrupt resource and market changes. Nat Commun. 2017; 8: 14042. https://doi.org/
10.1038/ncomms14042 PMID: 28091534
28. Green KM, Selgrath JC, Frawley TH, Oestreich WK, Mansfield EJ, Urteaga J, et al. How adaptive
capacity shapes the Adapt, React, Cope response to climate impacts: insights from small-scale fisher-
ies. Clim Change. 2021; 164: 15. https://doi.org/10.1007/s10584-021-02965-w
29. Mason JG, Eurich JG, Lau JD, Battista W, Free CM, Mills KE, et al. Attributes of climate resilience in
fisheries: From theory to practice. Fish Fish. 2022; 23: 522–544. https://doi.org/10.1111/faf.12630
30. Barnes ML, Wang P, Cinner JE, Graham NAJ, Guerrero AM, Jasny L, et al. Social determinants of
adaptive and transformative responses to climate change. Nat Clim Change. 2020; 1–6. https://doi.
org/10.1038/s41558-020-0871-4
31. Cinner JE, Barnes ML. Social Dimensions of Resilience in Social-Ecological Systems. One Earth.
2019; 1: 51–56. https://doi.org/10.1016/j.oneear.2019.08.003
32.
Fulton EA. Interesting times: winners, losers, and system shifts under climate change around Austra-
lia. ICES J Mar Sci. 2011; 68: 1329–1342. https://doi.org/10.1093/icesjms/fsr032
33. Papaioannou EA, Selden RL, Olson J, McCay BJ, Pinsky ML, St. Martin K. Not All Those Who Wander
Are Lost–Responses of Fishers’ Communities to Shifts in the Distribution and Abundance of Fish.
Front Mar Sci. 2021; 8: 741. https://doi.org/10.3389/fmars.2021.669094
34. Rooper CN, Ortiz I, Hermann AJ, Laman N, Cheng W, Kearney K, et al. Predicted shifts of groundfish
distribution in the Eastern Bering Sea under climate change, with implications for fish populations and
fisheries management. ICES J Mar Sci. 2021; 78: 220–234. https://doi.org/10.1093/icesjms/fsaa215
35. Payne MR, Kudahl M, Engelhard GH, Peck MA, Pinnegar JK. Climate risk to European fisheries and
coastal communities. Proc Natl Acad Sci. 2021; 118: e2018086118. https://doi.org/10.1073/pnas.
2018086118 PMID: 34583987
36. Aragão GM, Lo´ pez-Lo´pez L, Punzo´n A, Guijarro E, Esteban A, Garcı´a E, et al. The importance of
regional differences in vulnerability to climate change for demersal fisheries. ICES J Mar Sci. 2022; 79:
506–518. https://doi.org/10.1093/icesjms/fsab134
37. Warlick A, Steiner E, Guldin M. History of the West Coast groundfish trawl fishery: Tracking socioeco-
nomic characteristics across different management policies in a multispecies fishery. Mar Policy.
2018; 93: 9–21. https://doi.org/10.1016/j.marpol.2018.03.014
38.
Fuller EC, Samhouri JF, Stoll JS, Levin SA, Watson JR. Characterizing fisheries connectivity in marine
social–ecological systems. ICES J Mar Sci. 2017; 74: 2087–2096. https://doi.org/10.1093/icesjms/
fsx128
39. Russell SM, Oostenburg MV, Vizek A. Adapting to Catch Shares: Perspectives of West Coast Ground-
fish Trawl Participants. Coast Manag. 2019; 0: 1–18. https://doi.org/10.1080/08920753.2018.1522491
40. Hilborn R, Amoroso RO, Anderson CM, Baum JK, Branch TA, Costello C, et al. Effective fisheries
management instrumental in improving fish stock status. Proc Natl Acad Sci. 2020 [cited 14 Jan 2020].
https://doi.org/10.1073/pnas.1909726116 PMID: 31932439
41. McQuaw K, Hilborn R. Why are catches in mixed fisheries well below TAC? Mar Policy. 2020; 117:
103931. https://doi.org/10.1016/j.marpol.2020.103931
42. Errend MN, Pfeiffer L, Steiner E, Guldin M, Warlick A. Economic Outcomes for Harvesters under the
West Coast Groundfish Trawl Catch Share Program: Have Goals and Objectives Been Met? Coast
Manag. 2018; 46: 564–586. https://doi.org/10.1080/08920753.2018.1522489
43. Guldin M, Anderson CM. Catch Shares and Shoreside Processors: A Costs and Earnings Exploration
into the Downstream Sector. Mar Resour Econ. 2018; 33: 289–307. https://doi.org/10.1086/698200
44. Guldin M, Warlick A, Errend MN, Pfeiffer L, Steiner E. Shorebased Processor Outcomes Under Catch
Shares. Coast Manag. 2018; 46: 587–602. https://doi.org/10.1080/08920753.2018.1522490
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
25 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
45. Speir C, Lee M-Y. Geographic Distribution of Commercial Fishing Landings and Port Consolidation
Following ITQ Implementation. J Agric Resour Econ. 2021; 46: 152–169. https://doi.org/10.22004/ag.
econ.303606
46.
Thiault L, Gelcich S, Marshall N, Marshall P, Chlous F, Claudet J. Operationalizing vulnerability for
social–ecological integration in conservation and natural resource management. Conserv Lett. 2020;
13: e12677. https://doi.org/10.1111/conl.12677
47. Kuhlicke C, Madruga de Brito M, Bartkowski B, Botzen W, Doğulu C, Han S, et al. Spinning in circles?
A systematic review on the role of theory in social vulnerability, resilience and adaptation research.
Glob Environ Change. 2023; 80: 102672. https://doi.org/10.1016/j.gloenvcha.2023.102672
48.
49.
50.
Li Y, Sun M, Kleisner KM, Mills KE, Chen Y. A global synthesis of climate vulnerability assessments on
marine fisheries: methods, scales and knowledge co-production. Glob Change Biol. n/a. https://doi.
org/10.1111/gcb.16733
Jacox MG, Alexander MA, Bograd SJ, Scott JD. Thermal displacement by marine heatwaves. Nature.
2020; 584: 82–86. https://doi.org/10.1038/s41586-020-2534-z PMID: 32760046
Leonard J, Watson P. Description of the Input-Output Model for Pacific Coast Fisheries. NOAA Tech
Memo NMFS-NWFSC-111. 2011; 81.
51. Hodgson EE, Kaplan IC, Marshall KN, Leonard J, Essington TE, Busch DS, et al. Consequences of
spatially variable ocean acidification in the California Current: Lower pH drives strongest declines in
benthic species in southern regions while greatest economic impacts occur in northern regions. Ecol
Model. 2018; 383: 106–117. https://doi.org/10.1016/j.ecolmodel.2018.05.018
52. NGDC. U.S. Coastal Relief Model—Central Pacific (Vol. 7). Boulder, CO: National Geophysical Data
Center, NOAA; 2003.
53. NGDC. U.S. Coastal Relief Model—Northwest Pacific (Vol. 8). Boulder, CO: National Geophysical
Data Center, NOAA; 2003.
54. NGDC. U.S. Coastal Relief Model—Southern California vers. 2 (1 arc-sec). National Geophysical
Data Center, NOAA; 2012. Available: https://doi.org/10.7289/V5V985ZM
55. Pebesma E, Bivand R. Classes and methods for spatial data in R. R News. 2005; 5: 9–13.
56. Calenge C. The package “adehabitat” for the R software: A tool for the analysis of space and habitat
use by animals. Ecol Model. 2006; 197: 516–519. https://doi.org/10.1016/j.ecolmodel.2006.03.017
57. Pinsky ML, Worm B, Fogarty MJ, Sarmiento JL, Levin SA. Marine Taxa Track Local Climate Velocities.
Science. 2013; 341: 1239–1242. https://doi.org/10.1126/science.1239352 PMID: 24031017
58. Burgess MG, Becker SL, Langendorf RE, Fredston A, Brooks CM. Climate change scenarios in fisher-
ies and aquatic conservation research. ICES J Mar Sci. 2023; fsad045. https://doi.org/10.1093/
icesjms/fsad045
59. Drenkard EJ, Stock C, Ross AC, Dixon KW, Adcroft A, Alexander M, et al. Next-generation regional
ocean projections for living marine resource management in a changing climate. ICES J Mar Sci.
2021; 78: 1969–1987. https://doi.org/10.1093/icesjms/fsab100
60. Pebesma E. Simple Features for R: Standardized Support for Spatial Vector Data. R J. 2018; 10: 439–
446.
61.
Levina E, Tirpak D. Adaptation to Climate Change: Key Terms. COMENVEPOCIEASLT20061 OECD
Paris Fr. 2006.
62. Cardona OD, van Aalst MK, Birkmann J, Fordham M, McGregor G, Perez R, et al. Determinants of
risk: exposure and vulnerability. In: Managing the Risks of Extreme Events and Disasters to Advance
Climate Change Adaptation. In: Field CB, Barros V, Stocker TF, Qin D, Dokken DJ, Ebi KL, et al., edi-
tors. A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change
(IPCC). Cambridge, UK, and New York, NY, USA: Cambridge University Press; 2012. pp. 65–108.
Available: https://www.ipcc.ch/report/managing-the-risks-of-extreme-events-and-disasters-to-
advance-climate-change-adaptation/determinants-of-risk-exposure-and-vulnerability/
63. Harvey CJ, Garfield T, Williams G, Tolimieri N, editors. 2021–2022 California Current Ecosystem Sta-
tus Report. Report to the Pacific Fishery Management Council. 2022. Available: https://www.pcouncil.
org/documents/2022/02/h-2-a-cciea-team-report-1-2021-2022-california-current-ecosystem-status-
report-and-appendices.pdf/
64. Wasserman S, Faust K. Social Network Analysis: Methods and Applications. Cambridge: Cambridge
University Press; 1994. https://doi.org/10.1017/CBO9780511815478
65. Kasperski S, Holland DS. Income diversification and risk for fishermen. Proc Natl Acad Sci. 2013; 110:
2076–2081. https://doi.org/10.1073/pnas.1212278110 PMID: 23341621
66. Pita I, Mouillot D, Moullec F, Shin Y-J. Contrasted patterns in climate change risk for Mediterranean
fisheries. Glob Change Biol. 2021; 27: 5920–5933. https://doi.org/10.1111/gcb.15814 PMID:
34309958
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
26 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
67. Green KM, Selgrath JC, Frawley TH, Oestreich WK, Mansfield EJ, Urteaga J, et al. How adaptive
capacity shapes the Adapt, React, Cope response to climate impacts: insights from small-scale fisher-
ies. Clim Change. 2021; 164: 15. https://doi.org/10.1007/s10584-021-02965-w
68. Cinner JE. Social-ecological traps in reef fisheries. Glob Environ Change. 2011; 21: 835–839. https://
doi.org/10.1016/j.gloenvcha.2011.04.012
69. Kauer K, Bellquist L, Gleason M, Rubinstein A, Sullivan J, Oberhoff D, et al. Reducing bycatch through
a risk pool: A case study of the U.S. West Coast groundfish fishery. Mar Policy. 2018; 96: 90–99.
https://doi.org/10.1016/j.marpol.2018.08.008
70. Hennig J. Economic outlook survey: west coast groundfish industry 2021. San Francisco, CA, USA:
Positively Groundfish; 2022.
71. Wilen JE. Stranded Capital in Fisheries: The Pacific Coast Groundfish/Whiting Case. Mar Resour
Econ. 2009; 24: 1–18. https://doi.org/10.1086/mre.24.1.42629642
72. Beckensteiner J, Boschetti F, The´ baud O. Adaptive fisheries responses may lead to climate maladap-
tation in the absence of access regulations. Npj Ocean Sustain. 2023; 2: 1–5. https://doi.org/10.1038/
s44183-023-00010-0
73. Young T, Fuller EC, Provost MM, Coleman KE, St. Martin K, McCay BJ, et al. Adaptation strategies of
coastal fishing communities as species shift poleward. Makino M, editor. ICES J Mar Sci. 2019; 76:
93–103. https://doi.org/10.1093/icesjms/fsy140
74. Speir C, Phillips A, Mamula A, Norman K. A measure of port-level resilience to shocks in commercial
fisheries. Mar Policy. 2023; 151: 105575. https://doi.org/10.1016/j.marpol.2023.105575
75. Miller RR, Field JC, Santora JA, Schroeder ID, Huff DD, Key M, et al. A Spatially Distinct History of the
Development of California Groundfish Fisheries. PLoS ONE. 2014; 9: e99758. https://doi.org/10.1371/
journal.pone.0099758 PMID: 24967973
76.
Jardine SL, Fisher MC, Moore SK, Samhouri JF. Inequality in the Economic Impacts from Climate
Shocks in Fisheries: The Case of Harmful Algal Blooms. Ecol Econ. 2020; 176: 106691. https://doi.
org/10.1016/j.ecolecon.2020.106691
77. Keller AA, Ciannelli L, Wakefield WW, Simon V, Barth JA, Pierce SD. Species-specific responses of
demersal fishes to near-bottom oxygen levels within the California Current large marine ecosystem.
Mar Ecol Prog Ser. 2017; 568: 151–173. https://doi.org/10.3354/meps12066
78. Essington TE, Anderson SC, Barnett LAK, Berger HM, Siedlecki SA, Ward EJ. Advancing statistical
models to reveal the effect of dissolved oxygen on the spatial distribution of marine taxa using thresh-
olds and a physiologically based index. Ecography. 2022; 2022: e06249. https://doi.org/10.1111/ecog.
06249
79. Branch TA, Hilborn R, Haynie AC, Fay G, Flynn L, Griffiths J, et al. Fleet dynamics and fishermen
behavior: lessons for fisheries managers. Can J Fish Aquat Sci. 2006; 63: 1647–1668. https://doi.org/
10.1139/f06-072
80.
van Putten IE, Kulmala S, The´ baud O, Dowling N, Hamon KG, Hutton T, et al. Theories and beha-
vioural drivers underlying fleet dynamics models. Fish Fish. 2012; 13: 216–235. https://doi.org/10.
1111/j.1467-2979.2011.00430.x
81. Girardin R, Hamon KG, Pinnegar J, Poos JJ, The´baud O, Tidd A, et al. Thirty years of fleet dynamics
modelling using discrete-choice models: What have we learned? Fish Fish. 2017; 18: 638–655.
https://doi.org/10.1111/faf.12194
82. Kuriyama PT, Holland DS, Barnett LAK, Branch TA, Hicks RL, Schnier KE. Catch shares drive fleet
consolidation and increased targeting but not spatial effort concentration nor changes in location
choice in a multispecies trawl fishery. Can J Fish Aquat Sci. 2019; 76: 2377–2389. https://doi.org/10.
1139/cjfas-2019-0005
83. Crozier LG, McClure MM, Beechie T, Bograd SJ, Boughton DA, Carr M, et al. Climate vulnerability
assessment for Pacific salmon and steelhead in the California Current Large Marine Ecosystem.
PLOS ONE. 2019; 14: e0217711. https://doi.org/10.1371/journal.pone.0217711 PMID: 31339895
84.
Liu Owen, Ward Eric, Anderson Sean, Andrews Kelly, Barnett Lewis, Brodie Stephanie, et al. Species
redistribution creates unequal outcomes for multispecies fisheries under projected climate change. 5
Jan 2023 [cited 8 Aug 2023]. https://doi.org/10.1126/sciadv.adg5468 PMID: 37595038
85. Smith JA, Pozo Buil M, Muhling B, Tommasi D, Brodie S, Frawley TH, et al. Projecting climate change
impacts from physics to fisheries: A view from three California Current fisheries. Prog Oceanogr.
2023; 211: 102973. https://doi.org/10.1016/j.pocean.2023.102973
86. Rogers LA, Griffin R, Young T, Fuller E, Martin KS, Pinsky ML. Shifting habitats expose fishing com-
munities to risk under climate change. Nat Clim Change. 2019; 9: 512. https://doi.org/10.1038/
s41558-019-0503-z
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
27 / 28
PLOS CLIMATE Climate risk for fishing fleets that adapt in-place or on-the-move
87. Cavole LM, Demko AM, Diner RE, Giddings A, Koester I, Pagniello CMLS, et al. Biological Impacts of
the 2013–2015 Warm-Water Anomaly in the Northeast Pacific: Winners, Losers, and the Future.
Oceanography. 2016; 29: 273–285.
88. Walker HJ, Hastings PA, Hyde JR, Lea RN, Snodgrass OE, Bellquist LF. Unusual occurrences of
fishes in the Southern California Current System during the warm water period of 2014–2018. Estuar
Coast Shelf Sci. 2020; 236: 106634. https://doi.org/10.1016/j.ecss.2020.106634
89.
Free CM, Anderson SC, Hellmers EA, Muhling BA, Navarro MO, Richerson K, et al. Impact of the
2014–2016 marine heatwave on US and Canada West Coast fisheries: Surprises and lessons from
key case studies. Fish Fish. 2023; 24: 652–674. https://doi.org/10.1111/faf.12753
90. Hastings RA, Rutterford LA, Freer JJ, Collins RA, Simpson SD, Genner MJ. Climate Change Drives
Poleward Increases and Equatorward Declines in Marine Species. Curr Biol. 2020; 30: 1572–1577.e2.
https://doi.org/10.1016/j.cub.2020.02.043 PMID: 32220327
91. Holland DS, Sutinen JG. Location Choice in New England Trawl Fisheries: Old Habits Die Hard. Land
Econ. 2000; 76: 133–149. https://doi.org/10.2307/3147262
92. Diedrich A, Benham C, Pandihau L, Sheaves M. Social capital plays a central role in transitions to
sportfishing tourism in small-scale fishing communities in Papua New Guinea. Ambio. 2019; 48: 385–
396. https://doi.org/10.1007/s13280-018-1081-4 PMID: 30066124
93. Sunderlin WD. Resource decline and adaptation through time: Fishers in San Miguel Bay, Philippines,
1980–1993. Ocean Coast Manag. 1994; 25: 217–232. https://doi.org/10.1016/0964-5691(94)90057-4
94. Etongo D, Arrisol L. Vulnerability of fishery-based livelihoods to climate variability and change in a trop-
ical island: insights from small-scale fishers in Seychelles. Discov Sustain. 2021; 2: 48. https://doi.org/
10.1007/s43621-021-00057-4 PMID: 35425911
95. Norman K, Holland D, Abbott J, Phillips A. Community-level fishery measures and individual fishers:
Comparing primary and secondary data for the U.S. West Coast. Ocean Coast Manag. 2022; 224:
106191. https://doi.org/10.1016/j.ocecoaman.2022.106191
96. Holland DS, Abbott JK, Norman KE. Fishing to live or living to fish: Job satisfaction and identity of west
coast fishermen. Ambio. 2020; 49: 628–639. https://doi.org/10.1007/s13280-019-01206-w PMID:
31161600
97. Powell F, Levine A, Ordonez-Gauger L. Fishermen’s perceptions of constraints on adaptive capacity
in the California market squid and California spiny lobster fisheries. Front Mar Sci. 2022;9. Available:
https://www.frontiersin.org/articles/10.3389/fmars.2022.1028280
98. Nelson LK, Cullen AC, Koehn LE, Harper S, Runebaum J, Bogeberg M, et al. Understanding percep-
tions of climate vulnerability to inform more effective adaptation in coastal communities. PLOS Clim.
2023; 2: e0000103. https://doi.org/10.1371/journal.pclm.0000103
99.
Thompson KR, Heyman WD, Peckham SH, Jenkins LD. Key characteristics of successful fisheries
learning exchanges. Mar Policy. 2017; 77: 205–213. https://doi.org/10.1016/j.marpol.2016.03.019
100. Sepez J, Norman K, Poole A, Tilt B. Fish Scales: Scale and Method in Social Science Research for
North Pacific and West Coast Fishing Communities. Hum Organ. 2006; 65: 280–293.
101. Clay PM, Olson J. Defining “Fishing Communities”: Vulnerability and the Magnuson-Stevens Fishery
Conservation and Management Act. Hum Ecol Rev. 2008; 15: 143–160.
102. Pahl-Wostl C. A conceptual framework for analysing adaptive capacity and multi-level learning pro-
cesses in resource governance regimes. Glob Environ Change. 2009; 19: 354–365. https://doi.org/10.
1016/j.gloenvcha.2009.06.001
103. Shephard S, List CJ, Arlinghaus R. Reviving the unique potential of recreational fishers as environ-
mental stewards of aquatic ecosystems. Fish Fish. 2023; 24: 339–351. https://doi.org/10.1111/faf.
12723
104. Holland DS, Speir C, Agar J, Crosson S, DePiper G, Kasperski S, et al. Impact of catch shares on
diversification of fishers’ income and risk. Proc Natl Acad Sci. 2017; 114: 9302–9307. https://doi.org/
10.1073/pnas.1702382114 PMID: 28808006
105. Christensen A-S, Raakjær J. Fishermen’s tactical and strategic decisions: A case study of Danish
demersal fisheries. Fish Res. 2006; 81: 258–267. https://doi.org/10.1016/j.fishres.2006.06.018
106. Planque B, Mullon C, Arneberg P, Eide A, Fromentin J-M, Heymans JJ, et al. A participatory scenario
method to explore the future of marine social-ecological systems. Fish Fish. 2019; 20: 434–451.
https://doi.org/10.1111/faf.12356
107. Star J, Rowland EL, Black ME, Enquist CAF, Garfin G, Hoffman CH, et al. Supporting adaptation deci-
sions through scenario planning: Enabling the effective use of multiple methods. Clim Risk Manag.
2016; 13: 88–94. https://doi.org/10.1016/j.crm.2016.08.001
PLOS Climate | https://doi.org/10.1371/journal.pclm.0000285 February 9, 2024
28 / 28
PLOS CLIMATE
|
Could not heal snippet
|
10.1371_journal.pone.0256592.pdf
|
Data Availability Statement: The data underlying
the results presented in this study are from the
Centers for Medicare and Medicaid Services and
CMS does not permit data sharing as per their
legally binding and standard data use agreements.
The exact data used in this study can be purchased
directly from the Centers for Medicare and
Medicaid Services (https://www.cms.gov/
Research-Statistics-Data-and-Systems/Research-
Statistics-Data-and-Systems).
|
The data underlying the results presented in this study are from the Centers for Medicare and Medicaid Services and CMS does not permit data sharing as per their legally binding and standard data use agreements. The exact data used in this study can be purchased directly from the Centers for Medicare and Medicaid Services ( https://www.cms .
|
RESEARCH ARTICLE
A comparison of prediction approaches for
identifying prodromal Parkinson disease
Mark N. WardenID
Brad A. RacetteID
1, Susan Searles Nielsen1, Alejandra Camacho-Soto1, Roman Garnett2,
1,3*
1 Department of Neurology, Washington University School of Medicine, Saint Louis, Missouri, United States
of America, 2 Department of Computer Science and Engineering, Washington University in Saint Louis, Saint
Louis, Missouri, United States of America, 3 Faculty of Health Sciences, School of Public Heath, University of
the Witwatersrand, Johannesburg, South Africa
* [email protected]
Abstract
Identifying people with Parkinson disease during the prodromal period, including via algo-
rithms in administrative claims data, is an important research and clinical priority. We sought
to improve upon an existing penalized logistic regression model, based on diagnosis and
procedure codes, by adding prescription medication data or using machine learning. Using
Medicare Part D beneficiaries age 66–90 from a population-based case-control study of inci-
dent Parkinson disease, we fit a penalized logistic regression both with and without Part D
data. We also built a predictive algorithm using a random forest classifier for comparison. In
a combined approach, we introduced the probability of Parkinson disease from the random
forest, as a predictor in the penalized regression model. We calculated the receiver operator
characteristic area under the curve (AUC) for each model. All models performed well, with
AUCs ranging from 0.824 (simplest model) to 0.835 (combined approach). We conclude
that medication data and random forests improve Parkinson disease prediction, but are not
essential.
Introduction
Parkinson disease (PD) is a progressive, neurodegenerative disorder that is diagnosed when
patients experience motor symptoms such as resting tremor, bradykinesia, rigidity, and pos-
tural instability. However, before these motor symptoms fully manifest, patients may experi-
ence a variety of non-motor symptoms, including cognitive and mood dysfunction, sleep
disorders, and varying degrees of autonomic dysfunction [1–5]. This period of disease is
termed the “prodromal period” and may provide a critical window of opportunity during
which providers could identify PD patients. In particular, earlier recognition of PD might both
facilitate the identification of disease-modifying medications, as well as their initiation, when
available. Moreover, even without such treatments yet available, earlier identification of PD is
essential. During the prodromal disease window, many PD patients experience potentially pre-
ventable fall-related morbidity, including substantial excesses of both traumatic brain injuries
[6, 7] and fractures [8, 9] relative to comparable individuals without PD.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Warden MN, Searles Nielsen S,
Camacho-Soto A, Garnett R, Racette BA (2021) A
comparison of prediction approaches for
identifying prodromal Parkinson disease. PLoS
ONE 16(8): e0256592. https://doi.org/10.1371/
journal.pone.0256592
Editor: Thippa Reddy Gadekallu, Vellore Institute of
Technology: VIT University, INDIA
Received: January 14, 2021
Accepted: August 10, 2021
Published: August 26, 2021
Copyright: © 2021 Warden et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The data underlying
the results presented in this study are from the
Centers for Medicare and Medicaid Services and
CMS does not permit data sharing as per their
legally binding and standard data use agreements.
The exact data used in this study can be purchased
directly from the Centers for Medicare and
Medicaid Services (https://www.cms.gov/
Research-Statistics-Data-and-Systems/Research-
Statistics-Data-and-Systems).
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
1 / 13
PLOS ONE Funding: BAR: Michael J. Fox Foundation grant
#10289 (https://www.michaeljfox.org/); National
Institute of Environmental Health Sciences
K24ES017765 (https://www.niehs.nih.gov/);
Department of Defense PD190057 (https://cdmrp.
army.mil/default); SSN: National Institute of
Environmental Health Sciences K01ES028295
(https://www.niehs.nih.gov/).The funders had no
role in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
Competing interests: I have read the journal’s
policy and the authors of this manuscript have the
following competing interests: Dr. Racette serves
on the National Advisory Environmental Health
Sciences Council for the National Institute for
Environmental Health Sciences (NIEHS) for which
he is reimbursed for his time. The NIEHS had no
input or influence on the content of this
manuscript.
Parkinson disease predictive algorithms
Towards these ends, researchers have begun to move beyond traditional predictive model-
ing approaches by applying machine learning methods to a wide variety of data. Several inves-
tigators have used machine learning methods to distinguish PD patients from controls, using
data obtained from both wearable and non-wearable sensors [10, 11]. While these methods
have primarily been used to distinguish newly diagnosed PD patients from controls, other
studies were able to distinguish people with potential prodromal PD symptoms, such as hypos-
mia, from controls [11, 12]. Although these people do have a greater risk of developing PD,
this group remains heterogeneous, and there is no “ideal” prodromal PD population. In con-
trast, retrospective cohort studies using predictor data from the prodromal PD time window
afford an opportunity to confirm the PD diagnosis, while providing potentially extensive vari-
ables to include in predictive models.
Medicare claims are a rich source of population-based data to predict which patients will be
diagnosed eventually with PD. We previously developed a PD predictive model using Medi-
care claims data, specifically diagnosis and procedure codes, from the five years prior to PD
diagnosis [13]. This model contained 536 diagnoses and medical procedures as predictors and
achieved an AUC of 0.857, much higher than the AUC of 0.670 achieved with known demo-
graphic and medical predictors of PD. At the optimal cut point, sensitivity was 73.5% and
specificity was 83.2%. While this least absolute shrinkage and selection operator (LASSO)
penalized regression model performed well, the addition of Medicare Part D prescription med-
ication data or the use of other analytic methods, such as machine learning methods, may have
the potential to improve model performance. The current study builds upon our previous
work by considering whether the addition of prescription medication data improves discrimi-
nation and whether a random forest classifier could perform better or help improve the origi-
nal penalized regression approach [13]. Attempting to improve the model is the logical next
step, since we recently validated our original predictive model in a population-based sample
followed forward for PD [14]. We hypothesized that inclusion of prescription medication data
would improve model performance for four reasons: 1) these medication data offer an alterna-
tive way to capture information available from diagnosis codes, which could be incomplete; 2)
medication data might provide diagnostic confirmation and evidence of disease severity; 3)
medications might serve as proxies for biologic pathways that might be predictive of PD; and
4) some medications might increase or decrease risk of PD, regardless of the indication for the
medication, and thus could be independently predictive. Random forest classifiers use a
completely different methodology than penalized regression. Therefore, we sought to deter-
mine if this innovative approach could outperform or possibly enhance the previous penalized
regression model by introducing the probability from the random forest as a predictor in the
penalized logistic regression model. We were able to demonstrate modest improvements in
model performance.
Methods
Standard protocol approvals
This study was approved by the Washington University School of Medicine Human Research
Protection Office and the Centers for Medicare and Medicaid Services.
Study participants
This was a population-based case-control study using Medicare administrative claims data.
Briefly, all participants were U.S. residents age 66–90 years old and relying solely on Medicare
in 2009. Medicare is the only nationwide health insurance coverage universally available in the
U.S., specifically among those age 65 and older. In this age group >98% of Americans
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
2 / 13
PLOS ONE Parkinson disease predictive algorithms
participate in Medicare Part A/B, which provides medical coverage. From all of these benefi-
ciaries, we identified those who met all study eligibility criteria (age 66–90, no non-Medicare
insurance coverage, and U.S. residence) for the year 2009 using the Medicare “base file.” We
then included all incident PD cases and a random sample of comparable beneficiaries as con-
trols who also had Medicare Part D (pharmacy) coverage. We determined PD case status from
complete Part A and B Medicare claims data for 2004–2009, with cases identified as having at
least one International Classification of Diseases, Ninth Revision, Clinical Modification
(ICD9) code for PD (332 or 332.0) in 2009 but no prior year, and no code for atypical parkin-
sonism or Lewy body dementia. Controls met these same study eligibility criteria, except that
they had no ICD9 code for PD, and were alive in 2009 prior to their randomly assigned refer-
ence date (comparable to the cases’ diagnosis dates). The original study included 89,790 cases
and 118,095 controls. From this original group of participants, we further restricted to the
48,295 (54%) of cases and 52,324 (44%) of controls who were also enrolled in Medicare Part D
and had at least one medication filled under this coverage in 2008–2009. After review of medi-
cations taken by the PD patients, we excluded 12,354 cases who had filled a prescription for a
medication known to cause secondary parkinsonism (aripiprazole, chlorpromazine, fluphen-
azine, haloperidol, loxapine, metoclopramide, molindone, olanzapine, paliperidone, perphe-
nazine, pimozide, prochlorperazine, promethazine, quetiapine [if > 100 mg], reserpine,
risperidone, tetrabenazine, thioridazine, thiothixene, trifluoperazine, trimethobenzamide and/
or ziprasidone) within the 6 months prior to their PD diagnosis in 2009 [15]. This left a total of
35,941 PD cases and 52,324 controls for the present work. We formally divided these partici-
pants into a 90% training dataset and 10% test dataset by stratified random sampling (by case
status), such that we had 90% cases and 90% controls in our training set for developing the
models, and 10% cases and 10% controls in our test set for assessing model performance.
Calculation of predictor variables
We calculated predictor variables, as previously [13, 16]. In total, during the development of
the original predictive model there were 26,468 valid codes (11,063 diagnoses and 15,405 pro-
cedures, including ICD9 procedure codes and Healthcare Common Procedure Coding System
[HCPCS] codes mainly comprised of Current Procedural Terminology [CPT] codes). CPT
codes are part of a formal coding system for billing that encompasses surgical and more minor
procedures that physicians perform in the office, along with some radiology and laboratory
tests, in contrast to ICD9 procedure codes used by hospitals. HCPCS codes are similar to CPT
codes but are specific to Medicare. For ICD9/procedure codes recorded for > 10 PD cases, the
median time between receiving the code and PD diagnosis was 2.41 years. This period was
nearly identical to the median time for the 536 ICD9/procedure codes selected for our original
predictive model ultimately (2.42 years), However, the median time for diagnosis codes indica-
tive of cardinal signs of PD was shorter: 1.51 years for ICD9 333.1 (tremor), 1.98 years for
ICD9 781.2 (gait abnormality), 1.09 years for ICD9 781.0 (abnormal involuntary movement),
and 1.44 years for ICD9 781.3 (lack of coordination). We calculated age and obtained sex and
race/ethnicity from the 2009 beneficiary annual summary file. Given the importance of smok-
ing on PD risk [17], we derived a probability of ever having regularly smoked for each partici-
pant using a logistic regression model built from nationwide data [13, 16]. We previously also
identified that overall use of medical care is an important predictor of PD and included this
variable in our models [13, 18].
Building upon the above data from the beneficiary annual summary file and Part A and B
claims that were available to us when we developed our original PD predictive model, we
obtained Medicare Part D prescription data from 2008–2009, i.e., in the one to two years prior
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
3 / 13
PLOS ONE Parkinson disease predictive algorithms
to PD diagnosis, for use in our predictive models. We derived prescription data from a shorter
pre-diagnosis period than for our other claims data because Part D coverage first became avail-
able in late 2006. For each medication, we identified all associated active ingredients and cre-
ated a dichotomous variable representing whether a pharmacy filled a prescription claim for a
medication containing the active ingredient at any time during this period prior to the PD
diagnosis/control reference date. There were 880 active ingredients represented in these pre-
scription claims data. We did not include 31 active ingredients that could be used to treat PD
(carbidopa-levodopa, pramipexole, ropinirole, entacapone, tolcapone, selegiline, rasagline, tri-
hexiphenidyl, benztropine) or that could cause secondary parkinsonism (22 listed above).
Model building approach
We built all models within the training set (90% stratified random sample) using R version
3.5.0. For all models, we used a two-step model building approach with the same first step for
all. In this first step, we identified diagnosis/procedure codes and active ingredients associated
with PD using multivariable logistic regression. For each code and active ingredient, we fit a
logistic regression model adjusting a priori for age (modeled as a two-part linear spline with a
knot at age 85), sex, race/ethnicity (7 categories [6 dummy variables]), probability of ever
smoking (continuous), and number of unique diagnosis codes (continuous) [18]. These con-
stitute the 11 forced demographic predictors. We then used the Bonferroni correction for mul-
tiple comparisons to select a subset of all codes and active ingredients still significantly
associated with PD to consider in the second step of the model building. This prescreening
retained 983 codes and active ingredients, after we excluded ten that effectively were sex-spe-
cific, i.e. acting as a proxy for the patient’s sex.
Starting with the preselected set of predictor variables from the first step, i.e. the 983 codes/
active ingredient variables and the 11 forced demographic variables, we proceeded to the sec-
ond step, which differed for each model. We produced three models (fit three predetermined
classifiers): two penalized logistic regression models [13] (one with and one without prescrip-
tion medications) and a random forest that considered the prescription medications.
For the penalized logistic regressions, we built the models using only the LASSO regression
using the R package glmnet [19, 20]. In our previous work, we determined that LASSO alone
(i.e., α = 1) produced the optimal model as part of the elastic net algorithm [13]. This proce-
dure selects variables and regularizes coefficients based on penalties for possible overfitting.
The method is particularly suitable for high dimensional data, using ten-fold cross validation
to determine the shrinkage parameter (λ), and improves external validity. We used the area
under the receiver operator characteristic curve (AUC) as the measure of model quality for
selecting λ.
For the random forest, we used the R packages randomForest [21] and varSelRF [22],
which is a variable selection package designed for random forests. Specifically, we used a previ-
ously developed variable selection procedure [23]. Briefly, one large random forest was trained
on the full 90% training set using all 983 predictors and 11 demographic variables. The predic-
tor importance matrix, which contained the mean, un-scaled decrease in prediction accuracy
after variable permutation, was estimated once. Then, the 20% of predictors with the lowest
importance were dropped, and a new forest was trained on this smaller subset. The process
was repeated iteratively, while always using the original importance matrix, until only two pre-
dictor variables remained, i.e., 96 times in the present work. Each smaller subset is contained
within all larger subsets, and the predictor subset that generated the lowest “out of bag” error
was used to construct the final, predetermined random forest classifier. Random forests have
several strengths compared with support vector machines that are beneficial in this
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
4 / 13
PLOS ONE Parkinson disease predictive algorithms
application, including: 1) a useful, published feature selection method comparable to the
LASSO approach [23]; 2) the ability to handle many categorical and/or irrelevant features; 3)
automatic feature relevance determination; and 4) an exceptional generalization performance
on a wide range of tasks [24]. The first three of these are critical for our data and goals with
this study. Additionally, in other machine learning applications in PD, random forests have
consistently performed well [10, 25].
After we completed both the random forest and penalized logistic regression models, we
also experimented with using both approaches (penalized regression and random forest)
simultaneously to produce a single, combined classifier. For this, we fit a penalized logistic
regression model that also used the probability of PD generated by the final random forest as a
predictor. The random forest’s probability might be able to act like a case preprocessing filter,
allowing the penalized regression to detect more complex relationships akin to the strategy of
convolution neural networks [26] and the strategy used in Amoroso et al. (2018) [27]. We
again started with the preselected set of predictor variables from the first step but included the
prediction probabilities from the final random forest classifier as a variable that could be
selected.
Finally, given how close to PD diagnosis the cardinal signs were first coded, we repeated all
analyses while utilizing predictor variables that we calculated as of the timepoint one year
prior to PD diagnosis/control reference. Specifically, we applied a one-year lag.
Assessment of model performance
We formally assessed the performance of all models in the test set (10% stratified random sam-
ple). We were able to separate the model building step from the model diagnostic step in this
way because of the size of the available data, allowing for a clean and straightforward interpre-
tation of the test set, as if it were an external dataset. We applied each of the above models
(three primary models and one combined model) to this test dataset. Then, with PD case status
in this test set as the gold standard, we used R to calculate three summary measures of model
performance [28]: the sensitivity at the cut point that correctly classified the most beneficiaries
in the test set, the specificity at that cut point, and the AUC. We also repeated these calcula-
tions at Youden’s Index [29], the point at which the sum of sensitivity and specificities is maxi-
mized, which is not data dependent. We estimated 95% confidence intervals (CIs) using
bootstrapping with 2,000 replicates within the R package pROC [30] and validated the results
using the Stata command roctab [31]. We also calculated the percent of records in the test set
classified correctly. As further validation for all models, we calculated Spearman’s rho in the
test set between the predicted probabilities of PD for each patient derived from each model.
This inter-method reliability approach does not require a true gold standard in order to
attempt to validate both methods [32]. We compared the AUCs from the penalized regression
with Part D to the one without Part D, to assess whether the inclusion of prescription medica-
tion data improved discrimination [33]. Using the same method, we also compared the AUCs
from the random forest classifier, as well as the combined model, to the penalized regression
with Part D data, to assess whether the application of machine learning improved model
performance.
Results
Characteristics of cases and controls
We observed all known associations [13] between PD and age, sex, race/ethnicity, and smok-
ing (Table 1). On average, cases were 78.8 years old, and controls were 78.1 years old. Cases
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
5 / 13
PLOS ONE Parkinson disease predictive algorithms
Table 1. Characteristics of Parkinson disease cases and controls with Medicare Part D coverage, U.S. Medicare
2009, %.
Cases
N = 35,941
Controls
N = 52,324
Age, years
Female
Race/ethnicity
66–69
70–74
75–79
80–84
85–90
White
Black
Pacific Islander/other
Asian
Hispanic
Native American
Unknown
8.1
19.5
24.2
27.3
21.0
64.7
86.3
6.0
1.2
2.9
3.1
0.3
0.1
16.7
28.3
22.3
19.2
13.4
54.0
83.7
7.8
1.6
3.4
2.9
0.4
0.1
Smoking index � mediana
Age, years, mean (SD)
Number of unique ICD9 codes, mean (SD)
41.1
78.8 (6.1)
99.7 (52.4)
51.5
78.1 (6.2)
76.3 (46.0)
a Predicted probability of ever smoking divided by the person’s total number of unique diagnosis codes.
Abbreviations: ICD9 = International Classification of Diseases, Ninth Revision, Clinical Modification; SD = standard
deviation.
https://doi.org/10.1371/journal.pone.0256592.t001
had substantially more unique ICD9 codes in the five years prior to PD diagnosis as compared
to controls up to their comparable reference date.
Characteristics of the models
In the present dataset, the initial penalized logistic regression model, without prescription
medications, selected 183 ICD9/procedure codes, in addition to the 11 forced demographic
variables for a total of 194 predictors (S1 Table). The second model, which repeated the penal-
ized logistic regression, while including the prescription medications, contained all but two of
the ICD9/procedure codes from the first model, as well as 50 additional ICD9/procedure
codes and 28 prescription medications for a total of 270 predictors (S1 Table). Insofar as the
predictors were the same in both of the penalized regression models, the respective ORs were
generally similar.
For the random forest classifier model, the optimal subset of predictors contained 272 pre-
dictors: 248 ICD9/procedure codes, 18 active ingredients, and 6 of the 11 basic demographic
variables (the two age spline variables, sex, smoking, total count of ICD9 codes, black race)
(S1 Table).
Although 121 predictors in the random forest classifier model were not selected into either
penalized regression model, there was substantial overlap between the three models in terms
of the selected predictors, with 117 predictors (111 ICD9/procedure codes and the above 6
demographic variables) appearing in all three models (Fig 1 and S1 Table). Notably, when we
reviewed the non-overlapping codes it was clear that the random forest favored common diag-
noses/procedures, including those with modest magnitudes of association with PD, whereas
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
6 / 13
PLOS ONE Parkinson disease predictive algorithms
Fig 1. Comparison of distinct and shared predictors between models for predicting Parkinson disease, U.S.
Medicare 2009.
https://doi.org/10.1371/journal.pone.0256592.g001
the penalized logistic regression favored rare diagnoses/procedures if the magnitude of the
association was relatively large or other uncommon codes. For example, the penalized regres-
sion included gout (specifically ICD9 274.9), but the random forest did not.
When we joined the penalized regression and random forest approaches into a combined
model, 232 predictors were selected (S2 Table). These predictors included 193 ICD9/proce-
dure codes and 27 prescription medications in addition to the 11 demographic variables and
the one variable that captured the predicted probability of PD from the random forest. As
expected, we observed the largest OR for the single predictor that represented the random for-
est PD prediction probability. The combined model included 10 codes not selected by any of
the three primary models (S1 and S2 Tables). However, all these codes had ORs close to one.
Model performance
When we applied each of the three primary models to the test set, the AUC was quite similar
for each of the three models (Table 2). Accordingly, the AUC was not significantly improved
either by the addition of the Part D data to the penalized regression, or by using random forest
methods instead of penalized regression. We achieved a slightly greater AUC with the com-
bined model, in which the penalized regression model with Part D predictors also included the
probability of PD for each participant produced by the random forest as a predictor. However,
the AUC was not significantly better as compared to the similar model without this predictor.
When we applied a one-year lag to the claims data, the lagged penalized logistic regression
with Part D data contained 199 ICD9/procedure codes and no medications, while the random
forest contained 155 ICD9/procedure codes and five medications. The lagged penalized regres-
sion had an AUC of 0.742 (95% CI 0.731–0.753) and the random forest had an AUC of 0.740
(95% CI 0.729–0.751).
The three primary models had similar sensitivity and specificity. At the cut point that maxi-
mized the percent of subjects classified correctly, the combined model had greater sensitivity
but slightly less specificity than the penalized regression models (Table 2). At the cut point that
maximized the sum of sensitivity and specificity (Youden’s index) [29], all models had
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
7 / 13
PLOS ONE Parkinson disease predictive algorithms
Table 2. Performance of models for predicting Parkinson disease in the test dataset.
Cut point that maximizes
percent accurately classifieda
Specificity
Sensitivity
Cut point at Youden’s indexa Overall performance
Relative performanceb
Sensitivity
Specificity
AUC(95% CI)
Penalized regression without Part
D
Penalized regression with Part D
Random forest (with Part D)
Combined model (with Part D)c
(95% CI)
65.5 (63.9–
67.1)
67.2 (65.6–
68.7)
66.3 (64.7–
67.8)
72.9 (71.5–
79.6)
(95% CI)
83.4 (82.4–
84.4)
82.6 (81.6–
83.7)
82.8 (81.8–
83.9)
79.6 (78.4–
80.7)
(95% CI)
78.0 (76.7–
79.3)
78.6 (77.2–
79.9)
76.8 (75.4–
78.1)
76.3 (74.9–
77.6)
(95% CI)
73.2 (71.9–
74.4)
73.3 (72.1–
74.6)
75.0 (73.9–
76.2)
76.3 (75.0–
77.4)
0.824 (0.815–0.832)
Reference
model
0.827 (0.818–0.836)
p = 0.61
0.826 (0.818–0.835)
0.835 (0.826–0.843)
–
–
–
Reference
model
p = 0.90
p = 0.23
a Percent sensitivity or specificity, at selected cut points: The cut point that maximizes the percent accurately classified (data dependent) and the cut point at Youden’s
index [29] (not data dependent).
b The AUC is a measure of overall model performance, and the presented p-value assesses relative performance of the specified model as compared to the stated
reference model using the method of DeLong et al. [33] to obtain the p-value. A p-value < 0.05 indicates that the two AUCs being compared are significantly different.
The first comparison tests whether there is a difference in AUC when including Part D prescription medication data in the penalized regression model. The other
comparisons test whether there is a difference in the AUCs across the different approaches in which Part D data were included.
c Random forest classifier’s case prediction probability included as a predictor in a new penalized regression model with Part D prescription medication data.
Abbreviations: AUC = area under the receiver operator characteristic curve; CI = confidence interval.
https://doi.org/10.1371/journal.pone.0256592.t002
sensitivity and specificity estimates that were fairly similar (73.2–78.6%), with the combined
model maximizing specificity. The number of records correctly classified in the test set was
very similar across all models (76.1% for the penalized regression without medications, 76.4%
for the penalized regression with medications, 76.0% for the random forest, and 76.9% for the
combined model).
Agreement between predicted probabilities
For each Medicare beneficiary in our dataset, the two penalized regressions’ probabilities were
in very close agreement, despite the second model including prescription medication data
(Spearman’s rho = 0.995). When we compared the random forest predicted probabilities to
those generated by the penalized regression methods, agreement was still high (Spearman’s
rho = 0.915 with the model without Part D data and rho = 0.912 with the model with Part D
data used as predictors). The combined model had Spearman’s rho’s of 0.96 with all three
models.
Discussion
Identification of people with PD during the prodromal period represents an urgent research
priority due to the need to implement neuroprotective therapies earlier in the neurodegenera-
tive process and to prevent disease related morbidity associated with treatable motor symp-
toms. Our recent, complementary study [14] validated the previous PD predictive model [13],
providing evidence that the model is effective and a possible strategy to identify those in the
prodromal stage of PD. The current study continues to build upon this work by assessing the
value of adding medication data from Medicare Part D to an ICD9/procedure code-based pre-
dictive model, as well as applying machine learning methods to further validate and enhance
our previous work [13, 14]. The current study suggests prescription medication data would
not improve performance of our original predictions had pharmacy data been available for all
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
8 / 13
PLOS ONE Parkinson disease predictive algorithms
of the beneficiaries in that sample, because the AUCs between the models with and without
pharmacy data were quite similar and not statistically different. However, adding a random
forest classifier might slightly improve our model, which had already performed well. Even
though the combined model did not have a statistically significantly higher AUC, such a small
gain might be difficult to detect even in this large dataset. The latter method, which uses an
independent analytic paradigm, also provided confirmation that our previous modeling
approach was well suited to developing a predictive algorithm of undiagnosed PD. In addition,
the high correlations between model predictions and the consistency of the discriminative
ability to detect PD provide evidence that our previous and current models approach the best
possible classifier given the Medicare data structure used in this study. Taken together, this fur-
ther validates our previous predictive model [13].
Interestingly, the addition of medications to the predictive model did not improve the over-
all model performance consequentially. The addition of medications resulted in a model with
27% more diagnosis/procedure codes. In fact, the addition of prescription medications com-
plicated the model without greatly improving prediction, suggesting that the diagnoses for
which the medications were used sufficiently distinguished PD cases from controls. Moreover,
generating hypotheses about the point estimate associations with PD for the medications
selected by our model may be difficult, since some medications can be used for a variety of
medical conditions which may have directionally opposite associations with PD. Nevertheless,
most medications identified in the models consistently aligned with potential pharmacological
treatment options of medical conditions shared by all models. Our penalized regression model
with Medicare Part D confirmed the recently published “protective” association for albuterol
(salbutamol) [34]. However, this might reflect the strong inverse association between tobacco
smoking and PD [35], given that carvedilol, which has the opposite pharmacologic effect on β2
adreonoreceptors, also was selected as a negative predictor, and both medications are indicated
for smoking-related conditions. The random forest did not select these or similar medications
related to smoking but alternatively selected chronic ischemic heart disease and a history of
myocardial infarction, both strongly associated with smoking. The medications positively
associated with PD that remained in the penalized regression model, beyond what was cap-
tured via the diagnosis and procedure codes, were primarily those used to treat depression
(fluoxetine, duloxetine, mirtazapine, paroxetine, sertraline, and citalopram), reflecting the
importance of the non-motor symptoms during the prodromal PD period.
There were some consistent themes to the predictors selected by the different algorithms.
Both random forest and penalized regression models highlighted the importance of key pre-
dictors of PD, such as age, sex, white vs. black race, smoking, the cardinal motor signs of PD,
and dementia/cognitive impairment. The random forest and the respective penalized logistic
regression models (with medication data) shared approximately 43% of the predictors, and
these models were comprised almost entirely of ICD9/procedure codes. All models identified
diagnosis and procedure codes which were suggestive of both motor and non-motor symp-
toms and medical conditions associated with PD. Motor signs and/or symptoms, such as
“abnormal involuntary movement”, “tremor”, “lack of coordination”, and “abnormality of
gait” were recognized by all models as important predictors of PD, as expected. Procedure
codes shared among all three models included various brain and spine imaging codes, physical
therapy, and a variety of non-specific diagnostic tests. These codes likely reflect a combination
of diagnostic workup for prodromal PD symptoms and an attempt to treat progressive motor
problems with non-pharmacological approaches. The codes indicative of non-motor symp-
toms that appeared to identify patients with a high probability of PD reflected gastrointestinal
dysfunction (constipation), dysautonomia (orthostatic hypotension, dizziness), and cognitive/
psychiatric impairments other than general anxiety (memory loss, altered mental status,
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
9 / 13
PLOS ONE Parkinson disease predictive algorithms
mental disorder, and depression). Overall, the codes that were common between the three
models demonstrate a prodromal disease state characterized by non-motor symptoms, tremor,
gait impairment, and an attempt by health care providers to treat or identify the cause of the
symptoms.
The random forest tended to select more common predictors with lower magnitude associ-
ations. In contrast, the penalized logistic model selected conditions that were uncommon but
with a known association with PD, such as gout. Similarly, in our original predictive model
using the same regression method but larger sample size, this approach also selected condi-
tions that are rare but have large magnitude associations with PD, such as REM sleep behavior
disorder. The random forest model identified a greater number of unique codes than the
penalized regression models, yet the conditions/procedures represented by these codes had
weaker associations with PD. Many variables with the highest rank in the importance matrix
included common medical conditions that may reflect the importance of health care utiliza-
tion in being diagnosed with PD [18]. Categories distinguishing the random forest model
from the penalized regression models included: 1) prescription medications commonly pre-
scribed for bowel and bladder disorders, cognitive impairment/dementia, and psychiatric dis-
orders (e.g., depression and anxiety); 2) codes indicating head and other body trauma,
previously identified comorbidities of PD [8]; and 3) codes indicating health care utilization
prior to PD diagnosis. These codes provide interesting insight into an alternative approach to
predicting PD. The distinct methodologies we used in our study clearly identify marked clini-
cal differences between prodromal PD patients and the general population.
A strength of the study is that there were approximately 133 cases and 194 controls for each
predictor considered during the model fitting process. Theoretically, the large sample size to
predictor ratio in our models caused our predictions to approach the asymptotically minimum
achievable error [36, 37] for classifying PD. For this reason, and because the penalized regres-
sion and random forest machine learning are independent analytic approaches, we also com-
bined these into one model by feeding the PD probability from the random forest into the
penalized regression model. This approach increased the AUC by approximately 1% in abso-
lute terms. Although this difference may appear small, a 1% improvement might have a mean-
ingful impact on the absolute number of individuals further screened for PD, when applying
the predictive algorithm to a large dataset. Additionally, this improvement may be relatively
substantial considering the models may already be close to the asymptotic prediction limit.
Interestingly, the combined model’s incorporation of the random forest predictions resulted
in a discrimination gain by improving its sensitivity, reinforcing the idea that the random for-
est captured slightly different information about the cases than the penalized regressions. That
is, this model gained greater discrimination by improving case identification, and did so only
at a small cost to control identification. This is reasonable because the random forest probabil-
ity acts like a PD case preprocessing filter, improving sensitivity. In practice, all of these models
have the advantage of offering users complete flexibility in their application, such that one can
balance sensitivity and specificity to customize to each situation.
Despite the many study strengths, there are several potential limitations. First, Medicare is
only a population-based health care program for individuals older than 65; therefore, applica-
tion of this predictive model to younger individuals would not be appropriate. Second, Medi-
care data are limited to medical claims data, which are filed upon delivery of medical services
or filling of prescriptions. Other datasets, such as electronic medical record systems, may have
greater data granularity that could be leveraged for even greater model performance. With that
said, electronic medical record systems present substantial data quality challenges, as well [38].
Additionally, we only had pharmacy data for the final two years of the five year period prior to
PD diagnosis, which may have limited the usefulness of these data. However, these later years
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
10 / 13
PLOS ONE Parkinson disease predictive algorithms
are likely to be predictive due to the prodromal period of PD, insofar as patient symptoms lead
to new medications being prescribed or patients discontinuing medications due to side effects.
Non-pharmacy data in these later years were quite important to our predictive model. Notably,
we found that motor signs of PD had large ORs in the penalized regressions and high impor-
tance in the random forest. Because these signs and symptoms tend to occur in the later pro-
dromal period, relatively close to PD diagnosis, application of a one-year lag did materially
reduce the AUCs for all of our models. These reductions were similar across all models, but
discrimination remained quite good. We also note that ICD9 codes in the final three months
before PD diagnosis probably were particularly influential in achieving such high AUCs in the
unlagged model. There is an increase in the number of diagnoses (ICD9 codes) assigned to
patients around the time of PD diagnosis, as patients seek out care for either their symptoms
of PD or other medical conditions. The overall number of unique ICD9 codes is an important
predictor, in part because of this phenomenon. In addition, we and others have observed a
marked spike in traumas, likely due to falls, in the three months prior to PD diagnosis [9], but
that increased risk of fractures is evident for six to seven years prior to PD diagnosis. In addi-
tion, non-motor symptoms of PD frequently precede the motor symptoms [13]. Thus, we
believe that additional lagging would have a diminished influence on AUCs. As such, predic-
tion of PD more than five years prior to diagnosis will be an important goal for future studies.
The present work provides a useful foundation for this future work by demonstrating that
these predictive models should be attempted in larger datasets, as utilized in our original pre-
dictive model of PD, rather than restricted to individuals with pharmacy coverage.
Supporting information
S1 Table. Three primary predictive models, PD predictive model, U.S. Medicare 2009.
�HCPCS codes are similar to CPT codes but are specific to Medicare; Abbreviations:
CPT = Current Procedural Terminology; HCPCS = Healthcare Common Procedure Coding
System�; ICD9 = International Classification of Diseases, Ninth Revision; PD = Parkinson dis-
ease.
(PDF)
S2 Table. Combined model, PD predictive model, U.S. Medicare 2009. �HCPCS codes are
similar to CPT codes but are specific to Medicare; Abbreviations: CPT = Current Procedural
Terminology; HCPCS = Healthcare Common Procedure Coding System�;
ICD9 = International Classification of Diseases, Ninth Revision; PD = Parkinson disease.
(PDF)
Author Contributions
Conceptualization: Susan Searles Nielsen, Roman Garnett, Brad A. Racette.
Formal analysis: Mark N. Warden, Susan Searles Nielsen, Alejandra Camacho-Soto.
Funding acquisition: Brad A. Racette.
Methodology: Roman Garnett.
Supervision: Susan Searles Nielsen, Roman Garnett, Brad A. Racette.
Validation: Mark N. Warden, Alejandra Camacho-Soto.
Writing – original draft: Mark N. Warden.
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
11 / 13
PLOS ONE Parkinson disease predictive algorithms
Writing – review & editing: Susan Searles Nielsen, Alejandra Camacho-Soto, Roman Garnett,
Brad A. Racette.
References
1. Siderowf A, Jennings D, Eberly S, Oakes D, Hawkins KA, Ascherio A, et al. Impaired olfaction and other
prodromal features in the Parkinson At-Risk Syndrome Study. Mov Disord. 2012; 27(3):406–412.
https://doi.org/10.1002/mds.24892 PMID: 22237833
2. Plouvier AO, Hameleers RJ, van den Heuvel EA, Bor HH, Olde Hartman TC, Bloem BR, et al. Prodro-
mal symptoms and early detection of Parkinson’s disease in general practice: a nested case-control
study. Fam Pract. 2014; 31(4):373–378. https://doi.org/10.1093/fampra/cmu025 PMID: 24869632
3. Goldman JG, Postuma R. Premotor and nonmotor features of Parkinson’s disease. Curr Opin Neurol.
2014; 27(4):434–441. https://doi.org/10.1097/WCO.0000000000000112 PMID: 24978368
4. Schrag A, Horsfall L, Walters K, Noyce A, Petersen I. Prediagnostic presentations of Parkinson’s dis-
ease in primary care: a case-control study. Lancet Neurol. 2015; 14(1):57–64. https://doi.org/10.1016/
S1474-4422(14)70287-X PMID: 25435387
5. Gustafsson H, Nordstrom A, Nordstrom P. Depression and subsequent risk of Parkinson disease: A
nationwide cohort study. Neurology. 2015; 84(24):2422–2429. https://doi.org/10.1212/WNL.
0000000000001684 PMID: 25995056
6. Camacho-Soto A, Warden MN, Searles Nielsen S, Salter A, Brody DL, Prather H, et al. Traumatic brain
injury in the prodromal period of Parkinson’s disease: A large epidemiological study using medicare
data. Ann Neurol. 2017; 82(5):744–754. https://doi.org/10.1002/ana.25074 PMID: 29024046
7. Kenborg L, Rugbjerg K, Lee PC, Ravnskjaer L, Christensen J, Ritz B, et al. Head injury and risk for Par-
kinson disease: results from a Danish case-control study. Neurology. 2015; 84(11):1098–1103. https://
doi.org/10.1212/WNL.0000000000001362 PMID: 25681453
8. Nystrom H, Nordstrom A, Nordstrom P. Risk of Injurious Fall and Hip Fracture up to 26 y before the
Diagnosis of Parkinson Disease: Nested Case-Control Studies in a Nationwide Cohort. PLoS Med.
2016; 13(2):e1001954. https://doi.org/10.1371/journal.pmed.1001954 PMID: 26836965
9. Camacho-Soto A, Gross A, Searles Nielsen S, Miller AN, Warden MN, Salter A, et al. Fractures in the
prodromal period of Parkinson disease. Neurology. 2020; 94(23):e2448–e2456. https://doi.org/10.
1212/WNL.0000000000009452 PMID: 32345729
10. Khoury N, Attal F, Amirat Y, Oukhellou L, Mohammed S. Data-Driven Based Approach to Aid Parkin-
son’s Disease Diagnosis. Sensors (Basel). 2019; 19(2). https://doi.org/10.3390/s19020242 PMID:
30634600
11. Cavallo F, Moschetti A, Esposito D, Maremmani C, Rovini E. Upper limb motor pre-clinical assessment
in Parkinson’s disease using machine learning. Parkinsonism Relat Disord. 2019; 63:111–116. https://
doi.org/10.1016/j.parkreldis.2019.02.028 PMID: 30826265
12. Pena-Nogales O, Ellmore TM, de Luis-Garcia R, Suescun J, Schiess MC, Giancardo L. Longitudinal
Connectomes as a Candidate Progression Marker for Prodromal Parkinson’s Disease. Front Neurosci.
2018; 12:967. https://doi.org/10.3389/fnins.2018.00967 PMID: 30686966
13. Searles Nielsen S, Warden MN, Camacho-Soto A, Willis AW, Wright BA, Racette BA. A predictive
model to identify Parkinson disease from administrative claims data. Neurology. 2017; 89(14):1448–
1456. https://doi.org/10.1212/WNL.0000000000004536 PMID: 28864676
14.
15.
Faust IM, Racette BA, Searles Nielsen S. Validation of a Parkinson Disease Predictive Model in a Popu-
lation-Based Study. Parkinsons Dis. 2020; 2020:2857608. https://doi.org/10.1155/2020/2857608
PMID: 32148753
Lopez-Sendon J, Mena MA, de Yebenes JG. Drug-induced parkinsonism. Expert Opin Drug Saf. 2013;
12(4):487–496. https://doi.org/10.1517/14740338.2013.787065 PMID: 23540800
16. Centers for Disease Control and Prevention (CDC). BRFSS 2009 Survey Data and Documentation
Atlanta: CDC; 2009 [updated December 4, 2014; accessed May 17, 2016, 2016]. Available from: http://
www.cdc.gov/brfss/annual_data/annual_2009.htm.
17. Ritz B, Ascherio A, Checkoway H, Marder KS, Nelson LM, Rocca WA, et al. Pooled analysis of tobacco
use and risk of Parkinson disease. Arch Neurol. 2007; 64(7):990–997. https://doi.org/10.1001/archneur.
64.7.990 PMID: 17620489
18. Gross A, Racette BA, Camacho-Soto A, Dube U, Searles Nielsen S. Use of medical care biases associ-
ations between Parkinson disease and other medical conditions. Neurology. 2018; 90(24):e2155–
e2165. https://doi.org/10.1212/WNL.0000000000005678 PMID: 29743207
19.
Zou H, Hastie T. Regularization and variable selection via the elastic net. Journal of the Royal Statistical
Society Series B-Statistical Methodology. 2005; 67:301–320.
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
12 / 13
PLOS ONE Parkinson disease predictive algorithms
20.
Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate
Descent. J Stat Softw. 2010; 33(1):1–22. PMID: 20808728
21.
Liaw A, Wiener M. Classification and Regression by RandomForest. R News. 2002; 2(3):18–22.
22. Diaz-Uriarte R. GeneSrF and varSelRF: a web-based tool and R package for gene selection and classi-
fication using random forest. BMC Bioinformatics. 2007; 8:328. https://doi.org/10.1186/1471-2105-8-
328 PMID: 17767709
23. Diaz-Uriarte R, Alvarez de Andres S. Gene selection and classification of microarray data using random
forest. BMC Bioinformatics. 2006; 7:3. https://doi.org/10.1186/1471-2105-7-3 PMID: 16398926
24.
Fernandez-Delgado M, Cernadas E, Barro S, Amorim D. Do we Need Hundreds of Classifiers to Solve
Real World Classification Problems? J Machine Learning Research. 2014; 15:3133–3181.
25. Rovini E, Maremmani C, Moschetti A, Esposito D, Cavallo F. Comparative Motor Pre-clinical Assess-
ment in Parkinson’s Disease Using Supervised Machine Learning Approaches. Ann Biomed Eng. 2018;
46(12):2057–2068. https://doi.org/10.1007/s10439-018-2104-9 PMID: 30030773
26.
LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015; 521(7553):436–444. https://doi.org/10.
1038/nature14539 PMID: 26017442
27. Amoroso N, La Rocca M, Monaco A, Bellotti R, Tangaro S. Complex networks reveal early MRI markers
of Parkinson’s disease. Med Image Anal. 2018; 48:12–24. https://doi.org/10.1016/j.media.2018.05.004
PMID: 29807313
28. Steyerberg EW, Vickers AJ, Cook NR, Gerds T, Gonen M, Obuchowski N, et al. Assessing the perfor-
mance of prediction models: a framework for traditional and novel measures. Epidemiology. 2010; 21
(1):128–138. https://doi.org/10.1097/EDE.0b013e3181c30fb2 PMID: 20010215
29. Youden WJ. Index for rating diagnostic tests. Cancer. 1950; 3(1):32–35. https://doi.org/10.1002/1097-
0142(1950)3:1<32::aid-cncr2820030106>3.0.co;2-3 PMID: 15405679
30. Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC, et al. pROC: an open-source package
for R and S+ to analyze and compare ROC curves. BMC Bioinformatics. 2011; 12:77. https://doi.org/10.
1186/1471-2105-12-77 PMID: 21414208
31. StataCorp. Stata MC 14.2. MC 14.2 ed. College Station, TX: StataCorp LP; 2015.
32. White E, Armstrong BK, Saracci R. Principles of Exposure Measurement in Epidemiology: Collecting,
Evaluating, and Improving Measures of Disease Risk Factors. Oxford: Oxford University Press; 2008.
33. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated
receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988; 44(3):837–845.
PMID: 3203132
34. Mittal S, Bjornevik K, Im DS, Flierl A, Dong X, Locascio JJ, et al. beta2-Adrenoreceptor is a regulator of
the alpha-synuclein gene driving risk of Parkinson’s disease. Science. 2017; 357(6354):891–898.
https://doi.org/10.1126/science.aaf3934 PMID: 28860381
35. Searles Nielsen S, Gross A, Camacho-Soto A, Willis AW, Racette BA. beta2-adrenoreceptor medica-
tions and risk of Parkinson disease. Ann Neurol. 2018; 84(5):683–693. https://doi.org/10.1002/ana.
25341 PMID: 30225948
36.
37.
Tarca AL, Draghici S, Romero R. Developing classifiers for the detection of cancer using multi-analytes.
Methods Mol Biol. 2009; 520:259–272. https://doi.org/10.1007/978-1-60327-811-9_19 PMID:
19381961
Figueroa RL, Zeng-Treitler Q, Kandula S, Ngo LH. Predicting sample size required for classification per-
formance. BMC Med Inform Decis Mak. 2012; 12:8. https://doi.org/10.1186/1472-6947-12-8 PMID:
22336388
38. Harrington L. Copy-Forward in Electronic Health Records: Lipstick on a Pig. Jt Comm J Qual Patient
Saf. 2017; 43(8):371–374. https://doi.org/10.1016/j.jcjq.2017.04.007 PMID: 28738981
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
13 / 13
PLOS ONE
| null |
10.1038/s41467-022-34431-1
|
Data availability
The data that support this study are available from the corresponding
author upon reasonable request. The crystal structure described in this
study has been deposited in the Protein Data Bank under the accession
number 7TD5. The LC-MS/MS data files have been deposited to the Pro-
teomeXchange Consortium (http://proteomecentral.proteomexchange.
org) via the MassIVE partner repository with the dataset identifier
MSV000088683. Source Data are provided with this paper.
|
Data availability The data that support this study are available from the corresponding author upon reasonable request. The crystal structure described in this study has been deposited in the Protein Data Bank under the accession number 7TD5. The LC-MS/MS data files have been deposited to the Pro-teomeXchange Consortium ( http://proteomecentral.proteomexchange. org ) via the MassIVE partner repository with the dataset identifier MSV000088683. Source Data are provided with this paper.
|
Article
https://doi.org/10.1038/s41467-022-34431-1
CK2-mediated phosphorylation of SUZ12
promotes PRC2 function by stabilizing
enzyme active site
Received: 4 February 2022
Accepted: 25 October 2022
Lihu Gong 1,4, Xiuli Liu1,4, Lianying Jiao1,3,4, Xin Yang1, Andrew Lemoff
Xin Liu 1
2 &
Check for updates
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
Polycomb repressive complex 2 (PRC2) plays a key role in maintaining cell
identity during differentiation. Methyltransferase activity of PRC2 on histone
H3 lysine 27 is regulated by diverse cellular mechanisms, including post-
translational modification. Here, we report a unique phosphorylation-
dependent mechanism stimulating PRC2 enzymatic activity. Residue S583 of
SUZ12 is phosphorylated by casein kinase 2 (CK2) in cells. A crystal structure
captures phosphorylation in action: the flexible phosphorylation-dependent
stimulation loop harboring S583 becomes engaged with the catalytic SET
domain through a phosphoserine-centered interaction network, stabilizing
the enzyme active site and in particular S-adenosyl-methionine (SAM)-binding
pocket. CK2-mediated S583 phosphorylation promotes catalysis by enhancing
PRC2 binding to SAM and nucleosomal substrates and facilitates reporter gene
repression. Loss of S583 phosphorylation impedes PRC2 recruitment and
H3K27me3 deposition in pluripotent mESCs and compromises the ability of
PRC2 to maintain differentiated cell identity.
Polycomb repressive complex 2 (PRC2) is a key epigenetic enzyme
complex involved in the maintenance of cell identity during stem cell
differentiation1,2. PRC2 catalyzes methylation of histone H3 lysine 27
(H3K27); trimethylated H3K27 (H3K27me3) is a hallmark of gene
silencing3–6. PRC2 plays roles in both oncogenesis and tumor sup-
pression in a cell context-dependent manner by, for example, con-
ferring transcriptional repression of cell cycle checkpoint genes and
proliferation genes, respectively2,7. The PRC2 core complex consists of
four subunits: EZH2 (or its paralog EZH1) serves as the catalytic sub-
unit; other core subunits include EED, SUZ12, and RBBP4 (or its paralog
RBBP7). EZH2, EED, and the C-terminal VEFS (VRN2, EMF2, FIS2, and
SU(Z)12) domain of SUZ12 (SUZ12(VEFS)) assemble into the minimally
active catalytic module8,9, whereas RBBP4 and the N-terminal region of
SUZ12 are together folded into the accessory subunit-binding module,
which associates with a series of developmentally regulated accessory
subunits in PRC2 holo complexes, modulating chromatin binding10–13.
A focal point of the cellular regulation of PRC2 function is
methyltransferase activity. The PRC2 core complex displays limited
basal activity. The existing H3K27me3 histone mark engages with the
aromatic cage of EED and allosterically stimulates PRC2 enzymatic
activity8,9,14. PRC2 stimulation by H3K27me3 is thought to at least in
part account
the spreading of H3K27me3 on repressive
chromatin14. For genomic loci devoid of H3K27me3, JARID2 with tri-
methylated lysine 116 (JARID2K116me3) can initiate H3K27me3
deposition by activating PRC2 through a similar allosteric
mechanism15. Local chromatin compaction accompanied by a distinct
linker DNA length represents another cellular process leading to PRC2
activation, although the underlying molecular basis is not completely
for
1Cecil H. and Ida Green Center for Reproductive Biology Sciences, University of Texas Southwestern Medical Center, Dallas, TX 75390, USA. 2Department of
Biochemistry, University of Texas Southwestern Medical Center, Dallas, TX 75390, USA. 3Present address: Department of Biochemistry and Molecular Biology,
School of Basic Medical Sciences, Xi’an Jiaotong University Health Science Center, Xi’an, Shaanxi 710061, China. 4These authors contributed equally: Lihu
Gong, Xiuli Liu, Lianying Jiao.
e-mail: [email protected]
Nature Communications |
(2022) 13:6781
1
Article
https://doi.org/10.1038/s41467-022-34431-1
understood16. In comparison, Y641F/N/S/H and A677G cancer muta-
tions of EZH2 found in human B-cell
lymphomas cause hyper-
trimethylation of H3K27 in a heterozygous genetic background by
directly remodeling the active site and changing product specificity17,18.
PRC2 enzymatic activity is also subjected to inhibition by distinct
cellular mechanisms. As a notable example, oncogenic H3K27M
mutant histone identified in diffuse midline gliomas globally dimin-
ishes H3K27me3 level19, by blocking the histone substrate-binding
channel of PRC2 in a SAM-dependent manner9,20,21. Interestingly, EZHIP
expressed normally in gonads—and abnormally in posterior fossa
ependymoma—restricts PRC2 activity with a protein sequence
mimicking H3K27M22–25.
In addition to the methyltransferase activity, the establishment of
cell type-specific H3K27me3 patterns depends on accurate chromatin
targeting of PRC2. There are two classes of PRC2 holo complexes,
PRC2.1 and PRC2.2, in mammalian cells, which colocalize at many
target sites in mouse embryonic stem cells (mESCs). PRC2.1 and
PRC2.2 are defined based on the types of accessory subunits bound to
the core complex: PHF1/MTF2/PHF19 (a.k.a. PCL1/2/3), EPOP, and
PALI1/PALI2 are components of PRC2.1, whereas AEBP2 and JARID2
belong to PRC2.226–28. Combined genetic ablation of the accessory
subunits from both holo complexes obliterates chromatin enrichment
of PRC2 and results in a dispersed H3K27me3 pattern throughout the
genome29,30. In human-induced pluripotent stem cells (iPSCs), PRC2.1
and PRC2.2 compete for overlapping target sites; these holo com-
plexes correlate with disparate H3K27me3 levels and varying degrees
of gene repression, possibly due to differences in chromatin binding
affinity31.
PRC2 subunits undergo extensive posttranslational modification
(PTM), such as reversible phosphorylation, which couples cell signal-
ing to PRC2-mediated epigenetic gene silencing32. For example,
phosphorylation of residue S21 of EZH2 by AKT kinase hampers H3K27
methylation and causes derepression of developmental genes in sev-
eral cancer cell lines33. Cyclin-dependent kinase 1 (CDK1) phosphor-
ylates residue T345 of EZH2, promoting PRC2 recruitment and
H3K27me3 deposition at target loci34,35. AMP-activated protein kinase
(AMPK) is responsible for phosphorylation of residue T311 of EZH2
upon energy deprivation, which suppresses H3K27 trimethylation and
inhibits tumor cell growth36. Much less is known about posttransla-
tional modification of SUZ12, except that phosphorylation of residues
S539, S541, and S546 by polo-like kinase 1 (PLK1) has been found to
facilitate proteasomal degradation of PRC2 in liver tumors37.
CK2 is a conserved, ubiquitously expressed protein kinase, which
displays broad substrate specificity38,39. Active CK2 in mammalian cells
adopts a 2:2 tetrameric structure, containing two catalytic subunits,
CK2α/CK2α′, and two regulatory subunits, CK2β38,39. CK2 is a compo-
nent of two variant Polycomb repressive complex 1 (PRC1), PRC1.3 and
PRC1.528,40. CK2 inhibits monoubiquitination of histone H2A lysine 119
(H2AK119) mediated by PRC1.541. Notably, monoubiquitinated
H2AK119 (H2AK119ub) has recently been shown to play a direct role in
the chromatin recruitment of PRC242–48. CK2 expression and activity
positively correlate with proliferation and survival of cancer cells, and
host cell CK2 is exploited by several viruses, including COVID-19, to
promote viral life cycle38,39,49,50; inhibition of CK2 enzymatic activity by
chemical compounds is being tested in clinical trials for the treatment
of coronavirus disease caused by COVID-19 and of various cancer
types, including cholangiocarcinoma, basal cell carcinoma (BCC), and
recurrent medulloblastoma (clinicaltrials.gov).
Although the catalytic mechanism of PRC2 in both basal and
H3K27me3-stimulated states has been subjected to extensive bio-
chemical and structural studies8,9, our understanding of how the
enzyme may be regulated in cells remains far from complete. Here, we
report a unique phosphorylation-dependent mechanism that pro-
motes PRC2 function in cells. CK2 mediates SUZ12 phosphorylation at
a serine residue located in the SUZ12(VEFS) domain. A crystal structure
captures the phosphorylated SUZ12 in action: it induces structural
remodeling of an otherwise flexible acidic loop region in the SUZ12(-
VEFS) domain, establishing a set of molecular interactions with the
catalytic SET [Su(var)3–9, Enhancer-of-zeste and Trithorax] domain to
stabilize the enzyme active site and in particular SAM-binding pocket.
SUZ12 phosphorylation increases PRC2 enzymatic activity, enhances
PRC2 binding to nucleosomes, and promotes reporter gene repres-
sion. Loss of this phosphorylation in mESCs not only reduces PRC2
enrichment and H3K27me3 deposition, but also impairs the ability of
PRC2 to maintain a differentiated state of mESCs.
Results
Residue S583 of human SUZ12 is phosphorylated in vivo
Phosphorylation of residue S583 of human SUZ12 (SUZ12S583) and its
mouse equivalent mSUZ12S585 has been previously noted in several
untargeted phosphoproteomics studies (Fig. 1a)51. To confirm this
phosphorylation in a targeted low throughput assay, we purified
endogenous PRC2 from mESCs using an anti-SUZ12 affinity column
and carried out PTM analysis using liquid chromatography-tandem
mass spectrometry (LC-MS/MS). MS/MS spectra clearly indicated the
presence of phosphorylated mSUZ12S585 (mSUZ12S585p) in vivo
(Supplementary Fig. 1). Semi-quantitative assessment based on the
abundance of the peptides with or without the PTM indicated that the
majority of mSUZ12 is phosphorylated at this site in mESCs (Fig. 1b).
To characterize SUZ12S583 phosphorylation in human cell lines,
we raised a rabbit polyclonal anti-SUZ12S583p antibody using a syn-
thetic peptide encompassing SUZ12S583p. The purified antibody dis-
played at least 32-fold discrimination between phospho- and apo
peptides (Fig. 1c). In addition, phospho- but not apo peptide blocked
antibody binding to phosphorylated SUZ12 from HEK293T nuclear
extracts (Supplementary Fig. 2), indicative of phospho-specific recog-
nition of SUZ12 by the antibody. In line with this result, antibody signals
were greatly diminished by either treatment of a five-member PRC2
holo complex (PRC2-5m), EZH2–EED–SUZ12–RBBP4–AEBP2, with λ
protein phosphatase or introduction of an S583A mutation on SUZ12 in
the same complex (Fig. 1d).
Using the developed antibody, we examined SUZ12S583 phos-
phorylation in various cancer cell lines in a semi-quantitative manner.
We found that SUZ12S583 phosphorylation is a widespread phenom-
enon (Fig. 1e and Supplementary Fig. 3). Compared to the total cellular
line-
SUZ12, the SUZ12S583 phosphorylation level displayed cell
specific variations, with some cell lines showing distinctly less phos-
phorylation (Fig. 1e and Supplementary Fig. 3), which may be
accounted for by different kinase activities accessible to SUZ12 in
these cells.
CK2 mediates phosphorylation of residue S583 of SUZ12
To identify the kinase responsible for SUZ12S583 phosphorylation in
cells, we first performed a search with two web servers, NetPhos 3.1 and
PhosphoNET52 (www.phosphonet.ca), both of which predicted protein
kinase CK2 as the top candidate (Fig. 2a). Manual inspection also
indicated the existence of a potential CK2 substrate motif based on a
compilation of known phosphorylation motifs (Fig. 1a)53. To experi-
mentally validate the prediction, we purified two versions of CK2, α2β2
and α′2β2, and carried out in vitro phosphorylation assay on bacte-
rially expressed SUZ12. CK2-α2β2 and CK2-α′2β2 were able to
phosphorylate GST-tagged SUZ12 equally well, as indicated by an anti-
phosphoserine antibody recognizing all phosphorylated serine resi-
dues (Fig. 2b). An S583A mutation nearly abolished phosphorylation,
whereas alanine mutation of two other nearby serine residues, S546
and S604, only moderately reduced phosphorylation (Fig. 2b), sug-
gesting S583 is the primary target of CK2 kinase activity on SUZ12.
To study CK2-mediated SUZ12 phosphorylation in vivo, we used
shRNAs to knock down the CK2 subunit α, α′, or β in an embryonic
carcinoma cell line NT2/D1. Knockdown efficiency of two different
Nature Communications |
(2022) 13:6781
2
Article
https://doi.org/10.1038/s41467-022-34431-1
S583 phosphorylation stabilizes enzyme active site
Residue S583 is located in the VEFS domain of SUZ12, which associates
with the SET domain of EZH2 and is essential for the enzymatic
activity8,9. The highly conserved acidic sequence surrounding S583 on
SUZ12 was previously implicated in the stimulation of PRC2 enzymatic
activity (Fig. 3a)16,54,55. However, this acidic loop region is not well
defined in the known structures of 2.6–3.0 Å resolution in the absence
of S583 phosphorylation (Supplementary Fig. 4a, b)9,56, making it dif-
ficult to predict the impact of S583 phosphorylation on PRC2 function.
In search for constructs suitable for structural studies, we over-
expressed a truncated minimally active EZH1-containing PRC2,
EZH1–EED–SUZ12(VEFS),
in Saccharomyces cerevisiae for crystal-
lization. Unexpectedly, we found the majority of SUZ12 from the
purified complex is phosphorylated at residue S583 according to the
mass spectrometry result (Supplementary Fig. 5), likely by endogenous
yeast CK2. In addition, human CK2 was able to specifically phosphor-
ylate S583 within the truncated PRC2-EZH1 minimal complex pre-
treated by λ protein phosphatase, confirming the CK2 kinase
specificity in this context (Supplementary Fig. 6).
We determined the 3.0 Å crystal structure of this minimal com-
plex, which successfully captures residue S583 in the phosphorylated
state (Fig. 3b, Supplementary Fig. 4c and Supplementary Table 1).
Upon phosphorylation, the flexible loop harboring S583 and neigh-
boring acidic residues dramatically change conformation, becoming
engaged with the SET domain of EZH1 (Fig. 3b and Supplementary
Movie 1). In parallel, phosphoserine induces self-packing of the
N-terminal portion of the SUZ12(VEFS), which contacts EED and the
SET domain simultaneously (Fig. 3b and Supplementary Movie 1). EZH1
and EZH2 share a nearly identical SET domain (Supplementary Fig. 7),
and therefore structural analysis on the EZH1-containing PRC2 here
likely applies to the equivalent EZH2-containing complex.
The core of the SUZ12 loop undergoing phosphorylation-induced
conformational change is a motif of three acidic residues, D582-S583p-
E584, which makes extensive interactions with one lysine residue,
K684, protruding from the SET domain: both the phosphate group of
S583p and the carboxyl group of D582 side chain form hydrogen
bonds with the amino group of K684 side chain, whereas the carboxyl
group of E584 side chain mediates an additional hydrogen bonding
interaction with the main chain amine of K684 (Fig. 3c, d). Residues
H567 and S568 of the VEFS domain of SUZ12 also contact the phos-
phate group (Fig. 3c, d). Other residues helping shape the local con-
formation include K612 and L615 of the SET domain and E586, D588,
and R593 of the VEFS domain (Fig. 3d). Notably, residue K684 of the
SET domain belongs to a single turn helix partially lining the SAM-
binding pocket at the enzyme active site (Fig. 3c). We predicted that
the interaction network around K684 organized by the phosphoserine
may enhance PRC2 enzymatic activity by stabilizing the SET domain
and facilitating SAM binding. Accordingly, the S583-containing reg-
ulatory loop of SUZ12 is hereinafter referred to as the phosphorylation-
dependent stimulation (PDS) loop (Figs. 1a and 3b).
S583 phosphorylation enhances enzymatic activity and nucleo-
some binding of PRC2 in vitro
Mutations of EZH2 residue K683 (the equivalent of EZH1 residue K684)
and SUZ12 residues H567, S568, D582, and S583 were all found in
cancer cells, including established cancer cell lines and patient samples
(Supplementary Fig. 8)57 (cancer.sanger.ac.uk/cosmic), suggesting the
molecular interactions mediated by these residues may help maintain
normal PRC2 function (Fig. 3c, d). In consistence, minimal complexes
containing a K684A single mutation on EZH1 or an H567A/S568A
double mutation on SUZ12 displayed exceedingly reduced methyl-
transferase activities towards mononucleosome substrates (Fig. 4a),
likely due to disruption of the phosphoserine-centered interactions.
Similar results were obtained for the same set of mutations in the
context of the EZH2-containing minimal PRC2 complex (Fig. 4b).
Fig. 1 | Residue S583 of human SUZ12 is phosphorylated. a Domain structure of
SUZ12. Structurally characterized SUZ12 domains are represented by gray blocks
except that the VEFS domain included in the current study is colored in green. The
PDS loop harboring S583 is highlighted in orange with the amino acid sequence
shown above. b Peptides identified for SUZ12 by LC-MS/MS which contain S583.
Peptide sequence, modifications, number of peptide spectrum matches (PSMs),
and peptide abundance are listed. The phosphorylated residue that was unam-
biguously assigned is shown in orange. The percentage of phosphorylation was
calculated based on a comparison of the abundances of the phosphorylated and
unphosphorylated peptides. c Dot blot. Apo and phosphorylated peptides were
applied on a nitrocellulose membrane with a serial dilution. Phospho-specific
reactivity of the developed anti-SUZ12S583p antibody was analyzed. A repre-
sentative of three independent experiments is shown. d Effect of λ phosphatase
treatment and serine mutation on ectopically expressed PRC2-5m,
EZH2–EED–SUZ12–RBBP4–AEBP2. The total amount of PRC2-5m is indicated by
anti-SUZ12 signals. S583 phosphorylation level is indicated by signals of the anti-
SUZ12S583p antibody developed in this study (uncropped gel images of this figure
are shown in Supplementary Fig. 16). A representative of three independent
experiments is shown. e Levels of S583 phosphorylation in stem cells and cancer
cells. Anti-SUZ12S583p signals were generated using immunoprecipitates of anti-
SUZ12 antibody to avoid a non-relevant contaminating band (also see Supple-
mentary Fig. 2). A representative of three independent experiments is shown.
Source data are provided as a Source Data file.
shRNAs in each case was confirmed by respective antibodies (Fig. 2c).
The SUZ12S583 phosphorylation level was markedly decreased by the
loss of the CK2 catalytic subunit α or α′ and, to a larger extent, the
shared regulatory β subunit (Fig. 2d). CX4945 (silmitasertib) is a potent
and highly selective chemical inhibitor of CK2 that is being clinically
tested in anti-cancer and anti-virus
therapies. Treatment of
HEK293T cells, mESCs, and a panel of cancer cell lines by CX4945
resulted in a dose-dependent diminution of SUZ12S583 phosphoryla-
tion (Fig. 2e), further supporting the role of CK2 as the specific kinase
for SUZ12S583 phosphorylation in these cells.
Nature Communications |
(2022) 13:6781
3
Article
https://doi.org/10.1038/s41467-022-34431-1
Fig. 2 | CK2 is the kinase for the phosphorylation of S583 of SUZ12. a Kinase
prediction by web servers NetPhos 3.1 and PhosphoNET. The peptide motif around
S583 was used for the prediction. The top three hits are listed in each case with CK2
highlighted in blue. b In vitro phosphorylation assay. CK2 complexes were
expressed in HEK293T cells and GST-tagged full-length SUZ12 WT and mutants
were expressed in bacteria. Total serine phosphorylation was measured by an anti-
phosphoserine antibody (uncropped gel images of this figure are shown in Sup-
plementary Figs. 16 and 17). A representative of three independent experiments is
shown. c Stable knockdown of CK2 subunits a, a′, and b in NT2/D1 cells. Two
independent shRNAs were tested for knockdown efficiency. A representative of two
independent experiments is shown. d S583 phosphorylation in NT2/D1 in the pre-
sence of CK2 knockdown. A representative of three independent experiments is
shown. In d and e, immunoprecipitates of the anti-SUZ12 antibody were used for
the detection of S583 phosphorylation. e Effect of chemical inhibition of CK2 kinase
activity on S583 phosphorylation. Cell lines were treated with indicated con-
centrations of CX4945 for 24 h. Source data are provided as a Source Data file.
To examine the contribution of the interacting residues to enzy-
matic activity in a more complete system, we purified ectopically
expressed EZH2-containing wild-type (WT) and mutant PRC2-5m
complexes from HEK293T cells (Supplementary Fig. 9). No endogen-
ous phosphorylated SUZ12 was detected in the purified SUZ12S583A
mutant complex (Supplementary Fig. 10). When SUZ12 harbors the
S583A single mutation and thus lacks phosphorylation at this site,
histone methylation was severely compromised (Fig. 4c). All methy-
lation states were affected by the S583A mutation (Supplementary
Fig. 11). In comparison, the S583D phosphomimetic mutant complex
did not display a defect in catalysis (Fig. 4c). In addition, the K683A
mutation of EZH2 and the H567A/S568A mutation of SUZ12 also
noticeably impaired the enzymatic activity in this context (Fig. 4c).
More directly, CK2-mediated in vitro re-phosphorylation of λ
phosphatase-treated WT PRC2-5m pronouncedly enhanced histone
methylation (Fig. 4d).
To dissect how S583 phosphorylation facilitates catalysis, we
performed a steady-state enzymology study with PRC2-5m containing
WT or S583A mutant SUZ12 using histone peptide substrates (Fig. 4e).
Assays were conducted under both histone peptide-saturating and
SAM-saturating conditions (Fig. 4e). As indicated by the Km values
changing from 0.5 to 2.9 μM, loss of S583 phosphorylation most pro-
foundly affected SAM binding to PRC2, whereas histone peptide
binding was only moderately weakened (Fig. 4e). This is in line with the
structural observation that S583 phosphorylation stabilizes the SAM-
binding pocket (Fig. 3c). In comparison, enzyme turnover kcat did not
seem to be affected by the mutation (Fig. 4e).
To check if phosphorylation of S583 of SUZ12 plays a role in PRC2
binding to nucleosomes, we assembled mononucleosomes with a
biotinylated DNA and performed avidin bead pulldown assays. Com-
pared to the WT counterpart, PRC2-5m containing the S583A mutation
displayed markedly reduced interaction with nucleosomes; however,
in the absence of histone H3 tail (residues 1–27), nucleosome binding
was equally diminished for the WT and mutant PRC2-5m (Fig. 4f),
suggesting S583 phosphorylation may be necessary for optimal bind-
ing of enzyme active site to the histone tail in the nucleosomal context,
especially when SAM concentration is not saturating but likely limiting.
Congruently, nucleosomes were bound less tightly by λ phosphatase-
treated WT PRC2-5m, compared to the same complex re-
phosphorylated by CK2 in vitro (Fig. 4g). To gain a quantitative view
of nucleosome binding, we performed native gel shift assays. The
nucleosome binding affinity of the S583A mutant PRC2-5m complex
was reduced by roughly two folds compared to that of the WT complex
(Fig. 4h and Supplementary Fig. 12a, b). Correspondingly, nucleosome
binding by PRC2-5m was also impaired in the absence of the N-terminal
tail of histone H3 (Supplementary Fig. 12c).
S583 phosphorylation promotes reporter gene repression
A transient expression luciferase gene reporter system was pre-
viously established to recapitulate PRC2-dependent gene repression
in cells58. Here, we used a similar system to examine the role of S583
phosphorylation in reporter gene repression in an engineered
knocked out
line with endogenous SUZ12
HEK293T cell
(HEK293TΔSUZ12)11. Specifically, a “6×GAL4UAS” cassette was inserted
upstream of the thymidine kinase (TK) promoter that controls the
luciferase reporter gene. SUZ12 protein fused to the GAL4 DNA
expressed in
binding domain (GAL4DBD) was
HEK293TΔSUZ12 cells together with the reporter plasmid. GAL4DBD
transiently
Nature Communications |
(2022) 13:6781
4
Article
https://doi.org/10.1038/s41467-022-34431-1
Fig. 3 | S583 phosphorylation stabilizes PRC2 active site. a Alignment of
SUZ12 sequences around residue S583 in several model organisms. b Structure of
the minimal PRC2-EZH1 complex with a phosphorylated S583. The overall structure
is provided on the left with a close-up view on the right. Protein subunits, peptides,
and the cofactor included in the crystal structure are color-coded and labeled. A
previously reported structure of the minimal PRC2-EZH2 complex that lacks S583
phosphorylation (PDB 5HYN) is superimposed on the current structure in the close-
up view and is colored in gray. Conformational change of the PDS loop induced by
S583 phosphorylation is indicated by the red arrow. c Phosphoserine-centered
interaction network. Interacting residues are shown as sticks. The red arrow indi-
cates the single turn helix of the SAM-binding pocket. Some interacting residues are
omitted for clarity. d 2D schematic of the interaction network. Interacting residues
from the SET domain are colored in blue, the DSpE core motif from the PDS loop is
colored in green, and the rest are colored in black.
recruits ectopically expressed SUZ12 in complex with other endo-
genous PRC2 subunits to the TK promoter (Fig. 5a).
We first tested the dependence of reporter gene repression on
PRC2 enzymatic activity. Compared to the GAL4DBD alone control
construct, full-length SUZ12 was sufficient to confer reporter gene
repression (Fig. 5b and Supplementary Fig. 13). The VEFS domain of
SUZ12 essential for the assembly of the minimally active PRC2,
EZH2–EED–SUZ12(VEFS), mediated comparable gene repression
(Fig. 5b and Supplementary Fig. 13), suggesting that accessory subunits
of PRC2 are largely dispensable for this artificial targeting system and
thus will not complicate data interpretation. A highly specific PRC2
enzyme inhibitor EPZ6438 relieved reporter gene repression in a dose-
dependent manner in both contexts (Fig. 5b and Supplementary
Fig. 13), indicating the observed reporter gene repression was corre-
lated with PRC2 enzymatic activity in cells.
A W555C mutation within the VEFS domain of Drosophila SU(Z)12
was previously shown to cause a dramatic decrease in PRC2 enzymatic
activity in vitro55. In the current assay, the equivalent W591C mutation
of human SUZ12 led to reporter gene derepression (Fig. 5c and Sup-
plementary Fig. 13). Similar to this positive control, the S583A mutation
of SUZ12 also derepressed the reporter gene when present in either the
full-length or minimal construct (Fig. 5c and Supplementary Fig. 13),
which suggests S583 phosphorylation can directly promote reporter
gene repression in cells, likely by enhancing PRC2 enzymatic activity.
In support of the role of S583 phosphorylation, the S583D phospho-
mimetic mutation was not found to compromise the reporter gene
repression (Supplementary Fig. 14).
Loss of S583 phosphorylation disturbs PRC2 targeting and
H3K27me3 deposition in mESCs and impairs cell identity main-
tenance during mESC differentiation
PRC2 is known to be required for proper differentiation of mESCs, but
dispensable for self-renewal and pluripotency of these cells59,60. In
mESCs, mSUZ12 is substantially phosphorylated at residue S585, the
Nature Communications |
(2022) 13:6781
5
Article
https://doi.org/10.1038/s41467-022-34431-1
equivalent of residue S583 of human SUZ12 (Fig. 1b). To study how
S583 phosphorylation impacts PRC2 function in vivo, we re-expressed
3×FLAG-tagged human SUZ12 WT (SUZ12WT) and S583A (SUZ12S583A)
mutant that eliminates phosphorylation in a mSUZ12 knockout (KO)
mESC line61, using lentiviral vectors (Fig. 6a). Pluripotent mESCs were
maintained in serum-free 2i media62,63. An equal amount of WT and
mutant SUZ12 was bound to EZH2 in an anti-EZH2 co-immunopreci-
pitation (Co-IP) assay (Fig. 6b), indicating the phosphoserine-centered
interactions between SUZ12 and EZH2 are not essential for PRC2
assembly. In addition, PRC2 containing SUZ12S583A displayed a slightly
weaker association with bulk chromatin in mESCs than the WT PRC2
(Fig. 6c), suggesting a possible chromatin binding defect.
Nature Communications |
(2022) 13:6781
6
Article
https://doi.org/10.1038/s41467-022-34431-1
Fig. 4 | S583 phosphorylation promotes PRC2 function in vitro. a Radioactive
methyltransferase assay with the PRC2-EZH1 ternary complex (EZH1–EED–SUZ12
(VEFS)) and mononucleosome substrates. Assays were performed using 150 and
450 nM of the WT and mutant enzymes (uncropped gel images of this figure are
shown in Supplementary Figs. 16 and 17). A representative of three independent
experiments is shown. b The same as a, except that the PRC2-EZH2 ternary complex
(EZH2–EED–SUZ12(VEFS)) was used. A representative of three independent
experiments is shown. c Radioactive methyltransferase assay with mononucleo-
some substrates and PRC2-5m WT and mutant holo complexes expressed in
HEK293T cells. 50 and 100 nM of WT and mutant enzymes were used. A repre-
sentative of three independent experiments is shown. d Radioactive methyl-
transferase assay with λ phosphatase and CK2-treated PRC2-5m. WT PRC2-5m used
in this assay was expressed in Sf9 cells. λ phosphatase-treated PRC2-5m was sub-
jected to size exclusion chromatography to remove λ phosphatase. Depho-
sphorylated PRC2-5m was re-phosphorylated by human CK2 in vitro. 50 and
100 nM of the dephosphorylated and re-phosphorylated PRC2-5m were used for
the methyltransferase assay and compared. A representative of two independent
experiments is shown. e Steady-state enzymology study of PRC2-5m WT and S583A
mutant. Assays performed under the substrate peptide-saturating condition are
shown on the left and assays under the SAM-saturating condition are on the right.
GraphPad Prism was used to fit the data and derive Km and kcat values. n = 3 inde-
pendent enzymatic reactions. Error bars represent mean ± SEM. f Nucleosome
binding assay. Biotinylated nucleosomal DNA was generated by PCR with a biotin-
labeled primer. Bound WT and mutant PRC2-5m expressed in HEK293T cells are
indicated by anti-EZH2 signals. Anti-H3 signals for H3 and H3ΔN are controls for the
bait. H3ΔN lacks residues 1–27 of histone H3. A representative of two independent
experiments is shown. g The same as f, except that dephosphorylated and re-
phosphorylated Sf9-expressed PRC2-5m were used for the binding assay. Two
amounts of the bound PRC2-5m (1× and 3×) were loaded on the gel. A repre-
sentative of two independent experiments is shown. h Native gel shift nucleosome
binding assay. Mononucleosomes and HEK293T-expressed PRC2-5m WT and
mutant were used for the binding assay. Kd values were calculated based on n = 3
independent gel shift assays. Error bars represent mean ± SEM. Source data are
provided as a Source Data file.
Fig. 5 | S583 phosphorylation facilitates reporter gene repression. a Schematic
of GAL4-based reporter gene repression assay. b PRC2 enzymatic activity-
dependent reporter gene repression. GAL4DBD-HA-tagged SUZ12-FL or SUZ12(-
VEFS) was transiently expressed in HEK293T cells that lack endogenous SUZ12.
EPZ6438 is a selective enzyme inhibitor of PRC2. In b and c, assays were performed
on three different days with the measurement of two replicate wells recorded each
time. Signals were normalized to. GAL4DBD-HA negative control. p values were
derived from two-sided t-tests performed in Microsoft Excel. n = 6 biologically
independent experiments. Error bars represent mean ± SEM. c Effect of S583A loss-
of-phosphorylation mutation on reporter gene repression. Assays were performed
in the context of SUZ12-FL or SUZ12(VEFS). W591C is a known mutation within the
SUZ12(VEFS) that disrupts PRC2 enzymatic activity. Source data are provided as a
Source Data file.
To assess PRC2 recruitment and H3K27me3 deposition on indi-
vidual gene loci in mESCs expressing SUZ12WT or SUZ12S583A, we carried
out chromatin immunoprecipitation (ChIP)-qPCR experiments focus-
ing on known PRC2 targets. The active pluripotent gene NANOG served
as a non-target negative control. As shown by the anti-FLAG ChIP data,
the chromatin recruitment of SUZ12 was impaired by the S583A
mutation on members of HOX gene clusters, HOXA7 and HOXD12,
where PRC2 is highly enriched (Fig. 6d). Similar reduction in chromatin
binding was also observed for the mutant on other lineage marker
genes with varying degrees of PRC2 enrichment, including GATA4,
FGF5, and NESTIN (Fig. 6d). In line with the defect in chromatin binding,
H3K27me3 levels were also affected by the mutation on many of these
gene loci (Fig. 6e).
We next investigated the ability of PRC2 to maintain the dif-
ferentiated state of mESCs, using a recently reported replating
assay64. mESCs expressing SUZ12WT or SUZ12S583A were differentiated
to form embryoid bodies (EBs), which were subsequently dis-
sociated into single cells; these single cells were next replated in 2i
Nature Communications |
(2022) 13:6781
7
Article
https://doi.org/10.1038/s41467-022-34431-1
media, a growth condition that challenges the maintenance of the
differentiated cell identity (Fig. 6f). EBs formed by the WT and
mutant mESCs were indistinguishable in morphology (Fig. 6g),
indicating mESCs lacking S583 phosphorylation retains the capacity
to differentiate, despite the apparent defect in PRC2 targeting and
H3K27me3 deposition in the pluripotent state of mESCs (Fig. 6d, e).
A drastic phenotype appeared when differentiated cells from these
EBs were replated in 2i media: a large number of SUZ12S583A-con-
taining cells were reverted to a pluripotent stem cell state as shown
by alkaline phosphatase (AP) staining, whereas cell identity rever-
sion was only sporadic for SUZ12WT-containing cells (Fig. 6h, i and
Supplementary Fig. 15), suggesting the phosphorylation of S583 of
Nature Communications |
(2022) 13:6781
8
Article
https://doi.org/10.1038/s41467-022-34431-1
Fig. 6 | S583 phosphorylation is important for PRC2 recruitment, H3K27me3
deposition, and cell identity maintenance. a SUZ12 expression levels in the
parental and engineered mESCs. SUZ12 from the parental mESC line and engi-
neered mESC lines with the re-expression of 3×FLAG-SUZ12-FL-WT or 3×FLAG-
SUZ12-FL-S583A was checked by western blot (uncropped gel images of this figure
are shown in Supplementary Fig. 18). A representative of two independent
experiments is shown. b Integrity of PRC2 assembly. Anti-EZH2 antibody was used
to capture re-expressed WT and mutant SUZ12 by co-immunoprecipitation. Both
bound and unbound fractions were analyzed by western blot for SUZ12 (prey),
EZH2 (bait), and GAPDH (loading control). Rabbit IgG was a negative control. A
representative of two independent experiments is shown. c PRC2 binding to bulk
chromatin. FLAG immunoprecipitation was used to capture FLAG-tagged SUZ12
and associated chromatin fragments generated by sonication. Bound chromatin is
indicated by anti-H3 signals. A representative of two independent experiments is
shown. d Anti-FLAG ChIP-qPCR. Binding of WT and S583A mutant SUZ12 to known
PRC2 targets was compared. In d and e, two independent ChIP experiments were
performed each with three qPCR replicates. p values were derived from two-sided t-
tests performed in Microsoft Excel. NANOG is a negative control. n = 6 independent
experiments. Error bars represent mean ± SEM. e Anti-H3K27me3 ChIP-qPCR.
H3K27me3 deposition at known PRC2 targets in mESCs expressing WT or S583A
mutant SUZ12 was compared. f Schematic of the replating assay. g EB formation.
Morphology of EBs differentiated from mESCs expressing WT or S583A mutant
SUZ12 was compared. Scale bar stands for 1 mm. A representative of three inde-
pendent experiments is shown. h Reversion of the differentiated cell identity. Cells
dissociated from EBs expressing WT or S583A mutant SUZ12 were replated in 2i
media and checked for pluripotency by AP staining. Replating assays were per-
formed three times using cells from three independent EB formation experiments.
Scale bar stands for 1 mm. i Quantification of AP staining. Relative areas stained by
AP were quantified in ImageJ. p values were derived from two-sided t-tests per-
formed in Microsoft Excel. n = 3 biologically independent experiments. Error bars
represent mean ± SEM. Source data are provided as a Source Data file.
Fig. 7 | A model of PRC2 function promoted by the phosphorylation of S583 of
SUZ12. Cartoons illustrate how SUZ12 phosphorylation stabilizes enzyme active
site and promotes PRC2 function. S583 phosphorylation induces conformational
change of the PDS loop of SUZ12, stabilizes the SAM-binding pocket, and converts a
weak binding state of SAM to a strong binding state. This also facilitates histone
substrate H3K27 binding. PRC2 recruitment and H3K27me3 deposition are
enhanced in this way. Cell identity maintenance is compromised when differ-
entiated mESCs are challenged in 2i media in the absence of S583 phosphorylation.
Created with BioRender.com.
SUZ12 is essential for PRC2 function in maintaining cell identity
during mESC differentiation.
AEBP2 of PRC2.2 and directly mediates PRC2 dimerization crucial for
chromatin binding10,11.
Discussion
PRC2 sets an epigenetic threshold for maintaining cell identity2. In
supporting this pivotal function, PRC2 enzymatic activity is subjected
to complex cellular regulation. In the current work, we reveal a unique
phosphorylation-dependent mechanism that stimulates PRC2 enzy-
matic activity. Our structural study provides direct evidence for how a
posttranslational modification of a PRC2 core subunit may regulate
enzyme function. Upon phosphorylation of residue S583 in the
SUZ12(VEFS) domain, the PDS loop undergoes a dramatic conforma-
tional change:
it transitions from a partially disordered state to
become engaged with the catalytic SET domain, stabilizing the enzyme
active site (Fig. 7). The PDS loop is an addition to a collection of flexible
structural elements dictating distinct functional states of PRC2. Other
notable examples include the stimulation-responsive motif (SRM) of
EZH2 that bridges the stimulating signal from H3K27me3 to the SET
domain8,9, the bridge helix of EZH2 that connects nucleosomal sub-
strates and the SET domain48, and the C2 domain of SUZ12 that
associates with the accessory subunits MTF2 and PHF19 of PRC2.1 and
In analyzing the structural plasticity and phosphorylation-
dependent interactions of the PDS loop, we noticed that residue
K684 of the SET domain of EZH1 (the equivalent of residue K683 of
EZH2) close to the SAM-binding pocket is stabilized by an acidic motif
of SUZ12 centering on the phosphoserine (Fig. 7). Consistently, our
enzymology data using the WT and S583A mutant PRC2-5m confirmed
that SAM binding was severely compromised for the mutant (Fig. 4e).
In addition, when SAM concentration is limiting, PRC2 binding to
nucleosomal substrates is also impaired in the absence of S583 phos-
phorylation (Fig. 4f–h), likely due to the structural coupling of SAM
and histone H3 tail binding to the enzyme active site. Accordingly,
diminished PRC2 enzymatic activity caused by disruption of the
phosphoserine-centered interactions can be readily rationalized by
weakened binding of PRC2 to SAM and histone tail (Fig. 4a–d). Intra-
cellular availability of SAM as a critical metabolite is known to influence
histone methylation and gene regulation65. In this regard, phosphor-
ylation of S583 of SUZ12 may serve as a cellular mechanism to maintain
chromatin occupancy and enzymatic activity of PRC2 in case of
metabolic perturbations.
Nature Communications |
(2022) 13:6781
9
Article
https://doi.org/10.1038/s41467-022-34431-1
mESC differentiation provides a valuable system for studying
PRC2 function in vivo. Self-renewal and pluripotency of mESCs are not
changed even by some extreme alterations of PRC2, including partial
or full deletion of SUZ12, which results in redistribution or complete
loss of H3K27me3, respectively61. We found that the majority of SUZ12
in mESCs are phosphorylated at residue S583 and that the S583A
mutation is sufficient to reduce PRC2 enrichment on target genes,
which is also accompanied by a decrease in H3K27me3 deposition
(Figs. 1b and 6d, e). A prominent cell identity reversion phenotype
arises when differentiated mESCs dissociated from EBs are replated in
2i media promoting pluripotency64. The number of SUZ12S583A-
expressing mESCs reverting to the pluripotent state greatly exceeds
that of SUZ12WT-expressing mESCs (Fig. 6h, i), suggesting that PRC2
function in cell identity maintenance is compromised by the lack of
S583 phosphorylation (Fig. 7). A hypomorphic mutation of EZH2 also
impedes cell identity maintenance during mESC differentiation, and it
is proposed that full methylation of H3K27 is required for stable
commitment to differentiation64. In this regard, S583 phosphorylation
can be a missing piece of the puzzle of cell identity maintenance by
PRC2. It is not impossible that defects in cell differentiation not
revealed by the visual inspection of EBs from the SUZ12S583A-expressing
mESCs may also exist. In addition, it remains to be studied if the level of
S583 phosphorylation changes during early differentiation or in other
developmental stages, although it does appear to vary in some cancer
cell lines (Fig. 1e and Supplementary Fig. 3).
SUZ12 was previously found in the CK2 interactome in mitotic
HEK293T cells66. In this study, we showed that CK2 is the kinase
responsible for phosphorylation of S583 of SUZ12 (Fig. 7). This finding
connects a widespread cell signaling event known to regulate cell
proliferation and apoptosis to a key epigenetic mechanism preserving
cell identity. Our data also predict that clinically relevant CK2 inhibi-
tors may impair PRC2 function indirectly by inhibiting CK2-mediated
S583 phosphorylation. CK2 is a ubiquitous and constitutively active
kinase, and CK2 expression is often elevated in cancer cells39. This
raises the question of whether and how S583 phosphorylation is
regulated under physiological conditions. In addition, given that
CK2 serves as a subunit of PRC1.3 and PRC1.5 and that PRC1 and PRC2
co-occupy target loci in Polycomb chromatin domains, it would be
interesting to explore if S583 of SUZ12 is phosphorylated in the context
of these variant PRC1 complexes, which would add another mechan-
istic link between the two major complexes of the Polycomb repressive
system.
Methods
Cell culture
HEK293T, A172, MDA-MB-231, and U118MG cell lines were cultured in
DMEM (Sigma, Cat No. D5796) supplemented with 10% FBS (Sigma, Cat
No. 2442) and 1× penicillin-streptomycin (Sigma, Cat No. P0781).
LNCaP and 22RV1 cells were cultured in RPMI 1640 (ATCC, Cat No.
30–2001) supplemented with 10% FBS and 1× penicillin-streptomycin.
NT2/D1 cells were cultured in DMEM (ATCC, Cat No. 30–2002) sup-
plemented with 10% FBS and 1× penicillin-streptomycin. MCF-7 cells
were cultured in EMEM (ATCC, Cat No. 30–2003) supplemented with
10 µg/ml human insulin (Sigma, Cat No. 91077C), 10% FBS, and 1×
penicillin-streptomycin. MCF10A cells were cultured in the Mammary
Epithelial Cell Growth Medium (Sigma, Cat No. C-21010) supplemented
with 1× penicillin-streptomycin. BT-474 cells were cultured in RPMI
1640 supplemented with 20% FBS, 10 µg/ml human insulin, 2 mM L-
glutamine, and 1× penicillin-streptomycin. mESCs were cultured in 2i
media, containing a 1:1 mix of DMEM/F12 (GIBCO, Cat No. 11320033)
and Neurobasal media (GIBCO, Cat No. 21103049), 1× penicillin-
streptomycin (Sigma, Cat No. P0781), 0.05% BSA (Fisher, Cat No.
15260037), 100 μM BME (Sigma, Cat No. M3148), 0.5× GlutaMax
(GIBCO, No. 35050061), 0.5% N-2 supplement (GIBCO, Cat No.
17502048), 1% B-27 Supplement (GIBCO, Cat No. 17504044), 3 μM GSK
inhibitor CHIR99021 (Cayman Chemical, Cat No. 131225), 1 μM MEK
inhibitor PD0325901 (Cayman Chemical, Cat No. 130345), and LIF
produced in the lab. The activity of the homemade LIF was assayed
based on marker gene expression and morphology of mESC colonies.
rabbit
Antibodies
The following commercial antibodies were used in this study: rabbit
anti-SUZ12 (Cell Signaling, Cat No. 3737, 1:1000 dilution for western
blot), rabbit anti-CK2α (GeneTex, Cat No. GTX107897, 1:500 dilution
for western blot), rabbit anti-CK2α′ (Bethyl, Cat No. A300-199A, 1:500
rabbit anti-CK2β (Bethyl, Cat No.
dilution for western blot),
A301–984A,
anti-
1:500 dilution for western blot),
phosphoserine (Abcam, Cat. No. ab9332, 1:500 dilution for western
blot), mouse anti-GAPDH (Invitrogen, Cat No. MA515738, 1:1000 dilu-
tion for western blot), rabbit anti-EZH2 (Cell Signaling, Cat No. 5246,
1:1000 dilution for western blot), rabbit anti-H3 (Cell Signaling, Cat No.
4499, 1:5000 dilution for western blot), rabbit anti-HA tag (Cell Sig-
naling, Cat No. 3724, 1:1000 dilution for western blot), mouse anti-
FLAG tag (Sigma, Cat No. F1804, 1:1000 dilution for western blot and
1:500 dilution for ChIP), rabbit anti-H3K27me3 (Cell signaling, Cat No.
9733, 1:1000 dilution for western blot and 1:200 dilution for ChIP),
rabbit anti-H3K27me2 (Millipore, Cat No. 07–452, 1:500 dilution for
western blot), rabbit anti-H3K27me1 (Millipore, Cat No. 07–448, 1:500
dilution for western blot), and rabbit anti-β-Tubulin (Cell Signaling, Cat
No. 2128, 1:1000 dilution for western blot). Rabbit antibody specific for
SUZ12S583p was generated by the Animal Resource Center (ARC) of
UT Southwestern Medical Center using the KLH conjugated peptide:
KLH-CQEMEVD-[phospho-S]-EDEKDPE. The anti-SUZ12S583p antibody
in rabbit sera was purified by peptide affinity columns containing
crosslinked apo or phosphoserine peptides. The anti-SUZ12S583p
antibody was diluted by 1000 folds for western blot.
Re-expression of SUZ12 in SUZ12 knockout mESCs with lentiviral
vectors
SUZ12 knockout mESC line is a generous gift from Dr Kristian Helin
(Institute of Cancer Research)61. SUZ12 was re-expressed in the knockout
cell line using lentiviral vectors. cDNA sequence encoding human WT or
S583A mutant SUZ12 with an N-terminal 3×FLAG tag was subcloned into
the pCDH-EF1α-MCS-IRES-Puro vector using XbaI and EcoRI restriction
sites. For lentivirus production, the pCDH-EF1α−3×FLAG-SUZ12 WT or
S583A plasmid (5 μg), psPAX2 (5 μg), and pVSV-G (0.5 μg) were co-
transfected into HEK293T cells at ∼70% confluence. The medium con-
taining lentivirus particles was harvested 48 h post transfection and
centrifuged at 200 g for 10 min. The supernatant was passed through a
0.45-μm filter and precipitated by 1/3 volume of Lenti-X concentrator
(Takara, Cat No. 631231), followed by mixing on a nutator for 30 min at
4 °C and then centrifugation at 1500 g for 45 min. The Lentivirus parti-
cles were resuspended in the 2i condition medium, aliquoted, flash-
frozen by liquid nitrogen, and stored at −80 °C till transduction.
For transduction, 1 × 105 SUZ12 knockout mESCs were seeded at a
6-well plate, 24 h before transduction. Cells were transduced by len-
tiviruses expressing respective SUZ12 constructs together with 10 μg/
ml polybrene (Sigma, Cat No. TR-1003-G). 1 μg/ml puromycin (Sigma,
Cat No. P8833) was supplemented to the growth media 48 h post
transduction. After 72 h, mESCs were diluted and seeded into 96-well
plates in the presence of 1 μg/ml puromycin. Single-cell colonies
expressing comparable amounts of WT and S583A mutant SUZ12 were
identified by western blotting and were frozen for downstream
analysis.
Stable knockdown of CK2
Lentiviruses for stable knockdown of CK2 components, CK2α, CK2α′,
and CK2β, were generated using the pLKO.1 lentiviral vector that
expresses corresponding shRNAs (Sigma). shRNA sequences are pro-
vided in Supplementary Table 2. The same lentivirus production
Nature Communications |
(2022) 13:6781
10
Article
https://doi.org/10.1038/s41467-022-34431-1
protocol described above was followed. NT2/D1 cells were seeded
onto 6-well plates at 30% confluence. Lentiviruses were added into the
cell culture together with 10 μg/ml polybrene 24 h post cell seeding.
After 48 h of transduction, cells were selected in the growth medium
containing 1 μg/ml puromycin for 6 days with a medium change every
48 h. Cells resistant to puromycin were lysed for western blot to detect
the knockdown efficiency and stored for downstream analysis.
Recombinant protein expression and purification
The ternary human PRC2-EZH1 and PRC2-EZH2 complexes (EZH1/
2–EED–SUZ12(VEFS)) used for enzymatic assays contained a full-length
EZH1/2 (residues 1–747 and 1–746) fused to the VEFS domain of SUZ12
(residues 543–695) and a full-length EED (residues 1–441). cDNA cor-
responding to the His6−2×Protein A-TEV-EZH1/2-LVPRGS-SUZ12(VEFS)
fusion construct was subcloned into the p416GAL1 vector with a URA
marker. EED was subcloned into the p415-GAL1 vector (LEU marker).
The minimal complex used for crystallization contained the following
modifications: residues 188–229 of EZH1 were replaced by a
GGGSGGGSGGGS linker sequence, residues 353–413 of EZH1 were
deleted, residues 492–496 of EZH1 were replaced by a GGSGG linker
sequence, and residues 1–77 of EED was replaced by a StrepII tag. The
two plasmids were co-transformed into an S. cerevisiae CB010 strain,
followed by selection on a synthetic drop-out medium plate lacking
uracil and leucine. Starters of transformed yeast cells were grown in
synthetic drop-out media with 2% raffinose. Protein expression was
induced by 2% galactose in YP media for about 20 h. The minimal
complex was purified by IgG-sepharose and eluted from the resin by
TEV protease cleavage. The protein complex was further purified by
size exclusion chromatography on Superdex 200. Protein complex
purity was assessed by SDS–PAGE.
WT and mutant human PRC2-5m complex (EZH2–EED–
SUZ12–RBBP4–AEBP2) was expressed in HEK293T cells. Briefly, cDNAs
of HA-EZH2, His6-EED, and HA-RBBP4 were inserted into the pCS2+
vector. cDNA corresponding to 2×Protein A-TEV-SUZ12-HA was inserted
into the pCS2+ vector. cDNA corresponding to 2×Protein A-3C-AEBP2
(residues 1–295) was inserted into the pCS2+ vector. These five plasmids
were co-transfected into HEK293T cells at ∼70% confluence by poly-
ethylenimine (PEI). Cells were harvested 48 h post transfection. Protein
complexes were purified by IgG affinity resin and released by TEV and
HRV-3C protease cleavage overnight at 4 °C. Protein complex purity was
confirmed by SDS–PAGE. WT human PRC2-5m complex expressed in
Sf9 cells was purified as described previously11.
CK2α and CK2α′ cDNAs were tagged at the 5′ ends with a 2×Pro-
tein A tag followed by a TEV protease site and were inserted into the
pHEK293 ultra expression vector (Takara). CK2β was tagged with a
SUMO tag at the 5′-end and was cloned into the pHEK293 ultra vector
as well. HEK293T cells were co-transfected by PEI with the plasmids
expressing CK2α plus CK2β or CK2α′ plus CK2β. Cells were harvested
48 h post transfection. CK2 complexes were purified by IgG affinity
column, and protein purity was assessed by SDS–PAGE.
To prepare GST-SUZ12 proteins from bacterial expression, the
cDNA sequence encoding full-length human SUZ12 (1–739) was sub-
cloned into the pGEX-4T-1 vector. Alanine mutations were introduced
by site-directed mutagenesis. Rosetta 2(DE3) cells transformed with the
expression plasmid were induced with 0.5 mM IPTG at OD600 of 0.6 for
16 h at 18 °C. Cells were harvested and lysed in cell lysis buffer (50 mM
Tris-HCl pH 8.0, 300 mM NaCl, 10% glycerol, and 3 mM dithiothreitol
(DTT)) by sonication. After clarification by centrifugation, glutathione
agarose beads (Thermo Scientific) were added to the supernatant and
incubated at 4 °C for 2 h with mixing. The beads were washed thor-
oughly with cell lysis buffer supplemented with 0.1% NP40. The bound
GST-SUZ12 was eluted with the cell lysis buffer supplemented with
20 mM glutathione. Eluted proteins were further purified on a Superdex
200 size exclusion column (GE Healthcare) equilibrated with 20 mM
Tris-HCl pH 8.0, 100 mM NaCl, and 2 mM DTT.
Crystallization and structure determination
The truncated minimal EZH1–EED–SUZ12(VEFS) complex at 10 mg/ml
was pre-incubated with 0.5 mM H3K27M peptide, 0.5 mM H3K27me3
peptide, and 1 mM SAM for 1 h on ice before crystallization. The initial
crystallization conditions were screened by the sitting drop vapor
diffusion method at 22 °C. Conditions obtained from the initial screens
were optimized using the hanging-drop vapor diffusion method.
Crystals were grown by mixing 1 μl protein solution at 10 mg/ml with
1 μl of the reservoir solution containing 10% PEG3350, 100 mM
ammonium sulfate, and 50 mM HEPES pH 6.8. Diffraction-quality
crystals were cryoprotected with the reservoir solution supplemented
by 15% glycerol and flash-frozen in liquid nitrogen. Diffraction data
were collected at a synchrotron light source and processed with
HKL200067. Scaled data were imported and used for molecular
replacement with PDB 5WG6 as the search model68,69. The structure
was refined by REFMAC5 and autoBUSTER, and refinement statistics
were generated by PHENIX70–72. Model building and iterative refine-
ment were carried out using Coot73. Structure figures were generated
by PyMOL74.
The crystal had a C2 space group, and two copies of complexes
were found in one asymmetric unit, with one copy displaying a
noticeably higher degree of mobility.
Nucleosome reconstitution
Reconstitution of mononucleosomes was performed using the salt
dialysis method. Briefly, Xenopus laevis histone octamers and 147-bp
“601” DNA were mixed for 2 h in a buffer containing 2 M NaCl, 10 mM
Tris-HCl pH 7.5, 0.1 mM EDTA, and 1 mM 2-mercaptoethanol (BME).
The mixture was subjected to sequential salt dialysis in the same buffer
with reduced salt concentration like this: 1 M NaCl for 2 h, 0.8 M NaCl
for 2 h, 0.6 M NaCl for 2 h, 0.3 M NaCl for 2 h, 0.15 M NaCl for over-
night, and 0 M NaCl for 4 h. Mononucleosomes with a tailless histone
H3 lacking residue 1–27 were reconstituted following the same
procedure.
Histone methyltransferase assay
For the enzymology study, the reaction buffer contains 25 mM Tris pH
8.0, 10 mM NaCl, 1 mM EDTA, 2.5 mM MgCl2 and 2.5 mM DTT. In each
20 μl reaction system, 50 nM PRC2-5m was incubated with indicated
concentrations of biotin-labeled H3 (residues 21–44) peptide (Anaspec,
Cat No. AS-64440), Adenosyl-L-Methionine, S-[methyl-3H]- (SAM-3H)
(PerkinElmer, Cat No. NET155H001MC), and SAM (NEB, Cat No.
B9003S) at 30 °C for 1 h. 100 μM peptide was used for the substrate
peptide-saturating condition, and 64.6 μM SAM (64 μM cold SAM plus
0.6 μM hot SAM) was used for the SAM-saturating condition. For
quantification by a scintillation counter, the reaction system was stop-
ped by adding 1 mM cold SAM. 10 μl stopped reaction mixture was then
spotted onto P81 phosphocellulose paper (Reaction Biology Corpora-
tion) and air-dried for 3 h. P81 paper was washed with 50 ml of 50 mM
Na2CO3/NaHCO2 at pH 9.0 for 5 times, briefly rinsed with acetone, air-
dried for 1 h, and immersed in 4 ml of scintillation fluid. The radioactive
activity was quantified according to disintegrations per minute (DPM).
The reaction of the ternary complexes, EZH1–EED–SUZ12(VEFS)
and EZH2–EED–SUZ12(VEFS), with nucleosomal substrates was carried
out following the same protocol, except that 150 and 450 nM enzymes,
300 nM mononucleosomes, and 640 nM SAM were used in each
reaction. In the case of the reaction using PRC2-5m, 50 and 100 nM
enzymes were used. For quantification by autoradiography, the reac-
tion was quenched by adding 7 μl of 4× sample loading dye and boiling
at 85 °C for 5 min. The reaction mixture was separated by SDS–PAGE,
followed by exposure to X-ray film to detect the methylation level.
In vitro phosphorylation of GST-SUZ12
For the in vitro phosphorylation assay, the reaction buffer contains
50 mM Tris-HCl pH 7.5, 10 mM MgCl2, 0.1 mM EDTA, 2 mM DTT, and
Nature Communications |
(2022) 13:6781
11
Article
https://doi.org/10.1038/s41467-022-34431-1
0.01% Brij35. In each 20 μl reaction system, 2 μg GST-SUZ12 was incu-
bated with 100 ng CK2α2β2 or CK2α′2β2, supplemented with 200 μM
ATP, at 30 °C for 30 min. The reaction was quenched by adding 7 μl of
4× sample loading dye and boiling at 85 °C for 5 min. The reaction
mixture was separated by SDS–PAGE, followed by western blotting to
detect the phosphorylation level.
In vitro dephosphorylation and re-phosphorylation of PRC2
complexes
For the dephosphorylation and re-phosphorylation of the PRC2-EZH1
minimal complex, 200 μg PRC2-EZH1 was incubated with 1 μl λ phos-
phatase (NEB, Cat No. P0753) in 50 μl reaction buffer (50 mM HEPES
7.5, 100 mM NaCl, 2 mM DTT, 0.01% Brij35, supplemented with 1 mM
MnCl2) at 30 °C for 1 h, followed by the purification using Superdex
200 in the gel filtration buffer (100 mM NaCl, 20 mM Tris 8.0, and
2 mM DTT).
re-
phosphorylated by 2 μl CK2 (NEB, Cat No. P6010) in 50 μl reaction
buffer (50 mM Tris 7.5, 10 mM MgCl2, 0.1 mM EDTA, 2 mM DTT, 0.01%
Brij35, supplemented with 200 μM ATP) at 30 °C for 15 min or 30 min.
The dephosphorylation and re-phosphorylation efficiencies were
checked by western blot. WT Sf9-expressed human PRC2-5m
was treated in the same way, except that 300 μg of the complex was
dephosphorylated and 150 μg of the dephosphorylated complex
was re-phosphorylated.
100 μg dephosphorylated PRC2-EZH1 was
Mass spectrometry analysis of phosphorylated SUZ12
mESCs were harvested in ice-cold PBS containing PMSF and protease
inhibitor cocktail. Pelleted cells were lysed by hypotonic buffer (10 mM
HEPES pH 7.9, 1.5 mM MgCl2, 10 mM KCl, 0.5% NP40, 2 mM DTT, 1 mM
PMSF, 1× protease inhibitor cocktail) on ice for 30 min and centrifuged
at 1000 g for 10 min to collect nuclei. The pelleted nuclei were lysed
with nuclear extraction buffer (20 mM HEPES pH 7.9, 1.5 mM MgCl2,
420 mM KCl, 20% glycerol, 2 mM DTT, 1 mM PMSF, 1× Protease inhi-
bitor cocktail) by rotating at 4 °C for 1 h, followed by centrifuging at
17,000 g for 10 min. Nuclear extracts were diluted with 1 volume of
hypotonic buffer and immunoprecipitated by anti-SUZ12 resins made
with cyanogen bromide (CNBr)-activated Sepharose-4B (Sigma, Cat
No. 9142). Captured materials were separated by SDS–PAGE and
stained with Gel-Code Blue (Thermo Scientific, Cat No. 24594). The gel
band containing SUZ12 was excised and submitted for MS/MS analysis.
Samples were digested overnight with trypsin (Pierce) following
reduction
iodoacetamide
(Sigma–Aldrich). The samples then underwent solid-phase extraction
cleanup with an Oasis HLB plate (Waters), and the resulting samples
were injected onto an Orbitrap Fusion Lumos mass spectrometer
coupled to an Ultimate 3000 RSLC-Nano liquid chromatography sys-
tem. Samples were injected onto a 75 μm i.d., 75-cm long EasySpray
column (Thermo) and eluted with a gradient from 0–28% buffer B over
90 min. Buffer A contained 2% (v/v) ACN and 0.1% formic acid in water,
and buffer B contained 80% (v/v) ACN, 10% (v/v) trifluoroethanol, and
0.1% formic acid in water. The mass spectrometer operated in positive
ion mode with a source voltage of 1.5 kV and an ion transfer tube
temperature of 275 °C. MS scans were acquired at 120,000 resolution
in the Orbitrap, and up to 10 MS/MS spectra were obtained in the ion
trap for each full spectrum acquired using higher-energy collisional
dissociation (HCD) for ions with charges 2–7. Dynamic exclusion was
set for 25 s after an ion was selected for fragmentation.
alkylation with DTT
and
and
Raw MS data files were analyzed using Proteome Discoverer v2.4
SP1 (Thermo), with peptide identification performed using Sequest HT
searching against the mouse protein database from UniProt or the
human protein database from UniProt with the sequence of the fusion
protein EZH1-SUZ12 included. Fragment and precursor tolerances of
10 ppm and 0.6 Da were specified, and three missed cleavages were
allowed. Carbamidomethylation of Cys was set as a fixed modification,
with oxidation of Met and phosphorylation of Ser, Thr, and Tyr set as a
variable modification. The false-discovery rate (FDR) cutoff was 1% for
all peptides. Peptide abundances are defined as the peak intensity of
the most abundant charge state for the peptide ion.
Native gel shift nucleosome binding assay
0.5 nM nucleosomes were incubated with PRC2-5m (2-fold serial dilu-
tion from 2 μM) in a 20 μl reaction system (10 mM Tris 8.0, 50 mM
NaCl, and 10% Glycerol) on ice for 30 min. Each 10 μl reaction mixture
was separated with a 4% native polyacrylamide gel (Acrylacrylamide/
Bis 60:1) in 1× TGE buffer (25 mM Tris, 190 mM Glycine, 1 mM EDTA) at
100 V for 1 h on ice. The native gel was stained by SYBR Gold. Binding
assays were performed in three replicates for both WT and mutant
PRC2-5m complexes, which were expressed in HEK293T cells. The gel
band was quantified in ImageJ, and the dissociation constant Kd was
calculated by fitting binding curves in GraphPad Prism.
Chromatin binding assay in mESCs
mESCs expressing 3×FLAG-SUZ12 (WT or S583A) were harvested with
ice-cold PBS containing PMSF and 1× protease inhibitor cocktail. Pel-
leted cells were lysed by the hypotonic buffer on ice for 30 min and
centrifuged at 1000 g for 10 min to collect nuclei. The nuclei were
sonicated in binding buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl,
2 mM DTT, 10% glycerol, 0.1% NP40, 2 mM DTT, 1 mM PMSF, 1× pro-
tease inhibitor cocktail) and clarified by centrifugation at 17,000 g for
10 min. The clarified supernatant was incubated with anti-FLAG beads
(Thermo Scientific, Cat No. PIA36797) at 4 °C for 1 h, followed by
washing with binding buffer for three times. Captured chromatin
fragments were eluted from the beads by 1.5 mg/ml FLAG peptide and
analyzed by western blot to detect histone H3.
Reporter gene repression assay
SUZ12 knockout HEK293T cells were made in the lab previously using
the CRISPR/Cas9 gene-editing system11. The reporter vector with
6×GAL4UAS-TK-luciferase (G6-TK-luc) was also previously generated58.
DNA fragments encoding GAL4DBD-HA-SUZ12 were cloned into the
pCS2+ vector between its EcoRI and XhoI sites. SUZ12 knockout
HEK293T cells were plated at a density of ∼0.35 × 106 cells per well in
6-well plates and cultured for 20 h before transfection. After growing
for 20 h, cells were co-transfected with 200 ng G6-TK-luc reporter
plasmid, 200 ng pCS2+ plasmids expressing GAL4DBD-HA-SUZ12 (WT
or mutant) or GAL4DBD-HA control protein, and 100 ng pCMV-β-
galactosidase vector. Cells were harvested 48 h post transfection. The
luciferase activity was then measured using the Luciferase Assay Sys-
tem kit (Promega, Cat No. E4030). Luciferase signals were normalized
by β- galactosidase activity using the β-galactosidase Enzyme Assay
System (Promega, Cat No. E2000). Western blot using the anti-HA
antibody was performed to compare the GAL4-HA-SUZ12 expression
level. GAPDH or Tubulin served as the protein loading control.
ChIP-qPCR
mESCs were crosslinked with 1% formaldehyde for 10 min at room
temperature. Formaldehyde was quenched with 0.125 M glycine, and
cells were washed twice with ice-cold PBS. Cell lysates were prepared
with Farnham lysis buffer (5 mM PIPES pH 8.0, 85 mM KCl, 0.5% NP40,
1 mM DTT and 1× protease inhibitor cocktail) to collect nuclei. Nuclei
were resuspended with lysis buffer (50 mM Tris-HCl pH 7.9, 10 mM
EDTA, 1% SDS, 1 mM DTT, and 1× protease inhibitor cocktail), and
chromatin was sheared to an average size of 200–600 bp using the
Covaris M220 Focused Ultrasonicator. The sheared chromatin was
diluted 10-fold with ChIP dilution buffer (20 mM Tris-HCl pH 7.9, 2 mM
EDTA, 150 mM NaCl, 0.5% Triton X-100, 1 mM DTT and 1× protease
inhibitor cocktail). The chromatin solution was clarified by centrifuga-
tion at 15,000 g at 4 °C for 10 min. 20 μg of chromatin was used for
H3K27me3-ChIP and 50 μg for FLAG ChIP. Chromatin was incubated
with 5 μg of antibody overnight at 4 °C with rotation and then 80 μl of
Nature Communications |
(2022) 13:6781
12
Article
https://doi.org/10.1038/s41467-022-34431-1
Protein A (Invitrogen, Cat No. 10002D) or Protein G (Invitrogen, Cat No.
10004D) Dynabeads were added to the antibody-chromatin complex.
After incubation at 4 °C for 2 h, beads were sequentially washed with low
salt (20 mM Tris-HCl pH 8.0, 2 mM EDTA, 1% Triton X-100, 0.1% SDS,
150 mM NaCl), high salt (20 mM Tris-HCl pH 8.0, 2 mM EDTA, 1% Triton
X-100, 0.1% SDS, 500 mM NaCl), LiCl (10 mM Tris-HCl pH 8.0, 1 mM
EDTA, 1% NP40, 1% sodium deoxycholate, 250 mM LiCl), and TE (20 mM
Tris-HCl pH 8.0, 1 mM EDTA) wash buffers. All washes were carried out at
4 °C for 10 min with rotation. The immunoprecipitated chromatin was
eluted with elution buffer (1% SDS and 100 mM NaHCO3). To reverse the
crosslinks, samples were incubated in a 65 °C water bath for 8–12 h.
RNase A and proteinase K treatment were performed before phenol:-
chloroform:isoamyl alcohol (25:24:1) extraction. Quantitative PCR at
specific loci was performed to analyze the enrichment of FLAG-SUZ12
and H3K27me3. Primers used for qPCR are listed in Supplementary
Table 3.
EB formation and replating assay
mESCs were induced to differentiate to EBs in hanging drops. Trypsi-
nized cells were resuspended in EB differentiation medium (DMEM,
15% FBS, 1× MEM-NEAA, 50 μM BME, 1× sodium pyruvate, 1× Pen/
Strep), and 30 μl droplets of the suspension (300 cells/drop) were
deposited on the lid of a 15 cm petri dish (120 drops/lid) for 48 h. Each
culture plate was filled with 15 ml of 1× PBS. The EBs were then trans-
ferred to uncoated 10 cm Petri dishes and cultured on an orbital shaker
at 50 rpm. EBs were harvested on day 4 and dissociated with trypsin to
form single-cell suspensions. The cell suspensions were seeded in the
2i ES cell medium at a density of 30,000 cells/ml in 12-well plates and
incubated for 5 days. Culture medium was changed every day. Cell
colonies were stained using the Stemgent AP staining Kit II (Stemgent,
Cat No. 00-0055) following the manufacturer’s protocol. Experiments
were performed in three replicates. Stained colonies were quantified in
ImageJ, and statistics were generated in GraphPad Prism.
Reporting summary
Further information on research design is available in the Nature
Research Reporting Summary linked to this article.
Data availability
The data that support this study are available from the corresponding
author upon reasonable request. The crystal structure described in this
study has been deposited in the Protein Data Bank under the accession
number 7TD5. The LC-MS/MS data files have been deposited to the Pro-
teomeXchange Consortium (http://proteomecentral.proteomexchange.
org) via the MassIVE partner repository with the dataset identifier
MSV000088683. Source Data are provided with this paper.
References
1. Margueron, R. & Reinberg, D. The Polycomb complex PRC2 and its
mark in life. Nature 469, 343–349 (2011).
2. Comet, I., Riising, E. M., Leblanc, B. & Helin, K. Maintaining cell
identity: PRC2-mediated regulation of transcription and cancer.
Nat. Rev. Cancer 16, 803–810 (2016).
3. Cao, R. et al. Role of histone H3 lysine 27 methylation in Polycomb-
group silencing. Science 298, 1039–1043 (2002).
4. Czermin, B. et al. Drosophila enhancer of Zeste/ESC complexes
have a histone H3 methyltransferase activity that marks chromo-
somal Polycomb sites. Cell 111, 185–196 (2002).
Kuzmichev, A., Nishioka, K., Erdjument-Bromage, H., Tempst, P. &
Reinberg, D. Histone methyltransferase activity associated with a
human multiprotein complex containing the Enhancer of Zeste
protein. Genes Dev. 16, 2893–2905 (2002).
5.
6. Muller, J. et al. Histone methyltransferase activity of a Drosophila
Polycomb group repressor complex. Cell 111, 197–208
(2002).
7.
8.
9.
Schwartz, Y. B. & Pirrotta, V. A new world of Polycombs: unexpected
partnerships and emerging functions. Nat. Rev. Genet 14,
853–864 (2013).
Jiao, L. & Liu, X. Structural basis of histone H3K27 trimethylation by
an active polycomb repressive complex 2. Science 350,
aac4383 (2015).
Justin, N. et al. Structural basis of oncogenic histone H3K27M
inhibition of human polycomb repressive complex 2. Nat. Commun.
7, 11316 (2016).
10. Chen, S., Jiao, L., Shubbar, M., Yang, X. & Liu, X. Unique structural
platforms of Suz12 dictate distinct classes of PRC2 for chromatin
binding. Mol. Cell 69, 840–852 e5 (2018).
11. Chen, S., Jiao, L., Liu, X., Yang, X. & Liu, X. A dimeric structural
scaffold for PRC2-PCL targeting to CpG island chromatin. Mol. Cell
77, 1265–1278.e7 (2020).
12. Kasinath, V. et al. Structures of human PRC2 with its cofactors
AEBP2 and JARID2. Science 359, 940–944 (2018).
13. Liu, X. A structural perspective on gene repression by Polycomb
Repressive Complex 2. Subcell. Biochem. 96, 519–562 (2021).
14. Margueron, R. et al. Role of the polycomb protein EED in the pro-
pagation of repressive histone marks. Nature 461, 762–767 (2009).
15. Sanulli, S. et al. Jarid2 methylation via the PRC2 complex regulates
H3K27me3 deposition during cell differentiation. Mol. Cell 57,
769–783 (2015).
16. Yuan, W. et al. Dense chromatin activates Polycomb repressive
complex 2 to regulate H3 lysine 27 methylation. Science 337,
971–975 (2012).
17. Sneeringer, C. J. et al. Coordinated activities of wild-type plus
mutant EZH2 drive tumor-associated hypertrimethylation of lysine
27 on histone H3 (H3K27) in human B-cell lymphomas. Proc. Natl
Acad. Sci. USA 107, 20980–20985 (2010).
18. McCabe, M. T. et al. Mutation of A677 in histone methyltransferase
EZH2 in human B-cell lymphoma promotes hypertrimethylation of
histone H3 on lysine 27 (H3K27). Proc. Natl Acad. Sci. USA 109,
2989–2994 (2012).
19. Lewis, P. W. et al. Inhibition of PRC2 activity by a gain-of-function H3
mutation found in pediatric glioblastoma. Science 340,
857–861 (2013).
20. Diehl, K. L. et al. PRC2 engages a bivalent H3K27M-H3K27me3
dinucleosome inhibitor. Proc. Natl Acad. Sci. USA 116,
22152–22157 (2019).
21. Stafford, J. M. et al. Multiple modes of PRC2 inhibition elicit global
chromatin alterations in H3K27M pediatric glioma. Sci. Adv. 4,
eaau5935 (2018).
22. Ragazzini, R. et al. EZHIP constrains Polycomb Repressive Complex
2 activity in germ cells. Nat. Commun. 10, 3858 (2019).
23. Jain, S. U. et al. PFA ependymoma-associated protein EZHIP inhibits
PRC2 activity through a H3 K27M-like mechanism. Nat. Commun.
10, 2146 (2019).
24. Hubner, J. M. et al. EZHIP/CXorf67 mimics K27M mutated onco-
histones and functions as an intrinsic inhibitor of PRC2 function in
aggressive posterior fossa ependymoma. Neuro Oncol. 21,
878–889 (2019).
25. Piunti, A. et al. CATACOMB: An endogenous inducible gene that
antagonizes H3K27 methylation activity of Polycomb repressive
complex 2 via an H3K27M-like mechanism. Sci. Adv. 5,
eaax2887 (2019).
26. Grijzenhout, A. et al. Functional analysis of AEBP2, a PRC2 Polycomb
protein, reveals a Trithorax phenotype in embryonic development
and in ESCs. Development 143, 2716–2723 (2016).
27. Conway, E. et al. A family of vertebrate-specific polycombs enco-
ded by the LCOR/LCORL genes balance PRC2 subtype activities.
Mol. Cell 70, 408–421.e8 (2018).
28. Hauri, S. et al. A high-density map for navigating the human Poly-
comb complexome. Cell Rep. 17, 583–595 (2016).
Nature Communications |
(2022) 13:6781
13
Article
https://doi.org/10.1038/s41467-022-34431-1
29. Hojfeldt, J. W. et al. Non-core subunits of the PRC2 complex are
collectively required for its target-site specificity. Mol. Cell 76,
423–436.e3 (2019).
30. Healy, E. et al. PRC2.1 and PRC2.2 synergize to Coordinate H3K27
Trimethylation. Mol. Cell 76, 437–452.e6 (2019).
52. Blom, N., Sicheritz-Ponten, T., Gupta, R., Gammeltoft, S. & Brunak,
S. Prediction of post-translational glycosylation and phosphoryla-
tion of proteins from the amino acid sequence. Proteomics 4,
1633–1649 (2004).
53. Amanchy, R. et al. A curated compendium of phosphorylation
31. Youmans, D. T., Gooding, A. R., Dowell, R. D. & Cech, T. R. Com-
petition between PRC2.1 and 2.2 subcomplexes regulates PRC2
chromatin occupancy in human stem cells. Mol. Cell 81,
488–501.e9 (2021).
32. Yang, Y. & Li, G. Post-translational modifications of PRC2: signals
directing its activity. Epigenetics Chromatin 13, 47 (2020).
33. Cha, T. L. et al. Akt-mediated phosphorylation of EZH2 suppresses
methylation of lysine 27 in histone H3. Science 310,
306–H10 (2005).
34. Chen, S. et al. Cyclin-dependent kinases regulate epigenetic gene
silencing through phosphorylation of EZH2. Nat. Cell Biol. 12,
1108–1114 (2010).
35. Kaneko, S. et al. Phosphorylation of the PRC2 component Ezh2 is
cell cycle-regulated and up-regulates its binding to ncRNA. Genes
Dev. 24, 2615–2620 (2010).
motifs. Nat. Biotechnol. 25, 285–286 (2007).
54. Ketel, C. S. et al. Subunit contributions to histone methyltransferase
activities of fly and worm polycomb group complexes. Mol. Cell
Biol. 25, 6857–6868 (2005).
55. Rai, A. N. et al. Elements of the polycomb repressor SU(Z)12 needed
for histone H3-K27 methylation, the interface with E(Z), and in vivo
function. Mol. Cell Biol. 33, 4844–4856 (2013).
56. Brooun, A. et al. Polycomb repressive complex 2 structure with
inhibitor reveals a mechanism of activation and drug resistance.
Nat. Commun. 7, 11384 (2016).
57. Tate, J. G. et al. COSMIC: the Catalogue Of Somatic Mutations In
Cancer. Nucleic Acids Res. 47, D941–D947 (2019).
58. Jiao, L. et al. A partially disordered region connects gene repression
and activation functions of EZH2. Proc. Natl Acad. Sci. USA 117,
16992–17002 (2020).
36. Wan, L. et al. Phosphorylation of EZH2 by AMPK suppresses PRC2
59. Chamberlain, S. J., Yee, D. & Magnuson, T. Polycomb repressive
methyltransferase activity and oncogenic function. Mol. Cell 69,
279–291.e5 (2018).
complex 2 is dispensable for maintenance of embryonic stem cell
pluripotency. Stem Cells 26, 1496–1505 (2008).
37. Zhang, H. et al. PLK1 and HOTAIR accelerate proteasomal degra-
dation of SUZ12 and ZNF198 during hepatitis B virus-induced liver
carcinogenesis. Cancer Res. 75, 2363–2374 (2015).
60. Pasini, D., Bracken, A. P., Hansen, J. B., Capillo, M. & Helin, K. The
polycomb group protein Suz12 is required for embryonic stem cell
differentiation. Mol. Cell Biol. 27, 3769–3779 (2007).
38. Guerra, B. et al. CK2: a protein kinase in need of control. Pharm.
61. Hojfeldt, J. W. et al. Accurate H3K27 methylation can be established
Ther. 82, 303–313 (1999).
39. Litchfield, D. W. Protein kinase CK2: structure, regulation and role in
cellular decisions of life and death. Biochem J. 369, 1–15
(2003).
40. Gao, Z. et al. PCGF homologs, CBX proteins, and RYBP define
functionally distinct PRC1 family complexes. Mol. Cell 45,
344–356 (2012).
41. Gao, Z. et al. An AUTS2-Polycomb complex activates gene
expression in the CNS. Nature 516, 349–354 (2014).
42. Blackledge, N. P. et al. Variant PRC1 complex-dependent H2A ubi-
quitylation drives PRC2 recruitment and polycomb domain forma-
tion. Cell 157, 1445–1459 (2014).
43. Cooper, S. et al. Targeting polycomb to pericentric hetero-
chromatin in embryonic stem cells reveals a role for H2AK119u1 in
PRC2 recruitment. Cell Rep. 7, 1456–1470 (2014).
44. Kalb, R. et al. Histone H2A monoubiquitination promotes histone H3
methylation in Polycomb repression. Nat. Struct. Mol. Biol. 21,
569–571 (2014).
45. Cooper, S. et al. Jarid2 binds mono-ubiquitylated H2A lysine 119 to
mediate crosstalk between Polycomb complexes PRC1 and PRC2.
Nat. Commun. 7, 13661 (2016).
46. Blackledge, N. P. et al. PRC1 catalytic activity is central to Polycomb
system function. Mol. Cell 77, 857–874.e9 (2020).
47. Tamburri, S. et al. Histone H2AK119 mono-ubiquitination is essential
for Polycomb-mediated transcriptional repression. Mol. Cell 77,
840–856.e5 (2020).
48. Kasinath, V. et al. JARID2 and AEBP2 regulate PRC2 in the presence
of H2AK119ub1 and other histone modifications. Science 371,
eabc3393 (2021).
49. St-Denis, N. A. & Litchfield, D. W. Protein kinase CK2 in health and
disease: From birth to death: the role of protein kinase CK2 in the
regulation of cell proliferation and survival. Cell Mol. Life Sci. 66,
1817–1829 (2009).
50. Borgo, C., D’Amore, C., Sarno, S., Salvi, M. & Ruzzene, M. Protein
kinase CK2: a potential therapeutic target for diverse human dis-
eases. Signal Transduct. Target Ther. 6, 183 (2021).
51. Hornbeck, P. V. et al. PhosphoSitePlus, 2014: mutations, PTMs and
recalibrations. Nucleic Acids Res. 43, D512–D520 (2015).
de novo by SUZ12-directed PRC2. Nat. Struct. Mol. Biol. 25,
225–232 (2018).
62. Ying, Q. L. et al. The ground state of embryonic stem cell self-
renewal. Nature 453, 519–523 (2008).
63. Silva, J. et al. Promotion of reprogramming to ground state plur-
ipotency by signal inhibition. PLoS Biol. 6, e253 (2008).
64. Miller, S. A., Damle, M., Kim, J. & Kingston, R. E. Full methylation of
H3K27 by PRC2 is dispensable for initial embryoid body formation
but required to maintain differentiated cell identity. Development
148, dev196329 (2021).
65. Ducker, G. S. & Rabinowitz, J. D. One-carbon metabolism in health
and disease. Cell Metab. 25, 27–42 (2017).
66. Rusin, S. F., Adamo, M. E. & Kettenbach, A. N. Identification of
candidate casein kinase 2 substrates in mitosis by quantitative
phosphoproteomics. Front. Cell Dev. Biol. 5, 97 (2017).
67. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data col-
lected in oscillation mode. Methods Enzymol. 276, 307–326 (1997).
68. Winn, M. D. et al. Overview of the CCP4 suite and current devel-
opments. Acta Crystallogr D. Biol. Crystallogr. 67, 235–242 (2011).
69. McCoy, A. J. et al. Phaser crystallographic software. J. Appl Crys-
tallogr. 40, 658–674 (2007).
70. Adams, P. D. et al. PHENIX: a comprehensive Python-based system
for macromolecular structure solution. Acta Crystallogr D. Biol.
Crystallogr. 66, 213–221 (2010).
71. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. Refinement of mac-
romolecular structures by the maximum-likelihood method. Acta
Crystallogr D. Biol. Crystallogr. 53, 240–255 (1997).
72. Bricogne, G. et al. BUSTER version 2.10.4 (Global Phasing Ltd,
Cambridge, United Kingdom, 2017).
73. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and
development of Coot. Acta Crystallogr D. Biol. Crystallogr. 66,
486–501 (2010).
74. The PyMOL Molecular Graphics System, Version 1.8.6 Schrö-
dinger, LLC.
Acknowledgements
The SUZ12 KO mESC line was a gift from Dr Kristian Helin from Institute of
Cancer Research, London. The cDNAs of human PRC2 core components
Nature Communications |
(2022) 13:6781
14
Article
https://doi.org/10.1038/s41467-022-34431-1
were kindly provided by Dr Robert E. Kingston. The authors acknowl-
edge the UTSW Proteomics Core facility for assistance with the phos-
phopeptide LC-MS/MS experiments. This research was supported by
Welch Foundation research grant I-1790 and NIH grants GM121662 and
GM 136308 to Xin L. Xin L. is a W.W. Caruth, Jr, Scholar in Biomedical
Research. This research also received support from the Cecil H. and Ida
Green Center Training Program in Reproductive Biology Sciences
Research. L.G. was supported by American Heart Association Post-
doctoral Fellowship 19POST34450043. L.J. was supported by National
Natural Science Foundation of China grant 32071213. Results shown in
this report are derived from work performed at Argonne National
Laboratory, Structural Biology Center (SBC) at the Advanced Photon
Source. SBC-CAT is operated by UChicago Argonne, LLC, for the U.S.
Department of Energy, Office of Biological and Environmental Research
under contract DE-AC02-06CH11357. Use of the Stanford Synchrotron
Radiation Lightsource, SLAC National Accelerator Laboratory, is sup-
ported by the U.S. Department of Energy, Office of Science, Office of
Basic Energy Sciences under Contract No. DE-AC02-76SF00515. The
SSRL Structural Molecular Biology Program is supported by the DOE
Office of Biological and Environmental Research, and by the National
Institutes of Health, National Institute of General Medical Sciences
(including P41GM103393). The contents of this publication are solely the
responsibility of the authors and do not necessarily represent the official
views of NIGMS or NIH.
Author contributions
Xin L. conceived the study. L.G., Xiuli L., L.J., and Xin L. designed the
experiments. L.G., Xiuli L., and L.J. performed the experiments with
assistance from X.Y. A.L. analyzed the mass spectrometry data. L.G.,
Xiuli L. and Xin L. wrote the manuscript.
Additional information
Supplementary information The online version contains
supplementary material available at
https://doi.org/10.1038/s41467-022-34431-1.
Correspondence and requests for materials should be addressed to
Xin Liu.
Peer review information Nature Communications thanks the anon-
ymous reviewer(s) for their contribution to the peer review of this work.
Peer review reports are available.
Reprints and permissions information is available at
http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to jur-
isdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as
long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate if
changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons license and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.
Competing interests
The authors declare no competing interests.
© The Author(s) 2022
Nature Communications |
(2022) 13:6781
15
| null |
10.1088_1361-6463_ad0ac1.pdf
|
Data availability statement
All data that support the findings of this study are included
within the article (and any supplementary files).
|
Data availability statement All data that support the findings of this study are included within the article (and any supplementary files).
|
J. Phys. D: Appl. Phys. 57 (2024) 075101 (7pp)
Journal of Physics D: Applied Physics
https://doi.org/10.1088/1361-6463/ad0ac1
Enhanced performance of AlGaN-based
deep-UV LED by incorporating carrier
injection balanced modulation
layer synergistically with
polarization-regulating structures
Xun Hu1,2,3, Lijing Kong1,3, Pan Yang1, Na Gao1,2,∗, Kai Huang1, Shuping Li1,
Junyong Kang1,∗
and Rong Zhang1
1 Fujian Key Laboratory of Semiconductor Materials and Applications, College of Physical Science and
Technology, Xiamen University, Xiamen 361005, People’s Republic of China
2 Jiujiang Research Institute of Xiamen University, Jiujiang 332000, People’s Republic of China
E-mail: [email protected] and [email protected]
Received 11 July 2023, revised 22 October 2023
Accepted for publication 8 November 2023
Published 16 November 2023
Abstract
A comparable concentration of carriers injected and transported into the active region, that is,
balanced hole and electron injection, significantly affects the optoelectronic performance of
AlGaN-based deep ultraviolet light-emitting diodes (DUV LEDs). In this study, we introduce a
novel structure characterized by a carrier injection balanced modulation layer, incorporating a
polarization-regulating gradient p-AlGaN in a DUV LED. We conducted a systematic
examination of its impact on the carrier injection and transport processes. Theoretical
simulations demonstrate the mitigation of abrupt variations in Al content at the interface
between electron blocking layer/p-AlGaN and p-AlGaN/p-GaN within the valence bands.
Consequently, holes are more likely to be injected into the active region rather than
accumulating at these interfaces. Meanwhile, due to the reduced barrier height at the top of the
valence band, the holes were efficiently transported into the quantum well and confined with
comparable and balanced concentrations of electrons by suppressing overflow, thereby
promoting the radiative recombination rate. Compared with the conventional DUV LED, the
hole concentration and radiative recombination rate of the designed structure in the final
quantum well are significantly increased to 179.8% and 232.3%, respectively. The spontaneous
emission intensity achieves nearly twice at the same current injection density. Moreover, the
efficiency droop is significantly suppressed when operated at a gradually increasing current
density. This study presents a promising approach that can serve as a reference for achieving
high-efficiency AlGaN-based DUV LEDs.
Supplementary material for this article is available online
Keywords: DUV LED, carrier injection banlanced modulation layer, holes transport,
radiative remobination rate, efficiency droop
3 These authors contributed equally to this work.
∗
Authors to whom any correspondence should be addressed.
1361-6463/24/075101+7$33.00
1
© 2023 IOP Publishing Ltd
J. Phys. D: Appl. Phys. 57 (2024) 075101
X Hu et al
1. Introduction
In recent years, AlGaN-based deep ultraviolet light-emitting
diodes (DUV-LEDs) have garnered significant interest for
their potential applications in water and air purification, med-
ical treatment, and biochemistry [1]. Notably, they demon-
strate promising utility in rapid disinfection and steriliza-
tion processes, particularly in response to the widespread
COVID-19 pandemic [2, 3]. Due to the wide and tunable
direct bandgap (3.4–6.2 eV), high breakdown voltage, and
mercury-free non-toxic nature, the AlGaN semiconductor is
considered the preferred candidate for DUV LEDs. It rep-
resents an important research direction in the field of wide-
bandgap semiconductors [4, 5]. In addition, the AlGaN-based
DUV LEDs are highly desirable in other versatile applications,
such as solar-blind optical communication, marine antifoul-
ing, and display color conversion [6–8]. However, the external
quantum efficiency (EQE) of the contemporary AlGaN-based
DUV LEDs is below 10%, significantly hindering the large-
scale applications of AlGaN-based DUV LEDs [4, 9, 10].
A major constraint leading to the low EQE is the strong
piezoelectricity of the AlGaN semiconductor, which gener-
ates polarization fields and causes the segregation of electrons
and holes within the multiple quantum wells (MQWs). The
quantum confinement Stark effect (QCSE) further exacerbates
this issue by causing poor carrier injection and resulting in
low radiative recombination [11]. Therefore, charge carrier
confinement in quantum structures plays a crucial role in the
overlap of charge carriers against polarization fields and in
the operation of optoelectronic devices. On the one hand, the
MQWs are plagued by a significant number of electrons that
may overflow and escape the final quantum well. On the other
hand, low hole injection into the active region poses a mis-
match with the injected electrons, as the concentration of holes
is 1–2 orders of magnitude lower than that of electrons. The
unbalanced carrier injection, i.e., electrons and holes injec-
ted into the active region, results in a markedly low internal
quantum efficiency (IQE) [12]. Thus, it is necessary to regu-
late the carrier injection balance by promoting hole injection
and transport with a comparable concentration between elec-
trons and holes that are mainly confined to the active region,
to attain high-performance AlGaN-based DUV LEDs [13].
Recently, various strategies for controlling carrier injec-
tion have been proposed to improve the performance of
DUV LEDs, in particular, including numerous new structural
designs of the p-type electron blocking layer (EBL) [14]. Hu
et al, for example, have used the p-type superlattice EBL struc-
ture to slow down the electron injection and suppress the elec-
tron overflow, which improved the radiative recombination
rate in the MQWs [15]. Gu et al have designed an undoped
BAlN film instead of the conventional AlGaN EBL, reducing
the electron leakage and increasing the output light power of
the DUV LEDs [16]. Recent research on EBL-free AlGaN
UV LED has shown that the polarization-engineered structure
with Al content in each quantum barrier (QB) could signific-
antly reduce electron leakage and enhance the optical power
and wall-plug efficiency [17]. Furthermore, many researchers
focus on modulating carrier injection at the final QB/EBL
interface to simultaneously enhance the injection and trans-
port of holes into the MQWs while suppressing electron leak-
age. A recent study by Jamil et al has reported that the IQE
and radiative recombination rate were enhanced by using an
AlInN alloy as the final QB, revealing the suppression of elec-
tron leakage and facilitating the injection of holes into the
active region [18]. Another study has inserted an extremely
thin AlN layer between the final QB and EBL to improve
hole injection by intraband tunneling [19]. However, these Al-
rich p-type EBL or quantum structures are susceptible to indu-
cing polarization effects due to lattice mismatch and Mg dop-
ing diffusion, which is unfavorable for IQE enhancement of
AlGaN-based DUV LEDs [20–22]. Additionally, the complic-
ated insertion of new materials is challenging to control exper-
imentally. Thus, a design that accurately controls the carrier
injection and transport processes to achieve a balance in the
active region is urgently needed [14, 23]. Such a design can
significantly coordinate the injection and transport of holes in
the active region, while preventing electron overflow.
In this study, we propose a structure incorporating a car-
rier injection balanced modulation (CIBM) layer synergistic-
ally with a polarization-regulating gradient p-AlGaN layer that
enhances the injection and transportation of holes within the
active region and simultaneously mitigates electron overflow,
thereby achieving balanced carrier injection in the MQWs of
AlGaN-based DUV LEDs. The mechanism of the CIBM layer
on the injection and transport of holes and electrons for the
band structure, the effective barrier height, and the concentra-
tion distribution were systematically investigated. Moreover,
the radiative recombination rate and the overlap of wavefunc-
tions in the quantum well were quantitatively determined.
Theoretical results show the realization of comparable injected
control of holes and electrons, i.e., balanced carrier injection
into the active region, is expected to significantly promote the
radiative recombination rate and spontaneous emission intens-
ity of DUV LEDs.
2. Model and methods
Based on the SimuLED method, we identified three typical
structures of AlGaN-based DUV LEDs, which we have des-
ignated as LED-1, LED-2, and LED-3. LED-1 is a conven-
tional DUV LED structure with a peaked emission wavelength
of approximately 278 nm. As shown in figure 1(a), LED-1
consists of a sapphire substrate, an AlN buffer layer, a
500 nm thick n-Al0.6Ga0.4N layer, a four-period MQWs with
Al0.45Ga0.55N quantum well and Al0.6Ga0.4N QB thicknesses
of 3 and 8 nm, respectively; a 15 nm thick p-Al0.75Ga0.25N
EBL; and a 20 nm thick p-Al0.6Ga0.4N hole injection layer
from bottom to up. The top surface was covered with a 100 nm
thick p-GaN contact layer, and the n-type and p-type regions
were doped with Si (doping concentration: 1 × 1018 cm–3) and
Mg (doping concentration: 5 × 1019 cm–3), respectively.
Using LED-1 as the basis, LED-2 and LED-3 were
developed by modifying the interconnecting layers on both
sides of the EBL structure. Specifically, in LED-2, the hole
injection layer of p-type AlGaN was replaced with a gradient
2
J. Phys. D: Appl. Phys. 57 (2024) 075101
X Hu et al
Figure 1. Simulation design: (a) schematic diagram of a conventional LED-1 structure; (b) distributions of Al content with respect to depth
for LED-1, LED-2, and LED-3 in different regions, respectively.
AlGaN structure. The Al content gradually decreases from
0.75 to 0, in the same ratio as the thickness increases. LED-
3 incorporated a CIBM layer associated with the same p-type
gradient layer in LED-2. In LED-3, the CIBM layer was intro-
duced between the MQWs and EBL, exhibiting downward-
and upward-graded Al content in AlxGa1−xN layers. The Al
content decreased linearly from 0.6 to 0.5 and then increased
linearly to 0.6. Notably, the maximum linear difference of Al
content in AlGaN material achieves 10%. In addition, a sys-
tematic study of detailed structures by modifying the CIBM
geometry and material composition difference to 15%, 20%
and 25% were described in figure S1 in the supplementary file.
The distribution of Al content and the schematic component in
different areas of these three DUV LED structures were sim-
ulated numerically (figure 1(b)).
For the simulated parameters, the energy band offset ratio
between the conduction and valence bands for the AlGaN
material was assumed to be 0.7 and 0.3, respectively. The
Shockley–Read–Hall recombination lifetime was 10 ns, and
−1. The sim-
the Auger coefficient was set to 1 × 10–30 cm6 s
ulated operating temperature of the devices was at 300 K,
and the electron and hole mobility rates achieved 100 and
−1, respectively. Further details on the paramet-
10 cm2 V
ers and equations used in these simulation models can be found
in the [24, 25].
−1 s
3. Results and discussions
To probe the effect of charge carrier injection and trans-
port, the band structures of the three DUV LEDs were first
analyzed. Figures 2(a) and (b) show the band structures of
−2. An
LED-1 and LED-2 at a current density of 130 A cm
abruptness in the valance bands of LED-1 can be seen at
the interfaces of EBL/p-AlGaN and p-AlGaN/p-GaN, respect-
ively. However, the abruptness in the valance bands mitig-
ates in LED-2. LED-1 exhibits a significant variation in Al
content between the EBL/p-AlGaN and p-AlGaN/p-GaN lay-
ers, resulting in abrupt changes at both interfaces. As a res-
ult, holes were blocked, trapped, and could not be efficiently
injected into the MQWs. In the p-AlGaN structure with a
gradient of Al content, the holes at the EBL/p-AlGaN and
p-AlGaN/p-GaN interfaces could be injected into the act-
ive region by mitigating the abruptness of the valence bands
in LED-2.
On the basis of band structures shown in figure 2, the poten-
tial difference was calculated between the quasi-Fermi energy
level and the energy band to determine the effective barrier
heights, which affect the transport behavior of charge carri-
ers towards the active region. This analysis was particularly
important for examining the transport of electrons and holes
at the interface between the final quantum well and the CIBM
layer. In figure 2(b), the effective barrier heights in the con-
duction and valence bands of LED-2 are equal to 14 meV and
335 meV for electrons and holes, respectively. It is possible
to increase the non-radiative recombination due to electron
overflow. Meanwhile, holes cannot be efficiently injected and
transported into the active region. To overcome this issue, we
proposed the structure LED-3, which featured a CIBM layer
that gradually controlled the Al content in the active region.
Interestingly, the effective barrier height for holes decreases
to 228 meV, while the barrier height for electrons increases to
83 meV, almost six times as much as in LED-2, as shown in
figure 2(d). Therefore, it was essential to optimize the thick-
ness of the gradient p-AlGaN layer, as shown in figure 2(c).
One can see that the EL intensity and the IQE peak increase
when the thickness of the p-AlGaN layer increases from 10 nm
to 20 nm and then decreases as the thickness increases to
60 nm. Because the series resistance becomes larger due to
the substantially increased film thickness of p-AlGaN, the
IQE and EL intensity values tend to be reduced [26]. Thus, a
gradient p-AlGaN layer with 20 nm thickness was chosen for
the following model to ensure the performance of the newly
developed structure.
3
J. Phys. D: Appl. Phys. 57 (2024) 075101
X Hu et al
Figure 2. (a) and (b) are the energy band distributions with respect to depth for LED-1 and LED-2, respectively; (c) the properties of
LED-2 varying with different gradient p-AlGaN layer thicknesses; (d) LED-3 energy band distribution with respect to depth when the
p-AlGaN layer optimized in (c) is adopted.
Figure 3. (a) and (b) show the concentration distributions of holes and electrons (relative displacement: 3 nm), respectively. The insets
display the concentration distributions in the active region. (c) Magnitude of electrostatic field in LED-1, LED-2 and LED-3. Figures (d) and
(e) show electron and hole current distribution at an injection current density of 130 A cm
−2.
Subsequently, the concentration of holes and electrons was
explored spatially versus the distance nearby the active region
in these three DUV LED structures (the current injection
−2). As can be seen in figure 3(a),
density fixed at 130 A cm
the hole concentration in LED-1 is extremely high at the
EBL/p-AlGaN and the p-AlGaN/p-GaN interfaces. This is
mainly due to the polarization-induced positive charges at
the interface [22]. However, the hole concentration at the
4
J. Phys. D: Appl. Phys. 57 (2024) 075101
X Hu et al
interfaces distributes comparably in LED-2 and LED-3, indic-
ating the reduced positive sheet polarization charges at the
interface and in agreement with the previous results. The con-
centration of electron leakage into the p-AlGaN layer is depic-
ted in figure 3(b). As can be seen, the electron concentration
in the p-type region is decreased by almost two orders of mag-
nitude in LED-2 and LED-3 compared to LED-1. Notably, the
electron concentration within the final quantum well, exhibits
a remarkable increase, as shown in the inset. When incorpor-
ating the CIBM structure into LED-3, the concentrations of
holes and electrons increase to 179.8% and 120% in compar-
ison with LED-1, which are approximately at the same mag-
nitude to achieve a balance. Moreover, the spatial concentra-
tion distributions of holes and electrons with the higher Al
content difference in the different incorporated CIBM struc-
tures are shown in figure S3. The hole concentration gradually
increases in the CIBM region, which can also be deduced from
the energy band diagrams displayed in figure S2. Although
numerous holes have accumulated in the CIBM region, con-
trolling a comparable carrier injection is unfavorable in the
active region by increasing the CIBM layer with a higher Al
content difference (see supplementary file).
To gain the underlying mechanism of carrier concentra-
tion distributions, we investigated the electrostatic field in the
MQWs and the p-type region for these DUV structures, as
illustrated in figure 3(c). The electrostatic field of LED-3 is
lower at the EBL/p-AlGaN and the p-AlGaN/p-GaN interfaces
than those in LED-1 and LED-2. This indicates the weakened
electric field enables more holes to be injected and transported
in the active region [27]. At the same time, a slight decrease of
the electrostatic field in the MQWs is also observed, for which
the carrier confinement is improved in LED-3. Moreover,
figures 3(d) and (e) show the current density distributions of
electrons and holes varying with the distance around the act-
ive region. As the distance approaches the p-type layer, the
electron leakage current diminishes while the hole injection
current increases. Incorporating the CIBM layer into LED-3
results in a significantly lower electron leakage current than
in LED-1 and LED-2. As previously mentioned, the reduced
electric field accelerates more holes to inject into the active
region, thereby producing the hole injection current in LED-3
as high as twice that in LED-1. In this way, the current density
of holes and electrons is injected symmetrically to achieve a
balance in LED-3.
Figure 4(a) presents the carrier recombination rate of the
three DUV LEDs to identify the confinement of carriers within
the active region with the CIBM layer into LED-3. Compared
to LED-1, the radiative recombination rates of LED-2 and
LED-3 are increased to 157.9% and 232.3% at a current
−2, respectively. When enough
injection density of 130 A cm
holes with a comparable number of electrons are injected
into the active region, the trapped carriers will recombine
for an increased radiative rate in the MQWs, especially in
the final quantum well closer to the CIBM layer and p-type
AlGaN, which agrees well with the above results. We con-
clude that the synergistic CIBM control of carrier injection
Figure 4. (a) Carrier recombination rate (relative displacement:
3 nm) at 130 A cm
−2; (b) J–V curves of the three LEDs.
and transport in LED-3 accelerates the efficient recombination
process toward the final quantum well. However, it is worth
mentioning that once the CIBM layer with Al content differ-
ence increased to 15%, the structure will not promote radiat-
ive recombination. As shown in figures S3 and S4 in the sup-
plementary file, we can see that a steeper CIBM layer with
higher Al content difference causes additional carrier confine-
ment and leads to the emission by complicated carrier recom-
bination processes. Therefore, the CIBM layer should not
be designed with a higher composition difference, adversely
affecting the balanced carrier injection in the active region.
Moreover, the J–V characteristics of the three DUV LEDs are
shown in figure 4(b). It is evident that the threshold voltages of
LED-1 and LED-3 almost overlap. When the biased forward
voltage increases to approximately 4.78 V, the correspond-
ing current densities of LED-1, LED-2 and LED-3 achieve
−2, respectively. Owing to the
14.71, 10.89 and 14.84 A cm
increased series resistance of the gradient p-AlGaN layer in
LED-2, the current density of this DUV LED structure is
5
J. Phys. D: Appl. Phys. 57 (2024) 075101
X Hu et al
Figure 5. (a) Variations in the internal quantum efficiencies of LED-1, LED-2, and LED-3 for different injection current densities; (b)
−2; (d) squared overlap
spontaneous emission spectra at 130 A cm
integrals of individual quantum well wave functions.
−2; (c) wave function distributions of electrons and holes at 130 A cm
lower than that of LED-1 and LED-3. It is found that elec-
trical characteristics are not hampered by adding the CIBM
layer synergistically with polarization-regulating p-AlGaN
structure.
Furthermore, we analyzed the efficiency to evaluate the
optoelectronic conversion performance of all LEDs.
In
figure 5(a), the maximum IQE values for LED-1, LED-2, and
LED-3 are 56.6%, 59.8%, and 62.4%, respectively. It is found
that the maximum IQE in LED-3 peaks at a relatively higher
current density and decreases slowly with an increasing cur-
− IQE130) /IQEpeak, the
rent density. According to (IQEpeak
efficiency droop ratios were determined as 66.7%, 52.3%, and
30.2% for LED-1, LED-2, and LED-3, respectively. We attrib-
ute the notable drop in efficiency to the increase of the active
region, for which the incorporated CIBM is designed as a sym-
metric structure constituting a linear decreased and increased
AlGaN layer. Combing the CIBM structure into LED-3 will
accommodate more holes and electrons with better confine-
ment. Consequently, the IQE peak shifts to a higher current
density, and the ratio of efficiency droop dramatically reduces.
Moreover, the spontaneous emission spectra of the three DUV
−2) in figure 5(b). In con-
LEDs were obtained (@130 A cm
trast to the conventional LED-1, the maximum intensities of
LED-2 and LED-3 rise to 154.6% and 194.8%, respectively.
Additionally, the peak wavelength of LED-3 is 276 nm, rep-
resenting a blue shift of approximately 2 nm due to the sup-
pression of the QCSE. When operating at a current density
−2, the injected charge carriers will mitigate the
of 130 A cm
polarization fields within the quantum well, increasing the
band gap.
As previously mentioned, the CIBM design with the gradi-
ent p-AlGaN structure significantly contributes to the mod-
ulation of the polarization field in DUV LEDs. To deeply
understand the effect, we focused on the wavefunction dis-
tribution profiles of the active region, particularly in the
final quantum well, which are highlighted in darker colors
in figure 5(c). The shadow region represents the normal-
ized overlapping of the electron and hole wavefunctions. It
is noticeable that the shadow region of the final quantum
well achieves the highest value among the three structures.
The square of the integrals of the overlapping wavefunc-
tions for each quantum well was calculated, indicating the
probability that the wavefunctions of the electrons and holes
overlap in space. In figure 5(d), the squared overlap integ-
rals for the final quantum well of LED-3 reach as high as
28.5%, approximately twice of the LED-1 (13.6%) and LED-
2 (14.7%). Our results suggest that the CIBM layer in the
gradient p-AlGaN structure of LED-3 substantially enhances
6
J. Phys. D: Appl. Phys. 57 (2024) 075101
X Hu et al
carrier injection and transport by controlling the polariza-
tion field, thereby improving the confinement of electrons
and holes in the active region. Consequently, the spatial
overlap of electron and hole wave functions is significantly
increased.
4. Conclusions
To summarize, we proposed a novel structure featuring a
CIBM layer in a polarization-regulating gradient p-AlGaN
DUV LED. The energy band simulations show that the injec-
tion and transport of holes toward the quantum well are excel-
lently supported by passing through the reduced barrier height
in the valence band. At the same time, the electrons leaking
from the final quantum well are suppressed in the conduction
band. Thus, a comparable concentration of the injected elec-
trons and holes in the active region is approximately achieved.
The balanced injection and confinement of injected carriers
within the MQWs are achieved, resulting in increased radi-
ative recombination and IQE. In contrast to the conventional
LED-1, the recombination rate of the active region in the
design of LED-3 increased to 232.3%. With the peaked IQE
at 62.4%, the IQE of LED-3 decreases more slowly under the
higher current injection density, and the efficiency droop ratio
−2. The spontaneous emission
is reduced to 30.2% at 130 A cm
intensity increased to almost twice that of LED-1. The pro-
posed structure provides a novel approach to control remark-
ably efficient carrier injection and transport for high-efficiency
AlGaN-based DUV LEDs.
Data availability statement
All data that support the findings of this study are included
within the article (and any supplementary files).
Acknowledgments
This work was partly supported by the National Key Research
and Development Program (2021YFB3600102), the National
Natural Science Foundation of China (62135013, 62234001,
62174141),
the Natural Science Foundation of Jiangxi
Province of China (20212BAB202027), and the Fundamental
Research Funds for the Central Universities (20720210028).
Conflict of interest
The authors declare that they have not known any compet-
ing financial interests or personal relationships that could have
influenced the work reported in this paper.
ORCID iD
Na Gao https://orcid.org/0000-0002-0630-1328
References
[1] Lu S Q, Jiang X J, Wang Y Z, Huang K, Gao N, Cai D J,
Zhou Y H, Yang C C, Kang J Y and Zhang R 2022
Nanoscale 14 653–62
[2] Raeiszadeh M and Adeli B 2020 ACS Photonics 7 2941–51
[3] Gerchman Y, Mamane H, Friedman N and Mandelboim M
2020 J. Photochem. Photobiol. B 212 112132
[4] Li J, Gao N, Cai D J, Lin W, Huang K, Li S P and Kang J Y
2021 Light Sci. Appl. 10 129
[5] Li D, Jiang K, Sun X J and Guo C L 2018 Adv. Opt. Photonics
10 43
[6] Li D et al 2022 Adv. Mater. 34 2109765
[7] Feng F, Liu Y B, Zhang K, Lin Y H, Chan K W, Kwok H S
and Liu Z J 2022 J. Soc. Inf. Disp. 30 556–66
[8] Wu M C and Chen I-T 2021 Adv. Photonics Res. 2 2100064
[9] Amano H et al 2020 J. Phys. D: Appl. Phys. 53 503001
[10] Liang S H and Sun W H 2022 Adv. Mater. Technol. 7 2101502
[11] Leroux M, Grandjean N, Laugt M, Massies J, Gil B,
Lefebvre P and Bigenwald P 1998 Phys. Rev. B 58 13371
[12] Cho J, Schubert E F and Kim J K 2013 Laser Photonics Rev.
7 408–21
[13] Kneissl M, Seong T Y, Han J and Amano H 2019 Nat.
Photonics 13 233–44
[14] Ren Z J, Yu H B, Liu Z L, Wang D H, Xing C, Zhang H C,
Huang C, Long S B and Sun H D 2020 J. Phys. D: Appl.
Phys. 53 072002
[15] Hu J H, Zhang J, Zhang Y, Zhang H X, Long H L, Chen Q,
Shan M C, Du S D, Dai J N and Chen C Q 2019 Nanoscale
Res. Lett. 14 347
[16] Gu W, Lu Y, Lin R Y, Guo W Z, Zhang Z H, Ryou J H,
Yan J C, Wang J X, Li J M and Li X H 2021 J. Phys. D:
Appl. Phys. 54 175104
[17] Velpula R T, Jain B, Velpula S, Nguyen H D and Hieu P T N
2020 Opt. Lett. 45 5125–8
[18] Jamil T, Usman M and Jamal H 2021 J. Appl. Phys.
129 213102
[19] Chu C S, Tian K K, Che J M, Shao H, Kou J Q, Zhang Y H,
Li Y, Wang M Y, Zhu Y H and Zhang Z H 2019 Opt.
Express 27 A620–28
[20] Kolbe T, Stellmach J, Mehnke F, Rothe M A, Kueller V,
Knauer A, Einfeldt S, Wernicke T, Weyers M and
Kneissl M 2016 Phys. Status Solidi a 213 210–4
[21] Hirayama H, Fujikawa S and Kamata N 2015 Electron.
Commun. Japan 98 1–8
[22] Simon J, Protasenko V, Lian C X, Xing H L and Jena D 2010
Science 327 60–64
[23] Shatalov M et al 2014 Semicond. Sci. Technol. 29 084001
[24] Vurgaftman I and Meyer J R 2003 J. Appl. Phys. 94 3675–96
[25] Piprek J 2004 Simulation of GaN-based light emitting devices
Int. Conf. on Simulation of Semiconductor Processes and
Devices 2004 pp 101–8
[26] Mohsen A, Qing L, Manoj S and William S W 2019
Nanotechnology 30 504001
[27] Bui H Q T, Dang H A, Doan T T, Velpula R T, Jain B,
Nguyen H P T and Nguyen H D 2021 Phys. Status Solidi a
218 2100003
7
| null |
10.1186_s12866-020-01881-w.pdf
|
Availability of data and materials
All data generated or analyzed during this study are included in this
published article [and its supplementary information files].
|
Availability of data and materials All data generated or analyzed during this study are included in this published article [and its supplementary information files].
|
Negrete-González et al. BMC Microbiology (2020) 20:213
https://doi.org/10.1186/s12866-020-01881-w
R E S E A R C H A R T I C L E
Open Access
High prevalence of t895 and t9364 spa
types of methicillin-resistant Staphylococcus
aureus in a tertiary-care hospital in Mexico:
different lineages of clonal complex 5
C. Negrete-González1, E. Turrubiartes-Martínez1,2, O. G. Galicia-Cruz3, D. E. Noyola4, G. Martínez-Aguilar5,
L. F. Pérez-González6, R. González-Amaro7 and P. Niño-Moreno1,8*
Abstract
Background: Staphylococcus aureus is a leading cause of broad-spectrum infections both in the community and
within healthcare settings. Methicillin-resistant Staphylococcus aureus (MRSA) has become a global public health
issue. The aim of this study was to examine the clinical and molecular characteristics of Staphylococcus aureus
isolates and to define the population structure and distribution of major MRSA clones isolated in a tertiary-care
hospital in Mexico.
Results: From April 2017 to April 2018, 191 Staphylococcus aureus isolates were collected. The frequency of MRSA
was 26.7%; these strains exhibited resistance to clindamycin (84.3%), erythromycin (86.2%), levofloxacin (80.3%), and
ciprofloxacin (86.3%). The majority of MRSA strains harbored the SCCmec type II (76.4%) and t895 (56.8%) and t9364
(11.7%) were the most common spa types in both hospital-associated MRSA and community-associated MRSA
isolates. ST5-MRSA-II-t895 (New York /Japan clone) and ST1011-MRSA-II-t9364 (New York /Japan-Mexican Variant
clone) were the most frequently identified clones. Furthermore, different lineages of Clonal Complexes 5 (85.4%)
and 8 (8.3%) were predominantly identified in this study.
Conclusion: Our study provides valuable information about the epidemiology of MRSA in a city of the central
region of Mexico, and this is the first report on the association between t895 and t9364 spa types and ST5 and
ST1011 lineages, respectively. These findings support the importance of permanent surveillance of MRSA aimed to
detect the evolutionary changes of the endemic clones and the emergence of new strains.
Keywords: Methicillin-resistant Staphylococcus aureus, Spa-typing, SCCmec type II, Clonal complex 5-ST1011, Spa
type t895, Spa type t9364, New York/Japan-Mexican variant clone
* Correspondence: [email protected]
1Sección de Genómica Médica, Centro de Investigación en Ciencias de la
Salud y Biomedicina, Universidad Autónoma de San Luis Potosí, San Luis
Potosí, Mexico
8Laboratorio de Genética, Facultad de Ciencias Químicas, Universidad
Autónoma de San Luis Potosí, San Luis Potosí, Mexico
Full list of author information is available at the end of the article
© The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if
changes were made. The images or other third party material in this article are included in the article's Creative Commons
licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons
licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain
permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the
data made available in this article, unless otherwise stated in a credit line to the data.
Negrete-González et al. BMC Microbiology (2020) 20:213
Page 2 of 11
Background
Staphylococcus aureus (S. aureus) is a commensal and a
pathogen in humans; approximately 30–50% of the popu-
lation are transient nasal carriers and 10–20% of individ-
uals are persistently colonized with this organism [1, 2].
Furthermore, colonization of the skin or mucosa with S.
aureus may increase the risk of invasive infections [3]. In
addition, S. aureus has been recognized as an extremely
versatile pathogen in humans, causing three major syn-
dromes: superficial
lesions, such as impetigo and skin
wound infections; deep and systemic infections, such as
osteomyelitis, endocarditis, pneumonia, and bacteremia;
and toxemic infections, such as toxic shock syndrome,
scalded skin syndrome, and food poisoning [4].
In 1961, one year after the introduction of methicillin
into medical practice,
the first methicillin-resistant
Staphylococcus aureus (MRSA) strain was identified;
methicillin resistance is mediated by the Staphylococcal
Cassette Chromosome mec (SCCmec) genetic element
[5]. This element includes the mec and ccr gene com-
plexes, which are flanked by three junkyard regions.
SCCmec is inserted into a unique site of the bacterial
chromosome by the action of Ccr proteins (encoded by
the ccr gene complex), which induce the specific recom-
bination between the attB sequence at the 3′ end and
the attS homologous sequence of SCCmec [6]. Variations
in the genetic content and structural organization of
these elements result in 13 different types and subtypes
of SCCmec [7–9].
An increasing number of MRSA strains were identified
initially in hospital centers (HA-MRSA) and, several
years later, cases of community-associated MRSA infec-
tions (CA-MRSA) were reported. In this regard, the epi-
demiology of MRSA infections has changed significantly
with the global emergence and expansion of CA-MRSA
strains [10].
The most frequently reported MRSA isolates belong to
major Clonal Complexes (CC) CC1, CC5, CC8, CC22,
CC30, CC45, and CC80 [11–13]. The most representa-
tive HA-MRSA clones are ST5-I/EMRSA-3/Cordobes-
Chilean and ST5-II/USA100/New York/Japan clones
(CC5), ST36-II/USA200 clone (CC30), ST45-II/USA600
clone (CC45), and ST239 III/ Brazilian/Hungarian clone
(CC8), while the most representative CA-MRSA are
ST1-IV/USA400 (CC1), ST5-IV/Pediatric clone (CC5),
(CC8),
ST8-IV/USA300
EMRSA-15 clone (CC22), ST30-IV/Southwest Pacific
clone (CC30), and ST80-IV/European clone (CC80) [12,
14]. The distribution of these clones varies in different
countries and regions of the world; in Mexico, ST5-II/
New York Japan and USA300 clones have been
described [15, 16].
and USA300-LA variant
According to the World Health Organization (WHO),
the
surveillance of MRSA is essential
for
global
identification of international transmission routes and
the subsequent development of effective prevention and
control strategies of this pathogen [17]. For this purpose,
molecular typing methods are a valuable tool for the
successful characterization of S. aureus isolates. In this
regard, Next Generation Sequencing (NGS) has been
used to identify S. aureus CCs and is considered the best
laboratory technique for identification of DNA diversity
in any organism. However, this methodology remains
technically demanding and requires robust software to
analyze the results [12]. Traditional typing methods in-
clude Multiple Locus Sequence Typing (MLST), Pulsed-
Field Gel Electrophoresis (PGFE), and spa-typing. MLST
is a great tool for evolutionary investigations and strain
identification and is based on the allelic profile of the
seven housekeeping genes. PFGE is based on the diges-
tion of DNA with restriction endonucleases and the de-
tection of the banding patterns. Although these two
methods show a high discriminatory power, they are la-
borious and require intra-laboratory standardization
protocols [12]. On the other hand, spa-typing is based
on the detection of sequence variation in repeats at the
X region of the staphylococcal protein A spa gene. This
typing technique exhibits high discriminatory power, has
a standardized nomenclature, is cost-effective, and shows
an excellent reproducibility. Spa-typing can be used for
the investigation of hospital outbreaks and to analyze
the evolution of S. aureus [18]. However, this method-
ology has some limitations, mainly in regions where a
particular clone or a small number of clones are en-
demic [11, 19].
The aim of this study was to estimate the prevalence
of MRSA and to analyze the molecular characteristics,
and antibiotic resistance profiles of CA- and HA-MRSA
genotypes in San Luis Potosi, a large city (approximately
1.1 million inhabitants) in the center of Mexico.
Results
Sample collection
S. aureus strains were obtained from one hundred
ninety-one patients from the emergency department
(n = 62), surgery (n = 47), intensive care unit (n = 31), in-
ternal medicine (n = 35), gynecology (n = 6), burn unit
(n = 2), and outpatient service (n = 8); patients in whom
samples were obtained in the outpatient service were
subsequently admitted to the hospital. The clinical speci-
mens were obtained from infections in skin and soft tis-
sues (n = 79), respiratory tract (n = 53), blood (n = 36),
bone and joints (n = 20), and cerebrospinal fluid (n = 3).
Seventy-seven percent (147 out of 191) of strains were
considered as HA and 23 % (44 out of 191) were
classified as CA.
One hundred fourteen patients were male
and
seventy-seven were female. Forty isolates were identified
Negrete-González et al. BMC Microbiology (2020) 20:213
Page 3 of 11
in children, and 151 in adults. The median age was 44
years. The mean length of hospital stay was 18.4 ± 19.5
days (range 1–105 days). Table 1 shows comorbidities,
surgical procedures, and history of hospital admission in
the last two years before infection of participants in the
study. The majority of patients (84.4%) were discharged
due to clinical improvement, 2% of patients were trans-
ferred to another hospital, 1.6% of patients requested
voluntary discharge, and 11.5% of patients infected with
S. aureus died.
Ten (45.5%) of the 22 patients who died had an MRSA
infection compared to 41 (24.3%) of the 169 patients
who survived (P = 0.034). Patients who died were also
older (mean 46.3 years) than those who survived (mean
35.2 years; P = 0.034). In contrast, there were no signifi-
cant differences in the prevalence of underlying condi-
tions (such as diabetes, malignancy, or renal disease)
between patients with a fatal outcome and those who
did not die (Additional file 1: Table S1).
Table 1 Clinical and demographic characteristics of the patients
with S. aureus infection included in the study
Sex
Male
Female
Age (years)
Infants 0–1
Children 2–10
Adolescents 11–17
Young adults 18–35
Adults 36–60
Seniors > 60
Length of stay (days)
Mean
SD
Range
Underlying disease
Diabetes mellitus
Hypertension
Renal disease
Neoplasms
Surgical procedures
Prior hospitalization
Hospital discharge
Clinical improvement
Death
Transfer
Voluntary discharge
N = 191
114
77
12
13
15
59
62
30
18.45
19.52
1–105
51
45
21
10
84
135
162
22
4
3
(%)
59.7
40.3
6.2
6.8
7.8
30.9
32.5
15.7
26.7
23.6
11
5.2
44
70.7
84.4
11.5
2.1
1.6
two of
Identification of MRSA strains
The mecA gene was detected in 51 out of 191 isolates
(26.7%), and 45 of them showed resistance to oxacil-
lin and were positive on cefoxitin-based screening.
The study was carried out between epidemiological
week (as defined by WHO) 14, 2017 and epidemio-
logical week 17, 2018. The weekly number of S. aur-
eus
infections varied between 1 and 8 cases. As
shown in Fig. 1, the largest number of cases was ob-
served at week 37 (eight,
them MRSA),
whereas in weeks 25, 35, 38, 50, 9 and 12, six cases
were identified. Moreover, the highest weekly num-
ber of MRSA cases was 4,
in weeks 38 and 41,
followed by weeks 20, 50, and 9 with 3 cases. Two
MRSA cases were identified in weeks 18, 19, 23, 37,
39, 45, 46, 48, 5, and 13, whereas a single case was
detected in weeks 14, 15, 17, 21, 25, 26, 29, 34, 35,
42, 51, 6, 8, 11, 12, and 14. No MRSA isolates were
observed during weeks 16, 22, 24, 27, 28, 30–33, 36,
40, 43, 44, 47, 49, 52, 1–4, 7, 10, 15–17. Three dif-
ferent periods of MRSA detection were identified
during the study. The first period occurred between
weeks 14 and 29 (2017), the second between weeks
34 and 51(2017), and the third between weeks 5 and
14 (2018) Fig. 1.
Antimicrobial susceptibility
The antibiotic resistance pattern differed significantly
between MRSA and MSSA isolates (P < 0.001 in most
cases). Thus, most MRSA strains showed resistance to
clindamycin (84.3%), erythromycin (86.2%), levofloxacin
(80.3%), and ciprofloxacin (86.3%), with low resistance to
gentamicin (13.7%) and rifampin (9.8%). In contrast,
MSSA strains showed minimal resistance to clindamycin
(7.1%), erythromycin (9.3%), ciprofloxacin (3.5%), levo-
floxacin (1.4%), and gentamicin (1.4%). None of MRSA
or MSSA strains were resistant to vancomycin, linezolid,
tigecycline,
trimethoprim/sulfamethoxazole, and tetra-
cycline (Table 2). Additional file 2: Table S2, shows the
minimum inhibitory concentration for each antibiotic.
SCCmec typing
Thirty-nine MRSA strains were classified as SCCmec
type II (four CA-MRSA, and thirty-five HA-MRSA) and
the SCCmec subtype IIb was identified in four strains
(HA-MRSA). Two isolates harbored SCCmec type IVc/E
and SCCmec type IVa was identified in one isolate (one
CA-MRSA and two HA-MRSA). In five isolates it was
not possible to identify the SCCmec types.
Spa-typing
MRSA isolates were classified in 11 different spa types,
including t895 (n = 29, 56.8%), t9364 (n = 6, 11.7%), t008
(n = 2, 3.9%), t003 (n = 3, 5.8%), t4229, t002, t012, t040,
Negrete-González et al. BMC Microbiology (2020) 20:213
Page 4 of 11
Fig. 1 Number of cases of S. aureus infection in each epidemiological week. Black bars correspond to MRSA isolates and grey bars to
MSSA strains
t304, t111, and t509 (n = 1). The spa type t895 was the
most common spa type among HA-MRSA and CA-
MRSA isolates. In one strain, we identified a spa type
not previously reported (spa type unknown). In three
isolates, the PCR employed by us did not amplify the
spa gene.
Table 2 Antibiotic resistant pattern of MSSA and MRSA isolates
MSSA N = 140 MRSA N = 51
P
Antibiotic
Benzylpenicillin
(n / %)
109 (77.8)a
Clindamycin
Erythromycin
Levofloxacin
Ciprofloxacin
Moxifloxacin
Rifampin
Gentamicin
Oxacillin
Vancomycin
Tetracyclin
Linezolid
Tigecycline
10 (7.1)
13 (9.3)
2 (1.4)
5 (3.5)
0 (0)
0 (0)
2 (1.4)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
Trimethoprim/sulfametoxazole
0 (0)
(n / %)
51 (100)
43 (84.3)
44 (86.2)
41 (80.3)
44 (86.3)
18 (35.2)
5 (9.8)
7 (13.7)
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001
0.001
0.002
45 (88.2)
< 0.001
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
NA
NA
NA
NA
NA
aThe penicillinase test was not performed in the 31 benzylpenicillin
susceptible MSSA isolates
Dendrogram of MRSA strains
A dendrogram was constructed to analyze the relation
among S. aureus strains based on their spa type (Fig. 2).
The spa type t895 (cluster 1) was identified in twenty-
nine isolates in patients from the surgery ward (n = 15),
emergency department (n = 4), internal medicine (n = 3),
intensive care unit (n = 1), and outpatient service (n = 1).
In children, t895 was identified in the pediatric ward
(n = 4) and neonatal intensive care unit (n = 1). Twenty-
seven isolates harbored the SCCmec type II, one isolate
harbored SCCmec type IIb, and in one strain the
SCCmec type was not identified. Twenty-seven isolates
of this cluster showed resistance to beta-lactams, fluoro-
quinolones (levofloxacin and ciprofloxacin), clindamycin,
and erythromycin. The B-796 strain was
rifampin-
resistant, and the B-766 strain was gentamicin-resistant.
The origin of MRSA strains in this cluster was predom-
inantly HA-MRSA (n = 27), and only two cases were
CA-MRSA (C-706 and C-708); the latter isolates were
identified in the surgery ward, in weeks 13 and 14. In
this cluster seven patients died.
The cluster 2 (t9364) included six isolates, three from
the surgery ward and one from the intensive care unit,
the burn unit and the internal medicine ward. Five iso-
lates in this cluster harbored the SCCmec type II, and
one the SCCmec type IIb; strains from both SCC types
showed resistance to fluoroquinolones, clindamycin, and
erythromycin. Four of these strains (A-747, A-786, B-
713, and B-773) were resistant to rifampin, and another
Negrete-González et al. BMC Microbiology (2020) 20:213
Page 5 of 11
Fig. 2 Dendrogram of MRSA strains. Dendrogram constructed using the unweighted pair group method with arithmetic average (UPGMA) based
on pairwise similarity values of spa types from 48 characterized MRSA strains. The scale corresponds to the percent of similarity. Blue branch
corresponds to cluster 1, red branch to cluster 2, green branch to cluster 3, and orange branch to cluster 4
(B-748) to gentamicin. The percent of similarity between
spa types t895 and t9364 was 99.5%. The A-792 (t002)
and B-706 (t509) isolates showed more than 98% of
similarity with t985 and t9364, and A-701 (t111) showed
97.6% of similarity to spa types t895, t9364 and t003.
The cluster 3 (t003) included three isolates, two of
them were detected in children (A-736 and C-703), and
one (B-770) from a patient in the surgery ward. In this
group, two isolates harbored SCCmec type IVc/E and
were resistant to beta-lactams. In the other isolate we
identified SCCmec type II.
The spa type t040 had a 92.7% similarity with the spa
types t895, t9364, t003 and t002, and had 91.5% similar-
ity with an unknown spa type that was identified in B-
700. This strain was only resistant to beta-lactams.
The cluster 4 had 90.8% similarity with the previously
mentioned spa types. This cluster included four isolates
with t008, t4229 and t304 spa types. Two patients from
the surgery and burn wards were infected with HA-
MRSA-t008, A-713 strain was isolated in the eighteen
week and harbored the SCCmec type II, and B-705 was
isolated in the thirty-nine week and harbored the
SCCmec type IVa.
Multilocus sequence typing
Spa types t895 and t9364, the major spa types identified
in this study, have not previously been associated to se-
quence types (ST). In order to analyze this, we selected
six isolates, and these were identified as ST1011 (n = 4)
and ST5 (n = 2). The association analysis of spa types of
clusters 3 to 4 with ST was performed in the Spa server.
Discussion
In the present study, we have assessed the epidemio-
logical characteristics of S. aureus isolates during one
year of intra-hospital surveillance and we analyzed the
molecular characteristics of MRSA strains. The most fre-
quent S. aureus infections were those affecting the skin
and soft tissues (n = 48, 25.1%) and bacteremia (n = 31,
17%). In contrast, the most frequent type of infection
caused by MRSA isolates was surgical site infection (n =
14, 27%).
The mortality associated with staphylococcal infections
in our study (11.5%) was lower than that previously re-
ported (approximately 15 years ago) in Mexico (50%)
[20]. MRSA infections were detected more frequently in
fatal cases than in patients who survived. Study partici-
pants who died were also older than those who survived.
Of note, while the presence of underlying diseases, his-
tory of surgical procedures, and health-care exposure
have previously been reported to be associated with fatal
infections [21, 22], we did not find significant differences
for these conditions between patients who died and
those who survived.
The recent epidemiology of S. aureus has focused on
in the
the increase and spread of MRSA strains
Negrete-González et al. BMC Microbiology (2020) 20:213
Page 6 of 11
In contrast,
healthcare setting and the community. In Denmark and
Scandinavian countries the prevalence of MRSA is less
than 1%.
in the east and southeast of
Europe, the prevalence of MRSA is greater than 30%
[22]. Peru has the highest reported prevalence in Latin
America (80%) [17]. In Mexico, there are a limited num-
ber of studies about MRSA and the available information
shows an increase in the prevalence of MRSA ranging
from 7 to 53% between 1989 and 2017 [15, 23–25]. In
our study, the prevalence of MRSA identified by molecu-
lar methods was higher than the prevalence identified by
the oxacillin resistance phenotype, which has been the
most used method in our country [17, 25, 26]. The
26.7% prevalence registered in our study is higher than
that reported, between January and June 2018, in 47 hos-
pitals in 20 states of Mexico (21.4%) [26], a study that
did not include information from the state of San Luis
Potosi. In a previous study, performed between 2005 and
2006 in six Mexican hospitals, the prevalence of MRSA
ranged from 1 to 43%. The highest prevalence was re-
corded at the Hospital Central Dr. Ignacio Morones
Prieto (HCIMP)
the prevalence of
[27]. In 11 years,
MRSA decreased to 26.7% in this hospital. This fact can
be explained by the infection control actions imple-
mented. A study that highlights the importance of intra-
hospital surveillance of MRSA was carried out between
January 1997 and May 2003 at the Pediatric Hospital of
the Centro Medico Nacional-Siglo XXI (Mexico City).
At this hospital, the annual frequency of methicillin re-
sistance ranged from 17 to 23% between 1997 and 2001,
and dramatically decreased in 2002 (4%) and 2003 (0%),
due to the intervention of the infection control commit-
tee at the end of 2001 [28].
Until February 2020 the Spa server has recorded 19,
255 different spa types [29]. According to a literature re-
view, in the last decade, the spa types t032/t008/t002 are
the most prevalent in Europe, t037/t002 in Asia, t008/
t002/t242 in America, t037/t084/t064 in Africa, and t020
in Australia [5]. Interestingly, in our study the preva-
lence of the spa types commonly described in America
was lower than expected, and we mainly detected the
spa types t895 and t9364.
Compared to other spa types, t895 has a low frequency
(0.01%); however, in the last two years its detection has
increased. Between 2017 and 2019, eight strains were re-
ported in USA and another in Germany, according to
the Spa Server [29]. Although data is scarce, previous re-
ports have associated the t895 spa type with CC5 [30].
In this regard, our data suggest an apparent association
between t895 spa type and the ST5 lineage of CC5. In
97% of the MRSA-t895 isolates, the SCCmec type II cas-
sette was identified; these molecular characteristics cor-
respond to the New York / Japan/ USA100 clone [15].
However, a limitation of this study is that the sequence
types (MLST) of several of the t895 strains were not de-
termined, which precluded a proper statistical analysis.
Of note, SCCmec type I, type II, and type IV have been
reported previously in MRSA-t895 strains [31].
The identification of t895 as the predominant spa
type in our study is of relevance, since this may have
implications. Of interest,
clinical and epidemiological
isolated in
characterization of 21 MRSA strains
Estado de Mexico (Mexico) in 2013 also showed t895
to be the predominant spa type, accounting for 76.2%
of isolates [31]. In a study conducted in the United
t895 spa type was predictive for the weak-
States,
biofilm producing phenotype, compared to t008 spa
type, which was
the
strong-biofilm producing phenotype [30].
identified as a predictor of
The spa type t9364 was registered in 2011 and corre-
sponded to a strain detected in Mexico, in a region out-
side of the state of San Luis Potosi [29]. In this regard,
our data describe, for the first time, the association be-
tween the t9364 spa type and the ST1011 sequence type.
Sequence type ST1011 was registered in the MLST data-
base in 2006; the first report of this ST included four
clinical MRSA isolates which differed from the sequence
type ST5 by the replacement of a nucleotide in the arcC
gene. Three of these MRSA ST1011 isolates were identi-
fied at HCIMP and one at General Hospital of Durango
[27]. Between 2008 and 2017, 14 isolates have been re-
ported with the sequence type ST1011 and the SCCmec
type II [15, 16]; all of these isolates have been identified
in Mexico. In 2017, ST1011-II was classified as the
New York / Japan clone because of its similarity to
ST5-II [15], and in a subsequent phylogenetic analysis
of CC5, it was observed that the clones identified in
Mexico were grouped in a subclade that was subdi-
vided into two subclades: ST5-II and ST1011-II. This
suggests that ST1011-II is not a New York/ Japan
clone, but it may be a variant of it that originated in
the late 1990s, the period when the CC30 was re-
placed in Mexico [16]. In all, available data suggests
that ST1011-II-t9364 may be a Mexican variant of
the New York / Japan clone which has increased in
prevalence in the last 11 years; however, more studies
are required to determine the differences with ST5-II-
t895 [16].
t012,
t509,
t003,
Other spa types identified with lower frequency in this
t040,
study corresponded to t111,
t4229, and t304. The spa type t003 has been related to
ST225 and ST270 sequence types, which are part of
CC5 and includes the Rhine Hesse, EMRSA-3 and New
York/Japan clones [32]. In addition, the spa types t012
and t040 have been identified in strains belonging to
CC30 and CC45, respectively. Furthermore,
the spa
types t4229 and t304 have been associated with ST8,
ST247, ST250, and ST254 sequence types, which belong
Negrete-González et al. BMC Microbiology (2020) 20:213
Page 7 of 11
to CC8 and include the USA300, ORSA IV and Archaic/
Iberian clones [33].
II, and this might explain, in part, the absence of tetra-
cycline resistance.
Diverse lineages of CC5 were predominant in our
study. These strains are characterized by bearing the
SCCmec I, II, and IV type cassettes with subtypes IVa,
IVc/E, and IVg. In this regard, different studies have
shown that most strains of this CC are multi-resistant,
mainly to fluoroquinolones, aminoglycosides, macro-
lides, lincosamides, and streptogramins (as we detected
in the isolates of our study), except for those that carry
the IVc/E cassette that only show resistance to beta-
lactams [15]. Moreover, to determine the relationship
between MRSA strains, we classified them into clusters
and analyzed their clinical and molecular characteristics.
In this analysis, clusters 1 and 2 were distributed in all
areas of the hospital within the three periods described
previously; in this regard, it is possible that these three
periods could be due to different introductions of clones
into the hospital or be a consequence of intra-hospital
transmission [34]. Although these two possibilities are
plausible, the last one could have resulted from transfer
of patients between different hospital wards during their
stay. Moreover, all strains grouped in these two clusters
were multi-resistant, and the highest number of deaths
was recorded in cluster 1. Furthermore, most strains in
cluster 3 were only resistant to beta-lactams and the
methicillin resistance phenotype was not identified. Fi-
nally, three out of four isolates in cluster 4 were identi-
fied in weeks 38 and 39, the epidemiological weeks with
the highest number of S. aureus infections. The use of
efficient and accurate epidemiological typing methods is
a requisite for monitoring the spread of epidemic clones
within and between hospitals. In this case, spa-typing
was a good tool for differentiate into CC5 lineages, be-
cause t895 and t9364 are not widespread spa types [19].
It is worth mentioning that if t895 and t9364 clones be-
come endemic and spread to multiple regions of Mexico,
the discriminating power of spa-typing to analyze noso-
comial transmission would decrease. To overcome this
limitation, recent studies suggest the use of a combin-
ation of different typing techniques to increase the abil-
ity to discriminate isolates [35].
In our study, all S. aureus strains were susceptible to
tetracycline, doxycycline and minocycline [36], and
trimethoprim-sulfamethoxazole. This observation is of
relevance, since these are alternatives for ambulatory
treatment of MRSA infections, such as skin and soft tis-
sue infections [37]. In contrast, tetracycline resistance
was reported in 6% of MSSA strains and 17% of MRSA
strains collected globally between 1997 and 2016 [38].
Resistance to this antibiotic in S. aureus is encoded by
the tetK and tetM genes [39], mainly detected in
SCCmec III, IV, and V MRSA strains [9, 40, 41]. The
majority of MRSA strains in our study had SCCmec type
Conclusions
Our data indicate that the most prevalent clones in all
areas of our hospital were ST5-MRSA-II-t895 (New
York /Japan clone) and ST1011-MRSA-II-t9364 (New
York/Japan-Mexican Variant clone), which belong to
CC5. In the HCIMP, the dominance of two CC5 lineages
is evident; however, MRSA isolates with molecular char-
acteristics consistent with Irish (weeks 18, 38 and 39),
USA300 (week 39) and Pediatric (week 13) clones, that
are considered epidemic MRSA clones, were identified.
We consider that this study further supports continuous
molecular monitoring of S. aureus infections as a valu-
able tool for epidemiological surveillance of MRSA since
it allows the evaluation of evolutionary changes of en-
demic clones and the introduction of emerging clones
that can cause hospital outbreaks. In addition, subse-
quent studies that assess the correlation between the
phenotype and the MRSA genotype are required, as well
as characterization of additional features of these clus-
ters, including virulence factors and resistance genes.
Methods
Sample collection
This cross-sectional study was conducted at HCIMP in
San Luis Potosi, Mexico, after approval by the Research
Committee [COFEPRIS 14 CI 24028083] and the Re-
search Ethics Committee of the HCIMP [CONBIOE-
TICA-24-CEI-001-20,160,427]. The registration number
was 29–17. Informed consent was obtained from all par-
ticipants or legal guardians.
The city of San Luis Potosi
is located in central
Mexico and is the capital of the state of San Luis Potosi.
HCIMP provides medical services to mid- and low-
income populations from all over the state; it has 250
beds and 32 beds in the intensive care unit (ICU).
From April 2017 to April 2018, a total of 191 non-
repeated S. aureus isolates were obtained from different
patients in all hospital wards. These isolates were identi-
fied by using the Vitek 2C (bioMérieux) system and con-
firmed by PCR amplification of the nuc gene.
Demographic and clinical data, including sex, age, date
of hospitalization, type of infection, date of isolation,
underlying disease and outcome of infection were col-
lected from medical records. Patients were classified in
groups according to their age, as follows: infants (0 to 1-
year-old), children (2 to 10 years old), adolescents (11 to
17 years old), young adults (18 to 35 years old), adults
(36 to 60 years old), and seniors (more than 60 years
old). An infection was considered as CA when symp-
toms presented < 48 h of a patient’s hospital admission,
in the absence of previous healthcare exposure, whereas
Negrete-González et al. BMC Microbiology (2020) 20:213
Page 8 of 11
an infection was considered as HA when occurred 48 h
after patient admission or when it was associated with
the following risk factors: hospitalization in an acute care
unit for at least 48 h in the last year, chemotherapy ad-
ministration, hemodialysis, wound care, enteric nutrition
or specialized nursing care 30 days before the infection
[15, 42, 43].
Antimicrobial susceptibility
Antimicrobial susceptibility testing was performed using
Vitek 2C (bioMérieux) and results were interpreted
using the Clinical and Laboratory Standards Institute
guidelines. Antibiotics tested included benzylpenicillin,
clindamycin, erythromycin,
levofloxacin, ciprofloxacin,
moxifloxacin, rifampin, gentamicin, vancomycin, tetra-
cycline, linezolid, oxacillin, and cefoxitin test [36].
DNA extraction
Three colonies of an overnight culture were suspended
in 100 μL of DNase free water and incubated at 94 °C for
5 min and − 70 °C for additional 5 min. Then, tubes were
centrifuged at 13,000 rpm for 5 min and the supernatant
was used as DNA template.
nuc and mecA identification
All S. aureus strains were screened by targeting the nuc
and mecA genes by multiplex PCR (Table 3) [44, 45].
PCR reactions were performed in a 25 μL volume con-
taining 1x of Buffer (200 mM Tris-HCl pH 8.4, 500 mM
KCl), 4 mM of MgCl2, 10 pmol of each primer, 200 μM
of each dNTP’s, 1 U of Taq DNA polymerase and bac-
terial genomic DNA. The PCR conditions were main-
initial denaturation
tained at 95 °C for 5 min for
followed by 30 cycles of 94 °C for 30 s, 60 °C for 30 s, and
72 °C for 30 s. Then, 20 μL aliquots of each sample were
subjected to electrophoresis on 2% agarose gel.
SCCmec typing
Identification of SCCmec types was performed by multi-
plex PCR using the genomic DNA from each MRSA iso-
late, according to a previously described method and
primers (Table 3) [46, 47]. DNA amplification was car-
ried out with a 2 min denaturation step at 94 °C,
followed by 30 cycles of 60 s at 94 °C for denaturation,
60 s at 55 °C for annealing, and 60 s at 72 °C for exten-
sion, and then 5 min at 72 °C for final extension. Then,
20 μL aliquots of each sample were subjected to electro-
phoresis on 2% agarose gel.
Spa-typing
The X region of the spa gene of each MRSA isolate was
amplified by PCR with the primers 1095F and 1517R, as
described previously (Table 3) [48]. The amplified prod-
ucts were sequenced, and the results were analyzed
44
45
48
Table 3 PCR primers used in this study
Gene
nuc
Primer
nuc-F
nuc-R
Primer sequence 5′➔3′
Reference
GCGATTGATGGTGATACGGTT
AGCCAAGCCTTGACGAACTAAAGC
mecA
mecA147-F GTGAAGATATACCAAGTGATT
mecA147-R ATGCGCTATAGATTGAAAGGAT
spa
1095-F
1517-R
AGACGATCCTTCGGTGAGC
GCTTTTGCAATGTCATTTACTG
SCCmec
Type II-F
CGTTGAAGATGATGAAGCG
46, 47
Type II-R
CGAAATCAATGGTTAATGGACC
Type-IIb-F
TAGCTTATGGTGCTTATGCG
Type-IIb-R
GTGCATGATTTCATTTGTGGC
Type-IVa-F
GCCTTATTCGAAGAAACCG
Type-IVa-R
CTACTCTTCTGAAAAGCGTCG
Type IVE-F
CAGATTCATCATTTCAAAGGC
Type IVE-R
AACAACTATTAGATAATTTCCG
Type IVc-F
CCTGAATCTAAAGAGATACACCG
Type IVc-R
GGTTATTTTCATAGTGAATCGC
arcC
aro
glp
gmk
pta
tpi
yqi
arcC-F
arcC-R
aro-F
aro-R
glp-F
glp-R
gmk-F
gmk-R
pta-F
pta-R
tpi-F
tpi-R
yqi-F
yqi-R
TTG ATT CAC CAG CGC GTA TTG TC 50
AGG TAT CTG CTT CAA TCA GCG
ATC GGA AAT CCT ATT TCA CAT TC
GGT GTT GTA TTA ATA ACG ATA TC
CTA GGA ACT GCA ATC TTA ATC C
TGG TAA AAT CGC ATG TCC AAT TC
ATC GTT TTA TCG GGA CCA TC
TCATTAACTACAACGTAATCGTA
GTTAAAATCGTATTACCTGAAGG
GACCCTTTTGTTGAAAAGCTTAA
TCGTTCATTCTGAACGTCGTGAA
TTTGCACCTTCTAACAATTGTAC
CAGCATACAGGACACCTATTGGC
CGTTGAGGAATCGATACTGGAAC
using the Ridom Staph Type software version 1.4
(Ridom, GmbH, Wurzburg, Germany [http://spa.ridom.
de/index.shtml]) to determine the repeat profile and the
spa type of each isolate [29, 49].
Dendrogram of MRSA strains
Dendrogram was constructed based on spa types data
using a temporary BioNumerics evaluation license from
Applied Maths (version 7.6, bioMérieux).
Multilocus sequence typing
MLST was performed on six MRSA strains of the spa
types t9364 (n = 4) and t895 (n = 2). Seven housekeeping
genes (arcC, aroE, glpF, gmk, pta, tpi, and yqiL) of S.
aureus were used for MLST typing (Table 3). PCRs were
carried out in 50 μl reaction volumes containing 10 ng of
Negrete-González et al. BMC Microbiology (2020) 20:213
Page 9 of 11
chromosomal DNA, 10 pmol of each primer, 1 U of Taq
DNA polymerase, 5 μl of 10x PCR buffer, and 200 μM
each of dNTPs. PCR was performed with an initial de-
naturation at 95 °C for 5 min, followed by 37 cycles of
denaturation at 95 °C for 30 s, annealing at 55 °C for 30
s, extension at 72 °C for 30 s, followed by a final exten-
sion step of 72 °C for 5 min [50]. After amplification, the
PCR products were purified and sequenced by dideoxy-
nucleotides method (3500 Genetic Analyzer, Applied
Biosystems). The consensus sequences were assembled,
and the allelic profile was matched using the MLST
database (https://pubmlst.org/saureus/).
Statistical analysis
Comparisons between groups was carried using Fisher’s
exact test or the chi-squared test (for categorical vari-
ables) and Student’s t test of Mann-Whitney U test (for
continuous variables) using Statistical Package for Social
Sciences software for Mac OS, version 25.0 (SPSS, IBM,
Inc., Chicago, IL, USA). P value < 0.05 were considered
statistically significant.
Supplementary information
Supplementary information accompanies this paper at https://doi.org/10.
1186/s12866-020-01881-w.
Additional file 1: Table S1. Demographic and clinical characteristics of
patients with Staphylococcus aureus infections who died or survived.
Table S1 shows the demographic and clinical characteristics of patients
with Staphylococcus aureus infections who died or survived.
Additional file 2: Table S2. Minimum Inhibitory Concentration (μg/mL)
data for the MRSA strains. Table S2 shows the MIC for each antibiotic for
the MRSA strains.
Abbreviations
S. aureus: Staphylococcus aureus; MRSA: Methicillin-resistant Staphylococcus
aureus; HA-MRSA: Healthcare-associated MRSA; CA-MRSA: Community-
associated MRSA; MSSA: Methicillin-sensible Staphylococcus aureus; CC: Clonal
complex; ST: Sequence type; WHO: World Health Organization;
MLST: Multiple Locus Sequence Typing; PFGE: Pulsed-Field Gel
Electrophoresis; NGS: Next Generation Sequencing
Acknowledgments
We thank to María Anita de Lira Torres, Andrés Flores Santos, and Laura
Cerda Ramos, the staff of the Microbiology Laboratory of the HCIMP, for their
support in the collection of strains.
We thank to Adriana Rodríguez Martínez and Miriam Briano Macias for their
valuable technical support in this project.
Authors’ contributions
CNG conceived the study, acquired clinical data and samples, performed the
experiments, interpreted results, and drafted the manuscript. ETM and PNM
co-designed and supervised the study and interpreted the results of experi-
ments. DEN and OGC analyzed and interpreted data. DEN and RGA critically
revised and edited the manuscript. GMA and LPG analized and interpreted
the patient data. All authors have read and approved the manuscript.
Funding
This work was supported by the Grant 142334 from CONACyT-Salud, Mexico
to PNM. CNG was a recipient of a scholarship 443025 from CONACyT,
Mexico. The funding agency had no role in the study design, sample
collection, data collection and analysis, decision to publish, or preparation of
the manuscript.
Availability of data and materials
All data generated or analyzed during this study are included in this
published article [and its supplementary information files].
Ethics approval and consent to participate
This study was approved by the Research Committee [COFEPRIS 14 CI 24
028 083] and the Research Ethics Committee [CONBIOETICA-24-CEI-001-
20160427] of the HCIMP. The registration number was 29–17.
Written informed consent was obtained from all participants or legal
guardians/parents for those under the age of 16 years.
Consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests.
Author details
1Sección de Genómica Médica, Centro de Investigación en Ciencias de la
Salud y Biomedicina, Universidad Autónoma de San Luis Potosí, San Luis
Potosí, Mexico. 2Laboratorio de Hematología, Facultad de Ciencias Químicas,
Universidad Autónoma de San Luis Potosí, San Luis Potosí, Mexico.
3Departamento de Farmacología, Facultad de Medicina, Universidad
Autónoma de San Luis Potosí, San Luis Potosí, Mexico. 4Departamento de
Microbiología, Facultad de Medicina, Universidad Autónoma de San Luis
Potosí, San Luis Potosí, Mexico. 5Unidad de Investigación Biomédica, Instituto
Mexicano del Seguro Social, Durango, Mexico. 6Hospital Central “Dr. Ignacio
Morones Prieto”, San Luis Potosí, Mexico. 7Sección de Medicina Molecular y
Traslacional, Centro de Investigación en Ciencias de la Salud y Biomedicina,
Universidad Autónoma de San Luis Potosí, San Luis Potosí, Mexico.
8Laboratorio de Genética, Facultad de Ciencias Químicas, Universidad
Autónoma de San Luis Potosí, San Luis Potosí, Mexico.
Received: 29 February 2020 Accepted: 26 June 2020
References
1.
Von Eiff C, Becker K, Machka K, Stammer H, Peters G. Nasal carriage as a
source of Staphylococcus aureus bacteremia. N Engl J Med. 2001;344(1):11–6.
https://doi.org/10.1056/NEJM200101043440102.
Noble WC, Valkenburg HA, Wolters CHL. Carriage of Staphylococcus aureus
in random samples of a normal population. Epidemiol Infect. 1967;65(4):
567–73. https://doi.org/10.1017/S002217240004609X.
2.
3. Wertheim HFL, Melles DC, Vos MC, van Leeuwen W, van Belkum A,
4.
5.
6.
Verbrugh HA, et al. The role of nasal carriage in Staphylococcus aureus
infections. Lancet Infect Dis. 2005;5(12):751–62. https://doi.org/10.1016/
S1473-3099(05)70295-4.
Jarraud S, Mougel C, Thioulouse J, Lina G, Meugnier H, Forey F, et al.
Relationships between Staphylococcus aureus genetic background, virulence
factors, agr groups (alleles), and human disease. Infect Immun. 2002;70(2):
631–41. https://doi.org/10.1128/iai.70.2.631-641.2002.
Asadollahi P, Farahani NN, Mirzaii M, Khoramrooz SS, van Belkum A,
Asadollahi K, et al. Distribution of the most prevalent spa types among
clinical isolates of methicillin-resistant and-susceptible Staphylococcus aureus
around the world: a review. Front Microbiol. 2018;9:1–16. https://doi.org/10.
3389/fmicb.2018.00163.
International Working Group on the Classification of Staphylococcal
Cassette Chromosome Elements (IWG-SCC). Classification of staphylococcal
cassette chromosome mec (SCCmec): guidelines for reporting novel
SCCmec elements. Antimicrob Agents Chemother. 2009;53(12):4961–7.
https://doi.org/10.1128/AAC.00579-09.
7. Wu Z, Li F, Liu D, Xue H, Zhao X. Novel type XII staphylococcal cassette
chromosome mec harboring a new cassette chromosome recombinase,
CcrC2. Antimicrob Agents Chemother. 2015;59(12):7597–601. https://doi.org/
10.1128/AAC.01692-15.
Baig S, Johannesen TB, Overballe-Petersen S, Larsen J, Larsen AR, Stegger M.
Novel SCCmec type XIII (9A) identified in an ST152 methicillin-resistant
8.
Negrete-González et al. BMC Microbiology (2020) 20:213
Page 10 of 11
15. Arias CA, Reyes J, Carvajal LP, Rincon S, Diaz L, Panesso D, et al. A
32.
9.
Staphylococcus aureus. Infect Genet Evol. 2018;61(January):74–6. https://doi.
org/10.1016/j.meegid.2018.03.013.
Liu J, Chen D, Peters BM, Li L, Li B, Xu Z, et al. Staphylococcal chromosomal
cassettes mec (SCCmec): a mobile genetic element in methicillin-resistant
Staphylococcus aureus. Microb Pathog. 2016;101:56–67. https://doi.org/10.
1016/j.micpath.2016.10.028.
10. Deurenberg RH, Stobberingh EE. The evolution of Staphylococcus
11.
12.
aureus. Infect Genet Evol. 2008;8(6):747–63. https://doi.org/10.1016/j.
meegid.2008.07.007.
Stefani S, Chung DR, Lindsay JA, Friedrich AW, Kearns AM, Westh H, et al.
Methicillin-resistant Staphylococcus aureus (MRSA): global epidemiology and
harmonisation of typing methods. Int J Antimicrob Agents. 2012;39(4):273–
82. https://doi.org/10.1016/j.ijantimicag.2011.09.030.
Lakhundi S, Zhang K. Methicillin-resistant Staphylococcus aureus: molecular
characterization, evolution, and epidemiology. Clin Microbiol Rev. 2018;
31(4):e00020–18. https://doi.org/10.1128/CMR.00020-18.
13. Boswihi SS, Udo EE, Monecke S, Mathew B, Noronha B, Verghese T, et al.
Emerging variants of methicillin-resistant Staphylococcus aureus genotypes
in Kuwait hospitals. PLoS One. 2018;13(4):1–13. https://doi.org/10.1371/
journal.pone.0195933.
Lee AS, De Lencastre H, Garau J, Kluytmans J, Malhotra-Kumar S, Peschel A,
et al. Methicillin-resistant Staphylococcus aureus. Nat Rev Dis Prim. 2018;4:
18033. https://doi.org/10.1038/nrdp.2018.33.
14.
prospective cohort multicenter study of molecular epidemiology and
phylogenomics of Staphylococcus aureus bacteremia in nine Latin American
countries. Antimicrob Agents Chemother. 2017;61(10). https://doi.org/10.
1128/AAC.00816-17.
16. Challagundla L, Reyes J, Rafiqullah I, Sordelli DO, Echaniz-Aviles G,
Velazquez-Meza ME, et al. Phylogenomic classification and the evolution of
Clonal complex 5 methicillin-resistant Staphylococcus aureus in the Western
Hemisphere. Front Microbiol. 2018;9(AUG):1–14. https://doi.org/10.3389/
fmicb.2018.01901.
17. WHO. Antimicrobial Resistance Global Report on Surveillance. World Health
Organization. 2014; Available from: http://apps.who.int/iris/bitstream/
handle/10665/112642/9789241564748_eng.pdf;jsessionid=FAB85B56C04
08C73DBF4ED0A293ECD23?sequence=1. Accessed 15 Nov 2019.
18. Mazi W, Sangal V, Sandstrom G, Saeed A, Yu J. Evaluation of spa-typing of
19.
methicillin-resistant Staphylococcus aureus using high-resolution melting
analysis. Int J Infect Dis. 2015;38:125–8. https://doi.org/10.1016/j.ijid.2015.05.
002.
Strommenger B, Braulke C, Heuck D, Schmidt C, Pasemann B, Nübel U, et al.
Spa typing of Staphylococcus aureus as a frontline tool in epidemiological
typing. J Clin Microbiol. 2008;46(2):574–81. https://doi.org/10.1128/JCM.
01599-07.
20. Velázquez-Meza ME. Surgimiento y diseminación de Staphylococcus aureus
meticilinorresistente Staphylococcus aureus methicillin-resistant: emergence
and dissemination. Salud Publica Mex. 2005;47(5):381–7. https://doi.org/10.
1590/s0036-36342005000500009.
21. Bello-Chavolla OY, Bahena-Lopez JP, Garciadiego-Fosass P, Volkow P, Garcia-
Horton A, Velazquez-Acosta C, et al. Bloodstream infection caused by S.
aureus in patients with cancer: a 10-year longitudinal single-center study.
Support Care Cancer. 2018;26(12):4057–65. https://doi.org/10.1007/s00520-
018-4275-1.
van Hal SJ, Jensen SO, Vaska VL, Espedido BA, Paterson DL, Gosbell IB.
Predictors of mortality in Staphylococcus aureus bacteremia. Clin Microbiol
Rev. 2012;25(2):362–86. https://doi.org/10.1128/CMR.05022-11.
22.
23. Alpuche-Aranda C, Avila-Figueroa C, Espinoza-De los Monteros L, Gómez-
Barreto D, Santos-Preciado JI. Antimicrobial sensitivity profile of
Staphylococcus aureus at a pediatric hospital: prevalence of resistance to
methicillin. Bol Med Hosp Infant Mex. 1989;46:700–4.
24. Macías Hernández AE, Medina Valdovinos H, Gaona Reyes AD, Barret JM,
Guerrero Martínez FJ, Ramírez Barba EJ, et al. Estafilococo resistente a
meticilina en un hospital general de León Guanajuato. Enferm Infecc
Microbiol. 1993;13:123–7.
Salvatierra-González R, Benguigui Y. Resistencia antimicrobiana en las
Américas. Magnitud del problema y su contención. Washington: OPS; 2000.
p. 268. (OPS/HCP/163/2000). ISBN 92 75 32319 5.
25.
26. Garza-González E, Morfín-Otero R, Mendoza-Olazarán S, Bocanegra-Ibarias P,
Flores-Treviño S, Rodríguez-Noriega E, et al. A snapshot of antimicrobial
resistance in Mexico. Results from 47 centers from 20 states during a six-
month period. PLoS One. 2019;14(3):1–13. https://doi.org/10.1371/journal.
pone.0209865.
27. Martínez-Aguilar G. Análisis de genotipos y de los tiempos de duplicación
de cepas de Staphylococcus aureus resistente a meticilina aisladas de
infecciones nosocomiales y adquiridas en la comunidad. Diss. PhD thesis.
México: Universidad de Colima; 2010.
28. Velazquez-Meza ME, Aires De Sousa M, Echaniz-Aviles G, Solórzano-Santos F,
Miranda-Novales G, Silva-Sanchez J, et al. Surveillance of methicillin-resistant
Staphylococcus aureus in a pediatric hospital in Mexico City during a 7-year
period (1997 to 2003): clonal evolution and impact of infection control. J
Clin Microbiol. 2004;42(8):3877–80. https://doi.org/10.1128/JCM.42.8.3877-
3880.2004.
29. Ridom GmbH. Ridom SpaServer. Würzburg, Germany. 2005 Available from:
30.
https://www.spaserver.ridom.de/ Accessed Feb 2020.
Luther M, Parente D, Caffrey A, Daffinee K, Lopes V, Martin E. Clinical and
Genetic Risk Factors for Biofilm-Forming Staphylococcus aureus. Antimicrob
Agents Chemother. 2018;62(5). https://doi.org/10.1128/AAC.02252-17.
31. Paniagua-Contreras GL, Monroy-Pérez E, Vaca-Paniagua F, et al.
Implementation of a novel in vitro model of infection of reconstituted
human epithelium for expression of virulence genes in methicillin-resistant
Staphylococcus aureus strains isolated from catheter-related infections in
Mexico. Ann Clin Microbiol Antimicrob. 2014;13(6). https://doi.org/10.1186/
1476-0711-13-6.
Engelthaler DM, Kelley E, Driebe EM, Bowers J, Eberhard CF, Trujillo J, et al.
Rapid and robust phylotyping of spa t003, a dominant MRSA clone in
Luxembourg and other European countries. BMC Infect Dis. 2013;13(1):339.
https://doi.org/10.1186/1471-2334-13-339.
33. Grundmann H, Aanensen DM, Van Den Wijngaard CC, Spratt BG, Harmsen
D, Friedrich AW, et al. Geographic distribution of Staphylococcus aureus
causing invasive infections in Europe: A molecular-epidemiological analysis.
PLoS Med. 2010;7(1). https://doi.org/10.1371/journal.pmed.1000215.
34. Huenger F, Klik S, Haefner H, Krizanovic V, Koch S, Lemmen SW. P1326
35.
MRSA spa-typing reveals a newly imported hospital endemic strain. Int J
Antimicro Ag. 2007;29:S367. https://doi.org/10.1016/s0924-8579(07)71166-4.
Kuhn G, Francioli P, Blanc DS. Double-locus sequence typing using clfB and
spa, a fast and simple method for epidemiological typing of methicillin-
resistant Staphylococcus aureus. J Clin Microbiol. 2007;45(1):54–62. https://
doi.org/10.1128/JCM.01457-06.
36. CLSI. Performance Standards for Antimicrobial Susceptibility Testing;
37.
Twenty-Fifth Informational Supplement. CLSI document M100-S25. Wayne,
PA: Clinical and Laboratory Standards Institute; 2015.
Liu C, Bayer A, Cosgrove SE, Daum RS, Fridkin SK, Gorwitz RJ, et al. Infectious
Diseases Society of America. Clinical practice guidelines by the Infectious
Diseases Society of America for the treatment of methicillin-resistant
Staphylococcus aureus infections in adults and children. Clin Infect Dis. 2011;
52(3):e18–55. https://doi.org/10.1093/cid/ciq146.
38. Diekema DJ, Pfaller MA, Shortridge D, Zervos M, Jones RN. Twenty-year
trends in antimicrobial susceptibilities among Staphylococcus aureus from
the SENTRY antimicrobial surveillance program. Open Forum Infect Dis.
2019;6(Suppl 1):S47–53.
39. Partridge SR, Kwong SM, Firth N, Jensen SO. Mobile genetic elements
40.
associated with antimicrobial resistance. Clin Microbiol Rev. 2018;31. https://
doi.org/10.1128/CMR.00088-17.
Tenover FC, McDougal LK, Goering RV, Killgore G, Projan SJ, Patel JB, et.
al. Characterization of a strain of community-associated methicillin-
resistant Staphylococcus aureus widely disseminated in the United
States. J Clin Microbiol 2006;44:108–118. doi: https://doi.org/10.1128/
JCM.44.1.108-118.2006.
41. Côrtes MF, Botelho AM, Almeida LG, Souza RC, de Lima Cunha O, et al.
Community-acquired methicillin-resistant Staphylococcus aureus from ST1
lineage harboring a new SCCmec IV subtype (SCCmec IVm) containing the
tetK gene. Infect Drug Resist. 2018;11:2583–92. https://doi.org/10.2147/IDR.
S175079.
42. Gerber SI. Describing the methicillin-resistant Staphylococcus aureus
epidemic: a public health challenge. Expert Rev Anti-Infect Ther. 2006;4(6):
905–7. https://doi.org/10.1586/14787210.4.6.905.
43. Mekonnen SA, Palma Medina LM, Glasner C, Tsompanidou E, de Jong A,
Grasso S, et. al. Signatures of cytoplasmic proteins in the exoproteome
distinguish community-and hospital-associated methicillin-resistant
Staphylococcus aureus USA300 lineages. Virulence. 2017;8(6):891–907. doi:
https://doi.org/10.1080/21505594.2017.1325064.
Negrete-González et al. BMC Microbiology (2020) 20:213
Page 11 of 11
44. Brakstad OG, Aasbakk K, Maeland JA. Detection of Staphylococcus aureus by
polymerase chain reaction amplification of the nuc gene. J Clin Microbiol.
1992;30(7):1654–60. 1629319.
45. Zhang K, Mcclure J, Elsayed S, Louie T, Conly JM. Novel multiplex PCR assay
for characterization and concomitant subtyping of staphylococcal cassette
chromosome mec types I to V in methicillin-resistant Staphylococcus aureus.
J Clin Microbiol. 2005;43(10):5026–33. https://doi.org/10.1128/JCM.43.10.
5026-5033.2005.
46. Zhang K, McClure J-A, Conly JM. Enhanced multiplex PCR assay for typing
of staphylococcal cassette chromosome mec types I to V in methicillin-
resistant Staphylococcus aureus. Mol Cell Probes. 2012;26(5):218–21. https://
doi.org/10.1016/j.mcp.2012.04.002.
47. Okolie CE, Wooldridge KG, Turner DP, Cockayne A, James R. Development
of a new pentaplex real-time PCR assay for the identification of poly-
microbial specimens containing Staphylococcus aureus and other
staphylococci, with simultaneous detection of staphylococcal virulence and
methicillin resistance markers. Mol Cell Probes. 2015;29(3):144–50. https://
doi.org/10.1016/j.mcp.2015.03.002.
Shopsin B, Gomez M, Montgomery SO, Smith DH, Waddington M, Dodge
DE, et al. Evaluation of protein a gene polymorphic region DNA sequencing
for typing of Staphylococcus aureus strains. J Clin Microbiol. 1999;37(11):
3556–63 https://doi.org/10.1128/jcm.37.11.3556-3563.1999.
48.
49. Harmsen D, Claus H, Witte W, Claus H, Turnwald D, Vogel U. Typing of
methicillin-resistant Staphylococcus aureus in a university hospital setting by
using novel software for spa repeat determination and database
management. J Clin Microbiol. 2003;41(12):5442–8. https://doi.org/10.1128/
jcm.41.12.5442-5448.2003.
Enright MC, Day NPJ, Davies CE, Peacock SJ, Spratt BG. Multilocus sequence
typing for characterization of methicillin-resistant and methicillin-susceptible
clones of Staphylococcus aureus. J Clin Microbiol. 2000;38(3):1008–15 https://
doi.org/10.1128/jcm.38.3.1008-1015.2000.
50.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
| null |
10.1021/jacs.3c00612
| null | null |
pubs.acs.org/JACS
Article
Experimental Tests of the Virtual Circular Genome Model for
Nonenzymatic RNA Replication
Dian Ding, Lijun Zhou, Shriyaa Mittal, and Jack W. Szostak*
Cite This: J. Am. Chem. Soc. 2023, 145, 7504−7515
Read Online
ACCESS
Metrics & More
Article Recommendations
*sı Supporting Information
ABSTRACT: The virtual circular genome (VCG) model was
proposed as a means of going beyond template copying to indefinite
cycles of nonenzymatic RNA replication during the origin of life. In
the VCG model, the protocellular genome is a collection of short
oligonucleotides that map to both strands of a virtual circular
sequence. Replication is driven by templated nonenzymatic primer
extensions on a subset of kinetically trapped partially base-paired
configurations, followed by the shuffling of these configurations to
enable continued oligonucleotide elongation. Here, we describe
initial experimental studies of the feasibility of the VCG model for
replication. We designed a small 12-nucleotide model VCG and
synthesized all 247 oligonucleotides of
lengths 2 to 12
corresponding to this genome. We experimentally monitored the
labeled primers in the pool of VCG oligonucleotides following the addition of activated nucleotides and
fate of individual
investigated the effect of factors such as oligonucleotide length, concentration, composition, and temperature on the extent of primer
extension. We observe a surprisingly prolonged equilibration process in the VCG system that enables a considerable extent of
reaction. We find that environmental fluctuations would be essential for continuous templated extension of the entire VCG system
since the shortest oligonucleotides can only bind to templates at low temperatures, while the longest oligonucleotides require high-
temperature spikes to escape from inactive configurations. Finally, we demonstrate that primer extension is significantly enhanced
when the mix of VCG oligonucleotides is preactivated. We discuss the necessity of ongoing in situ activation chemistry for
continuous and accurate VCG replication.
■ INTRODUCTION
Nonenzymatic RNA replication is thought to have been an
essential early step that allowed the first RNA world protocells
to begin the process of Darwinian evolution. As an
intermediate stage between untemplated nucleotide polymer-
ization and ribozyme-catalyzed RNA replication,
template-
directed nonenzymatic replication could have enabled the
replication of protocells
seeded with initially random
sequences. Such a chemically driven exploration of sequence
space would have set the stage for the evolution of the first
functional ribozymes. Although recent advances have sug-
gested potential routes for extensive template copying by RNA
primer extension, going beyond template copying to cycles of
replication remains a significant challenge.1
The nonenzymatic copying of a long RNA strand would
result in a stable duplex that must be dissociated to allow for
the next round of replication. A variety of potential solutions to
the strand separation problem have been suggested. Thermal
denaturation can readily dissociate short RNA duplexes, but
this becomes increasingly challenging with strands long enough
ribozymes.1 Other environmental
to fold into functional
influences such as pH fluctuations,2 solvent viscosity cycles,3
microscale water evaporation/condensation cycles,4−6 or other
special geological properties can potentially couple with
thermocycling to facilitate strand separation. For example,
heat flux across a cylindrical pore can facilitate the periodic
shuffling of a complex mixture of oligonucleotides to enable
ribozyme-catalyzed RNA replication through ligation.7,8
However, in the absence of ribozymes, the rate of template
copying at reasonable concentrations is much slower than the
reannealing of the separated strands, which would block primer
extension. Small fractions of backbone 2′-5′ linkages or DNA
were shown to lower the melting temperature,9−11 but they
also increase the hydrolytic lability of the duplex and slow
down primer extension.12,13 As an alternative strategy, our lab
has previously demonstrated that RNA oligonucleotides can
lead to toehold-mediated branch migration that can open up a
the duplex, allowing for strand displacement
segment of
January 18, 2023
Received:
Published: March 24, 2023
© 2023 The Authors. Published by
American Chemical Society
7504
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
Journal of the American Chemical Society
pubs.acs.org/JACS
Article
synthesis by nonenzymatic primer extension.14 This approach
is closer to the helicase-catalyzed strand displacement that
occurs at replication forks in modern biology, but other
problems with nonenzymatic RNA replication remain.
the assembly of
The difficulties in replicating ribozyme-length sequences
recently led us to consider
functional
ribozymes by the ligation of shorter oligonucleotides that
would be easier to replicate. Our lab has recently demonstrated
that splinted ligation and loop-closing ligation can form
functional ribozymes from short oligonucleotides.15,16 How-
ever, even the replication of shorter oligonucleotides faces
problems that can lead to information loss at both ends of the
sequence. First, the nonenzymatic copying of the last base of a
sequence by primer extension is known to be very slow relative
to the copying of internal nucleotides,17 which could lead to
progressive loss of 3′-sequences over cycles of replication. This
notorious “last base addition problem” is now understood as
being due to the primary mechanism of nonenzymatic primer
extension, which requires the binding of an imidazolium-
bridged dinucleotide intermediate (N*N) to the template by
two base pairs.18 With only one base pair possible at the last
base of the template, binding of the bridged dinucleotide is
greatly weakened, thus reducing the rate of primer extension.
While an imidazole-activated mononucleotide can still perform
nonenzymatic primer extension, the reaction is much slower
and more error-prone.19,20
Maintenance of the genetic information at the 5′-end of a
sequence is even more problematic since this would require a
continuous supply of a specific primer, which is clearly not
information will be lost
prebiotically plausible. As a result,
when nonenzymatic primer extension is initiated at an internal
position on a template. Although ligation events could
potentially salvage some internally initiated strands,
this
process is slow and inefficient and would be completely
prevented if the 5′-end is unphosphorylated or is blocked by a
nucleotide 5′-5′-pyrophosphate cap.
These problems have led others to propose that primordial
genome replication occurred by a rolling circle process,
in
which primer extension continues many times around a
circular
template, spinning off a long multimeric single-
stranded product.21,22 As in modern viroid replication, this
linear product would have to be cleaved into unit-length
strands, which would then have to become circularized to
generate a circular template. The process would then have to
repeat for the other strand. Since this process would require
very extensive primer extensions in the face of the topological
difficulties of replicating a small circular RNA, as well as
requiring multiple ribozyme activities
for cleavage and
circularization, we do not consider rolling circle replication
to be a viable model for nonenzymatic RNA replication.
The above problems led us to propose the virtual circular
genome (VCG) model
for prebiotically plausible non-
enzymatic RNA replication.23 Under prebiotically plausible
conditions, spontaneous untemplated24−27 and templated
polymerization28 may give rise to a large diversity of short
oligonucleotides, small subsets of which could then become
encapsulated within lipid vesicles. As a result, each primordial
protocell genome would initially consist of a unique collection
of
such cases,
oligonucleotide overlaps would occur, such that the encapsu-
lated oligonucleotides would map onto one or both strands of
one or more virtual circular sequences (Figure 1A). Since a
circular genome does not have a defined start or end, copying
short oligonucleotides.
In a fraction of
Figure 1. (A) Schematic illustration of the virtual circular genome
model. The green circle represents the virtual genome that does not
correspond to any actual oligonucleotide. A subset of
the real
oligonucleotides in the VCG system is illustrated as the blue and red
arrows. Dotted lines, along with the bold arrows, showed how the
oligomers map onto the virtual circular genome. The two
complementary sequences selected for this study are shown inside
the green circle. The direction from 5′ to 3′ is clockwise for the blue
sequence and counterclockwise for the red sequence, which is the
same direction as the arrows. Adapted from Figure 3 of ref 23 with
permission under a Creative Commons Attribution 4.0 International
License. Copyright 2021 Zhou et al.; Cold Spring Harbor Laboratory
Press for
the RNA Society. (B) Examples of productive and
nonproductive configurations of annealed oligonucleotides.
can be initiated and terminated at any position. This genome is
not represented by any actual circular molecules but is instead
represented by all possible fragments from both strands of the
virtual sequence. In theory, every oligonucleotide in this
system can act as a primer, template, or as a downstream
helper due to stacking interactions or by forming an
imidazolium-bridged intermediate. Denaturation and reanneal-
ing induced by environmental fluctuations can generate
kinetically trapped partially base-paired configurations,29 of
which a productive fraction will enable primer extensions and
ligations to occur (Figure 1B). Shuffling of these base-paired
configurations would allow for additional elongation to occur,
and RNA-mediated branch migration could also open up base-
paired regions, allowing for primer extension by strand
displacement synthesis. In this model, the process of genetic
replication is distributed across all of the oligonucleotides of
the entire system through cycles of rearrangements of base-
paired configurations.
We envision the VCG system as, in effect, an assembly line
where newly generated or introduced short oligonucleotides
gradually become elongated to strands of roughly 10−20
nucleotides in length. Oligonucleotides of this length can then
be assembled into functional ribozymes, either by splinted
ligation15 or by iterated loop-closing ligation.16 These
ribozyme building blocks could be the end products of one
or potentially multiple virtual circular genomes replicating
together in a protocell in a prebiotically plausible environment.
7505
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
Journal of the American Chemical Society
pubs.acs.org/JACS
Article
Figure 2. Demonstration of extension inside the virtual circular genome system. (A) Comparison between the VCG system and a single-template
system, with schematic representations of the two experiments shown flanking the PAGE gel image. The VCG system contains 1 μM of all of the
VCG oligos listed in Table S1. The single-template system contains 1 μM of the template and 1 μM of the primer. Extensions were monitored
using trace 5′-32P-GUGAUG added to the reactions. The small VCG diagram is adapted from Figure 3 of ref 23 with permission under a Creative
Commons Attribution 4.0 International License. Copyright 2021 Zhou et al.; Cold Spring Harbor Laboratory Press for the RNA Society. (B)
Continuous VCG extension for 3 days, with and without periodic replenishment of 20 mM activated N*N and 90 °C heat pulses. The scheme on
the right demonstrates different treatments for the three reactions. All reactions were conducted at room temperature, with 50 mM MgCl2, 200 mM
Tris−HCl (pH 8.0), and 20 mM pre-equilibrated N*N.
Here, we explore a model VCG system with a 12-nt long
virtual genome represented by 247 different oligonucleotides,
which range from 2 to 12 nucleotides in length. Several dimers
and trimers occur multiple times in the sequence. Using
radiolabeling, we monitored the fate of individual oligonucleo-
tides in the system following the addition of activated
nucleotides or bridged dinucleotides. We investigated the
effect of
factors including oligonucleotide length, concen-
tration, and temperature on the primer extension yield. In the
course of these studies, we discovered a surprisingly prolonged
equilibration process of the oligonucleotide mix in the VCG
system that enables a considerable extent of
reaction.
Furthermore, we found that environmental fluctuations
would be essential for continuous and templated extension
of the entire VCG system across different oligo lengths. Finally,
we discuss the necessity of either a flow system or ongoing in
situ activation chemistry for continuous and accurate VCG
replication.
■ RESULTS
Primer Extension in the VCG Mix vs on a Single
Template. To begin to test the virtual circular genome model,
we first selected a 12-nt virtual circular genome sequence with
no secondary structure or kinetically severe stalling points such
as UU sequences that are difficult to copy (Figure 1A). The
sequence that we selected is represented by 247 different
oligonucleotides, ranging from 2 to 12 nucleotides in length,
that map to either strand of the virtual circular sequence
(Table S1). Every oligonucleotide in the system can,
in
principle, bind to many complementary oligonucleotides, but
the most
thermodynamically favored pairing will be the
formation of a fully base-paired duplex. To form kinetically
trapped partially base-paired configurations for
template
copying, we used a brief (10 s) initial 90 °C pulse to disrupt
all base-pairing. We expected subsequent fast cooling to trap a
fraction of the oligonucleotides in metastable configurations
that would allow complementary imidazolium-bridged dinu-
cleotides to bind to a template strand next to a primer and
react by primer extension (Figure 1B). Imidazolium-bridged
dinucleotides can extend a primer by one nucleotide, with an
activated mononucleotide displaced as the leaving group. All
ten possible intermediates were supplied at
the same
concentration (∼1.7 mM each) for all primer extension
reactions in the system (Figure S1).
We then set out to determine whether it is possible for
oligonucleotides to be elongated by primer extension in the
7506
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
Journal of the American Chemical Society
pubs.acs.org/JACS
Article
Figure 3. Virtual circular genome extension with different oligonucleotide compositions. (A) VCG extension when concentrated vs diluted. (i)
Concentration of each oligonucleotide at the indicated length. (ii) VCG extension measured by % unextended 5′-32P-GUGAUG. (iii) Melting
temperature of p-GUGAUG measured as a function of concentration in the primer extension buffer. (B) VCG extension at different concentration
gradients. The concentration gradient is expressed as [(pN)i]/[(pN)i+1], starting at 1 μM of each 12-mer. (i) Oligonucleotide concentrations in
each gradient. The concentrations of 2−5-mer in 1.41× exceed the y-axis limit. See Table S2 for all concentrations. (ii) VCG extension under
different concentration gradients. (C) Extension in partial VCG mixtures containing only the longer or shorter oligomers. (i) Oligonucleotide
composition in the partial VCG system. (ii) VCG extension in the partial system. All reactions were measured by the extension of 5′-32P-GUGAUG
(<0.05 μM) conducted at room temperature, with 50 mM MgCl2, 200 mM Tris−HCl (pH 8.0), and 20 mM pre-equilibrated N*N. See Table S2
for detailed oligomer concentrations in different VCG mixtures.
highly complex virtual circular genome system. We monitored
the extension of individual labeled primers occurring within
the mixture of 247 different VCG oligonucleotides (Figure
2A). We started by monitoring a single radiolabeled 6-mer
oligonucleotide added in trace concentration (<0.05 μM) to a
mixture of 1 μM of each VCG oligonucleotide, which we refer
to as the 1× VCG mixture. About half of
the initial
radiolabeled 6-mer was extended to the corresponding 7-mer
in 1 day. This rate of primer extension was much slower than
in the positive control, in which the same labeled primer was
incubated with only one complementary 12-mer template (1
μM). Nevertheless, this observation shows that a significant
to form
fraction of
configurations that are productive for primer extension and
that a fraction of these configurations exist for a time scale of
hours to days.
the VCG oligonucleotides anneal
fast equilibration of
We then asked what limits primer extension in the virtual
circular genome system compared to the single-template
system. One possibility is that
the
oligonucleotides depletes available templates as they become
sequestered in stable duplexes. Since only one heat pulse was
applied to dissociate duplexes and initialize the process, if all
oligonucleotides with melting points above room temperature
quickly equilibrated back to form stable duplexes with their
own complementary strands, then the labeled 6-mer would be
rapidly dissociated from any suitable templates for primer
extension. An alternative extreme possibility would be the
continued but very slow rearrangement of the initially formed
oligonucleotide complexes. If all oligonucleotide configurations
after the initial heat pulse were locked in place, then any
radiolabeled 6-mer trapped in an unproductive configuration
would not be able to shuffle into a productive configuration,
and primer extension would cease after all initially productive
configurations had become extended. However, it is unlikely
for a 6-mer with an estimated koff of ∼19 s−1
to its
complementary strand29 to bind so tightly that it could not
either spontaneously dissociate from its template or be strand
displaced by another longer complementary oligonucleotide.
The resulting free 6-mer could then anneal to a new template,
where it would have another opportunity to be extended.
Besides the equilibration and rearrangement rates, another
potential
limiting factor in the VCG system is simply the
proportion of productive configurations at any given time.
Since nonenzymatic templated extension requires at least two
open nucleotide binding sites downstream of a template-bound
primer for efficient reaction, any other kinetically trapped
configurations will block templated extension (Figure 1B).
Unlike the single-template system, where most primers can
form the appropriate primer−template complex and therefore
be extended, many of the oligonucleotides in the VCG system
will be at least initially bound in unproductive configurations.
Because the initial rate of primer extension in the VCG mix is
than the rate in the single-template control, we
slower
hypothesize that the initial
limiting factor for fast primer
extension is the proportion of productive configurations and
that slow equilibration in the complex virtual circular genome
system as well as the ongoing hydrolysis of the activated
species are responsible for the subsequent continuing decline
in the rate of primer extension.
Rearrangement and Equilibration of the Base-Paired
Configurations in the VCG System. To test the idea that
continued spontaneous shuffling of productive configurations
might be occurring, we allowed the same primer extension
reaction to continue for an extended time without any external
treatments. Remarkably, template-directed primer extension
continued for at least 3 days at an ever-declining rate (Figure
2B). This result suggests that at
the
oligonucleotide complexes were still shuffling and acting as
templates for primer extension after 3 days. However, we
suspected that the declining primer extension rate was also
partially due to a declining concentration of activated species
(N*N bridged dinucleotides) available at later times because
of their relatively rapid hydrolysis under primer extension
conditions (∼85% hydrolysis in 1 day) (Figure S2). Therefore,
least a fraction of
7507
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
Journal of the American Chemical Society
pubs.acs.org/JACS
Article
we performed a similar 3-day VCG reaction with replenish-
ment of N*Ns each day. These freshly supplied activated
species boosted the extent of primer extension in the VCG
system, suggesting that a significant proportion of productive
oligonucleotide configurations were still present in the system
after 3 days. Having established that replenishment of activated
nucleotides allows for continued primer extension, we then
asked whether additional thermal cycling at later time points
could improve primer extension by shuffling the base-paired
configurations of the VCG oligonucleotides. To our surprise,
when additional heat pulses were performed just prior to each
N*N replenishment, no significant improvement in primer
extension was observed. We speculate that the medium-sized
oligonucleotides in the VCG system were probably shuffling
well enough at room temperature to continuously generate
productive configurations that additional heat pulses to reset
the system did not induce significant improvement.
Given the remarkably prolonged equilibration process in the
VCG model, we asked if system-wide changes in oligonucleo-
tide concentrations would impact the observed extent and rate
of primer extension. Diluting or concentrating the entire VCG
oligo mixture will affect
the concentration of every
oligonucleotide complex in the system by affecting the
association rate for duplex formation. Although one might
expect that dilution and hence weaker binding of the short 6-
in reduced primer
mer primer to templates would result
extension, what we observed was the opposite. Under the same
reaction conditions, a less concentrated VCG mixture
exhibited faster primer extension and a greater yield of the
extended product (Figure 3A). We suggest that the lowered
concentration of short oligonucleotides allowed for a greater
initial fraction of productive configurations and that the slower
association rate for duplex formation allowed newly opened
templates to remain available for the primer extension for a
longer time. This result suggests that concentration fluctua-
tions could facilitate the continued rearrangement of
oligonucleotide configurations in the VCG mix. Changes in
oligonucleotide concentration can also be interpreted in terms
of concentration-dependent changes in duplex Tm. A more
dilute VCG mix implies a lower effective Tm for all
oligonucleotide duplexes, which could facilitate continued
shuffling of base-paired configurations. As a point of reference,
we measured the melting temperature of our 6-mer primer and
its complement at three different concentrations in a primer
extension buffer
to demonstrate this relationship (Figure
3A(iii)). A three-fold decrease in concentration led to a 1 °C
decrease in Tm, and even this modest effect was enough to lead
to a noticeable increase in primer extension.
In a further attempt
to manipulate the proportion of
productive configurations in the VCG system, we adjusted the
concentrations of
the VCG oligonucleotides in a length-
dependent manner. We reasoned that if, on average, elongation
by primer extension is slow, as we observed, then a length-
dependent concentration gradient might emerge, with shorter
oligonucleotides being more abundant than longer oligonu-
cleotides. For
the following experiments, we made the
simplifying assumption of an exponential gradient of length
distribution, where the concentration gradient is defined as
[(pN)i]/[(pN)i+1]. For example, a 1.41× VCG system with 1
μM of each 12-mer contains 1.41 μM of each 11-mer and 2
μM of each 10-mer. Table S2 lists the concentration of each
length in the different
oligonucleotide as a function of
the following
concentration gradients that we used for
experiments. As previously noted, with a 2× concentration
gradient, the primer extension of all oligonucleotides by one
nucleotide on average results in duplication of the entire
population, i.e., one round of replication. Similarly, a 1.41×
(≈√2) gradient requires an average of 2 nucleotides and a
1.2× (≈4√2) gradient requires approximately 4-nt of primer
extension for one round of replication.23
that
is able to bind to a longer
Experimentally, we observed that a steeper concentration vs
the length gradient leads to a significantly slower rate of
extension of a labeled 6-mer primer (Figure 3B). We interpret
this effect as being due to increased competition for binding to
the limited concentration of longer oligonucleotides, which are
expected to be better templates as they are long enough to
provide binding sites for a primer, a bridged dinucleotide
substrate, and a downstream helper. The ratio of a 6-mer
primer to a 12-mer template in a 1× concentration gradient is
1:1, but this ratio increases to 8:1 in the 1.41× and 64:1 in a
2× concentration gradient. As a result, the fraction of the 6-
mer primer
template
oligonucleotide is lower with a steeper gradient. Thus primer
is
extension, expressed as a fraction of the input primer,
decreased; however, it should be noted that the total amount
of the extended primer is increased. For example, while the 1×
gradient can produce 1 μM × 53% = 0.53 μM of newly
extended 7-mer in one day, the 1.41× gradient can produce up
to 8 μM × 18% = 1.44 μM, almost tripling the amount. The
effect of the concentration gradient on the extension rate is
seen with oligonucleotides of different lengths. We measured
the extension of 8-, 10-, and 12-nt primers, and in all cases, the
fraction of primer extended vs time was higher in a VCG mix
with a shallower concentration vs length gradient (Figure S3).
We also tested a 0.83× gradient, where longer oligonucleotides
are present at higher concentrations than shorter oligonucleo-
tides. With this reverse gradient, we observed a faster rate of
primer extension than in a 1× gradient, presumably due to the
higher availability of longer oligonucleotides as good templates.
To further investigate the factors controlling the rearrange-
ment of base-paired configurations, we explored partial VCG
systems where only the shorter or longer VCG oligonucleo-
tides were supplied. An optimal template for primer extension
requires sufficient complementarity to the primer for stable
binding and at
two additional unpaired nucleotides
downstream of the primer to act as the binding site for an
activated bridged dinucleotide. For our 6-mer primer, an
optimal template would need to be at least 8-nt long. We first
examined a partial VCG system consisting of only 2- to 8-mer
oligonucleotides. In this system, only one of the 24 8-mers
would be an optimal template for the radiolabeled 6-mer. We
observed a slower initial rate of primer extension and a lower
extent of primer extension at 24 h in the 2−8-mer partial VCG
system than in the complete system (∼36 vs 53%), presumably
because of the low proportion of the productively arranged 6-
mer primer at any given time point (Figure 3C). However,
even though the rate was low, this observation suggests that
even a VCG system with an 8-nt genome allows extensions. On
the other hand, the 9−12-mer partial VCG system, which
contains only the longer subset of oligonucleotides, shows
extremely good primer extension, with essentially complete
primer extension by one or more nucleotides in one day.
Because all of
these longer oligonucleotides are present
together with their complementary strands in the VCG system,
we initially expected that the rapid formation of stable duplexes
would prevent significant primer extension. Since not all of the
least
7508
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
Journal of the American Chemical Society
pubs.acs.org/JACS
Article
Figure 4. Length dependence and temperature effect on the primer extension in the 1× VCG system. (A(i)) Schematic representation of possible
base-paired configurations between radiolabeled oligonucleotides of varying lengths and complementary 12-mers. (ii) Extension of oligonucleotides
with different lengths in the 1× VCG oligo mix, represented by the percentage of unextended 5′-32P-labeled oligonucleotide over time. (iii)
Sequences of the labeled oligomers and their melting temperatures, measured in the primer extension buffer. (B) Heat pulses facilitate the
continuous VCG extension of a 10-mer (5′-32P-UGUGGUGAUG). (i) Experimental scheme. (ii) Measured extension with or without the heat
pulses. The heat pulses were performed by 10 s of 90 °C heating, followed by immediate 1 min cooling on ice. The replenishments were performed
by adding 10 mM of equilibrated and lyophilized N*N powder. (C) Lower temperature facilitates the VCG extension of a 4-mer (5′-32P-GAUG).
(i) A scheme showing that a tetramer in the VCG has a higher chance to anneal to a complementary strand at lower temperatures. (ii) 4-mer
extension in VCG at different temperatures. All reactions were conducted in 1× VCG with 200 mM Tris−HCl (pH 8.0), 50 mM MgCl2, and an
initial addition of 20 mM N*N.
radiolabeled 6-mer could be in a productive configuration after
the initial heat pulse, the fact that primer extension continued
until all of the 6-mer primer had been extended implies that
rearrangements of base-paired configurations were happening
in the VCG system for these 9−12-nt oligonucleotides at room
temperature.
Extension of Oligonucleotides of Different Lengths
in the VCG System. The length of an oligonucleotide in the
VCG system is likely to affect both its initial
likelihood of
annealing in a productive configuration as well as the dynamics
of the exchange processes that would allow for continued
primer extension. We, therefore, determined primer extension
rates for a series of oligonucleotides of different lengths (Figure
4A). To avoid the effects of differing sequences at the 3′-end of
the primer, we used a set of oligonucleotides with the same 3′-
end as the 6-mer primer used above and varied only the 5′-end.
Initially, we expected that longer oligonucleotides might show
faster initial rates of primer extension since they would be able
to bind more strongly to longer templates. We also expected
slower long-term rates of primer extension since they would be
more likely to become sequestered in stable, unproductive
configurations that would be unable to exchange into new
productive configurations. Surprisingly, we observed a
progressive decrease in both the initial and long-term rates
7509
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
Journal of the American Chemical Society
pubs.acs.org/JACS
Article
of primer extension as oligonucleotide length increased from 6
to 8, 10, and then to 12 nucleotides. We suggest that both of
these effects stem from a decreased probability of forming
productive configurations. The melting temperatures of these
oligonucleotides, when paired with their perfect complements,
increase significantly with length (Figure 4A(iii)). This greater
duplex stability is likely to decrease the spontaneous shuffling
of paired configurations in the VCG system, decreasing the
rate of primer extension at long times. Why longer primers are
extended more poorly initially is less clear but could potentially
be due to occupancy by pairs of shorter oligonucleotides,
preventing the formation of productive configurations.
Alternatively, toehold-mediated branch migration may lead to
rapid loss of productive configurations, thereby decreasing
primer extension even at early times.
To facilitate the shuffling of the longer oligonucleotides for
continuous elongation, we tested the effect of periodic
temperature fluctuations on the extension of
the 10-mer
primer. After an initial high-temperature pulse to initialize the
system, three additional high-temperature pulses (90 °C for 10
s) were applied every 2 h to shuffle the oligonucleotide
configurations. Fresh activated N*Ns were added after the
second and third high-temperature pulses to counter the effects
of hydrolysis
the initial bridged
dinucleotides had already hydrolyzed after 4 h (Figures S1
in the extent of 10-mer
and S2). A clear improvement
extension was observed with the extra high-temperature pulses,
demonstrating the importance of temperature fluctuations for
the continued elongation of longer oligonucleotides in the
VCG system (Figure 4B). As expected, the improvement in
primer extension was even greater when fresh N*N substrates
were added after each high-temperature pulse.
since about half of
In contrast
the time at
to the requirement of high-temperature
fluctuations for primer extension of longer oligonucleotides,
we have found that shorter oligonucleotides can only be
extended at lower temperatures. When a 4-mer primer is
radiolabeled and monitored in the VCG mix at
room
temperature (22 °C), we observe only minimal primer
extension. In addition to the low yield, many of the extended
products formed were incorrect (Figure S4B). We first
hypothesized that in the VCG mix, much of the 4-mer was
not bound to any template most of
room
temperature and that the observed extension arose primarily
through untemplated extension. However, a control experi-
ment showed that the 4-mer can extend efficiently on a single
template at room temperature (80% at 24 h), although the
extent of primer extension does improve markedly at lower
temperatures (Figure S4A). Therefore, the poor extension of
the 4-mer primer in the VCG system is not solely due to poor
binding. We speculated that rapid dissociation of the 4-mer
from a template strand, followed by template occupancy by a
competing oligonucleotide, would prevent primer extension
(Figure 4C(i)). We, therefore, tested the effect of reducing the
temperature on 4-mer extension yield in the VCG system. As
temperatures decreased, we observed increased correct
extension and decreased misincorporation (Figures 4C(ii)
and S4B). The remarkably improved yield and fidelity suggest
that primer extension of the shorter oligonucleotides in the
VCG system requires a lower temperature to prevent rapid loss
of productive configurations.
Fidelity in the Virtual Circular Genome Scenario. The
significant degree of misincorporation observed with the 4-mer
at room temperature drew our attention to the possibility of
untemplated extension in the VCG system. The untemplated
extension could result not only from unbound oligonucleotides
but also from some of the unproductive configurations in the
VCG system. As shown in Figure 1B, many unproductive
configurations have either an overhanging or blunt 3′-end that
can potentially be subject to untemplated extension. Moreover,
the 5′-phosphate of both free oligonucleotides and some
template-bound oligonucleotides may also react to form 5′-5′-
pyrophosphates. Indeed, polyacrylamide gel electrophoresis
(PAGE) analysis of the extension of the 6-mer primer in the
VCG system (Figure 2A) clearly shows that several products
are formed that are not seen in the single-template system,
suggesting that these misincorporations most likely derive from
processes other than templated primer extension. Interestingly,
increasing the concentration of
the bridged dinucleotides
enhanced the synthesis of these incorrect products but did not
significantly improve the correct templated primer extension
reaction (Figure S5).
in the presence of
To identify the sources of these misincorporations, we first
examined the untemplated extension of specific oligonucleo-
in the presence of all possible activated bridged
tides
dinucleotides. In the VCG system, because every oligonucleo-
tide exists
its partially and fully
complementary strands, blunt ends can form at either end of
the oligonucleotide. Therefore, we tested both single-stranded
RNAs of different lengths and the corresponding double-
stranded duplexes for untemplated extension. To our surprise,
we observed enhanced untemplated extension with blunt-
ended species (Figure S6).
Because of the limited ability of PAGE analysis to resolve
different products of untemplated extension, we determined
the extent and regioselectivity of untemplated extension by
supplying only one bridged homo-dinucleotide at a time
(Figure S7). The identity of each extended product was
determined by comparison with authentic radiolabeled samples
(see Materials and Methods for the synthesis of standards).
Untemplated oligonucleotide polymerization has long been
known to favor 2′- over 3′-extension due to the greater
nucleophilicity of the 2′-hydroxyl group, and the formation of
5′-5′ pyrophosphate products is known to be an unavoidable
byproduct of
reactions with nucleotide phosphorimidazo-
lides.30,31 In our examination of untemplated extension, we
also observed a predominance of products with nucleotides
added at either the 2′-OH or 5′-phosphate. Blunt-ended
duplex oligonucleotides appear to be particularly prone to
nucleotide addition to the 2′-hydroxyl, especially with G
(Figure S7). In addition to the untemplated extension of
single-stranded and blunt-end RNAs, we also examined the
primer extension of the labeled 6-mer primer in the VCG mix
in the presence of only one imidazolium-bridged homo-
dinucleotide at a time. Note that correct templated extension,
in this case, requires a C*G bridged dinucleotide. In the
absence of this fully complementary substrate, the products of
the primer extension were quite similar to those of
the
untemplated reactions, with most of the elongations being at
the 2′- or 5′-end. When supplied with a C*C bridged
the observed correct 3′-extension with C
dinucleotide,
probably resulted from the substrate binding to the template
with a downstream C:C mismatch. Interestingly, we observed
less extension with bridged homo-dinucleotides in the VCG
system than with an isolated duplex, especially when G*G was
the
supplied. This finding suggests
oligonucleotides in the VCG mix results in a low proportion
that annealing of
7510
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
Journal of the American Chemical Society
pubs.acs.org/JACS
Article
of blunt-ended duplexes, as might be expected since there are
many more annealed configurations with 5′- or 3′- overhangs
than blunt-ended configurations.
To better identify misincorporation events, we aligned the
gel-separated VCG extension products with the individual
untemplated reaction products; we also used phosphatase
digestion to distinguish between the 5′- (which protects the
32P-labeled 5′-phosphate from digestion) and 2′/3′-extension
(Figures 5 and S7). This assay showed that most of the
Figure 5. Detection of 5′-pyrophosphate-capped oligonucleotides.
(A) Schematic representation of the phosphatase deradiolabeling of
the extension products. The 5′-32P labels were shown as stars. The
5′-32P-oligonucleotides would be dephosphorylated while the 5′-
Np32P-oligonucleotides would be protected. (B) PAGE gel analysis of
the extension products with or without phosphatase digestion. The
VCG extension was performed with the 1× VCG mixture and 20 mM
N*N, while the untemplated reactions were performed with 1 μM 6-
mer and 20 mM of the indicated imidazolium-bridged homodimers.
See Figure S7B for more details. All reactions were run at room
temperature for 24 h. Phosphatase-digested products were loaded at
the same concentration as the untreated sample. Authentic samples
were run alongside the PAGE gel and are indicated in the figure.
apparent misincorporations in the VCG reaction were, in fact,
due to 5′-nucleotide-pyrophosphate formation. We could not
quantify how much of each pyrophosphate is formed because
primers with 5′-App-, 5′-Upp, and 5′-Cpp- have almost
identical gel mobilities. However, in the reactions with single
N*N substrates, A*A led to more formation of 5′-App-oligo
products than the corresponding products with C*C, U*U,
and G*G. It is also possible that the 5′-Gpp extension of the
specific radiolabeled primer we used could be template-
directed in the VCG mix. The 2′ + A and 2′ + U products have
similar gel mobility to the correct (templated) 3′ + C product.
However, we believe that there is little 2′-extension with A and
U because no significant amount of 2′ + C or 2′ + G products
was formed (Figure S7B). The small amount of slowly
migrating products in the VCG primer extension reaction that
survived the phosphatase digestion likely corresponds to the
5′-Npp extension of the correct 3′ + C product. As a result,
the misincorporations we observed in the VCG
most of
from the 5′-5′-pyrophosphate
systems appear
formation. We note that 5′-Npp-capped oligonucleotides can
to result
still act as fully functional primers and templates in the VCG
system; the accumulation of 5′-Npp-oligonucleotides could
also provide a selective advantage for the evolution of ribozyme
ligases that use such molecules as substrates.32
Potential Strategies to Enhance Extensions in a VCG
System. Having characterized the basic kinetics and fidelity of
primer extension in our model VCG system, we asked what
factors might further increase the rate and yield of primer
extension. Considering the rapid hydrolysis of imidazolium-
bridged substrates, an efficient method for in situ activation
would likely be extremely beneficial. This ideal approach would
lead to efficient activation of both monomers and oligonucleo-
tides, as this would allow for the formation of monomers
bridged to oligonucleotides, which we have previously shown
to be optimal substrates for primer extension.19 We, therefore,
asked whether preactivation of the VCG oligonucleotide mix
would enhance primer extension by allowing for the formation
of monomer-bridged-oligonucleotide intermediates.
We began by testing whether an activated trimer helper
could accelerate the extension of our labeled 6-mer primer in
the VCG system. We prepared the activated trimer *GUG and
doped it at increasing concentrations into the partial 9−12-mer
VCG system. Following the addition of activated monomers or
bridged dinucleotides, this trimer can form the highly reactive
C*GUG intermediate in situ. The higher affinity and greater
preorganization of this substrate facilitate the +C extension of
the 32P-labeled 6-mer primer. Previous kinetic measurements
have shown that a similar monomer-bridged-trimer (specifi-
cally, A*CGC) has a KM of 40 μM and a Vmax approaching 1
min−1 on a single template.19 When we added the *GUG
helper together with an equilibrated mix of
imidazolium-
to the partial 9−12-mer VCG
bridged dinucleotides
oligonucleotides, we observed significant acceleration of primer
extension when it was present at a concentration (∼50 μM)
closer to the estimated Kd of C*GUG. Moreover, primer
extension in the partial VCG system supplied with 100 μM
*GUG can be almost as fast as the one-template positive
control (Figure S8).
Encouraged by the observed benefit of adding a single
activated helper oligonucleotide, we asked whether activating
the entire set of VCG oligonucleotides would also help
monomer-bridged-oligonucleotides form in situ and thus
enhance primer extension. An important concern is that the
excess amount of 2-aminoimidazole required for efficient
activation will also reduce the formation of
imidazolium-
bridged substrates. To avoid this problem, we used
stochiometric 2AI to activate a concentrated set of VCG
oligonucleotides in a partially frozen reaction mixture at −15
°C and then thawed and diluted the mixture to allow primer
extension to occur at room temperature. The partial freezing
process served to concentrate the solutes in the liquid eutectic
phase between the pure ice crystals. We have previously used
in situ activation of
this approach to enable efficient
imidazolium-bridged species for nonenzymatic template copy-
ing.33 Here, we used the non-prebiotic 1-ethyl-3-(3-
dimethylaminopropyl)carbodiimide (EDC) as the coupling
reagent
for activation for ease of handling, but similar
activation chemistry can be performed using the more
prebiotically plausible methyl
isocyanide. A control NMR
experiment with a single dinucleotide demonstrated almost
complete activation under the same conditions (Figure S9).
After overnight eutectic phase activation, the reaction was
warmed to room temperature, diluted into the primer
7511
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
Journal of the American Chemical Society
pubs.acs.org/JACS
Article
Figure 6. Demonstration of possible strategies to improve VCG extension (A−B) Significantly enhanced VCG extension after preactivation with
either a 1.41× or U-shaped gradient. (i) Oligo concentration of each length. (ii) Comparison between the extensions of 5′-32P-GUGAUG inside
the VCG system with or without preactivation. (C) Faster extension in a U-shaped VCG mix with the more reactive 3′-NH2-2AIpddN modification
as a model system.
extension buffer, and a 32P-labeled 6-mer primer was added.
We started by activating the 1.41× VCG mix, in which short
oligonucleotides are present at higher concentrations than the
longer oligonucleotides. We observed significant enhancement
of primer extension (Figure 6A) even though the concen-
trations of the short oligonucleotides were still far below the Kd
the corresponding monomer-bridged-oligonucleotides.19
of
When we activated the 1× VCG system, we observed no
rate enhancement, probably because the concentration of the
short helper oligonucleotides was too low.
We then reasoned that the optimal concentration vs length
distribution might be more complex than a simple exponential
gradient. Clearly, short oligonucleotides must be present close
to their Kd to have a significant effect on primer extension. On
the other hand, medium-length oligonucleotides are elongated
most rapidly and therefore might be present at lower steady-
state concentrations, while longer oligonucleotides might
accumulate and reach higher concentrations. We therefore
prepared and activated a VCG mix with a U-shaped
concentration vs length distribution (Table S2). We were
pleased to observe improved primer extension in this system,
with about 70% of the labeled primer being extended by one or
more nucleotides in less than one day (Figure 6B).
Finally, we asked how primer extension in the VCG system
would be affected if the reaction kinetics were improved. To do
this, we employed a 32P-labeled 6-mer primer terminated with
a highly reactive 3′-amino-2′,3′-dideoxy-ribonucleotide, and
similarly modified 2-aminoimidazole activated mononucleo-
tides (3′-NH2-2AIpddNs). Although such nucleotides may not
be prebiotically plausible, they provide an excellent model
system for the simulation of nonenzymatic RNA copying under
conditions leading to enhanced rates of primer extension, such
as might be achieved, e.g., by a prebiotic catalyst or improved
conditions for chemical RNA copying. By employing the highly
nucleophilic 3′-amino group, we were able to observe ∼60%
+1 primer extension in just 1 h, with almost complete +1 or
greater extension by 4 h and a low fraction of misincorpora-
tions (Figure 6C). Remarkably, an average extension of ∼ +3
nucleotides was observed by 24 h, consistent with the
spontaneous shuffling of partially base-paired configurations
continuing for many hours.
■ DISCUSSION
We first proposed the virtual circular genome model23 as a
theoretical means of overcoming the barriers to prebiotically
plausible RNA replication. Replication in the VCG model does
not require the specific primers needed for replication of a
linear genome, and the distributed nature of the copying
processes is expected to impart resilience to chemical processes
that modify or block the 5′- or 3′- ends of
individual
oligonucleotides. Importantly, the repeated shuffling of base-
paired configurations of annealed oligonucleotides was
proposed as a means of overcoming the block to replication
imposed by rapid strand annealing. However, experimental
tests of this model were clearly needed, as template copying by
primer extension has previously been examined only in highly
simplified model systems. Our studies show that primers of
different lengths can indeed be extended by template copying
with a significant rate, extent, and fidelity in a model VCG
system, suggesting that under appropriate environmental
conditions, replication in the VCG mode may be possible.
Perhaps the most surprising aspect of our results is the
prolonged time scale (>1 day) over which primer extension in
the VCG mixture continues. We interpret the extended time
scale of primer extension as
reflecting the very slow
equilibration of the VCG oligonucleotides. The shuffling of a
simpler set of DNA oligonucleotides for template copying and
replication has been studied before,34,35 and our study provides
further insights into how a complex mixture of RNAs would
slowly equilibrate to enable nonenzymatic replication. The
very large number of competing base-paired configurations of
VCG oligonucleotides may prevent rapid equilibration to fully
base-paired duplexes, thus allowing for the continued shuffling
of partially base-paired configurations. At any given time, only
a fraction of these configurations is productive for primer
extension, while others are not. If unproductive configurations
can rearrange by dissociation, exchange, or strand displace-
ment, new productive configurations may continue to arise,
7512
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
Journal of the American Chemical Society
pubs.acs.org/JACS
Article
enabling the observed extended time course of primer
extension.
We have found that oligonucleotides that were both longer
(8, 10, and 12-nts) and shorter (4-nts) than the 6-mer primer
exhibited slower and less extensive primer extension. Short
pulses of high temperature partially rescued the poor extension
of the longer primers, suggesting that these oligonucleotides
tend to become trapped in stable unproductive configurations
that can be disrupted and exchanged during exposure to
elevated temperatures. In contrast, the shorter 4-nt oligonu-
cleotide required a lower temperature for optimal primer
in part due to the weaker binding to template
extension,
strands but also in part due to the greater lability of productive
configurations involving a base-paired 4-nt primer. These
divergent temperature requirements for the primer extension
of oligonucleotides of different lengths imply that repeated
cycles of RNA replication would only be possible in a
fluctuating environment. For example, changing temperature,
pH, or salt concentrations could trigger ongoing shuffling of
the annealed configurations of the VCG oligonucleotides.
The observed variation in the rate of primer extension with
primer length has implications for the steady-state length vs
concentration profile. In our original model, we assumed for
simplicity an exponential concentration vs length gradient, i.e.,
a constant [length n]/[length n + 1] ratio. However, if short
and long oligonucleotides are elongated more slowly than
medium-length oligonucleotides, both short and long
oligonucleotides would tend to accumulate, while medium-
length oligonucleotides would be rapidly extended, resulting in
a more U-shaped concentration vs length distribution. The
short oligonucleotides can be activated to form monomer-
bridged-oligonucleotides that will
facilitate faster extension,
while the longer oligonucleotides are good templates for
oligonucleotide extension. Further experiments will be
required to determine the steady-state distribution of VCG
oligonucleotides as a function of the length and sequence over
multiple cycles of replication.
Given the observed advantage of activating the VCG
oligonucleotides and the fast rate of hydrolysis of bridged
N*N intermediates, some means of
in situ activation will
clearly be required to enable continued oligonucleotide
elongation and thus complete cycles of
replication. Our
laboratory has recently demonstrated prebiotically plausible
activation and bridge-forming chemistry that allows one-pot
conversion of nucleotides to bridged dinucleotides with a high
yield; however,
this process requires repeated freeze-thaw
cycles, which are known to disrupt vesicles.33 Therefore, a less
disruptive process, compatible with vesicle integrity, may be
required for VCG replication within protocells. Alternatively, if
eutectic phase activation chemistry occurred in a distinct,
separate environment, periodic melting could potentially
release fresh activated nucleotides that could flow over a
population of protocells and diffuse into the vesicles while
hydrolyzed nucleotides diffuse out. In such a flow system, the
free 2AI generated from the formation of imidazolium-bridged
species could diffuse out of
shifting the
equilibrium inside the vesicles to favor the formation of 2AI-
bridged dinucleotides and monomer-bridged-oligonucleotides.
Template copying in the VCG system must proceed with
sufficient fidelity to allow the inheritance of useful amounts of
information. For a ribozyme on the order of 50 nucleotides in
length, this implies an error rate of roughly 2% or less.
the PAGE gels used to monitor primer
Examination of
the vesicles,
the 5′-end of
extension reactions in our model VCG system reveals the
presence of bands that do not correspond in mobility to the
correct products of primer extension. In principle, these bands
could correspond to products of primer extension with an
incorrect nucleotide, or to extension with a correct or incorrect
nucleotide at the 2′-hydroxyl of the primer, or to the addition
the primer via a 5′-5′
of a nucleotide at
pyrophosphate linkage, which could be formed by attack of the
5′-phosphate of the primer on the phosphate of an activated
monomer. Our experiments clearly show that 5′-5′ pyrophos-
phate-capped oligonucleotides are generated during primer
extension in the VCG system, especially from blunt-ended
duplexes. One of the major benefits of the VCG system is that
there is no defined start or end to the genomic sequence, and
oligonucleotides with a 5′-cap can still act as primers or
templates. Furthermore, the synthesis of 5′-5′ pyrophosphate-
capped oligonucleotides suggests a straightforward way in
which the evolution of ribozymes could potentiate replication.
Pyrophosphate-capped oligonucleotides can be substrates for
ligation by ribozyme ligases, much as modern DNA and RNA
ligases utilize an adenosine-5′-5′-pyrophosphate-activated
substrate.36 Our lab has previously evolved a ribozyme ligase
that catalyzes the ligation of adenylated RNAs to demonstrate
the prebiotic possibility of such a mechanism.32
In addition to mutations induced by 3′-misincorporations,
mispriming can also be a source of mutations. A vesicle
membrane that would encapsulate the VCG system and
separate it from the external environment could therefore be
extremely beneficial. The uptake of short oligonucleotides,
such as dimers and trimers, from the external environment
should not cause problems, as even a 50-nt VCG would be
likely to contain all di- and trinucleotide sequences. On the
other hand, the uptake of longer mismatched oligonucleotides
(5−8-nt) could be mutagenic. This may provide a useful
constraint in defining the desirable properties of protocell
membranes. Compartmentalizing each individual VCG system
inside a protocell is thus necessary to prevent contamination of
the VCG with random oligonucleotides that would lead to
extensive mispriming.
Finally, we note that genome replication via the VCG model
provides the raw materials necessary for spontaneous ribozyme
assembly from oligonucleotides with lengths of roughly 10−20
nts. Partially overlapping pairs of such oligonucleotides can
anneal with each other, after which loop-closing ligation can
lead to the formation of stem-loop structures.16 The iteration
of such processes could then lead to the assembly of complex
structured RNAs, including ribozymes. Furthermore, the short
oligonucleotides of the VCG could be substrates for ribozyme-
catalyzed ligation,7,8 facilitating a transition from nonenzymatic
replication to ribozyme-catalyzed RNA replication.
■ CONCLUSIONS
We initially proposed the virtual circular genome (VCG)
model as an approach to the nonenzymatic replication of RNA.
The distributed nature of template copying in the VCG model
circumvents problems associated with the replication of long
linear or circular genomes. Experimental tests of the rate,
extent, and fidelity of template copying are clearly required to
assess the viability of the VCG model. Our initial experiments
show that template-directed primer extension can indeed occur
within a complex synthetic VCG oligonucleotide mixture,
supporting our conjecture that a fraction of annealed
configurations of VCG oligonucleotides would be productive
7513
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
Journal of the American Chemical Society
pubs.acs.org/JACS
Article
for substrate binding and reaction. The surprisingly long time
course of primer extension suggests that
these annealed
configurations continue to rearrange spontaneously for
extended times, approaching the thermodynamic minimum
of
full base-pairing very slowly. Our hypothesis that an
exponential oligonucleotide concentration vs length profile
would facilitate rapid replication is not supported; rather, we
find that a U-shaped profile is optimal for template copying.
We conclude that very short oligonucleotides must be present
at high concentrations approaching their Kds for template
binding to act as effective primers and helpers, while a high
the longest oligonucleotides is beneficial
concentration of
because they are the best
templates. In contrast, a high
concentration of medium-length oligonucleotides is counter-
productive because they primarily act
to occlude needed
templates. We find that continued primer extension is
enhanced by replenishment of hydrolyzed substrates, strongly
suggesting that in situ activation will be required before cycles
of RNA replication can be demonstrated in a VCG system.
Overall, our experiments suggest that RNA replication via the
VCG model may be possible, given appropriate activation
chemistry and environmental fluctuations. Additional experi-
ments will be required to determine whether a replicating VCG
system can be maintained by feeding with activated monomers
or whether an input of activated oligonucleotides is also
required. We are currently exploring approaches to the
computational modeling of VCG replication and to the
experimental demonstration of VCG replication within
model protocells.
■ ASSOCIATED CONTENT
*sı Supporting Information
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/jacs.3c00612.
information; selection of
Abbreviations; general
the
VCG sequence; synthesis of activated nucleotides; NMR
equilibration and hydrolysis of activated nucleotides;
radiolabeled oligonucleotides; monitoring primer exten-
sion in the VCG oligonucleotide mixture; melting
temperature measurements; supplementary Figures S1
to S9; and supplementary Tables S1, S2 (PDF)
■ AUTHOR INFORMATION
Corresponding Author
Jack W. Szostak − Department of Chemistry and Chemical
Biology, Harvard University, Cambridge, Massachusetts
02138, United States; Department of Molecular Biology and
Center for Computational and Integrative Biology,
Massachusetts General Hospital, Boston, Massachusetts
02114, United States; Department of Genetics, Harvard
Medical School, Boston, Massachusetts 02115, United States;
Howard Hughes Medical Institute, Department of Chemistry,
The University of Chicago, Chicago, Illinois 60637, United
States;
Email: [email protected]
orcid.org/0000-0003-4131-1203;
Authors
Dian Ding − Department of Chemistry and Chemical Biology,
Harvard University, Cambridge, Massachusetts 02138,
United States; Department of Molecular Biology and Center
for Computational and Integrative Biology, Massachusetts
General Hospital, Boston, Massachusetts 02114, United
orcid.org/0000-0001-9046-7816
States;
Lijun Zhou − Department of Molecular Biology and Center for
Computational and Integrative Biology, Massachusetts
General Hospital, Boston, Massachusetts 02114, United
States; Department of Genetics, Harvard Medical School,
Boston, Massachusetts 02115, United States; Department of
Biochemistry and Biophysics, Perelman School of Medicine,
University of Pennsylvania, Philadelphia, Pennsylvania
19104, United States;
orcid.org/0000-0002-0393-4787
Shriyaa Mittal − Department of Molecular Biology and Center
for Computational and Integrative Biology, Massachusetts
General Hospital, Boston, Massachusetts 02114, United
States; Department of Genetics, Harvard Medical School,
Boston, Massachusetts 02115, United States;
0000-0003-3490-1969
orcid.org/
Complete contact information is available at:
https://pubs.acs.org/10.1021/jacs.3c00612
Author Contributions
The manuscript was written through contributions of all
authors.
Notes
The authors declare no competing financial interest.
■ ACKNOWLEDGMENTS
is an investigator at
J.W.S.
the Howard Hughes Medical
Institute. This work was supported in part by grants from the
Simons Foundation (290363) and the National Science
Foundation (2104708) to J.W.S. The authors thank Dr.
Marco Todisco for his helpful discussions and assistance
regarding oligonucleotide binding affinities and melting
temperatures. The authors also thank Drs. Longfei Wu and
Victor S. Lelyveld for helpful comments on the manuscript.
■ ABBREVIATIONS
virtual circular genome
VCG
2-AI or * 2-aminoimidazole or 2-aminoimidazolium
NMR
PAGE
EDC
nuclear magnetic resonance
polyacrylamide gel electrophoresis
1-ethyl-3-(3-dimethylaminopropyl)carbodiimide
■ REFERENCES
(1) Szostak, J. W. The Eightfold Path to Non-Enzymatic RNA
Replication. J. Syst. Chem. 2012, 3, No. 2.
(2) Mariani, A.; Bonfio, C.; Johnson, C. M.; Sutherland, J. D. pH-
Driven RNA Strand Separation under Prebiotically Plausible
Conditions. Biochemistry 2018, 57, 6382−6386.
(3) He, C.; Lozoya-Colinas, A.; Gállego, I.; Grover, M. A.; Hud, N.
V. Solvent Viscosity Facilitates Replication and Ribozyme Catalysis
from an RNA Duplex in a Model Prebiotic Process. Nucleic Acids Res.
2019, 47, 6569−6577.
(4) Ianeselli, A.; Mast, C. B.; Braun, D. Periodic Melting of
Oligonucleotides by Oscillating Salt Concentrations Triggered by
Microscale Water Cycles Inside Heated Rock Pores. Angew. Chem.
2019, 131, 13289−13294.
(5) Martin, C.; Frenkel-Pinter, M.; Smith, K. H.; Rivera-Santana, V.
F.; Sargon, A. B.; Jacobson, K. C.; Guzman-Martinez, A.; Williams, L.
D.; Leman, L. J.; Liotta, C. L.; Grover, M. A.; Hud, N. V. Water-Based
Dynamic Depsipeptide Chemistry: Building Block Recycling and
Oligomer Distribution Control Using Hydration−Dehydration
Cycles. JACS Au 2022, 2, 1395−1404.
7514
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
Journal of the American Chemical Society
pubs.acs.org/JACS
Article
(6) Ianeselli, A.; Atienza, M.; Kudella, P. W.; Gerland, U.; Mast, C.
B.; Braun, D. Water Cycles in a Hadean CO2 Atmosphere Drive the
Evolution of Long DNA. Nat. Phys. 2022, 18, 579−585.
(7) Kreysing, M.; Keil, L.; Lanzmich, S.; Braun, D. Heat Flux across
an Open Pore Enables the Continuous Replication and Selection of
Oligonucleotides towards Increasing Length. Nat. Chem. 2015, 7,
203−208.
(8) Salditt, A.; Keil, L. M. R.; Horning, D. P.; Mast, C. B.; Joyce, G.
F.; Braun, D. Thermal Habitat
for RNA Amplification and
Accumulation. Phys. Rev. Lett. 2020, 125, No. 048104.
(9) Engelhart, A. E.; Powner, M. W.; Szostak, J. W. Functional RNAs
Exhibit Tolerance for Non-Heritable 2′−5′ versus 3′−5′ Backbone
Heterogeneity. Nat. Chem. 2013, 5, 390−394.
(10) Gavette, J. V.; Stoop, M.; Hud, N. V.; Krishnamurthy, R.
RNA−DNA Chimeras in the Context of an RNA World Transition to
an RNA/DNA World. Angew. Chem., Int. Ed. 2016, 55, 13204−13209.
(11) Kim, S. C.; O’Flaherty, D. K.; Giurgiu, C.; Zhou, L.; Szostak, J.
W. The Emergence of RNA from the Heterogeneous Products of
Prebiotic Nucleotide Synthesis. J. Am. Chem. Soc. 2021, 143, 3267−
3279.
(12) Prakash, T. P.; Roberts, C.; Switzer, C. Activity of 2′,5′-Linked
RNA in the Template-Directed Oligomerization of Mononucleotides.
Angew. Chem., Int. Ed. Engl. 1997, 36, 1522−1523.
(13) Kim, S. C.; Zhou, L.; Zhang, W.; O’Flaherty, D. K.; Rondo-
Brovetto, V.; Szostak, J. W. A Model for the Emergence of RNA from
a Prebiotically Plausible Mixture of Ribonucleotides, Arabinonucleo-
tides, and 2′-Deoxynucleotides. J. Am. Chem. Soc. 2020, 142, 2317−
2326.
(14) Zhou, L.; Kim, S. C.; Ho, K. H.; O’Flaherty, D. K.; Giurgiu, C.;
Wright, T. H.; Szostak, J. W. Non-Enzymatic Primer Extension with
Strand Displacement. eLife 2019, 8, No. e51888.
(15) Zhou, L.; O’Flaherty, D. K.; Szostak, J. W. Assembly of a
Ribozyme Ligase from Short Oligomers by Nonenzymatic Ligation. J.
Am. Chem. Soc. 2020, 142, 15961−15965.
(16) Wu, L.-F.; Liu, Z.; Roberts, S. J.; Su, M.; Szostak, J. W.;
Sutherland, J. D. Template-Free Assembly of Functional RNAs by
Loop-Closing Ligation. J. Am. Chem. Soc. 2022, 144, 13920−13927.
(17) Wu, T.; Orgel, L. E. Nonenzymatic Template-Directed
Synthesis on Hairpin Oligonucleotides. 3. Incorporation of Adenosine
and Uridine Residues. J. Am. Chem. Soc. 1992, 114, 7963−7969.
(18) Walton, T.; Zhang, W.; Li, L.; Tam, C. P.; Szostak, J. W. The
Mechanism of Nonenzymatic Template Copying with Imidazole-
Activated Nucleotides. Angew. Chem., Int. Ed. 2019, 58, 10812−
10819.
(19) Ding, D.; Zhou, L.; Giurgiu, C.; Szostak, J. W. Kinetic
Explanations for the Sequence Biases Observed in the Nonenzymatic
Copying of RNA Templates. Nucleic Acids Res. 2022, 50, 35−45.
(20) Duzdevich, D.; Carr, C. E.; Ding, D.; Zhang, S. J.; Walton, T.
S.; Szostak, J. W. Competition between Bridged Dinucleotides and
Activated Mononucleotides Determines the Error Frequency of
Nonenzymatic RNA Primer Extension. Nucleic Acids Res. 2021, 49,
3681−3691.
(21) Tupper, A. S.; Higgs, P. G. Rolling-Circle and Strand-
Displacement Mechanisms for Non-Enzymatic RNA Replication at
the Time of the Origin of Life. J. Theor. Biol. 2021, 527, No. 110822.
(22) Kristoffersen, E. L.; Burman, M.; Noy, A.; Holliger, P. Rolling
Circle RNA Synthesis Catalyzed by RNA. eLife 2022, 11, No. e75186.
(23) Zhou, L.; Ding, D.; Szostak, J. W. The Virtual Circular Genome
Model for Primordial RNA Replication. RNA 2021, 27, 1−11.
(24) Ferris, J. P.; Ertem, G. Oligomerization of Ribonucleotides on
the 5′-Phosphorimidazolide of
Montmorillonite: Reaction of
Adenosine. Science 1992, 257, 1387−1389.
(25) Ferris, J. P.; Hill, A. R.; Liu, R.; Orgel, L. E. Synthesis of Long
Prebiotic Oligomers on Mineral Surfaces. Nature 1996, 381, 59−61.
(26) Monnard, P. A.; Kanavarioti, A.; Deamer, D. W. Eutectic Phase
Polymerization of Activated Ribonucleotide Mixtures Yields Quasi-
Equimolar Incorporation of Purine and Pyrimidine Nucleobases. J.
Am. Chem. Soc. 2003, 125, 13734−13740.
(27) Kanavarioti, A.; Monnard, P. A.; Deamer, D. W. Eutectic
Phases in Ice Facilitate Nonenzymatic Nucleic Acid Synthesis.
Astrobiology 2001, 1, 271−281.
(28) Weimann, B. J.; Lohrmann, R.; Orgel, L. E.; Schneider-
Bernloehr, H.; Sulston,
J. E. Template-Directed Synthesis with
Adenosine-5′-Phosphorimidazolide. Science 1968, 161, 387.
(29) Todisco, M.; Szostak, J. W. Hybridization Kinetics of Out-of-
Equilibrium Mixtures of Short RNA Oligonucleotides. Nucleic Acids
Res. 2022, 50, 9647−9662.
(30) Lohrmann, R.; Orgel, L. E. Preferential Formation of (2′-5′)-
Linked Internucleotide Bonds in Non-Enzymatic Reactions. Tetrahe-
dron 1978, 34, 853−855.
(31) Kanavarioti, A.; Lee, L. F.; Gangopadhyay, S. Relative
Reactivity of Ribosyl 2′-OH vs 3′-OH in Concentrated Aqueous
Solutions of Phosphoimidazolide Activated Nucleotides. Origins Life
Evol. Biospheres 1999, 29, 473−487.
(32) Hager, A. J.; Szostak, J. W. Isolation of Novel Ribozymes That
Ligate AMP-Activated RNA Substrates. Chem. Biol. 1997, 4, 607−
617.
(33) Zhang, S. J.; Duzdevich, D.; Ding, D.; Szostak, J. W. Freeze-
Thaw Cycles Enable a Prebiotically Plausible and Continuous
Pathway from Nucleotide Activation to Nonenzymatic RNA Copying.
Proc. Natl. Acad. Sci. U.S.A. 2022, 119, No. e2116429119.
(34) Edeleva, E.; Salditt, A.; Stamp, J.; Schwintek, P.; Boekhoven, J.;
Braun, D. Continuous nonenzymatic cross-replication of DNA strands
with in situ activated DNA oligonucleotides. Chem. Sci. 2019, 10,
5807−5814.
(35) Kühnlein, A.; Lanzmich, S. A.; Braun, D. tRNA sequences can
assemble into a replicator. eLife 2021, 10, No. e63431.
(36) Orgel, L. E. RNA Catalysis and the Origins of Life. J. Theor.
Biol. 1986, 123, 127−149.
7515
https://doi.org/10.1021/jacs.3c00612
J. Am. Chem. Soc. 2023, 145, 7504−7515
| null |
10.1016_j.isci.2022.104876.pdf
|
Data and code availability
d Microscopy images published in this paper will be shared by the lead contact upon request.
d Original code is uploaded in the supplementary documents and is publicly available as of the date of
publication.
Section 1: Data
All data reported in this paper will be shared by the lead contact upon request.
Section 2: Code
All original code is available in this paper’s supplemental information.
Section 3:
Any additional information required to reanalyze the data reported in this paper is available from the lead
contact upon request.
|
Materials availability This study did not generate new unique reagents. Data and code availability d Microscopy images published in this paper will be shared by the lead contact upon request. d Original code is uploaded in the supplementary documents and is publicly available as of the date of publication.
|
iScience
ll
OPEN ACCESS
Article
Scale space detector for analyzing spatiotemporal
ventricular contractility and nuclear
morphogenesis in zebrafish
Tanveer Teranikar,
Cameron
Villarreal, Nabid
Salehin, ..., Hung
Cao, Cheng–Jen
Chuong, Juhyun
Lee
[email protected]
Highlights
Cardiac defect genes in
humans have
corresponding zebrafish
orthologs
Light sheet modality is
very effective for non-
invasive, 4D modeling of
zebrafish
Hessian detector is robust
to varying nuclei scales
and geometric
transformations
Watershed filter is
effective for separating
fused cellular volumes
Teranikar et al., iScience 25,
104876
September 16, 2022
https://doi.org/10.1016/
j.isci.2022.104876
iScience
ll
OPEN ACCESS
Article
Scale space detector for analyzing spatiotemporal
ventricular contractility and nuclear
morphogenesis in zebrafish
Tanveer Teranikar,1 Cameron Villarreal,1 Nabid Salehin,1 Toluwani Ijaseun,1 Jessica Lim,1 Cynthia Dominguez,1
Vivian Nguyen,2 Hung Cao,3 Cheng–Jen Chuong,1 and Juhyun Lee1,4,5,*
SUMMARY
In vivo quantitative assessment of structural and functional biomarkers is essen-
tial for characterizing the pathophysiology of congenital disorders. In this regard,
fixed tissue analysis has offered revolutionary insights into the underlying
cellular architecture. However, histological analysis faces major drawbacks with
respect to lack of spatiotemporal sampling and tissue artifacts during sample
preparation. This study demonstrates the potential of light sheet fluorescence
microscopy (LSFM) as a non-invasive, 4D (3days + time) optical sectioning tool
for revealing cardiac mechano-transduction in zebrafish. Furthermore, we have
described the utility of a scale and size-invariant feature detector, for analyzing
individual morphology of fused cardiomyocyte nuclei and characterizing zebra-
fish ventricular contractility.
INTRODUCTION
Zebrafish (Danio rero) are emerging as potent vertebrate model’s for modeling human congenital heart
disorders (CHD) (Kula-Alwar et al., 2021, p. 2; Lee et al., 2018; Miura and Yelon, 2011; Vedula et al.,
2017a; Yu and Hwang, 2022; Zhao et al., 2020). This is due to numerous attractive traits such as embryonic
optical transparency, high fecundity, and ease in genetic or biomechanical modulation for mimicking the
human CHD pathophysiology (Choi et al., 2013; Lee et al., 2018; Miura and Yelon, 2011; Rafferty and Quinn,
2018; Tu and Chi, 2012). As a result, zebrafish enable access to phenotypic screening of dynamic biome-
chanical stimuli such as contractility and blood flow, responsible for modulating heart maturation (Kula-
Alwar et al., 2021; Lee et al., 2018; Miura and Yelon, 2011; Tu and Chi, 2012; Vedula et al., 2017a).
Previously conducted zebrafish studies have observed conserved cardiomyocyte count proportional to
atrial/ventricular mass or volume per developmental stage (Kula-Alwar et al., 2021, p. 2). Moreover,
recent studies suggest cardiomyocyte shape hypertrophy across three distinct ventricular regions—atrio-
ventricular (AV) canal, outer curvature (OC), and inner curvature (IC) regions—apart from distinct atrial
cardiomyocyte morphology (Kula-Alwar et al., 2021; Miura and Yelon, 2011; Tu and Chi, 2012).
In
addition, biologists have questioned the implications of cardiac mechano-transduction on enlarged
cardiomyocyte morphology in the OC region in front of AV canal, with respect to spherical (isotropic)
cardiomyocytes in the IC (Tu and Chi, 2012; Zhao et al., 2020). However, the ability to observe cardiomyo-
cyte morphogenesis in vivo is adversely affected by tissue birefringence, hindering characterization of
beforementioned cardiovascular phenotypes (Bensley et al., 2016; Bray et al., 2010; Ghonim et al.,
2017; Teranikar et al., 2020). In this regard, automated feature detectors are proving to be an indispens-
able tool for segmenting cellular volumes without human intervention to avoid gross inconsistencies and
produce refined datasets. (Bolo´ n-Canedo and Remeseiro, 2020; Sargent et al., 2009; Torres and Judson-
Torres, 2019).
Conventionally, invasive sectioning procedures have offered revolutionary insights into aberrant tissue up
to the cellular scale (Bensley et al., 2016; Javaeed et al., 2021; Teranikar et al., 2022). However, histopath-
ological analysis currently suffers from severe limitations, primarily disruption to tissue homeostasis (Klec-
zek et al., 2020; Teranikar et al., 2022). With respect to these drawbacks, the optical sectioning modality
light sheet fluorescence microscopy (LSFM) has proved instrumental in probing dynamic organogenesis
several millimeters inside tissue.(Ding et al., 2017; Fei et al., 2019; Lee et al., 2016; Teranikar et al., 2020;
1Joint Department of
Bioengineering, UT
Arlington/UT Southwestern,
Arlington, TX, USA
2Martin High School/ UT
Arlington, Arlington, TX, USA
3Department of Electrical
Engineering, UC Irvine, Irvine,
CA, USA
4Department of Medical
Education, TCU and
UNTHSC School of Medicine,
Fort Worth, TX 76107, USA
5Lead contact
*Correspondence:
[email protected]
https://doi.org/10.1016/j.isci.
2022.104876
iScience 25, 104876, September 16, 2022
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1
ll
OPEN ACCESS
iScience
Article
Vedula et al., 2017b). LSFM has tremendously benefitted embryologists to become cognizant of dynamic
phenomena such as mechano-transduction and undifferentiated precursor cell signaling pathways.
However, acquisition of dynamic organogenesis reported by endogenous fluorophores is a challenging
task owing to anisotropic contrast across the field of view (FOV). This is due to photon propagation through
heterogeneous tissue (Teranikar et al., 2020, 2021). Hence, precise orchestration of in vivo volumes requires
high sensitivity with respect to dynamic tissue motion and differing scales. As a result, optical aberrations
often induce redundancy in the imaging sample space, affecting interpretability of feature attributes.
Furthermore, cell studies are largely restricted to manual boundary demarcation due to the limited avail-
ability of binary classification methods impervious to heterogeneous contrast resolution (Astrakas and Ar-
gyropoulou, 2010; Marsh et al., 2018; Rajasekaran et al., 2016; Yin et al., 2014). Traditionally intensity-based
segmentation methods such as the Otsu’s method, adaptive thresholding, isodata thresholding, and
entropy-based thresholding have been used for automated cell tracking for their simplicity and speed
(Goh et al., 2018). However, these methods are incapable of separating attenuated objects in closeness,
proximity into meaningful biological regions (Xu et al., 2020). Another conventionally favored approach
for biomedical image segmentation is the watershed algorithm (Beucher and Mathmatique, 2000; Koyuncu
et al., 2012; Rajasekaran et al., 2016; Veta et al., 2013). However, the technique often causes over-segmen-
tation or false detection of non-existent objects in dense tissue (Rajasekaran et al., 2016; Xu et al., 2020).
Hence, there is a clear need for feature detectors impervious to low signal-to-noise ratio bioimages, for
aiding accurate cell segmentation within the tissue architecture.
In this study, we propose the application of a scale space feature detector for isolating fused, myocardial
nuclei blob morphology across distinct embryonic stages (Johnsen, 2000; Johnsen et al., 2011; Johnsen
and Widder, 1999; Teranikar et al., 2020; Zhang et al., 2013). The proposed segmentation framework
integrates Hessian difference of Gaussian (HDoG) feature detector with the watershed algorithm, for
enhancing sensitivity of localizing individual nuclei. In combining these two algorithms, provides for seam-
less and straightforward segmentation with respect to background noise (Bharodiya and Gonsai, 2019).
The proposed algorithm enabled in vivo characterization of wild-type zebrafish ventricular contractility
and morphological traits such as nuclei number, area, and sphericity and in the myocardium.
RESULTS
Isolating individual cardiomyocyte nuclei in dynamic zebrafish ventricular volumes
Distinguishing fused nuclei boundaries by manual contour segmentation or determining the intensity
threshold for distorted nuclei outside the light sheet confocal parameter (focus region) is a complex task
due to varying pixel intensities of overlapping nuclei (Figures 1A–1D). Furthermore, autofluorescence
and dynamic cardiac motion convolutes lateral and axial imaging planes (Figures 1I and 1J). To separate
nuclei clusters into disjoint regions, we implemented the difference of Gaussian (DoG) filter to enhance
cell boundaries followed by the watershed algorithm to separate merged boundaries affected by aniso-
tropic contrast. As a result, we were able to successfully quantify nuclei across different scales in a moder-
ately dense cell environment at 48 h postfertilization (hpf) (Figures 1E–1H). More importantly, integration of
the DoG scale space detector with the watershed algorithm enabled us to split longitudinally merged
nuclei (Figures 1K–1L). In this regard, photon travel through heterogeneous tissue and restrictions imposed
on resolvable sample depth are prone to induce sample de-focus (Figure 2A).
Integration of Hessian and difference of Gaussian (HDoG) to segment cardiomyocyte nuclei
from dense environment
Compared to pinhole-based microscopy techniques, a potential cause of concern for LSFM modality man-
ifests in the form of background contrast between adjacent cardiomyocytes (Figures 2B and 2C). As nuclei
move dynamically across the field of view (FOV), undesired fluorescence emitted from fluorophore-binding
sites outside the optical section beam waist affects accurate volumetric reconstruction (Figure 2A).
Furthermore, intensity attenuation caused by low numerical aperture (NA) objectives aggravates poor
signal-to-noise ratio (SNR).
Although we successfully separated longitudinally merged ventricular myocardial nuclei at 48 hpf using
DoG feature detector, we encountered inaccurate nuclei number quantification beyond 72 hpf. We assume
low pixel intensities produced by the DoG edge detector response prior binarization, resulted in under re-
porting of nuclei (Figures 2D and 2E). To overcome this, we applied the Hessian difference of Gaussian
2
iScience 25, 104876, September 16, 2022
iScience
Article
ll
OPEN ACCESS
Figure 1. Isolating and segmenting cardiomyocyte nuclei from contracting heart using the DOG (Difference of Gaussian) filter in combination with
the watershed algorithm
(A–D) 48 h postfertilization zebrafish ventricular volume was reconstructed using light sheet microscopy, in order to visualize time-dependent motion of
myocardial cardiomyocyte nuclei. Raw volume comprised of fused nuclei clusters (yellow highlighted boxes), exacerbated by tissue scattering (B–D) Zoomed
in regions demonstrate fused contours of nuclei, adversely affecting individual nuclei analysis (E) Ventricular volume was processed using the difference of
Gaussian (DoG) edge detector in conjunction with the watershed algorithm to distinguish individual nuclei from adjacent neighbors.
(F–H) Zoomed in regions show successful separation of nuclei for aiding cell tracking and counting.
(I–J) 2D lateral and axial views illustrate tissue birefringence resulting in merging of nuclei longitudinally (K-L) Segmented lateral and axial views were re-
constructed for qualitative assessment of contour separation of overlapping nuclei. (scale bar = 100 microns), a: atrium, v: ventricle.
detector (HDoG) in combination with the watershed algorithm, for accurate contour separation and assess-
ing the morphology of wild-type myocardial nuclei in vivo. The hessian determinant was used to localize
saddle points (Marsh et al., 2018). Saddle points can be defined as neither an intensity maximum nor min-
imum, that represent connecting nuclei edges. This approach improved detection sensitivity in the pres-
ence of multiple intensity peaks for a single biomarker (Figures 2F and 2G). The segmented labels were
further used for investigating nuclei shape and ventricular contractility, apart from nuclei counting.
Segmentation accuracy evaluation
Nuclei were detected for each distinct developmental phase: 48 hpf (Figures 3A–3C), 72 hpf (Figures 3D–
3F), and 96 hpf (Figures 3G–3I), to compare segmentation robustness for sparse nuclei distribution at 48 hpf
with respect to densely populated ventricle at 96 hpf. We used a segmentation ratio to evaluate segmen-
tation accuracy, by comparing Hessian DoG nuclei images to manual nuclei segmentation, with respect to
static 3D zebrafish heart confocal images. The segmentation ratio is the number of scale space segmented
nuclei divided by the number of cardiomyocyte nuclei manually counted in the confocal images as ground
truth. If the numerical value = 1, the segmentation is identical to the raw images. If the numerical value >1,
there is over-segmentation in the segmented images. If the numerical value <1, there is under segmenta-
tion in the segmented images (Figure S1). Our analysis found that the ideal segmentation was repeated
across developmental stages using the Hessian scale space (Figure S2).
Quantification of local contractility via tracking cardiomyocyte nuclei
After we processed the images to visualize individual nuclei, we performed contractility analysis by tracking
cardiomyocyte nuclei across 48 to 120 hpf to quantify the local cardiac contractility based on this novel seg-
mentation approach (Figures 4A–4G). We investigated the stretch level change of developing zebrafish
iScience 25, 104876, September 16, 2022
3
ll
OPEN ACCESS
iScience
Article
Figure 2. Isolating individual nuclei volumes among high-density cardiomyocyte clusters, at distinct phases of ventricular contraction cycle
(A) Illustration depicting zebrafish ventricular myocardial nuclei sections, scanned by a Gaussian light sheet (blue solid line). There exists a tradeoff between
the confocal parameter i.e. excitation lateral extent and beam waist (BW) i.e. light sheet axial resolution and hence, requires optimization of the Gaussian
focus spot to effectively sample embryogenesis across different growth stages. The detection objective lens modulates effective field of view (FOV). Samples
are scanned through the static optical section at discrete increments (dx) using mechanical transducers, to reconstruct complete in vivo 3days + time
volumes from individual sections. Red arrow represents the blood flow direction of zebrafish heart.
(B and C) Raw systolic and diastolic nuclei reconstruction at 96 h (about 4 days) post fertilization, consisted of closely packed nuclei blobs as compared to 48 h
postfertilization. Inaccurate nuclei localization is further exacerbated by dynamic contraction and relaxation.
(D and E) Application of difference of Gaussian (DoG) detector in conjunction with the watershed algorithm, exhibits reduced feature detection sensitivity
leading to inaccurate reporting of nuclei number.
(F and G) Hessian DoG feature detector exhibits improved sensitivity to local affine transformations experienced by nuclei pixel neighborhoods during
image acquisition. (scale bar = 50 micron), av: atrioventricular canal, v: ventricle, ot: outflow tract.
heart and normalized the temporally changing stretch values for the innermost and outermost curvatures at
each developmental stage (Figure S3). In addition to stretch, we calculated area ratio comparison between
innermost curvature and outermost curvature areas. The area ratio is a description of local deformation of
the area inside of three markers’ 2D stretch ratio. We analyzed the area ratio as a function of time, using
three cardiomyocytes as markers. We found area ratio of the outermost curvature area, where the opposite
side of the atrioventricular canal receiving blood directly from the atrium, has a higher area ratio than the
innermost curvature area of the ventricle (Figures 4H–4J).
Quantifying zebrafish cardiomyocyte nuclei development
The average values for number of nuclei in a developing zebrafish heart were 159 G 13, 222 G 17, 260 G 13,
and 284 G 10 for 48, 72, 96, and 120 hpf, respectively (Figure 5A) (n = 15). We observed cardiomyocyte
nuclei
in outermost curvature had larger systolic and diastolic volumes to innermost curvature area
(Figures 5B–5E, Table S1). Hence, we assume the outermost ventricular curvature experiences higher me-
chanical deformation (Figures 4H and 4I) due to direct inflow of blood from AV canal (Figure 4G), resulting
in larger nuclei volumes.
Contractility effect on morphology of cardiomyocyte nuclei
Apart from area characteristics, we also quantified the circularity of myocardial ventricular nuclei (Figure 6).
Isotropic/spherical nuclei in the innermost curvature region (Figures 6A and 6B) were evaluated to have an
average elongation index of 0.91, while more elongated/ellipsoid nuclei in the outer curvature region
nuclei had an average elongation index of 0.71, suggesting structural anisotropy. Interestingly, nuclei
exhibit distinct eccentricity (major/minor axis ratio) according to their ventricular location, despite dynamic
expansion and contraction across the cardiac cycle (Figures 5B–5E). In this regard, we observed spatially
4
iScience 25, 104876, September 16, 2022
iScience
Article
ll
OPEN ACCESS
Figure 3. Visualizing cmlc:GFPnuc zebrafish ventricular nuclei deformation at distinct developmental stages
(48 – 96 h postfertilization), across the cardiac contraction cycle
(A–I) The Hessian DoG scale space representation was used for localizing cardiomyocyte nuclei ranging from different
sizes, as a result of which we were able to assess ventricular contractility and complex nuclei morphology in vivo (scale bar
for A-C = 100-micron, scale bar for D-I = 50 micron). A:atrium, v:ventricle.
confined cardiomyocyte nuclei in the innermost curvature with shorter major and minor axis lengths,
compared to larger outer curvature nuclei (Figure S4). Hence, we hypothesize that different cardiomyocyte
shapes (Table S1) are modulated by varying contractility in different ventricular regions (Figures 4H and 4I).
This is in accordance with elongated nuclei volumes for accommodating greater mechanical stress in the
outer ventricular concave regions, as compared to smaller nuclei volumes in the inner convex region.
DISCUSSION
Scale space theory can be understood as a hierarchal set of 2D images produced for each optical section, ob-
tained by blurring the image from fine to coarser scale (Lindeberg, 1993; Marsh et al., 2018; Witkin, 1983). This
results in suppression of all image objects equal to the size of the Gaussian kernel. Each defocused image con-
tains a distinct number of edges obtained by blurring unresolved pixel subsets to a coarser resolution. Hence,
enabling multiscale edge visualization without any knowledge of nuclei sizes a priori (Lindeberg, 1999, 2013).
As zebrafish myocardial nuclei length varies spatiotemporally across 2–6 microns (Figure S4), we propose
the integration of scale space theory and watershed segmentation for robust scale-invariant edge
iScience 25, 104876, September 16, 2022
5
ll
OPEN ACCESS
iScience
Article
Figure 4. Selected markers utilized area ratio analysis
(A–F) represents the systolic reconstruction of ventricular myocytes at 48 hpf, 72 hpf, and 96 hpf, respectively, while (D–F) represents the diastolic
reconstruction of myocytes at different developmental stages.
(G and H) Schematic illustrating the nuclei region of interest. Blue windows represent light sheet sections. Zebrafish ventricular volumes were sampled to
compare the innermost curvature contractility (green markers), with respect to the outermost curvature (red markers) (H) Area ratio for innermost curvature
by tracking three cardiomyocytes highlighted green in the blue optical plane, which elucidate increasing contractility trend observed across distinct
developmental stages.
(I) The area ratio for outermost curvature calculated by tracking cardiomyocytes highlighted red in the blue optical plane, indicates the outermost curvature
has higher contractility compared to the innermost curvature. (J) Outermost curvature has a significantly higher area ratio compared to innermost curvature
after 72 hpf (n = 3, p = 0.05, one-tail t-test).
detection. Utilizing the inherent de-focus adaptation ability of the HDoG blob detector, we successfully
isolated individual centers of mass (Videos S1, Video S2) for tracking dynamic cardiomyocyte nuclei (Video
S3, Video S4). As a result, we were able to successfully characterize ventricular myocardial stretch post AV
valve specification to heart maturation (Kula-Alwar et al., 2021; Miura and Yelon, 2011). Although transpar-
ency was induced in zebrafish using PTU, tissue birefringence (RI(cid:1)1.3–1.5 (Jing et al., 2018)) results in
changes in optical path lengths of emitted photons (Johnsen, 2000; Johnsen et al., 2011; Johnsen and Wid-
der, 1999; Teranikar et al., 2020). Consequently, the light scatter compromises the optical modality pene-
tration capability, resulting in fusing of nuclei situated outside the confocal region (Figure 2A) (Teranikar
et al., 2020). Taking this into consideration, we sought to design an automated blob detection framework
that provides high sensitivity and repeatability for a singular Gaussian intensity peak detection correspond-
ing to each nuclei centroid. Furthermore, the proposed framework enabled in vivo quantification of
morphological descriptors such as nuclei volume, surface area, and shape.
6
iScience 25, 104876, September 16, 2022
iScience
Article
ll
OPEN ACCESS
Figure 5. Zebrafish cardiomyocyte nuclei analysis
(A and B) We observed an increase in the number of ventricular cardiomyocyte nuclei for successive developmental stages. Asterisk denotes statistically
significant difference with respect to previous time point. p % 0.05 (B) Systolic and diastolic nuclei volume expansion observed for the inner curvature.
(C) Systolic and diastolic volume trends observed for the outer curvature. (D and E) Systolic and diastolic nuclei surface area growth observed for the inner
curvature (E) Systolic and diastolic nuclei surface area observed for the outer curvature. n = 15.
Wild-type Tg(cmlc2:egfp) zebrafish have been observed to exhibit cuboidal cardiomyocyte morphology in linear
heart tube (24 hpf) and IC ventricular myocardium, with respect to elongated cardiomyocytes in the OC (Auman
et al., 2007; Kula-Alwar et al., 2021; Miura and Yelon, 2011). Studies indicate regionally distinct cardiomyocyte
phenotypes such as cell count, area, or sphericity are regulated by mechanical stimuli such as contractility or
blood flow during heart maturation (Auman et al., 2007; Miura and Yelon, 2011). This has been validated through
the mutation phenotype half-hearted (haf) mutation lacking ventricular contractility. The haf mutant exhibited
elongated cardiomyocytes with increased surface area across different parts of the ventricle including IC, result-
ing in a distended ventricle (Auman et al., 2007). Interestingly, cardiomyocyte count was observed to be consis-
tent between the haf mutant and wild-type zebrafish ventricle, suggesting contractility is responsible for moder-
ating the aberrant elongation of cardiomyocytes and not anomalous proliferation (Auman et al., 2007).
Furthermore, previously performed studies (Auman et al., 2007) indicate cardiomyocyte number reflects a
sigmoidal growth trend (Figure 5), subsequently plateauing at later stages (>96 hpf) thereby signaling specifica-
tion into the myocardium. In this regard, the proposed feature detector and cell tracking algorithm can prove
extremely beneficial for gaining insights into the effects of cardiac contraction on reducing proliferation and
its secondary effects on cardiomyocyte morphology in zebrafish. Unfortunately, currently we cannot conclude
that contraction is key to reducing the proliferation of cardiomyocytes due to lack of statistically significant
data. However, the intensity of mechanical workload experienced by cardiomyocytes in different parts of the
ventricle appears to be a regulatory mechanism for maturation into distinct shapes (Figures 4, Figure 6).
Analyzing the cardiomyocyte motion (Figure S3), we quantified the outermost curvature has higher area
ratio than the innermost curvature (Figures 4H and 4I), thereby experiencing greater mechanical workload.
Furthermore, we observed elongated cardiomyocyte nuclei morphology in the OC with respect to spher-
ical morphology in the IC. Although no phenotyping screening of nuclei morphology has been performed
with respect to modulation of contractility, our data indicate that cardiomyocyte nuclei shape and size cor-
responds to deformation experienced by distinct ventricular regions. Elongated nuclei (Figure 6) in the OC
suggest larger surface area is required to accommodate greater OC mechanical intensity. On the other
hand, IC consists of smaller, cuboidal nuclei due to lesser deformation compared to OC. In the develop-
mental biology aspect, researchers primarily focus on the ventricular OC where trabeculae form, but there
is lack of well-documented research regarding lack of trabeculae in the IC. Thus, the question remains how
different biomechanical or molecular signaling engenders a trabeculated OC and smooth IC. Our study has
iScience 25, 104876, September 16, 2022
7
ll
OPEN ACCESS
iScience
Article
Figure 6. Systole vs diastole circularity analysis
(A) Inner curvature nuclei are observed to have a circular shape, (symmetric circle elongation = 1, ellipse <1) with slightly higher values observed for the diastole.
(B) Outer curvature cardiomyocyte nuclei are observed to have an elongated shape with higher elongation observed in the diastole.
(C and D) Volumetric reconstructions of the circular shape of inner curvature and elliptic shape of outer curvature myocytes were visually presented,
respectively. In addition, the corresponding lateral and axial views are shown with binary images (scale bar = 15 um).
the potential to elucidate ventricular development in zebrafish orthologs, and aid cardiac pathophysiology
diagnosis or clinical translational of cardiac regeneration for pediatric population. However, further inves-
tigations will be required to validate this assumption. In this regard, nuclear morphology observed in car-
diomyocytes isolated from neonatal rat ventricles reports similar findings, regarding systolic and diastolic
heterogeneous cross-sectional surface areas due to deformation experienced by the cardiac cycle (Bray
et al., 2010). Hence, our novel study provides exciting avenues to characterize cell count, morphology,
and intercellular forces that may be responsible for cardiomyopathy in humans. Future studies will involve
modulation of contractility to characterize cardiomyocyte morphology in the IC and OC.
In summary, we have presented a scale-invariant feature detector for quantifying individual morphological
characteristics of merged nuclei and biomechanical analysis of the zebrafish ventricle. Our proposed blob
detection and cell tracking approach will prove to be extremely beneficial for analyzing cell count, volume,
area, sphericity, proliferation, or cardiac function for characterization of cardiomyopathy phenotypes.
Conclusion
In this report, we were able to successfully interrogate dynamic zebrafish cardiac tissue non-invasively using
bona fide biomarkers such as cell elongation, volume, and surface area. Moreover, we quantified the num-
ber of cells and the mechanical workload experienced by the ventricular inflow and outflow regions during
the systole and diastole, respectively.
Limitations of the study
Although we successfully separated merged nuclei clusters across varying scales and densities, the reproduc-
ibility of the Hessian DoG feature detector is highly dependent on appropriate identification of Gaussian blurring
weights (Figure S1). Moreover, Hessian scale space detector followed by watershed postprocessing is more
prone to over-segmentation with higher variability in nuclei count, in comparison to DoG feature detection
8
iScience 25, 104876, September 16, 2022
iScience
Article
ll
OPEN ACCESS
(Figure S1) if kernel weights are not selected appropriately. On the other hand, DoG scale space detector is
inherently prone to erosion of boundaries due to bandpass operation, resulting in reduction of nuclei volumes
and the object area affecting quantification. Other modality limitations include absence of peripheral nuclei dur-
ing diastole that may be present during the systole, due to ballooning of ventricle outside the light sheet
confocal region. As in vivo cardiomyocyte cell tracking and counting requires invariancy to sample translation
without distortion in shape, the Hessian DoG operation was effectively used to localize individual nuclei based
on pixel intensity gradients.
STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:
d KEY RESOURCES TABLE
d RESOURCE AVAILABILITY
B Lead contact
B Materials availability
B Data and code availability
d EXPERIMENTAL MODEL AND SUBJECT DETAILS
d METHOD DETAILS
B Light sheet microscope (LSFM) implementation
B Preparation of zebrafish for assessing cardiac function
B Image processing framework
B Cell counting and area measurements
B Cardiac myocyte nuclei tracking
B Contractility analysis
d QUANTIFICATION AND STATISTICAL ANALYSIS
SUPPLEMENTAL INFORMATION
Supplemental information can be found online at https://doi.org/10.1016/j.isci.2022.104876.
ACKNOWLEDGMENTS
The authors would like to express gratitude to Dr. Caroline Burns and Geoffrey Burns from Boston Chil-
dren’s Hospital for providing Tg(cmlc:nucGFP) for imaging and analysis. This study was supported by
grants from AHA 18CDA34110150 (J.L.) and NSF 1936519 (J.L.).
AUTHOR CONTRIBUTIONS
Methodology and visualization, T.T. and J.L. Conceptualization, investigation, software and validation,
T.T., C.L., N.S., CJ-C, and J.L. Writing – Original draft, T.T., C.L., and J.L. Writing - Review and editing,
T.T., T.I., J.L., C.D., V.N., H.C., CJ-C, and J.L. Supervision, T.T., H.C., CJ-C, and J.L. Funding acquisition, J.L.
Received: May 13, 2021
Revised: April 1, 2022
Accepted: July 29, 2022
Published: September 16, 2022
REFERENCES
Astrakas, L.G., and Argyropoulou, M.I. (2010).
Shifting from region of interest (ROI) to voxel-
based analysis in human brain mapping. Pediatr.
Radiol. 40, 1857–1867. https://doi.org/10.1007/
s00247-010-1677-8.
Auman, H.J., Coleman, H., Riley, H.E., Olale, F.,
Tsai, H.-J., and Yelon, D. (2007). Functional
modulation of cardiac form through regionally
confined cell shape changes. PLoS Biol. 5, e53.
https://doi.org/10.1371/journal.pbio.0050053.
Bensley, J.G., De Matteo, R., Harding, R.,
and Black, M.J. (2016). Three-dimensional
direct measurement of cardiomyocyte
volume, nuclearity, and ploidy in thick
histological sections. Sci. Rep. 6, 23756.
https://doi.org/10.1038/srep23756.
Bolo´ n-Canedo, V., and Remeseiro, B. (2020).
Feature selection in image analysis: a survey.
Artif. Intell. Rev. 53, 2905–2931. https://doi.org/
10.1007/s10462-019-09750-3.
Beucher, S., and Mathmatique, C. (2000).
The watershed transformation applied
to image segmentation. Scanning Microsc. 6.
Bharodiya, A.K., and Gonsai, A.M. (2019). An
improved edge detection algorithm for X-Ray
images based on the statistical range. Heliyon 5,
e02743. https://doi.org/10.1016/j.heliyon.2019.
e02743.
Bray, M.-A.P., Adams, W.J., Geisse, N.A.,
Feinberg, A.W., Sheehy, S.P., and Parker, K.K.
(2010). Nuclear morphology and deformation in
engineered cardiac myocytes and tissues.
Biomaterials 31, 5143–5150. https://doi.org/10.
1016/j.biomaterials.2010.03.028.
Choi, W.-Y., Gemberling, M., Wang, J., Holdway,
J.E., Shen, M.-C., Karlstrom, R.O., and Poss, K.D.
iScience 25, 104876, September 16, 2022
9
iScience
Article
the ventricle. Dev. Biol. 470, 95–107. https://doi.
org/10.1016/j.ydbio.2020.11.008.
APL Bioeng. 4, 036103. https://doi.org/10.1063/1.
5144613.
Lee, J., Fei, P., Packard, R.R.S., Kang, H., Xu, H.,
Baek, K.I., Jen, N., Chen, J., Yen, H., Kuo, C.C.J.,
et al. (2016). 4-Dimensional light-sheet
microscopy to elucidate shear stress modulation
of cardiac trabeculation. J. Clin. Invest. 126, 1679–
1690. https://doi.org/10.1172/JCI83496.
Teranikar, T., Villarreal, C., Salehin, N., Lim, J.,
Ijaseun, T., Cao, H., Chuong, C., and Lee, J.
(2021). Feature Detection to Segment
Cardiomyocyte Nuclei for Investigating Cardiac
Contractility. Preprint at bioRxiv. https://doi.org/
10.1101/2021.03.03.433810.
Lee, J., Vedula, V., Baek, K.I., Chen, J., Hsu, J.J.,
Ding, Y., Chang, C.-C., Kang, H., Small, A., Fei, P.,
et al. (2018). Spatial and temporal variations in
hemodynamic forces initiate cardiac
trabeculation. JCI Insight 3, e96672. https://doi.
org/10.1172/jci.insight.96672.
Lindeberg, T. (2013). Scale selection properties of
generalized scale-space interest point detectors.
J. Math. Imaging Vis. 46, 177–210. https://doi.
org/10.1007/s10851-012-0378-3.
Lindeberg, T. (1999). Principles for automatic
scale selection. In Handbook on Computer Vision
and Applications, B. Ja¨ hne, ed. (Academic Press),
pp. 239–274.
Lindeberg, T. (1993). Detecting salient blob-like
image structures and their scales with a scale-
space primal sketch: a method for focus-of-
attention. Int. J. Comput. Vis. 11, 283–318.
https://doi.org/10.1007/BF01469346.
Marsh, B.P., Chada, N., Sanganna Gari, R.R.,
Sigdel, K.P., and King, G.M. (2018). The
hessian blob algorithm: precise particle
detection in atomic force microscopy
imagery. Sci. Rep. 8, 978. https://doi.org/10.
1038/s41598-018-19379-x.
Miura, G.I., and Yelon, D. (2011). Chapter 7 - a
guide to analysis of cardiac phenotypes in the
zebrafish embryo. In Methods in Cell Biology,
H.W. Detrich, M. Westerfield, and L.I. Zon, eds.
(Academic Press), pp. 161–180. https://doi.org/
10.1016/B978-0-12-387036-0.00007-4.
Rafferty, S.A., and Quinn, T.A. (2018). A beginner’s
guide to understanding and implementing the
genetic modification of zebrafish. Prog. Biophys.
Mol. Biol. 138, 3–19. https://doi.org/10.1016/j.
pbiomolbio.2018.07.005.
Rajasekaran, B., Uriu, K., Valentin, G., Tinevez,
J.-Y., and Oates, A.C. (2016). Object
segmentation and ground truth in 3D embryonic
imaging. PLoS One 11, e0150853.
Sargent, D., Chen, C.-I., Tsai, C.-M., Wang, Y.-F.,
and Koppel, D. (2009). Feature detector and
descriptor for medical images. In Medical
Imaging 2009: Image Processing, J.P.W. Pluim
and B.M. Dawant, eds. (SPIE), pp. 991–998.
https://doi.org/10.1117/12.811210.
Teranikar, T., Lim, J., Ijaseun, T., and Lee, J.
(2022). Development of planar illumination
strategies for solving mysteries in the sub-cellular
realm. Int. J. Mol. Sci. 23, 1643. https://doi.org/10.
3390/ijms23031643.
Teranikar, T., Messerschmidt, V., Lim, J., Bailey,
Z., Chiao, J.-C., Cao, H., Liu, J., and Lee, J. (2020).
Correcting anisotropic intensity in light sheet
images using dehazing and image morphology.
Torres, R., and Judson-Torres, R.L. (2019).
Research techniques made simple: feature
selection for biomarker discovery. J. Invest.
Dermatol. 139, 2068–2074.e1. https://doi.org/10.
1016/j.jid.2019.07.682.
Tu, S., and Chi, N.C. (2012). Zebrafish models in
cardiac development and congenital heart birth
defects. Differentiation. 84, 4–16. https://doi.org/
10.1016/j.diff.2012.05.005.
Vedula, V., Lee, J., Xu, H., Kuo, C.-C.J., Hsiai, T.K.,
and Marsden, A.L. (2017a). A method to quantify
mechanobiologic forces during zebrafish cardiac
development using 4-D light sheet imaging and
computational modeling. PLoS Comput. Biol. 13,
e1005828. https://doi.org/10.1371/journal.pcbi.
1005828.
Vedula, V., Lee, J., Xu, H., Kuo, C.-C.J., Hsiai, T.K.,
and Marsden, A.L. (2017b). A method to quantify
mechanobiologic forces during zebrafish cardiac
development using 4-D light sheet imaging and
computational modeling. PLoS Comput. Biol. 13,
e1005828–24. https://doi.org/10.1371/journal.
pcbi.1005828.
Veta, M., van Diest, P.J., Kornegoor, R., Huisman,
A., Viergever, M.A., and Pluim, J.P.W. (2013).
Automatic nuclei segmentation in H&E stained
breast cancer histopathology images. PLoS One
8, e70221.
Witkin, A. (1983). Scale-Space Filtering (IJCAI).
Xu, Y., Wu, T., Gao, F., Charlton, J.R., and Bennett,
K.M. (2020). Improved small blob detection in 3D
images using jointly constrained deep learning
and Hessian analysis. Sci. Rep. 10, 326. https://doi.
org/10.1038/s41598-019-57223-y.
Yin, X., Ng, B.W.-H., He, J., Zhang, Y., and Abbott,
D. (2014). Accurate image analysis of the retina
using hessian matrix and binarisation of
thresholded entropy with application of texture
mapping. PLoS One 9. e95943–17. https://doi.
org/10.1371/journal.pone.0095943.
Yu, H.-L., and Hwang, S.-P.L. (2022). Zebrafish
integrin a3b is required for cardiac contractility
and cardiomyocyte proliferation. Biochem.
Biophys. Res. Commun. 595, 89–95. https://doi.
org/10.1016/j.bbrc.2022.01.083.
Zhang, H., Li, Y., Chen, H., Yuan, D., and Sun, M.
(2013). Perceptual contrast enhancement with
dynamic range adjustment. Optik 124, 5906–
5913. https://doi.org/10.1016/j.ijleo.2013.04.046.
Zhao, M.-T., Ye, S., Su, J., and Garg, V. (2020).
Cardiomyocyte proliferation and maturation: two
sides of the same coin for heart regeneration.
Front. Cell Dev. Biol. 8, 594226. https://doi.org/
10.3389/fcell.2020.594226.
ll
OPEN ACCESS
(2013). In vivo monitoring of cardiomyocyte
proliferation to identify chemical modifiers of
heart regeneration. Development 140, 660–666.
https://doi.org/10.1242/dev.088526.
Ding, Y., Lee, J., Ma, J., Sung, K., Yokota, T.,
Singh, N., Dooraghi, M., Abiri, P., Wang, Y.,
Kulkarni, R.P., et al. (2017). Light-sheet
fluorescence imaging to localize cardiac lineage
and protein distribution. Sci. Rep. 7, 42209.
https://doi.org/10.1038/srep42209.
Fei, P., Nie, J., Lee, J., Ding, Y., Li, S.,
Zhang, H., Hagiwara, M., Yu, T., Segura, T.,
Ho, C.-M., et al. (2019). Subvoxel
light-sheet
microscopy for high-resolution high-
throughput volumetric imaging of large
biomedical specimens. Adv. Photonics 1, 1.
Ghonim, S., Voges, I., Gatehouse, P.D., Keegan,
J., Gatzoulis, M.A., Kilner, P.J., and Babu-
Narayan, S.V. (2017). Myocardial architecture,
mechanics, and fibrosis in congenital heart
disease. Front. Cardiovasc. Med. 4, 30.
Goh, T.Y., Basah, S.N., Yazid, H., Aziz Safar, M.J.,
and Ahmad Saad, F.S. (2018). Performance
analysis of image thresholding: Otsu technique.
Measurement 114, 298–307. https://doi.org/10.
1016/j.measurement.2017.09.052.
Javaeed, A., Qamar, S., Ali, S., Mustafa, M.A.T.,
Nusrat, A., and Ghauri, S.K. (2021). Histological
stains in the past, present, and future. Cureus 13,
e18486. https://doi.org/10.7759/cureus.18486.
Jing, D., Zhang, S., Luo, W., Gao, X., Men, Y., Ma,
C., Liu, X., Yi, Y., Bugde, A., Zhou, B.O., et al. (2018).
Tissue clearing of both hard and soft tissue organs
with the PEGASOS method. Cell Res. 28, 803–818.
https://doi.org/10.1038/s41422-018-0049-z.
Johnsen, S. (2000). Transparent animals. Sci. Am.
282, 80–89.
Johnsen, S., Marshall, N.J., and Widder, E.A.
(2011). Polarization sensitivity as a contrast
enhancer in pelagic predators: lessons from in situ
polarization imaging of transparent zooplankton.
Philos. Trans. R. Soc. Lond. B Biol. Sci. 366,
655–670. https://doi.org/10.1098/rstb.2010.0193.
Johnsen, S., and Widder, E. (1999). The physical
basis of transparency in biological tissue:
ultrastructure and the minimization of light
scattering. J. Theor. Biol. 199, 181–198. https://
doi.org/10.1006/jtbi.1999.0948.
Kleczek, P., Jaworek-Korjakowska, J., and
Gorgon, M. (2020). A novel method for tissue
segmentation in high-resolution H&E-stained
histopathological whole-slide images. Comput.
Med. Imag. Graph. 79, 101686. https://doi.org/
10.1016/j.compmedimag.2019.101686.
Koyuncu, C.F., Arslan, S., Durmaz, I., Cetin-Atalay,
R., and Gunduz-Demir, C. (2012). Smart markers
for watershed-based cell segmentation. PLoS
One 7, e48664. https://doi.org/10.1371/journal.
pone.0048664.
Kula-Alwar, D., Marber, M.S., Hughes, S.M., and
Hinits, Y. (2021). Mef2c factors are required for
early but not late addition of cardiomyocytes to
10
iScience 25, 104876, September 16, 2022
iScience
Article
STAR+METHODS
KEY RESOURCES TABLE
ll
OPEN ACCESS
REAGENT or RESOURCE
SOURCE
IDENTIFIER
Chemicals, peptides, and recombinant proteins
0.0025% 1-phenyl 2-thoiurea
Sigma-Aldrich
0.05% tricaine (MS 222)
Sigma-Aldrich
P7629
E10521
Experimental models: Organisms/strains
Zebrafish: Tg(cmlc2:nucGFP)
Software and algorithms
Sharpe M et al.
Gifted by Dr Barnes at Boston children’s hospital, Harvard Medical.
ImageJ
Schneider et al., 2012
https://imagej.nih.gov/ij/
Hessian Determinant plugin
Sato, Y. et al.
https://imagescience.org/meijering/software/featurej/
Other
Cardiomyocyte nuclei tracking code
This paper
RESOURCE AVAILABILITY
Lead contact
Chuong CJ, Sacks MS, Templeton G, Schwiep F, Johnson RL Jr. Regional deformation,
and contractile function in canine right ventricular free wall. Am J Physiol. 1991;260(4
Pt 2):H1224-H1235. https://doi.org/10.1152/ajpheart.1991.260.4.H1224
Further information and requests for resources and reagents should be directed to and will be fulfilled by
the lead contact, Dr Juhyun Lee ([email protected]).
Materials availability
This study did not generate new unique reagents.
Data and code availability
d Microscopy images published in this paper will be shared by the lead contact upon request.
d Original code is uploaded in the supplementary documents and is publicly available as of the date of
publication.
Section 1: Data
All data reported in this paper will be shared by the lead contact upon request.
Section 2: Code
All original code is available in this paper’s supplemental information.
Section 3:
Any additional information required to reanalyze the data reported in this paper is available from the lead
contact upon request.
EXPERIMENTAL MODEL AND SUBJECT DETAILS
The animal experiments were performed in agreement with the UT Arlington Institutional Animal Care and
Use Committee (IACUC) protocol (#A17.014). The transgenic zebrafish line used in this particular study is
the Tg(cmlc2:nucGFP), with the cardiomyocyte nuclei labeled with GFP (Green Fluorescent Protein). The
zebrafish embryos were maintained at 28.5(cid:3)C in system water at the UT Arlington Aquatic Animal Core
Facility. 0.0025% 1-phenyl 2-thoiurea was added to the embryo medium starting around 20–24 hpf to sup-
press pigmentation. Prior to imaging, embryos were anesthetized in 0.05% tricaine (MS 222, E10521,
Sigma-Aldrich, St-Louis, MO) to avoid sample movement. Sex determination and segregation of zebrafish
was not performed in the embryonic stage (48–120 h postfertilization).
iScience 25, 104876, September 16, 2022
11
ll
OPEN ACCESS
iScience
Article
METHOD DETAILS
Light sheet microscope (LSFM) implementation
Our home built light sheet microscope consists of a single-side excitation pathway and a custom water
dipping lens (20x/0.5 NA UMPlanFL N, Olympus, Tokyo, Japan) detection. In the illumination pathway,
a cylindrical lens (LJ1695RM, Thorlabs) coupled with a 4x objective lens (4X Plan Apochromat Plan N,
light-sheet with (cid:1)4-5-micron thickness.
Olympus, Tokyo, Japan), are used to collimate a cylindrical
Furthermore, a mechanical slit aperture (VA100C, Thorlabs) is modulated across distinct developmental
to accommodate ventricular circumferential extent across the
stages (48-,72-,96- and 120 hpf),
light sheet confocal region (Figure 2). A DC servo motor actuator (Z825B, Thorlabs)
is used for
sample translation in the axial direction (z-step velocity and acceleration = 0.005 mm/s). The optical
detection pathway consisting of the water lens, infinity corrected tube lens (TTL 180-A, Thorlabs) and
sCMOS camera (ORCA flash 4.0, Hamamatsu, Japan, camera pixel size = [6.5 (um)^2] = [6.5/20x =
0.325 um], camera exposure time = 30–50 ms), is used for non-gated 4D (3D + time) cardiac volume
acquisition.
As the zebrafish ventricle undergoes periodic deformation during peak systole to end diastole, optical
sections were acquired at varying depths in the sample, covering 4–5 cardiac cycles20,23. Since triggering
of image slices is not synchronized to a particular phase in the cardiac cycle, we performed volumetric
reconstruction a posteriori to ensure alignment of adjacent optical sections. For this purpose, we
estimated the period of each individual cycle by minimization of the least squares intensity difference
criterion and calculated the relative period shift to ensure synchronization between independent cardiac
cycles20,23
Preparation of zebrafish for assessing cardiac function
The animal experiments were performed in agreement with the UT Arlington Institutional Animal Care
and Use Committee (IACUC) protocol (#A17.014). The transgenic zebrafish line used in this particular
study is the Tg(cmlc2:nucGFP), with the cardiomyocyte nuclei
labeled with GFP (Green Fluorescent
Protein)19. The zebrafish embryos were maintained at 28.5(cid:3)C in system water at the UT Arlington Aquatic
Animal Core Facility. 0.0025% 1-phenyl 2-thoiurea was added to the embryo medium starting at 20–24
hpf 4,32 to suppress pigmentation. Prior to imaging, embryos were anesthetized in 0.05% tricaine (MS
222, E10521, Sigma-Aldrich, St-Louis, MO) to avoid sample movement. Upon administering the anes-
thetic, alive embryos were embedded in 0.5% low-melt agarose gel inside a fluorinated ethylene propyl-
ene (FEP) tube (1677L, IDEX, Chicago,IL). Furthermore, the FEP tube was suspended in water within a
custom 3days printed ABS (Acrylonitrile Butadiene Styrene) cuvette (designed using solid works) housing
the water dipping lens, to ensure near isotropic refractive index between the water dipping lens and
sample inside the tube. (Refractive index of water = 1.33, refractive index of agarose and FEP tube =
1.34). Refractive index matching is necessary to avoid distortions and intensity attenuation in the optical
sections13.
Image processing framework
Haze removal using the dark channel prior (DCP) method
Introduction of haze by the ambient medium or scattering due to particulate matter, degrades the perfor-
mance of computer vision tasks(Lee et al., 2016; Teranikar et al., 2020). A haze free image can be retrieved
by using the image degradation model based on the Dark Channel Prior (DCP) algorithm.
IðxÞ = JðxÞ:tðxÞ + Að1 (cid:4) tðxÞÞ
(Equation 1)
(Lee et al., 2016; Teranikar et al., 2020)
where I(x) is the degraded image, J(x) is the original irradiance captured by the CMOS camera, t(x) repre-
sents the scene depth and A is the scattering introduced by the ambient light. Using the dehazing
algorithm, we estimated the intensity transmission map t(x) using the imreducehaze() MATLAB function
(Teranikar et al., 2020).
tðxÞ = e(cid:4) bdðxÞ
(Equation 2)
(Lee et al., 2016)
12
iScience 25, 104876, September 16, 2022
iScience
Article
ll
OPEN ACCESS
where b represents the scattering coefficient and d represents the scene depth. We used the estimated
intensity transmission map as a preprocessing step before performing the DoG operation. By estimating
the contrast attenuation with respect to distance, we were able to emphasize edges.
Intensity maxima localization at nuclei centers using the difference of Gaussian (DoG) filter
The DoG filter can be effectively used to enhance edge visualization for images suffering from poor
contrast. In this study, the greyscale bandpass operation is performed by subtracting a blurred version
of the transmission estimate from a lesser blurred version of itself,
tðxÞ (cid:5) g1ðxÞ (cid:4) tðxÞ (cid:5) g2ðxÞ = tðxÞ (cid:5) ðg1ðxÞ (cid:4) g2ðxÞÞ;
(Equation 3)
where g1(x) and g2(x) are the Gaussian kernels having different standard deviations. Using the DoG filter,
we were able to localize blobs to nuclei centers by isolating spatial frequencies correlating to the Gaussian
illumination maxima.
Precise contour delineation using the hessian scale space representation and watershed algorithm
The hessian scale space representation can be described by the convolution:
Dðx; y; tÞ = ½tðxÞ (cid:5) ðg1ðxÞ (cid:4) g2ðxÞÞ(cid:6) (cid:5) Gðx; y; tÞ;
(Equation 4)
(Marsh et al., 2018; Rajasekaran et al., 2016)
Where D(x,y,t) represents the family of images, derived from the original image. t represents the degree of
blurring. Hence, Equation (4) can be described as the convolution of the DoG blob maxima image with the
hessian blob detector Gaussian blur kernel G(x,y) at different degrees of blur (t > 0). The blurring scale
selection was based on the ratio t +1 = r *t (Marsh et al., 2018), where r is a constant.
The workflow involved for the hessian blob involves (Marsh et al., 2018),(Rajasekaran et al., 2016),
1) Computing the absolute magnitude of the intensity gradient image obtained by convolving the DoG
bandpass image with the derivative of Gaussian filter.
2) Computing the double derivative of the absolute magnitude image calculated in the previous step
3)
Imposing boundary conditions on the hessian determinant value [det D(x,y,t) < 0] (Rajasekaran et al.,
2016) at every pixel, for indicating saddle points.
The image arithmetic operation (OR – operation) results in the union of the DoG localized intensity maxima
and contour information from the Hessian blob, aiding the successful splitting of nuclei.
Preprocessing strategies
Images corrupted by noise or tissue scatter, were filtered by using a Gaussian kernel with an appropriate SD
followed by the background subtract operation in ImageJ. In addition, image processing code is enclosed
(Data S2, related to Figure 2).
Cell counting and area measurements
After converting raw optical images to binary images, we performed to count cardiomyocyte nuclei and
their area analysis by using the 3D object counter plugin33 in ImageJ. The plugin can be accessed by: Im-
ageJ – Analyze – 3D Object Counter. After cropping the ROI (ventricle in this case), we used the plugin to
quantify number of object voxels (volume), surface voxels of individual nuclei volumes and the number of
3D nuclei objects in the ventricular stack. The plugin can also be used to retrieve the centroid geometric
coordinates of object volumes. The user
is required configure 2 important parameters namely,
(a) intensity threshold to separate background and foreground pixel populations and (b) size threshold
to exclude smaller objects from the analysis. The plugin allows user to configure object counting based
on the presence or absence of touching edges.
iScience 25, 104876, September 16, 2022
13
ll
OPEN ACCESS
iScience
Article
Cardiac myocyte nuclei tracking
We utilized the segmented, processed, and time synchronized images to reconstruct three-dimensional
volumes through time for a cardiac cycle to perform this tracking. We then passed these images through
a custom MATLAB (Mathworks) code to perform for key steps (Data S3, related to Figure 4).
This MATLAB code performed following 4 steps.
1 The code compiles images into easily searchable 4D matrices.
2 The code resolves the 4D matrices of segmented images into centers of mass based on high pixel
concentration areas for each time step.
3 The user selects three markers to represent our plane for stretch calculations.
4 The code searches through the 4D stack of centers of mass to determine the closest center of mass in
the next time step and stores these points in a matrix of position values.
Each stored triplet value is the x, y, and z position of a particular nucleus at a particular time. This format is
easily searchable and allows for a multitude of calculations. This code assumes that there can be no erratic
motion of the nucleus with a high enough sampling frequency. The location at each time step depends on
the prior location. Imaging with a high sampling frequency supply data that meets this assumption require-
ment. Other works have utilized similar works, including Meijerling et al. (Meijering et al., 2009).Drawbacks
of this method include the requirement for user interaction. To verify that the cell tracking occurs appro-
priately, the user must analyze each vector to ensure the vector does not violate the small motion assump-
tion. This process can become time-consuming and increases the chance of human error. Subsequent work
can expand and refine this cell tracking method to include other parameters, including a probability net for
machine learning applications and size and orientation to decrease ambiguity and reduce the user input
requirement.
Contractility analysis
We selected and tracked three cardiomyocyte nuclei for both the innermost and outermost curvature. After
tracking the location of three cardiomyocytes through each time instance, we utilized the following method
to determine the deformation gradient with normalized one cardiac cycle as 0.5 s starting from ventricular
end-systolic stage. We determined the stretch ratio at each time instance into principal stretch values re-
ported as l1 and l2, or the longitudinal and circumferential principal direction followed by previous
methods41,42. These principal stretch vectors correspond to the first and second principal strain directions.
When viewed on Mohr’s circle, they correspond to the maximum and minimum normal strain values where
the shear strain is resolved to zero (Figure S6). These values are represented in the Cartesian coordinate
system as the x and y direction or in polar coordinates as the zero-degree rotation and 90-degree rotation.
We established the area ratio by multiplying the two principal stretch values. Area ratio provides a
description of the total in plane deformation from the initial undeformed state which was selected as
the start of filling.
QUANTIFICATION AND STATISTICAL ANALYSIS
For statistical analysis, we performed ad hoc pairwise comparisons for three morphological parameters to
characterize the maturity of the heart (p value = 0.05)10. We analyzed the number of visible nuclei, the total
volume, and total surface area. We estimated each of these parameters using built in functions in ImageJ
(NIH, Bethesda, MD) with n = 15. Additionally, we cleaned the data in excel utilizing Chauvenet’s criterion
to determine which values were outliers and should be removed. After removing outliers and cleaning the
datasets in excel to reduce the chance of error due to our sampling technique, we compared the data with
one-way ANOVA. If we detected a statistically significant difference for any comparison, we performed
Tukey’s test for multiple comparison of means. This test inherently compensates for multiple comparisons,
which allowed us to use an alpha value of 0.5. All values herein are reported as mean +/(cid:4) standard devi-
ation in the figures and respective figure legends.
14
iScience 25, 104876, September 16, 2022
| null |
10.1088_1361-6641_acfa1f.pdf
|
Data availability statement
All data that support the findings of this study are included
within the article (and any supplementary files).
|
Data availability statement All data that support the findings of this study are included within the article (and any supplementary files).
|
Semicond. Sci. Technol. 38 (2023) 115004 (6pp)
Semiconductor Science and Technology
https://doi.org/10.1088/1361-6641/acfa1f
MISHEMT intrinsic voltage gain under
multiple channel output characteristics
Bruno Godoy Canales1,∗, Welder Fernandes Perina1, Joao Antonio Martino1,
Eddy Simoen2, Uthayasankaran Peralagu2, Nadine Collaert2
and Paula Ghedini Der Agopian1,3
1 LSI/PSI/USP, University of Sao Paulo, Sao Paulo, Brazil
2 Imec, Leuven, Belgium
3 UNESP, Sao Paulo State University, Sao Joao da Boa Vista, Brazil
E-mail: [email protected]
Received 6 February 2023, revised 11 July 2023
Accepted for publication 15 September 2023
Published 22 September 2023
Abstract
In this paper the MISHEMT device (metal/Si3N4/AlGaN/AlN/GaN - metal–insulator–
semiconductor high electron mobility transistor) is studied focusing mainly on the impact of the
multiple conductions on the intrinsic voltage gain (Av). It is shown that the total drain current is
composed of three different drain current components, whereof one is related to the MIS
channel and the other two are related to high electron mobility transistor (HEMT) channels. The
device output characteristics present double drain voltage saturation that gives rise to a double
plateau in the saturation region of the output characteristics. This behavior relies also on the gate
voltage, so the output characteristics and analog parameters extraction are bias dependent. The
intrinsic voltage gain increases thanks to the early voltage increment in the second plateau
where HEMT conduction is dominant. Electron concentration profiles were simulated in order
to investigate the device saturation regime.
Keywords: MISHEMT, GaN, 2DEG, intrinsic voltage gain
(Some figures may appear in colour only in the online journal)
1. Introduction
Since 1983 the high electron mobility transistor (HEMT)
has been widely used in power electronics and high fre-
quency operations [1]. It presents a simple circuit config-
uration for power switch applications, simple design for
RF and microwave circuits [2, 3], and operates at harsh
environments [3, 4]. The HEMT is based on a heterostruc-
ture of AlGaN/GaN that gives rise to two types of internal
polarizations (spontaneous and piezoelectric polarizations),
which forms a two-dimensional electron gas (2DEG) [5],
providing high electron mobility and high electron density at
∗
Author to whom any correspondence should be addressed.
the interface of AlGaN/GaN, even though it usually generates
a normally-on transistor. In addition, normally-off GaN-based
transistors provide promising possibilities for digital circuits
applications operating at high temperatures [6].
However, the presence of severe self-heating effects in
HEMT devices degrades the performance [7, 8]. In addition,
the problem of high gate current leakage at scaled dimen-
sions and a consequent drain current collapse, reduces its per-
formance. As a solution to these issues, the metal–insulator–
semiconductor high electron mobility transistor (MISHEMT)
is presented as an alternative [9–11]. With the gate insu-
lator, the gate leakage reduces drastically and mitigates the
current collapse [10, 12, 13]. The MISHEMT is a promising
alternative for applications at high frequency, including 5G
applications [14–16], power electronics [11, 17], showing high
1361-6641/23/115004+6$33.00 Printed in the UK
1
© 2023 IOP Publishing Ltd
Semicond. Sci. Technol. 38 (2023) 115004
B G Canales et al
power gain at 10 GHz [18], with the possibility of achieving
higher RF performance of the device by adjusting the Al con-
centration in the AlGaN barrier [19].
There are many studies regarding the use of new materi-
als and different process steps in order to have a more stabil-
ized threshold voltage V t for a normally-off MISHEMT [3, 6,
9–11, 17, 20, 21], the implementation of different geometries
[22] and multiple channels [23]. A model for the 2DEG chan-
nel density has been created [24], and the study of low-
frequency noise has also been performed [13, 25, 26]. Lastly,
in most recent studies a 1 nm thin AlN spacer layer has been
placed between the AlGaN/GaN (AlGaN/AlN/GaN) in order
to increase the sheet density, mobility and decrease the sheet
resistance [27].
The focus of this work is to understand how the multiple
conduction channels of a MISHEMT, reported in [28–30],
impact on some analog parameters of these devices, mainly
the intrinsic voltage gain. This analysis is performed at room
temperature.
2. Device characteristics
The experimental data used in this work was obtained for a
MISHEMT fabricated in imec—Belgium. The device struc-
ture consists of a TiN/Si3N4 gate stack over a heterostructure
of AlGaN/AlN/GaN grown on a silicon platform. The device
has a width of 10 µm, a gate length (Lg) of 400 nm, an insu-
lator thickness (tSi3N4) of 2 nm, a barrier thickness (tAlGaN) of
15 nm, AlN layer thickness (tAlN) of 1 nm and a buffer thick-
ness (tGaN) of 200 µm. More fabrication details can be found
in [10].
The simulation data is based on a MISHEMT composed by
metal gate/Si3N4/AlGaN/AlN/GaN materials (figure 1) having
Lg = 400 nm, tSi3N4 = 2 nm, tAlGaN under the gate = 10 nm,
tAlN = 1 nm, tGaN = 300 µm and the gate to source and gate to
drain distances (LGS & LGD) of 1000 nm. The simulation was
performed using Synopsys Sentaurus Technology Computer
Aided Design [31], using material and region-wise models in
accordance with [32].
3. Results and analysis
As reported in [29], the transconductance curve of the metal
gate/Si3N4/AlGaN/AlN/GaN MISHEMT presents multiple
slopes due to the presence of multiple conduction channels
caused by the HEMT and MIS conductions. Since the stud-
ied MISHEMT has a negative threshold voltage, with a gate
voltage (V GS) of 0 V there are three high populations of elec-
trons, two of them are located at the 3rd and 2nd interfaces
due to III–V materials heterostructure, namely two 2DEG, and
one of them is located at the 1st interface, which mechanism is
similar to a depletion mode nMOSFET. These interfaces were
numbered according to their distance to the gate electrode,
the 1st interface (Si3N4/AlGaN) being the closest, and the 3rd
interface (AlN/GaN) the farthest. The farther the channel is
from the gate electrode, the more negative is the gate voltage
2
Figure 1. MISHEMT cross-section.
required to deplete the carriers and cut off the channel. We
will address these V GS values as different threshold voltages
for the different channels. It is possible to conclude that for
a distinct gate voltage, each channel will present a different
conduction condition regarding its electron concentration: for
V GS < V t3 all channels are cut off; for V t3 < V GS < V t2 the
2DEG channel at the 3rd interface is enabled and the channels
at 2nd and 1st interfaces are cut off; for V t2 < V GS < V t1 the
2DEG channels at the 3rd and 2nd interfaces are enabled and
the channel at the 1st interface is cut off; and for V GS > V t1 all
the channels are activated, including the MIS channel related
to the 1st interface. Because the drain current begins to rise at
V GS = V t3, then V t3 is considered the device effective threshold
voltage (V t).
Experimentally, the multiple transconductance slopes are
more easily seen when the device is operating at high temper-
atures, since there are two different transport mechanisms and
that each one responds differently to a temperature increase.
The MIS current, coming from the 1st interface, depends on
the Fermi level, which is lowered by the effect of high temper-
atures. The HEMT current, coming from the 2nd and 3rd inter-
faces, depends on the depletion depth, which is also affected by
temperature. In addition, the bandgap also has a dependency
on the temperature, playing an even major role on the HEMT
conduction [29]. Figure 2 shows the MISHEMT’s transfer
× V GS) and transconductance (gm) curve at 350 K
curve (IDS
and low V DS.
In figure 2 one can notice two peaks in the gm curve, the
first one is associated with the HEMT threshold voltage and
the second one, at higher V GS, is related to the MOS threshold
voltage. The difference between the two 2DEGs threshold
voltages (V t3 and V t2) is indistinguishable given that the chan-
nels related to them are physically separated by only 1 nm of
AlN layer.
Figure 3 shows the output characteristic (IDS
× V DS) at
C) for V GS ranging from −4.5 V to 0 V with
300 K (27
a V GS step of 100 mV. Four different overdrive voltages
(V GT = V GS
− V t3) are highlighted in blue in figure 3.
From figure 3 it is possible to notice the usual IDS beha-
vior for more negative gate bias, when only the 2DEG chan-
nels are activated. For a V GT of 1.5 V a MISHEMT kink effect
(MH kink effect) starts to occur in the drain current due to the
◦
Semicond. Sci. Technol. 38 (2023) 115004
B G Canales et al
Table 1. Electron concentration cross-sections for V DS of 2 V and
6 V and different gate overdrives (V GT) at 300 K.
V GT (V)
2.0
6.0
V DS (V)
0.9
1.9
2.4
3.9
(V GS = −2.8 V) and 2.4 V (V GS = −2.3) the depletion
region starts to move away from the 2DEG channels, enabling
the internal polarization to take place. For V GT = 3.9 V
(V GS = −0.8 V) the MIS channel is enabled by electric
field effect and the 2DEG channels are enabled by internal
polarization.
is
When the 1st
interface MIS channel
enabled
(V GT = 3.9 V), for V DS = 2 V the MIS channel is pinched-off
as can be seen at the 1st interface, but the HEMT channel is
fully formed and the drain voltage is not enough to impact
the HEMT conduction. However, when V DS = 6 V, the MIS
channel is pinched off over almost the entire channel and the
electron concentration cross section of HEMT conduction
shows that it is also influenced by the drain bias, entering the
saturation region. This second saturation results in a second
plateau in drain current causing a MH kink in the output char-
acteristics of the MISHEMT. Both plateaus only occurs when
both different conductions (MOS ad HEMT) are enabled, with
at a high enough V GT, as can be seen in figure 3.
Figure 2. Experimental drain current and transconductance as a
function of gate voltage at 350 K.
Figure 3. Experimental drain current as a function of drain voltage
for different gate voltages at room temperature. Blue curves are
related to gate overdrives V GT of 3.9, 2.4, 1.9, 0.9 V.
multiple conductions, and is responsible for the appearance of
a double plateau in the output characteristics. The IDS MH kink
shifts to higher drain voltage (V DS) as V GS increases.
In addition to the threshold voltage, the electron concentra-
tion at each interface gives rise to different saturation voltages
(V DSsat) related to the different channels. The saturation effect
of the HEMT conduction affects the output characteristics sim-
ilarly to the one that occurs in a MOSFET.
Table 1 presents the electron concentration cross-section
for V GT = [3.9, 2.4, 1.9, 0.9] V and V DS = [2.0, 6.0] V at
room temperature.
It can be seen from table 1 that varying V GT changes
the depletion depth, which works enabling or disabling the
multiple channels as it reaches out each one of them. For
V GT = 0.9 V (V GS = −3.8 V) the MIS channel is cut off
and the 2DEG channels are formed. For V GT between 1.9 V
3
Semicond. Sci. Technol. 38 (2023) 115004
B G Canales et al
Table 2. Experimental analog parameters for two values of overdrive voltage extracted in the first and second plateaus.
V GT (V) Output characteristics V DS (V)
gmsat (mS µm
−1)
gDsat (µS µm
−1) V EA (V) Av (dB)
3.9
3.4
First plateau
Second plateau
First plateau
Second plateau
3.90
7.25
3.00
6.75
0.185
0.164
0.195
0.187
4.60
3.33
4.62
3.75
177.6
252.8
154.7
200.7
32.1
33.8
32.5
34.0
When the depletion gets deeper (for V GT
⩽ 2.4 V) it starts
to compete with the internal polarization and to decrease the
electron concentration on both 2DEGs, while the MIS chan-
nel is cut off. In this condition, the saturation region of the
MISHEMT output characteristics becomes dominated by the
HEMT channels saturations.
The saturation on the 3rd interface (2DEG) is vertically
aligned with the gate electrode at its end closest to the drain.
In the worst case, when the 3rd interface is less populated by
carriers (V GT = 0.9 V), the HEMT conduction is most strongly
affected by the drain bias, but even in this case, the saturation
effect extends to only a small fraction of Lg. This means that
a channel length modulation-like effect takes place and that
the 3rd interface’s 2DEG effective channel length is minim-
ally affected, presenting a low gD.
It can be concluded that MIS conduction is more affected
by V DS (saturation) than HEMT conduction. In the latter case,
it can also be said that the lower the electron population of
the 2DEG (smaller V GS) is, the higher the V DS influence on
the HEMT conduction, resulting in a lower value of V DSsat.
For low V GS bias, the HEMT conduction is responsible for a
single plateau in the drain current curve, while at high applied
gate voltage, the MIS component is responsible for the first
× V DS curve and the HEMT component, for
plateau on IDS
the second one. In principle it is possible to observe similar
effects in others devices like GaN MISHEMT or other III–V
MISHEMT if the activation voltages of the different current
channels take place at enough distance of gate voltage.
Since MISHEMT behavior depends on MIS and HEMT
conductions, in order to extract the output conductance (gD),
the early voltage (V EA) and the intrinsic voltage gain (Av), two
× V DS were chosen
different points of the experimental IDS
for each applied V GS.
Table 2 shows the extracted analog parameters from exper-
imental data at room temperature.
From experimental data it is possible to note that the para-
meters for the lower V DS are related to the MIS channel satur-
ation (1st plateau), and the ones for higher V DS to the 2DEG
channel saturation (2nd plateau). For example, a higher gD is
obtained in the 1st plateau due to the high dependence with the
channel length modulation, while a lower gD is obtained in the
2nd plateau.
Knowing the early voltage is dependent on the drain cur-
rent level and on the output conductance, if two curves hav-
ing almost the same gD value and different IDS level were
taken, the one with a higher IDS level will have a higher
V EA, while that for two curves with the same IDSsat level
and different gD, the one with higher gD results in a smaller
Figure 4. Schematic figure of early voltage extraction of a common
MOSFET output curve (A), and of an output curve with double
plateau (B).
V EA. These V EA dependency characteristics are illustrated by
figure 4.
Since the MIS conduction is more dependent on the drain
bias, the output conductance (gD) extracted in the first plat-
eau tends to be higher than that extracted in the second one.
In addition, increasing V GT will make the current level to
increase due to the activation of each channel, together with
the increase in the electron concentration in the entire AlGaN
layer, which allows for conduction through the entire volume
of the layer, resulting in an early voltage improvement. It is
worth mentioning that MISHEMT devices present very high
V EA values when compared with MOSFET technology.
As the analog parameters depend on the gate and drain
bias, choosing a slightly different biasing condition can lead
to interesting new results for gD, V EA and Av.
Figure 5 shows the intrinsic voltage gain (Av) for different
biasing conditions at room temperature related to both plat-
eaus. These biasing conditions are slightly different from the
results shown in table 2.
For V GS of −1.5 V and −1.0 V, a double plateau appears
× V DS curve. The first plateau, for lower V DS, is
in the IDS
associated with the MIS channel. As V GS turns to more pos-
itive values, there is an increase of IDS, however, gD shows
higher values due to MIS channel length modulation. The high
IDS increase is provided by the AlGaN electron concentration
increase.
4
Semicond. Sci. Technol. 38 (2023) 115004
B G Canales et al
Data availability statement
All data that support the findings of this study are included
within the article (and any supplementary files).
Acknowledgments
authors would like
to thank CNPq (Processes
The
140223/2021-5 and 149902/2022-0) and CAPES for the fin-
ancial support.
ORCID iDs
Bruno Godoy Canales https://orcid.org/0000-0003-1013-
8073
Welder Fernandes Perina https://orcid.org/0000-0001-
6205-351X
Joao Antonio Martino https://orcid.org/0000-0001-8121-
6513
Eddy Simoen https://orcid.org/0000-0002-5218-4046
Paula Ghedini Der Agopian https://orcid.org/0000-0002-
0886-7798
References
[1] Mimura T 2018 Special contribution invention of high electron
mobility transistor (HEMT) and contributions to
information and communications field Fujitsu Sci. Tech. J.
54 3–8
[2] Wang H, Wang J, Liu J, He J, Wang M, Yu M and Wu W 2017
Normally-off Al2O3/GaN MOSFET: role of border traps on
the device transport characteristics Solid State Electron.
141 13–17
[3] He J, Wang Q, Zhou G, Li W, Jiang Y, Qiao Z, Tang C, Li G
and Yu H 2022 Normally-off AlGaN/GaN MIS-HEMTS
with low RON and Vth hysteresis by functioning in-situ
SiNx in regrowth process IEEE Electron Device Lett.
43 529–32
[4] Ghosh J, Das S, Mukherjee S, Ganguly S and Laha A 2019 A
study of electrical characteristics of Gd2O3/GaN and
Gd2O3/AlGaN/GaN MOS heterostructures Microelectron.
Eng. 216 111097
[5] Hamza K H, Nirmal D, Fletcher A S A, Arivazhagan L,
Ajayan J and Natarajan R 2021 Highly scaled graded
channel GaN HEMT with peak drain current of 2.48 A/mm
Int. J. Electron. Commun. 136 153774
[6] He L, Yang F, Yao Y, Zheng Y, Zhang J, Li L, He Z, Ni Y,
Gu X and Liu Y 2020 A review of selective area grown
recess structure for insulated-gate E-mode GaN transistors
Jpn. J. Appl. Phys. 59 SA0809
[7] Nigam A, Bhat T N, Rajamani S, Dolmanan B S, Tripathy S
and Kumar M 2017 Effect of self-heating on electrical
characteristics of AlGaN/GaN HEMT on Si (111) substrate
AIP Adv. 7 085015
[8] Guggenheim R and Rodes L 2017 Roadmap review for
cooling high-power GaN HEMT devices IEEE Int. Conf. on
Microwaves, Antennas, Communications and Electronic
Systems (COMCAS) pp 1–6
[9] Nahhas A M 2019 Review of AlGaN/GaN HEMTs based
devices Am. J. Nanomater. 7 10–21
[10] Peralagu U et al 2019 CMOS-compatible GaN-based devices
on 200 mm-Si for RF applications: integration and
Figure 5. Intrinsic voltage gain and early voltage for different bias
conditions at 300 K, related to both plateaus.
On the other hand, while still offering IDS increase for more
× V DS curve,
positive V GS, the second plateau on the IDS
related mainly to the 3rd interface’s HEMT channel, presents
an almost constant value of gD due to being less affected by
saturation effects. These factors make the Av to present a new
increase towards more positive V GS values.
4. Conclusions
The MISHEMT is analyzed under different bias conditions.
Different saturation effects due to multiple conduction mech-
anisms in saturation region output characteristics have been
noticed. The observed saturation effects are: (a) the pinch-off
in the MIS channel (1st interface); (b) 2DEG narrowing at the
2nd interface; and (c) the pinch-off-like effect of the 2DEG 3rd
interface. The MOS conduction channel showed to be more
affected by higher V DS.
As each channel presents different saturation voltages (first
the MIS channel and then the 2DEG channels), the output
characteristics show a double plateau in the saturation region
causing the MISHEMT kink effect, since the HEMT conduc-
tion predominates the drain current for higher V DS.
When the gate voltage is enough to enable the MIS con-
duction, the entire barrier layer has high carrier concen-
tration and the 2DEG channels are already formed. The
high carrier concentration is capable of maintaining an IDS
increase with V GS increase, and the 2DEG channels satur-
ations offer a slight degradation of the output conductance.
Both factors contribute to obtain a higher intrinsic voltage
gain (∼40 dB–2nd plateau at V GS of −1 V) than the res-
ults of the 1st plateau, thanks to the increase of the early
voltage (490 V–2nd plateau at V GS of −1 V). For these reasons
the MISHEMT appears as an interesting option for analogue
applications.
5
Semicond. Sci. Technol. 38 (2023) 115004
B G Canales et al
performance Int. Electron Devices Meeting (IEDM) vol
19 pp 398–401
[11] Whiteside M, Arulkumaran S and Ng G I 2021 Demonstration
of vertically-ordered h-BN/AlGaN/GaN
metal-insulator-semiconductor high-electron-mobility
transistors on Si substrate Mater. Sci. Eng. B 270 115224
[12] Agopian P G D, Carmo G J, Martino J A, Simoen E,
Peralagu U, Parvais B, Waldron N and Collaert N 2021
Gate dielectric material influence on DC behavior of
MO(I)SHEMT devices operating up to 150
Electron. 185 108091
◦
C Solid State
[13] Simoen E, Veloso A, O’Sullivan B, Takakura K and Claeys C
2021 Frontiers in low-frequency noise research in advanced
semiconductor devices China Semiconductor Technology
Int. Conf. (CSTIC) vol 21 pp 1–4
[14] Wu W M et al 2022 ESD HBM discharge model in RF
GaN-on-Si (MIS)HEMTs IEEE Trans. Electron Devices
69 2180–7
[15] Wu W M et al 2021 ESD Failures of GaN-on-Si D-Mode
AlGaN/GaN MIS-HEMT and HEMT devices for 5G
telecommunications Electrical Overstress/Electrostatic
Discharge Symp. Proc. vol 43 pp 1–7
[16] Nakajima S 2018 GaN HEMTs for 5G base station
applications IEEE Int. Electron Devices Meeting (IEDM)
vol 18 pp 320–3
[17] Grill A, Stampfer B, Im K, Lee J, Ostermaier C, Ceric H,
Waltl M and Grasser T 2019 Electrostatic coupling
and identification of single-defects in GaN/AlGaN
Fin-MIS-HEMTs Solid State Electron. 156 41–47
[18] Zine-eddine T, Zahra H and Zitouni M 2019 Design and
analysis of 10 nm T-gate enhancement-mode MOS-HEMT
for high power microwave applications J. Sci.: Adv. Mater.
Devices. 4 180–7
[19] Tunga A, Li X and Rakheja S 2021 Modeling-based design
and benchmarking of Al-rich AlGaN 3D nanosheet
MOSFET and MOSHEMTs for RF applications Device
Research Conf. (DRC) vol 21 pp 1–2
[22] Fatahilah M F, Strempel K, Yu F, Vodapally S, Waag A and
Wasisto H S 2019 3D GaN nanoarchitecture for field-effect
transistors Micro Nano Eng. 3 59–81
[23] Pal P, Pratap Y, Gupta M and Kabra S 2019 Comparative
analysis of oxides to improve performance of
DC-MOS-HEMTs IEEE Conf. on Modeling of Systems
Circuits and Devices (MOS-AK India)
vol 19 pp 82–85
[24] Amarnath G, Panda D K and Lenka T R 2019 Modeling and
simulation of DC and microwave characteristics of
AlInN(AlGaN)/AlN/GaN MOSHEMTs with different gate
lengths Int. J. Numer. Modelling, Electron. Netw. Devices
Fields 32 1–11
[25] Takakura K, Putcha V, Simoen E, Alian A R, Peralagu U,
Waldron N, Parvais B and Collaert N 2020 Low-frequency
noise investigation of GaN/AlGaN metal-oxide-
semiconductor high-electron-mobility field-effect transistor
with different gate length and orientation IEEE Trans.
Electron Devices 67 3062–8
[26] Hasan M R, Motayed A, Fahad M S and Rao M 2017
Fabrication and comparative study of DC and low
frequency noise characterization of GaN/AlGaN based
MOS-HEMT and HEMT J. Vac. Sci. Technol. B
35 052202
[27] Shrestha N M, Li Y and Chang E Y 2014 Simulation study on
electrical characteristic of AlGaN/GaN high electron
mobility transistors with AlN spacer layer Jpn. J. Appl.
Phys. 53 04EF08
[28] Canales B G and Agopian P G D 2022 Influence of multiple
conduction channels on MISHEMT’s intrinsic voltage gain
Symp. on Microelectronics Technology (SBMICRO) vol 36
pp 1–4
[29] Canales B G, Carmo G J and Agopian P G D 2021 The
conduction mechanisms analysis of AlGaN/GaN
MOSHEMTs with different source/drain electrode
configurations Symp. on Microelectronics Technology and
Devices (SBMICRO) vol 35 pp 1–4
[20] He L, Li L, Zhang J, Ni Y, Zhang J, Liu Z, Wu Q and Liu Y
[30] Perina W F, Martino J A and Agopian P G D 2022
2021 Threshold voltage engineering in Al2O3/AlGaN/GaN
MISHEMTs with thin barrier layer: MIS-gate charge
control and high threshold voltage achievement. Int. Symp.
on Power Semiconductor Devices and ICs (ISPSD) vol
33 pp 339–42
Experimental analysis of MISHEMT multiple conductions
from 200K to 450K Symp. on Microelectronics Technology
(SBMICRO) vol 36 pp 1–4
[31] Synopsys, Inc. 2022 Sentaurus™ Device User Guide
T-2022-03
[21] Lee H Y, Chang T W, Chang E Y, Rorsman N and Lee C T
2021 Fabrication and characterization of GaN-based
fin-channel array metal-oxide-semiconductor high-electron
mobility transistors with recessed-gate and GaO gate
insulator layer IEEE J. Electron Devices Soc. 9 393–9
[32] Strauss S, Erlebach A, Cilento T, Marcon D, Stoffels S and
Bakeroot B 2014 TCAD methodology for simulation of
GaN-HEMT power devices Int. Symp. on Power
Semiconductor Devices & IC’s (ISPSD) Proc. vol 26 pp
257–60
6
| null |
10.1371_journal.pone.0232007.pdf
|
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
|
All relevant data are within the manuscript and its Supporting Information files.
|
RESEARCH ARTICLE
Common mental disorders prevalence in
adolescents: A systematic review and meta-
analyses
Sara Arau´ jo SilvaID
Santos Gonc¸ alves1, Eliane Said Dutra1, Kênia Mara Baiocchi Carvalho1,2
1*, Simoni Urbano Silva2, De´ bora Barbosa Ronca1, Vivian Siqueira
1 Graduate Program in Human Nutrition, University of Brasilia, Federal District, Brasilia, Brazil, 2 Graduate
Program in Collective Health, University of Brasilia, Federal District, Brasilia, Brazil
* [email protected]
Abstract
An increasing number of original studies suggest the relevance of assessing mental health;
however, there has been a lack of knowledge about the magnitude of Common Mental Dis-
orders (CMD) in adolescents worldwide. This study aimed to estimate the prevalence of
CMD in adolescents, from the General Health Questionnaire (GHQ-12). Only studies com-
posed by adolescents (10 to 19 years old) that evaluated the CMD prevalence according to
the GHQ-12 were considered. The studies were searched in Medline, Embase, Scopus,
Web of Science, Lilacs, Adolec, Google Scholar, PsycINFO and Proquest. In addition, the
reference lists of relevant reports were screened to identify potentially eligible articles. Stud-
ies were selected by independent reviewers, who also extracted data and assessed risk of
bias. Meta-analyses were performed to summarize the prevalence of CMD and estimate
heterogeneity across studies. A total of 43 studies were included. Among studies that
adopted the cut-off point of 3, the prevalence of CMD was 31.0% (CI 95% 28.0–34.0; I2 =
97.5%) and was more prevalent among girls. In studies that used the cut-off point of 4, the
prevalence of CMD was 25.0% (CI 95% 19.0–32.0; I2 = 99.8%). Global prevalence of CMD
in adolescents was 25.0% and 31.0%, using the GHQ cut-off point of 4 and 3, respectively.
These results point to the need to include mental health as an important component of
health in adolescence and to the need to include CMD screening as a first step in the pre-
vention and control of mental disorders.
Introduction
Common Mental Disorders (CMD) refer to depressive and anxiety disorders and are distinct
from the feeling of sadness, stress or fear that anyone can experience at some moment in life.
Despite some methodological differences in the epidemiological studies, it is estimated that
4.4% and 3.6% of the world adult population suffers from depressive and anxiety disorders,
respectively [1]. CMD can affect health and quality of life, and it is noted that CMD affect peo-
ple at an early age [2].
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Silva SA, Silva SU, Ronca DB, Gonc¸alves
VSS, Dutra ES, Carvalho KMB (2020) Common
mental disorders prevalence in adolescents: A
systematic review and meta-analyses. PLoS ONE
15(4): e0232007. https://doi.org/10.1371/journal.
pone.0232007
Editor: Joel Msafiri Francis, University of the
Witwatersrand, SOUTH AFRICA
Received: August 6, 2019
Accepted: April 6, 2020
Published: April 23, 2020
Copyright: © 2020 Silva et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
1 / 19
PLOS ONE Common mental disorders prevalence in adolescents
The Global Burden of Diseases, Injuries, and Risk Factors (GBD) study is a comprehensive
study that evaluates incidence, prevalence, and years lived with disability (YLDs), which in its
most recent study evaluated the period from 1990 to 2017 for 195 countries and territories,
and identified that the burden of mental disorders is present for males and females and across
all age groups. The findings of the GDB indicate that mental disorders have consistently
formed more than 14% of age-standardized YLDs for nearly three decades, and have greater
than 10% prevalence in all 21 GBD regions [3]. Mental disorders are not often correctly identi-
fied and have negative consequences on people’s health.
At the population level the use of self-report psychiatric screening instruments, such as the
General Health Questionnaire (GHQ), has been recommended to track CMD, also known as
psychological distress/problems or psychiatric morbidity or non-psychotic mental illnesses
[4]. The GHQ-12 is a short and self-report form to identify people with psychological distress
or CMD [5,6]. This validated instrument comprising a multidimensional evaluation based in
three factors: anxiety and depression, social dysfunctions and loss of confidence [7] and can be
applied in individuals of different ages [8].
Adolescence, defined as a transitional phase between ages 10 and 19 [9] is generally per-
ceived as a phase of life with no health problems. However, approximately 20% of adolescents
experience a mental health problem, most commonly depression or anxiety [10].
Although there are preliminary data on the severity of these conditions among adolescents
[11], there has been a lack of knowledge about the magnitude of CMD in adolescents worldwide.
There was a systematic review of the global prevalence of CMD, published in 2014, which incor-
porated studies from 1980 to 2013 that surveyed people aged 16 to 65 and using diagnostic criteria
other than GHQ. In addition, from this study it was not possible to identify the prevalence of
CMD in adolescents [12]. In this context, a systematic review of the literature was carried out to
estimate the prevalence of CMD in adolescents around the world, from item 12 of the GHQ.
Materials and methods
This systematic review followed the Preferred Reporting Items for Systematic Review and
Meta-analyses PRISMA checklist [13] and for meta-analyses followed Meta-analysis of Obser-
vational Studies in Epidemiology (MOOSE) [14] guidelines.
Protocol and registration
The systematic review protocol was registered in the International Prospective Register of Sys-
tematic Reviews (PROSPERO), registration number CRD42018094763.
Eligibility criteria
The present study included observational studies. Only studies that assessed the prevalence of
CMD according to GHQ-12 in adolescents (10 to 19 years old) were considered for retention.
In studies that evaluated adolescents and also individuals outside the age group of interest for
this review, an attempt was made to identify only those eligible through the information con-
tained in the article or by contacting authors.
Moreover, no restrictions of language, publication date or status were applied. Studies of
specific groups such as obese or diabetic individuals, adolescents in treatment of any health
condition, college students, people who had traumatic experiences, pregnant teenagers and
people with physical disabilities were not eligible. The ineligibility criterion considered those
conditions that predispose to a higher risk of CMD, such as life events that presumably
increase the chances of having feelings of stress, depression or anxiety. For example, among
college students depression rates could be substantially higher than those found in the general
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
2 / 19
PLOS ONE Common mental disorders prevalence in adolescents
population, probably because they experience moments of stress related to studies or future
choices involving the profession phase of life [15]. Systematic reviews, interventional studies
or ecological estimates were also not included.
Information sources
A systematic search of the following databases was conducted to identify relevant studies:
Medline, Embase, Scopus, Web of Science, Lilacs and Adolec. A partial grey literature search
was also performed in Google Scholar, PsycINFO and Proquest Dissertation and Theses. The
Google Scholar search was limited to the first 200 most relevant articles. The search was con-
ducted on December 1, 2018 and updated in April 1, 2019. Additional articles, were hand-
searched in selected articles to identify potentially eligible studies not retrieved by the database
search. The search strategy was reviewed by two researchers, one of them with extensive expe-
rience in systematic reviews, according to the criteria of the checklist of the Peer Review of
Electronic Search Strategies (PRESS checklist) [16].
The following strategy was adapted for the databases: (Adolescent OR Teenager OR Child
OR Young OR Teen OR Youth OR Juvenile OR Adolescence OR Younger) AND (“General
Health Questionnaire” OR GHQ OR GHQ-12) AND (“common mental disorders” OR CMD
OR Anxiety OR anxious OR depression OR dysthymia OR “generalized anxiety disorder” OR
“panic disorder” OR phobia OR “social anxiety disorder” OR “obsessive-compulsive disorder”
OR “mental disorder” OR “mental health” OR "Psychological stress" OR "Life Stress" OR "Psy-
chologic Stress" OR "Mental suffering" OR Anguish OR "Emotional stress") AND (Survey OR
“Cross-sectional studies” OR Prevalence OR frequency OR "Cross-sectional" OR Observa-
tional). More information on the search strategies is provided in S1 Appendix. The Covidence
Software (Cochrane Collaboration software1, Melbourne, Australia) was used to remove
duplicate references and for the screening procedure, applied independently.
Data collection process
The study selection process was carried out in two stages. First, the articles were selected based
on their titles and abstracts, followed by a full text assessment. These two stages were carried
by two independent authors (SAS and SUS) and the records that did not meet the inclusion
criteria were discarded. The disagreements were resolved by consensus and counted on the
participation of a third author (DBR).
Data were extracted in duplicate by authors and discrepancies were resolved by consensus.
The following data were collected: authors, year of publication, year of research, country, study
design, age (mean or range), sample size (sex), GHQ cut-off point and outcome of the studies
(prevalence of CMD). The corresponding authors of the studies were contacted (at least two
attempts of contact) in case of unavailable data.
The 12-item version of the GHQ has psychometric properties comparable to those of the
longer versions of the questionnaire and the items of this instrument describe positive and
negative aspects of mental health in the last two weeks and present a scale with four response
options. The difference in the scale for positive and negative items indicates that the higher the
score, the higher level of psychiatric disorders. The studies show great variation in the scoring
methods for the GHQ, with scales ranging from zero to 12 or zero to 36.
Risk of bias within individual studies
The critical appraisal tool, recommended by The Joanna Briggs Institute for cross-sectional
studies, was used to assess the risk of bias. The purpose of this appraisal is to assess the method-
ological quality of a study and to determine the possibility of bias in its design, conduct and
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
3 / 19
PLOS ONE Common mental disorders prevalence in adolescents
analysis. This instrument consists of nine questions answered as “yes”, “no”, “unclear”, or “not
applicable” [17].
For this study, when all items were answered “yes”, the risk of bias were considered low,
and if any item were classified as “no” or “unclear”, a high risk of bias were expected. No scores
were assigned; results were expressed by the frequency of each classification of the evaluation
parameters. These ratings were not used as a criterion for study eligibility.
Summary measures and data analysis
The primary outcome was the prevalence of CMD, with a confidence interval of 95% (CI
95%). We estimated the summary measures for the total population and subgroups defined by
sex, risk of bias and income level according to the World Bank classification [18]. The meta-
analyses were calculated using a random-effect model and weighed by the inverse of the vari-
ance. The heterogeneity was evaluated by the Chi-square test with significance of p<0.10, and
its magnitude was determined by the I-squared (I2) [19].
Meta-regressions were performed in order to identify possible causes of heterogeneity
using the Knapp and Hartung test [20] with the following variables: risk of bias, sample size,
proportion of female adolescent, year of study and income level. The small-study effect by
visual inspection of the funnel graph and Egger’s test [21] was also evaluated.
Analyzes were performed with the "Metaprop" command of the Stata software (version
14.0), adopting p<0.05.
Results
Study selection
A total of 6 351 articles were initially found in the nine electronic databases, including grey liter-
ature. After removing the duplicates, the titles and abstracts of 3 783 articles were screened, and
197 potentially relevant studies were selected for full-text reading. An additional record was
selected from the reference lists of the fully read articles. A total of 126 articles were excluded for
nominated reasons (see S1 Table). Forty-three studies (reported in 72 articles) [22–93] were
therefore selected for inclusion in this review. The screening process is detailed in Fig 1.
Study characteristics
Table 1 shows a summary of the study characteristics. A total of 43 studies (200 980 partici-
pants; 19 countries) were included. The CMD prevalence studies were conducted in Asia
[26,27,34,39,40,45,48–50,52–54,57,70,89,90], America [38,41,44,84], Africa [22], Europe
[24,28,32,35–37,43,46,47,56,63,65,68,71,76,88,92] and Oceania [66,83]. The majority of studies
(n = 33) had a cross-sectional design.
For the purpose of comparing the studies, we selected only those that presented the score
scale from zero to 12, totaling 32 studies classified by 3 or 4 diagnostic cut-off points. Thus for
the set of studies that adopted the cut-off point of 3 or more symptoms of the GHQ-12, the
sample size varied from 145 adolescents in India [45] to 74 589 in Brazil [41], these studies
included 96 842 adolescents between the ages of 12 and 19 years. In the set of studies with cut-
off point of 4 or more symptoms, it ranged from 90 adolescents in Malaysia [90] to 17 920 in
Japan [57] and the total sample was 79 892 adolescents aged 12 to 19 years.
Results of individual studies and synthesis of results
Only six (18.8%) studies were considered to be of low risk of bias. Considering that the GHQ
is a self-administered instrument composed of validated questions and translated in several
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
4 / 19
PLOS ONE Common mental disorders prevalence in adolescents
Fig 1. Flow chart of systematic review procedure for illustrating search results, selection and inclusion of studies. �Adapted from PRISMA.
https://doi.org/10.1371/journal.pone.0232007.g001
languages, the parameter that deals with the identification of the outcomes measured in a valid
way was met by all the studies.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
5 / 19
PLOS ONE Table 1. Summary of characteristics of included studies.
Common mental disorders prevalence in adolescents
Country
Study design
Age (mean or range)
Sample size (sex)
Author, year
Amoran, 20051
Arun, 2009
Year of
research
NI
NI
Augustine, 2014
2009–2010
Nigeria
India
India
Ballbè, 20152
Bansal, 2009
Cheung, 2011
Czaba£a, 20053
Dzhambov, 20174
Emami, 2007
Fernandes, 2013
Gale, 20045
Geckova´, 20036
2011–2012
Spain
NI
NI
2002
2016
2004
2006
1986
1998
NI
China
Poland
Bulgaria
Iran
India
United
Kingdom
Slovakia
Glendinning, 2007
2002–2003
Russia
Gray, 2008
1998 and 2003
Green, 2018
2017–2013
Hamilton, 2009
Hori, 2016
Kaneita, 2009
Lopes, 20167
Ma¨kela¨, 2015
Mann, 2011
McNamee, 2008
Miller, 2018
Munezawa, 2009
Nakazawa, 2011
Nishida, 20088
2005
2011
2004
2013–2014
2008
2007
2005
2018
NI
2008
2006
United
Kingdom
United
Kingdom
Canada
Japan
Japan
Brazil
Finland
Canada
Ireland
United
Kingdom
Japan
Japan
Japan
Nur, 2012
2009–2010
Turkey
Cross-
sectional
Cross-
sectional
Cross-
sectional
Cross-
sectional
Cross-
sectional
Cross-
sectional
Cross-
sectional
Cross-
sectional
Cross-
sectional
Cross-
sectional
15 to 19
197
12 to 19
2 402 (boys = 1 371; girls = 1 031)
15 to 19
15 to 19
145 (all boys)
740 (boys = 396; girls = 344)
NI (9th grade students)
125
14.70±2.02
719 (boys = 434; girls = 285)
13.8
1 123 (boys = 521; girls = 600)
15 to 19
557 (boys = 408; girls = 149)
17 to 18
4 310 (boys = 1 923; girls = 2 387)
16 to 18
1 488
Longitudinal
16 (range not available)
5 187 (boys = 2 222; girls = 2 965)
GHQα cut-off
point
3b
3b
3b
3b
14c
11c
3b
3b
7b
5b
3b
Cross-
sectional
Cross-
sectional
Cross-
sectional
15 (range not available)
2 616 (boys = 1 369; girls = 1 243)
2/3b,c
14 to 15
13 to 15
626
1 253
Longitudinal
16 (range not available)
1 204 (boys = 619; girls = 585)
Cross-
sectional
Cross-
sectional
Longitudinal
Cross-
sectional
Cross-
sectional
Cross-
sectional
Cross-
sectional
Longitudinal
Cross-
sectional
Cross-
sectional
Cross-
sectional
Cross-
sectional
12 to 19
4 078 (boys = 2 092; girls = 1 986)
12 to 19
13 to 15
12 to 17
15 to 19
744 (boys = 373; girls = 371)
516 (boys = 294; girls = 222)
74 589 (boys = 33 364; girls = 41
225)
225 (boys = 102; girls = 123)
12 to 19
3 311 (boys = 1 566; girls = 1 745)
16 (range not available)
868 (boys = 352; girls = 516)
13 to 17
12 to 14
407 (boys = 204; girls = 203)
916 (boys = 568; girls = 348)
12 to 15
4 864 (boys = 2,429; girls = 2,435)
12 to 15
4 894 (boys = 2 523; girls = 2 371)
15 to 19
244 (all girls)
4b
4b
3b
6b
4b
4b
3b
4b
3b
4b
4b
4b
4b
4b
4b
(Continued )
6 / 19
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
PLOS ONE Common mental disorders prevalence in adolescents
Country
Study design
Age (mean or range)
Sample size (sex)
Table 1. (Continued)
Author, year
Ojio, 2016
Oshima, 20109
Year of
research
2006
2009
Oshima, 201210
2008–2009
Japan
Padro´n, 201211
2008–2009
Spain
Pisarska, 2011
Rickwood, 1996
Rothon, 201212
2004
1994
2005
Roy, 2014
2009–2010
Sweeting, 200913
Sweeting, 200913
Sweeting, 200913
1987
1999
2006
Thomson, 201814
1991–2014
Trainor, 2010
Trinh, 201515
Van Droogenbroeck,
2018
Yusoff, 2010
2001
2009
2008
NI
Japan
Japan
Poland
Australia
United
Kingdom
India
United
Kingdom
United
Kingdom
United
Kingdom
United
Kingdom
Australia
Canada
Cross-
sectional
Cross-
sectional
Cross-
sectional
Cross-
sectional
Cross-
sectional
Longitudinal
Longitudinal
12 to 18
12 to 18
12 to 18
15 637 (boys = 7 953; girls = 7
684)
341 (boys = 173; girls = 168)
17 920 (boys = 8 886; girls = 9
034)
15 to 17
4 054 (boys = 1 951; girls = 2 103)
15 to 16
16 to 19
14 to 15
Cross-
sectional
14 to 15 (around 80% of
sample)
Longitudinal
15.8±3.5 months
Longitudinal
15.5±3.6 months
Longitudinal
15.5±3.8 months
722 (boys = 383; girls = 335)
4 163 (boys = 1 988; girls = 2 175)
13 539 (boys = 7 852; girls = 7
579)
400 (boys = 200; girls = 200)
505
2 196
3 194
Cross-
sectional
Longitudinal
Cross-
sectional
16 to 19
13 to 17
15,8
11 397 (boys = 5 376; girls = 6
021)
947 (boys = 390; girls = 557)
2 660 (boys = 1 236; girls = 1 397)
Belgium
Cross sectional
15 to 19
680 (boys = 341; girls = 339)
Malaysia
Cross-
sectional
16 (range not available)
90 (boys = 40; girls = 50)
GHQα cut-off
point
4b
5b
4b
3b
3b
4b
4b
15c
2/3; 3/4;4/5b
2/3; 3/4;4/5b
2/3; 3/4;4/5b
4b
4b
3b
4b
4b
NI: Not informed.
αGHQ: General Health Questionnaire, 12 items.
bThe score range was 0–12.
cThe score range was 0–36.
1Amoran, 2007
2(Basterra, 2017; Gotsens, 2015)
3Bobrowski, 2007
4Dzhambov, 2018
5(Steptoe, 1996; Collishaw, 2010; Morgan, 2012)
6Geckova´, 2004
7Telo, 2018
8Nishida, 2010
9Yamasaki, 2018
10(Kinoshita, 2011; Ando, 2013; Shiraishi, 2014; Kitawaga, 2017; Morokuma, 2017)
11Padro´n, 2014
12Hale, 2014
13(West, 2003; Young, 2004; Sweeting, 2008; Sweeting 2010)
14(Fagg, 2008; Lang, 2011; Maheswaran, 2015; Pitchfort, 2016 and 2018)
15(Hamilton, 2011; Arbour-Nicitopoulos, 2012; Isaranuwatchai, 2014).
https://doi.org/10.1371/journal.pone.0232007.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
7 / 19
PLOS ONE Common mental disorders prevalence in adolescents
Fig 2. Risk of bias in the included studies (The Joanna Briggs Institute Critical Appraisal checklist for prevalence studies).
https://doi.org/10.1371/journal.pone.0232007.g002
Two parameters were not met by most studies: (1) appropriate statistical analysis; and (2)
study subjects and the setting described in detail (Fig 2 and Table 2). It is important to empha-
size that the critical appraisal tool recommends that the numerator and the denominator be
clearly reported, and that the percentages should be given with confidence intervals, so in the
methods section there must be enough details to identify the analytical technique used and
how specific variables were measured in the study. In addition, the study sample should be
described in enough detail so that other researchers can determine if it is comparable to the
population of interest to them. It is worth mentioning that some studies have reported the year
of data collection and characteristics of the study population.
Results of individual studies
Among those that adopted the cut-off point of 3 or more symptoms, the prevalence of CMD
was 31.0% (CI95% 28.0–34.0; I2 = 97.5%). In studies that used the cut-off point of 4 or more
symptoms, the prevalence of CMD was 25.0% (CI 95% 19.0–32.0; I2 = 99.8%) (Fig 3). In the
subgroup analysis, the heterogeneity remained high and it was observed that CMD is higher in
female adolescents when considered the cut-off point 3 (Table 3).
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
8 / 19
PLOS ONE Table 2. Risk of bias for each individual study assessed by Joanna Briggs Institute critical appraisal checklist for prevalence studies.
Common mental disorders prevalence in adolescents
Studies
Amoran, 2005
Arun, 2009
Augustine, 2014
Ballbè, 2015
Czaba£a, 2005
Droogenbroeck, 2018
Dzhambov, 2017
Fagg, 2008
Gale, 2004
Glendinning, 2007
Green, 2018
Hori, 2016
Kaneita, 2009
Lopes, 2016
Ma¨kela¨, 2014
Mann, 2011
McNamee, 2008
Miller, 2018
Munezawa, 2009
Nakazawa, 2011
Nishida, 2008
Nur, 2012
Ojio, 2016
Oshima, 2012
Padro´n, 2012
Pisarska, 2011
Rothon, 2012
Thomson, 2018
Trainor, 2010
Trinh, 2015
Yusoff, 2010
Rickwood, 1996
1�
2�
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
U
Y
Y
Y
Y
Y
Y
Y
Y
N
Y
Y
Y
Y
Y
Y
N
Y
3�
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
N
Y
Y
Y
Y
Y
Y
U
Y
4�
Y
Y
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
N
Y
N
N
N
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
N
Y
Criteria
5�
6�
7�
8�
9�
U
Y
Y
Y
Y
Y
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
U
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
N
Y
N
N
N
N
N
Y
N
N
N
N
Y
N
N
Y
N
Y
N
Y
Y
N
N
Y
N
Y
Y
Y
N
N
N
Y
N
N
Y
Y
U
Y
Y
N
Y
Y
Y
Y
U
Y
Y
Y
Y
Y
N
U
Y
Y
Y
Y
Y
Y
Y
Y
Y
U
U
Y
Y
Y
�Y = Yes, N = No, U = Unclear, NA = Not applicable
1�
The sample was appropriate to address the target population
2�
3�
4�
5�
6�
7�
8�
9�
Criteria for inclusion in the sample cleary defined
Adequate sample size
Study subjects and the setting described in detail
Analysis conducted with sufficient coverage of the identified sample
Outcomes measured in a valid way
Objective and standard criteria for measurement
Appropriate statistical analysis
Strategies for dealing with the response rate properly
https://doi.org/10.1371/journal.pone.0232007.t002
In the meta-regression, the high heterogeneity could not be explained by the studied vari-
ables: sex, income level and year of publication (p>0.05; data not shown).
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
9 / 19
PLOS ONE Common mental disorders prevalence in adolescents
Fig 3. Common mental disorders prevalence in adolescents in studies with cut-off point 3 or more symptoms (A) and cut-off
point 4 or more symptoms (B).
https://doi.org/10.1371/journal.pone.0232007.g003
The funnel graph was able to show the asymmetry between the studies, with greater repre-
sentation of large studies (Fig 4). Graph A shows the studies that adopted cut-off point 3 and
graph B, those that used cut-off point 4. Both illustrate that there is an effect of small studies
and these findings were confirmed by the Egger’s Test (p<0.001).
Table 3. Prevalence of common mental disorders, by subgroups, in adolescents.
Subgroups
Number of studies
Number of participants
Prevalence (%)
Confidence interval
95%
I2(%)
Cut-off 3 or more symptoms
Sex
Male
Female
Risk of bias
High
Low
Income Level
High income
Low income
Cut-off 4 or more symptoms
Sex
Male
Female
Risk of bias
High
Low
Income Level
High income
Low income
�p < 0.001.
10
9
8
5
8
5
9
9
18
1
16
3
https://doi.org/10.1371/journal.pone.0232007.t003
42 192
50 863
11 506
85 336
19 247
79 745
26 006
26 881
79 648
244
78 932
960
23.0
38.0
32.0
30.0
29.0
35.0
14.0
27.0
26.0
18.0
26.0
22.0
21.0–26.0
34.0–42.0
29.0–35.0
17.0–45.0
24.0–34.0
28.0–41.0
7.0–22.0
15.0–40.0
19.0–33.0
14.0–24.0
19.0–33.0
18.0–26.0
92.9�
96.9�
97.3�
98.2�
98.0�
96.9�
99.6�
99.8�
99.8�
-
99.8�
-
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
10 / 19
PLOS ONE Common mental disorders prevalence in adolescents
Fig 4. Funnel graph on the prevalence of common mental disorders in adolescents in studies with cut-off point 3 or more
symptoms (A) and cut-off point 4 or more symptoms (B). Egger´s test: p<0.001.
https://doi.org/10.1371/journal.pone.0232007.g004
Discussion
This systematic review was able to reveal the magnitude of CMD in adolescents from all over
the world. When presented at this stage of life, CMD can have negative consequences through-
out the future years. The problem is common and worrying, so much has been widely studied
since the 1980s [12] however, they refer to studies with diverse populations and with different
ways of identification of CMD.
Mental health can be influenced by several factors. Socioeconomic characteristics [38,94–
97]; characteristics of lifestyle [43,56,64,83,98–100] [43]and also characteristics related to affec-
tive relationships [101–103], have been the focuses of studies already performed in
adolescents.
Our meta-analysis revealed that very large studies were conducted in Japan and United
Kingdom. It was reported that children and adolescents in Japan have greater depressive ten-
dencies and this condition may be growing each year in several countries [104]. In the United
Kingdom, the assessment and monitoring of psychological distress among adolescents is a
common practice and generally performed in longitudinal studies for more than two decades
[105].The evidence indicates that the relationship between culture or personal values and men-
tal disorders differs across cultures and age groups [106]. An approach that takes into account
the differences in social and cultural contexts is necessary to understand the occurrence and
phenomenology of CMD in epidemiological studies, since there is a relationship between
them but that needs to be better clarifies in future studies.
Although with some degree of methodological issue in most studies, since less than 20% of
the studies presented low risk of bias, the results of this study indicate that CMD affect girls
more, considering only the studies that adopted cut-off point 3. Permanent concern with phys-
ical appearance, body dissatisfaction, exposure to sexualization may be one of the reasons that
affect girls’ mental health [107].
Another factor that apparently influences the presence of CMD is income level. Even
though the results presented in this systematic review showed no difference between income
level of the countries and CMD, further studies with this focus are needed in order to deepen
the knowledge about the subject. Longitudinal studies such as the British Household Panel
Survey (BHPS) and Longitudinal Study of Young People in England (LYSPE) demonstrate the
impact of economic recession and poverty in populations by strong associations between
socioeconomic variables and health outcomes [76,108–111].
Although the GHQ is a validated instrument for detecting CMD, the scoring scale and cut-
off point are not consensual, which impairs comparison among studies. Meta-analyses in the
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
11 / 19
PLOS ONE Common mental disorders prevalence in adolescents
present study were based on cut-off points 3 and 4, since they were more frequent among the
studies.
In relation to age, studies are commonly defined to be representative of the population aged
15 years or more, however, it is also important to investigate the phenomenon of CMD among
the younger population (10 to 14 years), since global epidemiological data consistently report
that up to 20% of children and adolescents suffer from a disabling mental illness [112]. Particu-
lar attention should be paid to the most vulnerable adolescent population in order to create
strategies based on scientific evidence [113]. This systematic review revealed the severity of the
problem by the worldwide high prevalence of CMD among adolescents, using a standardized
criterion of measurement, the GHQ-12.
Study limitations
In this review some of the eligible studies showed association data and did not present the
prevalence and the respective confidence intervals, nor did they present the description of the
evaluated population. It is possible that this review did not include all relevant publications,
either because the articles did not present sufficient information or because the authors were
not located or, finally, because of unanswered communication attempts.
It is observed that the different cut-off points for the GHQ-12 adopted in the original stud-
ies were a complicating factor in the identification of cases of CMD and in the comparison
among studies. Even if measures were taken to combine studies that were as comparable as
possible, this review included studies conducted at different times and places and with varying
methodologies. These characteristics are revealed in the heterogeneity between the studies,
typically found in cross-sectional studies and, therefore, we performed a subgroup analysis
and a meta-regression, but without success.
Strengths of the study
In the elaboration of this systematic review, some steps were considered as the registration of
protocol in PROSPERO, the use of the PRESS checklist, blind selection of studies, the adoption
of updated analytical methods and a search strategy that enabled the capture of a large num-
bers of studies. An extensive search for studies was carried out in the literature sources, the
grey literature, and the reference lists of the eligible articles. When necessary, the authors of
potentially eligible studies were contacted to obtain extra data to carry out the meta-analyses.
Moreover, this systematic review followed the PRISMA tool guide and the Meta-analysis of
Observational Studies in Epidemiology (MOOSE) [14].
Conclusion
The global prevalence of CMD in adolescents was 25.0% and 31.0%, using the GHQ cut-off
point of 4 and 3, respectively. CMD was more prevalent among girls when observing studies
that adopted a 3 cut-off point. These results point to the need to include mental health as an
important component of health in adolescence and to the need to include CMD screening as a
first step in the prevention and control of mental disorders.
Supporting information
S1 Appendix. PRISMA checklist.
(DOC)
S2 Appendix. Search strategy and databases.
(DOC)
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
12 / 19
PLOS ONE Common mental disorders prevalence in adolescents
S1 Table. Details of excluded studies.
(DOC)
S1 Data.
(XLSX)
Author Contributions
Conceptualization: Sara Arau´jo Silva, Simoni Urbano Silva, Vivian Siqueira Santos Gonc¸al-
ves, Kênia Mara Baiocchi Carvalho.
Data curation: Sara Arau´jo Silva, Simoni Urbano Silva, De´bora Barbosa Ronca.
Formal analysis: Sara Arau´jo Silva, Vivian Siqueira Santos Gonc¸alves.
Methodology: Sara Arau´jo Silva, Simoni Urbano Silva, De´bora Barbosa Ronca.
Project administration: Eliane Said Dutra, Kênia Mara Baiocchi Carvalho.
Supervision: Kênia Mara Baiocchi Carvalho.
Writing – original draft: Sara Arau´jo Silva.
Writing – review & editing: Sara Arau´jo Silva, Simoni Urbano Silva, De´bora Barbosa Ronca,
Vivian Siqueira Santos Gonc¸alves, Eliane Said Dutra, Kênia Mara Baiocchi Carvalho.
References
1. World Health Organization. Depression and other common mental disorders: global health estimates.
WHO World Heal Organ [Internet]. 2017;1–24. Available from: http://apps.who.int/iris/bitstream/
10665/254610/1/WHO-MSD-MER-2017.2-eng.pdf
2. UNICEF. The United NAtions Childre´s Fund. Adolescence: A Time That Matters [Internet]. 2002.
7–44 p. Available from: www.unicef.org
3.
James SL, Abate D, Abate KH, Abay SM, Abbafati C, Abbasi N, et al. Global, regional, and national
incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries
and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. Lan-
cet. 2018; 392(10159):1789–858. https://doi.org/10.1016/S0140-6736(18)32279-7 PMID: 30496104
4. Goldberg D. A bio-social model for common mental disorders. Acta Psychiatr Scand Suppl. 1994;
385:66–70. https://doi.org/10.1111/j.1600-0447.1994.tb05916.x PMID: 7740974
5. Goldberg DP, Gater R, Sartorius N, Ustun TB, Piccinelli M, Gureje O, et al. The validity of two versions
of the GHQ in the WHO study of mental illness in general health care. Psychol Med. 1997; 27(1):191–
7. https://doi.org/10.1017/s0033291796004242 PMID: 9122299
6. Gnambs T, Staufenbiel T. The structure of the General Health Questionnaire (GHQ-12): two meta-
analytic factor analyses. Health Psychol Rev [Internet]. 2018; 12(2):179–94. Available from: https://
doi.org/10.1080/17437199.2018.1426484 PMID: 29325498
7. Graetz B. Multidimensional properties of the General Health Questionnaire. Soc Psychiatry Psychiatr
Epidemiol. 1991; 26(3):132–8. https://doi.org/10.1007/bf00782952 PMID: 1887291
8. Goldberg D; Rickels K; Downing R. and Hesbacher P. A comparison of two psychiatric screening
tests. Br J Psychiatry. 1976;
9. World Health Organization. Adolescence: The Critical Phase. World Health Organization; 1997.
10. World Health Organization. Adolescent mental health—Mapping actions of nongovernmental organi-
zations and other international development organizations. WHO World Heal Organ. 2012; 50.
11. World Health Organization. Mental health atlas. Bulletin of the World Health Organization. 2014. 76
p.
12. Steel Z, Marnane C, Iranpour C, Chey T, Jackson JW, Patel V, et al. The global prevalence of common
mental disorders: A systematic review and meta-analysis 1980–2013. Int J Epidemiol. 2014; 43
(2):476–93. https://doi.org/10.1093/ije/dyu038 PMID: 24648481
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
13 / 19
PLOS ONE Common mental disorders prevalence in adolescents
13. Moher D, Liberati A, Tetzlaff J, Altman DG, Altman D, Antes G, et al. Preferred reporting items for sys-
tematic reviews and meta-analyses: The PRISMA statement. PLoS Med. 2009; 7(9):889–96.
14. Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al. Meta-analysis of observa-
tional studies in epidemiology: A proposal for reporting. J Am Med Assoc. 2000; 283(15):2008–12.
15.
Ibrahim AK, Kelly SJ, Adams CE, Glazebrook C. A systematic review of studies of depression preva-
lence in university students. J Psychiatr Res [Internet]. 2013; 47(3):391–400. Available from: https://
doi.org/10.1016/j.jpsychires.2012.11.015 PMID: 23260171
16. McGowan J, Sampson M, Salzwedel DM, Cogo E, Foerster V, Lefebvre C. PRESS Peer Review of
Electronic Search Strategies: 2015 Guideline Statement. J Clin Epidemiol [Internet]. 2016; 75:40–6.
Available from: https://doi.org/10.1016/j.jclinepi.2016.01.021 PMID: 27005575
17.
The Joanna Briggs Institute. The Joanna Briggs Institute Critical Appraisal tools for use in JBI System-
atic Reviews—Checklist for Prevalence Studies. Crit Apprais Checkl Preval Stud. 2017; 7.
18.
The World Bank Group. World Development Indicators 2017. 2017;10.
19. Rodrigues C L, Ziegelmann P K. Meta´ nalise: Um guia pra´ tico. Rev HCPA, editor. Rev HCPA [Internet].
2010;(1):54. Available from: http://hdl.handle.net/10183/24862
20. Knapp G, Hartung J. Improved tests for a random effects meta-regression with a single covariate. Stat
Med. 2003; 22(17):2693–710. https://doi.org/10.1002/sim.1482 PMID: 12939780
21. Sterne JAC, Sutton AJ, Ioannidis JPA, Terrin N, Jones DR, Lau J, et al. Recommendations for examin-
ing and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. BMJ.
2011; 343(7818):1–8.
22. Amoran OE, Lawoyin TO, Oni OO. Risk factors associated with mental illness in Oyo State, Nigeria: A
community based study. Ann Gen Psychiatry [Internet]. 2005; 4:19. Available from: http://www.
pubmedcentral.nih.gov/articlerender.fcgi?artid=1351179&tool=pmcentrez&rendertype=abstract
https://doi.org/10.1186/1744-859X-4-19 PMID: 16372910
23. Amoran O, Lawoyin T, Lasebikan V. Prevalence of depression among adults in Oyo State, Nigeria: A
comparative study of rural and urban communities. Aust J Rural Health [Internet]. 2007 [cited 2018
Apr 23]; 15(3):211–5. Available from: http://onlinelibrary.wiley.com/doi/10.1111/j.1440-1584.2006.
00794.x/full PMID: 17542795
24. Dzhambov A, Tilov B, Markevych I, Dimitrova D. Residential road traffic noise and general mental
health in youth: The role of noise annoyance, neighborhood restorative quality, physical activity, and
social cohesion as potential mediators. Environ Int [Internet]. 2017; 109(July):1–9. Available from:
https://doi.org/10.1016/j.envint.2017.09.009 PMID: 28917129
25. Dzhambov AM, Markevych I, Hartig T, Tilov B, Arabadzhiev Z, Stoyanov D, et al. Multiple pathways
link urban green- and bluespace to mental health in young adults. Environ Res [Internet]. 2018; 166
(May):223–33. Available from: https://doi.org/10.1016/j.envres.2018.06.004 PMID: 29890427
26. Emami H, Ghazinour M, Rezaeishiraz H, Richter J. Mental Health of Adolescents in Tehran, Iran. J
Adolesc Heal [Internet]. 2007 Dec [cited 2018 Apr 26]; 41(6):571–6. Available from: http://linkinghub.
elsevier.com/retrieve/pii/S1054139X07002704
27.
Fernandes AC, Hayes RD, Patel V. Abuse and other correlates of common mental disorders in youth:
a cross-sectional study in Goa, India. Soc Psychiatry Psychiatr Epidemiol [Internet]. 2013 Apr [cited
2018 Apr 26]; 48(4):515–23. Available from: http://link.springer.com/10.1007/s00127-012-0614-6
PMID: 23111769
28. Gale CR, Martyn CN. Birth weight and later risk of depression in a national birth cohort. Br J Psychia-
try. 2004; 184(January):28–33. https://doi.org/10.1192/bjp.184.1.28 PMID: 14702224
29. Steptoe A, Butler N. Sports participation and emotional wellbeing in adolescents. Lancet. 1996; 347
(9018):1789–92. https://doi.org/10.1016/s0140-6736(96)91616-5 PMID: 8667922
30. Collishaw S, Maughan B, Natarajan L, Pickles A. Trends in adolescent emotional problems in England:
A comparison of two national cohorts twenty years apart. J Child Psychol Psychiatry Allied Discip
[Internet]. 2010 [cited 2018 Apr 23]; 51(8):885–94. Available from: http://onlinelibrary.wiley.com/doi/
10.1111/j.1469-7610.2010.02252.x/full
31. Morgan Z, Brugha T, Fryers T, Stewart-Brown S. The effects of parent-child relationships on later life
mental health status in two national birth cohorts. Soc Psychiatry Psychiatr Epidemiol [Internet]. 2012
Nov 12 [cited 2018 Apr 23]; 47(11):1707–15. Available from: https://link.springer.com/article/10.1007/
s00127-012-0481-1 PMID: 22327406
32. Geckova´ A; Van Dijk JP, Stewart R; Groothoff J W; Post D. Influence of social support on health
among gender and socio-economic groups of adolescents. Eur J Public Health [Internet]. 2003 [cited
2018 Apr 23]; 13(1):44–50. Available from: https://academic.oup.com/eurpub/article-abstract/13/1/44/
488196 https://doi.org/10.1093/eurpub/13.1.44 PMID: 12678313
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
14 / 19
PLOS ONE Common mental disorders prevalence in adolescents
33. Geckova AM, van Dijk JP, Zezula I, Tuinstra J, Groothoff JW, Post D. Socio-economic differences in
health among Slovak adolescents. Soz Praventivmed. 2004; 49(1):26–35. https://doi.org/10.1007/
s00038-003-2050-6 PMID: 15040126
34. Arun P, Chavan B. Stress and suicidal ideas in adolescent students in Chandigarh. Indian J Med Sci
[Internet]. 2009; 63(7):281. Available from: http://www.indianjmedsci.org/text.asp?2009/63/7/281/
55112 https://doi.org/10.4103/0019-5359.55112 PMID: 19700908
35. Glendinning A, West P. Young people’s mental health in context: Comparing life in the city and small
communities in Siberia. Soc Sci Med. 2007; 65(6):1180–91. https://doi.org/10.1016/j.socscimed.2007.
05.012 PMID: 17576031
36. Gray L, Leyland AH. Overweight status and psychological well-being in adolescent boys and girls: A
multilevel analysis. Eur J Public Health. 2008; 18(6):616–21. https://doi.org/10.1093/eurpub/ckn044
PMID: 18663009
37. Green M J; Stritzel H; Smith C; Popham F; Crosnoe R. Timing of poverty in childhood and adolescent
health: Evidence from the US and UK. Soc Sci Med. 2018; 197:136–43. https://doi.org/10.1016/j.
socscimed.2017.12.004 PMID: 29232621
38. Hamilton HA, Noh S, Adlaf EM. Perceived financial status, health, and maladjustment in adolescence.
Soc Sci Med [Internet]. 2009 Apr [cited 2018 Apr 26]; 68(8):1527–34. Available from: http://linkinghub.
elsevier.com/retrieve/pii/S0277953609000471 https://doi.org/10.1016/j.socscimed.2009.01.037
PMID: 19246144
39. Hori D, Tsujiguchi H, Kambayashi Y, Hamagishi T, Kitaoka M, Mitoma J, et al. The associations
between lifestyles and mental health using the General Health Questionnaire 12-items are different
dependently on age and sex: a population-based cross-sectional study in Kanazawa, Japan. Environ
Health Prev Med. 2016; 21(6):410–21. https://doi.org/10.1007/s12199-016-0541-3 PMID: 27270412
40. Kaneita Y, Yokoyama E, Harano S, Tamaki T, Suzuki H, Munezawa T, et al. Associations between
sleep disturbance and mental health status: A longitudinal study of Japanese junior high school stu-
dents. Sleep Med [Internet]. 2009 Aug [cited 2018 Apr 23]; 10(7):780–6. Available from: http://
linkinghub.elsevier.com/retrieve/pii/S1389945708002700 https://doi.org/10.1016/j.sleep.2008.06.014
PMID: 19186103
41.
42.
Lopes CS; Abreu GA; Santos DF; Menezes PR; Carvalho KMB; Cunha CF; et al. ERICA: Prevalence
of common mental disorders in Brazilian adolescents. Rev Saude Publica. 2016; 50(supl 1):1s–9s.
Telo GH, Cureau F V, Lopes CS, Schaan BD. Common mental disorders in adolescents with and with-
out type 1 diabetes: Reported occurrence from a countrywide survey. Diabetes Res Clin Pract [Inter-
net]. 2018 Jan 1 [cited 2019 May 4]; 135:192–8. Available from: http://www.ncbi.nlm.nih.gov/pubmed/
29155124 https://doi.org/10.1016/j.diabres.2017.10.027 PMID: 29155124
43. Makela P; Raitasalo K; Wahlbeck K. Mental health and alcohol use: A cross-sectional study of the
Finnish general population. Eur J Public Health [Internet]. 2015 Apr 1 [cited 2018 Apr 26]; 25(2):225–
31. Available from: https://academic.oup.com/eurpub/article-lookup/doi/10.1093/eurpub/cku133
44. Mann RE; Paglia-Boak A; Adlaf EM; Beitchman J; Wolfe D; Wekerle C; et al. Estimating the Preva-
lence of Anxiety and Mood Disorders in an Adolescent General Population: An Evaluation of the
GHQ12. Int J Ment Health Addict [Internet]. 2011 Aug 27 [cited 2018 Apr 23]; 9(4):410–20. Available
from: https://link.springer.com/article/10.1007/s11469-011-9334-5
45. Augustine LF; Nair KM; Rao SF; Rao MV; Ravinder P; Balakrishna N; et al. Adolescent Life-Event
Stress in Boys Is Associated with Elevated IL-6 and Hepcidin but Not Hypoferremia. J Am Coll Nutr.
2014; 33(5):354–62. https://doi.org/10.1080/07315724.2013.875417 PMID: 25302670
46. McNamee H, Lloyd K, Schubotz D. Same sex attraction, homophobic bullying and mental health of
young people in Northern Ireland. J Youth Stud. 2008; 11(1):33–46.
47. Miller K, Wakefield J, Sani F. Identification with the school predicts better mental health amongst high
school students over time. Educ Child Psychol. 2018;(Special issue):21–9.
48. Munezawa T, Kaneita Y, Yokoyama E, Suzuki H, Ohida T. Epidemiological study of nightmare and
sleep paralysis among Japanese adolescents. Sleep Biol Rhythms. 2009; 7(3):201–10.
49. Nakazawa N, Imamura A, Nishida A, Iwanaga R, Kinoshita H, Okazaki Y, et al. Psychotic-like experi-
ences and poor mental health status among Japanese early teens. Acta Med Nagasaki [Internet].
2011; 56(2):35–41. Available from: http://www.scopus.com/inward/record.url?eid=2-s2.0-
84858141986&partnerID=tZOtx3y1
50. Nishida A, Tanii H, Nishimura Y, Kajiki N, Inoue K, Okada M, et al. Associations between psychotic-
like experiences and mental health status and other psychopathologies among Japanese early teens.
Schizophr Res [Internet]. 2008 Feb [cited 2018 Apr 26]; 99(1–3):125–33. Available from: http://
linkinghub.elsevier.com/retrieve/pii/S0920996407005610 https://doi.org/10.1016/j.schres.2007.11.
038 PMID: 18248792
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
15 / 19
PLOS ONE Common mental disorders prevalence in adolescents
51. Nishida A, Sasaki T, Nishimura Y, Tanii H, Hara N, Inoue K, et al. Psychotic-like experiences are asso-
ciated with suicidal feelings and deliberate self-harm behaviors in adolescents aged 12–15 years. Acta
Psychiatr Scand [Internet]. 2010 Apr [cited 2018 Apr 23]; 121(4):301–7. Available from: http://
onlinelibrary.wiley.com/doi/10.1111/j.1600-0447.2009.01439.x/full PMID: 19614622
52. Nur N. The effect of intimate partner violence on mental health status among women of reproductive
ages: A population-based study in a Middle Anatolian city. [Internet]. Journal of Interpersonal Violence
Sage Publications; Nov 30, 2012 p. 3236–51. Available from: http://journals.sagepub.com/doi/10.
1177/0886260512441255 PMID: 22550141
53. Ojio Y, Nishida A, Shimodera S, Togo F, Sasaki T. Sleep duration associated with the lowest risk of
depression/anxiety in adolescents. Sleep J Sleep Sleep Disord Res. 2016; 39(8):1555–62.
54. Oshima N, Nishida A, Fukushima M, Shimodera S, Kasai K, Okazaki Y, et al. Psychotic-like experi-
ences (PLEs) and mental health status in twin and singleton Japanese high school students. Early
Interv Psychiatry. 2010; 4(3):206–13. https://doi.org/10.1111/j.1751-7893.2010.00186.x PMID:
20712725
55. Yamasaki S, Usami S, Sasaki R, Koike S, Ando S, Kitagawa Y, et al. The association between
changes in depression/anxiety and trajectories of psychotic-like experiences over a year in adoles-
cence. Schizophr Res [Internet]. 2018; 195:149–53. Available from: https://doi.org/10.1016/j.schres.
2017.10.019 PMID: 29055569
56. Ballbè M, Martı´nez-Sa´ nchez JM, Gual A, Martı´nez C, Fu M, Sureda X, et al. Association of second-
hand smoke exposure at home with psychological distress in the Spanish adult population. Addict
Behav [Internet]. 2015 Nov [cited 2018 Apr 26]; 50:84–8. Available from: http://linkinghub.elsevier.
com/retrieve/pii/S0306460315002087 https://doi.org/10.1016/j.addbeh.2015.06.020 PMID: 26111658
57. Oshima N, Nishida A, Shimodera S, Tochigi M, Ando S, Yamasaki S, et al. The Suicidal Feelings,
Self-Injury, and Mobile Phone Use After Lights Out in Adolescents. J Pediatr Psychol [Internet]. 2012
Oct 1 [cited 2018 Apr 26]; 37(9):1023–30. Available from: https://academic.oup.com/jpepsy/article-
lookup/doi/10.1093/jpepsy/jss072 PMID: 22728900
58. Kinoshita Y, Shimodera S, Nishida A, Kinoshita K, Watanabe N, Oshima N, et al. Psychotic-like experi-
ences are associated with violent behavior in adolescents. Schizophr Res [Internet]. 2011 Mar [cited
2018 Apr 26]; 126(1–3):245–51. Available from: http://linkinghub.elsevier.com/retrieve/pii/
S0920996410014994 https://doi.org/10.1016/j.schres.2010.08.028 PMID: 20837390
59. Ando S, Yamasaki S, Shimodera S, Sasaki T, Oshima N, Furukawa TATA, et al. A greater number of
somatic pain sites is associated with poor mental health in adolescents: a cross-sectional study. BMC
Psychiatry [Internet]. 2013 Dec 17 [cited 2018 Apr 23]; 13(1):30. Available from: http://bmcpsychiatry.
biomedcentral.com/articles/10.1186/1471-244X-13-30
60. Shiraishi N, Nishida A, Shimodera S, Sasaki T, Oshima N, Watanabe N, et al. Relationship between
violent behavior and repeated weight-loss dieting among female adolescents in Japan. PLoS One.
2014; 9(9).
61. Kitagawa Y, Ando S, Yamasaki S, Foo JC, Okazaki Y, Shimodera S, et al. Appetite loss as a potential
predictor of suicidal ideation and self-harm in adolescents: A school-based study. Appetite [Internet].
2017; 111:7–11. Available from: https://doi.org/10.1016/j.appet.2016.12.026 PMID: 28011004
62. Morokuma Y, Endo K, Nishida A, Yamasaki S, Ando S, Morimoto Y, et al. Sex differences in auditory
verbal hallucinations in early, middle and late adolescence: Results from a survey of 17 451 Japanese
students aged 12–18 years. BMJ Open. 2017; 7(5):1–6.
63. Padro´ n A. Confirmatory factor analysis of the General Health Questionnaire (GHQ-12) in spanish ado-
lescents. Qual Life Res. 2012; 18(2):197–206.
64. Padro´ n A, Gala´ n I, Rodrı´guez-Artalejo F. Second-hand smoke exposure and psychological distress in
adolescents. A population-based study. Tob Control [Internet]. 2014 Jul [cited 2018 Apr 26]; 23
(4):302–7. Available from: http://tobaccocontrol.bmj.com/lookup/doi/10.1136/tobaccocontrol-2012-
050548 PMID: 23002180
65. Pisarska A, Ostaszewski K. Medicine use among Warsaw ninth-grade students. Drugs Educ Prev Pol-
icy [Internet]. 2011 Oct 12 [cited 2018 Apr 26]; 18(5):361–70. Available from: http://www.tandfonline.
com/doi/full/10.3109/09687637.2010.542786
66. Rickwood D, D’Espaignet E. Psychological distress among older adolescents and young adults in Aus-
tralia. Aust N Z J Public Health. 1996; 20(1):83–6. https://doi.org/10.1111/j.1467-842x.1996.tb01342.x
PMID: 8799073
67. Basterra V. Prevalence trends of high risk of mental disorders in the Spanish adult population: 2006–
2012. Gac Sanit [Internet]. 2017; 31(4):324–6. Available from: http://dx.doi.org/10.1016/j.gaceta.2017.
01.004 PMID: 28342634
68. Rothon C, Goodwin L, Stansfeld S. Family social support, community “social capital” and adolescents’
mental health and educational outcomes: a longitudinal study in England. Soc Psychiatry Psychiatr
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
16 / 19
PLOS ONE Common mental disorders prevalence in adolescents
Epidemiol [Internet]. 2012 May 10 [cited 2018 Apr 26]; 47(5):697–709. Available from: http://link.
springer.com/10.1007/s00127-011-0391-7 PMID: 21557090
69. Hale DR, Patalay P, Fitzgerald-Yau N, Hargreaves DS, Bond L, Go¨ rzig A, et al. School-Level Variation
in Health Outcomes in Adolescence: Analysis of Three Longitudinal Studies in England. Prev Sci.
2014;
70. Rupali R; Mukherjee S; Chaturvedi M; Agarwal K; Kannan A. Prevalence and predictors of psychologi-
cal distress among school students in Delhi. J Indian Assoc Child Adolesc Ment Heal. 2014; 10
(November 2009):150–66.
71. Sweeting H, Young R, West P. GHQ increases among Scottish 15 year olds 1987–2006. Soc Psychia-
try Psychiatr Epidemiol [Internet]. 2009 Jul 26 [cited 2018 Apr 23]; 44(7):579–86. Available from:
https://link.springer.com/article/10.1007/s00127-008-0462-6 PMID: 19037574
72. West P, Sweeting H. Fifteen, female and stressed: Changing patterns of psychological distress over
time. J Child Psychol Psychiatry Allied Discip. 2003; 44(3):399–411.
73. Young R, Sweeting H. Adolescent Bullying, Relationships, Psychological Well-Being, and Gender-
Atypical Behavior: A Gender Diagnosticity Approach. Sex Roles [Internet]. 2004; 50(7/8):525–37.
Available from: http://link.springer.com/10.1023/B:SERS.0000023072.53886.86
74. Sweeting H, West P, Young R. Obesity among Scottish 15 year olds 1987–2006: Prevalence and
associations with socio-economic status, well-being and worries about weight. BMC Public Health
[Internet]. 2008 Dec 9 [cited 2018 Apr 23]; 8(1):1–7. Available from: https://www.sciencedirect.com/
science/article/pii/S0140197101904221
75. Sweeting H, West P, Young R, Der G. Can we explain increases in young people’s psychological dis-
tress over time? Soc Sci Med [Internet]. 2010; 71(10):1819–30. Available from: https://doi.org/10.
1016/j.socscimed.2010.08.012 PMID: 20870334
76.
77.
Thomson RM, Katikireddi SV. Mental health and the jilted generation: Using age-period-cohort analy-
sis to assess differential trends in young people’s mental health following the Great Recession and
austerity in England. Soc Sci Med [Internet]. 2018; 214(June):133–43. Available from: https://doi.org/
10.1016/j.socscimed.2018.08.034
Fagg J, Curtis S, Stansfeld SA, Cattell V, Tupuola A-M, Arephin M. Area social fragmentation, social
support for individuals and psychosocial health in young adults: Evidence from a national survey in
England. Soc Sci Med [Internet]. 2008 Jan [cited 2018 Apr 26]; 66(2):242–54. Available from: http://
linkinghub.elsevier.com/retrieve/pii/S0277953607004170 https://doi.org/10.1016/j.socscimed.2007.
07.032 PMID: 17988774
78. Gotsens M, Malmusi D, Villarroel N, Vives-Cases C, Garcia-Subirats I, Hernando C, et al. Health
inequality between immigrants and natives in Spain: The loss of the healthy immigrant effect in times
of economic crisis. Eur J Public Health [Internet]. 2015 Dec [cited 2018 Apr 26]; 25(6):923–9. Available
from: https://academic.oup.com/eurpub/article-lookup/doi/10.1093/eurpub/ckv126 PMID: 26136466
79.
Lang IA, Llewellyn DJ, Hubbard RE, Langa KM, Melzer D. Income and the midlife peak in common
mental disorder prevalence. Psychol Med [Internet]. 2011 Jul 10 [cited 2018 Apr 26]; 41(7):1365–72.
Available from: http://www.journals.cambridge.org/abstract_S0033291710002060 https://doi.org/10.
1017/S0033291710002060 PMID: 21144109
80. Maheswaran H, Kupek E, Petrou S. Self-reported health and socio-economic inequalities in England,
1996–2009: Repeated national cross-sectional study. Soc Sci Med [Internet]. 2015 Jul [cited 2018 Apr
26]; 136–137:135–46. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0277953615003081
https://doi.org/10.1016/j.socscimed.2015.05.026 PMID: 26004207
81. Pitchforth JM, Viner RM, Hargreaves DS. Trends in mental health and wellbeing among children and
young people in the UK: a repeated cross-sectional study, 2000–14. Lancet [Internet]. 2016; 388:S93.
Available from: http://linkinghub.elsevier.com/retrieve/pii/S0140673616323297
82. Pitchforth J, Fahy K, Ford T, Wolpert M, Viner RM, Hargreaves DS. Mental health and well-being
trends among children and young people in the UK, 1995–2014: Analysis of repeated cross-sectional
national health surveys. Psychol Med. 2018;
83.
84.
Trainor S, Delfabbro P, Anderson S, Winefield A. Leisure time activities and adolescent psychological
well-being. J Adolesc [Internet]. 2010 [cited 2018 Apr 23]; 33(1):173–86. Available from: https://www.
sciencedirect.com/science/article/pii/S0140197109000396 https://doi.org/10.1016/j.adolescence.
2009.03.013 PMID: 19406463
Trinh L, Wong B, Faulkner GE. The independent and interactive associations of screen time and phys-
ical activity on mental health, school connectedness and academic achievement among a population-
based sample of youth. J Can Acad Child Adolesc Psychiatry. 2015; 24(1):17–24. PMID: 26336376
85. Hamilton HA, Paglia-Boak A, Wekerle C, Danielson AM, Mann RE. Psychological Distress, Service
Utilization, and Prescribed Medications among Youth with and without Histories of Involvement with
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
17 / 19
PLOS ONE Common mental disorders prevalence in adolescents
Child Protective Services. Int J Ment Health Addict [Internet]. 2011 Aug 30 [cited 2018 Apr 26]; 9
(4):398–409. Available from: http://link.springer.com/10.1007/s11469-011-9327-4
86. Arbour-Nicitopoulos KP, Faulkner GE, Irving HM. Multiple health-risk behaviour and psychological dis-
tress in adolescence. J Can Acad Child Adolesc Psychiatry [Internet]. 2012 [cited 2018 Apr 23]; 21
(3):171–8. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3413466/ PMID: 22876262
87.
Isaranuwatchai W, Rinner C, Hart H, Paglia-Boak A, Mann R, McKenzie K. Spatial Patterns of Drug
Use and Mental Health Outcomes Among High School Students in Ontario, Canada. Int J Ment Health
Addict [Internet]. 2014 Jun 16 [cited 2018 Apr 26]; 12(3):312–20. Available from: http://link.springer.
com/10.1007/s11469-013-9455-0
88. Van Droogenbroeck F, Spruyt B, Keppens G. Gender differences in mental health problems among
adolescents and the role of social support: Results from the Belgian health interview surveys 2008 and
2013. BMC Psychiatry. 2018; 18(1):1–9. https://doi.org/10.1186/s12888-017-1517-6
89. Cheung LM, WONG WS. The effects of insomnia and internet addiction on depression in Hong Kong
Chinese adolescents: An exploratory cross-sectional analysis. J Sleep Res [Internet]. 2011 Jun [cited
2018 Apr 23]; 20(2):311–7. Available from: http://doi.wiley.com/10.1111/j.1365-2869.2010.00883.x
90. Yusoff MS. Stress, stressors and coping strategies among secondary school students in a Malaysian
government secondary school: Initial findings. ASEAN J Psychiatry. 2010; 11(2):1–10.
91. Bansal V; Goyal S; Srivastava K. Study of prevalence of depression in adolescent students of a public
school. Ind Psychiatry J. 2009; 18(1):43. https://doi.org/10.4103/0972-6748.57859 PMID: 21234162
92. Czabała Czes£Aw J.; Brykczyńska Celina; Bobrowski Krzysztof; Ostaszewski K. Mental health prob-
lems in the population of high school students in Warsaw. Postêpy Psychiatr i Neurol. 2005; 14(1):1–
9.
93. Bobrowski KJ; Czabała JC; Brykczyńska C. Risk behaviors as a dimension of mental health assess-
ment in adolescents. Arch Psychiatry Psychother. 2007; 1–2:17–26.
94. Ahnquist J, Wamala SP, Lindstrom M. Social determinants of health—A question of social or eco-
nomic capital? Interaction effects of socioeconomic factors on health outcomes. Soc Sci Med [Inter-
net]. 2012 Mar [cited 2018 Apr 26]; 74(6):930–9. Available from: http://linkinghub.elsevier.com/
retrieve/pii/S0277953612000238 https://doi.org/10.1016/j.socscimed.2011.11.026 PMID: 22305807
95. Rocha KB; Pe´rez K; Rodrı´guez-Sanz M; Borrell C; Obiols J. Prevalence of mental health problems
and its association with socioeconomic, work and health variables: results of the National Health Sur-
vey of Spain. Psicothema [Internet]. 2010; 22(3):389–95. Available from: http://www.unioviedo.net/
reunido/index.php/PST/article/view/8867 PMID: 20667265
96. Weich S, Lewis G, Jenkins SP. Income inequality and the prevalence of common mental disorders in
Britain. Br J Psychiatry [Internet]. 2001 Mar 2 [cited 2018 Apr 26]; 178(03):222–7. Available from:
https://www.cambridge.org/core/product/identifier/S0007125000156454/type/journal_article
97. Okwaraji FE, Obiechina KI, Onyebueke GC, Udegbunam ON, Nnadum GS. Loneliness, life satisfac-
tion and psychological distress among out-of-school adolescents in a Nigerian urban city. Psychol
Heal Med [Internet]. 2018; 23(9):1106–12. Available from: https://doi.org/10.1080/13548506.2018.
1476726
98. De Vriendt T, Clays E, Huybrechts I, De Bourdeaudhuij I, Moreno LA, Patterson E, et al. European
adolescents’ level of perceived stress is inversely related to their diet quality: The Healthy Lifestyle in
Europe by Nutrition in Adolescence study. Br J Nutr. 2012; 108(2):371–80. https://doi.org/10.1017/
S0007114511005708 PMID: 22054044
99. Abu-omar K, Ru¨tten A, Lehtinen V. Mental health and physical activity in the European Union. Soz Pra-
ventivmed. 2004; 49:301–9. https://doi.org/10.1007/s00038-004-3109-8 PMID: 15497649
100. Kedzior KK, Laeber LT. A positive association between anxiety disorders and cannabis use or canna-
bis use disorders in the general population- a meta-analysis of 31 studies. BMC Psychiatry [Internet].
2014 Dec 10 [cited 2018 Apr 26]; 14(1):136. Available from: http://bmcpsychiatry.biomedcentral.com/
articles/10.1186/1471-244X-14-136
101. Young R, Sweeting H. Adolescent Bullying, Relationships, Psychological A Gender Diagnosticity
Approach. Sex Roles. 2004; https://doi.org/10.1023/B:SERS.0000037764.40569.2b
102. Rigby K, Slee PT, Martin G. Implications of inadequate parental bonding and peer victimization for
adolescent mental health. J Adolesc. 2007; 30(5):801–12. https://doi.org/10.1016/j.adolescence.
2006.09.008 PMID: 17113140
103. Glozah FN, Pevalin DJ. Social support, stress, health, and academic success in Ghanaian adoles-
cents: A path analysis. J Adolesc [Internet]. 2014; 37(4):451–60. Available from: https://doi.org/10.
1016/j.adolescence.2014.03.010 PMID: 24793393
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
18 / 19
PLOS ONE Common mental disorders prevalence in adolescents
104. Denda K, Kako Y, Kitagawa N, Koyama T. Assessment of depressive symptoms in Japanese school
children and adolescents using the birleson depression self-rating scale. Int J Psychiatry Med. 2006;
36(2):231–41. https://doi.org/10.2190/3YCX-H0MT-49DK-C61Q PMID: 17154151
105. Ross A, Kelly Y, Sacker A. Time trends in mental well-being: The polarisation of young people’s psy-
chological distress. Vol. 52, Social Psychiatry and Psychiatric Epidemiology. Ross, Andy: Research
Department of Epidemiology and Public Health, ESRC International Centre for Lifecourse Studies in
Society and Health, University College London , 1–19 Torrington Place, London, United Kingdom;
2017. p. 1147–58. https://doi.org/10.1007/s00127-017-1419-4 PMID: 28698927
106. Heim E, Maercker A, Boer D. Value Orientations and Mental Health: A Theoretical Review. Transcult
Psychiatry. 2019; 56(3):449–70. https://doi.org/10.1177/1363461519832472 PMID: 30924415
107. American Psychological Association. Report of the APA Task Force on the Sexualization of Girls.
Report of the APA Task Force on the Sexualization of Girls. Retrieved from http://www.apa.org/pi/
women/programs/girls/report-full.pdf. Washington; 2007.
108. Bayliss D, Olsen W, Walthery P. Well-being during recession in the UK. Vol. 12, Applied Research in
Quality of Life. Springer; 2017. p. 369–87. https://doi.org/10.1007/s11482-016-9465-8 PMID:
28580036
109. Oskrochi G, Bani-Mustafa A, Oskrochi Y. Factors affecting psychological well-being: Evidence from
two nationally representative surveys. PLoS One. 2018; 13(6):1–15.
110.
111.
Taylor MP, Pevalin DJ, Todd J. The psychological costs of unsustainable housing commitments. Psy-
chol Med [Internet]. 2007 Jul 16 [cited 2018 Apr 26]; 37(7):1027–36. Available from: http://www.
journals.cambridge.org/abstract_S0033291706009767 https://doi.org/10.1017/S0033291706009767
PMID: 17224094
Thomas H, Weaver N, Patterson J, Jones P, Bell T, Playle R, et al. Mental health and quality of resi-
dential environment. Br J Psychiatry. 2007; 191(DEC.):500–5. https://doi.org/10.1192/bjp.bp.107.
039438 PMID: 18055953
112. Belfer ML. Child and adolescent mental disorders: The magnitude of the problem across the globe. J
Child Psychol Psychiatry Allied Discip. 2008; 49(3):226–36.
113. World Health Organization. Strategic Guidance on Accelerating Actions for Adolescent Health in
South-east Asia Region (2018–2022). 2018.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232007 April 23, 2020
19 / 19
PLOS ONE
| null |
10.1007/s13187-021-02114-y
| null |
Data Availability Anonymized data will be made available upon request.
|
Journal of Cancer Education (2023) 38:292–300
https://doi.org/10.1007/s13187-021-02114-y
Using Protection Motivation Theory to Predict Intentions
for Breast Cancer Risk Management: Intervention Mechanisms
from a Randomized Controlled Trial
Claire C. Conley1 · Karen J. Wernli2 · Sarah Knerr3 · Tengfei Li4 · Kathleen Leppig5 · Kelly Ehrlich2 ·
David Farrell6 · Hongyuan Gao2 · Erin J. A. Bowles2 · Amanda L. Graham1,7 · George Luta4 · Jinani Jayasekera1 ·
Jeanne S. Mandelblatt1 · Marc D. Schwartz1 · Suzanne C. O’Neill1
Accepted: 1 November 2021
© The Author(s) 2021
/ Published online: 23 November 2021
Abstract
The purpose of this study is to evaluate the direct and indirect effects of a web-based, Protection Motivation Theory (PMT)–
informed breast cancer education and decision support tool on intentions for risk-reducing medication and breast MRI among
high-risk women. Women with ≥ 1.67% 5-year breast cancer risk (N = 995) were randomized to (1) control or (2) the PMT-
informed intervention. Six weeks post-intervention, 924 (93% retention) self-reported PMT constructs and behavioral inten-
tions. Bootstrapped mediations evaluated the direct effect of the intervention on behavioral intentions and the mediating role
of PMT constructs. There was no direct intervention effect on intentions for risk-reducing medication or MRI (p’s ≥ 0.12).
There were significant indirect effects on risk-reducing medication intentions via perceived risk, self-efficacy, and response
efficacy, and on MRI intentions via perceived risk and response efficacy (p’s ≤ 0.04). The PMT-informed intervention effected
behavioral intentions via perceived breast cancer risk, self-efficacy, and response efficacy. Future research should extend these
findings from intentions to behavior. ClinicalTrials.gov Identifier: NCT03029286 (date of registration: January 24, 2017).
Keywords Breast cancer · Prevention · Risk management · Risk-reducing medication · Magnetic resonance imaging
(MRI) · Protection Motivation Theory
* Suzanne C. O’Neill
[email protected]
1 Department of Oncology, Lombardi Comprehensive Cancer
Center, Georgetown University Medical Center, 2115
Wisconsin Avenue NW, Suite 300, Washington, DC 20007,
USA
2 Kaiser Permanente Washington Health Research Institute,
Seattle, WA, USA
3 Department of Health Services, University of Washington,
Seattle, WA, USA
4 Department of Biostatistics, Bioinformatics,
and Biomathematics, Georgetown University, Washington,
DC, USA
5 Washington Permanente Medical Group, Seattle, WA, USA
6 PeopleDesigns, Raleigh-Durham, NC, USA
7 Truth Initiative, Washington, DC, USA
1 3
Introduction
National guidelines present options for breast cancer risk
management among women with elevated risk [1]. Women
with an estimated 5-year risk of breast cancer ≥ 1.67% and at
low risk for adverse events may consider risk-reducing medi-
cation (tamoxifen or raloxifene). For high-risk women, these
medications reduce 5-year breast cancer risk by 30–55% [2].
Despite the potential benefits, uptake of risk-reducing medi-
cation remains low. In the USA, of the 65 million women
aged 35–79 without a history of breast cancer, about 10
million are eligible for risk-reducing medication; less than
500,000 use risk-reducing medication [3].
High-risk women with an estimated lifetime breast can-
cer risk ≥ 20% may also consider annual screening with
breast magnetic resonance imaging (MRI) [1]. For these
women, annual screening breast MRI is recommended in
addition to annual mammography. The limited research
on uptake of MRI among high-risk women provides esti-
mates ranging from 9 to 29% [4]. Thus, many high-risk
Journal of Cancer Education (2023) 38:292–300
293
women are not following guidelines for breast cancer risk
management or taking full advantage of the interventions
available to them.
Despite the low rates of risk-reducing medication and
MRI among high-risk women, efforts to increase uptake
have been few and have had limited success [5–8]. How-
ever, previously tested interventions have not been informed
by behavior change theories. To fill this gap, we developed
a web-based, breast cancer education and decision support
tool for women at an elevated risk of developing breast can-
cer. This tool was based on Protection Motivation Theory
(PMT) [9]. According to PMT, women are most likely to
adopt risk management behaviors when they believe that:
(1) they are at significant breast cancer risk, (2) risk-reduc-
ing medication and/or MRI could be effective at reducing or
managing their risk, and (3) risk-reducing medication and/or
MRI will be associated with few adverse effects.
A randomized controlled trial compared the PMT-
informed intervention to a control arm that directed par-
ticipants to relevant online health information [10, 11].
One year post-intervention, we found no improvement in
uptake of risk-reducing medication due to the intervention.
However, among women with ≥ 2.50% 5-year risk for breast
cancer, we did observe 4.5-fold increased odds of receipt of
breast MRI in the intervention arm compared to the control
arm [11]. The null intervention results may be due to the
time frame in which outcomes were assessed (1 year follow-
ing intervention delivery). In addition, risk-reducing medica-
tion and breast MRI are both physician-mediated behaviors,
in that they require a prescription or a referral.
Given these limitations, we wanted to examine the
intervention’s impact on a proximal outcome: intentions for
risk-reducing medication and breast MRI at 6 weeks post-
intervention. Intentions are an important necessary condition
for engaging in recommended health behaviors [12]; thus,
examining the intervention’s effects on behavioral intentions
would provide important information regarding the overall
null effects of the main trial. Additionally, we examined
PMT constructs as intervention process variables. Together,
these analyses would guide intervention modifications or
adaptations.
In the present study, a secondary data analysis examined
the direct and indirect effects of the PMT-informed inter-
vention on intentions for risk-reducing medication and/or
breast MRI at 6 weeks post-intervention. We hypothesized
that (1) the intervention would have a direct effect on inten-
tions for breast cancer risk management, such that women
in the intervention arm would report stronger intentions than
women in the control arm, and (2) PMT variables would
mediate the relationship between study arm and intentions
for breast cancer risk management. Our primary outcome
of interest was intentions for risk-reducing medication. We
also examined intentions for MRI as a secondary outcome.
Methods
Participants and Procedures
This two-arm randomized controlled trial (ENGAGED-2,
ClinicalTrials.gov identifier: NCT03029286) has been
described in detail elsewhere [10, 11]. The trial was approved
by the Georgetown University Institutional Review Board (IRB
#2015-0687). Briefly, eligible participants were women aged
40–69 and members of Kaiser Permanente Washington, an
integrated healthcare delivery system. All women had a normal
screening mammogram result between 2016 and 2018, and
had an elevated risk of an interval breast cancer per the Breast
Cancer Surveillance Consortium (BCSC) 5-Year Risk Calculator
[13]. Exclusion criteria included a personal history of cancer,
previous referral for cancer genetic counseling, and/or prior
genetic testing as documented in the electronic health records.
Women were randomized 1:1 to the intervention or control
arm at study sample identification (prior to recruitment).
Women randomized to usual care were instructed to review
information on the American Cancer Society website related
to breast cancer risk, prevention, and cancer screening. The
PMT-informed intervention is described below.
A total of 995 women provided verbal informed con-
sent, enrolled in the study, and completed a baseline inter-
view by telephone (intervention = 492, control = 503). Six
weeks later, 93% of participants (n = 924) completed a
follow-up survey (intervention = 459 [93%], control = 465
[92%]) and are included in the analyses presented here.
Intervention
The PMT-informed intervention has been previously described
[10]. In line with PMT [9], the intervention targeted threat
appraisals (perceived breast cancer severity and risk) and coping
appraisals (self-efficacy, response efficacy, and response cost).
Specifically, threat appraisals were targeted through presentation
of factual information about breast cancer and personalized 5- and
10-year breast cancer risk estimates. Self-efficacy was targeted
through allowing participants to create a tailored question prompt
list, and encouraging them to make an appointment with their
provider to discuss their questions and concerns. Response
efficacy and response cost were targeted through presentation of
tailored risks and benefits of risk-reducing medication and breast
MRI and an interactive values clarification exercise.
Measures
PMT constructs and intentions for breast cancer risk
management were assessed via self-report at the 6-week
follow-up time point.
1 3
294
PMT Constructs
Cancer Worry We adapted the 3-item Lerman Breast Can-
cer Worry Scale [14] to assess worry about getting breast
cancer in the future (e.g., “How often did you worry about
getting breast cancer during the past two weeks?). Partici-
pants rated each item on a 4-point Likert scale (1 = “never”
to 4 = “almost all the time/a lot”). Items were summed to
generate a total score ranging from 1 to 12, with higher
scores indicating greater worry.
Breast Cancer Severity Participants rated their agree-
ment with the statement “I believe that breast cancer is
severe” on a 5-point Likert scale (1 = “strongly disagree” to
5 = “strongly agree”).
Perceived Breast Cancer Risk Patients estimated their per-
sonal risk of experiencing breast cancer in the next 5 years
on a scale from 0% (no chance) to 100% (definitely will).
Self‑Efficacy Self-efficacy is an individual’s confidence in
performing a behavior; in the present study, participants
responded to items about self-efficacy of using risk-reducing
medication and MRI on a 5-point Likert scale (1 = “strongly
disagree” to 5 = “strongly agree”). The four items assessed
participants’ confidence in their ability to manage medica-
tion side effects, take a pill every day, manage discomfort
during an MRI, and have an MRI every year. Items were
averaged to generate separate self-efficacy scores for MRI
and risk-reducing medication. Total scores ranged from 1 to
5; higher scores indicate higher self-efficacy.
Response Efficacy Response efficacy is an individual’s
belief as to whether or not a behavior will avoid a health
threat. Participants responded to nine items (three each for
tamoxifen, raloxifene, and MRI) on a 5-point Likert scale
(1 = “strongly disagree” to 5 = “strongly agree”). Risk-reduc-
ing medication items assessed participants’ perceptions that
tamoxifen and raloxifene are effective in preventing breast
cancer, could significantly improve future health, and are
an effective way to reduce breast cancer risk. MRI items
assessed participants’ perceptions that MRI is effective in
finding breast cancer, could significantly improve future
health, and is an effective way to find breast cancer early.
Items were averaged to generate separate response efficacy
scores for MRI and risk-reducing medication. Total scores
ranged from 1 to 5; higher scores indicate higher response
efficacy.
Response Cost Response cost is an individual’s perceptions
of the downsides of a behavior. Participants responded to
three items assessing the costs of risk-reducing medication
1 3
Journal of Cancer Education (2023) 38:292–300
and four assessing the costs of MRI using a 5-point Likert
scale (1 = “strongly disagree” to 5 = “strongly agree”). Risk-
reducing medication items included side effects, taking a pill
daily, and cost. MRI items included lack of breast cancer
risk reduction, discomfort, cost, and potential additional,
unneeded tests or treatments. Items were averaged to gener-
ate separate response cost scores for MRI and risk-reducing
medication. Total scores ranged from 1 to 5; higher scores
indicate higher response cost.
Primary outcome: intentions for risk‑reducing medication
To measure participants’ intentions to use risk-reducing
medication, participants rated their likelihood of using
tamoxifen in the next year, and their likelihood of using
raloxifene in the next year on a 5-point Likert scale
(1 = “strongly disagree” to 5 = “strongly agree”). The two
items were averaged to create a single score representing
intentions for risk-reducing medication.
Secondary outcome: intentions for MRI
We measured intentions for MRI by asking participants
to rate their likelihood of having a breast MRI in the next
year using a 5-point Likert scale (1 = “strongly disagree” to
5 = “strongly agree”).
Statistical Analyses
Descriptive statistics were used to characterize the sam-
ple demographicsand the 6-week follow-up assessment of
PMT constructs and behavioral intentions. We described
categorical variables using frequencies and percentages, and
continuous variables using means and standard deviations.
Categorical variables were compared using chi-squared
tests; Student t-tests were used for the continuous variables.
To identify variables to include as mediators in bootstrapped
mediation models, we examined correlations between PMT
constructs and outcomes at the 6-week follow-up; only
potential mediators that were significantly correlated with
the outcomes of interest (p < 0.05) were included in primary
analyses.
Direct and indirect effects of PMT variables on intentions
for using risk-reducing medication or MRI were examined
using the PROCESS macro for SPSS (Model 4) [15]. The
PROCESS macro allows for the estimation of moderation
and mediation effects via a bootstrapping procedure. With
bootstrapping, effects are estimated based on a large number
of bootstrapped resamples (e.g., 10,000 resamples used here)
generated from the original data by random sampling with
replacement. If the 95% confidence interval (CI) for an effect
does not include zero, it indicates the significance of the
Journal of Cancer Education (2023) 38:292–300
295
effect at the 0.05 level. In the present analyses, treatment arm
(intervention v. control) was specified as the independent
variable. Threat appraisals (cancer worry, perceived breast
cancer severity, and perceived breast cancer risk) and cop-
ing appraisals (self-efficacy, response efficacy, and response
cost) were specified as parallel mediators. Finally, breast
cancer risk management intentions (risk-reducing medica-
tion and breast MRI) were specified as the outcome varia-
bles. Two models were run, one for risk-reducing medication
intentions and one for breast MRI intentions.
All analyses were conducted using IBM SPSS for Win-
dows, version 27 (IBM Corp., Armonk, NY, USA).
Results
The sample was primarily non-Hispanic White (95%), in
middle adulthood (M = 62 years, range = 40–69), with a
college degree or greater (74%) and an annual household
income ≥ $70,001 (56%) (see prior descriptions of this sam-
ple [10, 11]). The majority of the women were pre-menopau-
sal (93%). About half had a family history of breast cancer
(45%) or a prior breast biopsy (45%). Most participants had
heterogeneously dense breast tissue (56%) and high (66%)
or very high (9%) breast cancer risk.
Intentions for Risk‑Reducing Medication and Breast
MRI
Intentions for risk-reducing medication or MRI at 6 weeks
were low overall (Table 1). Compared to the control group,
the intervention group had significantly greater intentions for
risk-reducing medication (M = 1.8 versus 1.7, p = 0.03). The
intervention and control groups did not significantly differ
on intentions for breast MRI (M = 2.9 versus 2.9, p = 0.10).
Correlations Between PMT Constructs
and Behavioral Intentions
In bivariate analyses, intentions for risk-reducing medi-
cation and MRI were significantly correlated with cancer
worry, perceived breast cancer risk, self-efficacy for risk-
reducing medication, response efficacy for risk-reducing
medication, and response cost for risk-reducing medication
(all p’s ≤ 0.001) (Table 1). Perceived breast cancer severity
was not associated with intentions for risk-reducing medica-
tion (p = 0.97) or intentions for MRI (p = 0.42). Thus, boot-
strapped mediation analyses did not include perceived breast
cancer severity as a mediator.
Table 1 Descriptive statistics by intervention group and correlations between mediators and outcome variables at 6 weeks (n = 924)
Intervention
Control
(n = 459)
(n = 465)
p-value
Correlation with behavioral
intentions (r)
Risk-reducing
medication
MRI
Mediators
Cancer worry (M, SD)
Perceived breast cancer severity (M, SD)
Perceived 5-year breast cancer risk (M, SD)
Self-efficacy (M, SD)
Risk-reducing medication
MRI
Response efficacy (M, SD)
Risk-reducing medication
MRI
Response cost (M, SD)
Risk-reducing medication
MRI
Outcomes
Behavioral intentions (M, SD)
Risk-reducing medication
MRI
M, mean; SD, standard deviation
* p < 0.05
** p < 0.005
2.1 (1.69)
4.4 (0.87)
19.9 (19.48)
2.1 (1.64)
4.4 (0.82)
25.9 (21.38)
3.3 (0.88)
4.0 (0.98)
3.0 (0.62)
3.8 (0.77)
3.5 (0.77)
3.0 (0.79)
3.5 (0.83)
4.1 (0.94)
2.9 (0.60)
3.7 (0.78)
3.4 (0.80)
2.9 (0.79)
1.8 (0.91)
2.9 (1.07)
1.7 (0.91)
2.9 (0.98)
0.979
0.863
< 0.0001**
0.002**
0.107
0.008*
0.012*
0.078
0.092
0.029*
0.098
0.11**
-0.01
0.16**
0.04
0.22**
0.11**
0.32**
-0.08*
-0.21**
1
0.26**
0.14**
0.04
0.16**
0.25**
0.07*
0.31**
0.13**
-0.21**
-0.05
0.26**
1
1 3
296
Journal of Cancer Education (2023) 38:292–300
Mediating Effect of PMT Constructs on Intentions
for Risk‑Reducing Medication
Discussion
The bootstrapped mediation model predicting intentions for
risk-reducing medication explained 16% of the variance in
intentions for risk-reducing medication (R2 = 0.16) (Table 2,
Fig. 1a). Neither the total effect nor the direct effect of study
arm on intentions for risk reducing medication was signifi-
cant. There were significant indirect effects of study arm on
intentions for risk-reducing medication via perceived breast
cancer risk (p = 0.004), self-efficacy (p = 0.04), and response
efficacy (p = 0.01). Compared to women in the control arm,
women in the intervention arm reported lower perceived
breast cancer risk, lower self-efficacy, and higher response
efficacy. In turn, perceived breast cancer risk, self-efficacy,
and response efficacy were all positively associated with
intentions for risk-reducing medication.
Mediating Effect of PMT Constructs on Intentions
for Breast MRI
The bootstrapped mediation model predicting intentions for
breast MRI explained 15% of the variance in intentions for
breast MRI (R2 = 0.15) (Table 2, Fig. 1b). The direct effect
of study arm on intentions for breast MRI was not signifi-
cant (B = 0.0003, SE = 0.01, p = 0.996, 95% C.I. = [− 0.13,
0.13]). Neither the total effect nor the direct effect of study
arm on intentions for MRI was significant. There were sig-
nificant indirect effects of study arm on intentions for MRI
via perceived breast cancer risk (p = 0.02) and MRI response
efficacy (p = 0.01). Compared to women in the control arm,
women in the intervention arm reported lower perceived
breast cancer risk and higher response efficacy. In turn,
perceived breast cancer risk and response efficacy were all
positively associated with intentions for breast MRI.
We evaluated whether a web-based, Protection Motivation
Theory–informed breast cancer education and decision sup-
port tool could increase intentions for risk-reducing medica-
tion and breast MRI compared to an active control arm. The
data presented here demonstrate the important role of threat
appraisals, like cancer worry and perceived breast cancer
risk, on intentions to engage in breast cancer risk mitiga-
tion. Coping appraisals—including self-efficacy, response
efficacy, and response cost—were also related to women’s
intentions for breast cancer risk management.
We identified three significant mediators of the relation-
ship between study arm and intentions for breast cancer risk
management: perceived breast cancer risk, self-efficacy, and
response efficacy. Compared to women in the control arm,
women in the intervention arm reported significantly lower
perceived breast cancer risk at the 6-week follow-up. As
women tend to overestimate their risk of breast cancer [16],
it is likely that the PMT-informed intervention appropri-
ately decreased perceived risk via presentation of personal-
ized breast cancer risk estimates. Paradoxically, while the
intervention led to more accurate risk comprehension, it is
also possible that the reduction in perceived risk limited the
impact of the intervention on intentions for risk-reducing
medication. This may have been particularly salient for
women in this study who had not previously received breast
cancer risk information in routine clinical care. Thus, partic-
ipants may have been reassured by the lower than anticipated
risk that was conveyed by the intervention.
The intervention group also reported lower self-efficacy
for risk-reducing medication at the 6-week follow-up.
While there has been little research on the role of self-effi-
cacy in uptake of and adherence to risk-reducing medica-
tion, self-efficacy has been shown to play an important role
Table 2 Results of bootstrapped
mediation models
1 3
Model 1: intentions for risk-reducing
medication
(n = 901)
Model 2: intentions for breast MRI
(n = 896)
B
0.10
0.11
SE
0.06
0.06
p
95% CI
B
SE
p
95% CI
0.115
0.055
[− 0.02, 0.21]
[− 0.002, 0.22]
0.0003 0.07
0.06
0.01
0.996
0.879
[− 0.13, 0.13]
[− 0.12, 0.14]
0.001 0.004
0.740
− 0.03
− 0.02
0.04
0.01
0.01
0.02
0.005
0.043
0.009
[− 0.01, 0.01]
[− 0.06, − 0.01]
[− 0.04, − 0.004]
[0.01, 0.08]
0.001 0.01
0.01
0.01
0.02
− 0.03
− 0.01
0.04
0.864
0.020
0.192
0.013
[− 0.01, 0.02]
[− 0.05, − 0.01]
[− 0.04, 0.005]
[0.01, 0.07]
Total effect
Direct effect
Indirect effects
Cancer worry
Perceived risk
Self-efficacy
Response
efficacy
Response cost
− 0.01
0.01
0.192
[− 0.03, 0.004]
− 0.01
0.01
0.208
[− 0.03, 0.003]
B, unstandardized coefficient; SE, standard error; CI, confidence interval
Journal of Cancer Education (2023) 38:292–300
297
Fig. 1 Bootstrapped mediation models examining direct and indirect effects of the intervention on a intentions for risk-reducing medication and
b intentions for breast MRI. All coefficients are unstandardized, and asterisks indicate statistical significance (*p < 0.05, **p < 0.005)
in adherence to other types of medications [17]. Our inter-
vention targeted self-efficacy by encouraging participants
to make an appointment with their provider and providing
the opportunity to create a question prompt list to use
in that appointment. Our relatively short follow-up time
frame (6 weeks) may have limited participants’ ability to
1 3
298
Journal of Cancer Education (2023) 38:292–300
utilize these strategies. We have previously reported that
the proportion of women in the intervention group who
had “discussions” with their healthcare providers about
risk-reducing medication increased substantially from the
6-week follow-up (5%) to the 12-month follow-up (14%)
[11]. Thus, at the 6-week follow-up, participants’ self-
efficacy for risk-reducing medication may have reflected
the educational components of the intervention, which pro-
vided detailed information about tamoxifen and raloxifene.
This included the need to take the medication every day
and the common side effects for these medications. The
intervention’s impact on self-efficacy may be similar to
the paradoxical effect seen in prior studies that discussion
of the medication regimen and side effects can actually
lower self-efficacy for risk-reducing medication [18]. A
prior systematic review of adherence to risk-reducing med-
ication noted self-efficacy as a key barrier to adherence
[19]. Further examination of its role in initiation could be
warranted as well.
Compared to women in the control arm, women in the
intervention arm reported greater response efficacy for
risk-reducing medication and breast MRI at the 6-week
follow-up. Our intervention targeted response efficacy in
two ways: presenting tailored risks and benefits of risk-
reducing medication and breast MRI, and engaging partici-
pants in an interactive values clarification exercise. While
a dismantling study would be needed in order to assess the
relative effectiveness of these components, it is likely that
education about risk-reducing strategies played an important
role, given the demonstrated lack of knowledge about risk-
reducing medication [20] and supplemental breast screen-
ing [21] among women with elevated risk for breast cancer.
However, it should be noted that the group differences in
mean response efficacy scores were relatively small and may
not be clinically significant despite statistical significance.
These indirect effects must be interpreted in light of the
null total and direct effects of the intervention on intentions
for risk-reducing medication and breast MRI. Although tra-
ditional approaches to mediation require a direct effect in
order to estimate and test hypotheses about indirect effects,
current thinking about mediation analysis does not [22].
Instead, the relationship between two variables (i.e., the
total effect) is conceptualized as the sum of many different
paths of influence, including indirect effects (i.e., mediation)
and/or direct effects. Multiple indirect effects might cancel
out, resulting in a null direct effect. In the present study, we
observed both a negative indirect effect via perceived risk
and self-efficacy, and a positive indirect effect via response
efficacy. In other words, the intervention might have both
increased and decreased intentions, via different pathways,
resulting in no change overall.
Our results support the applicability of PMT to breast
cancer risk management. Of the six PMT constructs
1 3
examined, five were significantly related to intentions for
risk-reducing medication and breast MRI. In addition, the
direction of the relationships between PMT constructs and
behavioral intentions was theoretically consistent. Interest-
ingly, perceived breast cancer severity was not significantly
related to intentions for risk-reducing behaviors, and as a
result, was not included in the final models. This contrasts
with prior meta-analyses examining the relationship between
PMT variables and behavioral intentions that have demon-
strated a small but significant effect of perceived severity
[23]. The discrepancy between the results presented here and
prior findings may be due in part to differences in the meas-
urement of perceived severity. In the present study, over 90%
of participants “agreed” or “strongly agreed” with the state-
ment “breast cancer is severe” at baseline. The limited range
in perceived breast cancer severity may have resulted in a
“ceiling effect”, making it difficult to discriminate among
subjects reporting high levels of perceived severity.
These results have clinical implications for future inter-
ventions in this area. The tendency for women to overes-
timate their breast cancer risk is well-documented in the
literature, and prior risk communication interventions have
promoted more accurate breast cancer risk perceptions
through the provision of a personalized risk estimates [16].
Accurate risk perceptions are critical to making informed
health decisions, but the consequences of this reduction for
motivation of health-protecting behaviors requires further
consideration. While the current trial reported not only the
participant’s 5- and 10-year breast cancer risk, but also the
average risk for a woman her age and race, future studies
with individuals with clinically elevated cancer risk could
place accurate risk perceptions in the context of clinical
guidelines.
Promoting medication self-efficacy has become a focus of
interventions to promote adherence to oral medications, not
only in cancer but in other chronic conditions such as dia-
betes [24] and arthritis [25]. Unlike control of these chronic
conditions, where medication is prescribed to address symp-
toms, the use of medication for the reduction of breast can-
cer risk is more preference-sensitive and cannot be tied to an
observable metric. Therefore, support for self-efficacy may
be even more essential when women are making decisions
around initiation of the medication.
Our study had two key strengths. First, we examined theo-
retical constructs in the setting of a randomized controlled
trial. Second, we had a large study sample with a relatively
high retention rate; 93% of the baseline participants com-
pleted the 6-week follow-up assessment.
Study results must be interpreted in light of some limi-
tations. First, the specified models do not meet the criteria
for a “true” test of mediation as the PMT constructs and
behavioral intentions were both assessed at the 6-week fol-
low-up time point [22]. Second, the study sample excluded
Journal of Cancer Education (2023) 38:292–300
299
women who had prior cancer genetic counseling or test-
ing, a group most likely to be eligible for and amenable to
screening MRI. Third, prior publications have documented
that this sample was demographically homogenous [11].
Furthermore, women needed to access online information
to participate in the study. Thus, the generalizability of
the findings to other ethnic and minority groups or to the
underserved is unknown.
In summary, this trial evaluated a novel web-based
intervention informed by PMT that provides personal-
ized breast cancer risk communication and decision sup-
port. While the intervention did not have a direct effect
on intentions for risk-reducing medication or breast MRI,
we did observe significant indirect effects of the inter-
vention on breast cancer risk management intentions via
changes in perceived breast cancer risk, response efficacy,
and self-efficacy. Interventions that address perceived
risk and boost self-efficacy and response efficacy may be
particularly effective in the context of breast cancer risk
management.
Declarations
Ethics Approval All procedures were approved by the Georgetown
University Institutional Review Board (IRB #2015–0687). This study
confirms to the standards outlined in the Declaration of Helsinki and
US Federal Policy for the Protection of Human Subjects.
Consent to Participate All persons gave their informed consent prior
to study participation.
Conflict of Interest The authors have no conflicts of interest to report.
Open Access This article is licensed under a Creative Commons Attri-
bution 4.0 International License, which permits use, sharing, adapta-
tion, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source,
provide a link to the Creative Commons licence, and indicate if changes
were made. The images or other third party material in this article are
included in the article's Creative Commons licence, unless indicated
otherwise in a credit line to the material. If material is not included in
the article's Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a
copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
Author Contribution CCC: conceptualization, formal analysis, visu-
alization, writing—original draft; KJW: conceptualization, funding
acquisition, investigation, supervision, writing—review and editing;
SK: investigation, writing—review and editing; TL: data curation,
formal analysis, writing—original draft; KL: investigation, resources,
writing—review and editing; KE: data curation, investigation, project
administration, writing—review and editing; DF: data curation, inves-
tigation, resources, software, writing—review and editing; HG: data
curation, investigation, writing—review and editing; EJAB: data cura-
tion, investigation, writing—review and editing; ALG: methodology,
resources, writing—review and editing; GL: conceptualization, data
curation, formal analysis, supervision, writing—review and editing; JJ:
writing—review and editing; JSM: conceptualization, writing—review
and editing; MDS: conceptualization, writing—review and editing;
SCO: conceptualization, data curation, funding acquisition, investiga-
tion, supervision, writing—review and editing.
Funding This study was supported by the National Can-
cer Institute (R01CA190221, PI: O’Neill; R50CA211115, PI:
Bowles; K99CA241397, PI: Jayasekera; R03CA259896, PI:
Jayasekera; U01CA152958, PI: Mandelblatt; R35CA197289, PI: Man-
delblatt; and P30CA051008, PI: Weiner), the National Human Genome
Research Institute (K08HG010488, PI: Knerr), the Agency for Health-
care Research and Quality (K12HS022982; PI: Sullivan), the American
Society of Preventive Oncology and Breast Cancer Research Foun-
dation (ASPO-19–001, PI: Conley), and the American Cancer Soci-
ety (ACS IRG 92–152-20, PI: Atkins; and ACS IRG 17–177-23, PI:
Conley). Collection of breast cancer risk information is supported by
the National Cancer Institute–funded Breast Cancer Surveillance Con-
sortium (P01CA154292, PI: Miglioretti; U54CA163303, PI: Sprague;
and HHSN261201100031C). The content is solely the responsibility
of the authors and does not necessarily represent the official views of
the National Institutes of Health.
Data Availability Anonymized data will be made available upon
request.
References
1. National Comprehensive Cancer Network (NCCN). Breast cancer
risk reduction (Version 1.2020). NCCN Clinical Practice Guide-
lines in Oncology (NCCN Guidelines®) 2020 June 18, 2020];
Available from: https:// www. nccn. org/ profe ssion als/ physi cian_
gls/ pdf/ breast_ risk. pdf.
2. Nelson HD et al (2019) Medication use for the risk reduction of
primary breast cancer in women: updated evidence report and sys-
tematic review for the US Preventive Services Task Force. JAMA
322(9):868–886
3. Freedman AN et al (2003) Estimates of the number of US women
who could benefit from tamoxifen for breast cancer chemopreven-
tion. J Natl Cancer Inst 95(7):526–532
4. Wernli KJ et al (2014) Patterns of breast magnetic reso-
nance imaging use in community practice. JAMA Intern Med
174(1):125–132
5. Brewster AM et al (2018) A system-level approach to improve
the uptake of antiestrogen preventive therapy among women with
atypical hyperplasia and lobular cancer in situ. Cancer Prev Res
11(5):295–302
6. Kukafka R et al (2018) Pilot study of decision support tools on
breast cancer chemoprevention for high-risk women and health-
care providers in the primary care setting. BMC Med Inform
Decis Mak 18(1):134
7. Brinton JT et al (2018) Informing women and their physicians
about recommendations for adjunct breast MRI Screening: a
cohort study. Health Commun 33(4):489–495
8. Oeffinger KC et al (2019) Promoting Breast Cancer Surveillance:
The EMPOWER Study, a Randomized Clinical Trial in the Child-
hood Cancer Survivor Study. J Clin Oncol 37(24):2131–2140
9. Rogers RW (1975) A protection motivation theory of fear appeals
and attitude change. J Psychol 91(1):93–114
10. Knerr S et al (2017) A web-based personalized risk communi-
cation and decision-making tool for women with dense breasts:
Design and methods of a randomized controlled trial within an
integrated health care system. Contemp Clin Trials 56:25–33
1 3
300
Journal of Cancer Education (2023) 38:292–300
11 Wernli KJ et al (2021) Effect of personalized breast cancer risk
tool on chemoprevention and breast imaging: The Engaged-2 trial.
JNCI Cancer Spectrum 5:pkaa114
12. McEachan RRC et al (2011) Prospective prediction of health-
related behaviours with the theory of planned behaviour: A meta-
analysis. Health Psychol Rev 5(2):97–144
13. Tice JA et al (2015) Breast density and benign breast disease: Risk
assessment to identify women at high risk of breast cancer. J Clin
Oncol 33(28):3137–3143
14. Lerman C et al (1991) Psychological and behavioral implications
of abnormal mammograms. Ann Intern Med 114(8):657–661
15. Hayes AF (2017) Introduction to mediation, moderation, and con-
ditional process analysis: a regression-based approach, 2nd edn.
Guilford Press, New York
16. Xie Z et al (2019) Risk estimation, anxiety, and breast can-
cer worry in women at risk for breast cancer: a single-arm
trial of personalized risk communication. Psychooncology
28(11):2226–2232
17. Nafradi L, Nakamoto K, Schulz PJ (2017) Is patient empower-
ment the key to promote adherence? A systematic review of the
relationship between self-efficacy, health locus of control and
medication adherence. PloS one 12(10):e0186458
18. Juraskova, I. and C (2013) Bonner, Decision aids for breast cancer
chemoprevention. Springer.
19. Lin C et al (2017) Breast cancer oral anti-cancer medication
adherence: a systematic review of psychosocial motivators and
barriers. Breast Cancer Res Treat 165(2):247–260
20. Thorneloe RJ et al (2020) Knowledge of potential harms and
benefits of tamoxifen among women considering breast cancer
preventive therapy. Cancer Prev Res 13(4):411–422
21. Aminawung JA et al (2020) Breast cancer supplemental screen-
ing: women’s knowledge and utilization in the era of dense breast
legislation. Cancer Med 9(15):5662–5671
22. Hayes AF (2009) Beyond Baron and Kenny: statistical mediation
analysis in the new millennium. Commun Monogr 76(4):408–420
23. Milne S, Sheeran P, Orbell S (2000) Prediction and intervention
in health-related behavior: a meta-analytic review of protection
motivation theory. J Appl Soc Psychol 30(1):106–143
24. Gonzalez JS, Tanenbaum ML, Commissariat PV (2016) Psy-
chosocial factors in medication adherence and diabetes self-
management: implications for research and practice. Am Psychol
71(7):539
25. McCulley C et al (2018) Association of medication beliefs, self-
efficacy, and adherence in a diverse cohort of adults with rheuma-
toid arthritis. J Rheumatol 45(12):1636–1642
Publisher's Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
1 3
|
Could not heal snippet
|
10.3389/fmicb.2019.03155
|
DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article are made
available by the authors, without undue reservation, through
Zenodo (doi: 10.5281/zenodo.3582838).
|
DATA AVAILABILITY STATEMENT The raw data supporting the conclusions of this article are made available by the authors, without undue reservation, through Zenodo ( doi: 10.5281/zenodo.3582838 ).
|
fmicb-10-03155
January 13, 2020
Time: 16:53
# 1
ORIGINAL RESEARCH
published: 22 January 2020
doi: 10.3389/fmicb.2019.03155
Interacting Temperature, Nutrients
and Zooplankton Grazing Control
Phytoplankton Size-Abundance
Relationships in Eight Swiss Lakes
Francesco Pomati1,2*, Jonathan B. Shurin3, Ken H. Andersen4, Christoph Tellenbach1
and Andrew D. Barton3,5
1 Aquatic Ecology, Eawag: Swiss Federal Institute of Aquatic Science and Technology, Dübendorf, Switzerland, 2 Institute
of Integrative Biology, ETH-Zurich, Zurich, Switzerland, 3 Department of Ecology Behavior and Evolution, University
of California, San Diego, La Jolla, CA, United States, 4 Centre for Ocean Life, DTU Aqua, Technical University of Denmark,
Lyngby, Denmark, 5 Scripps Institution of Oceanography, La Jolla, CA, United States
Biomass distribution among size classes follows a power law where the Log-abundance
of taxa scales to Log-size with a slope that responds to environmental abiotic and
biotic conditions. The interactions between ecological mechanisms controlling the
locally realized size-abundance relationships (SAR) are however not well
slope of
understood. Here we tested how warming, nutrient levels, and grazing affect the slope
of phytoplankton community SARs in decadal time-series from eight Swiss lakes of
the peri-alpine region, which underwent environmental forcing due to climate change
and oligotrophication. We expected rising temperature to have a negative effect on
slope (favoring small phytoplankton), and increasing nutrient levels and grazing pressure
to have a positive effect (benefiting large phytoplankton). Using a random forest
approach to extract robust patterns from the noisy data, we found that the effects
of temperature (direct and indirect through water column stability), nutrient availability
(phosphorus and total biomass), and large herbivore (copepods and daphnids) grazing
and selectivity on slope were non-linear and interactive. Increasing water temperature
or total grazing pressure, and decreasing phosphorus levels, had a positive effect on
slope (favoring large phytoplankton, which are predominantly mixotrophic in the lake
dataset). Our results therefore showed patterns that were opposite to the expected
long-term effects of temperature and nutrient levels, and support a paradigm in which
(i) small phototrophic phytoplankton appear to be favored under high nutrients levels,
low temperature and low grazing, and (ii) large mixotrophic algae are favored under
oligotrophic conditions when temperature and grazing pressure are high. The effects
of temperature were stronger under nutrient limitation, and the effects of nutrients and
grazing were stronger at high temperature. Our study shows that the phytoplankton local
SARs in lakes respond to both the independent and the interactive effects of resources,
grazing and water temperature in a complex, unexpected way, and observations from
long-term studies can deviate significantly from general theoretical expectations.
Keywords: size spectra, bottom–up and top–down controls, main effects and interactions, non-linear effects,
random forests, eutrophication, climate change
Edited by:
Susana Agusti,
King Abdullah University of Science
and Technology, Saudi Arabia
Reviewed by:
Arda Özen,
Çankırı Karatekin University, Turkey
María Florencia Gutierrez,
National Institute of Limnology (INALI),
Argentina
*Correspondence:
Francesco Pomati
[email protected]
Specialty section:
This article was submitted to
Aquatic Microbiology,
a section of the journal
Frontiers in Microbiology
Received: 07 November 2019
Accepted: 30 December 2019
Published: 22 January 2020
Citation:
Pomati F, Shurin JB,
Andersen KH, Tellenbach C and
Barton AD (2020) Interacting
Temperature, Nutrients
and Zooplankton Grazing Control
Phytoplankton Size-Abundance
Relationships in Eight Swiss Lakes.
Front. Microbiol. 10:3155.
doi: 10.3389/fmicb.2019.03155
Frontiers in Microbiology | www.frontiersin.org
1
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 2
Pomati et al.
Environmental Drivers of Phytoplankton Size
INTRODUCTION
In aquatic ecosystems, the size of planktonic organisms is a key
determinant of community structure and food-web dynamics
(Marañón, 2015). The relationship between size and abundance
emerges from organism traits and ecological interactions, and
describes how biomass is partitioned among the biota within
an ecosystem (Sprules et al., 2016). Smaller phytoplankton are
typically more numerous than larger phytoplankton in freshwater
(Sprules and Munawar, 1986; Sprules et al., 1991; Gaedke et al.,
2004) and marine ecosystems (Sheldon et al., 1972; Rodriguez
and Mullin, 1986; Cavender-Bares et al., 2001; Huete-Ortega
et al., 2014; Marañón, 2015). This negative relationship between
abundance and body size is often called the size spectrum, and
can be generally described as:
Abundance = a · Sizeb
where a is a constant and b is the power spectral slope.
Observations from fresh and marine waters indicate that
community size spectra in aquatic ecosystems tend to conform to
the above power law, and exponent b is often close to −1 (larger
cells are scarce relative to smaller cells) (Cavender-Bares et al.,
2001; Gaedke et al., 2004; Marañón, 2015; Sprules et al., 2016).
A less negative slope indicates a more even distribution of large
and small cells. Alternatively, the size-abundance relationship
(SAR) can be depicted by plotting the density of taxa as a
function of their biovolume in a log–log space (Figure 1A). In
phytoplankton, transient states or variations in phytoplankton
SARs can have significant implications for aquatic food-webs and
biogeochemical cycling. For example, communities with a greater
proportion of larger phytoplankton cells are generally associated
with algal blooms and lower biomass transfer to the herbivore
food-chain (Behrenfeld and Boss, 2013; Yvon-Durocher et al.,
2015; Cloern, 2018).
Slopes of size spectra and SARs are strong indicators of
environmental and biotic impacts on community structure,
particularly temperature, resource supply, and size-selective
grazing. Competition theory in ecology predicts that temperature
should favor small organismal size relative to large, also
in phytoplankton (Reuman et al., 2014). This hypothesis
is consistent with experimental evidence that, under higher
temperature, smaller organisms maintain high metabolic rates
because they are more efficient in resource uptake (due to
more favorable surface to volume ratio) (Atkinson et al., 2003;
Andersen et al., 2016). Increasing temperature of water should
therefore decrease the slope (i.e., more negative values) of the
phytoplankton community SAR, by increasing the prevalence
taxa relative to large ones (Winder et al., 2009;
of small
Yvon-Durocher et al., 2011; Reuman et al., 2014; Marañón,
2015; Rasconi et al., 2015) (Figure 1B). Warming, however,
can also have indirect effects on phytoplankton community
structure. A strong effect of warming on phytoplankton
community composition may be related to changes in thermal
stratification and vertical mixing (Livingstone, 2003; Winder
and Sommer, 2012; Yankova et al., 2017). For example, mixing
processes determine changes in resource availability: enhanced
or prolonged stratification of the water column due to warming
can suppress the upward flux of nutrients from deep-waters
through vertical mixing, resulting in more nutrient-depleted
surface waters (Yankova et al., 2017; Lepori et al., 2018).
Smaller organisms, which possess advantageous surface to
volume ratio for nutrient uptake, should dominate in nutrient
depleted environments (Marañón, 2015). Increasing temperature
therefore has direct and indirect effects on phytoplankton
size distributions, which should all combine to promote the
dominance of small taxa and decrease (toward more negative
values) the slope of SAR (Figure 1B).
Resource supply has a key role in determining the slope
of SARs. Oligotrophic regions show steeper (more negative)
slopes, while nutrient rich or eutrophic environments present
flatter (less negative) slopes (Gaedke et al., 2004; Barton et al.,
2013; Marañón, 2015; Guiet et al., 2016; Sprules et al., 2016).
In principle, smaller phytoplankton should always outcompete
larger ones, not only under oligotrophic conditions, since they
have relatively high nutrient specific uptake affinity and growth
rates (Edwards et al., 2012). In the presence of grazers, however,
an increase in abundance of small phytoplankton is rapidly
balanced by grazing, leaving excess nutrients available for larger
phytoplankton forms, which gain an advantage through the
lagged growth and grazing of their smaller competitors (Stibor
et al., 2004; Ward et al., 2012; Barton et al., 2013; Marañón, 2015).
The grazers of small phytoplankton are mostly microzooplankton
(flagellates, dinoflagellates, ciliates and rotifers), which have
generation times of the same scale of their prey, while larger
phytoplankton have larger, slower-growing macrozooplankton
predators such as copepods or cladocerans (Hansen et al., 1997;
Sommer et al., 2001; Stibor et al., 2004; Wollrab and Diehl,
2015), whose generation times are orders of magnitude longer.
Large phytoplankton are therefore more likely to outgrow their
grazers under conditions of high nutrient supply, favoring larger
phytoplankton taxa (Cloern, 2018). This implies that adding
nutrients makes the SAR slope more positive (Armstrong, 1994;
Stibor et al., 2004; Ward et al., 2012; Marañón, 2015) (Figure 1C).
Light is also an important factor: light absorption decreases
in larger cells because self-shading by pigment molecules
(the package effect) increases with size, especially under light
limitation, when intracellular pigment concentrations are higher
(Finkel et al., 2010). Small cell size in phytoplankton is therefore
advantageous in resource poor environments, where light and/or
nutrient availability is low (Figure 1C). The ability to rapidly
uptake and store nutrients may favor large cells in fluctuating
environments, though it is not clear how this would uniformly
affect SARs (Verdy et al., 2009; Bonachela et al., 2011).
Consensus on the magnitude and variability of the slope
of SARs, as well as the causes of this variability, remains
elusive. We expect variability in the SAR to occur in natural
systems due to changes in environmental and ecological
conditions, and due to seasonal succession or disturbance,
but the mechanisms may be complex and multivariate. For
example, in a set of experimental aquatic ecosystems, warming
of ∼4◦C initially increased the steepness of the phytoplankton
community size spectra slope by increasing the prevalence of
small organisms (Yvon-Durocher et al., 2011). These results have
Frontiers in Microbiology | www.frontiersin.org
2
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 3
Pomati et al.
Environmental Drivers of Phytoplankton Size
FIGURE 1 | The SAR [local size-density relationship (Reuman et al., 2008) – A], and relative hypotheses of how changes in environmental conditions should influence
its slope in lake phytoplankton communities (B–D). In (D), the black solid line depicts the predicted effect of total grazing pressure, while the gray dashed lines the
effect of zooplankton selectivity (ratio of abundance between copepods and daphnids). Note that hypothesized trends in (B–D) are linear to simplify concepts. While
a linear fit in (A) is a prerequisite for power law scaling, identifying the shape of relationships in (B–D) is a goal of this study.
been attributed to greater competition among phytoplankton
for limiting resources due to temperature-induced increases in
metabolic rates. The effect of temperature was, however, reversed
over the long term in the same experiment when warming
was associated with dominance of large algal species (Yvon-
Durocher et al., 2015). This pattern was interpreted as emerging
from trophic interactions, with warming favoring taxa that
were more resistant to grazing (larger cell size and/or colony
formation), suggesting the importance of interacting bottom–up
and top–down controls on size distributions. Despite the above
evidence that temperature, resource supply and zooplankton
grazing impact size distributions, unequivocal evidence linking
main effects and interactions to changes in slopes from field
observations is lacking. The shape of such relationships are also
largely unknown. This gap may be due, in part, to the difficulty
of disentangling the effects of these co-variating drivers, and to
data limitations.
Here, we use long term phytoplankton community datasets
from eight lakes in Switzerland, sampled monthly for decades,
and spanning a range of environmental and ecosystem attributes
(Table 1). We quantify the temporal variability in phytoplankton
SARs, and test the above hypotheses about how abiotic and
biotic factors control
the relative abundances of different
phytoplankton sizes. We examine variations through time in
SARs and its drivers instead of the more commonly calculated
size spectra, where the abundance of organisms in log-spaced
bins is added together (White et al., 2008). We preferred the
SAR approach because we could retain all individual data points
from all taxa and adopt a robust slope estimation procedure
based on bootstrapping of the species pool. The eight Swiss lakes
provide an ideal setting for this study because of the quantity and
quality (standardization) of paired biological and environmental
observations (Table 1), and the lakes’ well-known history
of eutrophication, re-oligotrophication, and climate change
(Livingstone, 2003; Anneville et al., 2004; Pomati et al., 2012;
Monchamp et al., 2018). We explore the extent of variation in
phytoplankton SARs across ecosystems of contrasting conditions.
This included the investigation of the relative importance and
interactions among in-lake drivers such as nutrients, grazing
and temperature, and seasonal and inter-annual ecosystem
changes. For data analysis, we used a non-parametric machine
to find generalizable
learning approach (random forests),
predictive patterns in notoriously noisy data. Random forests
(RF) allowed us to overcome the most important constraints
of traditional statistical approaches: a priori specification of (i)
functional forms, (ii) interactions, and (iii) error distributions
(Thomas et al., 2018). We expect that the marked natural
and anthropogenic disturbances, particularly in temperature and
phosphorous supply, induced variations in abundances between
different size classes, which would allow us to quantitatively link
changes in phytoplankton SARs to environmental drivers.
MATERIALS AND METHODS
Data
The plankton dataset consists of microscopic counts of samples
collected between 1960 and 2016, mostly in monthly intervals
(occasionally biweekly), across 8 Swiss lakes (Table 1). The full
raw data and metadata are available from Zenodo (doi: 10.5281/
zenodo.3582838). Plankton microscopy data from Baldeggersee,
Greifensee, Hallwilersee, Sempachersee, and Lake Luzern were
collected by Eawag taxonomists, while data from Walensee,
upper Lake Zurich (location Lachen) and lower Lake Zurich
(location Thalwil) were collected by the Zurich Water Supply
Company (WVZ). Plankton samples have consistently been taken
in the same locations (with the exception of Lake Luzern in
which there was a change in sampling location in 1998, from
Kreuztrichter to Obermattbecken) and counted by the same
teams of taxonomists over the years, who have also exchanged
knowledge and attended the same taxonomy courses. For more
details about sampling methods and datasets, see references
reported in Table 1.
Samples for phytoplankton microscopy have been collected
as integrated samples over the epilimnion [with a Schröder
sampler (Mieleitner et al., 2008), where the lower depth varies
across lakes; Table 1] or at discrete depths (e.g., Pomati et al.,
2012), depending on lake and time period. Taxa abundances
were converted to total abundance (cells L−1) across the available
depths in the epilimnion (when discrete depth samples were
lakes (Table 1).
collected) to allow comparisons across all
Taxonomy of all phytoplankton species in the dataset was
harmonized according to modern phytoplankton classification
(for examples see Pomati et al., 2012, 2015). Biovolumes for each
Frontiers in Microbiology | www.frontiersin.org
3
January 2020 | Volume 10 | Article 3155
F
r
o
n
t
i
e
r
s
i
i
n
M
c
r
o
b
o
o
g
y
i
l
|
w
w
w
.
f
r
o
n
t
i
e
r
s
n
.
o
r
g
i
TABLE 1 | Lake and plankton metadata.
Phytoplankton
Chemistry
Zooplankton
Number of matching
dates (n)
Previous studies
i
n
o
i
t
a
v
e
r
b
b
A
e
m
u
o
V
l
)
3
m
k
(
)
m
(
e
t
i
s
g
n
i
l
p
m
a
s
t
a
h
t
p
e
D
)
n
(
s
e
t
a
d
e
v
i
t
c
e
f
f
E
)
m
(
e
g
n
a
r
h
t
p
e
D
n
a
p
s
e
m
T
i
a
x
a
t
f
o
r
e
b
m
u
N
)
i
d
e
n
g
s
s
a
n
u
(
n
a
p
s
-
e
m
T
i
)
n
(
s
e
t
a
d
e
v
i
t
c
e
f
f
E
)
m
(
e
g
n
a
r
h
t
p
e
D
Lake
Walensee
WA
2.5
144
1972–2000 350
0–20
158 (5)
1980–2000 261
0–144
Upper Lake Zurich UZ
0.6
36
1972–2000 349
0–20
174 (6)
1980–2000 253
0–36
)
n
(
s
e
t
a
d
e
v
i
t
c
e
f
f
E
)
m
(
e
g
n
a
r
h
t
p
e
D
n
a
p
s
-
e
m
T
i
l
e
v
e
l
*
*
)
L
/
g
µ
(
4
O
P
-
P
2
5
9
1980–2000 249
0–144
1980–2000 251
0–36
1975–2014 650 0–110 (150)*
Lake Lucerne*
4
Lower Lake Zurich
LU
LZ
11.8
110 (150)* 1968–2014 738 0–20 (30)* 384 (32) 1968–2004 327 0–110 (150)*
3.3
135
1976–2010 420
0–40
271 (14) 1976–2009 408
0–135
33
1977–2008 367
0–135
Sempachersee
SE
0.66
87
1984–2014 388
0–15
284 (1)
1982–1998 201
0–87
76
1984–2014 389
0–87
Hallwilersee
HA 0.285
45
1982–2014 415
0–13
314 (2)
1977–2001 172
0–45
86
1982–2014 414
0–45
124
Baldeggersee
BA 0.173
Greifensee
GR 0.148
67
30
1984–2014 403
0–15
295 (0)
1982–1998 205
0–67
87
1984–2014 388
0–67
1984–2016 429
0–20
401 (20) 1972–2015 519
0–30
96
1975–2015 565
0–30
157
286
Lakes are sorted in ascending levels of P-PO4. *Change in sampling location in 1998; ** median of the analyzed time series.
J
a
n
u
a
r
y
2
0
2
0
|
l
V
o
u
m
e
1
0
|
l
A
r
t
i
c
e
3
1
5
5
243
253
202
365
151
Anneville et al., 2004,
2005
Anneville et al., 2004,
2005
Finger et al., 2013
Anneville et al., 2004,
2005; Pomati et al.,
2012, 2015, 2017
Bürgi and
Stadelmann, 1991;
Monchamp et al.,
2018
Bürgi and
Stadelmann, 1991;
Monchamp et al.,
2018
Bürgi and
Stadelmann, 1991
Mieleitner et al.,
2008; Monchamp
et al., 2018
f
m
i
c
b
-
1
0
-
0
3
1
5
5
J
a
n
u
a
r
y
1
3
,
2
0
2
0
T
i
m
e
:
1
6
:
5
3
#
4
P
o
m
a
t
i
e
t
a
l
.
E
n
v
i
r
o
n
m
e
n
t
a
l
D
r
i
v
e
r
s
o
f
l
P
h
y
t
o
p
a
n
k
t
o
n
S
z
e
i
fmicb-10-03155
January 13, 2020
Time: 16:53
# 5
Pomati et al.
Environmental Drivers of Phytoplankton Size
phytoplankton species were recorded by the taxonomists that
counted the samples. Biovolumes represent the median of tens of
cells measured for each species over the years. This information
has been stored as a meta-database of species biovolumes (H.
R. Buergi, unpublished – see online Supplementary Table S1),
which was merged with information from the Zurich Water
Supply Company database. When species were missing in the
Eawag meta-database of species biovolumes, taxa biovolumes
were obtained by matching species names against the published
database by Kremer et al. (2014) (Supplementary Table S1).
From Kremer et al. (2014), we used median taxa biovolumes,
which were obtained by collecting data across studies from
the literature (Kremer et al., 2014). Biovolume (hereafter size)
was expressed in µm3 for each counted taxon, and reflects
individual cell volumes; for colony forming taxa such as diatoms
and cyanobacteria, the biovolume is for individual cells, not
the size of colonies. In this study, we linked taxa names
(at the species level) with a numeric taxon identifier, which
was then used to match each taxon with a corresponding
size in the meta-database (Supplementary Table S1). In this
way, every taxon in our microscopy data was assigned to a
species-specific cell size, with few exceptions of unassigned
taxa for which we could not find reliable cell biovolume
data (Table 1).
the net
Zooplankton samples were collected as a net tow from the
lake bottom to the surface; over time and across locations
and lakes (with different maximum depths) the depth span
of
sampling varied. Zooplankton densities were
therefore normalized across the database by expressing them as
individuals m−2 of surface area. In this study we considered
only two functional groups of grazers: the unselective filter
feeding daphnids and the selective feeding copepods (with
calanoids being current feeders and cyclopoids ambush feeders).
Consistent information on ciliates and rotifers was not available.
Specifically, we focused on the concentration of individuals
(juveniles included, but no eggs, ovaria or ephippia) of the
following four families: Bosminidae and Daphniidae (daphnids),
Diaptomidae and Cyclopidae (copepods). As potential drivers of
phytoplankton size spectra, we considered the total abundance
the above families in each sample, and the ratio
of all
between selective (Diaptomidae and Cyclopidae) and unselective
(Bosminidae and Daphniidae) grazers. We focused on these
four families as they represent
the dominant zooplankton
in Swiss lakes,
in terms of biomass, provide a strong top–
down constraint upon lake phytoplankton, and represent
grazing pressures on different size groups (all from daphnids
and large from copepods)
(Sommer et al., 1986, 2012;
Gaedke et al., 2004).
Chemical and physical water parameters were measured
monthly (occasionally biweekly) for all lakes in the same locations
in which phytoplankton samples were collected (Table 1).
The datasets included measurements over the water column
in discrete depths, from surface to bottom, with differences
in maximum depth and depth resolution depending on the
lake and sampling location (Table 1). Data from Walensee,
upper Lake Zurich (location Lachen) and lower Lake Zurich
(location Thalwil) were produced by WVZ, while data from
the other lakes were obtained from local Swiss Cantonal
environmental authorities. In this study we focus on the two
main environmental drivers of lake change in the Swiss peri-
alpine region, as previously assessed (Anneville et al., 2004,
2005; Pomati et al., 2012; Monchamp et al., 2018): water
temperature and free available dissolved phosphorus (P-PO4).
As noted previously, other variables such as light, turbulence,
and other nutrients (e.g., nitrogen) theoretically play important
ecological roles, but we focus on temperature and phosphate
because previous studies have shown these variables to be
the most significant drivers of ecological change in these
lakes (Monchamp et al., 2018, 2019). For example, nitrogen
levels have been steady over the past four decades and did
not correlate significantly with the changes phytoplankton
community structure detected in previous studies in the same
lakes (Pomati et al., 2012; Monchamp et al., 2018, 2019).
In situ physical measurement of temperature and laboratory
chemical analyses of P-PO4 were performed using standard
methods and are comparable across lakes (Pomati et al.,
2012; Monchamp et al., 2018). In many cases however, P-PO4
was below detection limits of
in such cases
we substituted the actual detection limit of the method for
the logical character “below detection limit.” For subsequent
statistical analyses we used the mean of temperature and mean
of P-PO4 over the water column (i.e., the average based on
available depths). Variability of temperature over depths was
used as an indicator of water column stability (i.e., high
variability over depth = strong stratification and therefore
stability), and estimated it as the coefficient of variation (standard
deviation divided by the mean value) over the sampled depths
(Pomati et al., 2012).
the method:
Data Analyses
The overarching goals of the data analysis were to: (a) calculate
the SAR for phytoplankton at each time in each lake and (b)
characterize the effects of key environmental drivers on the
slope of SARs across the lake database. All statistical analysis,
including RFs (see below), were performed in the R programing
environment (R-Development-Core-Team, 2018).
Calculation of Size-Abundance
Relationships
To analyze the phytoplankton SAR, we fit a linear model to
the raw data of taxa abundances relative to their size (both
variables in Log10, see Figure 1A) per each sampling date, in
each lake (no binning, histogram or distribution model was used).
Following the advice of White et al. (2008) and Duncanson
et al. (2015) we refrained from binning when estimating SAR
exponents, given that our phytoplankton size data are continuous
and binning introduces biases and arbitrary decisions (e.g.,
number and width of bins) in the estimation of the scaling
exponent. Our approach to study the SAR was based on
calculating local size-density relationships, which plot species
concentrations in each water sample relative to the mean species
biovolume (Reuman et al., 2008) (in this way we retained the
information from all the taxa in the database), and on fitting
Frontiers in Microbiology | www.frontiersin.org
5
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 6
Pomati et al.
Environmental Drivers of Phytoplankton Size
a linear regression to the data (Figure 1A). The model took
this form:
R-packages: randomForest (version 4.6-14), randomForestSRC
(version 2.7.0), and plotmo (for interactions plots).
Log10(density) = a + b · Log10(mean taxon biovolume)
where densities were expressed as cells L−1, and mean taxon
biovolume as µm3. We extracted from the generalized least
square linear fit the coefficient b, hereafter “slope” (Figure 1A).
We used this metric to examine how SARs vary across lakes
and over time. To reduce uncertainties in estimating the scaling
exponent, rather than using a maximum likelihood estimator,
we opted for a bootstrapping of the species pool. This allows
to account for potential biases in the estimation of b due to
(i) the linear assumption of the model and (ii) taxonomic
inconsistencies in the classification and counting of species in
the dataset, which spans across many lakes and decades (Straile
et al., 2013; Pomati et al., 2015). The linear fit held significant
for all lakes (Supplementary Figure S1) and all dates (data not
shown). To account for the potential effect of taxonomic biases,
we calculated the slope for each date and lake 999 times, by
resampling at each round of analysis only 70% of taxa present
in the species pool (jackknife bootstrapping). This allowed us
to calculate a median and 95% confidence intervals (CIs) for
each estimated slope. To confirm and interpret patterns observed
when studying changes in the slope of SARs across lakes and over
time, we also divided the proportion of biovolumes for all the
taxa in our database into three quartiles (Supplementary Figure
S2) and investigated patterns in the total abundance of species
composing the first (Q1, the smallest 25% of taxa) and third (Q4,
the largest 25% of taxa) quartiles.
Modeling of Size-Abundance
Relationships Based on Environmental
Drivers
We used RF, a non-parametric machine learning approach, to
test for the relative importance and direction of the effects
of hypothesized drivers of SARs (Figures 1B–D). RF are
a robust machine learning tool based on an ensemble of
regression (or decision) trees featuring bootstrap sampling,
random variable selection, and model averaging (Breiman, 2001).
When presented with complex environmental datasets, RF avoid
constraints inherent in traditional statistical approaches, namely
the a priori specification of functional forms, interactions, and
error distributions. In each regression tree within the “RF,” a
randomly selected subset of the data is recursively partitioned
based on the most strongly associated predictor. At each node,
a random subset of the total number of predictors is considered
for partitioning. This bootstrapping of both data and explanatory
variables minimizes problems associated with the presence of
data outliers or artifacts, and with variable collinearity. The
final tree prediction is given by the average value of the data
within each branch of the tree. By aggregating predictions across
trees in the forest, RF are able to reproduce arbitrarily complex
shapes and patterns without a priori functional form specification
(Breiman, 2001; Thomas et al., 2018). In our study, each forest
comprised 999 trees. For RF analyses, we used the following
We implemented a RF model for the prediction of estimated
median slopes after taxa resampling (see the section “Size-
abundance relationship analysis”). Modeling of observed slopes
instead of estimated medians, however, did not change the results
(see Supplementary Figures S3, S4). To explain variation in
the slope of size SARs across lakes and over time, we used the
following environmental drivers (see also the section “Data”):
- Temperature: average water column temperature (variable
name Tmean) and its coefficient of variation over the sampled
depths as a measure of stability (TCV ) (Pomati et al., 2012).
- Nutrients: mean P-PO4 levels over the sampled depths (P-
PO4mean), and total phytoplankton densities as measures of
total available resources (Phytoplanktontotal). While P-PO4 is
the limiting factor for phytoplankton growth in our panel of
lakes (Anneville et al., 2004; Pomati et al., 2012; Monchamp
et al., 2018), total phytoplankton abundances account for total
nutrients (phosphorus and nitrogen) available in the systems.
Additionally, high levels of phytoplankton densities anti co-
vary with light penetration in the water column, and causing
light limitation.
- Grazing:
of daphnids
total densities
and copepods
(Zooplanktontotal) as a measure of total grazing pressure,
and the ratio between abundances of copepods and daphnids
(Zooplanktonselectivity) as a proxy for the prevalence of
size-selective versus non-size-selective grazers, respectively
(Sommer et al., 2001; Gaedke et al., 2004; Stibor et al., 2004;
Wollrab and Diehl, 2015).
All environmental variables in the model were used without
any transformation, with the exception of phytoplankton and
zooplankton densities that were Log10 transformed. Missing
values (12 in total) were imputed automatically by the rfsrc
function (package randomForestSRC) or using the function
rfImpute (package randomForest): NAs are initially replaced by
data column medians, then a proximity matrix from a RF model
is used to update the imputation of NAs where the imputed
values is the weighted average of the non-missing observations.
To account for important differences in the morphometry of
lakes (Table 1), we included depth (Lake Depth) at the sampling
site and lake total water volume (Lake Volume) in the model.
This allowed us to separate the effects of in-lake environmental
conditions from lake characteristics in predicting phytoplankton
size spectra slopes. Additionally, to compare the magnitude of
effects of in-lake environmental drivers relative to the strength
of lake long-term temporal trends and seasonal changes, we
included as explanatory variables (i) the time sequence of
dates for every lake as a proxy for unaccounted time-varying
factors (Time-trend) and (ii) the sequence of months in the
year (Seasonality).
The RF model
for the median slope is a function of
variables: Tmean, TCV , P-PO4mean, Phytoplanktontotal,
all
Zooplanktontotal, Zooplanktonselectivity, Lake Depth, Lake Volume,
Time-trend, and Seasonality. This model explained 52% of the
variance in slopes. Including the response variable (slope) as
Frontiers in Microbiology | www.frontiersin.org
6
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 7
Pomati et al.
Environmental Drivers of Phytoplankton Size
an autoregressive term in the model only slightly increased the
variance explained (from 51.8 to 52.4%), without significantly
changing the effects of explanatory variables (Supplementary
Figures S5, S6), likely due to the inclusion of the temporal trend
as a predictor in the model.
The importance of each explanatory variable (e.g., TCV ) in
the RF model was assessed by permuting it across all generated
trees (the forest), and by quantifying the resulting change in
the model’s error rate. More important drivers lead to a greater
increase in error when omitted from the model (Breiman,
2001). The partial effect of any explanatory variable on the
response (slope) can be quantified by averaging, across the
forest, the variable values used in the trees to reach terminal
nodes. This property of RF allowed us to examine the functional
form of the relationship between environmental drivers and
slope values, which may be non-linear. We did not include
interaction (multiplicative) terms in the RF model: in linear
models, interaction terms might bring value by fixing non-
linearity or independence violations between the response and
the explanatory variables. RF do not have assumptions about
linearity and interactions emerge from predicting the response
variable over varying levels of a chosen pair of explanatory
variables (Breiman, 2001; Thomas et al., 2018).
RESULTS
Environmental Changes
Over the past five decades, the mean temperature of the water
lakes by an average of 0.85◦C
column has increased in all
(standard error = 0.13), and the mean dissolved phosphorus (P-
PO4) concentrations have decreased by an average of 99 µg/L−1
(standard error = 38) across all lakes, though the magnitude
and pace of change differed clearly by lake (Figure 2). The
unusual pattern in Lake Lucerne at the end of the time series is
likely due to change in sampling frequency, which has become
sporadic and irregular starting from the 1990s (due to complete
recovery of the lake from eutrophication, the sampling location
was changed and frequency reduced to 2–4 times per year).
Along with warming of surface waters, most lakes have shown an
increase in stability of the water column (coefficient of variation –
CV – of temperature over depths), which is consistent with an
increase in thermal stratification (Supplementary Figure S7).
P-PO4 levels differ among the lakes (Figure 2), ranging from a
high of almost 500 µg L−1 in Greifensee to below detection limits
(1 µg L−1) in Walensee, Upper Lake Zurich and Lower Lake
Zurich (Figure 2).
As a consequence of managing P-PO4 discharges and
subsequent recovery of lakes from eutrophication, phytoplankton
median population densities and total community densities
have decreased in all the studied lakes, with few exceptions
(namely total
in Lower Lake Zurich
and Baldeggersee, Supplementary Figure S8). Densities
of zooplankton (daphnids and copepods), and the ratio
between selective (copepods) and non-selective (daphnids)
grazers, did not show any strong or general pattern, with
lakes showing no change over time (Supplementary
most
abundances
algal
exceptions
in which total
Figure S9). Notable
and
are Lake Lucerne
Hallwilersee,
zooplankton densities have
decreased, and WA in which the ratio between copepods
and daphnids has increased over time (Supplementary Figure
S9). The effects of environmental changes on the median
size of taxa in phytoplankton communities were small and
inconsistent across lakes (Supplementary Figure S10).
In
the most oligotrophic lakes (Walensee, Upper Zurich and
Lucerne), we observed a slight decrease in median taxa size
over time, while the most productive lakes (Sempachersee,
Hallwilersee, Baldeggersee and Greifensee) showed a small
temporal increase in median size throughout the community
(Supplementary Figure S10).
Dynamics of SAR Slopes
SAR slopes varied across lakes and over time, as depicted in
Figure 3, showing observed and bootstrapped exponents of
SARs at each lake-date. The oligotrophic lakes (Walensee, Upper
Zurich, Lucerne) had a slightly less negative slope than the most
productive lakes (Hallwilersee, Baldeggersee, and Greifensee),
but the difference is small (and Baldeggersee and Lake Lucerne
have similar slopes). Note the ample variability of slopes within
lakes and within years, signaling potential fluctuations in the
mechanisms regulating phytoplankton SARs at the seasonal
scales. Additionally,
for five lakes out of eight (Walensee,
Upper Zurich, Lucerne, Lower Zurich, Greifensee), there was
a clear increasing long-term temporal trend, with a tendency
toward flattening of the slope (i.e., less negative) in the most
recent years (Figure 3). For lakes Sempachersee, Hallwilersee
and Baldeggersee, the pattern of slopes shows relatively large
fluctuations but no clear long-term trend.
Changes in slopes across lakes and over time corresponded
to variation in the absolute and relative abundances of small
and large phytoplankton taxa. We focused on the first (Q1)
and fourth (Q4) quartiles, respectively, of the distribution of
species biovolumes (Supplementary Figure S2). Lakes with
steeper slopes (e.g., list lakes here) were characterized by slightly
higher density of small taxa (Q1) compared to lakes with flatter
slopes (Figure 4). Similarly to patterns in SAR slopes, the
abundances of both large and small taxa groups was dynamic
within years and over the long term, with differences between
lakes. Overall, the average abundance of large taxa seemed to
be more stable over time compared to the density of small taxa,
which decreased slightly in time for all lakes with the exception
of Hallwilersee and Baldeggersee (Figure 4, solid thick lines).
An increase in the relative abundances of large versus small taxa
(Q4/Q1) was observed in lakes Walensee, Upper Zurich, Lucerne,
Lower Zurich and Greifensee (Figure 4), which helps explain the
flattening of the slope size spectra in these lakes (Figure 3).
The composition of small and large groups of phytoplankton
in our studied lakes is shown in Figures 5A,B. The small
taxa group was dominated by cyanobacteria, green algae, and
chrysophytes, while the most prevalent phytoplankton classes
in the large taxa group were dinoflagellates, Conjugatophyceae
the desmids),
(which includes
and cryptophytes. Note
that
taxa in our database corresponds to cell
the size of
biovolumes, since information in our database about the sizes
Frontiers in Microbiology | www.frontiersin.org
7
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 8
Pomati et al.
Environmental Drivers of Phytoplankton Size
FIGURE 2 | Time series of mean water column temperature (black lines, gray trend;◦C) and dissolved inorganic phosphorus (P-PO4, dark blue lines, light blue trend;
µg L−1), across our panel of lakes. Trend-lines were obtained by locally weighted scatterplot smoothing. Codes in panels represent the name of lakes as in Table 1.
colonies was not consistent among all lakes and time points
(Supplementary Table S1).
Effects of Environmental Drivers on
Slopes
To tease apart the relative effects of environmental drivers on
SARs we modeled slope values across lakes and over time using
a RF approach (see Section Materials and Methods). The most
important explanatory variables predicting the slope of SARs
lake volume,
across lakes and over time are those that, when omitted in the
RF model, more strongly reduce the performance of the model.
These were, respectively:
total phytoplankton
densities, time trend, and lake depth, followed by P-PO4, water
temperature, month of the year, total zooplankton density,
water column stability (CV of temperature over depths) and
zooplankton selectivity (ratio between abundance of copepods
and daphnids) (Figure 5C). Based on the analysis of partial effects
from the RF model, the time-invariant factors “lake volume”
and “depth at sampling site” had a similar consequence on
Frontiers in Microbiology | www.frontiersin.org
8
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 9
Pomati et al.
Environmental Drivers of Phytoplankton Size
FIGURE 3 | Time series of size spectra slopes across lakes (codes in panels represent lakes as in Table 1). Red dots = observed slopes in the monthly samples;
black line = median of bootstrapped slopes (see section Materials and Methods for details); gray lines = 95% confidence intervals of bootstrapping; blue
line = median slope of the whole time series.
slope: larger and deeper lakes had steeper (more negative) slopes
of the SARs (Supplementary Figure S11), with the exception
of Greifensee (which is the smallest lake but showed steep
slopes). RF-based partial effects of time-varying environmental
variables on slope of SAR exposed the importance of non-
linear dependencies and inconsistencies between theoretical
predictions (Figures 1B–D) and patterns in the data. Time trend,
included in the RF model to allow extracting the effects of all
unaccounted time-varying factors across lakes, showed a steady
increase of slope from values ranging −0.75 toward less negative
values during the 1970s and 1980s, with a peak of −0.68 in the
early 1990s (Figure 6A). The slope then decreased again in the
2000s and remained in the range of value of −0.70 at present.
Extracting seasonal succession from the data using the RF model
revealed steeper slopes in winter and spring, and flatter (less
negative) slopes in summer and autumn (Figure 6B).
The partial effects of water column thermal structure, nutrient
supply and grazing on the slopes of SARs are depicted in
Figures 6C–H. Patterns extracted from the RF analysis of
partial effects revealed evident non-linear responses of slope to
Frontiers in Microbiology | www.frontiersin.org
9
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 10
Pomati et al.
Environmental Drivers of Phytoplankton Size
FIGURE 4 | Time series of small and large phytoplankton densities, specifically of the first (Q1, smaller cells, blue) and fourth (Q4, larger cells, red) quartiles of the
distribution of species biovolumes, across lakes (codes represent lakes as in Table 1). The green lines represent the ratio Q4/Q1. Trend-lines were obtained by
locally weighted scatterplot smoothing.
these general environmental drivers. Increasing water column
temperature (average over depths) and stability (CV over depths)
showed positive and weak effects on slope up to values of 12◦C
and 0.2, respectively, after which the response curve appeared
to saturate (Figures 6C,D). Total phytoplankton cells density
(a measure of total productivity of the system) and dissolved
inorganic phosphorus (the main growth limiting factor) showed
a negative effect on slope, saturating on the low end at 8
Log10(counts L−1) and 100 µg L−1, respectively (Figures 6E,F).
The effects of nutrients on changes in SAR slope were stronger
compared to those of temperature (see the scale of the Y-axes in
Figures 6C,D compared to Figures 6E,F). A key finding of the
RF analysis is that the patterns in Figures 6C–F are the opposite
of what expected from theory and depicted in Figures 1B,C as
hypotheses. Total zooplankton grazing had the expected positive
effect on slope, starting from densities around 5 and saturating at
6 Log10(counts m−2) (Figure 6G). The effect of the ratio between
selective and non-selective grazers on slope was very weak and
potentially positive in its direction (Figure 6H).
The non-linear patterns in Figure 6, emerged from the RF
analysis, suggest possible multiplicative effects (e.g., interactions)
and threshold responses of SAR slope to in-lake environmental
Frontiers in Microbiology | www.frontiersin.org
10
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 11
Pomati et al.
Environmental Drivers of Phytoplankton Size
FIGURE 5 | (A,B) Relative abundance of phytoplankton classes, expressed as a percentage, in the first (Smaller cells, A) and fourth (Larger cells, B) quartiles of the
distribution of taxa biovolumes. Note that the color palette is consistent across the two charts, and not all groups are present in both size classes (e.g.,
Cyanophyceae). (C) Random forests ranking of predictors of SAR slopes over time and across lakes: the importance reflects the change in mean absolute error of
the model when the variable of interest is permuted (the color gradient has no specific meaning, it is only for display).
drivers. The RF model predicted effects of temperature, PO4
and total grazing on slope, for example, changed direction at
defined levels (Figures 6C,F,G). These potential interactions
are illustrated by color-coded contour plots generated by the
RF model, showing the jointed predicted effects of the main
ecological drivers on slope (Figure 7). P-PO4 levels and total
zooplankton densities show evident thresholds that influence
both their direct effects and the effects of co-varying drivers
(Figures 7A–C). The negative effects of increasing temperature
on slope were stronger under nutrient
limitation and low
zooplankton densities (see deep blue shades in Figures 7A,C),
and the positive effects of nutrients and of grazing were stronger
under high temperature (see bright red shades in Figures 7A,C).
Specifically, high temperature and high total zooplankton grazing
synergized with low P-PO4 to predict the least negative (flattest)
values of slope (bright red shades in Figure 7). The steepest slopes
(deep blue shades) are instead predicted for low temperature,
low zooplankton levels, and P-PO4 values between 110 and
200 µg L−1 (Figure 7). Note the abrupt change in predicted
response (slope) crossing the value of 100 µg L−1 of P-PO4
(Figures 6F, 7A,B), and 5.2 Log10 (counts m−2) of
total
zooplankton density (Figures 6G, 7B,C). The interactive effects
of water temperature and total zooplankton grazing showed low
slope values (bright red) for temperature comprised between 12
and 14◦C and zooplankton densities between 5.5 and 6, and high
slope values (deep blue) at 4◦C and low zooplankton densities
(Figure 7C). Note that low temperature and high phosphorus
are always associated with steep slopes (deep blue color), while
high temperature and low phosphorus correspond to flat slopes
(bright red color, Figure 7).
DISCUSSION
Random forests analysis allowed us to model and explain the
observed variation in the slope of size SARs across eight lakes
and over decades of time (Figure 3), based on abiotic and biotic
environmental drivers (Figure 2 and Supplementary Figures
S7–S9). As mentioned in Section “Materials and Methods,” this
machine learning approach is indifferent to outliers, not biased
by a priori specification of response functions (e.g., linearity),
and allows to extract robust patterns from noisy and high-
dimensional datasets (Thomas et al., 2018). The most striking
pattern emerging from our data analysis was the high prevalence
of small phytoplankton taxa in more nutrient rich environments,
signaled by steeper slopes of SARs under high nutrient levels. This
pattern is the opposite of our theoretical expectation (Figure 1C)
and is in contrast with what has been observed previously in
nutrient rich freshwater and marine environments (Cavender-
Bares et al., 2001; Gaedke et al., 2004; Marañón, 2015; Guiet et al.,
2016; Sprules et al., 2016). It is, however, the predominant pattern
in the data, consistent across lakes and over time (Figures 2–
4), and emerged unequivocally from the RF analysis of partial
effects of environmental drivers (Figures 6, 7). While deeper and
larger ecosystems tend to be characterized by a more oligotrophic
environment and higher dominance of small phytoplankton
taxa, as expected (Marañón, 2015) (Table 1 and Supplementary
Figure S11), eutrophic lakes in our dataset have steeper slopes
(Figure 3). The pattern in our data is mostly driven by changes in
abundance of small taxa, which decrease over time (Figure 4).
Over the temporal span of this study, lakes have undergone
a process of re-oligotrophication (Figure 2) (Anneville et al.,
2004; Monchamp et al., 2018). Concomitantly, we detected a
decrease in the slope of phytoplankton SARs toward less negative
values (i.e., a reduction of small phytoplankton forms over time)
(Figures 3, 4). This long-term trend in re-oligotrophication is
likely the strongest component of the effect of nutrient changes
on SARs: the decadal trend in nutrient levels covers a much
larger range than the seasonal fluctuations (Figures 2, 5). This
pattern of temporal decrease in slope values emerged in the RF
analysis as partial effect of the time trend, showing a minimum
of the SAR slope in the early 1990s, which is when most lakes
stabilized their decreasing trend in phosphorus levels (Figure 2).
This happened alongside with warming, causing stronger and
more stable stratification, which reinforced the oligotrophication
process in the upper water column where phytoplankton thrive
(Anneville et al., 2004; Pomati et al., 2012; Posch et al., 2012;
Yankova et al., 2017; Lepori et al., 2018).
The most representative taxonomic classes belonging to the
small phytoplankton group in our dataset are the cyanobacteria,
followed by the green algae (Figure 5A). Both of these classes
have unicellular and colonial forms, with cyanobacteria being
predominantly colonial while green algae are largely unicellular
(Reynolds, 2006). They all appear in our database as small
phytoplankton because our compiled information includes only
cell biovolumes: colony size was not available for all lakes and
Frontiers in Microbiology | www.frontiersin.org
11
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 12
Pomati et al.
Environmental Drivers of Phytoplankton Size
FIGURE 6 | Partial effects of time-varying environmental predictors of the slope of SARs across lakes, based on the RF model (Figure 5C). Red dots represent
partial values (the black dashed line follows these partial effects), and dashed red lines indicate a smoothed interval of ± two standard errors. Wile panels (A,B)
depict the partial effects of time trend and seasonal progression, comparing the direction of effects in panels (C–H) of this figure to the hypotheses in Figures 1B–D
exposes the importance of non-linear dependencies and inconsistencies between theoretical predictions and responses to environmental drivers in the data.
Frontiers in Microbiology | www.frontiersin.org
12
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 13
Pomati et al.
Environmental Drivers of Phytoplankton Size
FIGURE 7 | Interacting effects of environmental drivers on the slope of SARs. Color-coded contour plots in (A–C) depict the RF model inferred interactions, which
emerge from predicting slope (Z-axis) over varying levels of the chosen pair of explanatory variables (while holding others at their medians): temperature and
dissolved phosphorus (A), total zooplankton densities and dissolved phosphorus (B), and temperature and total zooplankton densities (C). (D) Conceptualized
interaction of temperature, resource availability and grazing effects on phytoplankton community composition and slope of SARs.
all dates (see Section Materials and Methods). We acknowledge
that the use of cell biovolume as a proxy for size, with no
consistent information about colony dimensions, might have
biased our results. Particularly, cyanobacterial diversity and
abundance have dramatically changed over the studied period
across the chosen lakes, due to interacting oligotrophication
and climate change. Temporal trends in taxonomic alpha and
beta diversity are consistent at the regional scale and have
favored in increase in richness and prevalence of colony forming
cyanobacteria (Monchamp et al., 2018, 2019). It is plausible
to hypothesize that changes in diversity and abundance of the
Cyanobacteria might have biased the data analysis toward an
increasing importance of small sized taxa, due to the strong
dynamics of
this (primarily colonial) phytoplankton group
over the past decades. We therefore tested for this bias by
excluding the entire class Cyanobacteria from the data. We then
estimated slopes and confidence intervals for each lake and
date by resampling the species pool without cyanobacteria, and
modeled the median of slope distributions using the same RF
approach reported in Methods and Results for the full dataset.
The RF model of cyanobacteria-free slopes showed slightly
different relative importance of explanatory variables, however
the directions of partial effects for environmental drivers matched
very closely those reported in Figure 6 and Supplementary
Figure S12. The trends we document therefore do not result
from changes in cyanobacteria only, advocating against strong
biases in the analyses due to a lack of information about
phytoplankton colony size.
The above test suggests that
the pattern of decreasing
abundance of small phytoplankton (flattening of the slope of
SARs) over declining nutrient levels (Figures 2–4, 6, 7) is robust.
Together with cyanobacteria, eukaryotic algae have declined
under oligotrophication, reinforced by climate warming, as
previously noted (Yankova et al., 2017; Lepori et al., 2018).
Decreasing nutrient levels appeared to penalize smaller taxa,
which are mostly phototrophic, more strongly than larger forms,
which in our data are predominantly mixotrophic (Figures 5, 7D)
(Reynolds, 2006). This is consistent with previous empirical and
Frontiers in Microbiology | www.frontiersin.org
13
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 14
Pomati et al.
Environmental Drivers of Phytoplankton Size
theoretical evidence suggesting that large mixotrophic species can
survive and thrive in nutrient depleted conditions by engaging in
heterotrophy and phagotrophy (Andersen et al., 2015; Ward and
Follows, 2016), while small phototrophs have higher growth rates
when carbon and inorganic nutrients are abundant (Edwards
the relative
et al., 2012). It has been recently noted that
importance of mixotrophic algae in lakes increases as nutrients
decrease (Waibel et al., 2019). Being large and mixotrophic
appeared in our study to be an advantageous strategy in lakes
undergoing oligotrophication and climate warming (Yankova
et al., 2017; Lepori et al., 2018).
are
In addition to nutrient uptake rates and resource uptake
strategies, phytoplankton SARs
also influenced by
susceptibility to general and selective zooplankton grazing,
which might co-vary with environmental conditions (Sommer
et al., 2001; Stibor et al., 2004; Barton et al., 2013; Marañón, 2015).
The impact of total zooplankton on the slope of phytoplankton
SARs matched the general expectations emerging from the
literature (compare Figure 1D with Figures 6G,H). The ratio
between selective and non-selective grazers (copepods/daphnids)
had a negligible effect on size distributions, likely due to a coarse
grouping of juvenile and adult forms of calanoids (current
feeders) and cyclopoids (ambush feeders), which might have
very different size-specific effects on phytoplankton. This could
have biased the RF analysis by adding noise to this variable,
and therefore reducing its predictive power. Our proxy for
zooplankton size-selectivity, the ratio between copepods and
daphnids, is also affected by the lack of data on a very important
group of size-selective grazers: ciliates and rotifers. Albeit not
being dominant in lakes in terms of biomass, they are very
significant drivers of changes in phytoplankton community
structure (Stibor et al., 2004; Sommer et al., 2012; Wollrab and
Diehl, 2015). On the other hand, total zooplankton abundance
had a clear positive effect on slope (Figure 6G). This result
was consistent with previous evidence of crustacean abundance
having a positive consequence on the slope of phytoplankton
size spectra in Lake Müggelsee (Gaedke et al., 2004). Our data
highlight a previously unnoticed non-linear (saturating) shape
of this effect. According to our hypotheses (outlined in the
Introduction), grazing pressure should also interact with the
effects of resource availability on the slope of phytoplankton
SARs, and the outcomes of our data analysis confirmed this
prediction – however, with the opposite direction. Specifically,
a combination of high grazing pressure and low (instead of
levels robustly favored large phytoplankton
high) nutrient
(Figures 7B,D). A positive interaction between zooplankton
grazing and warming was also detected in the data (Figure 7C),
indicating the prevalence of large phytoplankton under high
grazing pressure and high temperature conditions (Figure 7D).
This pattern is
that consumption
supported by findings
by herbivores (grazing rates) increases more strongly with
temperature than primary production (Rose and Caron,
2007), strengthening the top–down control from grazers on
phytoplankton abundance and community structure under
warming conditions (Winder and Sommer, 2012; Cloern, 2018).
The above consideration brings us the second most striking
pattern in our data, which is the positive effect of water
large
temperature on the slope of phytoplankton SARs:
phytoplankton taxa are more prevalent in warmer environments.
Given the monthly frequency of our sampled community data,
we note that the effects of temperature are necessarily linked to
changes at the monthly, seasonal, and inter-annual scale. The
pattern was in fact evident from partial effects of temperature and
stability (Figures 6C,D) that resembled the seasonal progression
(from winter to summer – Figure 6B) and the temporal trend
(climate warming – Figure 6A): they all drove slopes toward less
negative values (Figures 6A–D). While the effect of stability of the
water column was weak (Figures 5C, 6D), water temperature had
a clearly positive relationship with slope (Figure 6C). The direct
effect of water warming on plankton community SAR, predicted
to be negative (Figure 1B) and mediated by increase in metabolic
rates, has been previously detected under laboratory controlled
conditions, and after short-term warming of experimental
mesocosms (Atkinson et al., 2003; Yvon-Durocher et al., 2011).
It has been also noted, however, that the direct effects of
temperature are small and may be hard to estimate in natural
phytoplankton communities (Marañón et al., 2012; Mousing
et al., 2014) and, when detectable, might be minor compared to
the co-varying or interacting effects of seasonality and nutrient
levels (Marañón et al., 2015, 2018). Surveys (Mousing et al.,
2014), theoretical modeling (Sentis et al., 2017), and long-
term experimental studies (Yvon-Durocher et al., 2015), suggest
that the strongest effect of warming in aquatic communities is
mediated by indirect effects of temperature through changes in
grazing rates (as noted above), and resource availability (due
to suppressed vertical mixing) (Winder and Sommer, 2012).
The former might actually have the strongest effect of favoring
large phytoplankton due to increasing metabolic rates of grazers
and heavier grazing pressure under warming conditions (Yvon-
Durocher et al., 2015; Cloern, 2018). Our data analysis supports
this previous evidence and suggests a conceptual model of
the detected patterns in which phytoplankton size distributions
respond to interacting temperature, resource availability, and
grazing pressure by favoring small phototrophic algae under high
levels of nutrients and low temperature and grazing, and large
mixotrophs in oligotrophic conditions when temperature and
grazing are high (Figure 7D). This outlined concept matches
the predictions of the PEG model of phytoplankton seasonal
succession for spring and summer phytoplankton communities,
respectively (Sommer et al., 1986). Slopes of phytoplankton SARs
and community composition toward the end of lake time series,
in fact, resemble summer assemblages, supporting previous
reports of a temporal progression of lake ecosystems toward
a “summer-like” environment and phytoplankton community
structure (Anneville et al., 2004; Posch et al., 2012; Pomati et al.,
2017; Yankova et al., 2017; Monchamp et al., 2018).
CONCLUSION
In our analysis, each lake had a different baseline biomass
distribution among phytoplankton size classes, likely because
of different
food-web architectures. Our data indicate that
co-occurring seasonal and long-term environmental changes
Frontiers in Microbiology | www.frontiersin.org
14
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 15
Pomati et al.
Environmental Drivers of Phytoplankton Size
significantly control these structures. We highlight a three way
interaction between effects of warming, nutrient supply, and
grazing that might depend on seasonality and on the long-
term history of the analyzed ecosystems,
in this case lakes
experiencing climate warming and oligotrophication. Regardless
of the fact that cyanobacteria have increased in prevalence
within and between lakes, and occurrences of cyanobacterial
blooms have been increasingly reported, our data analysis
suggests that they are not the only group contributing to the
observed long-term changes in the phytoplankton community
SARs. While cyanobacterial fluctuations contributed a significant
proportion of the variation in the abundance of small sized
phototrophs over time, the increase in importance of large
mixotrophic species in recent monitoring data requires further
investigations. Some recent reports corroborate our findings
(Waibel et al., 2019), however more evidence is required to
confirm a generalized increase in prevalence of mixotrophs
relative to smaller phototrophs along oligotrophication and
warming gradients. Our results suggest changes in plankton
trophic interactions over the course of the past half century,
with potentially fundamental consequences for the functioning
of lake food-webs.
The main results of our analyses contrast with the starting
hypotheses based on previous reports, however we are not
the first authors to report inconsistencies between theoretical
expectations of environmental effects on phytoplankton size
distributions and observed patterns (Cermeño et al., 2006;
Marañón, 2015; Marañón et al., 2018). Our observations are
well supported by basic lake plankton ecology, and we speculate
that the inconsistencies between expected and detected effects
of environmental drivers are due to four main reasons. First,
the sampling frequency of our dataset (monthly) restricts the
detection of effects to seasonal and inter-annual scales, while
the direct effects of temperature on metabolic rates and the
effects of nutrient supply might have the strongest influence on
plankton dynamics and the daily and weekly scales (Thomas
et al., 2018). Second, previous studies did not specifically attempt
to address non-linearities and interactions in co-occurring
ecological mechanisms, and this might have confounded the
estimation of importance and direction of environmental effects.
Third, the data used in this study describe lakes that were
not at stationary state: strong effects of time-varying factors
like climate warming and re-oligotrophication had profound
but potentially transient effects on these ecosystems. The
patterns that we detected, therefore, might not be generalizable
to stationary state ecosystems. Fourth, since the majority of
previous studies come from the marine literature, our results
might suggest that there are fundamental differences in how
freshwater and marine phytoplankton communities respond
to bottom–up and top–down controls. Specifically, we note
that grazing by small herbivores such as ciliates and rotifers,
which control small phytoplankton under high nutrient supply
and were not counted in our datasets, might be weaker
in freshwater compared to marine planktonic environments.
This is currently an untested hypothesis and could explain
the dominance of
small algae under eutrophic or high
resource conditions.
DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article are made
available by the authors, without undue reservation, through
Zenodo (doi: 10.5281/zenodo.3582838).
AUTHOR CONTRIBUTIONS
CT and FP prepared the datasets. FP designed the study and
carried out the data analyses with feedbacks from AB, JS, and
KA. FP drafted the manuscript. All authors contributed to the
manuscript development and revisions, and approved the final
manuscript for publication.
FUNDING
This work was funded by the Swiss National Science Foundation
visiting grant IZK0Z3_173883 to FP. AB was funded by the
Simons Foundation.
ACKNOWLEDGMENTS
We thank O. Köster and M. Koss (Wasserversorgung Zürich) for
providing access and valuable insights to the Lake Zurich and
Walen data; B. Müller (Eawag) and the Office of Waste, Water,
Energy and Air (AWEL) of Canton Zürich, Abteilung für Umwelt
Kanton Aargau (A. Stöckli), Eawag/Kanton Luzern, and the Swiss
Federal Office for the Environment for providing chemistry data
for lakes LU, SE, HA, BA, and GR; the lab groups of H. R. Buergi
and P. Spaak for Eawag plankton data collection; M. Baggio
(University of Connecticut) for modeling advice; D. Bouffard for
discussions about water physics drivers; and M. K. Thomas for
advice on RF analysis.
SUPPLEMENTARY MATERIAL
for this article can be found
at: https://www.frontiersin.org/articles/10.3389/fmicb.
The Supplementary Material
online
2019.03155/full#supplementary-material
FIGURE S1 | Scaling of phytoplankton abundances (Log10 cells L−1) with size
(Log10 taxa biovolumes) in each lake dataset.
FIGURE S2 | Distribution of phytoplankton biovolumes across the whole dataset
(all lakes and all dates): blue lines depict the division of the distribution applied to
obtain Q1 (leftmost data, first quartile) and Q4 (rightmost data, fourth quartile)
used in Figure 4.
FIGURE S3 | Ranking of predictors for the RF model of observed slopes (as
opposed to the bootstrapped slopes as in Figure 5C).
FIGURE S4 | Partial effects of environmental predictors based on the RF model of
observed slopes (as opposed to the bootstrapped slopes as in Figure 6).
FIGURE S5 | Ranking of predictors for the RF model of bootstrapped slopes,
adding slope at time-lag 1 (previous month data) as a predictor (lag_slope).
FIGURE S6 | Partial effects of environmental predictors based on the RF model of
bootstrapped slopes, adding slope at time-lag 1 (previous month data) as a
predictor (lag_slope).
Frontiers in Microbiology | www.frontiersin.org
15
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 16
Pomati et al.
Environmental Drivers of Phytoplankton Size
FIGURE S7 | Time series of mean water column temperature (black lines, gray
trend) and temperature coefficient of variation (CV, blue line, light blue trend) over
the water column (i.e., meant temperature/standard deviation). Codes in panels
represent the name of lakes as in Table 1.
FIGURE S8 | Time series of total phytoplankton abundances (black lines, gray
trend) and median taxa abundances (green lines, gray trend). Codes in panels
represent the name of lakes as in Table 1.
FIGURE S9 | Time series of total zooplankton abundances (black lines, gray
trend) and selectivity (ratio between copepods / daphnids – blue lines, light blue
trend). Codes in panels represent the name of lakes as in Table 1.
FIGURE S10 | Time series of median phytoplankton size (i.e. biovolume – black
lines, blue trend): codes in panels represent the name of lakes as in Table 1.
FIGURE S11 | Partial effects of lake morphometry predictors in the RF model
(Figures 5C, 6).
FIGURE S12 | Ranking of predictors and partial effects of environmental drivers
for the RF model of bootstrapped slopes after exclusion of the class
Cyanobacteria from the dataset.
TABLE S1 | Phytoplankton meta-database, including code identifier of taxa, full
taxonomic classification and taxa biovolumes.
REFERENCES
Andersen, K. H., Aksnes, D. L., Berge, T., Fiksen, Ø, and Visser, A. (2015).
Modelling emergent trophic strategies in plankton. J. Plankton Res. 37, 862–868.
doi: 10.1093/plankt/fbv054
Andersen, K. H., Berge, T., Gonçalves, R. J., Hartvig, M., Heuschele, J., Hylander, S.,
et al. (2016). Characteristic Sizes of Life in the Oceans, from Bacteria to whales.
Annu. Rev. Mar. Sci. 8, 217–241. doi: 10.1146/annurev-marine-122414-034144
Anneville, O., Gammeter, S., and Straile, D. (2005). Phosphorus decrease and
climate variability: mediators of synchrony in phytoplankton changes among
European peri-alpine lakes. Freshw. Biol. 50, 1731–1746. doi: 10.1111/j.1365-
2427.2005.01429.x
Anneville, O., Souissi, S., Gammeter, S., Straile, D., and Wimereux, S. M. D. (2004).
Seasonal and inter-annual scales of variability in phytoplankton assemblages:
comparison of phytoplankton dynamics in three peri-alpine lakes over a period
of 28 years. Freshw. Biol. 49, 98–115. doi: 10.1046/j.1365-2426.2003.01167.x
Armstrong, R. A. (1994). Grazing limitation and nutrient limitation in marine
ecosystems: steady state solutions of an ecosystem model with multiple food
chains. Limnol. Oceanogr. 39, 597–608. doi: 10.4319/lo.1994.39.3.0597
Finkel, Z. V., Beardall, J., Flynn, K. J., Quigg, A., Rees, T. A. V., and Raven,
J. A. (2010). Phytoplankton in a changing world: cell size and elemental
stoichiometry. J. Plankton Res. 32, 119–137. doi: 10.1093/plankt/fbp098
Gaedke, U., Seifried, A., and Adrian, R. (2004). Biomass size spectra and plankton
diversity in a shallow eutrophic lake. Int. Rev. Hydrobiol. 89, 1–20. doi: 10.1002/
iroh.200310661
Guiet, J., Poggiale, J. C., and Maury, O. (2016). Modelling the community size-
spectrum: recent developments and new directions. Ecol. Modell. 337, 4–14.
doi: 10.1016/j.ecolmodel.2016.05.015
Hansen, P. J., Koefoed, P., and Winding, B. (1997). Zooplankton grazing and
growth: scaling within the 2-2,000-micrometer body size range. Limnol.
Oceanogr. 42, 687–704. doi: 10.4319/lo.1997.42.4.0687
Huete-Ortega, M., Rodríguez-Ramos, T., López-Sandoval, D., Cermeño, P.,
Blanco, J., Palomino, R., et al. (2014). Distinct patterns in the size-scaling
of abundance and metabolism in coastal and open-ocean phytoplankton
communities. Mar. Ecol. Prog. Ser. 515, 61–71. doi: 10.3354/meps11007
Kremer, C. T., Gillette, J. P., Rudstam, L. G., Brettum, P., and Ptacnik, R. (2014).
A compendium of cell and natural unit biovolumes for >1200 freshwater
phytoplankton species. Ecology 95, 2984–2984.
Atkinson, D., Ciotti, B. J., and Montagnes, D. J. S. (2003). Protists decrease in size
linearly with temperature: ca. 2.5% C-1. Proc. R. So. B Biol. Sci. 270, 2605–2611.
doi: 10.1098/rspb.2003.2538
Lepori, F., Roberts, J. J., and Schmidt, T. S. (2018). A paradox of warming in a
deep peri-Alpine lake (Lake Lugano. Switzerland and Italy). Hydrobiologia 824,
215–228. doi: 10.1007/s10750-018-3649-1
Barton, A. D., Pershing, A. J., Litchman, E., Record, N. R., Edwards, K. F., Finkel,
Z. V., et al. (2013). The biogeography of marine plankton traits. Ecol. Lett. 16,
522–534. doi: 10.1111/ele.12063
Livingstone, D. M. (2003). Impact of secular climate change on the thermal
structure of a large temperate central European lake. Clim. Change 57, 205–225.
doi: 10.1023/a:1022119503144
Behrenfeld, M.
J., and Boss, E. S.
(2013). Resurrecting the Ecological
Underpinnings of Ocean Plankton blooms. Annu. Rev. Mar. Sci. 6, 167–194.
doi: 10.1146/annurev-marine-052913-021325
Marañón, E. (2015). Cell size as a key determinant of phytoplankton metabolism
and community structure. Annu. Rev. Mar. Sci. 7, 241–264. doi: 10.1146/
annurev-marine-010814-015955
Bonachela, J. A., Raghib, M., and Levin, S. A. (2011). Dynamic model of flexible
phytoplankton nutrient uptake. Proc. Natl. Acad. Sci. U.S.A. 108, 20633–20638.
doi: 10.1073/pnas.1118012108
Marañón, E., Cermeño, P., Latasa, M., and Tadonléké, R. D. (2012). Temperature,
resources, and phytoplankton size structure in the ocean. Limnol. Oceanogr. 57,
1266–1278. doi: 10.4319/lo.2012.57.5.1266
Breiman, L. (2001). Random forests. Mach. Learn. 45, 5–32.
Bürgi, H. R., and Stadelmann, P. (1991). Plankton succession in lake sempach, lake
hallwil and lake baldegg before and during internal restoration measures. SIL
Proc. 24, 931–936. doi: 10.1080/03680770.1989.11898884
Cavender-Bares, K. K., Rinaldo, A., and Chisholm, S. W. (2001). Microbial size
spectra from natural and nutrient enriched ecosystems. Limnol. Oceanogr. 46,
778–789. doi: 10.4319/lo.2001.46.4.0778
Cermeño, P., Marañón, E., Harbour, D., and Harris, R. P. (2006). Invariant scaling
of phytoplankton abundance and cell size in contrasting marine environments.
Ecol. Lett. 9, 1210–1215. doi: 10.1111/j.1461-0248.2006.00973.x
Cloern, J. E. (2018). Why large cells dominate estuarine phytoplankton. Limnol
Oceanogr. 63, 392–409.
Duncanson, L. I., Dubayah, R. O., and Enquist, B. J. (2015). Assessing the general
patterns of forest structure: quantifying tree and forest allometric scaling
relationships in the United States. Global Ecol. Biogeogr. 24, 1465–1475. doi:
10.1111/geb.12371
Edwards, K. F., Thomas, M. K., Klausmeier, C. A., and Litchman, E. (2012).
Allometric scaling and taxonomic variation in nutrient utilization traits and
maximum growth rate of phytoplankton. Limnol. Oceanogr. 57, 554–566. doi:
10.4319/lo.2012.57.2.0554
Finger, D., Wüest, A., and Bossard, P. (2013). Effects of oligotrophication on
primary production in peri-alpine lakes. Water Resour. Res. 49, 4700–4710.
doi: 10.1002/wrcr.20355
Marañón, E., Cermeño, P., Latasa, M., and Tadonléké, R. D. (2015). Resource
supply alone explains the variability of marine phytoplankton size structure.
Limnol. Oceanogr. 60, 1848–1854. doi: 10.1002/lno.10138
Marañón, E., Lorenzo, M. P., Cermeño, P., and Mouriño-Carballido, B. (2018).
Nutrient limitation suppresses the temperature dependence of phytoplankton
metabolic rates. ISME J. 12, 1836–1845. doi: 10.1038/s41396-018-0105-1
Mieleitner, J., Borsuk, M., Bürgi, H.-R., and Reichert, P. (2008). Identifying
functional groups of phytoplankton using data from three lakes of different
trophic state. Aquat. Sci. 70, 30–46. doi: 10.1007/s00027-007-0940-z
Monchamp, M.-E., Spaak, P., Dubois, N., Domaizon, I., Bouffard, D., and Pomati,
F. (2018). Homogenization of lake cyanobacterial communities over a century
of climate change and eutrophication. Nat. Ecol Evolu. 2, 317–324. doi: 10.1038/
s41559-017-0407-0
Monchamp, M.-E., Spaak, P., and Pomati, F. (2019). High dispersal levels and
lake warming are emergent drivers of cyanobacterial community assembly in
peri-Alpine lakes. Sci. Rep. 9, 7366–7366. doi: 10.1038/s41598-019-43814-2
Mousing, E. A., Ellegaard, M., and Richardson, K. (2014). Global patterns in
phytoplankton community size structure-evidence for a direct temperature
effect. Mar. Ecol Prog. Ser. 497, 25–38. doi: 10.3354/meps10583
Pomati, F., Matthews, B., Jokela, J., Schildknecht, A., and Ibelings, B. W. (2012).
Effects of re-oligotrophication and climate warming on plankton richness and
community stability in a deep mesotrophic lake. Oikos 121, 1317–1327. doi:
10.1111/j.1600-0706.2011.20055.x
Frontiers in Microbiology | www.frontiersin.org
16
January 2020 | Volume 10 | Article 3155
fmicb-10-03155
January 13, 2020
Time: 16:53
# 17
Pomati et al.
Environmental Drivers of Phytoplankton Size
Pomati, F., Matthews, B., Seehausen, O., and Ibelings, B. W. (2017). Eutrophication
and climate warming alter spatial (depth) co-occurrence patterns of lake
phytoplankton assemblages. Hydrobiologia 787, 375–385. doi: 10.1007/s10750-
016-2981-6
Pomati, F., Tellenbach, C., Matthews, B., Venail, P., Ibelings, B. W., and Ptacnik,
R. (2015). Challenges and prospects for interpreting long-term phytoplankton
diversity changes in Lake Zurich (Switzerland). Freshw. Biol. 60, 1052–1059.
doi: 10.1111/fwb.12416
Posch, T., Köster, O., Salcher, M. M., and Pernthaler, J. (2012). Harmful filamentous
cyanobacteria favoured by reduced water turnover with lake warming. Nat.
Clim. Change 2, 1–5.
Rasconi, S., Gall, A., Winter, K., and Kainz, M. J. (2015). Increasing water
freshwater plankton. PloS One
temperature triggers dominance of small
13:e0140449. doi: 10.1371/journal.pone.0140449
R-Development-Core-Team (2018). R: A Language and Environment for Statistical
Computing. Vienna: R-Development-Core-Team.
Reuman, D. C., Holt, R. D., and Yvon-Durocher, G. (2014). A metabolic perspective
on competition and body size reductions with warming. J. Anim. Ecol. 83,
59–69. doi: 10.1111/1365-2656.12064
Reuman, D. C., Mulder, C., Raffaelli, D., and Cohen, J. E. (2008). Three allometric
relations of population density to body mass: theoretical integration and
empirical tests in 149 food webs. Ecol. Lett. 11, 1216–1228. doi: 10.1111/j.1461-
0248.2008.01236.x
Reynolds, C. S.
(2006). Ecology of Phytoplankton. Cambridge: Cambridge
University Press.
Rodriguez, J., and Mullin, M. M. (1986). Relation between biomass and body
weight of plankton in a steady state oceanic ecosystem. Limnol. Oceanogr. 31,
361–370. doi: 10.4319/lo.1986.31.2.0361
Rose, J. M., and Caron, D. A. (2007). Does low temperature constrain the growth
rates of heterotrophic protists? Evidence and implications for algal blooms in
cold waters. Limnol. Oceanogr. 52, 886–895. doi: 10.4319/lo.2007.52.2.0886
Sentis, A., Binzer, A., and Boukal, D. S. (2017). Temperature-size responses alter
food chain persistence across environmental gradients. Ecol. Lett. 20, 852–862.
doi: 10.1111/ele.12779
Sheldon, R. W., Prakash, A., and Sutcliffe, H. (1972). The size distribution of
particles in the ocean. Limnol. Oceanogr. 17, 327–340. doi: 10.4319/lo.1972.17.
3.0327
Sommer, U., Adrian, R., De Senerpont Domis, L., Elser, J. J., Gaedke, U., Ibelings,
B. W., et al. (2012). Beyond the plankton ecology group (PEG) model:
mechanisms driving plankton succession. Ann. Rev. Ecol.Evolu. Syst. 43, 429–
448. doi: 10.1146/annurev-ecolsys-110411-160251
Sommer, U., Gliwicz, Z. M., Lampert, W. I., and Duncan, A. (1986). The PEG-
model of seasonal succession of planktonic events in fresh waters. Arch.
Hydrobiol. 106, 433–471.
Sommer, U., Sommer, F., Santer, B., Jamieson, C., Boersma, M., Becker, C.,
et al. (2001). Complementary impact of copepods and cladocerans on
phytoplankton. Eco. Lett. 4, 545–550. doi: 10.1046/j.1461-0248.2001.00263.x
Sprules, W. G., Barth, L. E., and Giacomini, H. (2016). Surfing the biomass size
spectrum: some remarks on history, theory, and application. Can. J. Fishe.
Aquat. Sci. 73, 477–495. doi: 10.1139/cjfas-2015-0115
Sprules, W. G., Brandt, S. B., Stewart, D. J., Munawar, M., Jin, E. H., and Love, J.
(1991). Biomass size spectrum of the lake michigan pelagic food web. Can. J.
Fish. Aquat. Sci. 48, 105–115. doi: 10.1139/f91-015
Sprules, W. G., and Munawar, M. (1986). Plankton size spectra in relation to
ecosystem productivity. Size, and Perturbation. Can. J. Fish. Aquat. Sci. 43,
1789–1794. doi: 10.1139/f86-222
Stibor, H., Vadstein, O., Diehl, S., Gelzleichter, A., Hansen, T., Hantzsche, F., et al.
(2004). Copepods act as a switch between alternative trophic cascades in marine
pelagic food webs. Ecol. Lett. 7, 321–328. doi: 10.1111/j.1461-0248.2004.00580.x
Straile, D., Jochimsen, M. C., and Kümmerlin, R. (2013). The use of long-term
monitoring data for studies of planktonic diversity: a cautionary tale from two
Swiss lakes. Freshw. Biol. 58, 1292–1301. doi: 10.1111/fwb.12118
Thomas, M. K., Fontana, S., Reyes, M., Kehoe, M., and Pomati, F. (2018). The
predictability of a lake phytoplankton community, over time-scales of hours
to years. Ecol. Lett. 21, 619–628. doi: 10.1111/ele.12927
Verdy, A., Follows, M., and Flierl, G. (2009). Optimal phytoplankton cell size in an
allometric model. Mar. Ecol. Prog. Ser. 379, 1–12. doi: 10.3354/meps07909
Waibel, A., Peter, H., and Sommaruga, R. (2019). Importance of mixotrophic
flagellates during the ice-free season in lakes located along an elevational
gradient. Aquat. Sci. 81, 45–45. doi: 10.1007/s00027-019-0643-2
Ward, B. A., Dutkiewicz, S., Jahn, O., and Follows, M. J. (2012). A size-structured
food-web model for the global ocean. Limnol. Oceanogr. 57, 1877–1891. doi:
10.4319/lo.2012.57.6.1877
Ward, B. A., and Follows, M. J. (2016). Marine mixotrophy increases trophic
transfer efficiency, mean organism size, and vertical carbon flux. Proc. Natl.
Acad. S.U.A. 113, 2958–2963. doi: 10.1073/pnas.1517118113
White, E. P., Enquist, B. J., and Green, J. L. (2008). On estimating the exponent
of power-law frequency distributions. Ecology 89, 905–912. doi: 10.1890/07-
1288.1
Winder, M., Reuter, J. E., and Schladow, S. G. (2009). Lake warming favours
small-sized planktonic diatom species. Proc. R. Soc BBiol. Sci. 276, 427–435.
doi: 10.1098/rspb.2008.1200
Winder, M.,
and Sommer, U.
changing climate. Hydrobiologia 698, 5–16.
90-5_2
(2012). Phytoplankton response
to a
doi: 10.1007/978-94-007-57
Wollrab, S., and Diehl, S. (2015). Bottom-up responses of the lower oceanic
food web are sensitive to copepod mortality and feeding behavior. Limnol. nd
Oceanogr. 60, 641–656. doi: 10.1002/lno.10044
Yankova, Y., Neuenschwander, S., Köster, O., and Posch, T. (2017). Abrupt
stop of deep water turnover with lake warming: drastic consequences for
algal primary producers. Sci. Rep. 7, 13770–13770. doi: 10.1038/s41598-017-13
159-9
Yvon-Durocher, G., Allen, A. P., Cellamare, M., Dossena, M., Gaston, K. J.,
Leitao, M., et al. (2015). Five years of experimental warming increases the
biodiversity and productivity of phytoplankton. PLoS Biol. 13:e1002324. doi:
10.1371/journal.pbio.1002324
Yvon-Durocher, G., Montoya, J. M., Trimmer, M., and Woodward, G. (2011).
Warming alters the size spectrum and shifts the distribution of biomass in
freshwater ecosystems. Global Change Biol. 17, 1681–1694. doi: 10.1111/j.1365-
2486.2010.02321.x
Conflict of Interest: The authors declare that the research was conducted in the
absence of any commercial or financial relationships that could be construed as a
potential conflict of interest.
Copyright © 2020 Pomati, Shurin, Andersen, Tellenbach and Barton. This is an
open-access article distributed under the terms of the Creative Commons Attribution
License (CC BY). The use, distribution or reproduction in other forums is permitted,
provided the original author(s) and the copyright owner(s) are credited and that the
original publication in this journal is cited, in accordance with accepted academic
practice. No use, distribution or reproduction is permitted which does not comply
with these terms.
Frontiers in Microbiology | www.frontiersin.org
17
January 2020 | Volume 10 | Article 3155
| null |
10.1128_mbio.01039-23.pdf
|
DATA AVAILABILITY
Crystallographic data have been deposited to the RCSB protein data bank (accessions
8FZY, 8FZZ, and 8G0K). Metagenomic sequencing data were previously published (32)
and are publicly available at the NCBI sequence read archive (BioProject PRJNA398089).
Plasmids and bacterial strains generated in the study are listed in Table S3 and will be
available upon reasonable request to the corresponding author.
|
DATA AVAILABILITY Crystallographic data have been deposited to the RCSB protein data bank (accessions 8FZY, 8FZZ, and 8G0K ). Metagenomic sequencing data were previously published (32) and are publicly available at the NCBI sequence read archive (BioProject PRJNA398089 ). Plasmids and bacterial strains generated in the study are listed in Table S3 and will be available upon reasonable request to the corresponding author.
|
| Bacteriology | Research Article
Structural disruption of Ntox15 nuclease effector domains by
immunity proteins protects against type VI secretion system
intoxication in Bacteroidales
Dustin E. Bosch,1 Romina Abbasian,1 Bishal Parajuli,1 S. Brook Peterson,2,3 Joseph D. Mougous2,3,4
AUTHOR AFFILIATIONS See affiliation list on p. 13.
ABSTRACT Bacteroidales use type VI secretion systems (T6SS) to competitively colonize
and persist in the colon. We identify a horizontally transferred T6SS with Ntox15 family
nuclease effector (Tde1) that mediates interbacterial antagonism among Bacteroidales,
including several derived from a single human donor. Expression of cognate (Tdi1)
or orphan immunity proteins in acquired interbacterial defense systems protects
against Tde1-dependent attack. We find that immunity protein interaction induces a
large effector conformational change in Tde nucleases, disrupting the active site and
altering the DNA-binding site. Crystallographic snapshots of isolated Tde1, the Tde1/Tdi1
complex, and homologs from Phocaeicola vulgatus (Tde2/Tdi2) illustrate a conserved
mechanism of immunity inserting into the central core of Tde, splitting the nuclease fold
into two subdomains. The Tde/Tdi interface and immunity mechanism are distinct from
all other polymorphic toxin–immunity interactions of known structure. Bacteroidales
abundance has been linked to inflammatory bowel disease activity in prior studies,
and we demonstrate that Tde and T6SS structural genes are each enriched in fecal
metagenomes from ulcerative colitis subjects. Genetically mobile Tde1-encoding T6SS
in Bacteroidales mediate competitive growth and may be involved in inflammatory
bowel disease. Broad immunity is conferred by Tdi1 homologs through a fold-disrupting
mechanism unique among polymorphic effector–immunity pairs of known structure.
IMPORTANCE Bacteroidales are related to inflammatory bowel disease severity and
progression. We identify type VI secretion system (T6SS) nuclease effectors (Tde) which
are enriched in ulcerative colitis and horizontally transferred on mobile genetic elements.
Tde-encoding T6SSs mediate interbacterial competition. Orphan and cognate immunity
proteins (Tdi) prevent intoxication by multiple Tde through a new mechanism among
polymorphic toxin systems. Tdi inserts into the effector central core, splitting Ntox15 into
two subdomains and disrupting the active site. This mechanism may allow for evolution
ary diversification of the Tde/Tdi interface as observed in colicin nuclease–immunity
interactions, promoting broad neutralization of Tde by orphan Tdi. Tde-dependent T6SS
interbacterial antagonism may contribute to Bacteroidales diversity in the context of
ulcerative colitis.
KEYWORDS microbiome, Bacteroides, type VI secretion system, inflammatory bowel
disease, structural biology
T he Bacteroidota phylum is a major component of the healthy intestinal microbiome
community. Specific taxa within this phylum, and their relative abundances have
been linked to diverse diseases including components of the metabolic syndrome (1–3),
viral infection (4), and colorectal carcinogenesis (5). Members of the Bacteroidales order
may also have a role in severity and progression of inflammatory bowel disease (IBD) (6).
Editor Karina B. Xavier, Instituto Gulbenkian de
Ciência, Oeiras, Portugal
Address correspondence to Dustin E. Bosch, dustin-
[email protected].
The authors declare no conflict of interest.
See the funding table on p. 14.
Received 25 April 2023
Accepted 3 May 2023
Published 22 June 2023
Copyright © 2023 Bosch et al. This is an open-access
article distributed under the terms of the Creative
Commons Attribution 4.0 International license.
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 1
Research Article
mBio
IBD includes both Crohn’s disease (CD) and ulcerative colitis (UC), related diseases
with distinct pathophysiology. Crohn’s disease is characterized by “full-thickness”
inflammation extending through all layers at any location along the gastrointestinal
tract. In contrast, the inflammation of UC is confined to the superficial layers of the colon.
Active inflammation in Crohn’s disease correlates to Phocaeicola vulgatus abundance,
while reduced Bacteroides spp. were observed in UC patients with diarrhea and rectal
bleeding (7, 8). Several Bacteroidales were among the taxa with greatest fluctuation over
time in a large longitudinal IBD microbiome study, suggesting dynamic re-organization
of their niche(s) during disease development (9). In summary, correlative human studies
suggest that alterations among specific Bacteroidaceae may contribute IBD progres
sion, and patterns differ between Crohn’s disease and UC. The longitudinal changes in
Bacteroidales abundance observed in IBD may be influenced by competitive interactions
(10, 11).
Bacteroidales and other commensals in the intestinal microbiome engage in
contact-dependent interbacterial antagonism using toxin secretion systems (11, 12).
Type VI secretion system (T6SS) gene clusters encode at least 13 structural proteins that
assemble into a needle-like apparatus for delivery of effectors (toxins) into neighboring
bacteria (13). A contractile sheath composed of TssB and TssC propels an inner tubu
lar structure composed of hexameric hemolysin co-regulatory protein (Hcp) through
multi-protein baseplate and membrane complexes. Hcp and the T6SS tip structure
are secreted into recipient periplasm and/or cytoplasm, carrying payloads of effectors
(toxins) that promote cell death (14). T6SSiii gene clusters found in Bacteroidales are
distantly related to model systems in Pseudomonadota (T6SSi), encoding nine shared
core structural proteins and five core proteins restricted to Bacteroidales (15). Three
prototypic T6SS genetic architectures have been described in Bacteroidales, two found
on mobile elements a third largely restricted to Bacteroides fragilis (15). Conjugative
transfer of mobile GA1 and GA2 type T6SS among Bacteroidales has been observed
within the intestinal microbiome (16, 17). Bacteroides spp. utilize these T6SS to antago
nize non-immune Bacteroidales (12, 18) and establish and maintain colonization (11).
T6SS effector delivery frequently involves direct interaction with Hcp or the tip
structure (19, 20). However, some effector domains are translationally fused to secre
ted core components (21, 22). For example, pyocin and colicin type DNAse domain
fusions with Hcp mediate T6SS-dependent antagonism in Pseudomonadota (21). T6SS
effectors employ a striking array of activities to disrupt essential biologic functions,
spanning enzymatic degradation of key small molecules, post-translational modification
of essential proteins, and disruption of membrane and peptidoglycan layer barriers
(14). Effectors with novel toxin 15 (Ntox15) DNAse domains, also known as toxin_43
domains, degrade recipient genomic DNA (23). Ntox15 effectors in the soil bacterium
Agrobacterium tumefaciens, T6SS DNase effectors (Tde1-2), mediate competition in
planta. Secretion of A. tumefaciens Tde effectors requires loading onto the C-termini
of tip structure proteins with aid of adaptor/chaperone proteins (24, 25). Loading
of Tde1/2 onto the tip structure is required for efficient sheath assembly and T6SS
secretion (26). Cognate immunity proteins are encoded adjacent to T6SS effectors and
neutralize their activity to prevent intoxication of self and kin (27). Immunity proteins
usually prevent intoxication by direct occlusion of the effector active site (28, 29).
Less common mechanisms include enzymatic antagonism of effector activities, e.g.,
reversal of toxin-mediated ADP ribosylation (30). Arrays of immunity proteins are also
found encoded by gene clusters unassociated with a T6SS apparatus, termed “orphan”
immunity proteins. These AIDs are frequently on mobile genetic elements and hori
zontally transferred to confer protection from type VI attack, impacting competitive
colonization among Bacteroidales (27).
Bacteroidales T6SS have been implicated in mouse models of infectious colitis.
Commensal B. fragilis strains use T6SS to competitively exclude pathogenic enterotoxin-
producing strains and protect against colitis (10). We hypothesize that T6SS effector-
mediated competitive colonization underlies associations of Bacteroidales with IBD
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 2
Research Article
mBio
severity and progression (6, 8). In this study, we show that T6SS loci encoding Tde family
nuclease effectors are specifically enriched in ulcerative colitis metagenomes compared
to Crohn’s disease and healthy controls. We also show that immunity against Tde-medi
ated attack occurs by structural disruption of the effector domain, a mechanism unique
among polymorphic toxin–immunity pairs of known structure.
RESULTS
Bacteroidales T6SS, Ntox15 domains, and immunity proteins are enriched in
ulcerative colitis fecal metagenomes
Prior studies have implicated T6SS and specific effector–immunity pairs in enterotoxi
genic B. fragilis colitis (10). Based on these data, we asked whether T6SSiii loci and
particular effector types are enriched among bacterial communities of IBD patient fecal
samples. We constructed hidden Markov models (HMM) for the conserved Bacteroi
dales T6SS structural proteins, as well as ~150 Bacteroidales polymorphic toxin domain
families and associated immunity proteins (31). These HMMs were applied to a large
collection of publicly available shotgun metagenomic sequencing data from humans
with inflammatory bowel disease and healthy controls (32). This Integrative Human
Microbiome Project cohort included biweekly stool samples from 67 subjects with
Crohn’s disease, 38 with UC, and 27 non-IBD controls (9). Strong correlation of HMM hits
among the T6SS structural proteins was observed, as expected, because the correspond
ing genes are co-inherited in T6SS loci (Fig. S1A). HMM hit quantities per reads, corrected
for relative Bacteroidales abundance, of each T6SS structural protein were similar across
metagenomes (Fig. S1C), except for TssH, a AAA family ATPase which was excluded
from further analysis due to off-target HMM hits. T6SS structural genes were enriched
in fecal metagenomes from UC patients compared to CD (Fig. 1A). The enrichment of
T6SS structural gene hits in UC extends to comparison with non-IBD “healthy control”
specimens and is not explained by differential relative Bacteroidales abundance (Fig.
1B). Among the ~150 polymorphic toxin domain HMMs, greatest enrichment in UC was
for Ntox15 homologs (Fig. 1A). Ntox15 hits, corrected to Bacteroidales abundance were
enriched in UC relative to CD and controls, while the associated immunity gene did
not differ significantly across groups (Fig. 1C). There was relative enrichment of Ntox15
genes per T6SS structural gene (TssB) in ulcerative colitis samples compared to non-IBD
controls, and relative depletion in Crohn’s disease (Fig. S1D; linear fit slope 1.0 [0.9–1.1]
for UC, 0.5 [0.4–0.7] for non-IBD, 0.0 [-0.1–0.1] for CD). A subset of the metagenomic
data analyzed were time course samples from individual subjects. Multivariate analysis
indicated that T6SS hits per Bacteroidales abundance tended to increase over time
in subjects with ulcerative colitis (Fig. S1B). We conclude that T6SSiii and Ntox15-encod
ing genes are differentially abundant in the intestinal metagenomes of humans with
inflammatory bowel disease, and all are enriched in UC.
Bacteroidales from a single human intestinal community compete with T6SS
encoding Tde nuclease effectors
To identify strains for functional studies on Ntox15 effectors, we queried a human
intestinal commensal bacteria collection with whole genome sequencing (34). Several
Phocaeicola and Bacteroides strains contain nearly identical Ntox15-encoding T6SS of the
GA2 type architecture (15). These strains were all isolated from a single human donor,
and their T6SS loci are encoded with neighboring mobile genetic element-related genes,
highly suggestive of horizontal transfer events. Selection for specific T6SS effector and
immunity pairs has importance for competitive colonization and persistence in human
gut metagenomes (11). This T6SS encodes several Hcp proteins, a completely conserved
(100% amino acid identity) Hcp-effector fusion with C-terminal Ntox15 domain, and an
adjacent putative cognate immunity protein (Fig. 1D). This multispecies effector–
immunity pair is termed Tde1 and Tdi1 to conform with nomenclature in Agrobacterium
(23). Each genome also encodes putative effectors with rearrangement hotspot (RHS)
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 3
Research Article
mBio
FIG 1 Ntox15 domains enriched in IBD metagenomes mediate T6SS-dependent interbacterial antagonism among Bacteroidales. Metagenomic sequencing
reads with similarity to Bacteroidales T6SS, Ntox15 domains, and immunity proteins were detected with hidden Markov models [HMMer (33)]. (A) T6SSiii
structural genes and Ntox15 domain homologs are enriched in fecal metagenomes from patients with UC compared to CD (32). False discovery rate adjustment
for multiple comparisons was with the Benjamini–Hochberg method. (B, C) Aggregated T6SS structural genes and Ntox15 homologs, but not the associated
immunity are enriched in UC over CD and non-IBD controls after correction for relative Bacteroidales abundance. P-value reflects Kruskal–Wallis test. (D) A gene
structure diagram of a T6SS-encoding locus that is identical in several genetically diverse Bacteroidales isolates from a single human donor. In addition to other
T6SS structural genes (gray), there are five hcp genes (blue), including one fused with a C-terminal Ntox15 domain (tde1, green) and an immediately adjacent
immunity gene (tdi1, red). An HxxD motif is conserved at the putative active site, predicted to confer nuclease activity. (E) In competitive growth experiments
with P. vulgatus ATCC 8482, deletion of tde1 and tdi1 from MSK 16.10 or MSK 16.2 confers reduced relative fitness. Effector/immunity deletion is also a competitive
disadvantage relative to the isogenic parental strain. Thymidine kinase (tdk) is deleted to confer resistance to the selection agent floxuridine (FUdR). (F) tde1/tdi1
mediate competition between MSK 16.10 and MSK 16.2, isolates from a single human host. Statistical indicators reflect Student’s t-test: ** P < 0.01, *** P < 0.001.
(G) Tde1-dependent antagonism requires structural sheath proteins TssB and TssC. P-values reflect analysis of variance (ANOVA) tests for each recipient.
domains adjacent to mobile element genes, which have predicted structural similarity to
the Tre23 toxin of Photorhabdus laumondii (35).
We hypothesized that Tde1 mediates interbacterial competition among Bacteroidales.
Deletion of tde1 in two of these T6SSs, Phocaeicola vulgatus strains MSK 16.2 and MSK
16.10 (34), enhanced competitive survival of a recipient P. vulgatus strain ATCC 8482
that lacks immunity (Fig. 1E). Deletion of tde1 and tdi1 in MSK 16.10 also conferred
a competitive disadvantage relative to the isogenic parent strain, indicating that tdi1
likely protects against kin intoxication (Fig. 1E). The competitive disadvantage of tde1
deletion could be explained by requirement of the Hcp domain for T6SS assembly, but
the four other hcp-encoding genes may compensate. Horizontal transfer of this mobile
T6SS suggested that Tde1 may mediate cell killing among strains from a single host’s
microbiome. Indeed, there was tde1-dependent killing of MSK 16.10 by MSK 16.2 when
tde1 and tdi1 were removed from the recipient (Fig. 1F). Antagonism of MSK16.10 by MSK
16.2 required assembly of the T6SS apparatus with sheath proteins TssB and TssC (Fig.
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 4
Research Article
mBio
1G). We conclude that Hcp-Ntox15 effectors mediate T6SS-dependent competition with
non-immune Bacteroidales, including strains derived from a single host.
Bacteroidales Tde effectors are magnesium dependent DNAses with a
distinct α-helical fold
To identify mechanisms of Ntox15 effector toxicity, we characterized the structure
and enzymatic function of Tde1. The distantly homologous Ntox15 domain-containing
effector Tde1 in A. tumefaciens exhibited DNAse activity in vitro and in cells (23). To
examine enzymatic activity of Tde1 from P. vulgatus, we co-produced the Ntox15 domain
(Tde1tox) in E. coli with Tdi1 to circumvent toxicity, separated it from immunity under
denaturing conditions, and refolded it. Tde1tox exhibited DNAse activity on plasmid
dsDNA, which was abrogated by mutation of the HxxD active site (H279A) and strongly
inhibited by the presence of Tdi1 or chelation of divalent cations using EDTA (Fig. 2A).
EDTA-mediated inhibition was reversed by addition of molar excess magnesium salts, but
not other divalent cations (Fig. 2A). Slower migration of plasmid DNA in the presence
of Tde1tox H279A, excess ZnCl2 or CaCl2 suggest protein binding and/or effects on
supercoiling. Consistent with the toxin exhibiting non-specific DNAse activity, catalyt
ically inactive Tde1tox H279A/D282A directly interacted with 30-nucleotide single- or
double-stranded DNA oligomers of random sequence, with equilibrium binding affinities
near 500 nM (Fig. 2B; Fig. S2B). DNA binding affinity may be impacted by the dual point
mutations in the active site.
FIG 2 The DNAse Tde1 adopts an α-helical predominant fold with HxxD motif active site. (A) Refolded Tde1 Ntox15 domain degraded plasmid dsDNA. Nuclease
activity was impaired by mutation of the HxxD motif, addition of molar excess immunity protein, or chelation of divalent cations with EDTA. Tde1tox nuclease
activity impairment by EDTA was reversed by addition of molar excess magnesium, but not zinc or calcium. (B) Tde1tox with active site mutations interacted
with both double- and single-stranded biotinylated oligonucleotides of random sequence, measured with biolayer interferometry. (C) A crystal structure of
catalytically inactive Tde1tox H279A/D282A domain (Table S1) was obtained by molecular replacement using an AlphaFold2 prediction (36). Tde1tox adopts a
single domain fold with the predicted DNA binding surface (green). Mutation of key basic residues (green sticks) to alanine or acidic residues decreased DNA
binding affinity (Fig. S2F). The active site corresponds to the HxxD motif (red) and contains a modeled sulfate anion, present due to crystallization is high
concentrations of ammonium sulfate.
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 5
Research Article
mBio
A structural model of Tde1tox H279A/D282A Ntox15 domain was obtained by X-ray
crystallography with diffraction data extending to 2.9 Å resolution (Table S1; Fig. 2C).
Although no close homologs of known structure were available, phases were solved
by molecular replacement using an AlphaFold2 prediction model (Fig. S2C) (36). The
AlphaFold2 prediction model was very similar to the experimental crystal structure;
mean Cα r.m.s.d. among the eight monomers in the asymmetric unit was 1.0 Å (Fig. S2D).
The Ntox15 domain adopts a globular fold which is predominantly α-helical, forming a
short α-sheet between helices 5 and 6 immediately adjacent to the active site (Fig. 2C).
A structural similarity search with DALI (37) revealed no close homologs within the PDB,
including known nuclease structures (Z score 6.0 and Cα r.m.s.d 4.1 over 77 residues for
the top hit, two pore calcium channel PDB id 6NQ1). A cavity adjacent to the mutated
HxxD motif marks the active site. A sulfate ion is modeled within the active site, likely
an artifact of crystallization in high concentration of ammonium sulfate. However, it may
mimic accommodation of negatively charged moieties of the DNA substrate. The DNA
binding site predicted with ProNA2020 (38) maps to helices 4, 6, and 7, adjacent to
the active site (Fig. 2C). Coulombic surface rendering highlights relative positive surface
charge surrounding the active site pocket, consistent with favorable electrostatics for
interaction with negatively charged DNA (Fig. S2E). Point mutation of basic residues
at the predicted DNA binding surface decreased dsDNA binding affinity (Fig. 2C; Fig.
S2F). Charge reversal substitutions had greatest impact on DNA binding, supporting
likely importance of electrostatic interactions. We conclude that Ntox15 domains adopt a
globular fold, distinct from other nuclease families of known structure, with a structurally
well-defined active site that mediates DNAse activity.
Orphan tdi are frequent among human intestinal commensal bacteria
Ntox15 domain and core T6SS protein-encoding sequences were both enriched in UC
metagenomes while immunity-encoding sequences (tdi) were not (Fig. 1C), raising the
possibility of widespread tdi genes outside of T6SS loci. We assessed the distribution of
T6SS, Ntox15, and immunity protein encoding genes among a large collection of human
intestinal commensal genomes (34) using BLAST (39) and the Tde1-related T6SS genes as
queries (Fig. 3A). The core structural tssC gene was identified exclusively in Bacteroidota,
reflecting substantial sequence-level dissimilarity of the Bacteroidales T6SSiii relative to
Pseudomonadota. Ntox15 domain homologs were confidently identified (BLAST E-value
< 10−10) in 14 Bacteroidota strains, all with GA2 T6SS architecture. In contrast, 120 strains
encoded Tdi1 homologs, including all genetic architectures (Fig. 3A). Nine Bacteroidota
shared a similar gene structure with the Tde1-associated system query (Fig. S3), having
immediately adjacent Hcp-Ntox15 fusion and immunity proteins within the context
of a GA2 T6SS structural gene cluster. More distantly related Ntox15 domain-contain
ing proteins were encoded adjacent to Tdi1-like immunity proteins in five Firmicute
genomes (Roseburia intestinalis and Tyzzerella nexilis). The genomic context and domain
organization (e.g., an LXG domain fusion) suggest association with type VII secretion
systems.
Notably, most of the Tdi1 homolog encoding genes were found in organisms without
a Tde1 homolog, raising consideration of widespread orphan immunity among intestinal
commensal bacteria (Fig. 3A) (27). The order-of-magnitude higher frequency of tdi
compared to tde is consistent with the higher median frequency of tdi homolog sequen
ces in metagenomes (Fig. 1C) and suggests one explanation for lack of correlation
between tdi and disease state. Genomic context within 5 kb of these immunity genes
frequently contained other putative immunity genes, distinct in sequence and domain
structure, as well as genes associated with mobile genetic elements (Fig. S3). These
findings suggest that Tdi1 homolog genes are frequently found in arrays of diverse
immunity genes associated with mobile genetic elements, compatible with acquired
immune defense (AIDs) systems (27).
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 6
Research Article
mBio
FIG 3 Cognate and orphan immunity proteins protect against T6SS-mediated attack by inducing a
conformational shift in Tde1 to disrupt the DNA binding and active sites. (A) Query of Tde1, Tdi1, and
representative T6SS structural protein (TssC) against a collection of ~1,200 human intestinal commensal
genomes (40) with BLAST revealed predominant distribution of homologs within Bacteroidota. TssC
homologs from previously described genetic architectures (GA1-3) cluster together (15). Tde1, but not
Tdi1 homologs are exclusively in GA2 T6SS. Several Firmicutes harbor tde/tdi pairs not associated with
T6SS. Immunity encoding genes were more abundant than tde. (inset) A Venn diagram illustrates that all
identified tde1 homologs were accompanied by tdi. tde/tdi pairs were associated with a T6SS apparatus in
9 Bacteroidota and 5 Firmicutes. However, tdi genes were more frequently encountered than tde in both
phyla, indicating presence of orphan immunity genes. (B) Tde1tox • Tdi1 exhibited higher thermal stability
(melting temperature 67°C) than either component alone (55–55.5°C) in SYBR orange thermal melt
experiments. (C and D) Biolayer interferometry demonstrated comparable equilibrium binding affinities
of Tde1tox for Tdi1, as well as two homologous orphan immunity proteins (KD 18–24 nM). (E) Expression
of Tdi1, as well as two orphan immunity proteins from diverse Bacteroidota protect P. vulgatus ATCC
8482 against tde1-dependent attack by P. vulgatus MSK 16.10. (F) Crystal structures of two homologous
Tdetox (blues) and Tdi (gray, tan) complexes demonstrate a splitting of Ntox15 into two subdomains. The
subdomains are linked by the DNA binding site and the HxxD motif, which are partially disordered in
the crystal structures (dotted lines). The predicted DNA binding site is green, and basic residues required
for high affinity DNA interaction represented as sticks. There is high structural similarity among the
homologs, indicating a conserved mode of interaction.
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 7
Research Article
mBio
Cognate and orphan immunity proteins promiscuously engage Tde nucleases
to protect against killing
Frequent occurrence of Tdi homologs in AIDs suggests that orphan immunity toward Tde
toxins is an important mechanism of competition among Bacteroidales. Bacteroidales
orphan immunity and effector interactions have not been biochemically characterized
previously. We first characterized Tde1tox H279A/D282A and Tdi1 binding with multiple
biochemical platforms (Fig. 3). Tde1tox interaction with Tdi1 increases thermal stability
(melting temperature 67°C versus 55.5°C, Fig. 3B). A Tde1tox/Tdi1 dissociation constant of
18 nM was measured by biolayer interferometry (BLI, Fig. 3C and D). Two putative orphan
immunity proteins were selected for further study, based on their presence in several
intestinal commensal bacterial genomes, and gene structures compatible with AIDs (Fig.
S3). These two proteins, termed Tdi orphan A and B (TdioA and TdioB), share 61–65%
sequence identify with Tdi1. Both orphan immunity proteins, recombinantly produced
from E. coli, directly interacted with Tde1tox H279A/D282A (Fig. 3D). Affinities of TdioA and
TdioB for Tde1tox (24 and 19 nM) were very similar to that of the cognate immunity Tdi1.
Orphan immunity proteins co-expressed with the P. vulgatus dnLKV7 homolog Tde2tox
in E. coli also formed a stable 1:1 complex, as detected with analytical gel filtration
chromatography (Fig. S4).
We next examined protective effects of orphan immunity genes in competitive
growth experiments. Expression of Tdi1 from a chromosomally inserted transposon
(pNBU2) in P. vulgatus ATCC 8482 markedly reduced tde1-dependent killing by MSK
16.10 (Fig. 3E). Similarly, TdioA and TdioB were highly protective. We conclude that orphan
immunity proteins directly engage both Tde1 and Tde2 (Fig. 3B through D; Fig. S4).
Orphan immunity proteins have high affinity for Tde1 and provide competitive growth
advantage in co-culture with the Tde1-encoding strain P. vulgatus MSK 16.10.
Immunity proteins disrupt nuclease activity by inserting into the nuclease
central core: a new mechanism of polymorphic toxin immunity
We next sought a structural explanation for how promiscuous neutralization of Tde
effectors by diverse Tdi homologs is achieved. We therefore obtained crystal structures of
the Tde1 Ntox15 domain in complex with Tdi1, as well as a homologous complex from
P. vulgatus dnLKV7, Tde2tox and Tdi2 (Fig. 3F). The Tdetox/Tdi complex homologs exhibit
very similar structure despite 51% sequence identity between the Ntox15 domains,
indicating a conserved mode of effector–immunity interaction. Tdi1/2 have structural
homology to the Ntox15-associated immunity protein from A. tumefaciens (Atu4351, PDB
ID 6ITW), which has been crystallized in isolation (23). Tdi1 and Atu4351 align with a Cα
r.m.s.d. of 1.2 Å (Fig. S2E), although Bacteroidales Tdi1 exhibits a slightly more compact
overall structure with shortening of several loops (e.g., β8-α5). When bound to immunity
proteins, Tde1tox and Tde2tox split into two subdomains (Fig. 3F). Forty percent (17 of
43) of immunity-contacting Tde1/2tox residues in the effector immunity structures form
part of the central core in the globular Ntox15 domain alone structure, and many of
these are highly conserved (Fig. S5). There is an ~32 amino acid region disordered in
the crystal structure, corresponding to β2, α6, and the surrounding loops in the Tde1tox
only structure. Notably, this disordered region contains part of the HxxD active site motif
and most of the DNA binding site (Fig. S5). Superposition of the Tde1tox alone structure
with the Tde1tox/Tdi1 complex indicates a conformational shift characterized by a hinge
motion, as well as an ~180° relative rotation of the two Tde1tox subdomains (Fig. 4A).
We conclude that Tdi immunity proteins induce a marked conformational shift in Tde
effectors, driving a division into two subdomains with disruption of the enzymatic active
site and DNA binding motif.
Tdi1 and Tdi2 form extensive contacts with the conserved central cores of their Tdetox
counterparts (Fig. 3F). Upon immunity interaction, Tde effectors undergo a dramatic
conformational shift, highlighted by superposition of the Tde1tox alone and Tde1tox/Tdi1
complex structures (Fig. 4A). The immunity protein does not sterically occlude the active
site, but rather splits the effector into subdomains and structurally distorts the active site,
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 8
Research Article
mBio
FIG 4 Effector fold disruption is a new immunity mechanism among polymorphic toxins. (A) The Tde1tox alone structure (red) is superimposed on the Tde1tox/
Tdi1 complex structure. Upon immunity binding, the split subdomains of Tde1tox undergo a relative ~90° hinge motion and ~180° rotation. The DNA binding
site (including helix α6) and the active site (HxxD yellow) are disrupted by the conformational shift. (B) Solvation energy gains of effector/immunity interface
formation as percentages of monomer solvation energy were calculated with PDBePISA (41). Included structural models with PDB accession and PubMed IDs are
listed in Table S2. Tde1tox/orphan immunity calculations are derived from comparative homology models based on the Tde1tox/Tdi1 structure. (C) The “capping”
mechanism with non-disruptive steric occlusion of the effector active site is typified by the Pseudomonas aeruginosa T6SS-assocated peptidoglycan hydrolase
Tse1/Tsi1. (D) Several T6SS and other polymorphic toxin/immunity interactions involve insertion of the immunity protein into a pre-formed effector active site
crevice (“plugging”), typified by P. aeruginosa (P)ppApp synthetase Tas1/immunity. A predicted model of Tas1 alone, supported by an experimental structure of
homolog RelQ (not shown, PDB 5DEC), indicates lack of large conformational shift in the effector. (E) A structure of colicin E3 RNAse exhibits engagement of
immunity at an “exosite” separate from the enzymatic active site (42). Unlike Tde1tox/Tdi1, large effector conformational shifts are not predicted.
which is disordered in the crystal structures. Advances in deep learning have improved
prediction accuracy for protein-protein interfaces (43), leading us to ask whether the Tde
conformation shift mechanism of Tdi immunity is computationally predictable. However,
AlphaFold-Multimer predicted Tde1-2tox/Tdi1-2 complexes inaccurately in the absence of
an experimentally derived template structure (Fig. S6). The Tdetoxα4-α5 helices interface
with Tdi is approximated by the models, but effector conformational shifts and the
secondary immunity interface are not identified. Thus, the Tdi1 immunity mechanism
differs from previous structural investigations of T6SS-related effector–immunity pairs
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 9
Research Article
mBio
and cannot be reliably predicted from primary sequences with current deep learning
algorithms.
To identify similar immunity mechanisms among polymorphic toxins, we compared
the Tdetox/Tdi structure to all other polymorphic toxin–immunity pairs in the Protein Data
Bank. The hydrophobic nature of Tde’s interactions with Tdi are reflected numerically
in solvation energy calculations from the PDBePISA web server (41). Specifically, there
is a relatively large solvation energy gain upon complex formation as compared to the
Tdetox monomers alone (Fig. 4B). Comparative homology models of Tde1tox in complex
with orphan immunity proteins, using the Tde1tox/Tdi1 crystal structure as a template,
yielded similar solvation energy changes to the cognate immunity–effector pairs (44).
As numeric markers of interface hydrophobicity, solvation energy gains were likewise
calculated for each polymorphic toxin–immunity pair in the PDB (Fig. 4B). Most other
effector–immunity interfaces cluster with relatively low solvation energy gains for both
effector and immunity. Among the T6SS effector–immunity complexes, this pattern
corresponds to immunity “capping” for steric occlusion of the effector active site, typified
by the T6SS-associated Tse1/Tsi1 interaction in P. aeruginosa (Fig. 4C). Overlay of the
Tse1 only structure (PDB 4EQ8) with the Tse1/Tsi1 complex (PDB 4EQA) demonstrates the
absence of conformation shifts as found in Tde1 (Fig. 4A through C). A related mecha
nism of immunity, “plugging” or insertion of the immunity into a preformed effector
active site cleft is illustrated with the Tas1 and immunity complex structure (PDB 6OX6)
from P. aeruginosa (Fig. 4D). In contrast with Tdetox/Tdi, interactions of this type uniformly
occur at the active site and do not result in large conformational shifts. While a Tas1 only
structure is not available, an AlphaFold2 predicted model and structural homolog RelQ
from Bacillus subtilis (PDB 5DEC) exhibit similar conformations to the effector in complex
with immunity and an open active site crevice (Fig. 4D) (45). Several effector–immunity
interactions of this pattern produced relatively high immunity solvation energy gains
(Fig. 4B). The E. coli colicin E3 ribonuclease and immunity interfaces (PDB 1E44, 1JCH)
showed parallels to Tdetox/Tdi1 in having relatively high effector solvation energy gain
calculations (Fig. 4B) and an immunity interface that does not overlap with the effec-
tor active site (46) (Fig. 4E). Similar to other colicin nucleases, immunity is conferred
by high-affinity interaction at an “exosite” (42, 47). A model of the isolated colicin E3
effector domain, predicted with AlphaFold2, shows a highly similar fold to the immunity
complex, except for ~9 residues at the N-terminus. This region is predicted with low
confidence in the isolated colicin E3 and assumes a short helix with extensive immunity
contacts in the complex crystal structure (Fig. 4E). However, the marked conformational
shift and central core interactions observed in Tde1/Tdi1 are lacking.
We conclude that Tde conformational shift and active site disruption mediated by Tdi
differs from previously described polymorphic toxin–immunity interactions. Immunity
contacts with the effector central core are reflected in solvation energy calculations. In
contrast to the predominant active site occlusion immunity mechanisms, Tdi inserts into
the Tde central core, dividing the effector domain and disrupting the active site structure.
DISCUSSION
from a single human donor
Our finding of essentially identical T6SS apparatus genes and Tde1–Tdi1 within
is highly suggestive of
diverse Bacteroidales
intestinal micro
recent horizontal gene transfer, possibly within the donor’s
biome. Tde1-dependent competition among these strains implies selective pressure
favoring acquisition of T6SS. Consistent with prior literature, we find T6SS gene
clusters and acquired immune defense systems frequently associated with mobile
genetic elements (16, 27). Active exchange and selection for genetic material
relevant to T6SS-mediated attack supports previously described hypotheses that
interbacterial competition among the Bacteroidota is an important determinant of
the microbial community composition in individual hosts (11, 16).
Polymorphic toxins have been implicated in virulence of certain pathogenic bacteria,
with mechanisms including toxin delivery to host cells (48, 49). However, disease
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 10
Research Article
mBio
associations with human commensal bacterial polymorphic toxins have been less
thoroughly explored. In one prior study, T6SSs of commensal B. fragilis strains were
important for competitive exclusion of pathogenic enterotoxin-producing strains (10).
In this study, we find enrichment of T6SS structural genes and Ntox15 domains in
patients with ulcerative colitis, suggesting positive selection for this effector immunity
pair. T6SSs with tde homologs are found in P. vulgatus, and we demonstrate tde-mediated
antagonism among three intestinally derived strains. P. vulgatus abundance associates
with IBD disease activity (7). Furthermore, colonization with some strains of P. vulgatus
modulates inflammation severity in rodent colitis models, although none tested in these
model studies are known to encode tde–tdi homologs (50). Bacteroidales T6SSs and
Ntox15 effectors might contribute directly to the etiology of UC, or the disease process
(inflammation, epithelial disruption, etc.) may favor Bacteroidales with T6SS and tde. The
latter hypothesis is supported by significant increases in relative T6SS gene abundance
in time course metagenomic data from subjects with UC. Interestingly, UC and Crohn’s
disease metagenomes exhibited opposite patterns of Ntox15 gene abundance relative to
structural T6SS genes. This pattern raises the possibility that encoding Ntox15 domains
may be advantageous to bacteria in UC, but detrimental in Crohn’s disease. Alternatively,
there may be differential abundances of Bacteroidales with different T6SSiii genetic
architectures in the two disease states, which cannot be quantified with our HMM
approach.
The Tde–Tdi proteins investigated in our study bear distant homology to T6SS
effector–immunity pairs in A. tumefaciens (23). Like A. tumefaciens Tde1, the Bacteroidales
Ntox15 domain exhibits magnesium-dependent DNAse activity. These domains are likely
toxic due to non-targeted degradation of DNA in recipient cells. Given the enzymatic
similarity of the effectors and the structural similarity of the immunity proteins, the
immunity mechanism is very likely conserved. Mechanisms of secretion of the Bacteroi
dales Tde1/2 fused to Hcp are distinct from the non-covalent tip structure interactions
described in Agrobacterium Tde1/2 (25, 26). The adaptor/chaperone proteins Tap-1 and
Atu3641 required for Agrobacterium effector delivery are absent in Bacteroidales T6SS
(25). Similarly, Bacteroidales Tde lack the N-terminal glycine zipper motif described as
important for translocation of Agrobacterium Tde1 into recipient cells (51).
Most T6SS immunity proteins of known structure prevent intoxication of self and
kin by direct steric occlusion of the effector active site (30, 52, 53), although a subset
of immunity proteins also counteract effector-mediated intoxication though enzymatic
activity (30). In contrast, Tdi proteins in Bacteroidales induce a large conformational
change in cognate effectors, splitting the globular fold into subdomains and structur
ally disrupting the substrate binding and active sites. Possible mechanisms include an
inherent conformational flexibility in Tde1 with selection of a two-subdomain confor
mation for immunity interaction, or an induced fit model of interaction where initial
contacts with Tdi promote separation of the two Tdetox subdomains. One possible
consequence of the structural rearrangement induced in Tde could be increased
efficiency of toxin destruction in the immune recipient cell. For example, Tdi insertion
into the central core of Tde may facilitate proteolytic degradation of the effector.
Several parallels can be drawn between Tde/Tdi and colicin nuclease and immun
ity complexes. For example, colicins E3 and E9 engage immunity proteins at an
“exosite” separate from the active site (54). The mechanism of immunity in these
scenarios is thought to be steric and electrostatic repulsion of substrates (genomic
DNA or the ribosome) (42, 55), in contrast to central core insertion and structural
rearrangement of the active site seen in Tde/Tdi. Colicin nuclease immunity proteins
are structurally diverse, and a prevailing hypothesis is that exosite interactions allow for
evolutionary diversification at the interface, away from the conserved active site (56).
Prevalent cross-reactivity of nuclease colicins and immunity proteins (55) also parallels
the multi-effector interaction patterns of Tdi immunity proteins. The relatively broad
specificity of Tdi immunity interactions with the central core of Tde may have evolved
through exosite diversification as posited for colicin nuclease–immunity interactions.
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 11
Research Article
mBio
Promiscuous binding of multiple Tde by a single Tdi may be more advantageous to
recipient bacteria than highly specific Tde-directed interaction (i.e., 1:1 correspondence),
and may contribute to the high frequency of orphan Tdi in human commensal genome
collections.
As a class of T6SS effector–immunity pairs important for competition among
Bacteroidota, Tde nucleases are neutralized by unique mechanisms, including structural
disruption of the active site and substrate binding surface by an immunity-induced large
conformational shift. This novel immunity mechanism allows relatively broad neutraliza
tion of multiple Ntox15 domains by a single immunity protein. Further study will be
required to determine how Tde and Tdi influence Bacteroidales abudance in IBD and the
detailed mechanisms by which Tdi insert into the central core of Tde.
MATERIALS AND METHODS
T6SS gene quantitation in human intestinal metagenomes
See supplementary methods for detailed methods.
Cloning, plasmids, and Bacteroidales genetics
See supplementary methods for detailed methods.
Competitive growth
Bacteroidales were mixed to a final OD600 reading of 6.0 with 1:1 or 10:1 donor/recipient
ratios and plated on BHIS with gentamycin (60 mg/mL) (57) for ~24 h at 37°C in an
anerobic chamber (Anaerobe Systems, Morgan Hill, CA, USA). Bacteria were recovered
in BHIS liquid media, serially diluted, and quantitatively cultured with and without
5-fluorodeoxyuridine selection. Recipient competitive indices were calculated from
colony-forming units as (post-competition recipient/pre-competition recipient)/(post-
competition donor/pre-competition donor). For competitive growth experiments with
transposon-inserted immunity proteins, expression was induced (or mock in empty
transposon controls) with anhydrotetracycline for 3 h prior to co-culture with cell–
cell contact inducing conditions as above. All competitive growth experiments were
performed with at least biological triplicates and at least two independently replicated
experiments.
Protein purification, crystallization, and structure determination
See supplementary methods for protein purification and crystallization methods. See
Table S1 for diffraction data and refinement statistics.
Differential scanning fluorimetry
Tde1tox H279A/D282A, Tdi1, or the Tde1tox/Tdi1 complex were mixed at 10 µM concentra
tion with SYPRO Orange dye at 2× concentration in X1 buffer. Temperature was increased
at 0.5°C intervals every 10 s in a CFX real-time PCR detection instrument (BioRAD) with
detection of dye fluorescence. Melting temperatures were assigned at the fluorescence
curve inflection point. All data shown represent at least triplicate experiments.
Biolayer interferometry
BLI experiments were conducted on an Octet Red96 instrument (Sartorius). Nucleic
acid binding experiments were conducted with 30 base pair biotinylated synthetic
oligonucleotides, immobilized on streptavidin biosensors. For Ntox15/immunity binding
experiments, hexahistidine immunity proteins (5 mg/mL) were immobilized on NTA
biosensors. Equilibrium binding dose–response curves were generated with varying
concentrations of Tde1tox H279A/D282A, and additional mutations thereof, in Octet
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 12
Research Article
mBio
kinetics buffer (Sartorius). Association and dissociation intervals were 300 and 600 s,
respectively. Affinity constants were determined by one site binding curve fitting of
equilibrium binding data in Prism (GraphPad) after subtraction of non-specific binding
to an irrelevant surface control (biotin only). All data shown represent at least triplicate
experiments.
Nuclease activity
Plasmid DNA (2 µg of pcDNA3.1) was incubated at 37°C with Tde1tox or H279A mutant (1
µM), immunity protein (10 µM), EDTA (1 mM), and/or divalent cation and chloride salts
(10 mM) as indicated in a final volume of 50 µL. Reactions were halted by addition of
DNA electrophoresis loading dye, and nucleic acids assessed by 1% agarose electropho
resis and ethidium bromide staining.
Identification of T6SS, Ntox15, and immunity homologs
See supplementary methods for detailed methods.
Structural analysis and solvation energy calculations
Polymorphic toxin and immunity protein structures were identified in the PDB using
keyword searches and protein classification terms. Comparative homology models of
Tde1tox with TdioA or TdioB were constructed with SWISS-MODEL using the Tde1tox/Tdi1
crystal structure template (44). All structures were reviewed manually in Chimera (58)
to identify effector–immunity interfaces and classify immunity mechanism. Effector and
immunity solvation energy gain calculations were performed with PDBePISA (https://
www.ebi.ac.uk/pdbe/pisa/) (41).
ACKNOWLEDGMENTS
We thank Dr. Eric Pamer and Emily Waligurski at the Duchossois Family Institute for
access to an intestinal commensal bacteria strain collection, Dr. Ben Ross for an insightful
critique of the manuscript, and the University of Iowa Protein and Crystallography Core
for access to BLI instrumentation.
This work was supported by the NIH, K08 AI159619 (Bosch DE). This work was
supported by the NIH (AI080609 to JDM, etc.). J.D.M. is an HHMI Investigator and is
supported by the Lynn M. and Michael D. Garvey Endowed Chair. The Berkeley Center
for Structural Biology is supported in part by the Howard Hughes Medical Institute.
The Advanced Light Source is a Department of Energy Office of Science User Facility
under contract DEAC02-05CH11231. The Pilatus detector on 5.0.1. was funded under NIH
grant S10OD021832. The ALS-ENABLE beamlines are supported in part by the National
Institutes of Health, National Institute of General Medical Sciences, grant P30 GM124169.
D.E.B. – conceptualization, methodology, investigation, resources,data curation,
writing, visualization, supervision, funding acquisition; R.A. – investigation, writing; B.P. –
investigation, writing; S.B.P. – conceptualization, writing, supervision; J.D.M. – conceptu
alization, resources, writing, supervision, funding acquisition.
All authors declare no competing interests.
AUTHOR AFFILIATIONS
1Department of Pathology, Carver College of Medicine, University of Iowa, Iowa City,
Iowa, USA
2Department of Microbiology, University of Washington School of Medicine, Seattle,
Washington, USA
3Howard Hughes Medical Institute, University of Washington, Seattle, Washington, USA
4Microbial Interactions and Microbiome Center, University of Washington, Seattle,
Washington, USA
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 13
mBio
Research Article
AUTHOR ORCIDs
Dustin E. Bosch
http://orcid.org/0000-0002-7430-2939
FUNDING
Funder
HHS | NIH | National Institute of Allergy and
Infectious Diseases (NIAID)
HHS | NIH | National Institute of Allergy and
Infectious Diseases (NIAID)
AUTHOR CONTRIBUTIONS
Grant(s)
Author(s)
K08 AI159619 Dustin E. Bosch
AI080609
Joseph D. Mougous
Dustin E. Bosch, Conceptualization, Investigation, Supervision, Writing – original draft,
Writing – review and editing, Data curation, Methodology, Resources, Visualization |
Romina Abbasian, Investigation, Writing – original draft, Writing – review and editing |
Bishal Parajuli, Investigation, Writing – original draft, Writing – review and editing | S.
Brook Peterson, Conceptualization, Supervision, Writing – original draft, Writing – review
and editing | Joseph D. Mougous, Conceptualization, Funding acquisition, Resources,
Supervision, Writing – original draft, Writing – review and editing
DIRECT CONTRIBUTION
This article is a direct contribution from Joseph Mougous, a Fellow of the American
Academy of Microbiology, who arranged for and secured reviews by Arne Rietsch, Case
Western Reserve University, and Eric Cascales, Centre national de la recherche scientifi-
que, Aix-Marseille Université.
DATA AVAILABILITY
Crystallographic data have been deposited to the RCSB protein data bank (accessions
8FZY, 8FZZ, and 8G0K). Metagenomic sequencing data were previously published (32)
and are publicly available at the NCBI sequence read archive (BioProject PRJNA398089).
Plasmids and bacterial strains generated in the study are listed in Table S3 and will be
available upon reasonable request to the corresponding author.
ADDITIONAL FILES
The following material is available online.
Supplemental Material
Supplemental Material (mBio01039-23-s0001.pdf). Supplemental methods, Figures
S1-S6, and Tables S1-S3.
REFERENCES
1.
2.
3.
Turnbaugh PJ, Hamady M, Yatsunenko T, Cantarel BL, Duncan A, Ley RE,
Sogin ML, Jones WJ, Roe BA, Affourtit JP, Egholm M, Henrissat B, Heath
AC, Knight R, Gordon JI. 2009. A core gut microbiome in obese and lean
twins. Nature 457:480–484. https://doi.org/10.1038/nature07540
Pedersen HK, Gudmundsdottir V, Nielsen HB, Hyotylainen T, Nielsen T,
Jensen BAH, Forslund K, Hildebrand F, Prifti E, Falony G, Le Chatelier E,
Levenez F, Doré J, Mattila I, Plichta DR, Pöhö P, Hellgren LI, Arumugam M,
Sunagawa S, Vieira-Silva S, Jørgensen T, Holm JB, Trošt K, Kristiansen K,
Brix S, Raes J, Wang J, Hansen T, Bork P, Brunak S, Oresic M, Ehrlich SD,
Pedersen O, MetaHIT Consortium. 2016. Human gut microbes impact
host serum metabolome and insulin sensitivity. Nature 535:376–381.
https://doi.org/10.1038/nature18646
Vieira-Silva S, Falony G, Belda E, Nielsen T, Aron-Wisnewsky J, Chakaroun
R, Forslund SK, Assmann K, Valles-Colomer M, Nguyen TTD, Proost S,
4.
5.
Prifti E, Tremaroli V, Pons N, Le Chatelier E, Andreelli F, Bastard J-P, Coelho
LP, Galleron N, Hansen TH, Hulot J-S, Lewinter C, Pedersen HK, Quinquis
B, Rouault C, Roume H, Salem J-E, Søndertoft NB, Touch S, MetaCardis
Consortium, Dumas M-E, Ehrlich SD, Galan P, Gøtze JP, Hansen T, Holst JJ,
Køber L, Letunic I, Nielsen J, Oppert J-M, Stumvoll M, Vestergaard H,
Zucker J-D, Bork P, Pedersen O, Bäckhed F, Clément K, Raes J. 2020. Statin
therapy is associated with lower prevalence of gut Microbiota Dysbiosis.
Nature 581:310–315. https://doi.org/10.1038/s41586-020-2269-x
Stefan KL, Kim MV, Iwasaki A, Kasper DL. 2020. Commensal microbiota
modulation of natural resistance to virus infection. Cell 183:1312–1324.
https://doi.org/10.1016/j.cell.2020.10.047
Dejea CM, Fathi P, Craig JM, Boleij A, Taddese R, Geis AL, Wu X, DeStefano
Shields CE, Hechenbleikner EM, Huso DL, Anders RA, Giardiello FM, Wick
EC, Wang H, Wu S, Pardoll DM, Housseau F, Sears CL. 2018. Patients with
familial adenomatous polyposis harbor colonic biofilms containing
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 14
Research Article
mBio
6.
7.
8.
9.
tumorigenic bacteria. Science 359:592–597. https://doi.org/10.1126/
science.aah3648
Delday M, Mulder I, Logan ET, Grant G. 2019. Bacteroides thetaiotaomi
cron ameliorates colon inflammation in preclinical models of Crohn's
disease. Inflamm Bowel Dis 25:85–96. https://doi.org/10.1093/ibd/izy281
Gonzalez CG, Mills RH, Zhu Q, Sauceda C, Knight R, Dulai PS, Gonzalez DJ.
2022. Location-specific signatures of Crohn's disease at a multi-Omics
scale. Microbiome 10:133. https://doi.org/10.1186/s40168-022-01331-x
Hellmann J, Ta A, Ollberding NJ, Bezold R, Lake K, Jackson K, Dirksing K,
Bonkowski E, Haslam DB, Denson LA. 2023. Patient-reported outcomes
independent of
correlate with microbial community composition
Mucosal inflammation in pediatric inflammatory bowel disease. Inflamm
Bowel Dis 29:286–296. https://doi.org/10.1093/ibd/izac175
Lloyd-Price J, Arze C, Ananthakrishnan AN, Schirmer M, Avila-Pacheco J,
Poon TW, Andrews E, Ajami NJ, Bonham KS, Brislawn CJ, Casero D,
Courtney H, Gonzalez A, Graeber TG, Hall AB, Lake K, Landers CJ, Mallick
H, Plichta DR, Prasad M, Rahnavard G, Sauk J, Shungin D, Vázquez-Baeza
Y, White RA, IBDMDB Investigators, Braun J, Denson LA, Jansson JK,
Knight R, Kugathasan S, McGovern DPB, Petrosino JF, Stappenbeck TS,
Winter HS, Clish CB, Franzosa EA, Vlamakis H, Xavier RJ, Huttenhower C.
2019. Multi-Omics of the gut microbial Ecosystem in inflammatory
bowel diseases. Nature 569:655–662. https://doi.org/10.1038/s41586-
019-1237-9
12.
11.
10. Hecht AL, Casterline BW, Earley ZM, Goo YA, Goodlett DR, Bubeck
Wardenburg J. 2016. Strain competition restricts colonization of an
enteric pathogen and prevents colitis. EMBO Rep 17:1281–1291. https://
doi.org/10.15252/embr.201642282
Verster AJ, Ross BD, Radey MC, Bao Y, Goodman AL, Mougous JD,
Borenstein E. 2017. The landscape of type VI secretion across human gut
microbiomes reveals its role in community composition. Cell Host
Microbe 22:411–419. https://doi.org/10.1016/j.chom.2017.08.010
Chatzidaki-Livanis M, Geva-Zatorsky N, Comstock LE. 2016. Bacteroides
fragilis type VI secretion systems use novel effector and immunity
proteins to antagonize human gut Bacteroidales species. Proc Natl Acad
Sci U S A 113:3627–3632. https://doi.org/10.1073/pnas.1522510113
Cherrak Y, Flaugnatti N, Durand E, Journet L, Cascales E. 2019. Structure
and activity of the type VI secretion system. Microbiol Spectr 7:1–11.
https://doi.org/10.1128/microbiolspec.PSIB-0031-2019
Russell AB, Peterson SB, Mougous JD. 2014. Type VI secretion system
effectors: poisons with a purpose. Nat Rev Microbiol 12:137–148. https://
doi.org/10.1038/nrmicro3185
Coyne MJ, Roelofs KG, Comstock LE. 2016. Type VI secretion systems of
human gut Bacteroidales segregate into three genetic architectures, two
of which are contained on mobile genetic elements. BMC Genomics
17:58. https://doi.org/10.1186/s12864-016-2377-z
14.
13.
15.
16. García-Bayona L, Coyne MJ, Comstock LE. 2021. Mobile type VI secretion
system Loci of the gut Bacteroidales display extensive intra-Ecosystem
transfer, multi-species spread and geographical clustering. PLoS Genet
17:e1009541. https://doi.org/10.1371/journal.pgen.1009541
Coyne MJ, Zitomersky NL, McGuire AM, Earl AM, Comstock LE. 2014.
Evidence of extensive DNA transfer between Bacteroidales species
within the human gut. mBio 5:e01305–14. https://doi.org/10.1128/mBio.
01305-14
17.
18. Wexler AG, Bao Y, Whitney JC, Bobay LM, Xavier JB, Schofield WB, Barry
NA, Russell AB, Tran BQ, Goo YA, Goodlett DR, Ochman H, Mougous JD,
Goodman AL. 2016. Human symbionts inject and neutralize antibacterial
toxins to persist in the gut. Proc Natl Acad Sci U S A 113:3639–3644.
https://doi.org/10.1073/pnas.1525637113
Silverman JM, Agnello DM, Zheng H, Andrews BT, Li M, Catalano CE,
Gonen T, Mougous JD. 2013. Haemolysin coregulated protein is an
exported receptor and chaperone of type VI secretion substrates. Mol
Cell 51:584–593. https://doi.org/10.1016/j.molcel.2013.07.025
19.
20. Günther P, Quentin D, Ahmad S, Sachar K, Gatsogiannis C, Whitney JC,
Raunser S, Satchell KJF. 2022. Structure of a bacterial Rhs effector
exported by the type VI secretion system. PLoS Pathog 18:e1010182.
https://doi.org/10.1371/journal.ppat.1010182
21. Ma J, Pan Z, Huang J, Sun M, Lu C, Yao H. 2017. The Hcp proteins fused
with diverse extended-toxin domains represent a novel pattern of
Antibacterial Effectors in type VI secretion systems. Virulence 8:1189–
1202. https://doi.org/10.1080/21505594.2017.1279374
22. Wood TE, Howard SA, Förster A, Nolan LM, Manoli E, Bullen NP, Yau HCL,
Hachani A, Hayward RD, Whitney JC, Vollmer W, Freemont PS, Filloux A.
2019. The Pseudomonas aeruginosa T6SS delivers a periplasmic toxin
that disrupts bacterial cell morphology. Cell Rep 29:187–201. https://doi.
org/10.1016/j.celrep.2019.08.094
23. Ma L-S, Hachani A, Lin J-S, Filloux A, Lai E-M. 2014. Agrobacterium
tumefaciens Deploys a Superfamily of type VI secretion Dnase Effectors
as weapons for Interbacterial competition in Planta. Cell Host Microbe
16:94–104. https://doi.org/10.1016/j.chom.2014.06.002
24. Wettstadt S, Lai E-M, Filloux A. 2020. Solving the puzzle: Connecting a
heterologous Agrobacterium Tumefaciens T6Ss Effector to a Pseudomo
nas Aeruginosa spike complex. Front Cell Infect Microbiol 10:291. https://
doi.org/10.3389/fcimb.2020.00291
Bondage DD, Lin J-S, Ma L-S, Kuo C-H, Lai E-M. 2016. Vgrg C terminus
confers the type VI Effector transport specificity and is required for
binding with PAAR and Adaptor-Effector complex. Proc Natl Acad Sci U S
A 113:E3931–40. https://doi.org/10.1073/pnas.1600428113
25.
26. Wu C-F, Lien Y-W, Bondage D, Lin J-S, Pilhofer M, Shih Y-L, Chang JH, Lai
E-M. 2020. Effector loading onto the Vgrg carrier activates type VI
secretion system assembly. EMBO Rep 21:e47961. https://doi.org/10.
15252/embr.201947961
Ross BD, Verster AJ, Radey MC, Schmidtke DT, Pope CE, Hoffman LR,
Hajjar AM, Peterson SB, Borenstein E, Mougous JD. 2019. Human gut
bacteria contain acquired
interbacterial defence systems. Nature
575:224–228. https://doi.org/10.1038/s41586-019-1708-z
27.
28. Mok BY, de Moraes MH, Zeng J, Bosch DE, Kotrys AV, Raguram A, Hsu F,
Radey MC, Peterson SB, Mootha VK, Mougous JD, Liu DR. 2020. A
bacterial cytidine deaminase toxin enables CRISPR-free mitochondrial
base editing. Nature 583:631–637. https://doi.org/10.1038/s41586-020-
2477-4
32.
30.
31.
toxin
29. Hespanhol JT, Sanchez-Limache DE, Nicastro GG, Mead L, Llontop EE,
Chagas-Santos G, Farah CS, de Souza RF, Galhardo R da S, Lovering AL,
Bayer-Santos E. 2022. Antibacterial T6Ss Effectors with a VRR-Nuc
domain are structure-specific Nucleases. Elife 11:e82437. https://doi.org/
10.7554/eLife.82437
Ting S-Y, Bosch DE, Mangiameli SM, Radey MC, Huang S, Park Y-J, Kelly
KA, Filip SK, Goo YA, Eng JK, Allaire M, Veesler D, Wiggins PA, Peterson SB,
Mougous JD. 2018. Bifunctional immunity proteins protect bacteria
against ftsz-targeting ADP-ribosylating toxins. Cell 175:1380–1392.
https://doi.org/10.1016/j.cell.2018.09.037
Zhang D, de Souza RF, Anantharaman V, Iyer LM, Aravind L. 2012.
Polymorphic
systems: comprehensive characterization of
trafficking modes, processing, mechanisms of action, immunity and
Ecology using comparative Genomics. Biol Direct 7:18. https://doi.org/
10.1186/1745-6150-7-18
Integrative H. 2014. The
integrative human microbiome project:
dynamic analysis of microbiome-host omics profiles during periods of
human health and disease. Cell Host Microbe 16:276–289. https://doi.
org/10.1016/j.chom.2014.08.014
Potter SC, Luciani A, Eddy SR, Park Y, Lopez R, Finn RD. 2018. HMMER web
server: 2018 update. Nucleic Acids Res 46:W200–W204. https://doi.org/
10.1093/nar/gky448
Sorbara MT, Littmann ER, Fontana E, Moody TU, Kohout CE, Gjonbalaj M,
Eaton V, Seok R, Leiner IM, Pamer EG. 2020. Functional and genomic
variation between human-derived isolates of Lachnospiraceae reveals
inter- and intra-species diversity. Cell Host Microbe 28:134–146. https://
doi.org/10.1016/j.chom.2020.05.005
Jurėnas D, Rosa LT, Rey M, Chamot-Rooke J, Fronzes R, Cascales E. 2021.
Mounting, structure and Autocleavage of a type VI secretion-associated
Rhs polymorphic toxin. Nat Commun 12:6998. https://doi.org/10.1038/
s41467-021-27388-0
33.
34.
35.
36. Mirdita M, Schütze K, Moriwaki Y, Heo L, Ovchinnikov S, Steinegger M.
2021. Colabfold-making protein folding accessible to all. biorxiv:1–6.
https://doi.org/10.1101/2021.08.15.456425:2021.08.15.456425
37. Holm L. 2020. Using dali for protein structure comparison. Methods Mol
Biol 2112:29–42. https://doi.org/10.1007/978-1-0716-0270-6_3
38. Qiu J, Bernhofer M, Heinzinger M, Kemper S, Norambuena T, Melo F, Rost
B. 2020. ProNA2020 predicts protein-DNA, protein-RNA, and protein-
protein binding proteins and residues from sequence. J Mol Biol
432:2428–2443. https://doi.org/10.1016/j.jmb.2020.02.026
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 15
Research Article
mBio
39.
40.
41.
42.
43.
Camacho C, Coulouris G, Avagyan V, Ma N, Papadopoulos J, Bealer K,
Madden TL. 2009. BLAST+: architecture and applications. BMC
Bioinformatics 10:421. https://doi.org/10.1186/1471-2105-10-421
Forster SC, Kumar N, Anonye BO, Almeida A, Viciani E, Stares MD, Dunn
M, Mkandawire TT, Zhu A, Shao Y, Pike LJ, Louie T, Browne HP, Mitchell
AL, Neville BA, Finn RD, Lawley TD. 2019. A human gut bacterial genome
and culture collection for
improved metagenomic analyses. Nat
Biotechnol 37:186–192. https://doi.org/10.1038/s41587-018-0009-7
Krissinel E, Henrick K. 2007. Inference of macromolecular assemblies
from crystalline state. J Mol Biol 372:774–797. https://doi.org/10.1016/j.
jmb.2007.05.022
Kleanthous C, Walker D. 2001. Immunity proteins: enzyme inhibitors that
avoid the active site. Trends Biochem Sci 26:624–631. https://doi.org/10.
1016/s0968-0004(01)01941-7
Yin R, Feng BY, Varshney A, Pierce BG. 2022. Benchmarking Alphafold for
protein complex modeling reveals accuracy determinants. Protein Sci
31:e4379. https://doi.org/10.1002/pro.4379
44. Waterhouse A, Bertoni M, Bienert S, Studer G, Tauriello G, Gumienny R,
Heer FT, de Beer TAP, Rempfer C, Bordoli L, Lepore R, Schwede T. 2018.
SWISS-MODEL: homology modelling of protein structures and
complexes. Nucleic Acids Res 46:W296–W303. https://doi.org/10.1093/
nar/gky427
45. Ahmad S, Wang B, Walker MD, Tran H-KR, Stogios PJ, Savchenko A, Grant
RA, McArthur AG, Laub MT, Whitney JC. 2019. An interbacterial toxin
inhibits target cell growth by synthesizing (P) ppApp. Nature 575:674–
678. https://doi.org/10.1038/s41586-019-1735-9
47.
46. Walker D, Lancaster L, James R, Kleanthous C. 2004. Identification of the
catalytic motif of the microbial ribosome inactivating cytotoxin colicin
E3. Protein Sci 13:1603–1611. https://doi.org/10.1110/ps.04658504
Papadakos G, Wojdyla JA, Kleanthous C. 2012. Nuclease colicins and
their immunity proteins. Q Rev Biophys 45:57–103. https://doi.org/10.
1017/S0033583511000114
Ray A, Schwartz N, de Souza Santos M, Zhang J, Orth K, Salomon D. 2017.
Type VI secretion system MIX-effectors carry both antibacterial and anti-
eukaryotic activities. EMBO Rep 18:1978–1990. https://doi.org/10.15252/
embr.201744226
Brodmann M, Schnider ST, Basler M. 2021. Type VI secretion system and
its Effectors Pdpc, Pdpd, and Opia contribute to Francisella virulence in
48.
49.
50.
Galleria Mellonella larvae. Infect Immun 89:e0057920. https://doi.org/10.
1128/IAI.00579-20
Liu L, Xu M, Lan R, Hu D, Li X, Qiao L, Zhang S, Lin X, Yang J, Ren Z, Xu J.
2022. Bacteroides vulgatus attenuates experimental mice colitis through
modulating gut microbiota and immune responses. Front. Immunol
13:1036196. https://doi.org/10.3389/fimmu.2022.1036196
52.
54.
53.
51. Ali J, Yu M, Sung L-K, Cheung Y-W, Lai E-M. 2023. A glycine Zipper motif
governs translocation of type VI secretion toxic Effectors across the
cytoplasmic membrane of target cells. bioRxiv:1–47. https://doi.org/10.
1101/2022.07.12.499750:2022.07.12.499750
de Moraes MH, Hsu F, Huang D, Bosch DE, Zeng J, Radey MC, Simon N,
Ledvina HE, Frick JP, Wiggins PA, Peterson SB, Mougous JD. 2021. An
Interbacterial DNA Deaminase toxin directly Mutagenizes surviving
target populations. Elife 10:e62967. https://doi.org/10.7554/eLife.62967
Zhang H, Zhang H, Gao Z-Q, Wang W-J, Liu G-F, Xu J-H, Su X-D, Dong Y-H.
2013. Structure of the type VI effector-immunity complex (Tae4-Tai4)
provides novel insights into the inhibition mechanism of the effector by
its immunity protein. J Biol Chem 288:5928–5939. https://doi.org/10.
1074/jbc.M112.434357
Kühlmann UC, Pommer AJ, Moore GR, James R, Kleanthous C. 2000.
Specificity in protein-protein interactions: the structural basis for dual
recognition in endonuclease colicin-immunity protein complexes. J Mol
Biol 301:1163–1178. https://doi.org/10.1006/jmbi.2000.3945
Cascales E, Buchanan SK, Duché D, Kleanthous C, Lloubès R, Postle K,
Riley M, Slatin S, Cavard D. 2007. Colicin biology. Microbiol Mol Biol Rev
71:158–229. https://doi.org/10.1128/MMBR.00036-06
Riley MA. 1998. Molecular mechanisms of bacteriocin evolution. Annu
Rev Genet 32:255–278. https://doi.org/10.1146/annurev.genet.32.1.255
Bacic MK, Smith CJ. 2008. Laboratory maintenance and cultivation of
Bacteroides species. Curr Protoc Microbiol Chapter 13:Unit 13C.1. https:/
/doi.org/10.1002/9780471729259.mc13c01s9
Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng
EC, Ferrin TE. 2004. UCSF Chimera--a visualization system for exploratory
research and analysis. J Comput Chem 25:1605–1612. https://doi.org/10.
1002/jcc.20084
55.
57.
58.
56.
July/August Volume 14 Issue 4
10.1128/mbio.01039-23 16
| null |
10.1038_s41467-022-28771-1.pdf
|
Data availability
Data supporting the results of this study are presented within the article and
supplementary figures. NGS data are available in the NCBI Sequence Read Archive
database (BioProject accession code PRJNA803881). Additional details and data to
support the findings of this study are available from the corresponding authors upon
reasonable request. Source data for Figs. 1a, 2a, c, 3b, d–f, 4, 5b, d, S1, S2, S4, S5a are
provided as Source Data file. Source data are provided with this paper.
|
Data availability Data supporting the results of this study are presented within the article and supplementary figures. NGS data are available in the NCBI Sequence Read Archive database (BioProject accession code PRJNA803881). Additional details and data to support the findings of this study are available from the corresponding authors upon reasonable request. Source data for Figs. 1a, 2a, c, 3b, d-f, 4, 5b, d, S1, S2, S4, S5a are provided as Source Data file. Source data are provided with this paper.
|
ARTICLE
https://doi.org/10.1038/s41467-022-28771-1
OPEN
Harnessing DSB repair to promote efficient
homology-dependent and -independent prime
editing
Martin Peterka1✉
Burcu Bestas
Grzegorz Sienski
1, Jack Barr1, Stijn van de Plassche1, Patricia Mendoza-Garcia
1, Mike Firth3 & Marcello Maresca
, Nina Akrap1,4, Songyuan Li
1, Saša Šviković1,
1✉
1,4, Sandra Wimberger1,2,4, Pei-Pei Hsieh1, Dmitrii Degtev1,
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
Prime editing recently emerged as a next-generation approach for precise genome editing.
Here we exploit DNA double-strand break (DSB) repair to develop two strategies that install
precise genomic insertions using an SpCas9 nuclease-based prime editor (PEn). We first
demonstrate that PEn coupled to a regular prime editing guide RNA (pegRNA) efficiently
promotes short genomic insertions through a homology-dependent DSB repair mechanism.
it can rescue pegRNAs that
While PEn editing leads to increased levels of by-products,
perform poorly with a nickase-based prime editor. We also present a small molecule
approach that yields increased product purity of PEn editing. Next, we develop a homology-
independent PEn editing strategy, which installs genomic insertions at DSBs through the non-
homologous end joining pathway (NHEJ). Lastly, we show that PEn-mediated insertions at
DSBs prevent Cas9-induced large chromosomal deletions and provide evidence that con-
tinuous Cas9-mediated cutting is one of the mechanisms by which Cas9-induced large
deletions arise. Altogether, this work expands the current prime editing toolbox by leveraging
distinct DNA repair mechanisms including NHEJ, which represents the primary pathway of
DSB repair in mammalian cells.
1 Genome Engineering, Discovery Sciences, BioPharmaceuticals R&D Unit, AstraZeneca, Gothenburg, Sweden. 2 Department of Chemistry & Molecular
Biology, University of Gothenburg, Gothenburg, Sweden. 3 Data Sciences and Quantitative Biology, Discovery Sciences, AstraZeneca, Cambridge, UK. 4These
✉
email: [email protected]; [email protected]
authors contributed equally: Nina Akrap, Songyuan Li, Sandra Wimberger.
NATURE COMMUNICATIONS |
(2022) 13:1240 | https://doi.org/10.1038/s41467-022-28771-1 | www.nature.com/naturecommunications
1
ARTICLE
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28771-1
Proper utilization of cellular DNA repair mechanisms is
instrumental to any successful genome editing strategy. The
is highly
activity of different DNA repair pathways
dependent on tissue and cell type, chromatin context, and the
DNA sequence of the target locus1–4.
Targeted DNA insertions represent a particularly challenging
type of precise genome modification but have a considerable
therapeutic potential. About 25% of ClinVar human pathogenic
variants are deletions, the majority of which are <25 bp in length
and thus potentially actionable by prime editing5. A common
approach to introduce DNA insertions is to induce a targeted
DNA double-strand break (DSB) using a site-specific nuclease
combined with the delivery of a donor DNA repair template to
stimulate homology-directed repair (HDR) at the targeted locus.
A major disadvantage of this strategy is the limited activity of
homologous recombination, which is restricted to S/G2 phases of
the cell cycle and is generally absent in postmitotic cells6.
Unlike homologous recombination, DNA end joining repair
mechanisms such as non-homologous end joining (NHEJ) or
alternative end joining (a-EJ) pathways remain active through-
out the cell cycle and act as the major pathways of DSB repair in
mammalian cells7–9. While being typically considered error
prone, NHEJ can repair DSBs with high fidelity10,11. In contrast,
the homology-dependent a-EJ pathway leads to deletions, which
are highly predictable12,13. The respective precision and pre-
dictability of NHEJ and a-EJ have been successfully exploited for
precise genome modifications including DNA insertions14–17.
Harnessing DNA end joining pathways represents a valuable
genome editing strategy, because most adult tissues are com-
prised of postmitotic cells unable to perform homologous
recombination3,18.
The recently developed CRISPR-based prime editing can install
a wide spectrum of genomic modifications including deletions,
substitutions, and insertions without the need of a separate DNA
template and without introducing DSBs5, therefore offering a
major advantage over existing genome editing methods. The PE2
prime editor combines Cas9 (H840A) nickase with an engineered
reverse transcriptase (RT) to install an edit encoded directly in the
prime editing gRNA (pegRNA). The cascade of events leading to
a successful prime editing outcome is comprised of (1) Cas9-
mediated nicking of the target site, (2) hybridization of the
pegRNA-encoded primer binding site (PBS) to the 3’ end of
the nick, (3) pegRNA-templated extension of the primed 3’ end of
the nick by RT resulting in a “flap” containing the desired edit,
and (4) hybridization and ligation of the flap with the targeted
locus. Inhibition of mismatch repair was recently shown to
enhance prime editing efficiency19, but DNA repair mechanisms
responsible for the upstream steps of successful incorporation of
the 3’ flap remain to be described in detail and might not be
universally available in different cellular and genomic contexts,
potentially limiting the scope and efficiency of the nickase-based
PEs. Recent reports suggest a possible dependency of PE2 editing
on cell cycle progression20,21. Thus, a prime editing strategy
harnessing a wider spectrum of DNA repair pathways would be a
valuable addition to the prime editing toolbox.
Here we introduce Prime Editor nuclease (PEn), which com-
bines RT and the wild-type SpCas9 nuclease and show that prime
editing can be performed at DSBs by utilizing DNA end joining
repair pathways. We present two PEn strategies to robustly install
small insertions via distinct DNA repair mechanisms. The first
strategy utilizes regular pegRNAs to promote small DNA inser-
tions by a homology-dependent DSB repair mechanism. This
strategy worked robustly across different genomic loci as well as
with pegRNAs displaying inefficient editing when combined with
PE2. The second strategy relies on a modified sgRNA design to
install small insertions through precise NHEJ. We also present a
small molecule approach to decrease unintended by-products of
PEn editing. Finally, we show that unlike editing with Cas9 alone,
PEn does not induce large unintended on-target deletions, likely
because PEn-mediated insertions at DSBs prevent NHEJ-
mediated restoration of the wild-type sequence at the target
locus. This suggests that the futile cycle of nuclease-mediated cut
and NHEJ-mediated precise repair may be a possible cause of
DSBs genotoxicity associated with Cas9 treatment.
repair mechanism, we
the targeted sites revealed successful
Results
SpCas9 nuclease-based prime editing. To test if a Cas9 nuclease-
based prime editor can install small insertions at DSBs through a
DNA end joining
reverted the
Cas9(H840A)-based PE2 into wtCas9-PE, designated here as PEn.
We have constructed pegRNAs encoding small
insertions of
various sizes (6–18 bp) (Supplementary Data 1) against 10
genomic target sites and co-transfected HEK293T cells with a
pegRNA and either PEn or PE2. NGS analysis of the editing
outcomes at
intended
insertions with varying frequencies and product purities for both
PEn and PE2 (Fig. 1a). We classified the edited alleles into three
categories; (1) all prime edits, representing any type of RT-
templated insertions, (2) precise prime edits, that represent RT-
templated insertions of intended size, and (3) other indels. As
expected, PE2 editing resulted in high product purity, but also
showed large site-to-site variability of insertion efficiency. PEn
editing resulted in variable rates of precise prime edits but in
general higher as compared to PE2. At some tested sites (HEK3,
DPM2, AAVS1, EGFR) PE2 achieved similar or higher editing
efficiency compared to PEn and clearly outperformed PEn in
terms of precise editing purity. However, PEn installed insertions
efficiently even with pegRNAs that were suboptimal for PE2 in
our hands (Fig. 1a, PCSK9, FANCF, TRBC, PDCD1, CTLA4). We
observed a similar trend in HeLa and HCT116 cells, where we
tested PE2-optimal (AAVS1) as well as suboptimal (CTLA4)
pegRNAs (Supplementary Fig. 1). As expected, due to the use of
wild-type Cas9 that cuts both DNA strands, we also observed
variable levels of PEn-induced imprecise prime edits and indels.
Alignments of PEn-edited reads revealed that most of
the
imprecise prime edits represent additional integrations matching
RT templates (Fig. 1b).
Mechanism of PEn-based prime editing. We reasoned that the
integrated RT-templated homology tails might be products of
DSB repair mediated by non-homologous end joining (NHEJ),
while the precise insertions could occur through a homology-
dependent process (Fig. 1c). If true, the inhibition of NHEJ could
shift the outcomes of PEn editing by decreasing the frequency of
imprecise prime edits. To test this, we performed PEn editing in
HEK293T cells treated with AZD7648, a small molecule inhibitor
of DNA-PK, an essential mediator of NHEJ22. Indeed, upon
DNA-PK inhibition, the additional RT template integrations were
abolished, and the remaining prime edits represented almost
exclusively insertions of intended sizes (Fig. 2a, b). At several loci,
DNA-PK inhibitor treatment also led to an increase of total rates
of correct insertions (Fig. 2a – DPM2, AAVS1, EGFR).
responsible for
Having pinpointed the contribution of NHEJ to PEn editing,
we then investigated the mechanisms
the
homology-dependent DSB repair resulting in precise insertions.
We reasoned that the short homology tails used in our pegRNA
designs could utilize the a-EJ pathway, which typically uses DSB-
proximal homology regions ~2–20 bp in length8. To test this
hypothesis, we performed PEn editing using the AAVS1 pegRNA
with a 13 nt homology tail in HEK293T cells deficient in DNA
Polymerase θ (encoded by the POLQ gene), a crucial mediator of
2
NATURE COMMUNICATIONS |
(2022) 13:1240 | https://doi.org/10.1038/s41467-022-28771-1 | www.nature.com/naturecommunications
ARTICLE
prime edits - precise
prime edits - all
indels
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28771-1
)
%
(
t
i
d
e
d
e
t
a
c
d
n
i
i
h
t
i
w
s
d
a
e
r
S
G
N
100
80
60
40
20
0
a
b
HEK3
18 bp ins.
DPM2
11 bp ins.
50
40
30
20
10
0
**
PEn
PE2
ns
PEn
PE2
40
30
20
10
0
AAVS1
6 bp ins.
ns
PEn
PE2
EGFR
9 bp ins.
ns
PEn
PE2
80
60
40
20
0
20
15
10
5
0
TRAC
8 bp ins.
*
PEn
PE2
PCSK9
11 bp ins.
**
FANCF
3 bp ins.
60
ns
40
20
0
PEn
PE2
PEn
PE2
25
20
15
10
5
0
40
30
20
10
0
TRBC
8 bp ins.
**
PEn
PE2
40
30
20
10
0
PDCD1
12 bp ins.
**
PEn
PE2
CTLA4
8 bp ins.
***
PEn
PE2
50
40
30
20
10
0
RT template
PBS
A C T G T G G G G C A T C T T T G G A G G G G A C A G A
PEn
AAVS1
wild type
T C C C T A G T G G C C C C A C T G T G G G G T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
prime edits
T C C C T A G T G G C C C C A C T G T G G G G C A T C T T T G G A G G G G A C A G A T A A A A G T A C C C
T C C C T A G T G G C C C C A C T G T G G G G C A T C T T T G G A G G G G A C A G A G C T G G A G G G G A
T C C C T A G T G G C C C C A C T G T G G G G C A T C T T T G G A G G G G A C A G A G C A T G G A G G G G
T C C C T A G T G G C C C C A C T G T G G G G C A T C T T T G G A G G G G A C T G G A G G G G A C A G A T
T C C C T A G T G G C C C C A C T G T G G G G C A T C T T T G G A G G G G A C A G A G T G G A G G G G A C
T C C C T A G T G G C C C C A C T G T G G G G C A T C T T T T G G A G G G G A C A G A T A A A A G T A C C
T C C C T A G T G G C C C C A C T G T G G G G C A T C T T T G G A G G G G A C A G A T G G A G G G G A C A
T C C C T A G T G G C C C C A C T G T G G G G C A T C T T T G T G G A G G G G A C A G A T A A A A G T A C
T C C C T A G T G G C C C C A C T G T G G G G C A T C T T T G G A G G G G A T G G A G G G G A C A G A T A
T C C C T A G T G G C C C C A C T G T G G G G C A T C T T T G G A G G G G A C A T G G A G G G G A C A G A
- T G G A G G G G A C A G A T A A A A G T A C C C
T C C C T A G T G G C C C C A C T G T G G G G C A T -
indels
T C C C T A G T G G C C C C A C T G T G G G G C T G G A G G G G A C A G A T A A A A G T A C C C A G A A C
T C C C T A G T G G C C C C A C T G T G G -
- A G G G G A C A G A T A A A A G T A C C C A G A A C C
-
- T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
T C C C T A G T G G C C C C A C T G T G G -
- T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
T C C C T A G T G G C C C C A C T G T G -
-
T C C C T A G T G G C C C C A C T G T G G G - T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
T C C C T A G T G G C C C C A C T -
- T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
-
-
T C C C T A G T G G C C C C A C T G -
- G A G G G G A C A G A T A A A A G T A C C C A G A A C C
T C C C T A G T G G C C C C A C T G T G G G G A T G G A G G G G A C A G A T A A A A G T A C C C A G A A C
T C C C T A G T G G C C C C A C T G T -
- T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
A C C C T A G T G G C C C C A C T G T G G G G T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
-
-
-
-
-
-
-
-
-
-
-
-
c
intended insert
insertions
- deletions
cleavage position
substitutions
insert
homology
homology-dependent EJ
NHEJ
59.85%
10.98%
7.67%
1.00%
0.92%
0.58%
0.48%
0.47%
0.46%
0.42%
0.40%
0.38%
0.48%
0.38%
0.36%
0.33%
0.20%
0.16%
0.15%
0.14%
0.12%
0.11%
Fig. 1 SpCas9 nuclease-based prime editing. a NGS analysis of PEn or PE2-mediated targeted DNA insertions of indicated sizes using 10 different
pegRNAs targeting endogenous loci in HEK293T cells. Plots show mean ± SD of n = 3 biologically independent replicates. “prime edits – all” and “prime
edits – precise” categories are superimposed. P-values were determined using Student’s paired t test (two-tailed) *P < 0.05, **P < 0.01, ***P < 0.001.
Calculated P values: HEK3 = 0.0050, DPM2 = 0.3075, AAVS1 = 0.2569, EGFR = 0.0732, TRAC = 0.0433, PCSK9 = 0.0028, FANCF = 0.0733,
TRBC = 0.0021, PDCD1 = 0.0055, CTLA4 = 0.0003. b Representative alignment and allele frequencies of AAVS1 locus edited with PEn and the indicated RT
template in HEK293T cells. For each category, the top 10 variants are shown with a minimum frequency of 0.1%. c Model of homology-dependent and
NHEJ modes of PEn-mediated insertions at DSBs. Source data for Fig. 1a are provided as a Source Data file.
a-EJ8. First, to confirm a-EJ inhibition in POLQ-/- background,
we performed Cas9 editing of the AAVS1 locus in these cells with
or without DNA-PK inhibition. As expected, no indels were
treatment of
the targeted site upon DNA-PKi
detected at
POLQ-/- cells (Supplementary Fig. 2), suggesting both NHEJ
and e-EJ pathways were disabled. Despite this, PEn-mediated
editing still proceeded efficiently,
suggesting a mechanism
independent of a-EJ (Fig. 2c).
Altogether, our data reveal that the imprecise PEn prime edits
are mediated by NHEJ. Accordingly, DNA-PK inhibition improves
the purity of PEn editing and leads to increased efficiency in a
locus-dependent
Interestingly, PEn-mediated precise
insertions appear to be independent of the Pol θ-mediated a-EJ
pathway.
fashion.
PEn editing through NHEJ. The observation that PEn-mediated
imprecise edits were inserted via NHEJ prompted us to test whether a
pegRNA design encoding the intended insertion, but no homology
tail, could still perform precise primed insertions through NHEJ-
mediated integration (Fig. 3a). This strategy could only be exploited
to promote insertions at the cleavage site due to the end-to-end
joining mechanism. To test this PRimed INSertions strategy (PRINS),
we removed the homology region from the RT template of AAVS1
pegRNA, resulting in a gRNA design with an extension containing
only PBS and an intended insertion (Single PRimed INsertion gRNA,
springRNA). PRINS editing of AAVS1 using springRNA was able to
install the intended insertion in HEK293T cells and was completely
abrogated by DNA-PK inhibition, confirming that NHEJ is respon-
sible for the PRINS-mediated insertions (Fig. 3b). The imprecise
insertions constituted either truncated inserts or inserts longer than
the intended size due to integrations of the gRNA scaffold sequence
of various lengths (Fig. 3c). We have further tested this approach
using a panel of springRNAs against different targets in HeLa and
HCT116 achieving variable but robust editing with up to 50% effi-
ciencies (Fig. 3d, e). The unintended edits were sometimes prevalent
such as in the case of CTLA4, where the top variant contained
additional scaffold sequence (Fig. 3e and Supplementary Fig. 3). We
have also tested all four 1 nt insertions at the AAVS1 site and
observed variable ratios of intended/scaffold-containing editing pro-
ducts, suggesting an effect of RT template and targeted DNA
sequences on PRINS outcomes (Fig. 3d). Altogether, our results
demonstrate that PEn can efficiently install precise insertions through
NHEJ.
Off-target analysis of PEn editing. Integration of short double-
stranded DNA fragments at DSBs has been exploited for Cas9
off-target detection23 and integration of single-stranded DNA
fragments was shown to increase both on- and off-target editing
by Cas924. Based on these studies, we reasoned that PEn might
also show more pronounced off-target editing by actively mod-
ifying DSBs and in doing so preventing error-free DNA repair. To
investigate PEn-mediated off-target editing, we have targeted
three sites (FANCF, HEK3 and HEK4) with gRNAs that were
previously profiled for off-target editing with both Cas9 and
PE25,23. We used PEn and a matching PEn mutant (PEn-dRT)
carrying previously reported RT-disabling mutations5 with either
pegRNAs
in
HEK293T cells and analyzed both on-target editing and a total of
11 off-target sites by deep amplicon sequencing (Fig. 4). Com-
pared to PEn-dRT, PEn induced up to 2-fold higher total on-
target editing. Similarly, we observed that PEn increased off-
target editing across most target sites. The increase ranged from
moderate at HEK4 off-targets (1.4–2.3-fold), to high at FANCF
with off-target 1 reaching up to a 13-fold increase (Fig. 4).
Examination of editing outcomes at these sites revealed that the
increase was caused by RT-mediated insertions that constituted
springRNAs
against
targets
three
the
or
NATURE COMMUNICATIONS |
(2022) 13:1240 | https://doi.org/10.1038/s41467-022-28771-1 | www.nature.com/naturecommunications
3
ARTICLE
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28771-1
DPM2
AAVS1
***
ns
PEn PEn+i PE2
40
30
20
10
0
*
*
PEn PEn+i PE2
TRAC
ns
*
PEn PEn+i PE2
20
15
10
5
0
25
20
15
10
5
0
PCSK9
*
*
PEn PEn+i PE2
TRBC
ns
*
PEn PEn+i PE2
40
30
20
10
0
50
40
30
20
10
0
CTLA4
ns
**
PEn PEn+i PE2
PDCD1
**
*
PEn PEn+i PE2
40
30
20
10
0
80
60
40
20
0
EGFR
*
*
PEn PEn+i PE2
40%
50%
90%
70%
40%
60%
70%
30%
no indel
indel
AAVS1_Pool3_pAAVS1_ins6_9_13_PE0_1_1
−10 0
10
20
30
−15−10−5 0 5 10 15
−15 −10 −5 0
5 10 15
−10
0
10
20
−10 0
10 20 30 40
−10
0
10
20
−10 0
10
20
30
−15−10−5 0 5 10 15
100%
100%
100%
100%
100%
100%
100%
100%
a
t
i
d
e
d
e
t
a
c
d
n
i
i
h
t
i
w
s
d
a
e
r
S
G
N
f
o
%
%
s
e
c
n
e
u
q
e
S
50
40
30
20
10
0
O
S
M
D
i
-
K
P
A
N
D
−15 −10 −5 0
5 10 15
−15 −10 −5 0
5 10 15
−15 −10 −5 0
5 10 15
−15 −10 −5 0
5 10 15
−15−10 −5 0
5 10 15
−15 −10 −5 0
5 10 15
−15 −10 −5 0
5 10 15
−15−10 −5 0
5 10 15
size distribution of PEn-mediated prime edits (bp)
b
PEn + DNA-PKi
AAVS1
PBS
A
RT template
C T G T G G G G C A T C T T T G G A G G G G A C A G A
wild type
C C C A C T G T G G G G T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
prime edits
C C C A C T G T G G G G C A T C T T T G G A G G G G A C A G A T A A A A G T A C C C
C C C A C T G T G G G G C A T C T T T G G A G G G G A C A G A G C T G G A G G G G A
C C C A C T G T G G G G C A T C T T T G G A G G G G A C A G A -
-
C C C A C T G T G G G G C A T C T T T G G A G G G G A C A G A G A A A A G T A C C C
-
-
-
-
-
-
-
-
-
-
-
-
-
C C C A C T G T G G -
C C C A C T G T G G G G -
-
-
-
-
-
C C C A C T G T G G G -
-
C C C A C T G -
-
C C C A C T G T G G G G -
-
-
-
-
-
-
-
-
-
-
-
indels
- A G G G G A C A G A T A A A A G T A C C C A G A A C C
-
-
-
- A C A G A T A A A A G T A C C C A G A A C C
-
- G G A C A G A T A A A A G T A C C C A G A A C C
-
-
-
-
- A C A G A T A A A A G T A C C C A G A A C C
-
- G A G G G G A C A G A T A A A A G T A C C C A G A A C C
-
- C C A G A T A A A A G T A C C C A G A A C C
-
-
-
-
-
-
-
-
-
-
-
-
intended insert
insertions
- deletions
cleavage position
substitutions
59.43%
29.48%
0.33%
0.27%
0.13%
0.59%
0.44%
0.20%
0.17%
0.12%
0.12%
c
)
%
(
t
i
d
e
d
e
t
a
c
d
n
i
i
h
t
i
w
s
d
a
e
r
S
G
N
20
15
10
5
0
PEn
AAVS1
6 bp ins.
prime edits - precise
prime edits - all
indels
DNA-PKi:
-
WT
-
POLQ-/-
+
WT
+
POLQ-/-
Fig. 2 NHEJ mediates imprecise PEn editing. a Selected data from Fig. 1a with additional DNA-PK inhibitor treatments of PEn samples. Plots represent PEn or
PE2 editing using 8 different pegRNAs targeting endogenous loci in HEK293T. PEn edited cells were additionally pre-treated with DNA-PK inhibitor (PEn+i) or
DMSO (PEn). Plots show mean ± SD of n = 3 biologically independent replicates. “prime edits – all” and “prime edits – precise” categories are superimposed.
Histograms below the bar plots represent percentages and size distribution of PEn prime edited alleles for each target site with or without DNA-PK inhibition.
P-values were determined using Student’s paired t test (two-tailed) *P < 0.05, **P < 0.01, ***P < 0.001. Calculated P values: PEn vs. PEn+i (DPM2 = 0.00001,
AAVS1 = 0.0273, TRAC = 0.0536, PCSK9 = 0.0168, TRBC = 0.1189, CTLA4 = 0.1029, PDCD1 = 0.0091, EGFR = 0.0282), PEn+i vs. PE2 (DPM2 = 0.3229,
AAVS1 = 0.0222, TRAC = 0.0467, PCSK9 = 0.0149, TRBC = 0.0121, CTLA4 = 0.0020, PDCD1 = 0.0202, EGFR = 0.0342). b Representative alignment and allele
frequencies of AAVS1 locus edited with PEn and the indicated RT template in HEK293T cells treated with DNA-PK inhibitor. For each category, the top 10
variants are shown with a minimum frequency of 0.1%. c NGS analysis of PEn editing outcomes of AAVS1 locus in wild-type and POLQ-/- and HEK293T cells
with or without DNA-PK inhibitor treatment. The plot shows mean ± SD of n = 3 biologically independent replicates. Source data for Fig. 2a, c are provided as a
Source Data file.
the majority of edits across PEn-edited sites (Supplementary
these results show that efficient priming can
Fig. 4). Thus,
increase PEn-mediated off-target editing and highlight a need for
stringent peg/springRNAs and/or high fidelity Cas9 enzymes to
be used with PEn.
PEn-mediated insertions at DSBs mitigate Cas9-induced large
deletions. Cas9 editing has been shown to frequently cause large
deletions spanning kilobase-sized regions surrounding the Cas9
target site25,26. This unintended consequence of Cas9 editing poses
a potential roadblock for its therapeutic applications. As PEn gen-
erates DSBs, we wondered whether PEn editing also results in
similar unwanted on-target editing. To test this, we used a diph-
theria toxin (DT)-based selection system in HEK293T cells27 to
assay for large deletions induced by Cas9, PE2, and PEn. In this
system, the disruption of the HBEGF coding sequence generates
cells resistant to DT treatment, while cells carrying an intact copy of
the HBEGF coding sequence are efficiently killed by DT (Fig. 5a).
To monitor large deletions induced by different editors, we targeted
an intron of HBEGF with either Cas9, PE2 or PEn and subjected the
edited cells to DT selection. The percentage of colonies surviving
DT treatment normalized to the total HBEGF editing levels in each
condition can be used to approximate the levels of large deletions in
the cell population, as only cells carrying HBEGF deletions larger
than ~600 bp acquire DT resistance. As expected, Cas9 editing led
to a relatively high frequency of large deletions (Fig. 5b) confirming
previous observations25,26. In contrast, nickase-based PE2 editing
that does not induce DSBs only resulted in basal levels of large
deletions. Surprisingly, similar to PE2, PEn editing with pegRNA or
springRNA led to minimal levels of large deletions compared to
Cas9 editing (Fig. 5b), despite efficient editing at the target site
(Fig. S5a). We have analyzed large deletion patterns using PacBio
long-read DNA sequencing26 of the edited HEBGF locus prior to
DT selection. The alignment of HBEGF long reads confirmed the
presence of large deletions in Cas9-edited sample but not in PE2 or
PEn-edited samples (Fig. 5c).
We hypothesized that Cas9-induced large deletions might be a
result of cyclic targeted DNA cutting by Cas9 after precise DSB
4
NATURE COMMUNICATIONS |
(2022) 13:1240 | https://doi.org/10.1038/s41467-022-28771-1 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28771-1
b
PRINS
HEK293T
AAVS1
6 bp ins.
c
prime edits - precise
prime edits - all
indels
insert
NHEJ
)
%
(
t
i
d
e
d
e
t
a
c
d
n
i
i
h
t
i
w
s
d
a
e
r
S
G
N
60
40
20
0
DNA-PKi:
-
+
PRINS
HeLa
a
d
t
i
d
e
d
e
t
a
c
d
n
i
i
h
t
i
w
s
d
a
e
r
S
G
N
f
o
%
80
60
40
20
0
AAVS1 6 bp
PCSK9 1 bp A
PCSK9 1 bp C
PCSK9 1 bp G
PCSK9 1bp T
HBEGF 3 bp
HBEGF 6 bp
HBEGF 1 bp A
HBEGF 1 bp C
HBEGF 1 bp T
CTLA4 8 bp
-
RT template
PBS
A C T G T G G G G C A T C T T
PRINS
HEK293T
AAVS1
wild type
A G T G G C C C C A C T G T G G G G T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
prime edits
A G T G G C C C C A C T G T G G G G C A T C T T T G G A G G G G A C A G A T A A A A G T A C C C
A G T G G C C C C A C T G T G G G G C A T C T T - G G A G G G G A C A G A T A A A A G T A C C C
A G T G G C C C C A C T G T G G G G C A T -
- T G G A G G G G A C A G A T A A A A G T A C C C
- G G A G G G G A C A G A T A A A A G T A C C C
A G T G G C C C C A C T G T G G G G C A T C T -
A G T G G C C C C A C T G T G G G G C A T C T T T T G G A G G G G A C A G A T A A A A G T A C C
A G T G G C C C C A C T G T G G G G C A T G T -
- G G A G G G G A C A G A T A A A A G T A C C C
A G T G G C C C C A C T G T G G G G C A T C T G T G G A G G G G A C A G A T A A A A G T A C C C
prime edits + scaffold integrations
A G T G G C C C C A C T G T G G G G C A T C T T G C T G G A G G G G A C A G A T A A A A G T A C
A G T G G C C C C A C T G T G G G G C A T C T T G T G G A G G G G A C A G A T A A A A G T A C C
A G T G G C C C C A C T G T G G G G C A T C T T G C A C C T G G A G G G G A C A G A T A A A A G
A G T G G C C C C A C T G T G G G G C A T C T T G C A T G G A G G G G A C A G A T A A A A G T A
A G T G G C C C C A C T G T G G G G C A T C T T G C A C C G A C T C G G T G C C A C T T T T T C
indels
- T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
-
A G T G G C C C C A C T G T G -
-
A G T G G C C C C A C T G T G G -
- A G G G G A C A G A T A A A A G T A C C C A G A A C C
A G T G G C C C C A C T G T G G G G C T G G A G G G G A C A G A T A A A A G T A C C C A G A A C
A G T G G C C C C A C T G T G G -
- T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
A G T G G C C C C A C T G T G G G - T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
A G T G G C C C C A C T G T G G G G C A T G G A G G G G A C A G A T A A A A G T A C C C A G A A
A G T G G C C C C A C T -
- T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
- A C A G A T A A A A G T A C C C A G A A C C
A G T G G C C C C A C T G T G G G G -
- T G G A G G G G A C A G A T A A A A G T A C C C A G A A C C
A G T G G C C C C A C T G T -
-
-
-
-
-
-
-
-
-
-
-
-
-
-
ARTICLE
intended insert
insertions
- deletions
cleavage position
23.67%
6.21%
1.42%
0.83%
0.44%
0.25%
0.11%
10.11%
2.06%
0.19%
0.18%
0.11%
0.67%
0.65%
0.49%
0.39%
0.34%
0.32%
0.22%
0.16%
0.13%
PRINS
HCT116
30
20
10
e
t
i
d
e
d
e
t
a
c
d
n
i
i
h
t
i
w
s
d
a
e
r
S
G
N
f
o
%
0
AAVS1 6bp
HBEGF 1bp A
HBEGF 1bp C
PCSK9 1bp G
PRINS
HEK293T
AAVS1
1 bp ins.
prime edits - precise
prime edits - all
indels
f
t
i
d
e
d
e
t
a
c
d
n
i
i
h
t
i
w
s
d
a
e
r
S
G
N
f
o
%
80
60
40
20
0
template:
A
C
G
T
Fig. 3 PEn editing through NHEJ. a Model of PRimed INSertions (PRINS) – an NHEJ-mediated mode of PEn editing using springRNA. b NGS analysis of
PRINS-mediated editing at AAVS1 in HEK293T cells with or without DNA-PK inhibitor. c Representative alignment and allele frequencies of AAVS1 locus
edited with PRINS and the indicated RT template in HEK293T cells. For each category, the top 10 variants are shown with a minimum frequency of 0.1%.
d NGS analysis of PRINS-mediated editing at indicated loci using a panel of springRNAs in HeLa cells. e NGS analysis of PRINS-mediated editing at
indicated loci using a panel of springRNAs in HCT116 cells. f NGS analysis of PRINS-mediated 1 nt insertions at AAVS1 using springRNAs with four different
RT-templates in HEK293T cells. All plots show mean ± SD of n = 3 biologically independent replicates. “prime edits – all” and “prime edits – precise”
categories are superimposed. Source data for Fig. 3b, d–f are provided as a Source Data file.
FANCF
on-target
off-target 1
off-target 2
off-target 3
off-target 4
HEK3
on-target
off-target 1
off-target 2
off-target 3
off-target 4
HEK4
on-target
off-target 1
off-target 2
off-target 3
target sequence/PBS
GGAATCCCTTCTGCAGCACC
GGAAcCCCgTCTGCAGCACC
GGAtTgCCaTCcGCAGCACC
GGAgTCCCTcCTaCAGCACC
aGAggCCCcTCTGCAGCACC
target sequence/PBS
GGCCCAGACTGAGCACGTGA
caCCCAGACTGAGCACGTGc
GaCaCAGACcGgGCACGTGA
aGCtCAGACTGAGCAaGTGA
aGaCCAGACTGAGCAaGaGA
target sequence/PBS
GGCACTGCGGCTGGAGGTGG
tGCACTGCGGCcGGAGGaGG
GGCtCTGCGGCTGGAGGgGG
GGCAtcaCGGCTGGAGGTGG
33.63
PEn-dRT
PAM springRNA
TGG
AGG
TGG
AGG
AGG
0.04
0.07
0.16
1.50
89.96
PEn-dRT
PAM springRNA
TGG
TGG
GGG
GGG
GGG
0.59
1.54
0.35
0.01
80.30
PEn-dRT
PAM springRNA
GGG
TGG
TGG
AGG
48.13
33.74
24.58
PEn-dRT
pegRNA
PEn
springRNA
PEn
pegRNA
PE2
pegRNA
35.77
1.34
0.12
0.04
0.04
63.50
19.60
0.30
0.24
0.35
52.28
12.12
0.20
0.14
0.24
7.08
0.03
0.01
0.01
0.01
PEn-dRT
pegRNA
PEn
springRNA
84.00
94.81
PEn
pegRNA
91.10
PE2
pegRNA
43.42
0.39
1.10
0.28
0.01
2.28
1.72
0.47
0.02
0.74
2.46
0.69
0.02
PEn-dRT
pegRNA
PEn
springRNA
PEn
pegRNA
80.82
49.13
34.37
15.59
85.35
63.42
47.23
44.49
92.02
64.48
44.86
36.60
0.00
0.02
0.15
0.00
PE2
pegRNA
30.68
0.04
0.64
0.14
Fig. 4 Off-target analysis of PEn editing. NGS analysis of editing outcomes at three on-target and eleven off-target sites with indicated editors and
peg/springRNAs. Editing levels are shown as percentages of modified reads in each sample. The values represent the average of n = 3 biologically
independent replicates. Mismatches to the on-target gRNA sequence are highlighted in red, the PBS region is highlighted in blue. Source data for Fig. 5 are
provided as a Source Data file.
NATURE COMMUNICATIONS |
(2022) 13:1240 | https://doi.org/10.1038/s41467-022-28771-1 | www.nature.com/naturecommunications
5
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28771-1
ARTICLE
a
Cas9/PE2/PEn
HBEGF
exon 3
exon 4
small intronic indels: toxin-sensitive
large deletions: toxin-resistant
b
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
s
l
l
e
c
t
n
a
t
s
s
e
r
i
f
o
e
t
a
r
e
v
i
t
a
e
r
l
PE2
PEn + pegRNA
PEn + springRNA
c
0 Cas9
0
0
,
0
1
0
0
0
,
4
0
0
0
0
,
0
1
0
0
0
,
4
0
0
0
0
,
0
1
0
0
0
,
4
0
0
0
0
,
8
0
0
0
,
4
0
d
s
l
l
e
c
t
n
a
i
t
s
s
e
r
f
o
e
t
a
r
e
v
i
t
l
a
e
r
0
PEn + springRNA_PAMins
PEn + springRNA
0
0
0
,
8
0
0
0
,
4
0
0
0
0
,
0
1
0
0
0
,
4
0
0
0
0
,
0
1
0
0
0
,
4
0
0
0
0
,
8
0
0
0
,
4
0
Cas9
PE2
PEn + springRNA
PEn + pegRNA
h
t
p
e
d
d
a
e
r
h
t
p
e
d
d
a
e
r
5,000 bp
100 bp
e
0 PEn + springRNA
0
0
,
9
0
0
0
,
7
0
0
0
,
5
10
9
8
7
6
5
4
3
2
1
PEn + springRNA_PAMins
0
0
0
,
0
1
0
0
0
,
6
5,000 bp
5,000 bp
re-ligation by NHEJ. Since PEn does not rely on random indel
generation by endogenous DSB repair system, it could efficiently
disrupt this cycle by destroying the gRNA binding site upon
successful RT-templated DNA insertion. To test this model, we
have designed a
springRNA encoding an insertion that
reconstitutes the HBEGF gRNA binding site (PAMins), poten-
tially allowing multiple rounds of PEn-mediated cutting. Indeed,
we have observed ~8-fold higher rates of DT-sensitive clones after
PAMins springRNA editing relative to editing with a pegRNA
encoding a random non-PAM insertion (Fig. 5d). Long-read
sequencing of these two samples confirmed the more pronounced
presence of large deletions upon PAMins editing (Fig. 5e). We
have also performed long-read sequencing analysis post DT-
selection to examine large deletion patterns in more detail
6
NATURE COMMUNICATIONS |
(2022) 13:1240 | https://doi.org/10.1038/s41467-022-28771-1 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28771-1
ARTICLE
Fig. 5 Large on-target deletion induction by Cas9, PE2, or PEn editing. a Diphtheria toxin (DT) selection-driven assay to detect on-target large deletions
induced by different genome editing systems. b Relative rates of surviving colonies after DT selection of cells edited by indicated genome editors. Data
normalized to Cas9. The plot shows the mean of n = 2 biologically independent replicates. c Alignment of long HBEGF reads from samples targeted with
Cas9, PE2, or PEn and harvested before DT selection. Red lines denote Cas9 cleavage site. Scalebar 5000 bp. Panels on the right show window of 100 bp
around the cleavage site. d Relative comparison of the rates of surviving colonies after DT selection of cells edited with PEn and either springRNA with a
random insert of PAM-reconstituting springRNA. Data normalized to PEn + springRNA. The plot shows the mean of n = 2 biologically independent
replicates. e Alignment of long HBEGF reads from samples targeted with PEn and either springRNA with a random non-PAM insert of PAM-reconstituting
springRNA (PAMins) harvested before DT selection. The y-axis is set from minimal to maximal read depth for each sample. Source data for Fig. 5b, d are
provided as a Source Data file.
(Supplementary Fig. 5). PEn editing with a non-PAM-insert
springRNA revealed a deletion landscape with discrete transitions
in the coverage depth, suggesting that the detected large deletions
originated from a small number of resistant clones,
further
confirming the rarity of PEn-induced large deletions. On the
other hand, PAMins editing led to a complex and heterogeneous
large deletion pattern resembling that of the Cas9-edited sample
(Supplementary Fig. 5).
Altogether, our data show that PEn editing at the HBEGF locus
does not induce considerable levels of unwanted large on-target
deletions and thus might be a safer alternative compared to Cas9
editing. Additionally, we propose that multiple cycles of Cas9
cutting facilitated by precise repair of the target locus by NHEJ
might be one of the mechanisms responsible for unwanted large
deletions caused by Cas9 editing.
Discussion
In this work, we present two different strategies to introduce
precise genomic insertions using an SpCas9 nuclease-based
prime editor PEn. We showed that PEn promotes insertions
through distinct DNA repair mechanisms, expanding the cur-
rent nickase-based prime editing toolbox. In the first approach,
we combined PEn with canonical pegRNAs to promote a
homology-dependent DSB repair leading to precise insertions.
Using PEn, we efficiently introduced insertions even with
pegRNAs that performed poorly with PE2, suggesting that PEn
can promote a more efficient DNA editing mechanism at the
targeted locus. The highly efficient PEn editing also generated
undesired consequences of DSB repair, such as indels, shorter
prime edits and longer than intended prime edits that con-
tained additional RT-template integrations. Similar bystander
editing was also observed to various extents in the PE2 editing
approach28. While the presence of the unintended integrations
represents a downside of PEn editing, its high robustness and
efficiency might be advantageous over the existing methods
in situations where a seamless 3’ end of
the insertion to
maintain an open reading frame of the target is not necessary,
such as during the correction of frameshift mutations, gene
disruption by defined stop codon integration or exon–intron
junction editing. To control the DNA editing outcomes of PEn,
we devised a strategy to remove the unintended prime edits by
inhibiting DNA-PK, a crucial mediator of NHEJ8. For several
genomic targets, the DNA-PK inhibitor treatment also led to a
significant increase in precise editing levels. While this work
was in revision, a study was published demonstrating that
nuclease-based prime editing can outperform nickase-based
prime editing at hard-to-edit
targets as well as in mouse
independently confirming and complementing our
embryos,
observations29. Additionally, a recent study utilized nuclease-
based prime editing for the introduction of defined large
genomic deletions30.
The mechanism of precise pegRNA-dependent PEn editing is
likely a type of homology-dependent end joining DSB repair. We
tested the involvement of a-EJ, a pathway that is utilizing small
homologies (2–20 bp) to repair DSBs. Nevertheless, our current
data from a-EJ deficient cells suggest that Pol θ-mediated a-EJ is
not
involved. Different homology-dependent modes of DSB
repair such as single-strand annealing (SSA) or homologous
to
recombination might be involved, but
these are thought
require much longer homologies
(>50 bp and >100 bp
respectively)8 than those present in our pegRNAs. Nevertheless,
we cannot currently exclude those two possibilities. Future studies
of PEn editing in systems with selectively inhibited different DNA
into its molecular
repair
mechanism.
enzymes will provide
insights
The observation of NHEJ-mediated integrations of pegRNA
RT templates during PEn editing led us to the development of the
springRNAs. The springRNA does not require a homology
sequence in the RT template and the intended insertion is
installed through precise NHEJ. This mode of PEn editing
(PRINS) could be of particular utility because NHEJ is a preferred
type of DSB repair in most human cell types and acts indepen-
dently on cell cycle progression3,8. NHEJ-driven precise genome
editing has proved to be a valuable tool in the past, but unlike
PRINS, the existing approaches rely on either separately provided
dsDNA donors (larger than ~30 bp) or difficult-to-control indel
generation14,15,31. Thus, to our knowledge, PRINS represents a
unique way of installing small insertions via NHEJ.
Off-target analysis of PEn editing revealed that peg/spring-
RNA-priming can increase the total editing levels at off-target
sites to different extents. Further systematic investigation into
peg/springRNA design and optimal high fidelity Cas9 utilization
will be needed to fully understand and mitigate the off-target
activity of PEn.
Our surprising observation that PEn does not induce large on-
target deletions might provide a substantial advantage over Cas9
editing, where frequent large deletions can be of concern, espe-
cially in therapeutic applications25,26,32. Moreover, our data
suggest a potential mechanism by which large deletions arise
during Cas9-induced DSB generation. While the precision of
NHEJ is controversial10,33, our data provide further evidence that
NHEJ is inherently precise and possibly enables multiple cycles of
target cleavage by Cas9. This “persistent” DSB may then increase
the probability of faulty DNA repair leading to large deletion
generation. This is in line with the observation in human embryos
where long-lasting DSBs were suggested to be a potential cause of
chromosomal loss or rearrangements34.
In conclusion, PEn editing is an effective method for intro-
ducing small genomic insertions and expands the spectrum of
DNA repair mechanisms
that can support prime editing,
including NHEJ, which constitutes a major pathway of DSB
repair in humans.
Methods
DNA constructs. PE2, PEn, and SpCas9 plasmids were generated by gene synthesis
(GenScript). PE2 sequence including the backbone corresponds to the published
CMV-PE2 construct (Addgene #132775). To generate PEn, the H840A Cas9
mutation in the PE2 construct was reversed to the original histidine. To generate
the SpCas9 construct, RT in PEn was replaced with eGFP. PEn dead-RT construct
NATURE COMMUNICATIONS |
(2022) 13:1240 | https://doi.org/10.1038/s41467-022-28771-1 | www.nature.com/naturecommunications
7
ARTICLE
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28771-1
was generated by introducing reported mutations in RT (M3-deadRT M-MLV
RT(R110S, K103L, D200N, T330P, L603W)5. pegRNA constructs were generated
by customizing protospacer, PBS and RT template in the target pMA-U6-pegRNA
vector (GeneArt). Briefly, PCR fragments encoding pegRNAs flanked by 20 bp
homology sequence matching pMA (Invitrogen) target backbone were generated
by template-free PCR using two partially overlapping oligonucleotides. After PCR
cleanup, the fragments were assembled into a linearized pMA backbone using HiFi
DNA Assembly Master Mix (NEB) according to the manufacturer’s protocol. All
pegRNA sequences used in this work are listed in Supplementary Data 1.
Cell culture, drug treatments, and transfections. HEK293T (ATCC CRL-3216),
HEK293T POLQ-/- (Synthego CRISPR KO pool, >90% indels), HCT116 (ATCC
CCL-247) and HeLa (ATCC CCL-2) cells were cultured at 37 °C with 5% CO2 in
Dulbecco’s modified Eagle’s medium (Invitrogen) supplemented with 10% fetal
bovine serum. All cell lines were authenticated and regularly tested for myco-
plasma. For gene editing experiments, cells were transfected using FuGENE HD
reagent (Promega) as per manufacturer’s instructions. For 96-well plate format,
cells were seeded 24 h prior to transfection at 20,000 (HEK293T) or 10,000 (HeLa,
HCT116) cells per well. Cells were transfected with 110 ng of plasmid DNA per
well (55 ng of pegRNA/gRNA + 55 ng of PEn/PE2/Cas9). FuGENE:DNA ratio
used for all transfections was 3:1. For larger wells, cell seeding numbers and
transfected DNA amounts were scaled up accordingly. Cells were harvested for
gene editing analysis after 72 h. In DNA-PKi experiments, AZD7648 (Med-
ChemExpress, CAS No: 2230820-11-6) dissolved in DMSO was added to the
growth medium 5 h prior transfection to the final concentration of 1 µM.
Genomic DNA extraction and sequencing analysis. Cells were harvested using
Quick Extract solution (Lucigen) according to manufacturer’s instructions.
Amplicons were generated using Phusion Flash High-Fidelity PCR Mastermix
(F548, Thermo Scientific) in a 15 µL reaction, containing 1.5 µL of genomic DNA
extract and 0.5 µM of target-specific primers with NGS adapters (primers #1-50, as
listed in the Supplementary Data 1). Applied PCR cycling conditions: 98 °C for
3 min, 30x (98 °C for 10 s, 60 °C for 5 s, 72 °C for 5 s). PCR products were purified
using HighPrep PCR Clean-up System (MagBio Genomics). Size, purity, and
concentration of amplicons were determined using a fragment analyzer (Agilent).
Amplicons were subjected to the second round of PCR to add unique Illumina
indexes. Indexing PCR was performed using KAPA HiFi HotStart Ready Mix
(Roche), 1 ng of PCR template and 0.5 µM of indexed primers in the total reaction
volume of 25 µL. PCR cycling conditions: 72 °C for 3 min, 98 °C for 30 s, 10x (98 °C
for 10 s, 63 °C for 30 s, 72 °C for 3 min), 72 °C for 5 min. Indexed amplicons were
purified using HighPrep PCR Clean-up System (MagBio Genomics) and analyzed
using a fragment analyzer (Agilent). Samples were quantified using Qubit 4
Fluorometer (Life Technologies) and subjected to sequencing using Illumina
NextSeq system according to manufacturer’s instructions. For off-target analysis,
amplicons were generated using Q5 Hot Start High-Fidelity 2x Master Mix
(M0494, NEB). Amplicons for long-read sequencing were generated with Q5 High-
Fidelity polymerase (M0492S, NEB) using primers #51-52 (Supplementary Data 1)
and the following PCR protocol: 98 °C 30 s 30x (98 °C 10 s 70 °C 10 s 72 °C 6 min)
72 °C 6 min.
Bioinformatic analysis. Demultiplexing of the NGS sequencing data was per-
formed using bcl2fastq software. The fastq files were analyzed using
CRISPResso235 in the prime editing mode with the quantification window of
5 starting from the 3’ end of intended inserts. Detailed parameters are listed in the
Supplementary Data 1. Prime edited override sequences were used for each site. To
generate the representative alignments, the window was extended to 30 to visualize
homology arm integrations of different lengths. Histograms in Fig. 2a were gen-
erated using CRISPResso2. Barplots were generated using GraphPad Prism 9
(GraphPad Software, Inc) or JMP 14.1.0 (SAS Institute Inc.). Long-read sequencing
was performed by GeneWiz using PacBio platform. Resulting CCS reads were
aligned to the reference sequence using minimap236 (2.2.15 with “--MD -a -xsplice
-C5 -O6,24 -B4” options). The resulting sam files were processed using a custom
python3 script to extract the read depth and location of deletions. The coverage
plots were produced using R (3.4.2).
Diphtheria toxin selection assay. To assess the rate of large deletions induced by
genome editing, HEK293T cells were transfected with different combinations of
PEn/PE2/Cas9 and gRNA/springRNA/pegRNA followed by a survival assay
based on DT selection. In the survival assay, transfected cells (>50% confluence)
were treated with DT (Sigma-Aldrich) at 20 ng/mL. Cell viability was measured
using the AlamarBlue cell viability reagent (ThermoFisher) before and after DT
selection. The ratio of cell viability before/after the selection was calculated to
indicate the rate of large deletions. Genomic DNA was harvested from each
sample before and after DT selection and indel rates for each sample were
analyzed by NGS.
Reporting summary. Further information on research design is available in the Nature
Research Reporting Summary linked to this article.
Data availability
Data supporting the results of this study are presented within the article and
supplementary figures. NGS data are available in the NCBI Sequence Read Archive
database (BioProject accession code PRJNA803881). Additional details and data to
support the findings of this study are available from the corresponding authors upon
reasonable request. Source data for Figs. 1a, 2a, c, 3b, d–f, 4, 5b, d, S1, S2, S4, S5a are
provided as Source Data file. Source data are provided with this paper.
Received: 3 September 2021; Accepted: 7 February 2022;
References
1.
2.
Sun, S., Osterman, M. D. & Li, M. Tissue specificity of DNA damage response
and tumorigenesis. Cancer Biol. Med. 16, 396–414 (2019).
Schep, R. et al. Impact of chromatin context on Cas9-induced DNA double-
strand break repair pathway balance. Mol. Cell 81, 2216–2230.e10 (2021).
3. Yeh, C. D., Richardson, C. D. & Corn, J. E. Advances in genome editing
through control of DNA repair pathways. Nat. Cell Biol. 21, 1468–1478
(2019).
4. Allen, F. et al. Predicting the mutations generated by repair of Cas9-induced
double-strand breaks. Nat. Biotechnol. 37, 64–72 (2019).
5. Anzalone, A. V. et al. Search-and-replace genome editing without double-
strand breaks or donor DNA. Nature 576, 149–157 (2019).
6. Cox, D. B. T., Platt, R. J. & Zhang, F. Therapeutic genome editing: prospects
and challenges. Nat. Med. 21, 121–131 (2015).
7. Yun, M. H. & Hiom, K. CtIP-BRCA1 modulates the choice of DNA double-
strand-break repair pathway throughout the cell cycle. Nature 459, 460–463
(2009).
Zhao, B., Rothenberg, E., Ramsden, D. A. & Lieber, M. R. The molecular basis
and disease relevance of non-homologous DNA end joining. Nat. Rev. Mol.
Cell Biol. 21, 765–781 (2020).
8.
9. Taleei, R. & Nikjoo, H. Biochemical DSB-repair model for mammalian cells in
G1 and early S phases of the cell cycle. Mutat. Res./Genet. Toxicol. Environ.
Mutagen. 756, 206–212 (2013).
10. Bétermier, M., Bertrand, P. & Lopez, B. S. Is non-homologous end joining
really an inherently error-prone process? PLoS Genet. 10, e1004086 (2014).
11. Bhargava, R. et al. C-NHEJ without indels is robust and requires synergistic
function of distinct XLF domains. Nat. Commun. 9, 2484 (2018).
12. Martínez-Gálvez, G., Manduca, A. & Ekker, S. C. MMEJ-based precision gene
editing for applications in gene therapy and f unctional genomics. bioRxiv
https://doi.org/10.1101/2020.04.25.060541 (2020).
13. van Overbeek, M. et al. DNA repair profiling reveals nonrandom outcomes at
Cas9-mediated breaks. Mol. Cell 63, 633–646 (2016).
14. Maresca, M., Lin, V. G., Guo, N. & Yang, Y. Obligate ligation-gated
recombination (ObLiGaRe): custom-designed nuclease-mediated targeted
integration through nonhomologous end joining. Genome Res. 23, 539–546
(2013).
15. Suzuki, K. et al. In vivo genome editing via CRISPR/Cas9 mediated homology-
independent targeted integration. Nature 540, 144–149 (2016).
16. Shen, M. W. et al. Predictable and precise template-free CRISPR editing of
pathogenic variants. Nature 563, 646–651 (2018).
17. Cristea, S. et al. In vivo cleavage of transgene donors promotes nuclease-
mediated targeted integration. Biotechnol. Bioeng. 110, 871–880 (2013).
18. Saha, K. et al. The NIH somatic cell genome editing program. Nature 592,
195–204 (2021).
19. Chen, P. J. et al. Enhanced prime editing systems by manipulating cellular
determinants of editing outcomes. Cell 184, 5635–5652.e5629 (2021).
20. Schene, I. F. et al. Mutation-specific reporter for the optimization and
enrichment of prime editing. bioRxiv https://doi.org/10.1101/
2021.05.08.443062 (2021).
21. Wang, Q. et al. Broadening the reach and investigating the potential of prime
editors through fully viral gene-deleted adenoviral vector delivery. Nucleic
Acids Res. 49, 11986–12001 (2021).
22. Fok, J. H. L. et al. AZD7648 is a potent and selective DNA-PK inhibitor that
enhances radiation, chemotherapy and olaparib activity. Nat. Commun. 10,
5065 (2019).
23. Tsai, S. Q. et al. GUIDE-seq enables genome-wide profiling of off-target
cleavage by CRISPR-Cas nucleases. Nat. Biotechnol. 33, 187–197 (2015).
24. Richardson, C. D., Ray, G. J., Bray, N. L. & Corn, J. E. Non-homologous DNA
increases gene disruption efficiency by altering DNA repair outcomes. Nat.
Commun. 7, 12463 (2016).
8
NATURE COMMUNICATIONS |
(2022) 13:1240 | https://doi.org/10.1038/s41467-022-28771-1 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28771-1
ARTICLE
25. Adikusuma, F. et al. Large deletions induced by Cas9 cleavage. Nature 560,
E8–E9 (2018).
26. Kosicki, M., Tomberg, K. & Bradley, A. Repair of double-strand breaks
induced by CRISPR–Cas9 leads to large deletions and complex
rearrangements. Nat. Biotechnol. 36, 765–771 (2018).
27. Li, S. et al. Universal toxin-based selection for precise genome engineering in
human cells. Nat. Commun. 12, 497 (2021).
28. Petri, K. et al. CRISPR prime editing with ribonucleoprotein complexes in
zebrafish and primary human cells. Nat. Biotechnol. 40, 189–193 (2021).
29. Adikusuma, F. et al. Optimized nickase- and nuclease-based prime editing in
human and mouse cells. Nucleic Acids Res. 49, 10785–10795 (2021).
Jiang, T., Zhang, X.-O., Weng, Z. & Xue, W. Deletion and replacement of long
genomic sequences using prime editing. Nat. Biotechnol. 40, 227–234 (2022).
31. Román-Rodríguez, F. J. et al. NHEJ-mediated repair of CRISPR-Cas9-induced
DNA breaks efficiently corrects mutations in HSPCs from patients with
fanconi anemia. Cell Stem Cell 25, 607–621.e607 (2019).
30.
32. Alanis-Lobato, G. et al. Frequent loss of heterozygosity in CRISPR-
Cas9–edited early human embryos. Proc. Natl Acad. Sci. USA 118,
e2004832117 (2021).
33. Brinkman, E. K. et al. Kinetics and fidelity of the repair of Cas9-induced
double-strand DNA breaks. Mol. Cell 70, 801–813.e806 (2018).
34. Zuccaro, M. V. et al. Allele-specific chromosome removal after Cas9 cleavage
in human embryos. Cell 183, 1650–1664.e1615 (2020).
35. Clement, K. et al. CRISPResso2 provides accurate and rapid genome editing
sequence analysis. Nat. Biotechnol. 37, 224–226 (2019).
36. Li, H. Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics
34, 3094–3100 (2018).
Acknowledgements
We thank Steve Rees, Mohammad Bohlooly, and Mike Snowden for supporting this
work. We thank Amelia Smith for proofreading the manuscript. This project has received
funding from the European Union’s Horizon 2020 research and innovation program
under the Marie Skłodowska-Curie grant agreement no. 765269 (S.W). M.P. is a PostDoc
fellow of the AstraZeneca R&D PostDoc program.
Author contributions
M.P. and M.M. conceptualized the study. M.P., N.A., S.L., and S.W. performed most of
the experimental work with help from P.H., D.D., J.B., S.v.d.P., P.M-G., S.Š, and G.S.;
M.F. performed bioinformatical analyses. M.P. prepared the manuscript with input from
all authors. M.M. supervised the study.
Competing interests
M.P., N.A., S.L., S.W,. P.H., D.D., J.B., S.v.d.P., P.M-G., S.Š., G.S., M.F., and M.M. are
employees and shareholders of AstraZeneca. B.B. is a former employee of AstraZeneca.
M.M. is listed as inventor in an AstraZeneca patent application (WO2021204877A2)
related to this work.
Additional information
Supplementary information The online version contains supplementary material
available at https://doi.org/10.1038/s41467-022-28771-1.
Correspondence and requests for materials should be addressed to Martin Peterka or
Marcello Maresca.
Peer review information Nature Communications thanks the anonymous reviewer(s) for
their contribution to the peer review of this work.
Reprints and permission information is available at http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
material in this article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not included in the
article’s Creative Commons license and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.
© The Author(s) 2022
NATURE COMMUNICATIONS |
(2022) 13:1240 | https://doi.org/10.1038/s41467-022-28771-1 | www.nature.com/naturecommunications
9
| null |
10.1073_pnas.2301121120.pdf
|
Data, Materials, and Software Availability. Cryo-EM maps and atomic mod-
els for all structures described in this work have been deposited to the Electron
Microscopy Data Bank (EMDB) and the Protein Data Bank (PDB), respectively.
Accession codes are as follows: PLCβ3 in solution-8EMV and EMD-28266,
PLCβ3 in complex with Gβγ on vesicles-8EMW and EMD-28267, and PLCβ3 in
complex with Gβγ on nanodiscs-8EMX and EMD-28268.
| null |
RESEARCH ARTICLE |
BIOCHEMISTRY
OPEN ACCESS
Gβγ activates PIP2 hydrolysis by recruiting and orienting PLCβ on
the membrane surface
Maria E. Falzonea,b
and Roderick MacKinnona,b,1
Edited by James Hurley, University of California, Berkeley, CA; received January 19, 2023; accepted April 6, 2023
Phospholipase C-βs (PLCβs) catalyze the hydrolysis of phosphatidylinositol 4,
5–bisphosphate (PIP2) into inositoltriphosphate (IP3) and diacylglycerol (DAG). PIP2
regulates the activity of many membrane proteins, while IP3 and DAG lead to increased
intracellular Ca2+ levels and activate protein kinase C, respectively. PLCβs are regulated
by G protein–coupled receptors through direct interaction with G𝛼q and G𝛽𝛾 and are
aqueous-soluble enzymes that must bind to the cell membrane to act on their lipid sub-
strate. This study addresses the mechanism by which G𝛽𝛾 activates PLCβ3. We show that
∼ 0.43 mol % )
PLCβ3 functions as a slow Michaelis–Menten enzyme ( kcat
on membrane surfaces. We used membrane partitioning experiments to study the
solution-membrane localization equilibrium of PLCβ3. Its partition coefficient is such
that only a small quantity of PLCβ3 exists in the membrane in the absence of G𝛽𝛾 . When
G𝛽𝛾 is present, equilibrium binding on the membrane surface increases PLCβ3 in the
membrane, increasing Vmax in proportion. Atomic structures on membrane vesicle surfaces
show that two G𝛽𝛾 anchor PLCβ3 with its catalytic site oriented toward the membrane
surface. Taken together, the enzyme kinetic, membrane partitioning, and structural data
show that G𝛽𝛾 activates PLCβ by increasing its concentration on the membrane surface
and orienting its catalytic core to engage PIP2 . This principle of activation explains rapid
stimulated catalysis with low background activity, which is essential to the biological pro-
cesses mediated by PIP2, IP3, and DAG.
∼ 2 s−1, KM
PLCβ | Gβγ | PIP2 | GPCR signaling | membrane recruitment
Phospholipase C-β (PLCβ) enzymes cleave phosphatidylinositol 4,5-bisphosphate ( PIP2 )
into inositoltriphosphate ( IP3 ) and diacylglycerol ( DAG ) (1, 2). Their activity is controlled
by G protein–coupled receptors (GPCRs) through direct interaction with G proteins
(3–5). IP3 increases intracellular calcium, DAG activates protein kinase C, and levels of
PIP2 regulate numerous ion channels. Therefore, the PLCβ enzymes under GPCR regu-
lation are central to cellular signaling (Fig. 1A) (6–8). There are four PLCβs (1–4) in
humans: PLC 𝛽4 is activated by G𝛼q, and PLCβ1–3 are activated by both G𝛼q and G𝛽𝛾.
PLCβ2/3 are also activated by the small GTPases Rac1/2 (9–15).
What do we know about PLCβs and their regulation by G proteins? PLCβs are cytoplasmic
enzymes that must access the membrane where their substrate PIP2 resides in the inner leaflet.
They contain a catalytic core, a proximal C-terminal domain (CTD) with autoinhibitory
activity, and a distal CTD with structural homology to a bin-amphiphysin-Rvs domain
important for membrane binding (3, 4). At the active site, an X–Y linker exerts additional
autoinhibitory regulation by direct occlusion (9, 15–17). G𝛼q binds to the proximal and
distal CTDs, displacing the autoinhibitory proximal CTD from the catalytic core and Rac1
binds to the PH domain of PLCβ2 (9, 18–21). Notably, in both cases the autoinhibitory
X–Y linker still occludes the active site. Less is known about regulation of PLCβs by G𝛽𝛾 .
Potential binding sites have been described, but no structures have been determined (3, 4).
The focus of this study is regulation of PLCβ3 by G𝛽𝛾.
The mechanism of PLCβ activation by G𝛽𝛾 is unknown. In vitro studies have con-
cluded that locally concentrating PLCβ on the membrane is not the basis of activation
and this still dominates thinking in the field (3, 4, 22–28). However, the requirement
of the lipid group on G𝛽𝛾 to achieve activation and the demonstration that over
expression of G proteins in cells increases PLCβ in the membrane fraction suggests
that a localization mechanism needs revisiting (13, 29). Part of the challenge in char-
acterizing PLCβ enzymes is precisely the membrane involvement. PLCβs reside in 3
dimensions (the cytoplasm) but catalyze on a two-dimensional surface (the membrane).
Functional measurements must account for this and at the same time permit sufficient
time resolution, unlike the standard radioactive assay used in the field until now. To
overcome the challenge, we have developed new functional methods, including a rapid
kinetic analysis of PLCβ3 enzyme activity that employs a direct read-out of PIP2
concentration as a function of time, a membrane partitioning assay to quantify
Significance
GPCRs are major mediators of
transmembrane signal
transduction, responding to a
wide range of stimuli including
hormones and
neurotransmitters. Important
targets of GPCR signaling,
PLCβ enzymes catalyze the
hydrolysis of PIP2 into IP3 and
DAG, leading to increased
intracellular Ca2+ levels and
activation of PKC, respectively.
PLCβs exhibit very low basal
activity through multiple
mechanisms of autoinhibition
and are activated by both G𝛼q
and G𝛽𝛾 . In this study, we
demonstrate that G𝛽𝛾 activates
PLCβ by recruiting it to the
membrane where its substrate
PIP2 resides and by orienting its
active site. This activation
mechanism permits robust and
rapid activation of PLCβ upon
GPCR stimulation in the setting of
low background activity during
GPCR quiescence.
Author affiliations: aLaboratory of Molecular Neuro
biology and Biophysics, The Rockefeller University, New
York, NY 10065; and bHHMI, The Rockefeller University,
New York, NY 10065
Preprint servers: Deposited as a preprint on bioRxiv.
Author contributions: M.E.F. and R.M. designed research;
M.E.F. performed research; M.E.F. and R.M. analyzed
data; and M.E.F. and R.M. wrote the paper.
The authors declare no competing interest.
This article is a PNAS Direct Submission.
Copyright © 2023 the Author(s). Published by PNAS.
This open access article is distributed under Creative
Commons Attribution License 4.0 (CC BY).
1To whom correspondence may be addressed. Email:
[email protected].
This article contains supporting information online at
https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.
2301121120//DCSupplemental.
Published May 12, 2023.
PNAS 2023 Vol. 120 No. 20 e2301121120
https://doi.org/10.1073/pnas.2301121120 1 of 11
membrane recruitment, and atomic structures on lipid mem-
brane surfaces, to analyze the mechanism by which G𝛽𝛾 acti-
vates PLCβs.
Results
To explain with accuracy our data analysis, we present a series
of equations and their rationale. At least a qualitative under-
standing of these equations is required to fully appreciate the
meaning and wider significance of the data, and what it implies
about the molecular mechanisms crucial for PLCβ3 function.
Some of the analysis and associated equations are, to our knowl-
edge, unfamiliar to biochemical analysis. In particular, when
analyzing both the kinetics of PIP2 hydrolysis on a membrane
surface and the equilibrium binding reaction between proteins
on a membrane surface, we encountered the complex issue of
processes occurring in 2 dimensions that involve components
in 3 dimensions. We dealt with this issue in a particular way,
which we describe thoroughly to stimulate debate and invite
critique. We appreciate that many readers will want to grasp
the biological implications of this work without getting bogged
down by equations. For this reason, we have explained the
meaning of each equation in words, which should be sufficient
to understand the main conclusions of this work.
Development of a Planar Lipid Bilayer Assay for PLCβ3 Function.
We developed a detergent-free, planar lipid bilayer assay to measure
PLCβ3 function using a PIP2-dependent ion channel to report its
concentration over time (Fig. 1 B–D). Briefly, two chamber cups were
connected in the vertical configuration by a ~250 𝜇m hole in a 100 𝜇m
piece of Fluorinated ethylene propylene copolymer (30). A ground
electrode was placed in the Cis chamber and a reference electrode in
the Trans chamber (Fig. 1B). Lipids dispersed in decane were used to
paint a bilayer over the hole separating the two chambers. We used
a 2:1:1 mixture of 1, 2-dioleoyl-sn-glycero-3-phosphoethanolamine
(DOPE): 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC): 1-
palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine (POPS) lipids and
included a predetermined mole fraction of long-chain PIP2 inside
the membrane to set its starting concentration. Ion channels and G
proteins were incorporated by proteolipid vesicle application to the
bilayer, and the current from reconstituted ion channels was measured
(30). We added PLCβ3 to the Cis chamber, which was subjected to
continuous mixing to ensure homogeneity of the chamber.
The PIP2-dependent, G protein-dependent inward rectifier K+
channel-2 (GIRK2, specified as GIRK) was used as a readout of PIP2
concentration. This channel is well characterized in vitro, strictly
depends on PIP2 for channel opening, and is amenable to measuring
large macroscopic currents using planar lipid bilayers (31, 32).
Further, GIRK exhibits fast rates of association and disassociation of
A
B
C
D
Fig. 1. Development of a planar lipid bilayer assay for PLCβ activity using a PIP2 dependent ion channel as readout. (A) Cartoon summary of G𝛼idependent
signaling to PLCβ through G𝛽𝛾 . (B) Cartoon schematic of planar lipid bilayer setup used to measure PLCβ function. (C) Representative current decay upon PLCβ
dependent depletion of PIP2 . (D) Representative current recovery upon reactivation of incorporated channels with shortchain C8PIP2. This experiment was
carried out under subsaturating longchain PIP2 (1.0 mol % ) in the bilayer, which correlates to ~30% of maximal GIRK current. In D, saturating C8PIP2 was added,
which leads to ~3× the amount of starting current.
2 of 11 https://doi.org/10.1073/pnas.2301121120
pnas.org
PIP2 , which permits the measurement of PLCβ3 catalytic activity
that is not filtered by a slow channel response (31). Experiments were
carried out in the presence of symmetric MgCl2 to ensure blockage
of channels with their PIP2 binding sites facing the Trans chamber,
which is not accessible to PLCβ3 (Fig. 1B) (33). This ensures that
when positive voltage is applied to the reference relative to the ground,
the current is derived only from channels accessible to PLCβ3 added
to the Cis chamber.
GIRK also requires G𝛽𝛾 for channel activity. To separate the
effects of G𝛽𝛾 on channel function and PLCβ3 activity we used the
ALFA nanobody system (34) to tether soluble G𝛽𝛾 to GIRK (35).
We tagged GIRK with the short ALFA peptide on the C-terminus
and G𝛾 with the ALFA nanobody on the N-terminus in the back-
ground of the C68S mutant, which prevents lipidation of G𝛾 .
Nanobody-tagged G𝛾 assembled normally with G𝛽 and was able to
bind to other effectors (35). Because the ALFA nanobody binds to
the ALFA tag with ~30 pM affinity (34), at 30 nM concentration,
the ALFA nanobody-tagged G𝛽𝛾 fully activates ALFA peptide-tagged
GIRK. In addition, the nanobody-tagged G𝛽𝛾 does not activate
PLCβ3 due to its lack of a lipid anchor (13, 29).
Human PLCβ3 was used to establish our assay owing to its
significant activation by both G𝛽𝛾 and G𝛼q (14, 36, 37). The
addition of PLCβ3 to membranes already containing lipidated
G𝛽𝛾 , following an equilibration period of about 2 s, led to a rapid
current decay that was complete in ~20 s (Fig. 1C). Subsequent
addition of 32 𝜇M C8PIP2, an aqueous-soluble, short chain ver-
sion of PIP2, rescued the current to a maximum level (Fig. 1D)(31),
indicating that the current decay was due to PIP2 depletion from
the bilayer by the PLCβ3 enzyme. The PLCβ3 mediated current
decay was slower than when C8PIP2 is rapidly removed by perfu-
sion (31). Furthermore, the rate of PLCβ3-mediated current decay
depends on the PLCβ3 concentration (SI Appendix, Fig. S1). These
findings indicate that the decay measures the rate of PLCβ3 cata-
lytic activity rather than PIP2 unbinding from the channel. No
change in the current was observed following PLCβ3 addition in
the absence of CaCl2 (2 mM EGTA), which is required for enzy-
matic function (SI Appendix, Fig. S1A). Repetitions of these exper-
iments yielded consistent results with very similar time courses of
current decay. These observations indicate that we can measure
PLCβ3 catalytic activity using this system and that the addition
of PLCβ3 does not induce artifacts to the bilayer or to reconsti-
tuted GIRK channels.
Kinetic Analysis of PIP2 Hydrolysis by PLCβ3. The interfacial nature
of PLCβ3 activity presents a challenge to the study of its function
because PLCβ3 is a soluble enzyme that must associate with the
membrane to carry out catalysis. To describe the reaction occurring at
the two-dimensional membrane surface, which must account for the
exchange of PLCβ3 with the three-dimensional water phase, we give
concentrations as dimensionless mole fraction × 100 ( mf , expressed
as mol % ) using square brackets, [quantity], unless specified as molar
units using square brackets with subscript molar, [quantity]molar.
Furthermore, to simplify expressions, we approximate mf within each
solvent phase, water or lipid, as moles solute per moles solvent rather
than moles solute per moles solvent plus solute. This approximation
introduces into the kinetic analysis a maximum error in mf of 1.0
% for the PIP2 concentration in membranes and less than 1.0 %
for all other components. For PIP2 in membranes, the initial mf is
predetermined through the bilayer lipid composition. For PLCβ3,
the mf in membranes is calculated from that in three-dimensional
solution using its partition coefficient, which is described below.
The measured current decays can be converted to PIP2 decays
using the PIP2 concentration dependence of the channel, which we
determined using titration experiments. Bilayers were formed with
varying concentrations of long chain PIP2 from 0.1 to 4.0 mol % ,
GIRK-containing vesicles were fused, the current was measured, and
water-soluble C8PIP2 (32 𝜇M) was added to the Cis chamber to
activate the channels maximally (SI Appendix, Fig. S1 B and C) (31).
The measured current was normalized to the maximally activated
current, Imax , for each PIP2 concentration and fit to a modified Hill
equation, Eq. 1, to determine values A, k, and r (Fig. 2A):
I
Imax
= A
[PIP2]r
kr + [PIP2]r
.
[1]
Eq. 1 is an empirical function whose utility is to convert GIRK
current into PIP2 concentration. In subsequent experiments with
PLCβ3, bilayers initially contain 1.0 mol % PIP2 , which corre-
sponds to ~30% of the maximal current (Fig. 2A).
The PLCβ3/Gβγ-dependent current decays were corrected by
subtracting a constant current value representing nonspecific leak,
then normalized to the starting PIP2 concentration (1.0 mol % ),
and converted to PIP2 concentration decays using Eq. 1 with the
predetermined values for k , A , and r (Fig. 2B). After an approxi-
mately 2 s delay associated with mixing of PLCβ3, PIP2 decays
contained two components: an initial, approximately linear com-
ponent followed by a slower, approximately exponential compo-
nent (Fig. 2C). The linear component is consistent with PLCβ3
catalysis occurring as a 0th order reaction, where the catalytic rate
is independent of the PIP2 concentration, suggesting that at our
starting concentration (1.0 mol % PIP2 ), the active site of PLCβ3
is nearly fully occupied by substrate ( PIP2 ). The second, expo-
nential, component is consistent with the PIP2 concentration
becoming limiting to catalysis, a first-order reaction, as the decay
progresses and the concentration of PIP2 decreases. In the example
shown, for illustrative purpose, we estimated the rate within six
intervals along the decay curve, demarcated with different colored
circles (Fig. 2C), by measuring the slope to approximate d [PIP2]
within each interval, and then plotted the slope’s absolute value
against the average PIP2 concentration for the corresponding
interval (Fig. 2D). A Michaelis–Menten equation (Eq. 2, below)
fit the data points with R2 ~ 0.99, indicating that PLCβ3 catalytic
activity can be described by this kinetic rate equation (Fig. 2D).
The graphical procedure described above and in Fig. 2 C and
D was used as an example to place the PIP2 hydrolysis data into
a familiar form of rate as a function of substrate concentration.
For processing all data, we took a more direct approach to analyze
the time-dependent decays within the Michaelis–Menten frame-
work. Expressing the Michaelis–Menten rate equation as
dt
d [PIP2]
dt
= − Vmax
KM
[PIP2]
+ [PIP2] ,
[2]
and integrating from t = 0, we obtain for the PIP2 concentration
as a function of time
[PIP2(t )] = KM ProductLog
e
([PIP2(0)]−t Vmax)
KM
[PIP2(0)]
,
[3]
KM
where [PIP2(0)] is the PIP2 concentration at t = 0 and KM and
Vmax are the Michaelis–Menten parameters. Eq. 3 derived here
contains a well-known function called the Lambert W function
or ProductLog function (38). It describes for the PIP2 concen-
tration an initially linear decay followed by an exponential decay.
Substituting Eq. 3 into Eq. 1, we obtain an expression for GIRK
current decay as a function of time due to PIP2 hydrolysis,
I
Imax
= C + A
[PIP2(t )]r
kr + [PIP2(t )]r ,
[4]
PNAS 2023 Vol. 120 No. 20 e2301121120
https://doi.org/10.1073/pnas.2301121120 3 of 11
A
C
B
D
E
Fig. 2. Extraction of values for kinetic parameters for PLCβ3 catalysis in the presence of lipidated G𝛽𝛾 from current decay curves. (A) PIP2 activation curve for GIRK
varying the mole % of PIP2 in the bilayer and maximally activating with C8PIP2. Green diamonds are average values, open circles are values from each experiment,
and error bars are SEM. Each point is from 3–5 experiments. The normalized current (I/Imax) is fit to a modified hill equation, Eq. 1 (dashed red curve). R2 =
0.994. (B) Demonstration of using the PIP2 activation curve (Right) to convert the current decay (Left) to PIP2 decay. Points on the normalized current decay are
matched to mol % PIP2 and time. (C) Resulting PIP2 decay over time. Circles denote regions used for measuring the rates graphed in D. (D) Plot of d[PIP2]mf∕dt at
regions demarcated in C vs [PIP2] fit to the Michaelis–Menten equation, Eq. 2. R2 = 0.993. (E) Direct fit (shown as red curve) of the normalized current decay with
Vmax, KM, and C as free parameters (Eq. 4). The gray dashed line denotes where the fit starts, which excludes an initial equilibration period. R2 = 0.975.
which permits direct fitting of the normalized current decays to
estimate Vmax and KM (Fig. 2E). [PIP2(t )] in Eq. 4 is given by
Eq. 3, and a third free parameter, C , accounts for the level of
background leak in bilayer experiments; this is visible as the small
residual current (typically ≤ 5% of the GIRK current) at long
times in Figs. 2E and 3A. [PIP2(0)], the initial PIP2 concentra-
tion, is specified by the bilayer composition and A , k, and r are
predetermined through the fit of Eq. 1 to the data shown in Fig. 2A.
Eq. 4 fits the current decay data accurately after ~2 s (Fig. 2E) and
yields consistent results for Vmax (0.17 ± 0.02 mol % ∕ sec ) and KM
(0.42±0.05 mol % ) across repeated experiments (Fig. 3C).
The Role of Gβγ in the Function of PLCβ3. In the experiments
described above, G𝛽𝛾 was added to the planar lipid bilayers by
equilibrating lipid vesicles containing G𝛽𝛾 with the bilayer surface
prior to the application of PLCβ3. When G𝛽𝛾 is not added to
the bilayer, PLCβ3 produces a much slower current decay, as
shown (Fig. 3A and SI Appendix, Fig. S1 D and E). Similarly,
in the presence of 1 µM aqueous-soluble G𝛽𝛾 without a lipid
anchor, which does not partition onto the membrane surface (31),
PLCβ3 catalyzed current decay is also slow (Fig. 3 B and C). Seven
experiments were carried out in the absence of G𝛽𝛾 and the rmsd
between the current decay curves and Eq. 4 were minimized to
yield Vmax (0.0026 ± 0.0007 mol % ∕ sec ) and KM (0.43 ± 0.05
mol % ) (Fig. 3C). Thus, G𝛽𝛾 in the membrane increases Vmax ~65-
fold without affecting KM (Fig. 3C).
Because PLCβ3 is soluble in aqueous solution but must localize
to the membrane surface to catalyze PIP2 hydrolysis, we next
examined whether G𝛽𝛾 in the membrane influences PLCβ3 mem-
brane localization. As detailed by White and colleagues, protein
association with membranes cannot be considered as a simple
binding equilibrium due to the fluid nature of the membrane
without discrete binding sites (39). Instead, membrane association
must be treated as a partitioning process between two immiscible
solvents, the membrane and the aqueous solution. The equilib-
rium partition coefficient, Kx, is the ratio of the mole fraction of
PLCβ3 in the membrane (subscript m) to that in aqueous solution
(subscript w) (39),
Kx
= [PLC 𝛽3m]
[PLC 𝛽3w]
.
[5]
To determine the value of Kx , detergent-free liposomes were recon-
stituted using 2DOPE:1POPC:1POPS lipids to match the lipid
composition of the bilayer experiments, and H+ NMR was used
to measure the lipid concentration at the end of the detergent
removal process (SI Appendix, Fig. S2A). Large unilamellar vesicles
(LUVs) were prepared from the reconstituted liposomes using
freeze–thaw cycles and extrusion through a 200 nm membrane.
The LUVs were incubated with PLCβ3 and pelleted using ultra-
centrifugation to separate the membrane-bound and aqueous
protein fractions. This method allows direct measurement of both
the bound and free protein using fluorescently labeled PLCβ3,
which facilitates determining the partition coefficient from each
experiment individually (39). The membrane-associated fraction
of PLCβ3, fraction partitioned ( Fp), is
=
Fp
[PLC 𝛽3m] [L]molar
[PLC 𝛽3m] [L]molar
+ [PLC 𝛽3w] [W ]molar
=
Kx [L]molar
Kx [L]molar
+ [W ]molar
.
[6]
[W ]molar , the molar concentration of water, is ~55 M and [L]molar ,
the molar concentration of lipid, is set for each experiment using
a stock measured by NMR. Thus, Eq. 6 is a function of the single
4 of 11 https://doi.org/10.1073/pnas.2301121120
pnas.org
A
C
E
B
D
F
mol %
s
Fig. 3. G𝛽𝛾 activates PLCβ3 by increasing its concentration at the membrane. (A) Comparison of normalized current decay in the presence (pink) and absence
, C = −0.03 ± 0.0004, KM = 0.43 ± 0.0008 mol % , R2 =
(gray) of lipidated G𝛽𝛾 fit to Eq. 4 (black curves). Results from the fit without G𝛽𝛾 : Vmax =0.0023 ± 0.6E6
, C = 0.0074 ± 5E5, KM = 0.37 ± 0.0006 mol % . (B) Normalized current decay in the presence of 1 𝜇M soluble G𝛽𝛾 fit to
0.992. With G𝛽𝛾 : Vmax = 0.22 ± 0.0001
Eq. 4 (red curve). R2 = 0.955. (C) Comparison of Vmax , KM , and kcat for PLCβ3 alone, with lipidated G𝛽𝛾 ( G𝛽𝛾 (l)) and with soluble G𝛽𝛾 (G𝛽𝛾 (s)). (D) Membrane
partitioning curve for PLCβ3 alone (black) or in the presence of lipidated G𝛽𝛾 (pink) for 2DOPE:1POPC:1POPS LUVs with Fraction Partitioned ( Fp ) on the Y axis.
Data for 0 G𝛽𝛾 were fit to Eq. 6 for Kx (dashed black curve) and data for +G𝛽𝛾 were fit to Eq. 7 to determine Keq (39). Error bars are range of mean from two
experiments for each lipid concentration. R2 = 0.96 in the absence of G𝛽𝛾 and R2 = 0.95 in the presence of G𝛽𝛾 . (E) Cartoon representation of PLCβ3 activation by
G𝛽𝛾 through membrane recruitment. G𝛽𝛾 significantly increases the membrane association of PLCβ, and accordingly [PLCβ]membrane, which amplifies PIP2 hydrolysis.
[PLCβ]membrane was calculated from Eq. 8 using [PLCβw]=5.3E8 mol%, [Gβγ]=[Gtot]=0.34 mol%, and Kx and Keq, which were determined through the fits in panel D.
(F) Calculated Michaelis–Menten curves (from Eq. 2) for PLCβ3 alone (black), in the presence of 1 𝜇M soluble G𝛽𝛾 (blue) or in the presence of lipidated G𝛽𝛾 (pink)
using the values for KM and Vmax determined from our fits.
mol %
s
free parameter, Kx , which we determine by fitting Eq. 6 to the
partitioning data, yielding Kx ~2.9 ⋅ 104 (Fig. 3D, black curve).
Partitioning experiments carried out with unlabeled PLCβ3 quan-
tified using sodium dodecyl sulfate–polyacrylamide gel electro-
phoresis (SDS-PAGE) analysis yielded a similar value of Kx
( ∼ 4 ⋅ 104 ) (SI Appendix, Fig. S2 B and C), confirming that the
fluorescent label does not alter the partitioning behavior of PLCβ3.
LUVs with the same lipid composition were also prepared con-
taining G𝛽𝛾 , which is exclusively membrane bound, at a protein
to lipid ratio of 1:5 (wt:wt), corresponding to 0.34 mol % , to
match the concentration of G𝛽𝛾 in vesicles equilibrated with pla-
nar lipid bilayers in the kinetic experiments. At this concentration
of G𝛽𝛾 , we observe that PLCβ3 binds to vesicles much more
readily than in the absence of G𝛽𝛾 (Fig. 3D). This observation is
explicable if, when PLCβ3 partitions onto the membrane surface,
it binds to G𝛽𝛾 . Writing the binding reaction on the membrane
+ G𝛽𝛾 ⇌ PLC 𝛽3 ⋅ G𝛽𝛾 , we have Keq
=
surface as PLC 𝛽3m
[PLC 𝛽3m][G𝛽𝛾]
(Fig. 3E). (Note that subscript m indicates
[PLC 𝛽3 ⋅ G𝛽𝛾]
PLC 𝛽3 on the membrane. Since G𝛽𝛾 only resides on the mem-
brane, a subscript is not used for [G𝛽𝛾] and [PLC 𝛽3 ⋅ G𝛽𝛾] ).
When equilibrium is reached, the membrane surface will contain
a quantity of PLC 𝛽3 in the membrane that is not bound to G𝛽𝛾 ,
set by Kx and the aqueous solution concentration of PLC 𝛽3 , as
well as a quantity of PLC 𝛽3 in the membrane that is bound to
G𝛽𝛾 (i.e., PLC 𝛽3 ⋅ G𝛽𝛾) , set by the membrane concentrations of
PLC 𝛽3 , G𝛽𝛾 and Keq . Therefore, in the presence of a total quantity
of G𝛽𝛾 on the membrane, [Gtot] = [G𝛽𝛾] + [G𝛽𝛾⋅PLC 𝛽3] , the
fraction of PLC 𝛽3 on the membrane surface, unbound plus bound
to G𝛽𝛾 , is given by (SI Appendix 2)
(+G𝛽𝛾) =
Fp
Kx
[L]
molar (f (x) + 2 [Gtot] [W ]
+ [W ]
molar) f (x)
(Kx
[L]
molar
molar)
,
[7]
2}1∕2 , where
p = Kx
with f (x) = p + q + x + {4 q x + (p − q + x)
[L]molar [Gtot] , q = Kx [PLCtot]molar , and x = Keq ([W ]
+
Kx [L]molar) . Because [PLCtot]molar (the molar concentration of
PLC 𝛽3 ( PLC 𝛽3w and PLC 𝛽3m ) plus PLC 𝛽3 ⋅ G𝛽𝛾 ), [L]molar and
[W ]molar (molar concentrations of lipid and water) and [Gtot]
( mf G𝛽𝛾 plus PLC 𝛽3 ⋅ G𝛽𝛾 in the membrane) are established in
the experimental setup, and Kx is determined through partition
molar
PNAS 2023 Vol. 120 No. 20 e2301121120
https://doi.org/10.1073/pnas.2301121120 5 of 11
measurements in the absence of G𝛽𝛾 (Fig. 3D), the right-hand side
of Eq. 7 contains a single free parameter, Keq , for the binding of
PLC 𝛽3 to G𝛽𝛾 on the lipid membrane surface. The red dashed
curve in Fig. 3D corresponds to Keq = 0.0090 mol % . It may seem
at first surprising that the series of partitioning experiments in the
presence of G𝛽𝛾 , with knowledge of Kx for PLC 𝛽3 in the absence
of G𝛽𝛾 , uncovers the equilibrium reaction between PLC 𝛽3 and
G𝛽𝛾 on the membrane surface. Nevertheless, the binding reaction
is discernable by this approach, and the inescapable conclusion is
that G𝛽𝛾 concentrates PLC 𝛽3 on the membrane surface (Fig. 3E).
The PLC 𝛽3 -concentrating effect of G𝛽𝛾 has obvious implica-
tions for interpreting the kinetic data reported above, which show
that G𝛽𝛾 increases Vmax by a factor ~65, without affecting KM
very much (Fig. 3C). From Eq. 2, Vmax is the asymptotic rate of
PIP2 hydrolysis when [ PIP2 ] far exceeds KM . In this limit, the
maximum rate of hydrolysis, Vmax , is given by the total membrane
concentration of PLC 𝛽3 times kcat , the turnover rate of a
PLC 𝛽3 ⋅ PIP2 complex. In the bilayer chamber used for the
kinetic experiments, the volume of the aqueous solution is so large
compared to the small area of the lipid bilayer that surface binding
does not significantly alter [PLC 𝛽3w] . Under this condition, we
have
Vmax
= Kx [PLC 𝛽3w]
(
1 +
[Gtot]
+ Kx [PLC 𝛽3w]
)
kcat,
Keq
[8]
where Kx [PLC 𝛽3w] is the membrane concentration of PLC 𝛽3
( [Gtot] = 0 ) and Kx [PLC 𝛽3w]
in
(
)
1 +
is the membrane concentration in its pres-
the absence of G𝛽𝛾
[Gtot]
+ Kx [PLC 𝛽3w]
Keq
)
Keq
1 +
is a mul-
[Gtot]
+ Kx [PLC 𝛽3w]
(
ence ( [Gtot] > 0 ). Thus, the term
tiplier giving the fold-increase in total membrane PLC 𝛽3
concentration due to the presence of G𝛽𝛾 at concentration [Gtot] .
When the known quantities are entered for our experimental con-
ditions, this factor is ~33. In the kinetic experiments, we observed
a 65-fold increase in Vmax in the presence of G𝛽𝛾 . Eq. 8 predicts
a 33-fold increase through G𝛽𝛾′s ability to increase the local con-
centration of PLC 𝛽3 on the membrane surface. A mere two-fold
increase in kcat produced by G𝛽𝛾 binding to PLC 𝛽3 would
account for the full enhancement of Vmax in the kinetic experi-
ments (Fig. 3C). The important conclusion is that most of the
increase in Vmax (within a factor of ~2) is explained by the ability
of G𝛽𝛾 to concentrate PLC 𝛽3 on the membrane surface. Indeed,
it seems very possible that the ~two-fold shortfall is accountable
by the ability of G𝛽𝛾 to orient PLC 𝛽3 , in addition to concentrat-
ing it. Using a conventional Michaelis–Menten plot, with the Vmax
and KM values derived experimentally, we observe that at concen-
trations in our assay, G𝛽𝛾 essentially switches the PLC 𝛽3 enzyme
on (Fig. 3F), and this effect is due largely to the ability of G𝛽𝛾 to
concentrate PLC 𝛽3 on the membrane surface.
In summary, the kinetic studies show that PLC 𝛽3 catalyzes
PIP2 hydrolysis with a substrate concentration dependence like
that of a Michaelis–Menten enzyme (Fig. 2 C–E). We note that
KM corresponds to the mid-range of known PIP2 concentrations
in cell membranes (Figs. 2D and 3F) (40, 41). PLC 𝛽3 aqueous-
membrane partition studies show that G𝛽𝛾 concentrates PLC 𝛽3
on the membrane surface, enough to account for most of
the effect on Vmax (Fig. 3 C and D). To a smaller extent (~two-
fold), G𝛽𝛾 augments Vmax through kcat (Fig. 3C). Next, we
evaluate the structural underpinnings of these functional
properties.
Structural Studies of PLCβ3 in Aqueous Solution by Cryo-EM.
We next determined the structure of PLCβ3 in aqueous solution
using cryo-EM. The structure, consisting of the PLCβ3 catalytic
core at 3.6 Å resolution, contained the PH domain, EF hands,
X and Y domains, the C-terminal part of the X-Y linker, the
C2 domain, and the active site with a Ca2+ ion bound (Fig. 4 A
and B and SI Appendix, Fig. S3 and Table S1). The autoinhibitory
Hα2′ element in the proximal CTD was also resolved, bound
to the catalytic core between the Y domain and the C2 domain,
as proposed by Lyon and colleagues (Fig. 4 A and B) (16, 21)
but not the distal CTD. We also obtained several low-resolution
reconstructions with varying levels of density corresponding to the
catalytic core and distal CTD with differing arrangements between
the two domains (SI Appendix, Fig. S3F). This observation suggests
that the distal CTD is disordered rather than proteolyzed in our
final reconstruction and that the two domains are flexible with
respect to each other, as previously proposed (19). The catalytic
core resolved by cryo-EM is very similar to the crystal structure
with a Cα rmsd of 0.6 Å if the Hα2′ helix is excluded (Fig. 4C). We
note that, as in the crystal structure, the autoinhibitory X–Y linker
occludes the active site (Fig. 4C). We attempted to determine
a structure of PLCβ3 in complex with G𝛽𝛾 in solution, in the
presence or absence of detergent, without success. Furthermore,
we were unable to detect the formation of a complex in solution
by size-exclusion chromatography (SI Appendix, Fig. S3G).
Structural Studies of PLCβ3 Associated with Liposomes. We next
determined the structure of PLCβ3 bound to liposomes consisting
of 2DOPE:1POPC:1POPS. PIP2 was omitted from these samples
because it would have been degraded by PLCβ3 prior to grid
preparation. We obtained a low-resolution reconstruction with the
distal CTD associated with the membrane and the catalytic core
located away from the membrane surface (Fig. 4C and SI Appendix,
Fig. S4 and Table S1). Although the map was low resolution,
previously determined structures fit into the density for each domain
and all reconstructions showed the same orientation of the protein on
the membrane surface (SI Appendix, Fig. S4). The interaction of the
distal CTD with the membrane is consistent with previous reports
of its involvement in membrane association (3). The position of the
catalytic core indicates that significant rearrangements of PLCβ3 with
respect to the membrane must be involved in activation because the
active site is too far from the membrane to access PIP2 . Activating
rearrangements could be mediated by interactions of lipid-anchored
G proteins with the PLCβ3 catalytic core.
The PLCβ3 · Gβγ Complex on Liposomes Reveals Two Gβγ Binding
Sites. We reconstituted G𝛽𝛾 into liposomes consisting of
2DOPE:1POPC:1POPS at a protein to lipid ratio of 1:15 (wt:wt)
and incubated the liposomes with purified PLCβ3 prior to grid
preparation. We determined the structure of the PLCβ3· Gβγ
complex to 3.5 Å and observed two Gβγs bound to the catalytic
core of PLCβ3 (Fig. 5 A–C and SI Appendix, Fig. S5 and Table S1).
The distal CTD is not resolved in our reconstructions, suggesting
that it might adopt many different orientations on the plane of the
membrane relative to the catalytic core, in agreement with previous
studies showing that heterogeneity in the distal CTD increases upon
G𝛽𝛾 binding (42). The catalytic core is very similar to our cryo-
EM structure without membranes, with a Cα rmsd of 0.7 Å. Only
small rearrangements occur at the G𝛽𝛾 binding sites (SI Appendix,
Fig. S6A). Both autoinhibitory elements, the Hα2' and the X–Y
linker, are engaged with the catalytic core (Fig. 5C) consistent with
previous proposals that G𝛽𝛾 does not play a role in relieving this
autoinhibition (15, 16, 21).
6 of 11 https://doi.org/10.1073/pnas.2301121120
pnas.org
Fig. 4. Structures of PLCβ3 in solution and on vesicles without G𝛽𝛾 . (A) primary structure arrangement of PLCβ enzymes. Sections are colored by domain as in C.
Domains in gray (CTD linker and Distal CTD) are not observed in our structures. pCTD is proximal CTD, of which only the Hα2′ is resolved. (B) Sharpened, masked
map of PLCβ3 catalytic core obtained from a sample in solution without membranes or detergent. (C) Structural alignment of the catalytic core of PLCβ3 from the
crystal structure of the fulllength protein bound to G𝛼q [colored in gray, PDBID: 4GNK, (19)] and the structure determined using cryoEM without membranes
(colored by domain). Cα rmsd is 0.6 Å. Calcium ion from the cryoEM structure is shown as a yellow sphere, and the active site is denoted with an asterisk. The PH
domain is pink, the EF hand repeats are blue, the C2 domain is light blue, the Y domain is green, the X domain is teal, and the X–Y linker and the Hα2’ are red. (D)
Unsharpened reconstruction of PLCβ3 bound to lipid vesicles containing 2DOPE:1POPC:1POPS. PLCβ3 is colored in yellow and the membrane is colored in gray.
One G𝛽𝛾 is bound to the PH domain and the first EF hand,
referred to as G𝛽𝛾 1, and the other is bound to the remaining EF
hands, referred to as G𝛽𝛾 2 (Fig. 5C and SI Appendix, Fig. S6
A and B). Both interfaces are extensive, with the G𝛽𝛾 1 interface
burying ~800 Å2 and involving 34 residues, (16 from PLCβ3 and
18 from G𝛽𝛾 ) and the G𝛽𝛾 2 interface burying ~1,100 Å2 and
involving 44 residues (21 from PLCβ3 and 23 from G𝛽𝛾 ) (Fig. 5
D and E and SI Appendix, Fig. S6B and Table S2). The G𝛽𝛾 1
interface is mostly composed of hydrophobic interactions, with
three hydrogen bonds (Fig. 5F and SI Appendix, Fig. S6C),
whereas the G𝛽𝛾 2 interface is mostly composed of electrostatic
interactions, including 10 hydrogen bonds spanning the length
of the interface (Fig. 5 G and H). Both interfaces involve the same
region of G𝛽𝛾 that interacts with G𝛼 and several residues on Gβ
shown to be important for PLCβ activation are involved (43)
(Fig. 5 E and F). Specifically, L117 and W99 on Gβ 1 form hydro-
phobic interactions with L40, I29, and V89 on PLCβ3
(Fig. 5F and SI Appendix, Table S2) (43). On Gβ 2, W99 forms
a hydrogen bond with E294 on PLCβ3, W332 forms an anion-edge
interaction with D227, M101 and L117 form hydrophobic inter-
actions with P239 and F245 on PLCβ3, and D186 forms a hydro-
gen bond with Y240 on PLCβ3 (Fig. 5 G and H and SI Appendix,
Table S2) (43).
We also determined the structure of the PLCβ3 · Gβγ complex
using lipid nanodiscs. We reconstituted G𝛽𝛾 into nanodiscs formed
using the MSP2N2 scaffold protein (44) and 2DOPE:1POPC:
1POPS lipids and incubated them with purified PLCβ3 prior to
grid preparation. We observed only reconstructions with two Gβγs
bound and determined the structure of the complex to 3.3 Å
(SI Appendix, Fig. S7). The two G𝛽𝛾s are bound in the same loca-
tions as was observed in liposomes with comparable interfaces
(SI Appendix, Fig. S6D). A model for this structure aligns well to
the model built using the lipid vesicle reconstruction with a Cα
rmsd of 0.8 Å for all proteins (SI Appendix, Fig. S6D). These struc-
tures suggest that the PLCβ3 · Gβγ complex depends on a mem-
brane environment as we were unable to form a stable complex in
solution with or without detergent, which highlights the importance
of the membrane in Gβγ-dependent activation of PLCβ3.
Gβγ Mediates Membrane Association and Orientation of
the PLCβ3 Catalytic Core. Unmasked refinement of our final
subset of particles from the liposome structure yielded a 3.8 Å
reconstruction showing the PLCβ3 · Gβγ assembly and density
from the membrane (Figs. 5B and 6A). The two Gβγs and the
region of PLCβ3 between them are closely associated with the
membrane and the remainder of the catalytic core, including
the active site, tilts away from the membrane (Fig. 6A). Despite
the tilting, the structure reveals significant rearrangement of the
catalytic core with respect to the membrane compared to its
position in the absence of G𝛽𝛾 , where it was separated from the
membrane surface by a larger distance (Fig. 6A).
Additional 2D and 3D classification without alignment revealed
heterogeneity in the position of the PLCβ3 · Gβγ assembly with
respect to the membrane (Fig. 6 B–D). 2D classes show large
variation in the orientation of the catalytic core with respect to
the membrane surface, with some classes showing the entire cat-
alytic core engaged with the membrane (Fig. 6B). The 2D classes
also reveal differences in membrane curvature originating from
differences in liposome size, which do not seem to be correlated
with the degree of membrane tilting (Fig. 6B). 3D classification
revealed four reconstructions capturing different degrees of tilting
of the catalytic core ranging from ~26° to ~36° (Fig. 6D). We note
that in a locally planar membrane, as opposed to a curved vesicle
membrane, the active site would be nearer the membrane surface
in all classes, but the variability in orientation would presumably
still exist. The protein components of these reconstructions are
like in the original reconstruction, with no internal conforma-
tional changes, indicating that the whole complex tilts on the
membrane as a rigid body.
The lack of conformational changes observed upon G𝛽𝛾 bind-
ing and the catalytic core membrane association are consistent
with our functional studies showing that activation by G𝛽𝛾 is
largely mediated by increasing membrane partitioning. Our struc-
tures suggest that the configuration of the two G𝛽𝛾 binding sites
maintains the catalytic core at the membrane and increases the
probability of productive engagement with PIP2 , potentially
mediated by orientation of the catalytic core observed in our
PNAS 2023 Vol. 120 No. 20 e2301121120
https://doi.org/10.1073/pnas.2301121120 7 of 11
A
D
F
B
C
E
G
H
Fig. 5. PLCβ3 · Gβγ complex on lipid vesicles and G𝛽𝛾 interfaces. (A) Example micrograph showing lipid vesicles with protein complexes. (B) Unsharpened map
from nonuniform refinement showing the PLCβ3 · Gβγ complex on the vesicle surface. Both the inner and outer leaflets of the vesicle are shown. (C) Sharpened,
masked map of the catalytic core of PLCβ3 in complex with two Gβγs on lipid vesicles containing 2DOPE:1POPC:1POPS. PLCβ3 is yellow, G𝛽 1 is dark teal, G𝛾 1 is
light purple, G𝛽 2 is light blue, and G𝛾 2 is light pink. The autoinhibitory elements Hα2′ and the X–Y linker are colored in red. Coloring is the same throughout.
DE: Surface representation of the PLCβGβγ 1 (D) or PLCβGβγ 2 (E) interfaces peeled apart to show extensive interactions. Residues on PLCβ3 that interact with
G𝛽𝛾 1 or 2 are colored according to the corresponding G𝛽 coloring and residues on the G𝛽 s that interact with PLCβ3 are colored in yellow. Interface residues were
determined using the ChimeraX interface feature using a buried surface area cutoff of 15 Å2. (F and G) Interactions of residues on G𝛽 that have been shown to
be important for PLCβ activation with residues from PLCβ3 in the PLCβGβγ 1 interface (F) or the PLCβGβγ 2 interface (G) (43). All labeled interactions are < ~4 Å.
Interacting residues are shown as sticks and colored by heteroatom. Interactions are denoted by black dashed lines. (H) Extensive hydrogen bond network in
the PLCβGβγ 2 interface including both sidechain and backbone interactions. All labeled hydrogen bonds are between ~2.3 and ~3.8 Å. Interacting residues are
shown as sticks and colored by heteroatom. Interactions are denoted by black dashed lines.
reconstructions. Taken together, our kinetic, binding, and struc-
tural studies lead us to conclude that G𝛽𝛾 activates PLC𝛽 mainly
by bringing it to the membrane and orienting the catalytic core
so that the active site can access the PIP2-containing surface
(Fig. 7).
Discussion
This study aims to understand how a G protein, G𝛽𝛾 , activates
the PLC 𝛽3 phospholipase enzyme. We developed and applied
three new technical approaches to study this process. First, because
kinetic analyses of PLC 𝛽 enzymes historically have been limited
to relatively slow radioactivity-based or semiquantitative fluores-
cence assays, we have developed a new higher resolution assay
using a modified, calibrated PIP2-dependent ion channel to pro-
vide a direct read out of membrane PIP2 concentration as a func-
tion of time. This assay is employed in a reconstituted system in
which all components are defined with respect to composition
and concentration. Second, we have used a membrane-water par-
tition assay to study a surface equilibrium reaction between two
proteins ( PLC 𝛽3 and G𝛽𝛾 ) on membranes. Third, we have deter-
mined structures of a protein complex ( PLC 𝛽3 and G𝛽𝛾 ) assem-
bled on the surface of pure lipid vesicles. We also determined the
structures using lipid nanodiscs; however, the lipid vesicles per-
mitted structural analysis of the enzyme-G protein complex on
lipid surfaces unperturbed by the scaffold proteins required to
make nanodiscs. The membrane in our nanodisc reconstructions
is poorly resolved and the complex appears to be associated at
nonphysiological orientations; therefore, we cannot gain any infor-
mation regarding the positioning of the complex on the membrane
from those reconstructions.
We list our essential findings. 1) PLC 𝛽3 catalyzes PIP2 hydrol-
ysis in accordance with Michaelis–Menten enzyme kinetics. 2)
G𝛽𝛾 modifies Vmax , leaving KM essentially unchanged. Under our
experimental conditions, Vmax increases ~65-fold. 3) G𝛽𝛾 increases
membrane partitioning of PLC 𝛽3 , an effect accountable through
equilibrium complex formation between G𝛽𝛾 and PLC 𝛽3 on the
membrane surface. Under our experimental conditions, partition-
ing increases the membrane concentration of PLC 𝛽3 ~33-fold.
4) The G𝛽𝛾 -mediated increase in PLC 𝛽3 partitioning can account
for most of the increase in Vmax , with a smaller, ~two-fold, effect
on kcat . Thus, G𝛽𝛾 regulates PLC 𝛽3 mainly by concentrating it
on the membrane. 5) Two G𝛽𝛾 proteins assemble to form a com-
plex with PLC 𝛽3 on vesicle surfaces. One G𝛽𝛾 binds to the PH
domain and one EF hand of PLC 𝛽3 , while the other binds to the
remaining EF hands. Both G𝛽𝛾 orient their covalent lipid groups
toward the membrane so that the PLC 𝛽3 catalytic core is firmly
anchored on the membrane surface. 6) The PLC 𝛽3 ⋅ G𝛽𝛾 assem-
bly holds the PLC 𝛽3 catalytic core with its active site, as if on the
end of a stylus, poised to sample the membrane surface. Assemblies
on lipid vesicles reveal multiple orientations of the catalytic core
with respect to the surface.
8 of 11 https://doi.org/10.1073/pnas.2301121120
pnas.org
Fig. 6. Tilting of the PLCβ3 · Gβγ complex with respect to the membrane. (A) Consensus unmasked refinement with density for the PLCβ3 · Gβγ complex and the
membrane colored by protein. The membrane is gray, PLCβ3 is yellow, G𝛽 1 is dark cyan, and G𝛾 1 is light purple. The X–Y linker is colored red to highlight the
active site. (B) 2D class averages of the final subset of particles determined without alignment showing side views of the complex on the membrane. Different
membrane curvatures and positions of the complex with respect to the membrane are demonstrated. (C) 2D projections of 3D classes of the PLCβ3 · Gβγ complex
on the membrane. (D) 3D reconstructions of four 3D classes with different positions of the complex on the membrane arranged by degree of tilting with the
most tilted on the left and least tilted on the right.
We described the formation of a complex between PLCβ and
G𝛽𝛾 as a two-step process: first, partitioning of PLCβ from aque-
ous solution into the membrane, and second, binding to G𝛽𝛾 on
the membrane surface. We explicitly consider two steps rather
than one in which PLCβ binds directly to G𝛽𝛾 for the following
reasons. We measured partitioning of PLCβ into membranes with-
out G𝛽𝛾 and measured the corresponding catalysis of PIP2 in the
absence of G𝛽𝛾 . Thus, we know that PLCβ partitions onto the
membrane surface without G𝛽𝛾 . Furthermore, we find that PLCβ
and G𝛽𝛾 do not form a complex in the absence of a membrane,
Fig. 7. G𝛽𝛾 activates PLC𝛽 by increasing its concentration at the membrane and orienting the catalytic core to engage PIP2 . Upon activation of a G𝛼icoupled
receptor, GTP is exchanged for GDP in the G𝛼i subunit and free G𝛽𝛾 is released to bind PLCβ, which increases the concentration of PLCβ at the membrane and
orients the active site for catalysis. The kcat is limited by the X–Y linker (shown in red), which occludes the active site and is only transiently displaced from the
active site to allow catalysis. The distal CTD of PLCβ was omitted for clarity.
PNAS 2023 Vol. 120 No. 20 e2301121120
https://doi.org/10.1073/pnas.2301121120 9 of 11
neither as evaluated by size exclusion chromatography (SI Appendix,
Fig. S3G) nor on cryo-EM grids. It was also shown previously that
G𝛽𝛾 does not activate PLCβ in the absence of membranes (17).
Taken together, this set of findings support the conclusion that
PLCβ partitioning is a required first step in the two-step process
of PLC 𝛽3 ⋅ G𝛽𝛾 complex formation on membranes. We hypoth-
esize that partitioning orients PLCβ with respect to G𝛽𝛾 , defines
a local surface concentration, and thus permits a binding equilib-
rium process that occurs in 2 dimensions, rather than in a
three-dimensional aqueous phase.
We modeled the second step, the equilibrium reaction between
PLC 𝛽 and G𝛽𝛾 on the membrane surface, as bimolecular (1:1 sto-
ichiometry) characterized by a single Keq . In our structural analysis,
however, we discovered two binding sites for G𝛽𝛾 on PLC 𝛽3 .
Additional binding data, using multiple concentrations of G𝛽𝛾, for
example, might reveal two distinct binding constants and whether
they interact with each other (i.e., behave cooperatively). Such a
finding would be important because multiple binding sites could
shape the PLC 𝛽3 activity response to GPCR stimulation. But for
purposes of the present study, the binding model treating a single
site is sufficient. This is because using a single site model when two
sites exist introduces an uncertainty in how PLC 𝛽3 is distributed
over G𝛽𝛾 , not how much PLC 𝛽3 is present in the membrane. The
kinetics depend on how much PLC 𝛽3 is present, and this we have
measured directly with experiment.
The conclusion that G𝛽𝛾 concentrates PLC 𝛽3 on the mem-
brane in our assay is unequivocal. To what extent do these con-
clusions apply to cell membranes? From Eq. 8, we saw that the
increase in membrane PLC 𝛽3 concentration due to the fraction
bound to G𝛽𝛾 is proportional to total G𝛽𝛾 concentration, [Gtot] .
In our assay, [Gtot] is 0.34 mol % , which corresponds to ~5,000
G𝛽𝛾∕𝜇m2 . In cells, we have previously estimated the concentra-
tion of G𝛽𝛾 near GIRK2 channels in dopamine neurons during
GABAB receptor activation at ~1,200 G𝛽𝛾∕𝜇m2 (32). Applying
Eq. 8, this would produce an ~nine-fold increase in the membrane
concentration of PLC 𝛽3 . This is an estimate with certain
unknowns, especially the cytoplasmic concentration of PLC 𝛽3
( [PLC 𝛽3w] ), but the result suggests that the conditions of our
in vitro assay are applicable to cell membranes. Moreover, both
G𝛽𝛾 and G𝛼q have been shown to increase membrane association
of PLC 𝛽s in cells, consistent with our results (29).
We note that our demonstration that G𝛽𝛾 increases membrane
association of PLC 𝛽3 directly contradicts many previous biochem-
ical studies and the current consensus in the field that G proteins
do not increase the local concentration of PLCβs in the membrane
(3, 4, 22–24, 26–28). We suspect that the use of detergent solu-
bilized G𝛽𝛾 in past studies may have interfered with the control
of its concentration on the membrane (22–24).
While our results and mechanism contradict the notion that G
proteins do not concentrate PLC 𝛽s on the membrane, they are
consistent with many previous observations, some we list here. As
stated above, studies with cells have led to the conclusion that
G𝛽𝛾 and G𝛼q increase membrane association of PLC 𝛽s (29). The
lipid anchor is required for the activation of PLCβs by the small
GTPases and G𝛽𝛾 , and G proteins do not activate PLCβs in the
absence of a membrane environment (9–11, 13, 15, 17, 29). The
binding of Rac1 or G𝛼q do not induce conformational changes
around the active site, suggesting that activation is not mediated
by obvious allosteric changes (9, 18, 19). Likewise, we observe no
change in the PLC 𝛽3 active site conformation when G𝛽𝛾 is
bound, only that G𝛽𝛾 recruits PLC 𝛽3 to the membrane and ori-
ents its active site.
Several properties of the G𝛽𝛾 binding sites on PLC 𝛽3 offer
explanations of past observations. First, it has been shown that
G𝛽𝛾 and G𝛼q can activate PLC 𝛽3 simultaneously (36, 37, 45–49).
We find here that the G𝛽𝛾 sites do not occlude the G𝛼q binding
site (18, 19), and therefore both G proteins can in principle bind
to PLC 𝛽3 at the same time and activate PLC 𝛽3 (3, 36, 37, 45).
Second, several amino acids on G𝛽𝛾 that contact PLCβ3 in the
structure were previously shown to play a role in binding to G𝛼 ,
PLCβ, and other effectors (Fig. 5 C and D) (43). Third, the PH
domain was shown to play a role in G𝛽𝛾 binding and activation;
however, based on our structures, G𝛽𝛾 binding does not require
or induce rearrangement of the catalytic core as was previously
proposed (50, 51). Fourth, Rac1 was also shown to bind to the
PH domain of PLCβ2 (SI Appendix, Fig. S6D), and Rac1-activated
PLCβ was shown to be additionally activated by G𝛽𝛾 , leading to
a proposal that the two binding sites did not overlap (9, 10). Our
structures show that Rac1 and G𝛽𝛾 do indeed share an interface
within the PH domain (SI Appendix, Fig. S6D); however, the sec-
ond G𝛽𝛾 binding site can explain the dual activation (9, 10).
An intriguing aspect of PLC 𝛽 enzymes is that all wild-type
structures show that the active site is occluded by the inhibitory
X–Y linker. This includes complexes with G𝛼q, Rac1 and, now,
G𝛽𝛾 (9, 16, 18, 19). It has been proposed that lipids are required
to remove the X–Y linker to achieve catalysis (3, 16, 17). This
must be true to some extent because unless the linker is dis-
placed, even if only rarely, catalysis cannot occur. From our data,
we put forth an alternative proposal that the active site is pre-
dominantly autoinhibited, accounting for a small kcat , even in
the presence of lipids. Consequently, in the absence of GPCR
stimulation, the baseline partitioning of PLC 𝛽 enzyme from the
cytoplasm to the membrane, determined by Kx and the cyto-
plasmic concentration of PLC 𝛽 , will produce very little PIP2
hydrolysis. Only upon GPCR stimulation, when a large quantity
of PLC 𝛽 partitions into the membrane, determined by Keq and
the G𝛽𝛾 concentration generated by GPCR stimulation, is there
enough PLC 𝛽 enzyme in the membrane, even though kcat
remains low, to catalyze PIP2 hydrolysis. In other words, a small
kcat combined with an ability to enact large changes in membrane
enzyme concentration upon GPCR stimulation permits a strong
signal when the system is stimulated and a minimal baseline
when it is not.
Materials and Methods
Protein Expression, Purification, and Reconstitution. All proteins were
purified according to previously established protocols using affinity chroma-
tography and size exclusion chromatography. Detailed methods are described
in SI Appendix, Materials and Methods: Protein Expression and Purification and
Protein Reconstitution.
PLCβ3 Functional Assay. PLCβ activity was measured using a planar lipid bilayer
setup and a PIP2-dependent ion channel to report PIP2 concentration in the mem-
brane over time. Detailed methods are described in SI Appendix, Materials and
Methods: Bilayer Experiments and Analysis.
Membrane Partitioning Experiments. Fluorescently labeled PLCβ3 was mixed
with LUVs and pelleted. Protein in the pellet and supernatant were quantified
using fluorescence. Detailed methods are described in SI Appendix, Materials
and Methods: PLCβ3 Vesicle Partition Experiments.
PLCβ3 Structure Determination. PLCβ3 was mixed with liposomes with or
without Gβγ prior to sample vitrification. Cryo-EM data were collected using a
Titan Krios with a Gatan K3 direct electron detector according to the parameters
in SI Appendix, Table S1 and analyzed according to the procedures outlined in
SI Appendix, Figs. S3–S5 and S7. Atomic models from previously determined
structures were fit into our density maps, refined using PHENIX real-space refine
(52), and manually adjusted. Detailed methods are described in SI Appendix,
10 of 11 https://doi.org/10.1073/pnas.2301121120
pnas.org
Materials and Methods: Cryo-EM Sample Preparation and Data Collection, Cryo-EM
Data Processing, and Model Building and Validation.
Data, Materials, and Software Availability. Cryo-EM maps and atomic mod-
els for all structures described in this work have been deposited to the Electron
Microscopy Data Bank (EMDB) and the Protein Data Bank (PDB), respectively.
Accession codes are as follows: PLCβ3 in solution-8EMV and EMD-28266,
PLCβ3 in complex with Gβγ on vesicles-8EMW and EMD-28267, and PLCβ3 in
complex with Gβγ on nanodiscs-8EMX and EMD-28268.
ACKNOWLEDGMENTS. We thank Chen Zhao for developing and characterizing
the ALFA nanobody-mediated tethering of G𝛽𝛾 to GIRK and for insightful discus-
sions. We thank Venkata S. Mandala for assistance with protein reconstitution and
NMR experiments. We thank Christoph A. Haselwandter for insightful discussion
and comments on the manuscript. We thank Yi Chun Hsiung for assistance with
tissue culture. We thank members of the MacKinnon lab, Jue Chen and members
of her lab for helpful discussions. This work was supported by National Institute of
General Medical Sciences (NIHF32GM142137 to M.E.F.). R.M. is an investigator
in the Howard Hughes Medical Institute. We thank Rui Yan and Zhiheng Yu at
the HHMI Janelia Cryo-EM Facility for help in microscope operation and data
collection. We thank Mark Ebrahim, Johanna Sotiris, and Honkit Ng at the Evelyn
Gruss Lipper Cryo-EM Resource Center of Rockefeller University for assistance with
cryo-EM data collection. Some of this work was performed at the Simons Electron
Microscopy Center and National Resource for Automated Molecular Microscopy
located at the New York Structural Biology Center, supported by grants from the
Simons Foundation (SF349247), NYSTAR (Empire State Development Division of
Science, Technology and Innovation), and the NIH National Institute of General
Medical Sciences (GM103310) with additional support from Agouron Institute
(F00316) and NIH (OD019994).
2.
1.
P. Kemp, G. Hübscher, J. Hawthorne, Phosphoinositides. 3. Enzymic hydrolysis of inositol-containing
phospholipids. Biochem. J. 79, 193–200 (1961).
R. Rodnight, Cerebral diphosphoinositide breakdown: Activation, complexity and distribution in
animal (mainly nervous) tissues. Biochem. J. 63, 223–231 (1956).
A. M. Lyon, J. J. G. Tesmer, Structural insights into phospholipase C-β function. Mol. Pharmacol. 84,
488–500 (2013).
4.
G. Kadamur, E. M. Ross, Mammalian phospholipase C. Annu. Rev. Physiol. 75, 127–154 (2013).
5. M. J. Berridge, Inositol trisphosphate and diacylglycerol: Two interacting second messengers. Annu.
3.
6.
7.
8.
Rev. Biochem. 56, 159–193 (1987).
J. A. Poveda et al., Lipid modulation of ion channels through specific binding sites. Biochim. et
Biophys. Acta 1838, 1560–1567 (2014).
C. V. Robinson, T. Rohacs, S. B. Hansen, Tools for understanding nanoscale lipid regulation of ion
channels. Trends Biochem. Sci. 44, 795–806 (2019).
S. B. Hansen, Lipid agonism: The PIP2 paradigm of ligand-gated ion channels. Biochim. et Biophys.
Acta 1851, 620–628 (2015).
9. M. R. Jezyk et al., Crystal structure of Rac1 bound to its effector phospholipase C-β2. Nat. Struct. Mol.
Biol. 13, 1135–1140 (2006).
10. D. Illenberger et al., Rac2 regulation of phospholipase C-β2 activity and mode of membrane
interactions in intact cells. J. Biol. Chem. 278, 8645–8652 (2003).
11. D. Illenberger et al., Stimulation of phospholipase C-β2 by the Rho GTPases Cdc42Hs and Rac1.
EMBO J. 17, 6241–6249 (1998).
12. M. Camps et al., Isozyme-selective stimulation of phospholipase C-β2 by G protein βγ-subunits.
Nature 360, 684–686 (1992).
13. A. Katz, D. Wu, M. I. Simon, Subunits βγ of heterotrimeric G protein activate β2 isoform of
phospholipase C. Nature 360, 686–689 (1992).
14. A. V. Smrcka, P. C. Sternweis, Regulation of purified subtypes of phosphatidylinositol-specific
phospholipase C β by G protein α and βγ subunits. J. Biol. Chem. 268, 9667–9674 (1993).
15. S. N. Hicks et al., General and versatile autoinhibition of PLC isozymes. Mol. Cell 31, 383–394
(2008).
16. A. M. Lyon, V. G. Taylor, J. J. G. Tesmer, Strike a pose: Gαq complexes at the membrane. Trends
Pharmacol. Sci. 35, 23–30 (2014).
17. T. H. Charpentier et al., Membrane-induced allosteric control of phospholipase C-β isozymes. J. Biol.
Chem. 289, 29545–29557 (2014).
18. G. L. Waldo et al., Kinetic scaffolding mediated by a phospholipase C–b and Gq signaling complex.
Science 330, 974–980 (2010).
19. A. M. Lyon, S. Dutta, C. A. Boguth, G. Skiniotis, J. J. G. Tesmer, Full-length Gαq-phospholipase C-β3
structure reveals interfaces of the C-terminal coiled-coil domain. Nat. Struct. Mol. Biol. 20, 355–362
(2013).
20. A. M. Lyon, J. A. Begley, T. D. Manett, J. J. G. Tesmer, Molecular mechanisms of phospholipase C β3
autoinhibition. Structure 22, 1844–1854 (2014).
21. A. M. Lyon et al., An autoinhibitory helix in the C-terminal region of phospholipase C-β mediates
Gαq activation. Nat. Struct. Mol. Biol. 18, 999–1005 (2011).
22. J. M. Jenco, K. P. Becker, A. J. Morris, Membrane-binding properties of phospholipase C-β1 and
phospholipase C-β: Role of the C-terminus and effects of polyphosphoinositides, G-proteins and
Ca2+. Biochem. J. 327, 431–437 (1997).
23. V. Romoser, R. Ball, A. V. Smrcka, Phospholipase C β2 association with phospholipid interfaces
assessed by fluorescence resonance energy transfer. J. Biol. Chem. 271, 25071–25078 (1996).
24. L. W. Runnels, J. Jenco, A. Morris, S. Scarlata, Membrane binding of phospholipases C-β1 and C-β2
is independent of phosphatidylinositol 4,5-bisphosphate and the α and βγ subunits of G proteins.
Biochemistry 35, 16824–16832 (1996).
25. S. Scarlata, Regulation of the lateral association of phospholipase Cβ2 and G protein subunits by
lipid rafts. Biochemistry 41, 7092–7099 (2002).
26. E. E. Garland-Kuntz et al., Direct observation of conformational dynamics of the PH domain in
phospholipases Cε and β may contribute to subfamily-specific roles in regulation. J. Biol. Chem.
293, 17477–17490 (2018).
27. K. Muralidharan, M. M. Van Camp, A. M. Lyon, Structure and regulation of phospholipase Cβ and ε
at the membrane. Chem. Phys. Lipids 235, 105050 (2021).
28. B. N. Hudson, R. E. Jessup, K. K. Prahalad, A. M. Lyon, Gαq and the phospholipase Cβ3 X-Y linker regulate
adsorption and activity on compressed lipid monolayers. Biochemistry 58, 3454–3467 (2019).
29. O. Gutman, C. Walliser, T. Piechulek, P. Gierschik, Y. I. Henis, Differential regulation of phospholipase
C-β2 activity and membrane interaction by Gαq, Gβ1γ 2, and Rac2. J. Biol. Chem. 285, 3905–3915
(2010).
30. C. Miller, Integral membrane channels: Studies in model membranes. Physiol. Rev. 63, 1209–1242
(1983).
31. W. Wang, M. R. Whorton, R. MacKinnon, Quantitative analysis of mammalian GIRK2 channel
regulation by G proteins, the signaling lipid PIP2 and Na+ in a reconstituted system. eLife 3,
e03671 (2014).
32. W. Wang, K. K. Touhara, K. Weir, B. P. Bean, R. MacKinnon, Cooperative regulation by G proteins and
Na+ of neuronal GIRK2 K+ channels. eLife 5, 1–15 (2016).
33. C. G. Nichols, S.-J. Lee, Polyamines and potassium channels: A 25-year romance. J. Biol. Chem. 293,
18779–18788 (2018).
34. H. Götzke et al., The ALFA-tag is a highly versatile tool for nanobody-based bioscience applications.
Nat. Commun. 10, 1–12 (2019).
35. C. Zhao, R. MacKinnon, Structural and functional analyses of a GPCR-inhibited ion channel TRPM3.
Neuron 1 (2022).
36. F. Philip, G. Kadamur, R. G. Silos, J. Woodson, E. M. Ross, Synergistic activation of phospholipase
C-β3 by Gαq and Gβγ describes a simple two-state coincidence detector. Curr. Biol. 20, 1327–1335
(2010).
37. R. A. Rebres et al., Synergistic Ca2+ responses by Gαi- and Gαq-coupled G-protein-coupled
receptors require a single PLCβ isoform that is sensitive to both Gβγ and Gαq. J. Biol. Chem. 286,
942–951 (2011).
38. M. Goličnik, On the Lambert W function and its utility in biochemical kinetics. Biochem. Eng. J. 63,
116–123 (2012).
39. S. H. White, W. C. Wimley, A. S. Ladokhin, K. Hristova, "[4] Protein folding in membranes:
Determining energetics of peptide-bilayer interactions" in Methods in Enzymology (Academic Press,
1998), vol. 295, pp. 62–87.
40. S. McLaughlin, J. Wang, A. Gambhir, D. Murray, PIP2 and proteins: Interactions, organization, and
information flow. Annu. Rev. Biophys. Biomol. Struct. 31, 151–175 (2002).
41. K. Mandal, Review of PIP2 in cellular dignaling, functions and diseases. IJMS 21, 8342 (2020).
42.
I. J. Fisher, M. L. Jenkins, G. G. Tall, J. E. Burke, A. V. Smrcka, Activation of phospholipase C β by Gβγ
and Gαq involves C-terminal rearrangement to release autoinhibition. Structure 28, 1–10 (2020).
43. C. E. Ford et al., Molecular basis for interactions of G protein βγ subunits with effectors. Science 280,
1271–1274 (1998).
44. Y. V. Grinkova, I. G. Denisov, S. G. Sligar, Engineering extended membrane scaffold proteins for self-
assembly of soluble nanoscale lipid bilayers. Protein Eng. Des. Sel. 23, 843–848 (2010).
45. T. I. A. Roach et al., Signaling and cross-talk by C5a and UDP in macrophages selectively use PLCβ3
to regulate intracellular free calcium. J. Biol. Chem. 283, 17351–17361 (2008).
46. R. C. Carroll, A. D. Morielli, E. G. Peralta, Coincidence detection at the level of phospholipase C
activation mediated by the m4 muscarinic acetylcholine receptor. Curr. Biol. 5, 536–544 (1995).
47. F. Okajima, K. Sato, K. Sho, Y. Kondo, Stimulation of adenosine receptor enhances α1 -adrenergic
receptor-mediated activation of phospholipase C and Ca2+ mobilization in a pertussis toxin-
sensitive manner in FRTL-5 thyroid cells. FEBS Lett. 248, 145–149 (1989).
48. B. H. Shah et al., Co-activation of Gi and Gq proteins exerts synergistic effect on human platelet
aggregation through activation of phospholipase C and Ca 2+ signalling pathways. Exp. Mol. Med.
31, 42–46 (2009).
49. J. A. Ware, M. Smith, E. W. Salzman, Synergism of platelet-aggregating agents. Role of elevation of
cytoplasmic calcium. J. Clin. Invest. 80, 267–271 (1987).
50. G. Kadamur, E. M. Ross, Intrinsic pleckstrin homology (PH) domain motion in phospholipase C-β
exposes a Gβγ protein binding site. J. Biol. Chem. 291, 11394–11406 (2016).
51. G. Drin, D. Douguet, S. Scarlata, The Pleckstrin homology domain of phospholipase Cβ transmits
enzymatic activation through modulation of the membrane-domain orientation. Biochemistry 45,
5712–5724 (2006).
52. P. V. Afonine, J. J. Headd, T. C. Terwilliger, P. D. Adams, PHENIX news. Comput. Crystallogr. Newsletter
4, 43–44 (2013).
PNAS 2023 Vol. 120 No. 20 e2301121120
https://doi.org/10.1073/pnas.2301121120 11 of 11
| null |
10.1103_physrevb.106.235128.pdf
| null | null |
PHYSICAL REVIEW B 106, 235128 (2022)
Thermal critical points from competing singlet formations in fully frustrated
bilayer antiferromagnets
Lukas Weber
,1,2,* Antoine Yves Dimitri Fache,3 Frédéric Mila ,3 and Stefan Wessel
4
1Center for Computational Quantum Physics, Flatiron Institute, 162 5th Avenue, New York, New York 10010, USA
2Max Planck Institute for the Structure and Dynamics of Matter, Luruper Chaussee 149, 22761 Hamburg, Germany
3Institute of Physics, Ecole Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland
4Institute for Theoretical Solid State Physics, RWTH Aachen University, JARA Fundamentals of Future Information Technology, and JARA
Center for Simulation and Data Science, 52056 Aachen, Germany
(Received 25 October 2022; accepted 8 December 2022; published 15 December 2022)
We examine the ground-state phase diagram and thermal phase transitions in a plaquettized fully frustrated
bilayer spin-1/2 Heisenberg model. Based on a combined analysis from sign-problem free quantum Monte
Carlo simulations, perturbation theory, and free-energy arguments, we identify a first-order quantum phase
transition line that separates two competing quantum-disordered ground states with dominant singlet formations
on interlayer dimers and plaquettes, respectively. At finite temperatures, this line extends to form a wall of
first-order thermal transitions, which terminates in a line of thermal critical points. From a perturbative approach
in terms of an effective Ising model description, we identify a quadratic suppression of the critical temperature
scale in the strongly plaquettized region. Based on free-energy arguments we furthermore obtain the full phase
boundary of the low-temperature dimer-singlet regime, which agrees well with the quantum Monte Carlo data.
DOI: 10.1103/PhysRevB.106.235128
I. INTRODUCTION
Geometric frustration in quantum magnets can give rise to
a variety of nonclassical ground states, including quantum-
disordered states that are dominated by the formation of local
spin singlets on particular subclusters [1–4]. Examples in-
clude dimer singlet and plaquette singlet states, where spin
singlets form predominantly among two- and four-spin sub-
clusters, respectively. Such quantum-disordered regions are
often separated by discontinuous (first-order) quantum phase
transition lines in the parameter space of the system. Thermal
fluctuations may replace the discontinuous quantum phase
transition by a continuous thermal crossover between these
different regimes, but it is also possible that the discontinuous
behavior remains stable at low temperatures. In recent years,
several instances were indeed reported in strongly frustrated
quantum magnets in which a discontinuous quantum phase
transition line extends beyond the zero-temperature limit,
forming a boundary of first-order thermal transitions in the
thermal phase diagram [5–9]. It was found that such a “wall
of discontinuities” terminates along a line of thermal critical
points. In the two-dimensional (2D) models studied in these
references, these critical points belong to the universality class
*lweber@flatironinstitute.org
Published by the American Physical Society under the terms of the
Creative Commons Attribution 4.0 International license. Further
distribution of this work must maintain attribution to the author(s)
and the published article’s title, journal citation, and DOI. Open
access publication funded by the Max Planck Society.
of the 2D Ising model. This reflects the fact that a single
scalar quantity is sufficient to distinguish the phases, hence
to describe the critical fluctuations at the thermal critical
points [5].
A prominent example for this scenario is provided by
the layered compound SrCu2(BO3)2, a material that received
increasing attention recently in the field of frustrated quan-
tum magnetism [6]: In SrCu2(BO3)2, a pressure-induced
discontinuous quantum phase transition takes place between a
dimer singlet product phase and a plaquette singlet quantum-
disordered phase at about 20 kbar. The low-temperature
first-order transition line was found to terminate at a critical
point at a temperature of about 4 K, i.e., well below the
scale of the magnetic exchange interactions in this system.
Upon approaching the critical point, the specific heat further-
more exhibits characteristic critical enhancement, as in the 2D
Ising model.
Prior to its experimental observation in SrCu2(BO3)2, this
physics was identified [5] in a related basic 2D model of
strongly frustrated quantum magnetism, the fully frustrated
bilayer (FFB) spin-1/2 Heisenberg antiferromagnet (AFM)
[10]. In the FFB, a discontinuous quantum phase transition
takes place between a dimer singlet phase and an AFM or-
dered phase. Building on recent progress in designing minus
sign-problem free quantum Monte Carlo (QMC) approaches
for frustrated quantum magnets [11,12], it is now possible to
study this quantum phase transition and the critical point that
terminates the extended first-order transition line by unbiased
and large-scale QMC simulations.
In contrast to the case of SrCu2(BO3)2, the temperature
scale of the critical point in the FFB model turns out to be
of similar magnitude as the magnetic exchange interaction
2469-9950/2022/106(23)/235128(7)
235128-1
Published by the American Physical Society
WEBER, FACHE, MILA, AND WESSEL
PHYSICAL REVIEW B 106, 235128 (2022)
(a)
JD
(b)
JP
J
JP
J
FIG. 1. (a) The pFFB lattice with interlayer dimer bonds JD
(thick, red), plaquette bonds JP (thick, blue) and interplaquette bonds
J (thin, black). (b) The effective spin-1 model for the pFFB within
the dimer spin triplet regime.
strengths. Another difference between the FFB model and
SrCu2(BO3)2 is the fact that in the FFB the discontinuous
quantum phase transition takes place between an AFM ground
state and a quantum-disordered phase, while in SrCu2(BO3)2,
the phases on both sides of the quantum phase transition point
are nonmagnetic and quantum disordered. It would thus be
interesting to come up with an example of a discontinuous
quantum phase transition between two quantum-disordered
regions in a quantum spin model that is accessible to sign-
problem free QMC simulations.
Here we consider an extension of the original FFB model
that exhibits a line of discontinuous quantum phase transitions
between two quantum disordered phases with different singlet
patterns, and which can be studied by sign-problem free QMC
simulations. More specifically, we consider the plaquettized
fully frustrated bilayer (pFFB) spin-1/2 Heisenberg model, cf.
Fig. 1(a), and defined in detail below. From sign-problem free
QMC simulations combined with analytical results from per-
turbation theory as well as free-energy considerations, we find
that in this system the line of discontinuous quantum phase
transitions yields a finite-temperature wall of discontinuities
that terminates along a line of 2D Ising critical points, with a
critical temperature that is strongly suppressed with respect to
the magnetic exchange couplings, i.e., similar to the case of
SrCu2(BO3)2.
The remainder of this paper is organized as follows: In
Sec. II we introduce the pFFB model and present results from
sign-problem free QMC simulation of this model in Sec. III.
Next, we report our analytical findings in Sec. IV, before
giving final conclusions in Sec. V.
II. MODEL
The pFFB model that we consider in the following is a
spin-1/2 Heisenberg AFM on the plaquettized bilayer square
lattice, shown in Fig. 1(a). It is defined by the Hamiltonian
(cid:2)
(cid:2)
(cid:2)
H = JD
Si · S j + JP
Si · S j + J
Si · S j,
(1)
P
P−P
(cid:2)i, j(cid:3)
(cid:2)i, j(cid:3)
D
(cid:2)i, j(cid:3)
where Si denotes a spin-1/2 degree of freedom on the ith
lattice site, and the summations extend (from left to right) over
the interlayer dimer bonds, the plaquette bonds and the inter-
plaquette bonds, respectively, cf. Fig. 1(a). The four-site unit
cell, containing two JD-dimer bonds and four JP interdimer
bonds, is also referred to as a plaquette in the following.
The above Hamiltonian can be rewritten in terms of total
dimer spin variables. Namely, for each JD-dimer d, we define
the total dimer spin operator Td = Sd,1 + Sd,2, i.e., the sum of
the spin operators of the two sites that belong to the dth dimer.
In terms of these operators, the Hamiltonian H reads
H = JD
(cid:3)
(cid:2)
d
T2
d
2
− 3
4
(cid:4)
(cid:2)
+ JP
Td · Td (cid:4) + J
(cid:2)
Td · Td (cid:4),
(cid:2)d,d (cid:4)(cid:3)P
(cid:2)d,d (cid:4)(cid:3)P−P
(2)
where the summations extend (from left to right) over the
interlayer dimers, neighboring dimers coupled by plaque-
tte bonds, and neighboring dimers coupled by interplaquette
bonds, respectively.
This expression makes it clear that H has extensively many
local conserved quantities, namely each total dimer spin T2
d ,
which we may encode in additional quantum numbers Td ,
which take on the values 0 and 1 for dimer singlet and triplet
states, respectively. In the dimer triplet sector, where Td = 1
on all dimers, the Hamiltonian H then describes a spin-1
Heisenberg model on a square lattice with a columnar dimer-
ization pattern, cf. Fig. 1(b), i.e., the spin-1 columnar dimer
square lattice Heisenberg model.
In several limiting cases, the physics of the pFFB model
is readily accessible. If the couplings JD (JP) dominate,
the model will host a dimer (plaquette) singlet quantum-
disordered ground state, denoted DS (PS), in which singlets
predominantly form on the JD dimers (JP plaquettes), giving
rise to a finite triplet excitation gap in both cases. If the
coupling J dominates, the system decouples into a system of
weakly coupled one-dimensional spin tubes formed by the J
bonds [cf. Fig. 1(a)]. Along JP = J, the pFFB reduces to the
original FFB, where, if J/JD > 0.42957(2) [10], the ground
state hosts long-range AFM order, with each dimer forming
an effective S = 1 degree of freedom, while for lower values
of J, the FFB resides in the DS phase. In the following,
we will examine the full phase diagram of the pFFB model
in the antiferromagnetic regime, i.e., assuming all exchange
couplings to be positive.
III. QUANTUM MONTE CARLO RESULTS
Even though the Hamiltonian H is strongly frustrated, we
can obtain unbiased numerical results for its properties by em-
ploying sign-problem free stochastic series expansion (SSE)
QMC simulations [13–16] in the dimer spin basis [11,12].
Here we consider systems with periodic boundary conditions,
consisting of L × L unit cells with N = 4L2 spins.
Two basic observables that allow us to distinguish the dif-
ferent phases of the pFFB model are directly accessible in the
dimer spin basis: (i) The dimer triplet density nD = (cid:2)ND/N(cid:3),
where the operator ND counts the number of JD dimers that
are in a triplet state, and (ii) the AFM spin structure factor
S(π , π ) = 1
2L2
(cid:2)
i, j=1
(cid:3)i(cid:3) j (cid:2)Si · S j(cid:3),
(3)
where (cid:3)i = (−1)xi+yi , in terms of the coordinates of lattice
site i. This observable is susceptible to long-ranged AFM
correlations within each of the planes of the bilayer lattice.
QMC results for both observables are presented at a low
temperature of T /JD = 0.1 in Fig. 2 for an L = 12 system.
Both quantities are shown in the parameter plane spanned by
the coupling ratios J/JD and JP/JD. Combining the data from
235128-2
THERMAL CRITICAL POINTS FROM COMPETING …
PHYSICAL REVIEW B 106, 235128 (2022)
PS
1.5
1.0
D
J
/
P
J
0.5
DS
T /JD = 0.1
PS
0.0
1.5
1.0
D
J
/
P
J
0.5
DS
T /JD = 0.1
0.0
0.0
0.2
AFM
tubes
AFM
JP/JD
AFM
1.5
0.48
0.5
J/JD
tubes
50
)
π
,
π
(
S
0
0.0
0.4
J/JD
0.6
0.8
D
n
1.0
0.8
0.6
0.4
0.2
0.0
50
)
π
,
π
(
S
40
30
20
10
0
FIG. 2. Dimer triplet density nD (top panel) and AFM structure
factor S(π , π ) (bottom panel) of the pFFB model as a function of
J/JD and JP/JD at a fixed temperature of T /JD = 0.1 as obtained
from QMC for L = 12. Black lines denote the boundaries of the
AFM phase as obtained from the effective spin-1 model description.
Red (white) lines denotes the first-order quantum phase transition
line obtained from the free-energy comparison (perturbation theory
in the regime J (cid:5) JP ≈ JD). The different regions are labeled by the
corresponding ground-state phases. The inset shows two constant-
JP/JD cuts of S(π , π ).
the two panels, we can readily identify four regimes, denoted
DS, PS, AFM, and tubes, which we already introduced above
and which all appear as extended regions in the ground-state
phase diagram. For low values of both J/JD and JP/JD, the
dominant JD coupling forces the system into the DS phase,
with very small values of both nD and S(π , π ). Outside the
DS region, the triplet density nD is essentially saturated, and
the structure factor S(π , π ) allows us to separate the AFM
regime, with a sizable value of S(π , π ), from both the PS
region (for dominant JP) and the tube phase (for dominant J).
Since long-range AFM order is restricted to zero temperature,
the structure-factor data in Fig. 2, taken at low but finite tem-
perature, varies continuously across the corresponding phase
transition regimes.
As seen from the formulation of the Hamiltonian H in
terms of the dimer spin operators, cf. Eq. (2), inside the dimer
triplet dominated regime the pFFB model becomes an effec-
tive spin-1 Heisenberg model on the columnar dimer square
lattice, cf. Fig. 1(b). The ground-state phase diagram of this
spin-1 model has been determined by QMC simulations in
Ref. [17]. From those results we can extract the critical cou-
pling ratios (J/JP) = 0.18920(2) and (J/JP) ≈ 1/0.011 ≈ 91
for the continuous quantum phase transitions between the
AFM regime and the large-JD PS and the large-J tube phase,
respectively. The black lines in Fig. 2 indicate these transition
lines, which match very well the QMC results.
Along the line JP = J, where the pFFB reduces to the
original FFB, both quantities exhibit a pronounced jump as
the coupling J is tuned across the previously determined
position of the DS-to-AFM quantum phase transition at
J/JD = 0.42957(2) [10]. Indeed, in this regime, the simula-
tion temperature used for Fig. 2 is well below the critical
temperature Tc ≈ 0.22JD [5] of the FFB, i.e., at T = 0.1JD
the system is driven across the first-order thermal transition
line upon increasing J, which leads to the sudden jump in
both quantities, observed already in Ref. [5] (QMC data taken
at temperatures beyond Tc instead show a smooth crossover
behavior, cf. Appendix C). As seen from Fig. 2, the sudden
change in both quantities remains similarly sharp also upon
moving slightly off the JP = J line. However, in the transition
regime between the DS and the PS phase, the triplet density
nD exhibits a smooth crossover in contrast to its sharp jump
along the JP = J line. There are two possible explanations for
this observation: (i) there exists a finite-temperature first-order
transition between both phases, and the line of critical points,
along which the wall of discontinuities terminates, resides at
temperatures below those accessible to the finite-temperature
SSE QMC simulations, or (ii) there is no finite-temperature
phase transition between the DS and the PS regime, but only a
smooth crossover (which however appears unlikely to be real-
ized in a two-dimensional model). In the following section we
will provide arguments from perturbation theory calculations
(in J/JD) as well as free-energy considerations that strongly
support the first scenario, (i), and derive an explicit expression
for Tc along the DS-PS transition line within the perturbative
regime.
IV. PERTURBATION THEORY AND FREE-ENERGY
ARGUMENTS
A. Perturbation theory
Compared to the original FFB, the pFFB model exhibits
various weak coupling regimes where perturbation theory can
be performed. Here we are especially interested in the regime
where J (cid:5) JP ≈ JD, corresponding to the case of weakly cou-
pled plaquettes. Namely, this regime is contained within the
crossover region observed at finite temperature in QMC (cf.
Fig. 2). Our goal in the following will be to use perturbation
theory in order to understand the physics in this region at low
temperatures that are beyond reach of QMC.
(cid:2) = (
To start, we consider the spectrum of a single plaquette, cf.
Fig. 3. Based on the symmetries of the problem, the states
of the plaquette can be labeled (up to degeneracy) by the
(cid:5)
4
plaquette’s total spin S2
i=1 Si )2 and the dimer triplet
density nD. Around JP/JD ≈ 1, the low energy subspace is
made up of a dimer singlet state and a plaquette singlet state,
which exhibit a level crossing at JP/JD = 1. From this, we find
that in the decoupled limit, i.e., for J = 0, the pFFB model
indeed hosts a level-crossing first-order transition at T = 0.
However, at finite temperature, this transition immediately
softens into a crossover at J = 0.
The next question is what changes in this picture once
the interplaquette interactions J are included. On the level of
235128-3
WEBER, FACHE, MILA, AND WESSEL
PHYSICAL REVIEW B 106, 235128 (2022)
nD = 1
nD = 1
2
nD = 0
S(cid:2) = 2
S(cid:2) = 1
S(cid:2) = 0
D
J
/
(cid:2)
E
1
0
−1
−2
0.8
1.0
JP/JD
1.2
FIG. 3. Energy levels of a single plaquette E(cid:2) as a function of
JP/JD, splitting into two singlets, three triplets, and one quintuplet of
different dimer triplet density nD.
quantum numbers, we recall that nD remains a good quantum
number also in the fully coupled model. The same is not
true for the other quantum numbers, so the interplaquette
interactions will in general mix levels within the different nD
sectors. For the low-energy subspace, this means that the sole
nD = 0 dimer singlet level, having no other levels to be mixed
with, remains the same, while the plaquette singlet level gets
shifted, depending on the states on the neighboring plaquettes.
This physics results (see Appendix A for a detailed deriva-
tion) in an effective low-energy Hamiltonian devoid of any
off-diagonal terms,
(cid:2)
(cid:6)
(cid:7)
Heff =
(cid:2)
JP − JD + 5J 2
3JD
σ z
(cid:2)
− J 2
6JD
(cid:8)
4σ z
(cid:2)+ˆx
σ z
(cid:2)
+ σ z
(cid:2)+ˆy
(cid:6)
(cid:9)
+ O
(cid:7)
,
J 3
J 2
D
(4)
where we define σ z
= +1 (−1) if a plaquette is in the dimer
(cid:2)
(plaquette) singlet state, and (cid:2) + ˆx ((cid:2) + ˆy) denotes the neigh-
boring plaquette to the right (top). This Hamiltonian realizes a
classical Ising model with spatially anisotropic interactions in
an effective magnetic field. The effective Ising magnetization,
= 1 − 2nD, is exactly
in the region of validity of the model σ z
(cid:2)
related to nD, shown in Fig. 2 (top).
From the expression of the magnetic field, we can read off
the existence of a first-order transition along
(cid:7)
JP = JD − 5J 2
,
3JD
which extends from T = 0 up to a finite critical temperature
Tc, which is known from Onsager’s solution [18] to satisfy
J 3
J 2
D
+ O
(5)
(cid:6)
(cid:6)
(cid:7)
(cid:6)
(cid:7)
sinh
sinh
= 1,
(6)
2Jx
Tc
2Jy
Tc
with Jx = 2J 2/3JD and Jy = J 2/6JD in our case. This yields
Tc ≈ 0.826J 2/JD
(7)
in the perturbative regime. Thus, weak interplaquette cou-
plings are sufficient to stabilize a first-order transition at finite
temperature. The line of critical temperature at which the first-
order transitions terminate is however suppressed by a factor
of J 2, making it unfeasible to resolve in QMC simulations
(e.g., for a value of J/JD = 0.1, the above estimates gives
Tc ≈ 0.008JD). Nevertheless, we find that the estimate for the
first-order transition line extracted from (5) agrees very well
with the position of the T > Tc crossover observed in QMC
(cf. the white dashed lines in Fig. 2). I
B. Free-energy arguments
The previous perturbative approach was limited to the re-
gion of small J, but it was powerful enough to predict the
existence and shape of a first-order transition in the ther-
modynamic limit. In this section we change our viewpoint,
assuming that such a first-order transition exists in the first
place and that it happens between two specific quantum num-
ber sectors, nD = 0 and nD = 1. For the weakly coupled
regime we just showed that this is the case with the DS
(nD = 0) and PS phase. For the original FFB this fact is also
established with the nD = 1 state being an effective S = 1
AFM [5,10]. Making this assumption allows us to calculate
the first-order line as a level crossing in the free energy of the
two states involved,
FnD=0(T, JD, JP, J ) = FnD=1(T, JD, JP, J ),
(8)
which has been shown to be a very accurate estimate for the
shape of the first-order line below the critical temperature in
the FFB and related models [5–7]. We can further simplify the
argument by noting that in the original FFB case the shape
of the first-order line depends only weakly on temperature
and the same is true for the weakly coupled plaquette regime
where the effective Ising magnetic field is independent of
temperature. Therefore, we approximate the free energy in
Eq. (8) by the ground-state energy.
The ground-state energy of the nD = 0 dimer singlet prod-
uct state,
EnD=0/ND = − 3
4 JD,
(9)
is known exactly and the equivalent for nD = 1 can be written
as the sum of the dimer triplet energy and the ground-state
energy of the S = 1 columnar dimer Heisenberg model,
4 JD + E S=1
CD (J, JP).
EnD=1/ND = 1
The energy E S=1
CD (J, JP) is readily accessible to QMC simula-
tions (Appendix B). It is convenient to introduce an angular
parametrization
CD (J, JP ) = E S=1
E S=1
CD (Jr cos θ , Jr sin θ )
(10)
= JrE S=1
CD (θ ),
(11)
where due to the structure of the Heisenberg model, the fac-
tor Jr can be pulled out. Using these steps, the form of the
first-order transition line in the (J, JP) plane can be written in
“polar coordinates” as
Jr = − JD
E S=1
CD (θ )
.
(12)
The resulting first-order line Jr (θ ) is shown in Fig. 2 to match
the observed crossover and first-order transition across the
full phase diagram, in the absence of any fitting parameters.
This can be considered indirect evidence for the correctness of
235128-4
THERMAL CRITICAL POINTS FROM COMPETING …
PHYSICAL REVIEW B 106, 235128 (2022)
lattice, a rigorous mapping to a classical two-dimensional
Ising model could be derived in order to exactly calculate
the thermal transition lines, which also compared well to the
results from the free-energy arguments [9].
We finally note that the phenomenology observed in the
pFFB model, i.e., a discontinuous quantum phase transition
separating two different quantum-disordered regimes and ex-
tending up to a thermal critical point with a comparably
low temperature scale, is similar to the thermal physics in
SrCu2(BO3)2 [6]. Indeed, the FFB lattice may be considered
as a fully frustrated extension [5,10] of the Shastry-Sutherland
model [19] that underlies the magnetism in SrCu2(BO3)2.
The original FFB model does however not feature a PS
phase (in contrast to the Shastry-Sutherland model [20]). As
we have shown, its plaquettized generalization considered
here contains a PS phase and furthermore realizes a discontin-
uous DS-to-PS transition, while its symmetry still protects us
from the severe QMC sign problem that hampers QMC simu-
lations of the Shastry-Sutherland model beyond the DS regime
[21]. Even though the PS phase of the pFFB model involves
no spontaneous symmetry breaking, one may nevertheless
consider the pFFB model a sign-problem free designer model
[22] for the specific thermal physics observed in Ref. [6] on
SrCu2(BO3)2, and considered here.
ACKNOWLEDGMENTS
We thank P. Corboz, A. Honecker, and B. Normand for
numerous discussions and collaborations on related topics.
We acknowledge support by the Deutsche Forschungsge-
meinschaft (DFG) through Grant No. WE/3649/4-2 of the
FOR 1807 and through RTG 1995, the Swiss National Sci-
ence Foundation through Grant No. 182179, the IT Center
at RWTH Aachen University and JSC Jülich for access to
computing time through the JARA Center for Simulation and
Data Science, and the Scientific IT and Application Support
Center of EPFL. The Flatiron Institute is a division of the
Simons Foundation.
APPENDIX A: DETAILS ON THE PERTURBATIVE
CALCULATION
In this Appendix we outline the details of the perturbative
calculation in the J (cid:5) JD ≈ JP regime. In this regime the
model is well described by weakly coupled plaquettes, and we
perform a perturbative downfolding to the low-energy S(cid:2) = 0
sector of each plaquette, consisting of the states
|α(cid:3) = |0, 0; 0, 0(cid:3),
(A1)
|β(cid:3) = 1√
3
(|1, +; 1, −(cid:3) + |1, −; 1, +(cid:3) − |1, 0; 1, 0(cid:3)),
(A2)
in the dimer basis of the two JD dimers contained in this
plaquette, |l1, m1; l2, m2(cid:3). As discussed in the main text, there
are no virtual processes that can renormalize the |α(cid:3) states so
the effective low-energy Hamiltonian can then be written to
FIG. 4. Sketch (not to scale) of the finite temperature phase
diagram showing the wall of discontinuities (red) of the finite-
temperature first-order transitions in the pFFB model. Near the
decoupled plaquette limit, the dependence of the critical temperature
Tc (bold red line) on the interplaquette coupling is quadratic. Its
behavior for small JP is purely indicative based only on its limiting
value Tc = 0 at JP = 0. The black lines represent the T = 0 continu-
ous transitions of the AFM phase to the PS and tubes phases, which
are expected to terminate at the wall of discontinuities.
our assumptions, and hence for the existence of the first-order
transition out of the DS phase for the full range of couplings.
Another point of interest is that Eq. (12) exactly matches
the weak-coupling perturbative result from before, as also
seen in Fig. 2. This can be understood by remembering that the
effect of the perturbations was limited to mixing the different
nD = 1 levels in the model. These levels all have effective spin
S = 1 and the perturbation theory is thus equivalent to doing
perturbation theory for the S = 1 columnar dimer square lat-
tice model.
V. CONCLUSION
From a combined analysis using unbiased QMC sim-
ulations and perturbation theory as well as free-energy
arguments, we derived the ground-state phase diagram of the
pFFB spin-1/2 Heisenberg model, and explored in particular
the emergence of a line of critical points that terminate a wall
of first-order phase transitions between the DS and the PS
low-temperature regimes. A sketch containing both the zero
temperature phase diagram as well as the wall of discontin-
uous first-order transitions and its line of critical points is
shown in Fig. 4.
From the perturbative approach, we derived that the corre-
sponding critical temperature scale in this regime is strongly
suppressed by its quadratic dependence on the interplaquette
coupling J, implying critical scales that fall well below the
temperature regime that is accessible to the QMC approach.
For the future it might be interesting to provide a similar
perturbative approach also in the regime at low JP. Here a
quantum phase transition takes place between the DS and
the regime of one-dimensional J tubes, with Tc = 0 in the
decoupled tube limit JP = 0 (how this limiting value of Tc is
approached as JP → 0 would be interesting to extract from
such a perturbative approach). Based on the free-energy ar-
guments, we obtained an estimate for the phase boundary of
the DS phase that is in remarkable agreement with the results
from the QMC simulations. It is worthwhile to note that for a
spin-1/2 Ising-Heisenberg on the diamond-decorated square
235128-5
WEBER, FACHE, MILA, AND WESSEL
PHYSICAL REVIEW B 106, 235128 (2022)
D
J
/
P
J
D
J
/
P
J
1.5
1.0
0.5
0.0
1.5
1.0
0.5
0.0
T /JD = 0.2
T /JD = 0.3
T /JD = 0.2
T /JD = 0.3
0.0
0.5
0.0
J/JD
0.5
J/JD
1.00
0.75
0.50
D
n
0.25
0.00
40
20
)
π
,
π
(
S
0
FIG. 6. Dimer triplet density nD (top panels) and AFM structure
factor S(π , π ) (bottom panels) of the pFFB model as functions of J
and JP at fixed temperatures T /JD = 0.2 (left panels) and T /JD = 0.3
(right panels) on a system with L = 12.
JD dimers in plaquette (cid:2) ((cid:2)(cid:4)). The resulting matrix products
can be readily evaluated and—after writing the projectors on
|α(cid:3) and |β(cid:3) in terms of Pauli matrices—yield the Hamiltonian
in Eq. (4).
APPENDIX B: S = 1 COLUMNAR DIMER MODEL
In the energy argument for the first-order line, in Eq. (12),
the ground-state energy E S=1
CD of the columnar dimer model
(inset of Fig. 5) appears. This quantity is readily accessible in
quantum Monte Carlo simulations, which we present in the
main panel of Fig. 5. The QMC simulations were performed
using the standard SSE QMC algorithm [12–15] in the S =
1 Sz basis. The energy is found to be well converged at L =
60, where L is the linear system size of the N = 2L2 sites
spin-1 system.
APPENDIX C: HIGHER TEMPERATURES
In this Appendix we present in Fig. 6 additional results
for the dimer triplet denstiy nD and the AFM structure factor
S(π , π ) taken at higher temperatures.
FIG. 5. Energy per site E S=1
CD of the S = 1 Heisenberg model on
the columnar dimer lattice (inset) parametrized by the angle θ for
different system sizes L. The temperature was scaled as T = 1/2L
to probe ground-state properties. Black vertical lines denote the
boundaries of the central AFM regime from Ref. [17].
second order in J/JP as
(cid:2)
(cid:2)
Heff =
εp |p(cid:3)(cid:2)p|(cid:2) −
(cid:2)
p=α,β
(cid:2)
(cid:2)(cid:2),(cid:2)(cid:4)(cid:3)
|ββ(cid:3)(cid:2)ββ|(cid:2),(cid:2)(cid:4)
× (cid:2)ββ|H(cid:2),(cid:2)(cid:4) P
1
H(cid:2) + H(cid:2)(cid:4) − 2εβ
PH(cid:2),(cid:2)(cid:4) |ββ(cid:3) ,
(A3)
where
H(cid:2) = JPT1 · T2 + const
(A4)
is the single-plaquette Hamiltonian, with H(cid:2) |α(cid:3) = εα |α(cid:3),
H(cid:2) |β(cid:3) = εβ |β(cid:3), and
P = 1 − |ββ(cid:3)(cid:2)ββ|
(A5)
is a projector on the high-energy subspace (all in the nD = 1
sector), and lastly
(cid:10)
H(cid:2),(cid:2)(cid:4) = J
T1 · T1(cid:4) + T2 · T2(cid:4) ,
T1 · T2(cid:4) ,
if (cid:2)(cid:4) = (cid:2) + ˆx,
if (cid:2)(cid:4) = (cid:2) + ˆy.
(A6)
Here T1 and T2 (T1(cid:4) and T2(cid:4) ) denote the total spins of the two
[1] J. Richter, J. Schulenburg, and A. Honecker, Quantum mag-
netism in two dimensions: From semi-classical Néel order
to magnetic disorder, in Quantum Magnetism, edited by U.
Schollwöck, J. Richter, D. J. J. Farnell, and R. F. Bishop
(Springer, Berlin, 2004), pp. 85–153.
[2] L. Balents, Nature (London) 464, 199 (2010).
[3] C. Lacroix, P. Mendels, and F. Mila, Introduction to Frustrated
Magnetism: Materials, Experiments, Theory, Springer Series in
Solid-State Sciences (Springer, Berlin, 2011), Vol. 164.
[4] H. T. Diep, Frustrated Spin Systems, 2nd edition (World Scien-
[5] J. Stapmanns, P. Corboz, F. Mila, A. Honecker, B.
Normand, and S. Wessel, Phys. Rev. Lett. 121, 127201
(2018).
[6] J. Larrea Jiménez, S. P. G. Crone, E. Fogh, M. E. Zayed, R.
Lortz, E. Pomjakushina, K. Conder, A. M. Läuchli, L. Weber, S.
Wessel, A. Honecker, B. Normand, C. Rüegg, P. Corboz, H. M.
Rønnow, and F. Mila, Nature (London) 592, 370 (2021).
[7] L. Weber, A. Honecker, B. Normand, P. Corboz, F. Mila, and S.
Wessel, SciPost Phys. 12, 054 (2022).
[8] N. Caci, K. Karˇlová, T. Verkholyak, J. Streˇcka, S. Wessel, and
tific, Singapore, 2013).
A. Honecker, arXiv:2212.01278.
235128-6
THERMAL CRITICAL POINTS FROM COMPETING …
PHYSICAL REVIEW B 106, 235128 (2022)
[9] J. Streˇcka, K. Karˇlová, T. Verkholyak, N. Caci, S. Wessel, and
[16] F. Alet, S. Wessel, and M. Troyer, Phys. Rev. E 71, 036706
A. Honecker, arXiv:2212.01278.
(2005).
[10] E. Müller-Hartmann, R. R. P. Singh, C. Knetter, and G. S.
[17] M. Matsumoto, C. Yasuda, S. Todo, and H. Takayama, Phys.
Uhrig, Phys. Rev. Lett. 84, 1808 (2000).
Rev. B 65, 014407 (2001).
[11] A. Honecker, S. Wessel, R. Kerkdyk, T. Pruschke, F. Mila, and
B. Normand, Phys. Rev. B 93, 054408 (2016).
[18] L. Onsager, Phys. Rev. 65, 117 (1944).
[19] B. Sriram Shastry and B. Sutherland, Physica B+C 108, 1069
[12] F. Alet, K. Damle, and S. Pujari, Phys. Rev. Lett. 117, 197203
(1981).
(2016).
[13] A. W. Sandvik and J. Kurkijärvi, Phys. Rev. B 43, 5950 (1991).
[14] A. W. Sandvik, Phys. Rev. B 59, R14157 (1999).
[15] O. F. Syljuåsen and A. W. Sandvik, Phys. Rev. E 66, 046701
(2002).
[20] P. Corboz and F. Mila, Phys. Rev. B 87, 115144 (2013).
[21] S. Wessel, I. Niesen, J. Stapmanns, B. Normand, F. Mila, P.
Corboz, and A. Honecker, Phys. Rev. B 98, 174432 (2018).
[22] R. K. Kaul, R. G. Melko, and A. W. Sandvik, Annu. Rev.
Condens. Matter Phys. 4, 179 (2013).
235128-7
| null |
10.1371_journal.pntd.0009999.pdf
|
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
|
All relevant data are within the paper and its Supporting Information files.
|
RESEARCH ARTICLE
Multimodal biomarker discovery for active
Onchocerca volvulus infection
1*, Emmanuel Njumbe Ediage2, Dirk Van Roosbroeck2, Stijn Van Asten2,
Ole LagatieID
Ann Verheyen1, Linda Batsa Debrah3, Alex Debrah4, Maurice R. Odiere5, Ruben T’Kindt6,
Emmie DumontID
1
Filip Cuyckens2, Lieven J. StuyverID
6, Koen Sandra6, Lieve DillenID
2, Tom VerhaegheID
2, Rob VreekenID
2,
1 J&J Global Public Health, Janssen R&D, Beerse, Belgium, 2 Discovery Sciences, Janssen R&D, Beerse,
Belgium, 3 Department of Clinical Microbiology, School of Medicine and Dentistry, Kwame Nkrumah
University of Science and Technology, Kumasi, Ghana, 4 Faculty of Allied Health Sciences, Kwame
Nkrumah University of Science and Technology, Kumasi, Ghana, 5 Kenya Medical Research Institute,
Centre for Global Health Research, Kisumu, Kenya, 6 Research Institute for Chromatography (RIC), Kortrijk,
Belgium
* [email protected]
Abstract
The neglected tropical disease onchocerciasis, or river blindness, is caused by infection
with the filarial nematode Onchocerca volvulus. Current estimates indicate that 17 mil-
lion people are infected worldwide, the majority of them living in Africa. Today there are
no non-invasive tests available that can detect ongoing infection, and that can be used
for effective monitoring of elimination programs. In addition, to enable pharmacodynamic
studies with novel macrofilaricide drug candidates, surrogate endpoints and efficacy bio-
markers are needed but are non-existent. We describe the use of a multimodal untar-
geted mass spectrometry-based approach (metabolomics and lipidomics) to identify
onchocerciasis-associated metabolites in urine and plasma, and of specific lipid features
in plasma of infected individuals (O. volvulus infected cases: 68 individuals with palpable
nodules; lymphatic filariasis cases: 8 individuals; non-endemic controls: 20 individuals).
This work resulted in the identification of elevated concentrations of the plasma metabo-
lites inosine and hypoxanthine as biomarkers for filarial infection, and of the urine metab-
olite cis-cinnamoylglycine (CCG) as biomarker for O. volvulus. During the targeted
validation study, metabolite-specific cutoffs were determined (inosine: 34.2 ng/ml; hypo-
xanthine: 1380 ng/ml; CCG: 29.7 ng/ml) and sensitivity and specificity profiles were
established. Subsequent evaluation of these biomarkers in a non-endemic population
from a different geographical region invalidated the urine metabolite CCG as biomarker
for O. volvulus. The plasma metabolites inosine and hypoxanthine were confirmed as
biomarkers for filarial infection. With the availability of targeted LC-MS procedures, the
full potential of these 2 biomarkers in macrofilaricide clinical trials, MDA efficacy surveys,
and epidemiological transmission studies can be investigated.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Lagatie O, Njumbe Ediage E, Van
Roosbroeck D, Van Asten S, Verheyen A, Batsa
Debrah L, et al. (2021) Multimodal biomarker
discovery for active Onchocerca volvulus infection.
PLoS Negl Trop Dis 15(11): e0009999. https://doi.
org/10.1371/journal.pntd.0009999
Editor: Krystyna Cwiklinski, National University of
Ireland Galway, IRELAND
Received: September 24, 2021
Accepted: November 16, 2021
Published: November 29, 2021
Copyright: © 2021 Lagatie et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: The author(s) received no specific
funding for this work.
Competing interests: I have read the journal’s
policy and the authors of this manuscript have the
following competing interests: Ole Lagatie,
Emmanuel Njumbe Ediage, Dirk Van Roosbroeck,
Stijn Van Asten, Ann Verheyen, Lieve Dillen, Tom
Verhaeghe, Rob Vreeken, Filip Cuyckens and
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
1 / 20
PLOS NEGLECTED TROPICAL DISEASES Lieven J. Stuyver are current employees of
Janssen Pharmaceutica NV, a Johnson & Johnson
company, and may own stock or stock option in
that company.
Discovery of novel biomarkers for active Onchocerca volvulus infection
Author summary
Today’s diagnosis of infection with the filarial parasite Onchocerca volvulus mainly
depends on the microscopic analysis of skin biopsies and serological testing. The work
presented here describes the use of multiple mass spectrometry-based screening methods
(metabolomics and lipidomics) to search for biomarkers indicative of infection with
Onchocerca volvulus. This resulted in the identification of elevated concentrations of the
plasma metabolites inosine and hypoxanthine as biomarkers for filarial infection, and of
the urine metabolite cis-cinnamoylglycine as biomarker for O. volvulus. Further evalua-
tion of these biomarkers in a geographically distinct non-endemic population however
invalidated the use of urine cis-cinnamoylglycine. These findings are of utmost impor-
tance as it not only opens new avenues in the development of non-invasive diagnostic
tools for filarial infections, but also emphasizes the need for evaluation and validation of
newly discovered biomarkers in different populations from different geographies.
Introduction
Onchocerciasis, or river blindness, is an infectious disease caused by the filarial parasitic nema-
tode Onchocerca volvulus with an estimated prevalence of current infection of 17 million peo-
ple worldwide and 120 million people at risk. Although transmission occurs in the African
Region, the Region of the Americas and the Eastern Mediterranean Region, 99% of infected
people live in 31 African countries [1]. Life cycle stages of O. volvulus in the human host con-
sist of adult worms called macrofilaria, and microfilaria. While the macrofilaria accumulate in
subcutaneous onchocercomas, microfilaria migrate through the skin, eyes and other organs.
Symptoms of the disease; rash, itching, skin lesions and eye lesions that ultimately can lead to
blindness are the result of the host’s inflammatory response to dying microfilariae [2]. Treat-
ment of onchocerciasis is mainly based on mass drug administration (MDA) through Com-
munity Directed Treatment with Ivermectin (CDTi) aimed at breaking the transmission cycle
in affected communities [3,4]. Alternatively, the antibiotic doxycycline targets the bacterial
endosymbiont Wolbachia, resulting in sterilization and to some extent also death of adult
worms [5]. To be able to monitor and evaluate these MDA programs, epidemiological map-
ping is performed to identify all high-risk areas where ivermectin treatment is needed. These
mappings are mainly based on examination of individuals for the presence of palpable oncho-
cercomas, presence of microfilariae (mf) in skin biopsies, and also the rapid diagnostic test
(RDT) for the detection of IgG4 antibodies to the parasitic antigen Ov16 [6–14]. However, the
invasive nature of skin biopsies makes it increasingly unpopular while antibody-based tests
have their limitations [15]. To improve the sensitivity of the existing serological tests, a com-
bined test for Ov16 and OVOC3261 IgG4 detection was proposed [16]. Since repetitive annual
ivermectin treatment is required to prevent further pathology caused by newly produced
microfilariae, efforts have been undertaken to develop drugs with macrofilaricidal activity,
directly targeting the adult worms [17–20]. There is a need for surrogate markers of infection
and preferably of the presence of live and active adult worms [15].
Recently, a WHO report was made available describing the target product profile (TPP) to
support preventive chemotherapy [21]. The minimal target analyte to be detected is an antigen
or other biomarkers specific for live, adult female worms. A diagnostic clinical sensitivity of
�60% was deemed sufficient, while a clinical specificity of � 99.8% was found to be necessary.
To overcome the shortcomings of the currently available diagnostic tools, a number of studies
have been performed to identify metabolites in blood and urine that reflect infection status
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
2 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
and possibly also intensity of infection [22–25]. The most promising metabolite that was dis-
covered so far was the neurotransmitter derived N-acetyltyramine-O-glucuronide (NATOG)
in urine [23,26–30]. Parasite-derived DNA or microRNAs have also been proposed as possible
blood-based biomarkers for onchocerciasis, but were found to have limited utility [31–33].
Also new serological markers have been proposed, such as the peptide makers OvMP-1 and
OvMP-23, and OvNMP-48 [34–38]. The use of these peptide markers was subsequently found
to be limited (specificity significantly less than the required �99.8%), due to unexplained
cross-reactivity in a population of school-age children in a non-endemic area in southwest
Kenya [39]. In the work presented here, we conducted a study using both plasma and urine
samples from nodule positive individuals that had very low or negative mf counts due to treat-
ment with ivermectin. Mass spectrometric methods, specifically designed to detect a large
range of small molecules, i.e. metabolites or lipids, were applied to allow untargeted identifica-
tion of parasite derived molecules or host response markers. In a second phase, targeted liquid
chromatography coupled to mass spectrometry (LC-MS) methods were developed and used to
assess the concentrations of selected features in an extended validation sample set, leading to
the confirmation of 3 candidate biomarkers, namely plasma hypoxanthine, plasma inosine
and urine cis-cinnamoylglycine (CCG) [40]. The cinnamoylglycine candidate biomarker for
onchocerciasis was more recently also identified by others [41]. In this study, we describe the
full discovery process of these candidate biomarkers and evaluate the value of their perfor-
mance considering the TPP from WHO in a sample set collected in a geographically distinct
non-endemic region.
Results
Selection of sample sets for biomarker discovery and biomarker validation
We envisioned identifying biomarkers for Onchocerca volvulus infection and more in particu-
lar for the presence of macrofilaria. For all samples from Ghana (n = 253), an Ov16 RDT was
performed. Based on this test, 68 of the 98 (69.4%) nodule positive individuals (NP), 26 of the
51 (51.0%) endemic controls (EC), 9 of the 54 (16.7%) non-endemic controls (NEC), and 12 of
the 50 (24.0%) lymphatic filariasis patients (LF) were found to be seropositive. Given the high
specificity reported for this test (97–98%), these data demonstrated that in the LF and NEC
groups some onchocerciasis occurred, although at a lower prevalence than in the O. volvulus
endemic population [8,9,42].
The discovery sample set consisted out of biomaterials of 68 nodule positive individuals
that were Ov16 positive. The non-endemic control group consisted out of samples from 20
individuals that were Ov16 negative. The discovery set was further completed with 8 LF
infected individuals that were Ov16 negative. An overview of the samples that were selected to
be used in the discovery study is presented in S1 Table.
The validation sample set consisted out of the entire collection (n = 253) described above,
complemented with biomaterials of 50 Belgian healthy controls.
Biomarker discovery using untargeted approaches
All employed untargeted methodologies for comparative profiling of lipids (lipidomics) or
metabolites (metabolomics) in plasma or urine, resulted in several features that met the pre-
defined criteria (See S1 Supplementary Materials and Methods). All features detected using
LC-MS were subjected to recursion analysis to remove false positives and a final list of features
was prepared. For gas chromatography (GC)-MS analyses, only features that could be identi-
fied based on the available libraries were retained. Table 1 summarizes the number of features
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
3 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
Table 1. Number of features retained throughout the successive data-processing steps for the different methodologies employed. LC-MS based analyses were per-
formed both in the positive and negative electrospray ionization mode (referred to as ESI+ and ESI-, respectively).
Data processing step
Feature number
Plasma
Lipidomics
Metabolomics
Urine
Metabolomics
ESI +
ESI -
ESI +
ESI -
GC-MS
ESI +
ESI -
GC-MS
Features extracted from full cohort
23,247
14,182
36,116
34,344
167
14,757
35,143
200
Features highly upregulated in nodule positives individuals
Upregulated features in NP shared with LF patients
Features retained after recursion (for LC-MS) or identification (for GC-MS)
56
37
19
81
28
21
58
42
14
87
57
35
27
2
22
44
27
5
146
93
25
28
1
16
https://doi.org/10.1371/journal.pntd.0009999.t001
retained throughout the successive data-processing steps. The final lists of features that were
retained as candidate biomarkers for onchocerciasis are presented in S2–S6 Tables.
For each approach (urine metabolomics, plasma metabolomics and plasma lipidomics),
two features were selected for further investigation:
a) Plasma PI(16:0/14:0) and PC(12:0/14:0)
In the lipidomics analysis, among the 34 features retained (S2 Table), the occurrence of sev-
eral features with C12 and C14 fatty acid chains is notable. Two of them, namely phosphati-
dylinositol (PI)(16:0/14:0) and phosphatidylcholine (PC) (12:0/14:0) show more than
20-fold change between NEC and NP (Fig 1A). Clearly, these lipids are not specific for O.
volvulus infection but rather reflect infection with a filarial helminth (both LF and oncho-
cerciasis). Whereas PI(16:0/14:0) is markedly elevated in plasma from infected individuals,
it is also present in lower quantities in the non-endemic controls. The occurrence of PC
(12:0/14:0) in plasma however appears to be unique for individuals with filarial infection.
b) Plasma hypoxanthine and inosine
In the metabolomics analysis of the plasma samples, 71 features were retained (S3 and S5
Tables). Many of these have an unknown structure, but for several other features a
Fig 1. Selection of biomarkers associated to onchocerciasis. (A) Results for plasma lipids PI(16:0/14:0) and PC(12:0/14:0) in the different groups. (B) Results for plasma
metabolites hypoxanthine and inosine in the different groups. (C) Results for urine metabolites cinnamoylglycine and hippuric acid in the different groups. All results are
expressed in peak area. Abbreviations: NP: Nodule Positive; NEC: Non-Endemic Control; LF: Lymphatic Filariasis.
https://doi.org/10.1371/journal.pntd.0009999.g001
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
4 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
structure could be derived from the MS/MS spectrum. Two of the most discriminating
markers (based on fold change and statistical significance) could be identified as hypoxan-
thine and inosine. Both metabolites were elevated both in onchocerciasis and LF patients
(Fig 1B). Hypoxanthine levels are significantly higher in LF patients compared to oncho-
cerciasis (P = 0.0008). Inosine levels are similar in both patient populations (P = 0.06) but
markedly higher than in in the NEC (P<0.0001 both for onchocerciasis and LF).
c) Urine cinnamoylglycine and hippuric acid
Upon analysis of the urine metabolome, a total of 46 features were retained for further analy-
sis (S4 and S6 Tables). Besides several features with unknown molecular structure, also for
urine many could be identified based on the MS/MS spectrum. Two of the most discriminat-
ing markers were identified as hippuric acid and cinnamoylglycine (Fig 1C). These molecules
appear to be only elevated in urine from onchocerciasis patients and not from LF patients.
Biomarker confirmation in discovery sample set using targeted approaches
Targeted LC-MS methods were developed for (i) PI(16:0/14:0) and PC(12:0/14:0) in plasma,
(ii) for hypoxanthine and inosine in plasma, and (iii) for hippuric acid and cinnamoylglycine
in urine. To proof the structure of the identified biomarkers and to be able to determine them
quantitatively, synthetic reference material was obtained and used to develop the targeted
LC-MS methods. The same plasma or urine sample extracts as used in the discovery study
were re-analyzed using the targeted methods, and results of the comparison between both data
sets are shown in Table 2 and S1 Fig. Also, receiver operating characteristic (ROC) analysis
was performed on the data obtained using the targeted assays and Area Under Curve (AUC)
values were calculated (S2 Fig). Only markers that could be considered good or excellent
markers (AUC values above 0.80 or 0.90, respectively) were retained for further validation
[43].
a) Plasma PI(16:0/14:0) and PC(12:0/14:0)
For the plasma lipids PI(16:0/14:0) and PC(12:0/14:0), a moderate correlation was obtained
between both data sets (r2 = 0.48 and 0.57, respectively) resulting also in markedly reduced
AUC in the ROC analysis (AUC = 0.77 and 0.71, respectively). Although generally the
same trend as in the discovery data set is still apparent, based on the AUC values, these
markers could be considered only fair markers (AUC 0.70–0.80). It was therefore decided
not to further explore both lipids as markers for active O. volvulus infection.
Table 2. Comparison between untargeted and targeted analysis of candidate biomarkers. Correlation curves were prepared on log-transformed data based on peak
area from the untargeted analysis (X) and concentration (ng/mL) of the targeted analysis (Y) and correlation coefficients (r2) were calculated. Based on the data of the tar-
geted analysis, ROC analysis was performed with NP and NEC as cases and controls, respectively and AUC was calculated.
Plasma markers
PI(16:0/14:0)
PC(12:0/14:0)
Hypoxanthine
Inosine
Urine markers
Hippuric acid
Trans-cinnamoylglycine
Cis-cinnamoylglycine
https://doi.org/10.1371/journal.pntd.0009999.t002
Equation
Y = 0.7172�X - 1.953
Y = 0.7968�X - 2.976
Y = 0.9996�X - 3.027
Y = 1.165�X - 5.292
Y = 0.1184�X + 7.645
Y = 0.3664�X + 3.178
Y = 0.7487�X + 0.1196
r2
0.4813
0.5653
0.9263
0.9622
0.05429
0.4226
0.854
AUC [NP vs NEC]
0.7720
0.7148
0.8612
0.9465
0.7022
0.7222
0.9034
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
5 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
b) Plasma hypoxanthine and inosine
For the plasma metabolites hypoxanthine and inosine correlation between both data sets
were within preset acceptance criteria (r2 = 0.93 and 0.96, respectively), confirming the
identity of these molecules and validating the use of the targeted LC-MS method. With
AUC values of 0.86 and 0.95, for hypoxanthine and inosine respectively, both features were
considered of interest for further evaluation.
c) Urine cinnamoylglycine and hippuric acid
For the urine metabolites, the data of hippuric acid could not be confirmed using the tar-
geted method (r2 = 0.05). Cinnamoylglycine occurs both as cis- and trans-isomer and since
it could not be deduced from the discovery experiments which isomer was identified, both
synthetic molecules were included. Targeted LC-MS analysis demonstrated that the cis-iso-
mer was the urinary biomarker for onchocerciasis, as the data obtained for the cis-isomer
were concordant with the discovery data (r2 = 0.85) while for the trans-isomer this was
weak (r2 = 0.42). For the cis-isomer, eight samples that were undetectable in the untargeted
method, now had low but detectable levels of cis-cinnamoylglycine (CCG). This discrep-
ancy might be due to a difference in sensitivity of the targeted method compared to the
untargeted method. Exclusion of these discrepant samples would result in a r2 of 0.96, re-
confirming the identity of the discovered feature to be the cis-isomer of cinnamoylglycine
and warranting its further evaluation in the validation set. ROC analysis of the targeted
analysis data re-confirmed the diagnostic potential of CCG with an AUC of 0.90.
In conclusion, the conversion of the biomarker features from the untargeted approach into
confirmed features using a targeted LC-MS approach was successful for CCG, inosine, and
hypoxanthine. The PI(16:0/14:0), PC(12:0/14:0), and hippuric acid biomarker features did not
meet the preset acceptance criteria, and were hence not further evaluated in the validation set.
Biomarker validation in validation sample set
The levels of hypoxanthine and inosine in plasma, and of CCG in urine were further evaluated
in the validation sample set (Fig 2). Based on these data, ROC analysis was performed, bio-
marker specific cutoffs were defined, and diagnostic characteristics were determined (Table 3).
Since the discovery sample set confirmed that both plasma biomarkers hypoxanthine and ino-
sine are elevated in plasma of nodule positive individuals as well as in LF infected individuals,
ROC analysis for these markers was performed using the Belgian healthy controls and the
non-endemic controls as negative panel and the nodule positives, LF infected individuals and
endemic controls as positive panel. The urine biomarker CCG appeared to be specifically ele-
vated in the onchocerciasis group and not in the LF group, based on the discovery sample set,
and therefore ROC analysis was performed using the Belgian healthy controls, the non-
endemic controls and LF infected individuals as negative panel and the nodule positives and
endemic controls as positive panel (S3 Fig).
a) Plasma hypoxanthine and inosine
Based on the ROC analysis, cutoffs for plasma hypoxanthine and inosine were set at 1380
ng/mL, and 34.2 ng/mL, respectively. Based on these cutoffs, hypoxanthine had a sensitivity
of 86.2% and a specificity of 89.2%, while for inosine sensitivity was 74.5% and specificity
was 95.7%. The obtained quantitative results confirm our initial observation that both
hypoxanthine and inosine are elevated in plasma of nodule positive individuals, endemic
controls as well as in LF patients compared to non-endemic controls (P<0.0001). These
data also confirm the more pronounced elevation of hypoxanthine and inosine in LF
patients compared to onchocerciasis patients (P<0.0001). Based on hypoxanthine levels,
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
6 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
Fig 2. Validation of biomarkers associated to onchocerciasis. Results for the three new biomarkers (plasma hypoxanthine and inosine; and urine CCG) and for NATOG
that have been obtained on the nodule positive individuals (NP, blue), endemic controls (EC, purple), LF patients (LF, green), non-endemic controls (NEC, red) and
healthy controls from Belgium (HC, orange). For each plot, the grey zone indicates the zone between limit of detection (LOD) and limit of quantification (LOQ), the
dashed line indicates the biomarker-specific cutoff and the yellow zone indicates the zone between 0.5 log times the cut-off and the maximum value observed for the
specific biomarker. In case of NATOG, the cutoff (4.63 μg/mL = 13 μM) and maximum value (98.3 μg/mL = 276 μM) was derived from the data published by Globisch and
colleagues [27]. NATOG data were previously obtained on the same sample set (doi: 10.1186/s13071-016-1582-6) [28]. For each marker, the percentage of positive samples
in the group considered to be infected (i.e. sensitivity) and the percentage of negative samples in the group considered to be not infected (i.e. specificity), is indicated. For
CCG and NATOG, which are onchocerciasis specific biomarkers, the LF group was plotted separately from the other control samples to highlight specificity.
https://doi.org/10.1371/journal.pntd.0009999.g002
no difference could be observed between nodule positive individuals and endemic controls
(P = 0.2578). For inosine, levels are significantly lower in the endemic control group
(P = 0.0010), but with a large overlap between both groups.
Table 3. Diagnostic characteristics of the biomarkers, as determined on the validation sample set. Sensitivity for filarial markers was based on NP, LF and EC, while
specificity was based on NEC and HC. For onchocerciasis markers, sensitivity was based on NP and EC, with specificity based on NEC, HC and LF.
AUROC
Cutoff
(ng/mL)
Sensitivity
(%)
Specificity
(%)
NP positive
(%)
LF positive
(%)
EC positive
(%)
HC positive
(%)
NEC
positive (%)
Plasma filariasis markers
Number of samples included in each group (n)
Hypoxanthine
Inosine
0.93
0.91
1380
34.2
86.2%
74.5%
89.2%
95.7%
Urine onchocerciasis markers
Number of samples included in each group (n)
Cis-cinnamoylglycine
0.87
29.7
82.9%
82.2%
95
85.3%
74.7%
96
88.5%
50
100.0%
94.0%
48
29.2%
51
74.5%
52.9%
50
72.0%
49
4.1%
0.0%
44
0.0%
53
18.9%
9.4%
52
23.1%
Abbreviations: NP: nodule positive; LF: Lymphatic Filariasis; EC: Endemic Control; HC: Healthy Control; NEC: Non-Endemic Control
https://doi.org/10.1371/journal.pntd.0009999.t003
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
7 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
b) Urine CCG
The data obtained for urinary CCG in the validation sample set reinforce the potential of
this urinary metabolite as a biomarker for onchocerciasis. Based on the ROC analysis, a cut-
off was defined at 29.7 ng/mL, resulting in 82.9% sensitivity and 82.2% specificity. The NEC
samples had levels that were higher than the healthy control samples (P<0.0001), with 12 of
the 52 non-endemic control samples above the cutoff, hence considered false positive. This
observation might need further evaluation. Also in the LF group, 14 out of 48 samples
appeared to be positive for urinary CCG. None of the healthy control samples were positive
for CCG. Similar to plasma inosine, CCG levels are largely overlapping in the nodule posi-
tive and endemic control groups, with on average lower concentrations in the endemic con-
trol group (P = 0.0449). When assessing both groups separately, 88.5% of the nodule positive
individuals was positive while for the endemic control group this was only 72.0%.
Evaluation of the biomarkers in a non-endemic population from Kenya
In order to evaluate the specificity of the new biomarkers, the levels of hypoxanthine and ino-
sine in plasma, and of CCG in urine were evaluated in a sample set collected in the southwest
part of Kenya (Fig 3). Kenya is non-endemic for O. volvulus and lymphatic filariasis is mainly
confined to the coastal region, which is different from the region where this sample set has
been collected [44–47]. Out of 476 study participants, 3.8% and 4.7% were found to be positive
for plasma hypoxanthine and inosine, respectively. Also, Chi square analysis indicated that the
Kenyan population was indistinguishable from the negative validation set based on plasma
inosine levels (P > 0.9999). Based on plasma hypoxanthine levels, there was a weak statistical
difference with the negative validation set (P = 0.015) but this appears to be caused by the
rather high number of false positives in the validation set. These data confirm the biomarker
potential for plasma hypoxanthine and inosine as filarial markers. On the other hand, 51.6% of
the Kenyan participants were found to be positive for CCG in urine. The CCG levels in this
non-endemic population overlapped almost entirely with the levels detected in both the
Fig 3. Levels of plasma hypoxanthine, plasma inosine, and urine CCG in a non-endemic population from Kenya, compared to the positive and negative population
from the validation study. The dashed lines indicate the biomarker-specific cutoffs. The percentage of positive samples in each group is indicated as well as P-value of Chi
square analysis for each metabolite comparing the Kenyan population with the negative and positive validation set, respectively.
https://doi.org/10.1371/journal.pntd.0009999.g003
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
8 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
positive and negative sample set from the validation study. This observation suggests that
urine CCG is not a good biomarker for onchocerciasis. Since many of the study participants
were infected with soil-transmitted helminths or Schistosoma mansoni, we grouped samples
based on these infections. No significant difference between the non-infected and different
infection groups was observed (Pχ
are linked to elevated CCG levels. The urine CCG data were also compared with the peptide
serology data previously obtained from the same study population [39]. None of the 3 peptide
markers correlated with urine CCG, with P-values from Chi square analysis ranging from
0.2362 to >0.999 (Fig 4). This observation indicated that the elevated excretion of CCG and
positive peptide serostatus in this population are not caused by the same study- or region-spe-
cific factor.
2 = 0.7042), suggesting that none of these intestinal parasites
Discussion
Onchocerciasis remains an important health issue in several African countries, despite the
MDA programs that have been put in place. To better steer these programs, novel tools for epi-
demiological mapping are urgently needed, besides skin biopsies and antibody-based tests,
with specifications as presented in the TPP from WHO [21]. Also, for the development of
macrofilaricide drugs, good pharmacodynamic markers will be required to monitor the effect
of these drug candidates on the adult worms. In the work presented here, we have used urine
and plasma metabolomics and plasma lipidomics approaches to identify novel molecules that
have potential as diagnostic markers.
In plasma, 2 molecules were identified with promising diagnostic characteristics: the metab-
olites hypoxanthine and inosine. Both plasma markers were found to be non-specific for oncho-
cerciasis but were rather indicative of a filarial infection. With a sensitivity of 86.2% and 74.5%,
and a specificity of 89.2% and 95.7%, respectively for hypoxanthine and inosine, both markers
Fig 4. The percentage of individuals that were urine CCG positive stratified according to their peptide serology status OvMP-1, OvMP-23 and OvNMP-48 [39]. Red
bars indicate number of samples positive for urine CCG, blue bars indicate number of samples negative for urine CCG. For each peptide the P-value of Chi square analysis
is indicated.
https://doi.org/10.1371/journal.pntd.0009999.g004
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
9 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
might warrant further research into the clinical utility as filarial markers. For LF specifically, the
data indicate that hypoxanthine is a stronger biomarker than inosine as hypoxanthine had a
100% sensitivity for detecting LF, while for inosine only 94.0% of samples had levels above the
cutoff. Both metabolites are products of the purine degradation pathway, with inosine the first
step in the catabolism of adenosine, which is then being further degraded into hypoxanthine
[48,49]. In man, this is then further metabolized into uric acid by the enzyme xanthine oxidase,
and subsequently excreted in urine [50]. Helminths however lack this enzyme and consequently
have hypoxanthine as end product of their purine metabolism [51]. Both molecules were also
identified as being significantly upregulated in a metabolite profiling study that was performed
on plasma samples from microfilaridermic patients (> 50 mf/mg skin) [24]. The fact that also
in our study population with no or very low levels of mf in the skin, a similar increase is
observed, might be indicative that the adult worm is (partly) responsible for the accumulation
of both metabolites. Why plasma hypoxanthine levels–and to some extent also plasma inosine
levels–in LF patients are even further increased is not clear based on the currently available
data, but it is possible that differences in infection intensity play a role.
In urine from individuals residing in an endemic area, CCG was found to be specifically
upregulated in onchocerciasis patients and not in LF patients, with a sensitivity of 82.9% and a
specificity of 82.2%. Cinnamoylglycine is one of the metabolites that is produced upon degra-
dation of cinnamic acid or one of its derivatives, such as e.g. caffeic acid and ferulic acid [52].
Whereas its most abundant metabolite, hippuric acid, does not form stereoisomers, the minor
metabolite cinnamoylglycine can occur in both trans- and cis-configuration. Since most cin-
namic acid present in nature is trans-cinnamic acid, this will also give rise to trans-cinnamoyl-
glycine [53]. We found indeed that in the western healthy control population, all (100%) urine
samples contained very low levels of CCG, with maximal level detected at 21.6 ng/mL, which is
still substantially lower than the cutoff that was set at 29.7 ng/mL. In the onchocerciasis
endemic population that was investigated, 82.9% of all urine samples contained levels above
this cutoff. In the group of nodule positive individuals specifically, this was even 88.5%. How-
ever, investigation of a non-endemic population from Kenya, a country declared free of
onchocerciasis [44], demonstrated that urine CCG was also detected at similarly high levels in
more than half of the tested individuals, suggesting that CCG excretion in urine of individuals
in Kenya is not related to O. volvulus infection.
We have previously reported on the discovery and value of these biomarkers in Onchocerca
endemic areas [40]. In a more recent publication, Wewer et al. also identified cinnamoylgly-
cine as a candidate biomarker for onchocerciasis in endemic areas [41]. Although not further
investigated and hence not confirmed, it is very well possible that the marker they identified is
in fact also cis-cinnamoylglycine. It is difficult to compare the data from both studies as no
quantitative method has been used to determine cinnamoylglycine in the Wewer et al. study.
However, the authors observed that the difference between individuals with onchocerciasis
and non-infected controls was not significant and that only 17.2% of the onchocerciasis group
had urine cinnamoylglycine levels higher than the highest value of the control samples. Taken
together, the authors concluded that cinnamoylglycine is suitable to identify infected individu-
als with very high metabolite levels, but with a large variation. Our results add a further restric-
tion to that observation, namely that the CCG cannot be considered as a diagnostic marker for
onchocerciasis when outside of the endemic area. Given the WHO TPP requirements on spec-
ificity, CCG should not be considered as a useful contribution to the Onchocerca biomarker
armamentarium.
The levels of CCG might be influenced by specific dietary patterns in the study populations
investigated here, as cinnamic acid is a molecule that is widely present in plants. The cis-isomer
of cinnamic acid is produced in plants by photoisomerization of trans-cinnamic acid but is
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
10 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
typically detected only in trace amounts in plants [54,55]. It is possible that there are substan-
tial regional differences in baseline CCG levels, caused by different nutritional or living habits.
Since the cutoff for CCG positivity was only based on samples collected in Ghana, this might
explain the high number of CCG positive individuals in Kenya. It would however mean that
specific cutoffs need to be determined for different countries or regions, making it practically
very hard to use. Next to the possible dietary effect, other helminth infections might play a
role, but based on the data obtained in this study it appears that soil-transmitted helminthiasis
and schistosomiasis are not linked to the increased excretion of CCG.
A test for a metabolite biomarker for onchocerciasis such as urinary CCG could have been
a promising tool to identify those individuals that are currently missed based on clinical exam-
ination. However, the work here shows the importance of demonstrating clinical utility in bio-
marker research. Biomarker discovery studies, even when well-executed with a proper test and
validation set, are typically based on sample sets from one specific origin. Especially in the con-
text of tropical diseases, it is not always easy to have proper control groups. Ideally, these
should be as similar as possible to the infected group, but only differing in their infection sta-
tus. Often—also in this study—a negative control group is obtained from a city in the vicinity
of the endemic region. However, these individuals do not only differ in their infection status,
but also in their diet, specific exposures to other pathogens, occupation. . . Confirmation stud-
ies in geographically different populations are absolutely essential to ensure that the right bio-
markers are being selected for further development into diagnostic tools.
In our previous work on clinical utility testing of peptide biomarkers, we came to a similar
observation when investigating this population of children in Kenya [39]. More than 50% of
the children were indeed seroreactive to the peptide epitopes, without presenting any evidence
for being infected, or being exposed, or residing in an endemic area. This observation invali-
dated the peptide biomarker concept as an additional tool. In this study, the same population
of children also showed elevated levels of CCG, again without any evidence of O. volvulus
exposure or infection. It should also be noted that there is no correlation between the signals
observed on the peptide serology and elevated CCG levels. Both are independent observations
and again emphasize the need for confirmation studies.
To be useful as a pharmacodynamic (PD) marker to monitor the efficacy of new drugs in clini-
cal trials, it is important that a biomarker covers a sufficiently large dynamic range in the study
population. We reasoned that a metabolite would need to be present at least at a concentration
0.5 log higher than the defined cutoff to allow proper pharmacodynamic modeling upon treat-
ment. This permits detailed longitudinal monitoring of the drug’s effect on the disease and worm
activity. Based on the data for the candidate biomarkers described here, we suggest plasma ino-
sine as PD marker for onchocerciasis treatment (Fig 2) as for this marker, a sufficiently high num-
ber of infected individuals are found in the window above this PD cutoff, which is not the case for
hypoxanthine. To confirm their use as PD marker, retrospective analysis of previously executed
(animal) studies and prospective collection of samples from macrofilaricide treated individuals
will be required. Both the O. ochengi cow model and the SCID mouse O. ochengi implant model
might be ideally suited to follow the increase of the suggested biomarkers under controlled exper-
imental conditions and eventually also the decrease upon macrofilaricide treatment [56,57]. Also
retrospective analysis of samples from doxycycline field studies might be useful to study the bio-
marker levels as it’s been described that doxycycline has macrofilaricide properties [18, 58].
Previous studies setup to discover biomarkers for onchocerciasis have identified NATOG
as a urinary biomarker for onchocerciasis [23,26,27]. We included the NATOG data that were
previously obtained on the same sample set in this work in Fig 2 [28]. This allows proper com-
parison of the newly identified biomarkers with NATOG. As was already described before, no
increase in urinary NATOG levels were detected in the onchocerciasis group, with no samples
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
11 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
having NATOG levels above the 13 μM cutoff (i.e. 4.63 μg/mL) that was defined by Globisch
and colleagues [27]. Furthermore, no difference was observed with the control groups (non-
endemic controls and healthy controls). It’s important to emphasize that the samples used in
this study were collected from individuals with palpable nodules but without or with very low
levels of microfilaria in the skin. The data for NATOG, in contrast to those obtained for plasma
inosine, might suggest that NATOG should be considered a surrogate marker for the presence
of microfilaria, rather than a surrogate for the presence of live adult worms.
In conclusion, this work shows the potential of plasma inosine and hypoxanthine as mark-
ers for filarial infection. Furthermore, plasma inosine shows potential to be used as pharmaco-
dynamic marker for use in clinical trials investigating the efficacy of filaricides for treatment of
onchocerciasis or lymphatic filariasis.
Methods
Ethics statement
Field study performed in Ghana was approved by the Committee on Human Research, Publi-
cations and Ethics of the School of Medical Sciences of the Kwame Nkrumah University of Sci-
ence and Technology, Kumasi, Ghana and all study subjects signed an informed consent form.
Plasma and urine samples from Kenya were collected as part of a field study. The study was
approved by the KEMRI Scientific and Ethics Review Unit (SERU), Nairobi, Kenya (Protocol
Nr. # KEMRI/SERU/CGHR/102/3554). Since all study participants were minors, informed
consent forms were signed by parents/guardians of the study participants, and verbal assents
were obtained from all study participants. Collection of samples from healthy donors in Bel-
gium was approved by The Ethics Committee [“Commissie voor Medische Ethiek—Zieken-
huisNetwerk Antwerpen (ZNA) and the Ethics Committee University Hospital Antwerp] and
an Informed consent was signed by all subjects. All samples used in this study were anon-
ymized and were collected from adults (18 years or above) only.
Study samples
Plasma and urine samples used for biomarker discovery were collected as part of a field study
in Ghana as described before [28]. A total of 98 nodule positive subjects that donated plasma
and urine samples were included, as well as 51 endemic controls that had no visible signs of
onchocerciasis. Additionally, plasma and urine of samples from 54 non-endemic controls
(from Kumasi, Ashanti Region) and 50 lymphatic filariasis patients (from Ahanta West Dis-
trict, Western Region) were available for testing. As an additional control group, plasma and
urine samples from 50 Belgian healthy controls were included [59–63]. Samples for the bio-
marker evaluation study were collected in the former Nyanza province, in the southwest part
of Kenya, with collections in the Kisumu county (high S. mansoni prevalence area) and Siaya
county (high STH prevalence area). Parasitological information for this study sample set has
been published before [64,65]. Stool samples were collected in order to determine the STH
and S. mansoni infection status of these study participants. An overview of all study popula-
tions, including microfilarial (mf) load in the skin and mass drug administration information
is provided in S7 Table. All blood and urine samples were stored in cold boxes before being
processed in the lab. The plasma and urine samples were then stored at -80˚C until analysis.
Onchocerciasis IgG4 rapid test
The presence of IgG4 antibodies against the O. volvulus antigen Ov16 was determined using
the SD BIOLINE Onchocerciasis IgG4 test (Standard Diagnostics, Gyeonggi-do, Republic of
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
12 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
Korea), according to manufacturer’s instructions. Briefly, 10 μL of plasma was added to the
round sample well on the lateral flow strip, immediately followed by the addition of 4 drops of
assay diluent into the square assay diluent well. After 1 hour, tests were scored. Tests were con-
sidered positive only when both the test and control line were visible. Faint lines were consid-
ered positive, as recommended by the manufacturer.
Preparation of QC samples
A quality control (QC) pool was constructed by collecting 50 or 100 μL of all the plasma or
urine samples, respectively, that were used for the untargeted discovery approaches. Subse-
quently, this QC pool was divided into aliquots to acquire representative QC samples. QC
samples were prepared simultaneously along with study samples and were analyzed through-
out the LC-MS and GC-MS analysis sequences every five study samples. Since these samples
do not contain any biological variability, they can be considered as technical replicates. For
both plasma and urine, study and QC samples were prepared in random order. Blank extracts
were prepared simultaneously along with study samples and were analyzed before the LC-MS
and GC-MS analysis sequences to check the overall contamination in the analytical pipeline.
Sample preparation and analysis
All sample preparation procedures, as well as all sample analyses (both untargeted and targeted
approaches) are described in S1 Supplementary Materials and Methods [66–68]. All reference
materials used in the targeted analysis were purchased from commercial suppliers, except for
cis-cinnamoylglycine which was synthesized in-house. A detailed description of the synthesis
and quality control procedures is available in S1 Supplementary Materials and Methods.
Quality of analysis of the untargeted approaches
To monitor stability of the data during the analytical sequence, the total lipid or metabolite
intensity of the QC samples is monitored in function of analysis time. For the metabolomics
studies (both LC-MS and GC-MS), stable trends were observed for all sequences. For the lipi-
domics studies, an intensity drop was observed after QC sample 6. Therefore, all samples ana-
lyzed before this QC sample, were ruled out for further data processing, leaving only 49 NP, 12
NEC and 7 LF study samples for the comparative lipidomics study.
The validity of the performed analyses was monitored in both a targeted and a non-targeted
manner using the QC samples. For the LC-MS based metabolomics and lipidomics, targeted
monitoring was performed by determining the error of the measurement on signal intensity
(peak area), retention time and mass accuracy for a list of 18–22 randomly selected metabo-
lites. S8 Table summarizes the results of this targeted validity verification. Peak area fluctua-
tions, originating from both the sample preparation step and the LC-MS analysis, are typically
below 15% relative standard deviation, except for lipidomics, where these are typically below
30% relative standard deviation because of the more complex extraction procedure [69,70].
Chromatographic retention time reproducibility is in general satisfactory and less than 1 RSD
%. Also, high mass accuracy (< 5ppm) was obtained for all analyses. For GC-MS based meta-
bolomics, a normalization strategy was employed on all detected features. For plasma, next QC
normalization was employed for all features. For urine, normalization for the Total Metabolite
Content and Internal Standard was employed to compensate for the inherent dilutional differ-
ences between urine samples. Precision on peak area was calculated for a randomly selected
range of identified metabolite species measured in the QC samples. S9 and S10 Tables summa-
rize the results of this targeted validity verification. Peak area fluctuations are typically below
15% relative standard deviation for plasma and below 30% for urine.
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
13 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
Apart from this targeted approach, the reproducibility of the applied metabolomics analysis
was examined in a more comprehensive way by calculating the error on all detected features in
the QC samples and representing the acquired RSD distribution as depicted in S4 Fig. For all
analyses performed, > 75% of all features had an RSD below 30%, which can be defined as the
upper limit for untargeted or discovery metabolomics analysis [71]. For lipidomics in positive
ESI however, only 58% of all features had an RSD below 30%, which can be explained by the
large number of MS saturated lipids in positive ESI mode. The high lipid load was deliberately
chosen for the detection of lower abundant lipid markers. For GC-MS, the data confirm the
requirement for proper normalization procedures.
Statistical analysis
Statistical analyses used to analyze the data obtained in the untargeted discovery studies, have
been described in S1 Supplementary Materials and Methods. For evaluation of the correlation
between data obtained using the untargeted approach and the targeted approach, linear regres-
sion analysis was performed on log-transformed data. A minimal r2 of 0.9 was set as accep-
tance criterion to justify the use of the targeted method for subsequent validation studies. For
comparison of different groups in the validation studies, two-tailed unpaired t-test with
Welch’s correction on log-transformed data was performed. ROC analysis was performed
using specified sample sets as cases and controls, and cutoffs were determined as the point
with maximal Youden’s index ((Sensitivity + Specificity)-1). Based on these cutoffs, sensitivity
and specificity of each biomarker was determined, as well as percentage positives in specific
sample sets. To determine whether more samples were found to be positive in one group com-
pared to another group, contingency tables were prepared, and Chi square test was performed.
All analyses were performed using GraphPad Prism version 7.00.
Supporting information
S1 Fig. Correlation between data obtained in the untargeted -omics approach vs. data
obtained using targeted method. NEC samples have been indicated in red, LF samples in
green and NP samples in blue.
(TIFF)
S2 Fig. ROC analysis of the markers that were further investigated in a targeted LC-MS/
MS analysis. ROC analysis was performed with NP and NEC as cases and controls, respec-
tively.
(TIFF)
S3 Fig. ROC analysis of the filarial markers hypoxanthine and inosine and the onchocerci-
asis marker CCG based on the data obtained from the validation sample set. Cutoff defined
by maximal Youden’s index is indicated in red.
(TIFF)
S4 Fig. RSD distribution on all detected features in the QC samples.
(TIFF)
S1 Table. Overview of samples used in metabolomics and lipidomics discovery study.
(DOCX)
S2 Table. Characteristics of features selected from the comparative plasma lipid profiling
study.
(DOCX)
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
14 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
S3 Table. Characteristics of features selected from the comparative LC-MS based plasma
metabolite profiling study.
(DOCX)
S4 Table. Characteristics of features selected from the comparative LC-MS based urine
metabolite profiling study.
(DOCX)
S5 Table. Characteristics of features selected from the comparative GC-MS based plasma
metabolite profiling study.
(DOCX)
S6 Table. Characteristics of features selected from the comparative GC-MS based urine
metabolite profiling study.
(DOCX)
S7 Table. Overview of study population.
(DOCX)
S8 Table. Targeted validity verification of LC-MS based metabolomics and lipidomics.
(DOCX)
S9 Table. Targeted validity verification of GC-MS based metabolomics in plasma. Precision
obtained with different normalization strategies for the QC samples is shown.
(DOCX)
S10 Table. Targeted validity verification of GC-MS based metabolomics in urine. Precision
obtained with different normalization strategies for the QC samples is shown.
(DOCX)
S1 Supplementary Materials and Methods. Sample preparation, analytical procedures and
synthesis of cis-cinnamoylglycine.
(DOCX)
Acknowledgments
We thank Janssen Biobank for logistic support, Jonathan Vandenbussche from RIC for analyt-
ical support, and Benny Baeten and Marc Engelen from Janssen Global Public Health for pro-
grammatic support.
Author Contributions
Conceptualization: Ole Lagatie, Lieven J. Stuyver.
Formal analysis: Ole Lagatie, Dirk Van Roosbroeck, Stijn Van Asten, Ruben T’Kindt, Lieven
J. Stuyver.
Investigation: Ole Lagatie, Linda Batsa Debrah, Alex Debrah, Maurice R. Odiere, Lieven J.
Stuyver.
Methodology: Ole Lagatie, Emmanuel Njumbe Ediage, Ann Verheyen, Ruben T’Kindt,
Emmie Dumont, Lieve Dillen, Tom Verhaeghe, Rob Vreeken, Filip Cuyckens, Lieven J.
Stuyver.
Supervision: Koen Sandra, Lieve Dillen, Tom Verhaeghe, Rob Vreeken, Filip Cuyckens, Lie-
ven J. Stuyver.
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
15 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
Visualization: Ole Lagatie, Lieven J. Stuyver.
Writing – original draft: Ole Lagatie.
Writing – review & editing: Linda Batsa Debrah, Alex Debrah, Maurice R. Odiere, Ruben
T’Kindt, Emmie Dumont, Koen Sandra, Lieve Dillen, Tom Verhaeghe, Rob Vreeken, Filip
Cuyckens, Lieven J. Stuyver.
References
1. WHO. Progress report on the elimination of human onchocerciasis, 2016–2017. In: Cantey P, editor.
Releve epidemiologique hebdomadaire / Section d’hygiene du Secretariat de la Societe des Nations =
Weekly epidemiological record / Health Section of the Secretariat of the League of Nations. 92: WHO;
2017. p. 681–94.
2. Pearlman E, Gillette-Ferguson I. Onchocerca volvulus, Wolbachia and river blindness. Chemical immu-
nology and allergy. 2007; 92:254–65. https://doi.org/10.1159/000099276 PMID: 17264501.
3. Gardon J, Boussinesq M, Kamgno J, Gardon-Wendel N, Demanga N, Duke BO. Effects of standard
and high doses of ivermectin on adult worms of Onchocerca volvulus: a randomised controlled trial.
Lancet. 2002; 360(9328):203–10. https://doi.org/10.1016/S0140-6736(02)09456-4 PMID: 12133654.
4. Homeida M, Braide E, Elhassan E, Amazigo UV, Liese B, Benton B, et al. APOC’s strategy of commu-
nity-directed treatment with ivermectin (CDTI) and its potential for providing additional health services to
the poorest populations. African Programme for Onchocerciasis Control. Annals of tropical medicine
and parasitology. 2002; 96 Suppl 1:S93–104. https://doi.org/10.1179/000349802125000673 PMID:
12081254.
5. Hoerauf A. Filariasis: new drugs and new opportunities for lymphatic filariasis and onchocerciasis. Cur-
rent opinion in infectious diseases. 2008; 21(6):673–81. https://doi.org/10.1097/QCO.
0b013e328315cde7 PMID: 18978537.
6.
Lipner EM, Dembele N, Souleymane S, Alley WS, Prevots DR, Toe L, et al. Field applicability of a rapid-
format anti-Ov-16 antibody test for the assessment of onchocerciasis control measures in regions of
endemicity. J Infect Dis. 2006; 194(2):216–21. https://doi.org/10.1086/505081 PMID: 16779728.
7. Weil GJ, Steel C, Liftis F, Li BW, Mearns G, Lobos E, et al. A rapid-format antibody card test for diagno-
sis of onchocerciasis. J Infect Dis. 2000; 182(6):1796–9. https://doi.org/10.1086/317629 PMID:
11069258.
8. Steel C, Golden A, Stevens E, Yokobe L, Domingo GJ, de Los Santos T, et al. Rapid point-of-contact
tool for mapping and integrated surveillance of Wuchereria bancrofti and Onchocerca volvulus infection.
Clinical and vaccine immunology: CVI. 2015; 22(8):896–901. https://doi.org/10.1128/CVI.00227-15
PMID: 26018537.
9. Golden A, Steel C, Yokobe L, Jackson E, Barney R, Kubofcik J, et al. Extended result reading window
in lateral flow tests detecting exposure to Onchocerca volvulus: a new technology to improve epidemio-
logical surveillance tools. PLoS One. 2013; 8(7):e69231. https://doi.org/10.1371/journal.pone.0069231
PMID: 23935960; PubMed Central PMCID: PMC3720650.
10.
Lavebratt C, Dalhammar G, Adamafio NA, Nykanen-Dejerud U, Mingarini K, Ingemarsson K, et al. A
simple dot blot assay adaptable for field use in the diagnosis of onchocerciasis: preparation of an adult
worm antigen fraction which enhances sensitivity and specificity. Transactions of the Royal Society of
Tropical Medicine and Hygiene. 1994; 88(3):303–6. https://doi.org/10.1016/0035-9203(94)90090-6
PMID: 7974670.
11. Chandrashekar R, Ogunrinade AF, Weil GJ. Use of recombinant Onchocerca volvulus antigens for
diagnosis and surveillance of human onchocerciasis. Tropical medicine & international health: TM & IH.
1996; 1(5):575–80. https://doi.org/10.1111/j.1365-3156.1996.tb00082.x PMID: 8911441.
12.
13.
Lont YL, Coffeng LE, de Vlas SJ, Golden A, de Los Santos T, Domingo GJ, et al. Modelling anti-Ov16
IgG4 antibody prevalence as an indicator for evaluation and decision making in onchocerciasis elimina-
tion programmes. PLoS neglected tropical diseases. 2017; 11(1):e0005314. https://doi.org/10.1371/
journal.pntd.0005314 PMID: 28114304; PubMed Central PMCID: PMC5289624.
Zoure HG, Noma M, Tekle AH, Amazigo UV, Diggle PJ, Giorgi E, et al. The geographic distribution of
onchocerciasis in the 20 participating countries of the African Programme for Onchocerciasis Control:
(2) pre-control endemicity levels and estimated number infected. Parasites & vectors. 2014; 7:326.
https://doi.org/10.1186/1756-3305-7-326 PMID: 25053392; PubMed Central PMCID: PMC4222889.
14. Noma M, Zoure HG, Tekle AH, Enyong PA, Nwoke BE, Remme JH. The geographic distribution of
onchocerciasis in the 20 participating countries of the African Programme for Onchocerciasis Control:
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
16 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
(1) priority areas for ivermectin treatment. Parasites & vectors. 2014; 7:325. https://doi.org/10.1186/
1756-3305-7-325 PMID: 25053266; PubMed Central PMCID: PMC4223657.
15. Vlaminck J, Fischer PU, Weil GJ. Diagnostic Tools for Onchocerciasis Elimination Programs. Trends in
parasitology. 2015; 31(11):571–82. https://doi.org/10.1016/j.pt.2015.06.007 PMID: 26458784.
16. Bennuru S, Oduro-Boateng G, Osigwe C, Del Valle P, Golden A, Ogawa GM, et al. Integrating Multiple
Biomarkers to Increase Sensitivity for the Detection of Onchocerca volvulus Infection. J Infect Dis.
2020; 221(11):1805–15. Epub 2019/06/16. https://doi.org/10.1093/infdis/jiz307 PMID: 31201416;
PubMed Central PMCID: PMC7213562.
17. O’Connell EM, Bennuru S, Steel C, Dolan MA, Nutman TB. Targeting Filarial Abl-like Kinases: Orally
Available, Food and Drug Administration-Approved Tyrosine Kinase Inhibitors Are Microfilaricidal and
Macrofilaricidal. J Infect Dis. 2015; 212(5):684–93. https://doi.org/10.1093/infdis/jiv065 PMID:
25657255; PubMed Central PMCID: PMC4539898.
18. Walker M, Specht S, Churcher TS, Hoerauf A, Taylor MJ, Basanez MG. Therapeutic efficacy and
macrofilaricidal activity of doxycycline for the treatment of river blindness. Clinical infectious diseases:
an official publication of the Infectious Diseases Society of America. 2015; 60(8):1199–207. https://doi.
org/10.1093/cid/ciu1152 PMID: 25537873; PubMed Central PMCID: PMC4370165.
19. Geary TG, Mackenzie CD. Progress and challenges in the discovery of macrofilaricidal drugs. Expert
review of anti-infective therapy. 2011; 9(8):681–95. https://doi.org/10.1586/eri.11.76 PMID: 21819332.
20. Mackenzie CD, Geary TG. Flubendazole: a candidate macrofilaricide for lymphatic filariasis and oncho-
cerciasis field programs. Expert review of anti-infective therapy. 2011; 9(5):497–501. https://doi.org/10.
1586/eri.11.30 PMID: 21609260.
21. WHO. Onchocerciasis: diagnostic target product profile to support preventive chemotherapy. 2021.
22. Denery JR, Nunes AA, Hixon MS, Dickerson TJ, Janda KD. Metabolomics-based discovery of diagnos-
tic biomarkers for onchocerciasis. PLoS neglected tropical diseases. 2010; 4(10). https://doi.org/10.
1371/journal.pntd.0000834 PMID: 20957145; PubMed Central PMCID: PMC2950146.
23. Globisch D, Moreno AY, Hixon MS, Nunes AAK, Denery JR, Specht S, et al. Onchocerca volvulus-
neurotransmitter tyramine is a biomarker for river blindness. P Natl Acad Sci USA. 2013; 110
(11):4218–23. https://doi.org/10.1073/pnas.1221969110 PubMed PMID: WOS:000316238300022.
PMID: 23440222
24. Bennuru S, Lustigman S, Abraham D, Nutman TB. Metabolite profiling of infection-associated metabolic
markers of onchocerciasis. Molecular and biochemical parasitology. 2017; 215:58–69. https://doi.org/
10.1016/j.molbiopara.2017.01.008 PMID: 28188804; PubMed Central PMCID: PMC5474354.
25. Wewer V, Makepeace BL, Tanya VN, Peisker H, Pfarr K, Hoerauf A, et al. Lipid profiling of the filarial
nematodes Onchocerca volvulus, Onchocerca ochengi and Litomosoides sigmodontis reveals the
accumulation of nematode-specific ether phospholipids in the host. International journal for parasitol-
ogy. 2017; 47(14):903–12. https://doi.org/10.1016/j.ijpara.2017.06.001 PMID: 28743489; PubMed Cen-
tral PMCID: PMC5716430.
26. Globisch D, Specht S, Pfarr KM, Eubanks LM, Hoerauf A, Janda KD. Litomosoides sigmodontis: a jird
urine metabolome study. Bioorganic & medicinal chemistry letters. 2015; 25(24):5804–7. https://doi.
org/10.1016/j.bmcl.2015.10.037 PMID: 26573416.
27. Globisch D, Eubanks LM, Shirey RJ, Pfarr KM, Wanji S, Debrah AY, et al. Validation of onchocerciasis
biomarker N-acetyltyramine-O-glucuronide (NATOG). Bioorganic & medicinal chemistry letters. 2017;
27(15):3436–40. https://doi.org/10.1016/j.bmcl.2017.05.082 PMID: 28600214; PubMed Central
PMCID: PMC5510726.
28.
Lagatie O, Njumbe Ediage E, Batsa Debrah L, Diels L, Nolten C, Vinken P, et al. Evaluation of the diag-
nostic potential of urinary N-Acetyltyramine-O,beta-glucuronide (NATOG) as diagnostic biomarker for
Onchocerca volvulus infection. Parasites & vectors. 2016; 9(1):302. https://doi.org/10.1186/s13071-
016-1582-6 PMID: 27216752; PubMed Central PMCID: PMC4877973.
29. Shirey RJ, Globisch D, Eubanks LM, Hixon MS, Janda KD. Noninvasive Urine Biomarker Lateral Flow
Immunoassay for Monitoring Active Onchocerciasis. ACS Infect Dis. 2018; 4(10):1423–31. Epub 2018/
08/25. https://doi.org/10.1021/acsinfecdis.8b00163 PMID: 30141624; PubMed Central PMCID:
PMC6189908.
30. Hotterbeekx A, Dusabimana A, Mandro M, Abhafule GM, Deogratias W, Siewe Fodjo JN, et al. Urinary
N-acetyltyramine-O,beta-glucuronide in Persons with Onchocerciasis-Associated Epilepsy. Pathogens.
2020; 9(3). Epub 2020/03/11. https://doi.org/10.3390/pathogens9030191 PMID: 32151038; PubMed
Central PMCID: PMC7157704.
31.
Lagatie O, Batsa Debrah L, Debrah A, Stuyver LJ. Plasma-derived parasitic microRNAs have insuffi-
cient concentrations to be used as diagnostic biomarker for detection of Onchocerca volvulus infection
or treatment monitoring using LNA-based RT-qPCR. Parasitology research. 2017; 116(3):1013–22.
https://doi.org/10.1007/s00436-017-5382-5 PMID: 28111713; PubMed Central PMCID: PMC5313568.
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
17 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
32.
Tritten L, O’Neill M, Nutting C, Wanji S, Njouendoui A, Fombad F, et al. Loa loa and Onchocerca
ochengi miRNAs detected in host circulation. Molecular and biochemical parasitology. 2014; 198
(1):14–7. https://doi.org/10.1016/j.molbiopara.2014.11.001 PMID: 25461483.
33. Macfarlane CL, Quek S, Pionnier N, Turner JD, Wanji S, Wagstaff SC, et al. The insufficiency of circulat-
ing miRNA and DNA as diagnostic tools or as biomarkers of treatment efficacy for Onchocerca volvulus.
Scientific reports. 2020; 10(1):6672. Epub 2020/04/23. https://doi.org/10.1038/s41598-020-63249-4
PMID: 32317658; PubMed Central PMCID: PMC7174290.
34.
35.
36.
Lagatie O, Verheyen A, Nijs E, Van Dorst B, Batsa Debrah L, Debrah A, et al. Evaluation of the Diagnos-
tic Performance of Onchocerca volvulus Linear Epitopes in a Peptide Enzyme-Linked Immunosorbent
Assay. The American journal of tropical medicine and hygiene. 2018; 98(3):779–85. Epub 2018/01/10.
https://doi.org/10.4269/ajtmh.17-0756 PMID: 29313477; PubMed Central PMCID: PMC5930915.
Lagatie O, Van Dorst B, Stuyver LJ. Identification of three immunodominant motifs with atypical isotype
profile scattered over the Onchocerca volvulus proteome. PLoS neglected tropical diseases. 2017; 11
(1):e0005330. Epub 2017/01/27. https://doi.org/10.1371/journal.pntd.0005330 PMID: 28125577;
PubMed Central PMCID: PMC5295699 following competing interests:All authors are current employ-
ees of Janssen Pharmaceutica NV, a Johnson and Johnson Company and may own stock or stock
options in that company.
Lagatie O, Verheyen A, Nijs E, Batsa Debrah L, Debrah YA, Stuyver LJ. Performance evaluation of 3
serodiagnostic peptide epitopes and the derived multi-epitope peptide OvNMP-48 for detection of
Onchocerca volvulus infection. Parasitology research. 2019; 118(7):2263–70. Epub 2019/05/16.
https://doi.org/10.1007/s00436-019-06345-3 PMID: 31089811; PubMed Central PMCID:
PMC6611744.
37. Vlaminck J, Lagatie O, Verheyen A, Dana D, Van Dorst B, Mekonnen Z, et al. Patent infections with
soil-transmitted helminths and Schistosoma mansoni are not associated with increased prevalence of
antibodies to the Onchocerca volvulus peptide epitopes OvMP-1 and OvMP-23. Parasites & vectors.
2019; 12(1):63. Epub 2019/01/30. https://doi.org/10.1186/s13071-019-3308-z PMID: 30691520;
PubMed Central PMCID: PMC6350367.
38.
39.
40.
Lagatie O, Verheyen A, Van Dorst B, Batsa Debrah L, Debrah A, Stuyver LJ. Linear epitopes in Oncho-
cerca volvulus vaccine candidate proteins and excretory-secretory proteins. Parasite immunology.
2018; 40(11):e12587. Epub 2018/09/07. https://doi.org/10.1111/pim.12587 PMID: 30188578; PubMed
Central PMCID: PMC6220944.
Lagatie O, Granjon E, Odiere MR, Zrein M, Stuyver LJ. Assessment of multiplex Onchocerca volvulus
peptide ELISA in non-endemic tropical regions. Parasites & vectors. 2019; 12(1):570. Epub 2019/12/
01. https://doi.org/10.1186/s13071-019-3824-x PMID: 31783767; PubMed Central PMCID:
PMC6884800.
Lagatie O, Njumbe Ediage E, Van Roosbroeck D, Van Asten S, Verheyen A, Van Dorst B, et al. Ele-
vated plasma inosine and hypoxanthine concentrations, and urine cis-cinnamoyl glycine concentrations
as biomarkers for active Onchocerca volvulus infections. Annual Meeting American Society of Tropical
Medicine and Hygiene; New Orleans, LA, USA2018.
41. Wewer V, Peisker H, Gutbrod K, Al-Bahra M, Menche D, Amambo NG, et al. Urine metabolites for the
identification of Onchocerca volvulus infections in patients from Cameroon. Parasites & vectors. 2021;
14(1):397. Epub 2021/08/13. https://doi.org/10.1186/s13071-021-04893-1 PMID: 34380554; PubMed
Central PMCID: PMC8359580.
42. Golden A, Faulx D, Kalnoky M, Stevens E, Yokobe L, Peck R, et al. Analysis of age-dependent trends in
Ov16 IgG4 seroprevalence to onchocerciasis. Parasites & vectors. 2016; 9(1):338. https://doi.org/10.
1186/s13071-016-1623-1 PMID: 27296630; PubMed Central PMCID: PMC4907250.
43. Bozikov J, Lijana ZK. Test Validity Measures and Receiver Operating Characteristic (ROC) Analysis.
In: Lijana ZK, Bozikov J, editors. Methods and Tools in Public Health. Lage, Germany: Hans Jacobs;
2010.
44. Roberts JM, Neumann E, Gockel CW, Highton RB. Onchocerciasis in Kenya 9, 11 and 18 years after
elimination of the vector. Bulletin of the World Health Organization. 1967; 37(2):195–212. Epub 1967/
01/01. PMID: 5303323; PubMed Central PMCID: PMC2554347.
45. Crump A, Morel CM, Omura S. The onchocerciasis chronicle: from the beginning to the end? Trends in
parasitology. 2012; 28(7):280–8. https://doi.org/10.1016/j.pt.2012.04.005 PMID: 22633470.
46. WHO. Onchocerciasis and its control. Report of a WHO Expert Committee on Onchocerciasis Control.
World Health Organization technical report series. 1995;852:1–104. Epub 1995/01/01. PMID: 7541171.
47. Moraga P, Cano J, Baggaley RF, Gyapong JO, Njenga SM, Nikolay B, et al. Modelling the distribution
and transmission intensity of lymphatic filariasis in sub-Saharan Africa prior to scaling up interventions:
integrated use of geostatistical and mathematical modelling. Parasites & vectors. 2015; 8:560. Epub
2015/10/27. https://doi.org/10.1186/s13071-015-1166-x PMID: 26496983; PubMed Central PMCID:
PMC4620019.
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
18 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
48.
49.
Traut TW. Enzymes of nucleotide metabolism: the significance of subunit size and polymer size for bio-
logical function and regulatory properties. CRC critical reviews in biochemistry. 1988; 23(2):121–69.
https://doi.org/10.3109/10409238809088318 PMID: 3048887.
Traut TW. Physiological concentrations of purines and pyrimidines. Molecular and cellular biochemistry.
1994; 140(1):1–22. https://doi.org/10.1007/BF00928361 PMID: 7877593.
50. Maiuolo J, Oppedisano F, Gratteri S, Muscoli C, Mollace V. Regulation of uric acid metabolism and
excretion. International journal of cardiology. 2016; 213:8–14. https://doi.org/10.1016/j.ijcard.2015.08.
109 PMID: 26316329.
51. Berens RL, Krug EC, Marr JJ. Purine and Pyrimidine Metabolism. In: Marr JJ, Mu¨ ller M, editors. Bio-
chemistry and Molecular Biology of Parasites. London: Academic Press Ltd.; 1995.
52. Adisakwattana S. Cinnamic Acid and Its Derivatives: Mechanisms for Prevention and Management of
Diabetes and Its Complications. Nutrients. 2017; 9(2). https://doi.org/10.3390/nu9020163 PMID:
28230764; PubMed Central PMCID: PMC5331594.
53.
Turner LB, Mueller-Harvey I, McAllan AB. Light-induced isomerization and dimerization of cinnamic
acid derivatives in cell walls. Phytochemistry. 1993; 33:6.
54. Wong WS, Guo D, Wang XL, Yin ZQ, Xia B, Li N. Study of cis-cinnamic acid in Arabidopsis thaliana.
Plant physiology and biochemistry: PPB. 2005; 43(10–11):929–37. https://doi.org/10.1016/j.plaphy.
2005.08.008 PMID: 16310363.
55. Yang XX, Choi HW, Yang SF, Li N. A UV-light activated cinnamic acid isomer regulates plant growth
and gravitropism via an ethylene receptor-independent pathway. Australian journal of plant physiology.
1999; 26(4):325–35. https://doi.org/10.1071/pp99007 PMID: 11542914.
56.
Trees AJ, Graham SP, Renz A, Bianco AE, Tanya V. Onchocerca ochengi infections in cattle as a
model for human onchocerciasis: recent developments. Parasitology. 2000; 120 Suppl:S133–42.
https://doi.org/10.1017/s0031182099005788 PMID: 10874716.
57. Halliday A, Guimaraes AF, Tyrer HE, Metuge HM, Patrick CN, Arnaud KO, et al. A murine macrofi-
laricide pre-clinical screening model for onchocerciasis and lymphatic filariasis. Parasites & vectors.
2014; 7:472. https://doi.org/10.1186/s13071-014-0472-z PMID: 25338621; PubMed Central
PMCID: PMC4212127.
58. Debrah AY, Specht S, Klarmann-Schulz U, Batsa L, Mand S, Marfo-Debrekyei Y, et al. Doxycycline
Leads to Sterility and Enhanced Killing of Female Onchocerca volvulus Worms in an Area With Persis-
tent Microfilaridermia After Repeated Ivermectin Treatment: A Randomized, Placebo-Controlled, Dou-
ble-Blind Trial. Clinical infectious diseases: an official publication of the Infectious Diseases Society of
America. 2015; 61(4):517–26. https://doi.org/10.1093/cid/civ363 PMID: 25948064; PubMed Central
PMCID: PMC4518165.
59. Stuyver LJ, Verbeke T, Van Loy T, Van Gulck E, Tritsmans L. An antibody response to human poly-
omavirus 15-mer peptides is highly abundant in healthy human subjects. Virology journal. 2013;
10:192. https://doi.org/10.1186/1743-422X-10-192 PMID: 23758776; PubMed Central PMCID:
PMC3691923.
60. Van Loy T, Thys K, Tritsmans L, Stuyver LJ. Quasispecies analysis of JC virus DNA present in urine of
healthy subjects. PLoS One. 2013; 8(8):e70950. https://doi.org/10.1371/journal.pone.0070950 PMID:
23967139; PubMed Central PMCID: PMC3744523.
61.
62.
63.
64.
65.
Lagatie O, Van Loy T, Tritsmans L, Stuyver LJ. Circulating human microRNAs are not linked to JC poly-
omavirus serology or urinary viral load in healthy subjects. Virology journal. 2014; 11(1):41. https://doi.
org/10.1186/1743-422X-11-41 PMID: 24588811; PubMed Central PMCID: PMC3945012.
Lagatie O, Van Loy T, Tritsmans L, Stuyver LJ. Antibodies reacting with JCPyV_VP2 _167-15mer as a
novel serological marker for JC polyomavirus infection. Virology journal. 2014; 11:174. https://doi.org/
10.1186/1743-422X-11-174 PMID: 25273879; PubMed Central PMCID: PMC4194363.
Lagatie O, Van Loy T, Tritsmans L, Stuyver LJ. Viral miRNAs in plasma and urine divulge JC polyomavirus
infection. Virology journal. 2014; 11:158. https://doi.org/10.1186/1743-422X-11-158 PMID: 25178457;
PubMed Central PMCID: PMC4168162.
Lagatie O, Verheyen A, Van Hoof K, Lauwers D, Odiere MR, Vlaminck J, et al. Detection of Ascaris lum-
bricoides infection by ABA-1 coproantigen ELISA. PLoS neglected tropical diseases. 2020; 14(10):
e0008807. Epub 2020/10/16. https://doi.org/10.1371/journal.pntd.0008807 PMID: 33057357; PubMed
Central PMCID: PMC7591086 following competing interests: OL, AV, KVH and LJS are current employ-
ees of Janssen Pharmaceutica NV, being a Johnson and Johnson Companies and they may own stock
or stock options in that company.
Lagatie O, Verheyen A, Van Asten S, Odiere MR, Djuardi Y, Levecke B, et al. 2-Methyl-pentanoyl-carni-
tine (2-MPC): a urine biomarker for patent Ascaris lumbricoides infection. Scientific reports. 2020; 10
(1):15780. Epub 2020/09/27. https://doi.org/10.1038/s41598-020-72804-y PMID: 32978457; PubMed
Central PMCID: PMC7519643.
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
19 / 20
PLOS NEGLECTED TROPICAL DISEASES Discovery of novel biomarkers for active Onchocerca volvulus infection
66.
t’Kindt R, Telenga ED, Jorge L, Van Oosterhout AJ, Sandra P, Ten Hacken NH, et al. Profiling over
1500 lipids in induced lung sputum and the implications in studying lung diseases. Analytical chemistry.
2015; 87(9):4957–64. https://doi.org/10.1021/acs.analchem.5b00732 PMID: 25884268.
67. Boelaert J, t’Kindt R, Schepers E, Jorge L, Glorieux G, Neirynck N, et al. State-of-the-art non-targeted
metabolomics in the study of chronic kidney disease. Metabolomics. 2014; 10(3):425–42. https://doi.
org/10.1007/s11306-013-0592-z
68. Sandra K, Pereira Ados S, Vanhoenacker G, David F, Sandra P. Comprehensive blood plasma lipido-
mics by liquid chromatography/quadrupole time-of-flight mass spectrometry. Journal of chromatogra-
phy A. 2010; 1217(25):4087–99. https://doi.org/10.1016/j.chroma.2010.02.039 PMID: 20307888.
69. Cajka T, Smilowitz JT, Fiehn O. Validating Quantitative Untargeted Lipidomics Across Nine Liquid Chro-
matography-High-Resolution Mass Spectrometry Platforms. Analytical chemistry. 2017; 89(22):12360–
8. https://doi.org/10.1021/acs.analchem.7b03404 PMID: 29064229.
70. Dunn WB, Wilson ID, Nicholls AW, Broadhurst D. The importance of experimental design and QC sam-
ples in large-scale and MS-driven untargeted metabolomic studies of humans. Bioanalysis. 2012; 4
(18):2249–64. https://doi.org/10.4155/bio.12.204 PMID: 23046267.
71. Spagou K, Wilson ID, Masson P, Theodoridis G, Raikos N, Coen M, et al. HILIC-UPLC-MS for explor-
atory urinary metabolic profiling in toxicological studies. Analytical chemistry. 2011; 83(1):382–90.
https://doi.org/10.1021/ac102523q PMID: 21142126.
PLOS Neglected Tropical Diseases | https://doi.org/10.1371/journal.pntd.0009999 November 29, 2021
20 / 20
PLOS NEGLECTED TROPICAL DISEASES
| null |
10.1088_1361-6595_ad05f5.pdf
|
Data availability statement
The data cannot be made publicly available upon publication
because they are not available in a format that is sufficiently
accessible or reusable by other researchers. The data that sup-
port the findings of this study are available upon reasonable
request from the authors.
|
Data availability statement The data cannot be made publicly available upon publication because they are not available in a format that is sufficiently accessible or reusable by other researchers. The data that support the findings of this study are available upon reasonable request from the authors.
|
Plasma Sources Sci. Technol. 32 (2023) 115006 (10pp)
Plasma Sources Science and Technology
https://doi.org/10.1088/1361-6595/ad05f5
The breakdown characteristic of porous
dielectric discharge based on
percolation structure
Yuheng Hu, Libo Rao, Feiyu Wu, Kai Chen, Yilong Mao, Yue Chen,
Jialei Wang and Hao Wang
∗
State Key Laboratory of Power Transmission Equipment System Security and New Technology,
Chongqing University, Chongqing, People’s Republic of China
E-mail: [email protected]
Received 8 March 2023, revised 18 August 2023
Accepted for publication 23 October 2023
Published 6 November 2023
Abstract
Porous dielectrics have received increasing attention in plasma sterilization, all-solid-state
battery technology, and surface functionalization of biological tissue materials. Due to their
complex structure and discharge characteristics, the current researches are hard to quantify the
stochastic properties of porous dielectrics. In this paper, we used a percolation structure to
simulate the discharge process in a 2D porous dielectric. The simulation results of the 2D
percolation model are similar to that of 2D real porous slices, which can characterize the
physical properties of the porous dielectric well while greatly reducing the time required for
simulation. In addition, simulations on percolation models with different porosity and lattice
size are performed. When the porosity and lattice size remain constant, tortuosity and Debye
radius are the main factors affecting the breakdown of the percolation model. With the decrease
in porosity, the Pashcen curve shifts to the upper right. With the decrease in lattice size, the
Pashcen curve moves higher. The results show correlations between random parameters and
Paschen curves. This study presents a novel simulation approach for the theoretical analysis of
porous dielectric and improves the simulation efficiency at the same time. In addition, this new
model is also applied to quantify the impact mechanism of random parameters such as porosity
and lattice size on porous dielectric discharge.
Keywords: percolation structure, porous dielectric, porosity, lattice size, Paschen’s law
1. Introduction
Porous dielectrics are random non-uniform dielectrics with
a large number of tiny multi-skeleton gaps. The special
tiny pore structure contained in porous dielectric provides
it with a huge specific surface area, which helps to realize
many novel physical and chemical functions. Porous structure
material discharge under different pressures is widely used
in various industrial applications, such as chemical catalysis
[1–3], energy storage [4–6], biomedical materials [7–10], etc.
These applications have gained significant attention in both
∗
Author to whom any correspondence should be addressed.
industrial and scientific research fields. Due to the complex
nature of porous dielectrics and their discharge processes,
the mechanism of plasma discharge within porous dielectrics
remains relatively underdeveloped. Furthermore, the porous
dielectric discharge under atmospheric pressure is influenced
by a multitude of physical phenomena, including photoion-
ization and streamer discharge, thereby rendering it challen-
ging to elucidate the underlying physical principles. To exam-
ine the impact of random structural variations on the por-
ous dielectric, this study employs low-pressure DC condi-
tions to simplify the analysis and avoid the effect of other
factors.
Currently, research methods can be mainly classified into
experiments and simulations. For the experiment, Hensel,
1361-6595/23/115006+10$33.00 Printed in the UK
1
© 2023 IOP Publishing Ltd
Plasma Sources Sci. Technol. 32 (2023) 115006
Y Hu et al
Engeling, and Kruszelnicki et al studied the discharge of
porous dielectric under atmospheric pressure extensively
[11–14]. Porous dielectrics are difficult to diagnose with
probes or spectroscopy due to their uneven spatial distribu-
tion and opaque material. Moreover, the atmospheric pres-
sure conditions, where multiple complex discharge mechan-
isms are intertwined, present challenges in understanding the
key factors affecting discharge. Due to the complexity, cost,
and non-quantifiability of experimental studies, many schol-
ars opt for mathematics and simulation-based research instead.
Verboncoeur et al used a 2D fluid model to assess gas break-
down in microgaps with electrode surface protrusions. He
also investigated transition characteristics and electron kinet-
ics in microhollow cathode discharges [15, 16]. Go and his
team created a mathematical model for the modified Paschen’s
curve, focusing on the breakdown in microgaps. They also
studied the basic properties of microdischarges [17, 18]. Van
Laer and Annemie Bogaerts et al have studied DBD dis-
charges in packed bed plasma reactors by simulation. This
structure is mostly composed of regular spheres, which are
less stochastic than a porous dielectric [19–25]. The simula-
tions by Zhang et al mainly focused on the study of single
pores, which is difficult to reveal the plasma interactions in
a large number of pores [26–28]. Kruszelnicki, Engeling and
Kushner et al used stochastic multi-sphere models to sim-
plify the complex structure of real porous dielectric [13, 14,
29]. They simplified the model into a structured distribution
of regularity, which makes it difficult to reflect the stochastic
complexity. Additionally, these simulations have not fully
addressed the random characteristic parameters of porous
dielectrics.
Finite element analysis in previous simulations of por-
ous dielectric discharge poses difficulties in solving strongly
coupled nonlinear equations, particularly under non-uniform
dielectric conditions. Increased computational effort required
for simulations in real porous dielectric structures often leads
to poor efficiency of results. Due to computational limita-
tions and geometric asymmetry in the porous dielectric, cur-
rent simulations of porous dielectric discharges mainly use
2D sections, lacking the capability to accurately reconstruct
3D models through rotation and symmetry. Moreover, the
results are often hard to quantify. Based on these problems,
to overcome the shortcomings of previous experiments and
simulations, this paper proposes to use a percolation struc-
ture to describe the geometric properties of porous dielec-
tric. The percolation theory, introduced by Broadbent and
Hammersley [30], is a set of mathematical and statistical phys-
ics theories used to study the properties of clusters on ran-
dom graphs [31–33]. The site percolation, abstracts a com-
plex real material into an ideal mathematical model, where
individual square black cells represent the solid part and the
remaining white cells represent the gas part. It has previously
been applied in the field of fluid motion in porous dielectric
and has successfully explained two-phase flow phenomena
[34–37]. The porous dielectric can be equivalent to a per-
colation structure due to its disordered state. Some scholars
have used the percolation model to simulate porous dielectric
before. Andrade et al investigated the dynamics of viscous
penetration in percolation porous dielectric [38–40]. They
also simulated Navier-Stokes equations in percolation struc-
ture directly to study fluid flow through the disordered por-
ous dielectric [41, 42]. Hitherto, the percolation model has not
been used to study plasma discharge in a porous dielectric.
In this paper, DC plasma discharge simulations are per-
formed under low pressure for a macro-size percolation struc-
ture. The results with various porosity and lattice size are
explained by the capillary network model and gas discharge
theory. When porosity and lattice size are held constant, tor-
tuosity and Debye length become primary factors affecting
the breakdown of the percolation model. An increase in tortu-
osity leads to an upward displacement of the Paschen curve,
while an increase in the Debye radius causes the Paschen
curve to shift toward the left. A decrease in porosity res-
ults in an upward and rightward shift of the Paschen curve,
while a decrease in lattice size causes the Paschen curve to
shift upwards. From the simulation, we can conclude that
the percolation model can reduce the simulation difficulty
and improve the simulation efficiency. Most importantly, it
provides a quantitative understanding of the stochastic and
complex nature of porous dielectric. The porous dielectric
model based on percolation structure could be promising for
studying the properties of porous dielectric.
2. Experimental setup and methods
2.1. Geometrical modeling
We created a 2D model of a porous dielectric cross-section
using tomography scans from an aluminum oxide ceramics
porous dielectric sample. The sample, as shown in figure 1(a),
which had a 20 ppi pore density, 5 cm diameter, and 1 cm
height, was scanned using a 3D CT analyzer model CD-
130BX/uCT. The resulting data was used to construct a 3D
model, which was then cut longitudinally along its ca cyl-
indrical axis using Avizo software to obtain 2D slices. After
binarizing the slices and removing isolated holes, they were
imported into COMSOL Multiphysics software, as shown in
figure 1(b).
Then, the 2D porous dielectric section was imported into
ImageJ software. According to the percolation theory, we con-
verted it to black-and-white negative cells, making the pore
domain white and the dielectric domain black. Then the image
was binarized and divided. After reducing the resolution, the
percolation model of the porous dielectric section was finally
derived, which is presented in figure 2. The size of each cell
is smaller than the characteristic length of the plasma, namely
the Debye length, to ensure that the model has sufficient spa-
tial scale without disrupting physical processes [43].
Due to the time-consuming and tedious nature of perform-
ing tomography scans and analyses on real porous dielec-
tric, in order to better study plasma discharge in percolation
structures, we used Mathematica software to generate square
2
Plasma Sources Sci. Technol. 32 (2023) 115006
Y Hu et al
Figure 4. Percolation models with the same lattice size (a = 1/16)
at different porosities. (p = 0.9, 0.8, 0.7).
Figure 1. Real picture (a) and tomography section image (b) of the
porous dielectric. (length: 2.21 cm, height: 1.18 cm).
Figure 5. Percolation models with the same porosity (p = 0.8) at
different lattice sizes. (a = 1/16, 1/32, 1/48).
percolation matrix. The lattice size a is the reciprocal of height
h since the height is set to 1. By setting p = 0.8 and h = 16,
the matrix was visualized [30, 44, 45]. The random percolation
model plots are obtained (figure 3).
Keeping the height h = 16 and varying the porosity p, the
percolation model plots are generated (figure 4).
Keeping the porosity p = 0.8 and varying the height h to
change the side length of a single cell, the percolation model
plots are generated (figure 5).
In the subsequent simulations, we keep the side lengths of
all models at 2.51 cm.
2.2. Plasma modeling
The external circuit diagram for the plasma discharge, includ-
ing the DC power supply (Us), discharge current (I), dis-
charge voltage (U), and 1 kΩ ballast resistor (R), is presented
in figure 6. By adjusting the voltage (U), the gas eventually
reaches the breakdown state. This allows for the determina-
tion of discharge results. All the models connect their tops to
the anode and their bottoms to the cathode.
Gas discharge spaces containing particles such as electrons,
ions, and neutrals undergo various physical and chemical reac-
tions, including ion collisions, electron collisions, excitation,
and de-excitation. These reactions can be described by a set
of multi-physics field equations. Specifically, the continuity
equation for electron density is defined as follows:
∂
∂t
(ne) + ∇ · [−ne (µe
· E) − De
· ∇ne] = Re.
(1)
Figure 2. The percolation model of porous dielectric slices.
Figure 3. Percolation model with same porosity (p = 0.8) and
lattice size (a = 1/16). (No. p0.8a16-A, B, C, D).
Boolean random matrices. The resulting matrices were con-
verted into tables, with the matrix elements consisting of bin-
ary numbers. The porosity parameter p was used to adjust
the values of these binary numbers, which can be specified
as a percentage of their true values. The height parameter
h represents the number of cells on each side of the square
3
Plasma Sources Sci. Technol. 32 (2023) 115006
Y Hu et al
Assume the density and flux of electrons are ne and
−→
−→
Γe, the unit vector towards the dielectric wall is
n , elec-
tron thermal velocity is ve
th, the mean energy of emitted
electrons is eε and the secondary electron emission coeffi-
cient is γ. Γi represents the flux of ith species. The bound-
ary conditions for the electron and electron energy flux
on the dielectric walls and electrodes will be specified as
follows [56]:
( −→
·⃗n = 1
Γe
−→
Γε ·⃗n = 5
2 ve
6 ve
·⃗n
− δeγΓi
thne
·⃗n
thnε − δe ˜εγΓi
.
(5)
Figure 6. Porous dielectric discharge external circuit diagram. The
porous dielectric is placed between the anode and the cathode.
The continuity equation for the average electron energy is
defined as:
∂
∂t
(nε) + ∇ · [−nε (µε · E) − Dε · ∇nε]
· ∇ne] = Rε.
· E) − De
+ E · [−ne (µe
(2)
The two equations above involve several key variables,
including the electron density (ne), average electron energy
density (nε), electron mobility (µe), energy mobility (µε), elec-
tron diffusivity (De), and energy diffusivity (Dε). Additionally,
Re and Rlε represent the electron density and energy source
terms, respectively. Furthermore, the continuity equation for
other matter, such as ions and neutral particles, can be
expressed as follows:
ρ
∂
∂t
(wk) + ρ (u · ∇) wk = ∇ · jk + Rk.
(3)
In the above equation, ρ represents the mixture density, u
represents the average velocity of the fluid mass, wk represents
the mass fraction of substance k, and Rk is the source term for
substance k. The flux of substance k, denoted as ljk, can be
expressed as:
jk = ρwkDk
(cid:19)
(cid:18)
∇wk
wk
+
∇Mn
Mn
− ρwkZkµkE
(4)
where diffusion coefficients, charges, and mobilities of sub-
stance k are represented by Dk, Zk, and µk, respectively, while
the average molar mass of the mixture is represented by Mn.
The temperature of other heavy particles is assumed to be
300 K, and electron transport is described using the drift-
diffusion approximation. All reaction equations occur in the
porous dielectric voids and are considered in the plasma mod-
eling as shown in table 1.
if
In these equations,
−→
·⃗n > 0 then δe = 1, other-
Γe
wise δe = 0. According to the research from Swanson and
Kaganovich [57], it is assumed that ion bombardment on
dielectric walls will not result in secondary electron emission.
It is further assumed that the cathode wall will act as the sole
source of secondary electrons, with the coefficient being set to
0.1. Additionally, the 2D porous dielectric is considered to be
an ideal dielectric, without any net charge migration occur-
ring in or out of its walls, and the charge is therefore con-
served. The accumulation of charge on the dielectric surfaces
can be mathematically represented using the formula provided
below [58]:
(cid:16)−→
D1
−
(cid:17)
−→
D2
⃗n ·
= ρs
= ⃗n ·
−→
Ji +⃗n ·
−→
Je .
∂ρs
∂t
(6)
(7)
The initial reaction argon pressure is maintained at 1 torr,
under a gas temperature of 300 K, with an initial electron
−3, and an initial average electron energy
density of 1013 m
of 4 eV. The Maxwellian distribution function is used in this
simulation to model the energy distribution. To determine the
breakdown stage, we obtained a voltage-current curve. The
voltage is measured as the potential difference between the
electrodes, while the current is computed as the sum of elec-
tron and ion currents. The resulting voltage-current curve can
be divided into three distinct stages, as described by [54]: the
Geiger–Müller regime, the Townsend discharge regime, and
the subnormal glow discharge regime. The breakdown voltage,
defined as the voltage at which the discharge transitions
from the Townsend regime to the subnormal glow regime
and exhibits negative differential resistance, can be determ-
ined by identifying the characteristics of the voltage–current
curve [59].
As the plasma is not generated within the dielectric, we
have adopted a strategy of coarsening the mesh of the dielectric
and refining the mesh of the pore domain. In addition, bound-
ary layers have been added to all walls to enhance the accuracy
of the mesh.
4
Plasma Sources Sci. Technol. 32 (2023) 115006
Y Hu et al
Table 1. Argon reaction kinetic process with rate coefficients.
No.
Reaction
Rate coefficient
References
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
e + Ar → e + Ar
e + Ar → e + Ar (4s)
e + Ar (4s) → e + Ar
e + Ar → e + Ar (4p)
e + Ar (4p) → e + Ar
e + Ar → 2e + Ar+
e + Ar (4s) → e + Ar (4p)
e + Ar (4p) → e + Ar (4s)
e + Ar (4s) → 2e + Ar+
e + Ar (4p) → 2e + Ar+
Ar (4s) + Ar (4s) → e + Ar + Ar+
Ar (4p) + Ar (4p) → e + Ar + Ar+
Ar (4s) + Ar (4p) → e + Ar + Ar+
Ar (4s) + Ar → Ar + Ar
Ar (4p) + Ar → Ar (4s) + Ar
2e + Ar+ → e + Ar
e + Ar+ + Ar → e + Ar + Ar
Cross-section
Cross-section
Cross-section
Cross-section
Cross-section
Cross-section
Cross-section
Cross-section
Cross-section
Cross-section
k(m3s−1) = 1.62 × 10
k(m3s−1) = 1.62 × 10
k(m3s−1) = 1.62 × 10
k(m3s−1) = 2.3 × 10
−18
k(m3s−1) = 5 × 10
k(m6s−1) = 8.75 × 10
k(m6s−1) = 1.5 × 10
−16(Tg(K)1/2)
−16(Tg(K)1/2)
−16(Tg(K)1/2)
−21
−39T
−4.5
(eV)
e
−40(Tg (K) /300)
−2.5
[46]
[46]
[46]
[46]
[46]
[46]
[47]
[48]
[49]
[50]
[51]
[51]
[51]
[52]
[53]
[54]
[55]
to the bottom left. The electrons were found to be concentrated
in the porous region located at the bottom left.
By comparing the physical parameters of real and percola-
tion models in figure 8, it can be seen that the trends and crit-
ical points of change are very similar, and the important turn-
ing points of the curves are all located at 3.16 µs. The average
electron energies of both models decrease to final values of
10.504 eV and 10.784 eV with an error of 2.7%. The elec-
tron densities of both models reach their maximum with an
error of 16.9%. The final values of global currents are 0.161 A
and 0.152 A, respectively, with an error of 5.9%. The elec-
trode potentials of both models decrease from the same value
of 500 V, and eventually end at 338.7 V and 348.2 V, respect-
ively, with an error of 2.8%.
The solution time of
3.1.2. Decreased computational cost.
the real model is 7421 s while that of the percolation model
is only 3933 s for a pressure of 1 torr and a voltage of 500 V.
Therefore, the simulation speed of the percolation model is
about 2 times faster than that of the real model. The reduction
in the simulation time is mainly due to the improvement of the
mesh. The pore boundaries of the percolation model are par-
allel or perpendicular to the dielectric contact wall, then the
meshes become more regular. Besides, the reduction in resol-
ution also decreases the number of meshes.
The resolution of the percola-
3.1.3. Sensitivity analysis.
tion model needs to be kept at a certain limit to properly
exploit its advantages. An overdense percolation structure, as
in figure 7(b), has similar errors as model (c), but the simula-
tion time is 6487 s, which does not significantly improve the
simulation speed. A too-coarse percolation structure can cause
huge deviations in the simulation parameters and even change
the original breakdown channel, as in figure 7(d).
Figure 7. Electron density distributions of real (a) and percolation
−3, lattice size:
(b)–(d) models under the breakdown stage. (unit: m
(b) 0.0099 cm, (c) 0.0158 cm, (d) 0.0480 cm).
3. Results and discussions
3.1. Plasma discharge simulation comparison between real
and percolation model
The simulation was con-
3.1.1. Close physical parameters.
ducted under controlled experimental conditions, whereby the
pressure was 1 torr and a voltage of 500 V was applied to
the anode. This simulation used two sorts of distinct models,
namely the real porous dielectric model and the percolation
model. The resultant electron density distributions obtained
from the simulation are depicted in figure 7.
As shown in figure 7, it can be observed that the electron
distribution obtained from the percolation models (b) and (c)
closely resemble that obtained from the real model (a), with
both producing a breakdown channel that extends from the top
5
Plasma Sources Sci. Technol. 32 (2023) 115006
Y Hu et al
Figure 8. Average electron energy and electron density curves (a)
for real and percolation models. Electrode potential and global
current curves (b) for real and percolation models. (lattice size:
0.0158 cm).
Figure 9. Schematic diagrams (a) and Paschen curves of
percolation models with the same porosity and lattice size (b).
3.2. Plasma discharge simulation of the percolation models
with the same porosity
To investigate the percolation behavior of the models, we
maintain the same initial voltage and breakdown determin-
ation and perform simulations on four percolation models
(p0.8a1/16-A, B, C, D) with porosity p = 0.8 and lattice
size of 1/16. For each random model, we evaluate the crit-
ical breakdown voltage at 12 different pressure values ran-
ging from 0.3 torr to 50 torr, using the dichotomous method
for approximation. Given that the distance between the elec-
trodes remains constant, we plot the Paschen curve by using
the logarithmic coordinates of the pressure as the horizontal
axis and the breakdown voltage as the vertical axis, as shown
in figure 9(b).
The curves reveal that Paschen curves of percolation mod-
els with the same porosity and lattice size exhibit similar
trends, with minimum breakdown pressure values ranging
from 1 to 4 torr corresponding to the breakdown voltage.
We note two main differences in the Paschen curves. This
finding suggests that, while percolation models with similar
porosity and lattice size display similar Paschen curves, the
subtle differences in geometric and plasma characteristics can
still have a significant impact on breakdown behavior.
3.2.1. Vertical shift of Paschen curves caused by changes in
The lowest point of the curves varies between
tortuosity.
these models. The main factors for the generation of this phe-
nomenon can be explained by the capillary network theory
[59]. According to this theory, the breakdown voltage of por-
ous dielectric is controlled by three parameters: capillary tor-
tuosity, average line porosity, and radius. The tortuosity is the
ratio of the actual electron drift distance to the vertical distance
between the electrodes and is defined as:
τ =
Lt
L
.
(8)
The average line density is determined by both the channel
length Lp of the dielectric wall contact and the electron drift
channel length Lt, and is defined as:
6
Plasma Sources Sci. Technol. 32 (2023) 115006
ϕ l = 1 − Lp
Lt
.
(9)
¯λ =
√
kEav
2π Pda
2
Y Hu et al
(13)
We define the average line density as the percentage of
dielectric square blocks in the average unit vertical area. The
radius is calculated as the side length of the unit square, which
is set to 1. As the porosity is fixed, the average line density is
also constant. Consequently, the primary factor affecting the
Paschen curve is the variation in tortuosity. Our findings indic-
ate that changes in tortuosity play a critical role in shaping the
Paschen curve, even when the porosity and average line dens-
ity remain constant.
According to the Derivation of the breakdown model in por-
ous dielectric [59], the breakdown criterion is given:
0
dˆ
@
0
γ=
τ [δ (l) · α (E (l)) (1 − ϕ l) + α (E (l)) · ϕ l] dz
−1
1
A
.
(10)
In this equation, α and γ are the ionization coefficients. d
is the distance from the anode to the cathode. δ (l) is the loss
probability, which can be defined by:
"
δ (l) =
1 − 4
(cid:18)
Dr,e
APµeE (l)
(cid:19)
1/2 1
R
#
+ 4
Dr,e
APµeE (l)
1
R2
.
(11)
In this equation, Dr,e is the radial diffusion coefficient of
electrons. µeE (l) is the axial drift speed of electrons. P rep-
resents pressure and A is a constant. R is called lattice size, an
important parameter in the theory. E (l) represents the electric
field in the straight capillary through tortuosity, considering
any position l in the capillary.
Model A has low tortuosity and high minimum breakdown
voltage due to the absence of vertical breakdown channels. On
the other hand, model D has a highly inhomogeneous dielec-
tric distribution that blocks large tortuous channels, requir-
ing a very high voltage to break down, resulting in a severely
upward-shifted Paschen curve. Model B contains a vertical
channel distributed on the right side with a wider channel
width compared to model A, resulting in an overall elevated
Paschen curve compared to model C.
where P is the pressure, Eav is the temperature and da rep-
resents the effective diameter of particles. The Debye length
is proportional to the square root of the result of dividing the
electron kinetic energy by electron density, defined as follows:
r
λDe =
ε0kEav
ne0e2
(14)
where Eav is the electron kinetic energy and ne0 is the electron
density. ε0 represents the permittivity of free space, k is the
Boltzmann constant, and e is the elementary charge.
At the breakdown stage, the electron kinetic energy of mod-
els B and C are 5.33 eV and 5.23 eV, while the electron dens-
−3. The Debye
ities are 7.93 × 1015 m
lengths of models B and C are 0.0193 cm and 0.0243 cm,
respectively. Obviously, the Debye length λDe of model C is
larger.
−3 and 4.90 × 1015 m
To meet the conditions of critical breakdown, the mean free
path should increase with the Debye length. Theoretically, the
pressure needs to be reduced, so that the Paschen curve shifts
left.
In summary, the percolation model reveals that the Paschen
curves of porous dielectric with the same porosity tend to be
the same by merely changing the random distribution of por-
ous regions. However, because the specific distribution of por-
ous areas affects the discharge channel tortuosity and plasma
Debye length, the Paschen curve shifts. High tortuosity shifts
the Paschen curve upward and large Debye length shifts it to
the left.
3.3. Plasma discharge simulation of the percolation models
with different porosities
Similar to 3.2, the simulation was repeated for random percol-
ation models with varying porosities of p = 0.9, 0.8, 0.7, and
lattice size of 1/16 to obtain the critical breakdown voltages
and Paschen curves. Figure 10 illustrates a typical breakdown
path in these models.
3.2.2. Left shift of Paschen curves caused by changes in
The Paschen curve shows a shift in the
Debye length.
extreme point (Stoletov point), which is particularly pro-
nounced in model C. This may be caused by the change in
Debye length of plasma. When the Paschen curve reaches the
Stoletov point, the mean free path of the gas is close to the
Debye length, as shown in the following equation [60]:
λDe
¯λ
≈ 1.
(12)
The presence of a porous dielectric does not affect the mean
free path, which is the characteristic length describing the col-
lision between gas molecules, and can be expressed by the
equation:
3.3.1. Upper right shift of Paschen curves caused by changes
From the data presented in the curves from
in porosities.
figure 11 , it is inferred that a lower porosity level inhibits the
breakdown process. The minimum breakdown voltages cor-
responding to porosity levels of 0.9, 0.8, and 0.7 were found
to be 342 V, 571 V, and 782 V, respectively. A reduction of
10% in porosity value results in an increase in the minimum
breakdown voltage by approximately 220 V. Furthermore, a
decrease in porosity leads to an upward and rightward shift
of the Paschen curve. This phenomenon can be attributed to
the combined effects of ambipolar diffusion and recombina-
tion caused by the presence of porosity, which can be math-
ematically described by the Paschen curve equation based on
Boltzmann’s equation.
7
Plasma Sources Sci. Technol. 32 (2023) 115006
Y Hu et al
where the scaling factor Q and the porosity p1 are added. At
constant electrode spacing d, the above equation can be sim-
plified to:
Vb =
ap
ln (bp1p)
.
(17)
In equation (17), a and b are constants. By setting the deriv-
ative of (16) to 0. Since the pressure p is positive, we get:
p =
e
bp1
(18)
where e is the Euler’s number. The horizontal coordinate of the
extreme point moves to the right when porosity p1 decreases
and p increases. Substituting the value of p into equation (17),
we get
Vb =
ae
bp1
.
(19)
When p1 decreases, Vb increases and the vertical coordinate
of the extreme point shifts upwards. Therefore, it can explain
how the Paschen curve moves in the simulation.
The simulation of percolation models with different poros-
ities illustrates the accuracy of the percolation model once
more. It is concluded that the lower porosity leads to a higher
minimum breakdown voltage, shifting the Paschen curve to
the upper right. This can be explained by the Paschen curve
equation incorporating the ambipolar diffusion and recombin-
ation for the porous dielectric.
3.4. Plasma discharge simulation of the percolation models
with different lattice sizes
Similar to the above experiments, we simulated percolation
models with porosity p of 0.8 and pore side lengths of 1/16,
1/32, and 1/48 to obtain critical breakdown voltages and
Paschen curves. The simulation model has a clear breakdown
channel between the positive and negative electrodes.
3.4.1. Upward shift of Paschen curves caused by changes in
lattice sizes. As illustrated in figure 12, the influence of lat-
tice size on the percolation model primarily manifests as an
upward shift of the Paschen curve. According to the capillary
network theory, lattice size stands for capillary radius, which
is an important parameter that influences breakdown voltage in
equations (10) and (11) [59]. Decreasing the lattice size results
in a higher minimum breakdown voltage due to the alteration
of the average width of the discharge path. The average widths
of the discharge paths show a gradual decrease from 0.14 to
0.082 and 0.072, respectively. This decrease in the average
widths of the discharge paths restrains diffusion and induces
more electron loss. Finally, it poses a greater challenge to the
initiation of plasma discharge, ultimately leading to an elev-
ation of the Paschen curve and an increase in the breakdown
voltage.
Figure 10. Breakdown channel in model p0.8a1/48 under the
−3).
condition of 1400V, 5 torr (unit of electron density 1018 m
Figure 11. Paschen curves of percolation models with different
porosities.
The conventional Paschen curve equation is shown
below [61]:
Vb =
Bpd
(cid:18)
(cid:19)
ln
Apd
ln(1+ 1
γ )
(15)
where p is the pressure, d is the breakdown distance, γ is
the coefficient, and A and B are constants. Vb represents the
breakdown voltage. Ionization occurs due to the collision of
electrons and atoms, while both disappear in the dielectric
wall recombination because of ambipolar diffusion. Assuming
that the porous dielectric is absolutely random and the aver-
age energy of the ion is much smaller than the electron’s, the
Paschen curve equation applicable to the porous dielectric can
be obtained as:
Vb =
Bpd
(cid:18)
(cid:19)
ln
App1
103Q ln 1
γ
(16)
8
Plasma Sources Sci. Technol. 32 (2023) 115006
Y Hu et al
of porosity and lattice size on discharge in porous dielec-
tric through percolation modeling. Additionally, the study
reveals differing Paschen curve displacement trends between
discharges in porous dielectric and traditional glow discharges.
These findings suggest that discharge can be manipulated by
deliberately adjusting structural factors such as porosity and
lattice size. Moreover, practitioners can utilize 3D printing and
sintering techniques to actualize this method in creating real
3D porous dielectric, thereby applying the outcomes of this
research to regulate discharges. With future breakthroughs in
algorithms and arithmetic processing capabilities, it is expec-
ted to delve deeper and explore the mechanism of porous
dielectric discharge with 3D percolation structures.
Data availability statement
The data cannot be made publicly available upon publication
because they are not available in a format that is sufficiently
accessible or reusable by other researchers. The data that sup-
port the findings of this study are available upon reasonable
request from the authors.
Acknowledgments
research was
This
supported by the China National
Postdoctoral Program for Innovative Talents (BX20200069),
the National Natural Science Foundation of China for Key
Projects (52237010), the Fundamental Research Funds for the
Central Universities (2021CDJQY-043), and the Open Project
of State Key Laboratory (SKLIPR2103).
Conflict of interest
The authors declare no competing interests.
ORCID iDs
Yilong Mao https://orcid.org/0000-0003-1471-666X
Hao Wang https://orcid.org/0000-0002-9945-8155
References
[1] Wisser F M, Mohr Y, Quadrelli E A and Canivet J 2020
ChemCatChem 12 1270–5
[2] Perego C and Millini R 2013 Chem. Soc. Rev. 42 3956–76
[3] Liu L C and Corma A 2021 Nat. Rev. Mater. 6 244–63
[4] Xia Y D, Yang Z X and Zhu Y Q 2013 J. Mater. Chem. A
1 9365–81
[5] Li L B, Luo S, Zheng Z K, Zhong K, Huang W and Fang Z
2022 Ionics 28 161–72
[6] Wickramaratne N P, Xu J T, Wang M, Zhu L, Dai L and
Jaroniec M 2014 Chem. Mater. 26 2820–8
[7] He G, Liu P and Tan Q B 2012 J. Mech. Behav. Biomed.
Mater. 5 16–31
[8] Lei Q, Guo J M and Noureddine A 2020 Adv. Funct. Mater. 30
1–28
[9] Vallet-Regi M 2008 Chem. Eng. J. 137 1–3
Figure 12. Paschen curves of percolation models with different
lattice sizes.
In a nutshell, the lattice size variation alters the plasma
diffusion loss process in the porous dielectric, resulting in a
shift of the Paschen curve and a modification of the minimum
breakdown voltage.
4. Conclusion
Research of plasma discharge in the porous dielectric is still at
a preliminary stage due to challenges encountered in exper-
iments and simulations, specifically attributed to the com-
plex structural nature of porous dielectric. In this paper, we
introduce a percolation physical structure to model 2D porous
dielectric and subsequently perform a plasma discharge simu-
lation analysis. The findings of this study demonstrate that the
2D percolation model is comparable to the actual 2D porous
tomographic model concerning the global waveform, channel,
and particle density, as well as other parameters. Therefore, the
percolation model can effectively capture the physical proper-
ties of porous dielectric while significantly reducing simula-
tion time. Through an analysis of the Paschen curves gener-
ated from the discharge simulation of the percolation model
using varying porosity and lattice size parameters, the fol-
lowing conclusions have been drawn: firstly, Paschen curves
of percolation models with identical porosity values show a
similar pattern, but with a shifted peak. This shift is primar-
ily due to changes in tortuosity and plasma Debye length. An
increase in tortuosity results in an upward shift of the Paschen
curve, while an increase in the Debye radius leads to a leftward
shift of the Paschen curve. Secondly, as porosity decreases, the
breakdown voltage increases, resulting in an upward shift of
the Paschen curve due to ambipolar diffusion and recombin-
ation. Thirdly, as lattice size decreases, the discharge path’s
average width decreases, elevating the minimum breakdown
voltage and shifting the Paschen curve upwards. This study
presents a novel simulation approach for theoretical analyses
of porous dielectric, improving efficiency while minimizing
costs. Furthermore, it quantitatively investigates the impact
9
Plasma Sources Sci. Technol. 32 (2023) 115006
Y Hu et al
[10] Kumar P, Kim K H, Saneja A, Wang B and Kukkar M 2019 J.
Porous Mater. 26 655–75
[11] Hensel K, Martiˇsovitˇs V, Machala Z, Janda M, Leˇstinsk´y M,
Tardiveau P and Mizuno A 2007 Plasma Process. Polym.
4 682–93
[12] Hensel K 2009 Eur. Phys. J. D 54 141–8
[13] Kruszelnicki J, Engeling K W, Foster J E, Xiong Z and
[37] Blunt M, King M J and Scher H 1992 Phys. Rev. A 46 7680–99
[38] Andrade J S, Araujo A D and Buldyrev S V 2001 Phys. Rev. E
63 051403
[39] Andrade J S, Street D A and Shinohara T 1995 Phys. Rev. E
51 5725–31
[40] Costa U M S, Andrade J S, Makse H A and Stanley H E 1999
Physica A 266 420–4
Kushner M J 2017 J. Phys. D: Appl. Phys. 50 025203
[41] Andrade J S, Almeida M P, Mendes J, Havlin S, Suki B and
[14] Engeling K W, Kruszelnicki J and Kushner M J 2018 Plasma
Sources Sci. Technol. 27 085002
[15] Fu Y Y, Zheng B C and Zhang P 2021 J. Appl. Phys.
129 223903
[16] Fu Y Y, Krek J, Zhang P and Verboncoeur J P 2019 IEEE
Trans. Plasma Sci. 47 2011–9
[17] Go D B and Pohlman D A 2010 J. Appl. Phys. 107 103303
[18] Rumbach P and Go D B 2012 J. Appl. Phys. 112 103302
[19] Van Laer K and Bogaerts A 2016 Plasma Sources Sci.
Technol. 25 015002
Stanley H E 1997 Phys. Rev. Lett. 79 3901–4
[42] Andrade J S, Costa U M S, Almeida M P, Makse H A and
Stanley H E 1999 Phys. Rev. Lett. 82 5249–52
[43] Goebel D M and Katz I 2008 Basic Plasma Physics (Wiley)
[44] Li M, Liu R R and Lu L Y 2021 Phys. Rep. 907 1–68
[45] Saberi A A 2015 Phys. Rep. 578 1–32
[46] Hagelaar G and Pitchford L C 2005 Plasma Sources Sci.
Technol. 14 722
[47] Zapesochnyi I, Shimon L and Soshnikov A 1965 Opt.
Spectrosc. 19 480
[20] Van Laer K and Bogaerts A 2017 Plasma Sources Sci. Technol
[48] Zapesochnyi I, Semenyuk Y N, Dashchenko A, Imre A and
26 085007
[21] Van Laer K and Bogaerts A 2017 Plasma Process. Polym.
14 1600129
[22] Wang W Z, Kim H H, Van Laer K and Bogaerts A 2018 Chem.
Eng. J. 334 2467–79
[23] Van Laer K and Bogaerts A 2015 Energy Technol. 3 1038–44
[24] Wang W Z, Butterworth T and Bogaerts A 2021 J. Phys. D:
Zapesochnyi A 1984 JETP Lett. 39 141–3
[49] McFarland R H and Kinney J D 1965 Phys. Rev. 137 A1058
[50] Vriens L 1964 Phys. Lett. 8 260–1
[51] Saifutdinova A A, Saifutdinov A I, Gainullina S V and
Timerkaev B A 2022 IEEE Trans. Plasma Sci.
50 1144–56
[52] Yamabe C, Buckman S and Phelps A 1983 Phys. Rev. A
Appl. Phys. 54 214004
27 1345
[25] Liu Y K, Sabio J C and Hartman R L 2021 J. Phys. D: Appl.
[53] Lam S, Zheng C, Lo D, Dem’yanov A and Napartovich A
Phys. 54 194003
2000 J. Phys. D: Appl. Phys. 33 242
[26] Zhang Y R, Van Laer K, Neyts E C and Bogaerts A 2016 Appl.
[54] Raizer Y P 1991 Production and decay of charged particles
Catal. B 185 56–67
[27] Xu Y, Wang C L and Qin S C 2021 Acta Phys. Sin. 70 154101
[28] Zhang Y, Wang H-Y, Zhang Y-R and Bogaerts A 2017 Plasma
Sources Sci. Technol. 26 054002
Gas Discharge Physics (Springer)
[55] Dyatko N, Ionikh Y Z, Kochetov I, Marinov D,
Meshchanov A, Napartovich A, Petrov F and Starostin S
2008 J. Phys. D: Appl. Phys. 41 055204
[29] Kruszelnicki J, Ma R and Kushner M J 2021 J. Appl. Phys.
[56] Bogdanov E, Demidov V, Kudryavtsev A and Saifutdinov A
129 143302
[30] Essam J W 1980 Rep. Prog. Phys. 43 833–912
[31] Wierman J C 1982 Ann. Probab. 10 509–24
[32] Shante V K S and Kirkpatrick S 1971 Adv. Phys. 20 325
[33] Arrowsmith D K 1979 J. Math. Phys. 20 101–3
[34] Larson R G, Scriven L E and Davis H T 1981 Chem. Eng. Sci.
36 57–73
[35] Bryant S and Blunt M 1992 Phys. Rev. A 46 2004–11
[36] Chapuis O, Prat M, Quintard M, Chane-Kane E, Guillot O and
2015 Phys. Plasmas 22 024501
[57] Swanson C and Kaganovich I D 2018 J. Appl. Phys.
123 023302
[58] Fu Y, Krek J, Zhang P and Verboncoeur J P 2018 IEEE Trans.
Plasma Sci. 47 2011–9
[59] Chen K, Yao C, Mao Y, Wu F, Chen Y, Dong S and Wang H
2023 Plasma Sources Sci. Technol. 32 015001
[60] Zhang Q-Z and Bogaerts A 2018 Plasma Sources Sci. Technol.
27 035009
Mayer N 2008 J. Power Sources 178 258–68
[61] Burm K T A L 2007 Contrib. Plasma Phys. 47 177–82
10
| null |
10.1038_s41598-020-79151-y.pdf
|
Data availability
The raw data of the study are available at http://www.natur e.com/srep.
|
Data availability The raw data of the study are available at http://www.natur e.com/srep . Received: 7 August 2020; Accepted: 1 December 2020
|
OPEN
Longitudinal changes in superficial
microvasculature in glaucomatous
retinal nerve fiber layer defects
after disc hemorrhage
Yoko Okamoto, Tadamichi Akagi*, Kenji Suda, Takanori Kameda, Masahiro Miyake,
Hanako Ohashi Ikeda, Eri Nakano, Akihito Uji & Akitaka Tsujikawa
Glaucoma is a multifactorial optic neuropathy, possibly involving vascular dysfunction, leading to the
death of retinal ganglion cells and their axons. Disc hemorrhage (DH) is known to be closely associated
with the widening of retinal nerve fiber layer defect (NFLD); however, it has not been well elucidated
how DH affects retinal microvasculature. We aimed to investigate the association between DH history
and longitudinal changes in superficial retinal microvasculature in NFLD. We enrolled 15 glaucoma
patients with DH history (32 glaucomatous NFLD locations, with or without DH history). NFLD-angle,
superficial retinal vessel density (VD), and decreased superficial retinal microvasculature (deMv)-angle
were assessed using optical coherence tomography angiography for at least three times over time.
The mean follow-up period and OCT/OCTA scan interval were 21.3 ± 5.4 months (range, 12–28) and
6.8 ± 0.4 months (range, 2–18), respectively. Linear mixed-effects models showed that the presence
of DH history was significantly associated with an additional NFLD-angle widening of 2.19 degree/
year (P = 0.030), VD decrease of 1.88%/year (P = 0.015), and deMv-angle widening of 3.78 degree/year
(P < 0.001). These changes were significantly correlated with each other (P < 0.001). Thus, the widening
of NFLD was closely associated with deMv, and DH was associated with a subsequent decrease in
superficial retinal microvasculature in NFLD.
Glaucoma is a progressive optic neuropathy characterized by the degeneration of retinal ganglion cells and
their axons and results in visual field loss1. Widening of the retinal nerve fiber layer (RNFL) defect (NFLD) is an
important sign of glaucoma progression, leading to the functional deterioration of the visual field2,3.
Recently, optical coherence tomography angiography (OCTA) has enabled noninvasive assessments of retinal
microvasculature, and several studies using OCTA have revealed decreased superficial retinal vessel density (VD)
in eyes with glaucoma4–7. It was also reported that a region of decreased superficial retinal microvasculature
(deMv) in the parapapillary region was topographically associated with NFLD6–8.
Disc hemorrhage (DH) is well-known as an important risk factor for the progression of glaucoma, including
visual field defects9–12 and RNFL thinning13,14. Recently, some studies have reported close associations between
DH and parapapillary choroidal microvasculature dropout (CMvD) assessed using OCTA 15,16. However, it has
not been well elucidated how DH affects longitudinal change in retinal microvasculature.
In the present study, we investigated the longitudinal changes in NFLD, parapapillary superficial retinal VD,
and parapapillary deMv in patients with glaucomatous NFLD with or without DH history and the association
among these structural or vascular parameters.
Results
Demographic and clinical characteristics. Fifteen glaucoma patients with DH history were enrolled,
and a total of 32 NFLDs of 26 eyes were included in the analysis (Table 1). DH was detected within 3 years before
the first OCTA examination in 18 NFLDs (DH group) and not in 14 NFLDs (non-DH group). NFLD-angle, VD,
and deMv-angle were assessed in each NFLD quadrant, as shown in Fig. 1.
Department of Ophthalmology and Visual Sciences, Kyoto University Graduate School of Medicine, 54 Kawahara-cho,
Shogoin, Sakyo-ku, Kyoto 606-8507, Japan. *email: [email protected]
Scientific Reports | (2020) 10:22058
| https://doi.org/10.1038/s41598-020-79151-y
1
Vol.:(0123456789)www.nature.com/scientificreports By subjects (N = 15)
Age (years)
Sex (F/M), n
By eye (N = 26)
53.7 ± 9.1 (38–73)
8/7
Diagnosis (POAG/PPG), n
24/2
Intraocular pressure (mmHg)
13.9 ± 2.7 (10–21)
Axial length (mm)
25.1 ± 1.2 (22.9–27.0)
Central corneal thickness (μm)
525.0 ± 36.8 (436–583)
Visual field mean deviation (dB)
− 3.71 ± 3.34 (− 13.07 to − 0.22)
Medication-baseline
Medication-at last
1.4 ± 1.3 (0–4)
2.0 ± 1.1 (0–4)
OCTA follow-up period (month)
21.3 ± 5.4 (12–28)
Table 1. Demographic and clinical characteristics of included subjects (N = 26 eyes of 15 subjects). Data
(except sex and diagnosis) are presented as mean ± standard deviation with the minimum and maximum
values in parentheses. POAG primary open angle glaucoma, PPG preperimetric glaucoma, OCTA optical
coherence tomography angiography.
Figure 1. Measurements of NFLD-angle, VD, and deMv-angle. (a) Disc photograph. The white circle indicates
the boundary of the optic disc. (b) NFLD-angle (β) is determined in the OCT en face image. RPC = Radial
peripapillary capillary. (c) deMv-angle (α) is determined in the superficial OCTA image. (d) VD is assessed in
the superficial OCTA images.
Comparisons between DH and non-DH groups. Table 2 shows a comparison between DH and non-
DH groups. There were 7 superotemporal NFLDs without DH, 7 inferotemporal NFLDs without DH, 6 super-
otemporal NFLDs with DH, and 12 inferotemporal NFLDs with DH. The means ± standard deviations of follow-
up period and OCT/OCTA scan interval were 21.3 ± 5.4 months (range, 12–28) and 6.8 ± 0.4 months (range,
2–18), respectively. There were no significant differences in age, intraocular pressure (IOP), axial length, central
corneal thickness, the number of glaucoma medications, follow-up period, and the number of OCT and OCTA
examinations between groups (all P > 0.05). The mean NFLD-angle, VD, and deMv-angle at baseline were not
significantly different between the groups (all P > 0.05) (Table 2). The NFLD-angle, VD, and deMv-angle at base-
line were not significantly different between the superotemporal and inferotemporal quadrants (all P > 0.05)
(Supplementary Table 1). In the NFLDs with DH, the differences between the baseline and final measurements
of NFLD-angle and deMv-angle were statistically significant (NFLD-angle, 4.71 ± 5.97 degree, P < 0.001; VD,
− 1.21 ± 3.49%, P = 0.16; deMv-angle, 10.24 ± 8.66 degree, P < 0.001), whereas, in the NFLDs without DH, the
differences in measurements of NFLD-angle, VD, and deMv-angle were not significant (all, P > 0.05). The inter-
observer reproducibility of measurements of NFLD-angle and deMv-angle were intraclass correlation coefficient
(ICC)(2,1) = 0.978 and ICC(2,1) = 0.955, respectively, which were excellent.
Change rates of NFLD-angle, VD, and deMv-angle were evaluated using multivariable mixed-effects models
including the history of DH and mean IOP (Table 3). Signal strength index (SSI) was also included as a vari-
able in VD assessments. The models showed that the presence of DH history was associated with an additional
average NFLD-angle widening of 2.19 degree/year (95% confidence interval [CI], 0.33–4.06; P = 0.030), VD
decrease of 1.88%/year (95% CI, 0.40–3.35; P = 0.015), and deMv-angle widening of 3.78 degree/year (95% CI,
2.01–5.55; P < 0.001). Figure 2 shows the distribution of the change rates in NFLD-angle, VD, and deMv-angle
after adjusting for IOP. The rates of NFLD-angle widening, VD decrease, and deMv-angle widening were sig-
nificantly larger in the DH group than the non-DH group (NFLD-angle, 1.79 ± 0.75 degree/year vs. 0.27 ± 0.47
degree/year, P < 0.001; VD, − 0.82 ± 0.34%/year vs. 1.15 ± 0.43%/year, P < 0.001; deMv-angle, 4.78 ± 0.84 degree/
year vs. 1.08 ± 1.25 degree/year, P < 0.001). The rates of NFLD-angle widening, VD decrease, and deMv-angle
widening were not significantly different between the superotemporal and inferotemporal quadrants (all P > 0.05)
(Supplementary Table 2).
Scientific Reports | (2020) 10:22058 |
https://doi.org/10.1038/s41598-020-79151-y
2
Vol:.(1234567890)www.nature.com/scientificreports/ NFLD without DH (N = 14)
NFLD with DH (N = 18)
Mean ± SD
(Range)
Mean ± SD
(Range)
P value*
Age-baseline (years)
Mean IOP (mmHg)
Axial length (mm)
Central corneal thickness (μm)
Visual field mean deviation (dB)
Medication-baseline
Medication-at last
55.1 ± 9.1
(38–73)
52.8 ± 9.8
(38–73)
13.9 ± 2.9
(9.9–20.3)
13.7 ± 2.6
(9.9–19.0)
25.1 ± 1.3
(22.9–27.0)
24.9 ± 1.2
(22.9–27.0)
523.0 ± 39.6
(436–583)
523.0 ± 39.6
(467–581)
− 4.4 ± 3.8
(− 13.1 to − 0.22)
− 3.2 ± 1.9
(− 7.89 to 0.35)
1.5 ± 1.5
1.9 ± 1.3
(0–4)
(0–4)
1.6 ± 1.2
2.6 ± 1.0
(0–4)
(1–4)
OCTA follow-up period (month)
20.6 ± 5.6
(12–25)
22.6 ± 5.3
(12–28)
The frequency of OCTA exam
4.3 ± 1.2
(3–6)
4.2 ± 1.1
(3–6)
CMvD (+ /−), no
NFLD location (superotemporal/inferotemporal), no
4/10
7/7
10/8
6/12
deMv-angle-baseline (°)
deMv-angle-final (°)
VD-baseline (%)
VD-final (%)
NFLD angle-baseline (°)
NFLD angle-final (°)
32.9 ± 12.6
(12.7–54.6)
30.8 ± 12.7
(5.3–50.9)
34.3 ± 14.9
(11.1–60.3)
40.8 ± 13.4
(13.6–66.4)
54.3 ± 7.0
(42.4–63.0)
54.8 ± 5.5
(39.2–61.2)
> 0.99
55.9 ± 7.1
(47.5–67.4)
53.0 ± 6.8
(39.2–59.8)
17.4 ± 14.3
(5.2–52.7)
20.7 ± 13.3
(0–45.1)
18.9 ± 15.3
(5.9–54.9)
23.9 ± 11.6
(7.0–48.7)
0.24
0.51
0.30
0.18
0.99
0.48
0.77
0.31
0.72
0.48
> 0.99
> 0.99
0.14
0.36
0.65
0.20
Table 2. Comparison of clinical characteristics between NFLD with DH and NFLD without DH (N = 32
NFLD locations). Data are presented as mean ± standard deviation (SD) with the minimum and maximum
values in parentheses. CMvD peripapillary choroidal microvasculature dropout, deMv decreased superficial
retinal microvasculature, DH disc hemorrhage, IOP intraocular pressure, NFLD nerve fiber layer defect, OCTA
optical coherence tomography angiography, VD vessel density. *Analyzed using linear mixed-effect modeling
between NFLD without DH and NFLD with DH.
Rate of NFLD-angle change
Rate of VD change
Rate of deMv-angle change
Coefficients (95% CI)
P value Coefficients (95% CI)
P value Coefficients (95% CI)
P value
Time, year
− 1.63 (− 7.31, 4.05)
DH history, yes
2.20 (− 8.03, 12.43)
0.58
0.68
− 3.69 (− 13.44, 6.06)
− 0.33 (− 4.88, 4.22)
0.46
0.89
1.01 (− 4.29, 6.32)
− 0.44 (− 9.47, 8.58)
0.37
0.92
DH history × time
2.19 (0.33, 4.06)
0.030
− 1.88 (− 3.35, − 0.40)
0.015 3.78 (2.01, 5.55)
< 0.001
Mean IOP, per mmHg
0.09 (− 1.87, 2.05)
Mean IOP × time
0.13 (− 0.27, 0.54)
0.93
0.54
0.24 (− 0.79, 1.27)
− 0.10 (− 0.43, 0.23)
0.65
0.56
0.17 (− 1.58, 1.93)
− 0.00 (− 0.38, 0.38)
0.85
0.99
SSI, per unit
SSI × time
Intercept
n/a
n/a
0.11 (− 0.02, 0.24)
0.08 (− 0.04, 0.21)
0.094 n/a
0.20
n/a
16.88 (− 11.40, 45.17)
0.25
43.40 (26.75, 60.05)
< 0.001 29.99 (4.76, 55.22)
0.030
Table 3. Results of multivariable mixed effects model analysis for longitudinal changes in NFLD-angle, VD,
and deMv-angle in NFLD quadrants. Rates of changes in deMv-angle and NFLD-angle are adjusted by IOP,
and the rate of VD change is adjusted by IOP and SSI. CI confidence interval, deMv decreased superficial
retinal microvasculature, DH disc hemorrhage, IOP intraocular pressure, NFLD nerve fiber layer defect, SSI
signal strength index, VD vessel density, n/a not applicable.
Figure 3 shows a representative case with inferotemporal NFLD with DH history and superotemporal NFLD
without DH history. Longitudinal widenings of NFLD-angle and deMv-angle were observed in the inferotem-
poral NFLD, but were not apparent in the superotemporal NFLD.
Figure 4 shows the scatterplots of the linear associations among the change rates of NFLD-angle, VD, and
deMv-angle. These three parameters showed significant linear correlations (NFLD-angle and deMv-angle,
r = 0.643, P < 0.001; VD and deMv-angle, r = − 0.757, P < 0.001; NFLD-angle and VD, r = − 0.714, P < 0.001).
Discussion
The current study showed that a history of DH was a significant factor for site-specific progression of blood
flow impairment in NFLD. The decrease in VD and widening of deMv-angle were significantly correlated with
the widening of NFLD, suggesting that assessments of microvasculature using OCTA could be helpful to detect
glaucoma progression.
Previous studies have shown the association between glaucoma progressions according to function 9,11,12,17
and structure13,14,18. The location of DH also is known to relate spatially to the progressive localized thinning of
the RNFL13,14, widening of the NFLD19,20, and focal visual field progression9,11. Recent advancements in OCTA
Scientific Reports | (2020) 10:22058 |
https://doi.org/10.1038/s41598-020-79151-y
3
Vol.:(0123456789)www.nature.com/scientificreports/ Figure 2. Box plots illustrating the distribution of the rates of changes in NFLD-angle (degree/year), VD (%/
year), and deMv-angle (degree/year). Rates of changes in NFLD-angle and deMv-angle are adjusted by IOP, and
the rate of VD change is adjusted by IOP and SSI. The medians are represented by horizontal lines in the white
boxes. Boxes represent the interquartile range between the first and third quartiles. **P-value < 0.001.
have enabled assessments of the microvasculature. There have been some reports that VD measured using OCTA
can be useful to assess glaucoma progression21. Recently, Nitta et al.22 showed that a decrease in peripapillary VD
was significantly associated with DH occurrence in patients with normal tension glaucoma, which is consistent
with our results. However, the association between microvasculature reduction and RNFL thinning has not been
fully clarified. Longitudinal assessments may be necessary to elucidate this issue.
We used the deMv-angle to investigate the longitudinal decrease in superficial retinal microvasculature. This
parameter was first described by Lee et al.8 using the term “vascular impairment.” They showed that “vascular
impairment” (deMv) was almost identical to NFLD in primary open angle glaucoma (POAG) eyes having a
localized RNFL defect. In contrast, we found that the deMv was 14.5 degrees larger on average than the NFLD
measurements in the current study (deMv-angle, 35.74 ± 13.73; NFLD-angle, 21.23 ± 13.69; P < 0.001). The reason
for this discrepancy is not clear. One possible reason is the difference in the method used for the measurement
of NFLD. NFLD was evaluated using red-free fundus photographs in the previous report, whereas OCT en face
images were used in the current study. Another reason might be the difference in the threshold level between
these methods. In any case, the change rate of the widening of deMv-angle was significantly associated with that
of the widening of NFLD, which indicated that superficial microvasculature reduction was highly correlated
with structural damage in NFLD.
In the current study, the change rates of NFLD-angle, VD, and deMv-angle had significant correlations with
each other, which suggested that both OCTA and OCT parameters could be useful to detect glaucoma progres-
sion. Both OCTA and OCT could be valid and feasible assessments for glaucoma progression, although it was
not clarified in the current study. Shoji et al.21 reported that a significant decrease was more detectable in macula
VD than in ganglion cell complex thickness in some glaucomatous eyes. Moghimi et al.23 showed the possibility
that in OCTA, VD was less affected by floor effects, and no further structural change could be detected in OCT-
based thickness measurements. This evidence indicates the possibility of the different utility of these methods.
Further studies should be conducted to clarify this issue.
This study has several limitations. First, the number of eyes analyzable was small. Nonetheless, change rates
of all examined parameters (NFLD-angle, VD, and deMv-angle) were significantly different between the two
groups, and their close relationships were detected with this small sample size. Second, because the line between
the normal retina and the area of deMv to detect deMv-angle were manually evaluated, the values for deMv-angle
could be different between graders. However, because the ICCs for these parameters were excellent, we believe
that the results can be acceptable.
In conclusion, significant changes in NFLD-angle, VD, and deMv-angle were detected in NFLD with DH his-
tory more than in NFLD without DH history, and change rates of these parameters were significantly correlated
with each other. This suggests that OCTA and OCT measurements can be used to detect glaucoma progression.
Further studies are needed to determine the relationship between vascular and structural measurements and
the usefulness of OCTA measurements in clinical practice.
Methods
This longitudinal observational study was conducted at the Glaucoma Clinic of the Kyoto University Hospital.
The study adhered to the tenets of the Declaration of Helsinki and was approved by the Institutional Review
Board and Ethics Committee of the Kyoto University Graduate School of Medicine. Written informed consent
was obtained from all the patients.
Scientific Reports | (2020) 10:22058 |
https://doi.org/10.1038/s41598-020-79151-y
4
Vol:.(1234567890)www.nature.com/scientificreports/ Figure 3. Representative images on longitudinal OCTA examination of a glaucomatous eye. (a) OCTA
images obtained at the inner retinal layer show the longitudinal changes in deMv-angle of a 73-year-old man.
Inferotemporal NFLD had 13 episodes of DH. Superotemporal NFLD had no DH episode. The numbers in
the figure represent the angles of deMv in each NFLD. The deMv-angle of the inferotemporal NFLD changes
more than the superotemporal one. (b) OCT en face images show the longitudinal changes in NFLD of the
same eye. The numbers in the figure represent the angles of NFLD. The angle of inferotemporal NFLD changes
more than the superotemporal one. RPC = Radial peripapillary capillary. (c) Disc photograph at baseline shows
disc hemorrhage in inferotemporal NFLD. (d) Choroidal OCTA image of the same eye showing choroidal
microvascular dropout (white dotted line).
Participants. The participants in this study consisted of primary open angle glaucoma (POAG) and preperi-
metric glaucoma (PPG) patients with DH history in either eye, who had OCTA examination at Glaucoma Clinic
of the Kyoto University Hospital between March 1, 2015, and August 31, 2017. Inclusion criteria of this study
were: (1) open angles on gonioscopy and best-corrected visual acuity of 20/40 or better at baseline to ensure high
imaging quality. (2) Underwent more than three high-quality examinations of OCT and OCTA (signal strength
index: SSI > 50). (3) The presence of NFLD confirmed using red-free image of fundus photo at the first OCTA
examination. (4) Followed up with fundus photograph at least 3-month interval to detect the occurrence of DH.
NFLD without DH in subjects met the above inclusion criteria were included as a control group. Exclusion crite-
ria of this study were: (1) eyes with coexisting uveitis, retinal disease, or non-glaucomatous optic neuropathy. (2)
eyes with any history of intraocular surgery including cataract surgery and glaucoma surgery.
The patients had undergone a comprehensive ophthalmic examination including measurement of best-cor-
rected visual acuity (using a 5-m Landolt chart), slit-lamp examination, measurement of axial length (IOLMaster
500, Carl Zeiss Meditec, Dublin, CA), central corneal thickness (SP-3000, Tomay, Tokyo, Japan), Goldmann
applanation tonometry, gonioscopy, indirect ophthalmoscopy, dilated slit-lamp examination of the optic nerve
head, fundus photography, stereo disc photography (using a 3-Dx simultaneous stereo disc camera, Nidek,
Scientific Reports | (2020) 10:22058 |
https://doi.org/10.1038/s41598-020-79151-y
5
Vol.:(0123456789)www.nature.com/scientificreports/ Figure 4. Scatterplots illustrating the linear associations between the rates of changes in NFLD-angle, VD, and
deMv-angle. Rates of changes in NFLD-angle and deMv-angle are adjusted by IOP, and the rate of VD change is
adjusted by IOP and SSI. These 3 parameters show significant linear correlations (NFLD-angle and deMv-angle,
r = 0.643, P < 0.001; VD and deMv-angle, r = − 0.757, P < 0.001; NFLD-angle and VD, r = − 0.714, P < 0.001).
Gamagori, Japan), standard automated perimetry (Humphrey Visual Field Analyzer, Carl Zeiss Meditec) with
the 24-2 Swedish Interactive Threshold Algorithm standard program6.
Optic coherence tomography angiography. The optic nerve and peripapillary area were imaged using
a commercially available OCTA device (AngioVue; OptoVue, Fremont, CA, USA). Each image covered an area
of 4.5 × 4.5 mm and 3.0 × 3.0 mm centered on the optic disc. Each B-Scan contained 216 A- scan. To produce
images of perfused vessels, the Split Spectrum Amplitude Decorrelation Angiography software algorithm was
employed7,24. The OCTA images were coregistered with OCT B-scans that were obtained concurrently to enable
visualization of both the vasculature and structure in tandem. The area of deMv was assessed by determining the
presence of a region of decreased vasculature in the inner retina using the en face angiogram. Using the internal
limiting membrane as a plane of reference, a slab with a uniform thickness that included the RNFL, ganglion cell
layer, and inner plexiform layer was manually determined from the entire OCTA data sets using the coregistered
OCT B-scans in each eye8.
The peripapillary region was defined as a 500-μm-wide elliptical annulus extending from the optic disc bound-
ary, and segmentation of the peripapillary area was performed using the intrinsic software provided by Opto-
Vue. Vessel density was defined by the percentage area occupied by vessels, measured using the intensity-based
thresholding feature of the software, which adopted the same method of calculation as that previously reported24.
Measurement methods of NFLD-angle, VD, and deMV-angle.
In reference to previous reports8,19,
NFLD-angle was measured by identifying the two points at which the borders of an NFLD area met the clini-
cal optic disc margin (Fig. 1a), using 3.0 × 3.0 enface image of OCT. Lines were then drawn that connected the
disc center and the two points, and the angular distance between these two lines was defined as NFLD-angle
(Fig. 1b).
In the same way, deMv-angle was measured by identifying the two points at which the borders of the deMv
area met the clinical optic disc margin and the disc center, using 3.0 × 3.0 enface image of OCTA (Fig. 1c).
As for VD, the optic disc was divided into six areas (blue dotted lines) using the software. In this study, we
measured the VD of the superotemporal or inferotemporal areas (white dotted lines) in the location of NFLD
(Fig. 1d).
Statistical analysis. Values were presented as mean ± SD for continuous variables. Variables were com-
pared using linear-mixed effects modeling, where eyes were nested within subjects to properly adjust for eyes
from the same individual exhibiting similar measurements. The significance of differences between the groups
was determined after Bonferroni correction.
Linear mixed-effects modeling was used to evaluate the rates of changes in NFLD-angle (degree/year), VD (%/
year), and deMv-angle (degree/year). Details of the use of these models for assessment of longitudinal changes
in glaucoma have been previously described13,14,25–27. In brief, models were first fit with objective parameter
measurements (NFLD-angle, VD, or deMv-angle) as a response variable, whereas time, group (the presence of
DH history), and a time-group interaction were defined as fixed effects. The variable, time, was measured as time
in years from the first OCT and OCTA examination. Random intercepts and slopes were used to account for
repeated measurements over time, where eyes (right or left) and locations (superotemporal or inferotemporal)
were nested within subjects to properly adjust for NFLDs from the same individual exhibiting similar measure-
ments. Then, multivariable models were evaluated with mean IOP throughout the follow-up period and SSI (in
case of VD), which potentially affect rates of change in objective parameters, to evaluate relationships between
objective parameter measurements over time and the presence of DH history. Two-way interactions between time
and mean IOP and SSI (for VD) were used to determine whether these were significantly associated with change
Scientific Reports | (2020) 10:22058 |
https://doi.org/10.1038/s41598-020-79151-y
6
Vol:.(1234567890)www.nature.com/scientificreports/ in objective parameter measurements over time. IOP-adjusted rates of changes in NFLD-angle and deMv-angle,
and IOP-SSI-adjusted rates of VD change, were used for analyses. The statistical analyses were performed using
the R package ‘lme4′ with R version 3.3.1 (http://www.r-proje ct.org) and SPSS Version 24 software (IBM Corp.,
Armonk, New York, USA. P values less than 0.05 were considered statistically significant.
Interobserver reproducibility of NFLD-angle and deMv-angle. To evaluate the interobserver repro-
ducibility of NFLD-angle and deMv-angle, all NFLDs included in the current study were evaluated indepen-
dently by two examiners (YO and EN) blinded to any information other than OCT enface images or OCTA, and
intraclass correlation coefficients [ICCs (2,1)] were calculated.
Data availability
The raw data of the study are available at http://www.natur e.com/srep.
Received: 7 August 2020; Accepted: 1 December 2020
References
1. Weinreb, R. N., Aung, T. & Medeiros, F. A. The pathophysiology and treatment of glaucoma: A review. JAMA 311, 1901–1911
(2014).
2. Sommer, A. et al. The nerve fiber layer in the diagnosis of glaucoma. Arch. Ophthalmol. 95, 2149–2156 (1977).
3. Suh, M. H. et al. Patterns of progression of localaized retinal nerve fiber layer defect on red free fundus photographs innormal-
tension glaucoma. Eye. 24, 857–863 (2010).
4. Liu, L. et al. Optical coherence tomography angiography of peripapillary retina in glaucoma. JAMA Ophthalmol. 133, 1045–1052
(2015).
5. Yarmohammadi, A. et al. Relationship between optical coherence tomography angiography vessel density and severity of visual
field loss in glaucoma. Ophthalmology 123, 2498–2508 (2016).
6. Akagi, T. et al. Microvascular density in glaucomatous eyes with hemifield visual field defects: An optical coherence tomography
angiography study. Am. J. Ophthalmol. 168, 237–249 (2016).
7. Chen, C. L. et al. Peripapillary retinal nerve fiber layer vascular microcirculation in glaucoma using optical coherence tomography-
based microangiography. Investig. Ophthalmol. Vis. Sci. 57, 475–485 (2016).
8. Lee, E. J., Lee, K. M., Lee, S. H. & Kim, T. W. OCT angiography of the peripapillary retina in primary open-angle glaucoma. Investig.
Ophthalmol. Vis. Sci. 57, 6265–6270 (2016).
9. Ishida, K., Yamamoto, T., Sugiyama, K. & Kitazawa, Y. Disc hemorrhage is a significantly negative prognostic factor in normal-
tension glaucoma. Am. J. Ophthalmol. 129, 707–714 (2000).
10. Budenz, D. L. et al. Detection and prognostic significance of optic disc hemorrhages during the Ocular Hypertension Treatment
Study. Ophthalmology 113, 2137–2143 (2006).
11. De Moraes, C. G. et al. Spatially consistent, localized visual field loss before and after disc hemorrhage. Investig. Ophthalmol. Vis.
Sci. 50, 4727–4733 (2009).
12. Rasker, M. T., van den Enden, A., Bakker, D. & Hoyng, P. F. Deterioration of visual fields in patients with glaucoma with and
without optic disc hemorrhages. Arch. Ophthalmol. 115, 1257–1262 (1997).
13. Akagi, T. et al. Rates of local retinal nerve fiber layer thinning before and after disc hemorrhage in glaucoma. Ophthalmology 124,
1403–1411 (2017).
14. Akagi, T. et al. Association between rates of retinal nerve fiber layer thinning and previous disc hemorrhage in glaucoma. Oph-
thalmol. Glaucoma. 1, 23–31 (2018).
15. Park, H. L., Kim, J. W. & Park, C. K. Choroidal microvasculature dropout is associated with progressive retinal nerve fiber layer
thinning in glaucoma with disc hemorrhage. Ophthalmology 125, 1003–1013 (2018).
16. Rao, H. L. et al. Choroidal microvascular dropout in primary open-angle glaucoma eyes with disc hemorrhage. J Glaucoma. 28,
181–187 (2019).
17. Diehl, D. L., Quigley, H. A., Miller, N. R., Sommer, A. & Burney, E. N. Prevalence and significance of optic disc hemorrhage in a
longitudinal study of glaucoma. Arch. Ophthalmol. 108, 545–550 (1990).
18. Sugiyama, K. et al. The associations of optic disc hemorrhage with retinal nerve fiber layer defect and peripapillary atrophy in
normal-tension glaucoma. Ophthalmology 104, 1926–1933 (1997).
19. Nitta, K. et al. Does the enlargement of retinal nerve fiver layer defects relate to disc hemorrhage or progressive visual field low in
normal-tension glaucoma?. J. Glaucoma. 20, 189–195 (2011).
20. Lee, E. J., Han, J. C. & Kee, C. Location of disc hemorrhage and direction of progression in glaucoamatous retinal nerve fiber layer
defects. J. Glaucoma. 27, 504–510 (2018).
21. Shoji, T. et al. Progressive macula vessel density loss in primary open-angle glaucoma: A longitudinal study. Am. J. Ophthalmol.
182, 107–117 (2017).
22. Nitta, K., Sugiyama, K., Wajima, R., Tachibana, G. & Yamada, Y. Associations between changes in radial peripapillary capillaries
and occurrence of disc hemorrhage in normal-tension glaucoma. Graefes Arch. Clin. Exp. Ophthalmol. 257, 1963–1970 (2019).
23. Moghimi, S. et al. Measurement floors and dynamic ranges of OCT and OCT angiography in glaucoma. Ophthalmology 126,
980–988 (2019).
24. Jia, Y. et al. Split-spectrum amplitude-decorrelation angiography with optical coherence tomography. Opt. Express. 20, 4710–4725
(2012).
25. Miki, A. et al. Rates of retina nerve fiber layer thinning in glaucoma suspect eyes. Ophthalmology 121, 1350–1358 (2014).
26. Liu, T. et al. Rates of retinal nerve fiber layer loss in contralateral eyes of glaucoma patients with unilateral progression by conven-
tional methods. Ophthalmology 122, 2243–2251 (2015).
27. Medeiros, F. A. et al. Detection of progressive retinal nerve fiver layer loss in glaucoma using scanning laser polarimetry with
variable corneal compensation. Investig. Ophthalmol. Vis. Sci. 50, 1675–1681 (2009).
Acknowledgements
This research was supported in part by the Japan Society for the Promotion of Science (JSPS, KAKENHI Grant
Number 16K11267, TA). The funding organization had no role in the design and conduct of this research; col-
lection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript;
or decision to submit the manuscript for publication.
Scientific Reports | (2020) 10:22058 |
https://doi.org/10.1038/s41598-020-79151-y
7
Vol.:(0123456789)www.nature.com/scientificreports/ Author contributions
Conception and design of the study, Y.O. and T.A.; analysis and interpretation, Y.O., T.A., and A.T.; writing of
the article, Y.O. and T.A.; data collection, Y.O., T.A., K.S., T.K., M.M., H.O.I., E.N., and A.U.; final approval of
the article, all authors. T.A. had full access to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary Information The online version contains supplementary material available at https ://doi.
org/10.1038/s4159 8-020-79151 -y.
Correspondence and requests for materials should be addressed to T.A.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.
© The Author(s) 2020
Scientific Reports | (2020) 10:22058 |
https://doi.org/10.1038/s41598-020-79151-y
8
Vol:.(1234567890)www.nature.com/scientificreports/
| null |
10.1038_s41467-022-30736-3.pdf
|
Data availability
The data used in this paper are available at the following url: https://figshare.com/
articles/dataset/Manuscript_Data/16695592. In addition, Source Data are provided with
this paper, which can be used to reproduce figures without rerunning analyses. Source
data are provided with this paper.
Code availability
Analysis code used in this study is in the repository available at https://github.com/
Brody-Lab/dynamic_ephys36.
|
Data availability The data used in this paper are available at the following url: https://figshare.com/ articles/dataset/Manuscript_Data/16695592 . In addition, Source Data are provided with this paper, which can be used to reproduce figures without rerunning analyses. Source data are provided with this paper. Code availability Analysis code used in this study is in the repository available at https://github.com/ Brody-Lab/dynamic_ephys 36 .
|
ARTICLE
https://doi.org/10.1038/s41467-022-30736-3
OPEN
Stable choice coding in rat frontal orienting fields
across model-predicted changes of mind
1, Ahmed El Hady
1,2,4, Emily Jane Dennis
1,4, Alex T. Piet
1,5✉
&
J. Tyler Boyd-Meredith
1,3,5✉
Carlos D. Brody
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
During decision making in a changing environment, evidence that may guide the decision
accumulates until the point of action. In the rat, provisional choice is thought to be repre-
sented in frontal orienting fields (FOF), but this has only been tested in static environments
where provisional and final decisions are not easily dissociated. Here, we characterize the
representation of accumulated evidence in the FOF of rats performing a recently developed
dynamic evidence accumulation task, which induces changes in the provisional decision,
referred to as “changes of mind”. We find that FOF encodes evidence throughout decision
formation with a temporal gain modulation that rises until the period when the animal may
need to act. Furthermore, reversals in FOF firing rates can be accounted for by changes of
mind predicted using a model of the decision process fit only to behavioral data. Our results
suggest that the FOF represents provisional decisions even in dynamic, uncertain environ-
ments, allowing for rapid motor execution when it is time to act.
1 Princeton Neuroscience Institute, Princeton University, Princeton, NJ, USA. 2 Allen Institute, Seattle, WA, USA. 3 Howard Hughes Medical Institute,
Princeton University, Princeton, NJ, USA. 4These authors contributed equally: J. Tyler Boyd-Meredith, Alex T. Piet. 5These authors jointly supervised this work:
Ahmed El Hady, Carlos D. Brody.
email: [email protected]; [email protected]
✉
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
1
ARTICLE
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30736-3
When making decisions, animals must weigh and com-
bine the available evidence in favor of each alternative.
the
With each new observation, evidence about
underlying state of the environment gradually accumulates until
the animal is ready to act. This accumulation model successfully
describes a wide array of decisions1–3. Neural correlates of this
accumulation process are also present across many brain regions
in animals performing perceptual categorization tasks1,4. Not all
brain regions with neural correlates of evidence accumulation
play the same role in the decision making process4–6. For
example, regions important
for accumulation may represent
evidence in a continuous, graded fashion. On the other hand,
regions important for reading out choice and preparing motor
movements may have more categorical representations of the
accumulated evidence.
Hanks et al.7 characterized the neural representation of accu-
mulating evidence in rats performing accumulation of trains of
auditory click evidence. In the task, two streams of randomly-
timed auditory clicks were emitted from either side of a fixation
location and rats were trained to orient toward the side that played
a greater number of clicks. Presenting the evidence as discrete
pulses provided additional power to estimate the evolution of each
subject’s latent accumulated evidence variable on individual trials8,
increasing the resolution for estimating neural encoding of this
variable across brain regions7,9,10. Experimenters recorded from
the posterior parietal cortex (PPC) and the frontal orienting fields
term
(FOF), a frontal cortical structure implicated in short
memory and preparation of orienting movements11–13. They
found that FOF neurons encoded the instantaneous accumulated
evidence with sigmoidal tuning curves that remained stable during
accumulation7. These representations were more categorical than
representations found in PPC, providing a readout of the animal’s
provisional decision—the choice favored by the evidence pre-
sented so far—throughout accumulation7,10. While this study
could not differentiate between evidence representations resulting
from a role in motor preparation and motor-independent evi-
dence representations, temporally-precise perturbations of the
signals in FOF only impaired the animal’s choice when they
overlapped with the final time points of accumulation and not
when they occurred early in the evidence period. These results,
along with a two-node model of the FOF14, suggested that the
FOF is not involved in the accumulation of new pieces of evidence,
but provides a critical readout of the animal’s provisional decision
when it is time to act.
While these experiments were conducted using stationary
environments, many natural decisions unfold in dynamic envir-
onments. In stationary settings, all evidence samples in a trial
reflect the same underlying environmental state. This means
the best strategy is to equally weigh all samples of evidence
throughout stimulus presentation15. In this regime it is difficult to
dissociate the provisional from the final decision. In dynamic
environments, the state of the world can change while the animal
is deliberating. This means the animal should learn to discount
old evidence via leaky integration, weighing more recently pre-
information more heavily than older
sented samples of
samples16–20. Unlike stationary environments, adopting the
optimal strategy in a dynamic environment leads to frequent
fluctuations in the animal’s provisional decision.
Recent work has shown that rats and humans can learn to
adopt the optimal discounting rate in a dynamic environment16,18.
However, it is unknown whether the neural correlates of evidence
accumulation observed during putatively non-leaky integration in
stationary environments are preserved in animals performing
putatively leaky integration in dynamic environments. Here, we
recorded from FOF in rats during a dynamic accumulation of
evidence task. We tested whether the stable code observed in the
stationary environment persisted in the dynamic environment by
applying and extending a method developed to characterize neural
tuning to accumulated evidence7. The evolution of the latent
accumulation variable was estimated using a behavioral model fit
to the animal’s choice data18. In FOF, tuning to this accumulation
variable was described by a single sigmoidal tuning curve multi-
plied by a time varying gain modulation, which increased with
time early in the trial and stabilized at the time of the earliest
possible go cue.
We reasoned that if FOF neurons track the accumulated evi-
dence throughout the entire accumulation period, firing rates
should respond rapidly to changes in the provisional decision,
which in the literature are referred to for short as “changes of
mind”. Such “changes of mind” have been studied in stationary
environments when movement trajectories initiated toward one
target are subsequently revised, possibly due to continued pro-
cessing of the stimulus after initial decision commitment21,22.
They may also arise from noisy fluctuations in decision-related
neural activity23 and their timing may be inferred through neural
decoding24,25. (For clarity, we emphasize that we do not claim to
test whether the FOF encodes an abstract notion of “mind”, but
much more simply that changes in the provisional decision can be
read out from FOF activity). We used a behavioral approach to
predict the precise timing of changes of mind using the latent
state of the behavioral model fit to each rat’s choice data. We
found that FOF neurons responded to these model state change
events within 100 ms, reflecting the new provisional decision in
their activity. Recomputing the evidence tuning curves aligned to
model state changes, we confirmed that FOF neurons encode
evidence with a single tuning curve before and after changes of
mind. These results suggest that FOF maintains a stable readout
of the decision provisionally favored by the accumulated evidence
despite dynamic uncertainty in the environment and the
upcoming choice. Maintaining a stable representation of the
provisional decision may help ensure that the animal is ready
when it is time to act. Our study opens up the opportunity for
future work on the neural circuit level understanding of how
animals integrate and decide in a volatile environment.
Results
The dynamic evidence accumulation task. We trained rats
(n = 5) to perform a previously developed dynamic evidence
accumulation task18. This task requires the rat to report which of
two hidden states the environment is in at the time of a go cue. At
the beginning of each trial, the center port in an array of three
nose ports is illuminated by an LED. This invites the rat to poke
its nose into the center port, initiating presentation of an auditory
stimulus. The stimulus is composed of two trains of auditory
pulses (clicks) delivered in stereo from speakers positioned on
either side of the center port. The left and right click trains are
generated from different Poisson processes with rate parameters,
ri
R and ri
L, that depend on the state i. When the environment is in
state 1, the “go right” state, the generative click rate is higher for
¼ 2Hz). In state
the right speaker than the left (r1
R
2, the “go left” state, the click rates are reversed (r2
¼ 2Hz and
R
¼ 38Hz). Trials begin in either state with equal probability and
r2
L
switch stochastically between states with a fixed hazard rate
h = 1Hz. After a randomized duration, drawn from a uniform
distribution between 500 and 2000 ms, the stimulus ends and the
center LED turns off. This “go” cue signals the rat to withdraw
from the center port and poke its nose into one of two reward
delivery ports on either side. The animal receives a drop of water
(18 uL) if it chooses the side port corresponding to the final value
of the hidden state. Incorrect choices were signaled with a white
noise stimulus (Fig. 1a). In our dataset, roughly 33% of trials had
¼ 38Hz and r1
L
2
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30736-3
ARTICLE
Fig. 1 Rats accumulate and discount evidence in a dynamic accumulation task. a Schematic showing task events and timing. The center port is illuminated
by an LED. The rat pokes its nose into the port to initiate playback of randomly timed auditory clicks from speakers on either side. Clicks on each side are
generated with different underlying Poisson rate parameters that depend on a hidden environmental state. The stimulus duration is drawn from a uniform
distribution between 500 and 2000 ms. During that time the hidden state changes stochastically at a fixed hazard rate, h = 1Hz. At the end of the stimulus
presentation, the center LED turns off and reward is baited in the side port corresponding to the final state. b Schematic of the evolution of the
accumulation model on an example trial. Three example accumulation traces are shown for different instantiations of the noise applied at each time point
(σa) and the noise applied to each click (σs). Neighboring clicks can either depress or facilitate each other according to the adaptation parameters (ϕ and
τϕ). The evidence discounting rate (λ) determines how quickly the decision variable a decays back to zero. At the end of the trial, a choice is made by
comparing the decision variable to the decision boundary parameter B. c Frequency of state changes per trial across all rats' datasets. d Example
psychometric plot showing the probability that the rat chooses “go right” as a function of the ideal observer log-odds supporting a “go right” choice. Rat
data (black points) is overlaid on predictions of the accumulation model with parameters fit to this rat (red traces). Errorbars for rat data represent 95%
binomial confidence intervals around the mean (n = 92,468 trials from 252 sessions). e Example final state chronometric plot for the same dataset as in
(d). Accuracy (mean with 95% binomial confidence intervals) is plotted as a function of the duration of a trial’s final state and the number of state changes
in a given trial. f Psychophysical reverse correlation kernel for the same dataset as in (d) and (e). Green and blue patches indicate strength (mean ± s.d.)
of evidence favoring rightward choice as a function of time until the trial ends for rightward and leftward choices, respectively. The red patches are
corresponding predictions from the accumulation model. g Discounting parameters for each rat in this study (red points) compared to each rat in a
previously published stationary environment (lilac points; Brunton et al.8). Group medians are plotted as black horizontal lines. Source data are provided as
a Source Data file.
no state changes, 33% had one, and 34% had more than one
(Fig. 1c).
Behavioral model captures leaky integration strategy. We fit a
previously-developed behavioral model8,18 to rats’ choices using
an average of 108,126 trials per rat (63,494 to 185,091 trials each
from 118 to 308 sessions). The model (Fig. 1b) parameterizes the
process by which the evidence available in each auditory click is
integrated over time into a decision variable that guides the rat’s
choice.
The decision variable, referred to as the accumulation value a,
takes an initial value a0, drawn from a Gaussian with zero mean
i , which is fixed across trials. Each right
and an initial variance σ2
and left click increments or decrements the accumulation value,
subject to sensory adaptation governed by parameters ϕ and τϕ.
Each click also introduces additional noise with variance σ2
s .
Memory noise with variance σ2
a is introduced at each time step.
Evidence is discounted with rate λ, which parameterizes the rate
at which, in the absence of further input, a decays with time
(λ < 0) or increases with time (λ > 0). When λ < 0, older pieces of
evidence are discounted relative to newer evidence. While
decision makers in stationary environments perform best when
discounting is minimal (λ = 0), ideal observers in our task adopt a
high-level of discounting of old evidence (λ < 0), reducing the
impact of older clicks that may have been presented before a
change in the hidden state16–19. As previously described18, the
optimal discounting rate in a dynamic environment depends on
the quality of evidence, including the observer’s per-click noise
σ2
s . At the end of non-lapse trials, the rat chooses to go right if the
final accumulation value aN is greater than the decision boundary
B, and chooses to go left if aN < B. The ideal value of B is 0 and
any deviation from 0 reflects the animal’s side bias. On a fraction
of trials l, called “lapse” trials, the rat chooses randomly. Unlike
the model described by Brunton et al.8, there is no decision bound
setting the maximum magnitude of a at which the animal is fully
committed to a decision. Instead, the decision remains sensitive
to new information throughout the accumulation period. Best fit
values of this parameter were previously found to be effectively
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
3
ARTICLE
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30736-3
infinite and did not improve model fits8. For each rat, we used
maximum likelihood estimation to find the parameter set θ that
best described the rat’s choices across all behavioral trials (see
methods for mathematical details; see Supplementary Fig. 1 for all
rats’ best fit parameters). This model is highly flexible and can
capture many possible behavioral strategies8,18.
We present several assessments of task performance and model
validation. Psychometric curves show a rat’s choices as a function
of the ideal observer log-odds favoring a rightward choice, as well
as the correspondence with predictions from the behavioral
model fit to an example rat (Fig. 1d) and all rats used in this study
(Supplementary Fig. 2). Final state chronometric curves show that
performance increased with the final state duration, the elapsed
time between the final state change and the “go” cue (Fig. 1e and
Supplementary Fig. 3). Radillo et al.19 demonstrated the rate of
increase and saturation level of the chronometric curve for an
ideal observer depends only on the hazard rate and SNR of the
click rates. Psychophysical reverse correlation kernels quantify the
influence of clicks at each timepoint throughout the stimulus
period, providing an assay of the rats’ evidence discounting.
Reverse correlations for all rats in this study show heavier
weighting of clicks presented at the end of the trial compared to
the beginning (Fig. 1f and Supplementary Fig. 4).
The behavioral model parameter fits for each rat confirm that
rats used a leaky integration strategy (λ < 0). Best fit
all
discounting parameters were significantly different
from a
previously reported8 dataset of rats integrating in a stationary
environment (p < 0.01; two-tailed Wilcoxon rank-sum test, n = 5
in dynamic and n = 19 stationary environments) (Fig. 1g).
Consistent with previous work, rats adopted discounting rates
favor more recent evidence due to the environmental
that
volatility18. The best fit model parameters, along with the model-
independent behavioral assays described above, provide conver-
ging lines of evidence that the rats integrated evidence throughout
the trial, with hundreds of milliseconds and multiple sensory
clicks influencing their final decision.
FOF responses during dynamic accumulation. We recorded
from the frontal orienting fields (FOF) of rats performing the
dynamic evidence accumulation task. In 4 rats, we implanted
unilateral (n = 2 left FOF, 2 right FOF) microwire arrays at
coordinates (+2 AP; ± 1.3 ML) (Fig. 2a).
In a 5th rat, we
implanted a bilateral tetrode drive over the same coordinates.
Recordings from 69 sessions yielded 738 units across 5 animals.
See Supplementary Table 1 for a breakdown of data by rat
(Method, location). Cells were considered active and included for
further analysis if they had a mean firing rate of at least 1 Hz
during the trial (n = 579 active cells).
Individual cells show stereotyped temporal dynamics aligned to
both the onset of the trial (entering the center nose port), and the
movement following the end of the stimulus (nose out of center
port). Many individual cells had trial-averaged firing rates that
diverged throughout the trial, reaching a final value correspond-
ing to the animal’s choice (Fig. 2b; see Supplementary Fig. 5 for
spike rasters). These cells had more intermediate average values
throughout error trials, but eventually diverge according to the
animal’s choice, suggesting that firing rates reflect the animal’s
internal representation of the evidence or the motor plan. We
tested the timecourse of selectivity for single neurons to right
versus left choices by computing the area under the receiver
operating characteristic curve (AUC) and comparing it to a
permutation distribution computed by shuffling choice labels
across trials. For purposes of visualization, cells are sorted by
latency to 200ms (8 consecutive time bins) of significant AUC
values (2-tailed permutation test, 250 permutations, p < 0.05). We
present these plots for all active neurons and for a subset of pre-
movement side-selective neurons (Fig. 2c). Cells were defined as
pre-movement side-selective if their total spike counts during the
trial between the start of the stimulus and the movement away
from the fixation port were significantly different depending on
the animal’s side choice (2-tailed t test, p < 0.05). This subset
made up 17.8% of the active population (n = 103 selective). For
each neuron, the side associated with the higher spike count is
referred to as the cell’s preferred side. Following Hanks et al.7, we
focus on these pre-movement side-selective neurons because they
are most likely to play a role in decision formation.
Pre-movement side selectivity was slightly less common in
this dataset
than in previous studies of FOF in stationary
environments7. This may be a consequence of frequent changes
of mind, which create a dissociation between provisional and
final choice throughout trials in the dynamic task. Across pre-
movement side-selective neurons, we computed the average
activity conditioned on final state duration and cell preference
(Fig. 2d). We observe divergences at different latencies depending
on the final state duration (see Fig. S6 for population average
conditioned on side-choice and trial outcome as in Fig. 2b).
Stable accumulator tuning in dynamic environment. The
choice-selectivity metrics presented above reveal coding of the
final choice in average neural activity. However, during the trial,
the hidden state can change multiple times (1.22 ± 1.20 state
changes per trial). This creates frequent dissociations during the
trial between the animal’s provisional choice and the final choice,
which are rare in stationary environments. To better describe
encoding of
the provisional choice throughout the trial, we
applied and extended a method developed to quantify the tuning
of single neurons to the accumulation value at each moment
during the trial. Grouping firing rates according to the predicted
accumulation values at each timepoint, allows us to more infor-
matively combine information across trials with different hidden
state change timing, final choice, and trial outcome. Using this
method, Hanks et al.7 found that FOF neurons had a stable
encoding of evidence throughout accumulation in a stationary
environment. Here, we sought to test whether the FOF continues
to stably encode the evidence throughout trials when the envir-
onment is dynamic. We reasoned that choice-encoding might
emerge later in the dynamic environment when early provisional
decisions are less likely to be acted upon. For example, neurons
might only represent choice once the stimulus period has ended
and the animal has committed to a decision. Further, we asked
whether this encoding can still be captured by a single tuning
curve in the dynamic environment. While our cell selection
enforces that there be choice-encoding at some point in the trial,
it does not constrain the presence or stability of this encoding
over the course of the trial.
We used the approach described by Hanks et al.7, to produce a
map describing each neuron’s tuning to the accumulation value
over the course of
the trial. First, we computed the joint
distribution P(r, a, t) of each cell’s firing rate r, the instantaneous
accumulation value a, and time in the trial t. The evolution of the
distribution over a in response to right and left click trains δ
R and
δL, given by the behavioral model described above, was further
constrained using the animal’s choice y on each trial, giving the
posterior distribution
pðaÞ ¼ Pðajt; δ
; δ
; θ; a0
L
R
(cid:2) N ð0; σ2
i
Þ; yÞ:
ð1Þ
We improve on the method used by Hanks et al.7 by using an
analytical computation of the posterior distribution of accumu-
lated evidence, allowing for more accurate estimation of P(r, a, t)
(see methods). Firing rate maps are generated by computing the
4
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30736-3
ARTICLE
Fig. 2 FOF neurons encode the rat’s upcoming choice. a Coordinates used for FOF recordings (+2 AP; ± 1.3 ML). b Average firing rates for three example
FOF cells aligned to stimulus onset (left) and movement (right). Activity is conditioned on right (green) vs. left (blue) side choice, as well as hits (solid
lines) vs. errors (dashed lines). Shaded regions represent s.e.m. c Side-selectivity at each time point relative to movement for all active cells (left; firing
rate > 1 Hz) and for the subset of these cells that meet the spike count pre-movement side-selectivity criterion (right; 2-tailed t test p < 0.05). AUC is
computed on spike rates for right versus left choices. Plots are sorted by latency to 200 ms (8 consecutive time bins) of significant AUC values relative to a
distribution created by permuting choice labels across trials (2-tailed permutation test, 250 permutations, p < 0.05). d Average activity of all pre-
movement side-selective cells conditioned on final state duration and cells' side preferences. Grand-average firing rate at stimulus onset (11.6 Hz) is written
in brackets. Source data are provided as a Source Data file.
the firing rate as a function of
conditional expectation of
accumulated evidence and time, for each cell E[r∣a, t]. We present
this rate map for an example cell which is strongly tuned to the
accumulator throughout the trial, firing more when accumulated
evidence favors left choices (Fig. 3a). Because our neurons have
stereotyped temporal dynamics aligned to stimulus onset, we
subtract out the average temporal dynamics to isolate E[Δr∣a, t]
(Fig. 3b), the expected firing rate modulation by accumulated
evidence over time
E½Δrja; t(cid:3) ¼ E½rja; t(cid:3) (cid:4) E½rjt(cid:3):
ð2Þ
Following Hanks et al.7, a summary tuning curve was computed
by averaging over time to get E[Δr∣a] (Fig. 3c).
the example cell,
We extend the method by computing the rank 1 approxima-
tion of the residual firing rate map E[Δr∣a, t] using the singular
value decomposition (Fig. 3d). For
this
approximation captures 99.6% of the variance in the estimated
residual firing rate map. The mean variance explained by this
approximation for all pre-movement side-selective cells was
89.7% ± 9.8% (Fig. S10). Higher explained variance indicates that
a cell’s residual rate map can be accurately described by a single
tuning curve with linear scaling across time points. The fraction
of the variance captured by the rank 1 decomposition is positively
correlated with the total duration of side-selectivity favoring the
cell’s preferred side (Pearson’s correlation, ρ = 0.41, p < 0.01; Fig.
S10C). The approximation is equal to the outer product of the
first left singular vector u1 and the first right singular vector v1,
scaled by the first singular value s1. These terms can be rearranged
and interpreted as the outer product of a firing rate modulation,
^mðtÞ ¼ u1s1 range ðv1
Scaling by range(v1) gives
s. Our complete tuning curve approximation becomes:
^
Þ.
= range ðv1
f ðaÞ ¼ v1
^
f ðaÞ unit scale and ^mðtÞ units of spikes/
Þ and a tuning curve
rða; tÞ (cid:5) E½rjt(cid:3) þ ^mðtÞ (cid:6) ^
f ðaÞ:
ð3Þ
We computed a population average residual rate map across all
pre-movement side-selective cells by computing the residual firing
rate map E[Δr∣a, t] for each cell using z scored firing rates. The
accumulated value axis was inverted for left choice preferring cells
and then the residual firing rate maps were averaged together
(Fig. 3e). We computed the rank 1 approximation of
this
population residual rate map. This approximation explained
99.7% of the variance in the population residual rate map (Fig. 3e
middle, right). The population firing rate modulation curve ^mðtÞ
rises for the first 500 milliseconds and then plateaus at its maximum
value. Therefore, the population tuning can be described as a single
tuning curve whose modulation increases during the period of the
trial before a “go” cue is possible. The modulation stabilizes at its
maximum value during the period in which the trial may end and
the animal may need to report its decision. Despite the dynamic
environment, and changing provisional choice, we find FOF
neurons continue to stably encode the evidence with a single
tuning curve throughout evidence accumulation.
Neurons track model-predicted changes in provisional deci-
sion. If cells are stably tuned to the accumulated evidence
throughout deliberation, we should be able to see rapid responses
in their firing rates to changes in the animal’s provisional
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
5
ARTICLE
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30736-3
Fig. 3 FOF neurons encode the accumulated evidence throughout the trial despite a changing environment. a Firing rate map as a function of
accumulated evidence and time for an example neuron. Colors indicate accumulated evidence value with the same colors as in (b) and (c). b Residual rate
map in which the mean temporal trajectory is subtracted. c Tuning curve averaged over time (n = 33 time bins). Points indicate mean (±s.e.m.) across time
of the change in firing rate relative to temporal average as a function of accumulated evidence value a. d Rank 1 approximation of the residual rate map
E[Δr∣a, t] from (b). The approximation (left) is equal to the outer product of a modulation over time ^mðtÞ (middle) and a tuning curve ^rðaÞ (right). e Average
residual z scored firing rate map (left). This plot is produced by averaging over the residual z scored firing rate map of all pre-movement side-selective cells.
This map is approximated by the outer product of a modulation curve (middle) and a tuning curve (right). Source data are provided as a Source Data file.
decision. Unlike the previous analysis (Fig. 3), here we isolate
time points around model-predicted changes of mind, grouping
data only by the inferred provisional decision. This approach
allows us to confirm that the stable choice coding seen in Fig. 3 is
not an artifact produced by averaging in a subset of trials with
stronger coding and fewer changes of mind.
To look at responses to model-predicted changes of mind, we
computed each cell’s average deviation from its mean temporal
trajectory aligned to time points when the behavioral model
predicted a change in the animal’s estimate of the environmental
state (Fig. 4a). Following Hanks et al.7, we introduced a 100 ms
response lag between model-predictions and FOF responses. For
this analysis, model state changes were selected at time points
when a 100 ms running average of the posterior mean crossed the
decision boundary B. To avoid introducing noise into this
analysis, model state changes in the first and last 200 ms of the
trial were excluded, as were state changes that immediately
reversed to the previous state (see methods). For each cell, this
method produced two state-change triggered response curves
describing responses to changes into states 1 (STR1) and 2 (STR2).
STRs are also referred to as STRpref and STRnon-pref according to
cells’ previously determined side-preference. STRs are shown for
an example neuron (Fig. 4b). Discriminability before and after
model state changes was measured using d’ and tested for
significance by permuting the state-change labels across trials (2-
tailed permutation test, 250 permutations, p < 0.05).
response is
To visualize the state change triggered response across the
neural population, each cell’s
summarized by
computing the difference between the z scored STR for state
changes into the preferred state and into the non-preferred state
(STRpref − STRnon − pref). We present these data as a heat map
for all pre-movement side-selective cells (Fig. 4c). The z scored
STRs were averaged across these cells to give the average state-
change triggered response across the population (Fig. 4d). We
apply the permutation procedure described above to each cell and
compute the fraction of the included cells that significantly
encode state at each timepoint relative to the state change
(Fig. 4e). If cells are encoding the animal’s provisional decision,
6
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30736-3
ARTICLE
Fig. 4 FOF neurons track changes in the provisional decision. a Schematic explaining method used to compute state change triggered responses (STR). A
given trial has a hidden environmental state (blue and green bar) used to generate click trains from each speaker. We compute the posterior distribution of
accumulated evidence given the choice at each time point, p(a). We find time points where the smoothed posterior mean crosses the decision boundary
and label these model-inferred state changes. We then select the residual smoothed firing rates from the 550 ms before and after each state change and
average together the residual responses for changes into state 1 and changes into state 2. b STR (mean ± s.e.m.) for the example cell used in panel A.
Significance bars indicate time points when d0 for discriminating model state is different from chance (2-tailed permutation test, 250 permutations,
p < 0.05). The trace showing changes into the cell’s preferred state (state 2 for this cell) is labeled STRpref (solid line) and the trace for changes into the
cell’s non-preferred state is labeled STRnon-pref (dashed line). c Heat map showing difference between responses for changes into the preferred and non-
preferred state (STRpref − STRnon − pref) for each of the pre-movement side-selective cells. d Average z scored STR (mean ± s.e.m.) across all pre-
movement side-selective cells for state changes into cells' preferred states and non-preferred states. e Percentage of included cells (mean ± s.e.m.) with
significant encoding across time relative to model predicted state changes (red trace) and generative state changes (gray trace). Source data are provided
as a Source Data file.
we expect them to take intermediate firing rates during changes of
mind and not show significant encoding of either state. If our
behavioral model accurately predicts the timing of changes of
mind, these intermediate firing rates should coincide with model
state changes. As predicted, we find that the population reaches
its minimum fraction of cells differentiating between states at the
time of the model-predicted state change. We recomputed the
timecourse of discriminability across cells triggered on changes in
the veridical environmental
than the model-
predicted changes. When we do this, we find the time point at
which the minimum fraction of cells significantly discriminates
between states is delayed relative to generative state changes. This
is consistent with the FOF tracking changes in the sign of
accumulated evidence rather than simply responding to the
instantaneous stimulus. At the level of individual cells and across
the population, we see rapid responses to changes of mind,
providing further evidence that neurons track the animal’s
provisional decision throughout the accumulation process.
rather
state,
Stable evidence tuning before and after changes of mind. To
further characterize cell tuning to accumulated evidence during
changes of mind, we recomputed the tuning maps aligning time
to model-predicted state changes instead of the start of the trial.
This analysis is restricted to the time points around model state
changes as in Fig. 4, but also allows us to more closely examine
the stability of tuning before and after these events. The com-
putation and rank 1 decomposition of the tuning curves pro-
ceeded in the same manner as before except time in each trial was
aligned to model state changes:
rða; t (cid:4) tc
Þ (cid:5) E½rjt(cid:3) þ ^mðt (cid:4) tc
Þ (cid:6) ^
f ðaÞ
ð4Þ
where tc is the timing of model state changes. Consistent with the
state-change triggered responses and previous tuning curve ana-
lysis, we see that tuning in example neurons and the population is
well described by a single evidence tuning curve multiplied by a
temporal modulation before and after state changes (Fig. 5a, b).
The rank 1 approximation for the example cell presented in
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
7
ARTICLE
a
Cell 18181
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30736-3
Rank 1
Rank 1
a
≈
*
R
F
d
e
z
i
l
a
m
r
o
N
Time from model state change (s)
Time from model state change (s)
Accumulated value (a)
b
Average of selective cells
Rank 1
Rank 1
≈
*
R
F
d
e
z
i
l
a
m
r
o
N
Time from model state change (s)
Time from model state change (s)
Accumulated value (a)
Fig. 5 Stable tuning curve captures responses before and after model state changes. a Example cell tuning map triggered on model-predicted state
changes with rank 1 approximation derived temporal modulation and evidence tuning. Data is excluded from the 300 ms around the state change where
the accumulated value distribution is too narrow to estimate tuning (dotted lines). b Average of all pre-movement side-selective cells' tuning maps
computed with z scored firing rates and triggered on model-predicted state changes along with rank 1 approximation derived temporal modulation and
evidence tuning for the average map. Source data are provided as a Source Data file.
Fig. 5a explains 98.9% of the variance in the tuning map and the
average variance explained for all selective cells is 84.9% ± 9.8%.
The population average across z scored tuning maps for all pre-
movement side-selective cells is also well-described by the rank 1
approximation, which captures 89.2% of the variance (Fig. 5b).
This demonstrates that neurons encode the accumulated evidence
with a single tuning curve even at the times when the hidden state
and provisional decision fluctuate.
Discussion
We recorded neural activity from the frontal orienting fields
(FOF) of rats performing a dynamic decision-making task
designed to induce frequent changes of mind. In our study, rats
integrated sequential pieces of information, discounting older
evidence,
to track changes in a volatile hidden state. FOF
responses have been characterized previously during a similar
task in a stationary environment where rats learn to equally weigh
all evidence and changes of mind are rare7. This previous work
revealed categorical encoding of population activity to the accu-
mulated evidence, characterized by a single tuning curve
throughout the trial. This suggested that FOF encoded the pro-
visional decision during evidence accumulation. However, in a
stationary environment, the provisional decision rarely differs
from the final choice meaning that preparatory activity could
begin without needing to be reversed. In a dynamic environment,
where changes of mind are frequent, it might be advantageous to
suppress choice coding until the final decision is reached. It was
not clear whether FOF would play a similar role in representing
evidence during decision-making in a constantly-changing
environment and while the provisional decisions were still
highly flexible.
We found that FOF responses to accumulation in a dynamic
environment were similar to FOF responses during accumulation
in a stationary environment. First, a subpopulation of about 18%
of active neurons showed significant side-selectivity during the
pre-movement stimulus period. This was a smaller fraction than
previously reported, but was an expected result of a task with
more frequent stimulus-induced changes of mind. Using a
method developed by Hanks et al.7, we measured the encoding of
the decision variable in single neurons and across the population.
We improved this method by using a rank 1 approximation to
explain the evidence-encoding component of neural firing rates as
the product of a temporal modulation and an evidence tuning
curve. The rank 1 approximation supported the description of
FOF neurons with a single evidence tuning curve that was
modulated over the trial. Across the population, we found that
the temporal modulation increased until the timing of the earliest
possible “go” cue and then plateaus at a maximum modulation
strength during the rest of the trial.
The dynamic nature of the task allowed measurements that are
not possible in stationary tasks, where evidence is drawn from a
single distribution during each decision, and changes of mind are
rare. We used our behavioral model to estimate the rat’s provi-
sional decision throughout each trial. Fluctuations in this model
state variable provided an estimate of the timing of changes of
mind for analysis of neural activity. If the neurons use a single
evidence tuning curve throughout accumulation, we expect the
neural firing rates to encode the provisional decision before
and after changes of mind. Computing state change triggered
responses for each neuron, showed that FOF cells responded
rapidly to model state changes, reflecting the new provisional
decision in their firing rates. Critically, neurons encoded provi-
sional decisions both before and after these events, which implies
that provisional decisions are encoded even when they differ from
the final choice. Neuronal responses were better aligned to state
changes predicted by the behavioral model than to changes in the
8
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30736-3
ARTICLE
true environmental state, suggesting that these responses were not
simply reflecting a change in sensory experience. Combining this
approach with the method for computing accumulated evidence
tuning maps, we found, as described above, that the product of a
single evidence tuning curve and temporal modulation was still
sufficient to explain the evidence response across model state
changes (rank 1 approximation). We observed that, after the
moment when “go” cues could arrive, the temporal modulation of
evidence tuning was, on average, stable. Together, our results
demonstrate that FOF neurons encode the animal’s provisional
decision and respond rapidly, updating this representation fol-
lowing changes of mind.
One important limitation of our study is that our evidence
accumulation model only uses one fixed set of parameters to
describe each rat’s behavior in a trial in terms of the stimulus on that
is highly flexible and captures average
trial. While the model
behavior, it does not allow parameters to change over trials, nor does
it capture trial-to-trial history effects. Future work should develop
more flexible behavioral models to capture slow drifts and sudden
state changes in the parameters that describe the animals’ strategies.
This work will allow deeper investigation into neural coding.
Changes of mind are not unique to dynamic environments and
can also occur during evidence accumulation in stationary
environments. These events can occur during stimulus pre-
sentation due to noise in the decision making process and can be
predicted from neural activity25. Changes of mind may also occur
after the subject begins to execute their choice due to post-
processing delays21 or constraints placed on action26. Our work
differs from these studies, in that we use an environment designed
to induce changes of mind and ask how neurons respond to these
model-predicted events. To our knowledge, only one other
study27 has examined neural responses to behaviorally predicted
changes of mind during evidence accumulation in a dynamic
environment, and ours is the only such study in an animal model.
Previous inactivation studies suggest that while FOF is critical
for performing actions and reporting decisions, it is not necessary
for the integration of evidence7,28. This is consistent with the FOF
representing the evidence after categorization into a provisional
choice14. Work in mouse anterior lateral motor (ALM), a com-
parable cortical region, shows that categorical signals in this
region recover quickly following photoinhibition, suggesting
categorical input from other brain regions29. In a recent study,
Finkelstein et al.30 found that ALM choice signals were robust to
distractors delivered during a delay period after the typical evi-
dence presentation period, suggesting local circuitry maintained
the choice signal. Our study considered a similar brain structure
operating in a regime where, rather than ignoring distractors, it
needed to flexibly update provisional decisions in response to new
information. These studies, along with recent modeling work14,31,
suggest a common role for the FOF and the ALM in maintaining
choice signals that are either robust to or responsive to new
information according to task demands.
The dynamic decision-making task offers a complementary
approach to typical studies of evidence accumulation in static
in constantly-changing
environments. Here, we showed that
environments FOF neurons encode provisional choices and
respond rapidly to changes of mind predicted from our beha-
vioral model. Our quantitative methods and behavioral paradigm
will be useful tools for investigation of the brain circuitry sup-
porting evidence accumulation and the decision-making process
more generally.
Methods
Subjects. Animal use procedures were approved by the Princeton University
Institutional Animal Care and Use Committee and carried out in accordance with
NIH standards. All subjects were adult male Long Evans rats (Vendor: Taconic,
Hilltop and Harlan, USA). Rats were pair-housed prior to implantation with
recording electrodes and single-housed subsequently. Rats were placed on a water
restriction schedule to motivate them to perform the task for water rewards.
Behavioral training. We trained rats on the dynamic clicks task18 (Fig. 1). Rats
went through several stages of an automated training protocol. In the final stage of
training, each trial began with the illumination of a center nose port by an LED
light inside the port. This LED indicated that the rat could initiate a trial by placed
its nose into the center port. Rats were required to keep their nose in the center
port (nose fixation) until the light turned off as a “go” signal. During center
fixation, auditory cues were played indicating the current hidden state. The
duration of the stimulus period was drawn from a uniform distribution between
500 and 2000 ms. After the “go” signal, rats were rewarded for entering the side
port corresponding to the final value of the hidden state. The hidden state did not
change after the “go” cue. Correct choices were rewarded with 18 microliters of
water. Incorrect choices were signaled by a white noise stimulus (spectral noise of
1 kHz for a 0.7 s duration). The rats were put on a controlled water schedule where
they receive at least 3% of their weight every day. Rats trained each day in training
session of around 120 min. Training sessions were included for analysis if the
overall accuracy rate exceeded 70%, the center-fixation violation rate was below
25%, and the rat performed more than 50 trials. In order to prevent the rats from
developing biases towards particular side ports an anti-biasing algorithm detected
biases and probabilistically generated trials with the correct answer on the non-
favored side.
Psychometric and chronometric curves. Task performance was assessed using
psychometric curves, chronometric curves and psychophysical reverse correlations.
For all task performance plots, rat data was overlaid on predictions from the
accumulation model described below. These predictions were made by using the
probability of a right or correct choice on each trial given by the acummulation
model in place of the actual choice observed.
Psychometric plots show the probability that the rat chose to go right as a
function of the ideal observer log-odds supporting a “go right” choice. Final state
chronometric plots show the probability of a correct choice as a function of the
final state duration, the elapsed time between the final hidden state change (or the
beginning of the stimulus, if there are no state changes) and the end of the
stimulus. Data is plotted separately for trials with 0, 1, or more than 1 state
changes.
Psychophysical reverse correlation. The computation of the reverse correlation
curves was similar to methods previously reported7,8,28. An additional step was
included, as in Piet et al.18, to deal with the changing hidden state. First, the right
and left click trains were each smoothed using a causal Gaussian filter k with a
standard deviation of 5 msec. The smoothed left clicks were then subtracted from
the smoothed right clicks, creating one smooth click difference rate d for each trial:
dðtÞ ¼ ðδ
R
(cid:6) kÞðtÞ (cid:4) ðδ
L
(cid:6) kÞðtÞ:
ð5Þ
Here, the click train δ
R is a sum of delta functions with peaks at the time of each
right click and the value 0 everywhere else. Then, the expected click difference rate
given the current state of the environment, E[d(t)∣S(t)], was subtracted from d at
each timepoint on each trial. Here, S(t) is the current environmental state. This
gives us the deviation from the expected click difference rate for each trial. This is
called the excess click difference rate or just the excess click rate.
eðtÞ ¼ dðtÞ (cid:4) E½dðtÞjSðtÞ(cid:3)
ð6Þ
Finally, we compute the choice-triggered average of the excess click rate by aver-
aging over trials conditioned on the rat’s choice y ∈ { − 1, 1}.
excess-rate ðtjyÞ ¼ E½eðtÞjy(cid:3)
ð7Þ
The excess rate curves were then normalized to integrate to one. This was done
to remove distorting effects of a lapse rate, as well to make the curves more
interpretable by putting the units into effective weight of each click on choice.
L and δ
Accumulation model. The accumulation model characterizes the decision-making
process as the evolution over time t of an accumulation value a in response to left and
right click trains, δ
R, with dynamics governed by a parameter set θ. Each rat’s
behavioral data is used to find the parameter set that maximizes the probability under
the model of the rat’s choices y. Evaluating this model with the best fit parameters
produces a probability distribution over values of a at every timepoint in the trial. We
refer to this as the forward model distribution f ðaÞ ¼ Pðajt; δ
ÞÞ.
The forward model was described previously in Piet et al.18 and will be reviewed in
detail below. To characterize neural encoding of the accumulation value, we further
constrained the accumulation value distribution on trials where we had simultaneous
neural recordings by incorporating the rat’s choice, y, to find the posterior distribution
pðaÞ ¼ Pðajt; δ
that we refer to as the backward model distribution, which we describe in the next
section.
Þ; yÞ. To do this, we computed a distribution
(cid:2) N ð0; σ2
i
(cid:2) N ð0; σ2
i
; θ; a0
; θ; a0
; δ
; δ
R
R
L
L
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
9
ARTICLE
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30736-3
The accumulation model is a stochastic differential equation that describes the
We develop a novel method of computing the posterior distribution by taking
evolution of an accumulation value a, and a sensory adaptation value C:
(cid:1)
da ¼ δ
(cid:7) η
R
(cid:7) C (cid:4) δ
(cid:7) η
L
(cid:7) C
L;t
R;t
(cid:3)
dt (cid:4) λadt þ σ
dC
dt
¼ 1 (cid:4) C
τϕ
(cid:4)
(cid:5)
þ ϕ (cid:4) 1
(cid:1)
C δ
(cid:3)
:
þ δ
L;t
R;t
adW;
ð8Þ
ð9Þ
Each sensory click is scaled by the sensory adaptation value C and multiplicative
Gaussian noise η drawn from N ð1; σ2
Þ. The model parameters θ can be described
s
i , a per-click noise variance σ2
in words as an initial noise variance σ2
s , a memory
noise variance σ2
adaptation ϕ and τϕ, a decision boundary B, which captures the animal’s bias, and a
lapse rate l. If the adaptation strength parameter ϕ < 1, then consecutive clicks are
depressed. If ϕ > 1, then consecutive clicks are facilitated.
a, a discounting rate λ, the strength and time constant of
The forward model is the solution to Eqn. (8), assuming the initial
accumulation value a0 is Gaussian distributed with zero mean and variance σ2
each moment in the trial, the forward model f ðaÞ ¼ Pðajt; δ
predicts a Gaussian distribution of accumulation values with mean μ(t) and
variance σ2(t) given by:
i . At
ÞÞ
(cid:2) N ð0; σ2
i
; θ; a0
; δ
R
L
μðtÞ ¼ μ
0e
λt þ
(cid:7) CðsÞ (cid:4) δ
L;s
R;s
(cid:7) CðsÞ
ds
(cid:3)
Z t
(cid:1)
δ
0
#Rt
¼ ∑
i
ð
λ t(cid:4)RðiÞ
e
Þ
#Lt
CðRðiÞÞ (cid:4) ∑
i
e
ð
λ t(cid:4)LðiÞ
Þ
CðLðiÞÞ
ð10Þ
σ2ðtÞ ¼ σ2
i e2λt þ
¼ σ2
i e2λt þ
(cid:5)
(cid:4)
σ2
a
2λ e2λt (cid:4) 1
(cid:5)
(cid:4)
σ2
a
2λ e2λt (cid:4) 1
Z t
þ
(cid:1)
δ
σ2
s
(cid:7) CðsÞ (cid:4) δ
L;s
(cid:7) CðsÞ
R;s
(cid:3)
e2λtds
0
#Rt
þ ∑
i
s CðRðiÞÞe2λ t(cid:4)RðiÞ
σ2
ð
#Lt
Þ þ ∑
i
s CðLðiÞÞe2λ t(cid:4)LðiÞ
σ2
ð
Þ
ð11Þ
Where δ
R,t indicates whether there was a right click at time t and C(t) tells us the
effective adaptation for a click at time t. For the discrete case, #Rt is the number of
right clicks on this trial up to time t and R(i) is the time of the ith right click.
To determine the probability of a right versus left choice, we first integrate the
accumulation value distribution in the last timepoint tN of the trial from the
decision boundary parameter B to ∞
Pða>Bjt ¼ tN
; δ
; δ
; θ; a0
L
R
(cid:2) N ð0; σ2
i
ÞÞ ¼ 1
2
1 þ erf
(cid:7)
(cid:7)
(cid:4)
(cid:5)
(cid:8)
(cid:8)
Þ
(cid:4) B (cid:4) μðtN
p
ffiffiffi
σðtN
Þ
2
:
ð12Þ
On each trial, the rat makes a random choice with probability determined by lapse
rate l. Then, the probability of a “go right” choice is given by
(cid:4)
P y ¼ 1jθ
(cid:5)
(cid:4)
¼ ð1 (cid:4) lÞP
(cid:4)
a > Bjt ¼ tN
(cid:4)
; δ
; δ
; θ; a0
L
R
(cid:2) N
(cid:4)
(cid:5)(cid:5)
0; σ2
i
(cid:4)
þ l=2
(cid:5)
(cid:5)
(cid:5)
Pðy ¼ (cid:4)1jθÞ ¼ ð1 (cid:4) lÞ 1 (cid:4) P a > Bjt ¼ tN
; δ
; δ
; θ; a0
L
R
(cid:2) N
0; σ2
i
Where
(cid:9)
1;
(cid:4)1;
y ¼
if rat chooses right
if rat chooses left
Parameters θ were fit to each rat individually by maximizing the likelihood
function:
Y# trials
ð16Þ
i
L ¼
A half-Gaussian prior was included on the initial noise σ2
PðyijθÞ:
i and accumulation noise
parameters σ2
a. The priors were set to match the respective best fit values from Brunton
et al.8. The numerical optimization was performed in MATLAB, using the function
fmincon. To estimate the uncertainty on the parameter estimates, we used the
inverse hessian matrix as a parameter covariance matrix32. To compute the hessian of
the model, we performed automatic differentiation in julia to exactly compute the local
curvature33. See the Supplementary Information for parameter estimates and
uncertainty values. Brunton et al.8 extensively analyzed how well a similar model with
an additional bound parameter recovers generative parameters, finding the model
contains one maximum likelihood point in parameter space (See Section 2.3.3-6 of the
Supplement to Brunton et al.8). We compared parameter fits in this task to those
reported in Brunton et al.8, which developed the stationary version of this task.
Posterior model. The forward model described above gives us a probability dis-
tribution over accumulation values at each time point in each trial. It also gives an
estimated probability of the rat choosing to go right or left on that trial. Observing
the rat’s choice y at the end of each trial allows us to constrain the distribution of
possible trajectories that the accumulation value could have taken. The resulting
posterior distribution (referred to as the backward pass distribution in Brunton
et al.8) is useful for analyzing the neural encoding of accumulated evidence.
ð13Þ
þ l=2
ð14Þ
ð15Þ
the product of the forward distribution and a backward distribution. Again, we
note that while Brunton et al.8 refers to the posterior distribution as the backward
pass distribution, we use the term backward distribution to refer to a distinct
distribution which constrains the final state of the accumulation value distribution
in accordance with the animal’s choice, but does not constrain the initial state. As
described above, the forward distribution assumes that the initial accumulation
value a0 is normally distributed with mean 0 and variance σ2
distribution makes no assumption about the initial distribution, but assumes that
the final accumulation value aN is uniformly distributed on the side of the decision
boundary B that corresponds to the rat’s choice. Importantly, the forward and
backward distributions are conditionally independent, conditioned on the final
value of the accumulated evidence. Given that these distributions are independent,
their product gives the posterior distribution p(a) that combines the constraints on
the initial and final distributions of accumulation values:
(cid:4)
(cid:5)
(cid:4)
pðaÞ ¼ P ajt; δ
(cid:2) N 0; σ2
i
i . The backward
; θ; a0
ð17Þ
; y
; δ
(cid:5)
R
L
/ f ðaÞbðaÞ:
ð18Þ
Where f(a) is the forward model described above, which assumes a Gaussian initial
distribution of accumulation values depending on σ2
i :
(cid:2) N ð0; σ2
i
ð19Þ
L
and b(a) is the backward distribution, which assumes the final accumulation value
is on the side of the decision boundary B corresponding to the animal’s choice y:
ð20Þ
bðaÞ ¼ Pðajt; δ
f ðaÞ ¼ Pðajt; δ
; θ; a0
; θ; yÞ
; δ
; δ
ÞÞ
R
R
L
(cid:9)
¼ Pðajt; δ
Pðajt; δ
; δ
; δ
; θ; aN
; θ; aN
L
L
R
R
≥ BÞ;
≤ BÞ;
if y ¼ 1
if y ¼ (cid:4)1:
ð21Þ
We approximated the backward distribution as a mixture distribution over a
grid of final accumulation values with spacing Δa. A unit of probability mass is
initialized at each point in the grid and the solution is given by:
± 1
bðaÞ ¼ ∑
j¼0
wjPðajt; δ
R
; δ
; θ; aN
L
(cid:2) N ðB þ jΔa; 0ÞÞ
ð22Þ
The mixture weights wj are all equal if the bin spacing is uniform. Each unit of
probability mass evolves using the same solution as the forward model, but with
time reversed. This solution is exact as Δa → 0.
For tuning curve analyses we use the full posterior distribution, for the state
change triggered response analyses we use the mean of the posterior. See
the Supplementary Information for a detailed discussion on the derivation and
evaluation of the backward and posterior model.
Microwire array recordings. Microwire array implant surgery: Four rats were
implanted with microwire arrays in their left or right FOF (n = 2 in lFOF, n = 2 in
rFOF) The target region was accessed by craniotomy, using standard stereotaxic
techniques (centred 2 mm anterior to the bregma and 1.3 mm lateral to the mid-
line). Dura mater was removed over the entire craniotomy with a small syringe
needle. The remaining pia mater, even if not usually considered to be resistant to
penetration, nevertheless presents a barrier to the entry of the microelectrode
arrays because of the high-density arrangement of electrodes in the multi-channel
electrode arrays. This dimpling phenomenon, when the electrodes are pushing the
brain cortex down without penetrating, is more pronounced for arrays with larger
numbers of electrodes. In addition to potentially injuring the brain tissue, dimpling
is a source of error in the determination of depth measurements. Ideally, if dim-
pling could be eliminated, the electrodes would move in relation to the pial surface,
allowing for effective and accurate electrode placement. To overcome the dimpling
problem, we implemented the following procedure. After the craniotomy was
made, and the dura was carefully removed over the entire craniotomy, a
petroleum-based ointment (such as bacitracin ointment or sterile petroleum jelly
(Puralube Vet Ointment)) was applied to the exact site of electrode implantation.
The cyanoacrylate adhesive (Vetbond Tissue Adhesive) was then applied to the
zone of the pia surrounding the penetration area. This procedure fastens the pia
mater to the overlying bone and the resulting surface tension prevents the brain
from compressing under the advancing electrodes. Once the polymerization of
cyanoacrylate adhesive was complete, over a period of few minutes, the petroleum
ointment at the target site was removed, and the 32-electrode microwire array
(Tucker-Davis Technologies) was inserted by slowly advancing a Narishige
hydraulic micromanipulator. After inserting the array(s), the remaining exposed
cortex was covered with biocompatible silicone (kwik-sil), and the microwire array
was secured to the skull with C&B Metabond and dental acrylic.
During a ten-day recovery period, rats had unlimited access to water and food.
Recording sessions in the apparatus began thereafter, using Neuralynx acquisition
systems. Once rats had recovered from surgery, recording sessions were performed
in a behavioral chamber outfitted with a 32 channel recording system (Neuralynx).
Spiking data was acquired using a bandpass filter between 600 and 6000 Hz and a
spike detection threshold of 30 microV.
10
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30736-3
ARTICLE
For array recordings, clusters were manually cut (Spikesort 3D, Neuralynx), and
both single- and multi-units were considered.
Tetrode recordings. Tetrode drives were 3d printed from custom designs (design
files available upon request) on a Form2 3d printer in tough resin. Each drive
consists of a drive body, a cone and cap to protect the drive body, and four bundles
of 8 tetrodes in glass tubes. Each bundle was glued together and to a cannula. Each
cannula was attached to a screw using dental cement, and cured with UV light.
Each wire from each tetrode was fed through a unique channel in a 128 channel
Electrode Interface Board (SpikeGadgets) and pinned with a gold pin. After loading
all tetrodes, trimming, and building of the drive, the day-of or night before the
surgery, we electroplated the drive in gold using a nanoZ impedence tester (White
Matter LLC) and measured impedences.
Tetrode drive implant surgery proceded as described for microwire arrays,
except we did not need to vetbond the brain surface because each tetrode bundle
produced very little dimpling. A silver wire and skull screw were used to ground the
drive. Drives were secured with metabond and acrylic until secure. Tetrodes were
advanced 0.1 mm into the brain.
During a seven-day recovery period, animals had unlimited access to water and
food. Animals were then returned to training and water restriction. To acclimate
animals to the weight of the wireless apparatus, every three days, we replaced the
cap on the implant with a new cap 3 g heavier than the previous cap. If animals’
behavioral performance or weight dropped, or if we noticed any excess tilting of the
head from the weight, we returned the animal to the previous weight and waited an
additional 2 days before moving to the next weight. This process was repeated until
the animals were behaving well with caps weighing 27 g.
Once animals were acclimated to the weight, recordings could begin. Tetrodes
were advanced 0.25 mm at a time, at least 20 h before recording. For each recording
session, the animal’s cap was replaced with a 500 mAh lithium battery, 128 Gb
Sandisk extreme plus SD card, a 160-pin Amphenol Lynx connector, and
datalogger (SpikeGadgets). At the end of each session, the datalogger, SD card, and
battery were removed and the 27 g cap replaced.
The tetrode recordings were automatically clustered using Kilosort234.
Automatically determined clusters were manually curated using the Phy GUI
(https://github.com/kwikteam/phy).
Electrophysiological analysis. We computed the firing rates for all neurons
aligned to the time of stimulus onset (when the rat first broke the center port IR
beam triggering playback of the stimulus) and to movement (when the rat first
stopped breaking the center port IR beam to make its choice). Firing rates were
computed by binning spikes into 25 ms bins and smoothing them with a casual
Gaussian filter with a standard deviation of 100 ms. Stimulus onset aligned firing
rates were masked on each trial after the movement and movement aligned firing
rates were masked prior to stimlus onset. Firing rates for example cells were
averaged over trials conditioned on choice and outcome.
Cells were considered active if their average stimulus onset aligned firing rate was
greater than 1 Hz during the time from 1 s prior to the stimulus onset to the time of
movement onset. Cells were considered pre-movement side-selective if the spike
counts during the period between stimulus onset and movement were different on
trials that resulted in a left versus a right choice (2-tailed t test, p < 0.05). The side with
the higher firing rate is referred to as the cell’s preferred side.
A population-average PSTH was computed by averaging over all trials from all
pre-movement side-selective cells conditioned on final state duration and whether
the trial ended in a choice to the cell’s preferred side.
We analyzed the timecourse of choice-selectivity by computing the area under
the receiver operating characteristic curve (AUC) at each 25 ms time bin in for the
smoothed firing rates in left choice versus right choice trials. In this application, the
receiver operating characteristic curve (ROC) treats the spike rate in a time bin as a
classifier of right versus left choice, computing the true positive rate and false
positive rate as a function of the spike threshold. The area under this curve is
equivalent to the probability that in a randomly chosen pair of right left choice
trials the firing rate in that bin will be higher on the right choice trial than on the
left choice trial. AUC values significantly greater than 0.5 indicate a preference for
right choice trials and AUC values significantly less than 0.5 indicate a preference
for left choice trials35. To compute significance, we performed a permutation test
where the left/right choice labels were permuted relative to the firing rates across
trials. For visualization purposes, we sorted cells by latency to reach 8 significant
25 ms time bins in a row (2-tailed permutation test, 250 permutations, p < 0.05).
Evidence tuning curves. We compute evidence tuning curves using a method based
on the one used in Hanks et al.7. First, the posterior accumulation distribution p(a) for
each trial is computed, providing a distribution over the evolution of the accumula-
tion values for each trial that is consistent with the rat’s choice on that trial. The joint
distribution of p(a), the firing rate r, and time t, which we will call P(r, a, t) is
computed by binning time, accumulation values, and firing rates. For each trial and
each timepoint, the probability mass in each accumulation value bin in p(a) is added
to the bin in P(r, a, t) associated with that timepoint and that firing rate. Because the
shortest trials are 500 ms, not all trials contribute to each time point, so each time bin
is normalized according to the number of trials that contribute to it. When estimating
the joint distribution we discretized our data along three dimensions: time, firing rate,
and accumulation value. Time was binned into 25 ms bins. Each neuron’s firing rate
was divided into 100 bins spanning the minimum to the maximum firing rate of the
neuron. The firing rate bin was chosen by taking the average firing rate within the
time bin. Accumulation value bin size was divided into 10 bins with width set to 1.625
except the last bin which was larger to capture the tails of the distribution. The
posterior distribution was evaluated with 1 ms time bins and accumulation value bin
size of 0.1 and then downsampled to populate the joint distribution.
To estimate each cell’s firing rate map, we compute the conditional expectation of
rrP(r∣a, t). We then
the firing rate in each accumulation value and time bin E[r∣a, t] = ∑
computed the expected difference, at each time point, from the cell’s time-averaged
firing rate as a function of the accumulation value E[Δr∣a, t] = E[r∣a, t] − E[r∣t], where
E[r∣t] is the average of E[r∣a, t] over all values of a. Average accumulation value tuning
over the full trial is computed by averaging E[Δr∣a, t] over time to get E[Δr∣a]. The
same procedure is used to compute a map of z scored firing rates. And these maps are
averaged across pre-movement side-selective cells to produce a population average.
Rank 1 approximations of the E[Δr∣a, t] are computed using the singular value
decomposition. The approximation is equal to the outer product of the first left
singular vector u1 and the first right singular vector v1, scaled by the first singular
value s1. These terms are rearranged to give the outer product of a firing rate
modulation, ^mðtÞ ¼ u1s1 range ðv1
Þ.
Scaling by range(v1) gives
tuning curve approximation becomes:
^
f ðaÞ unit scale and ^mðtÞ units of spikes/s. Our complete
Þ and a tuning curve
^
f ðaÞ ¼ v1
= range ðv1
rða; tÞ (cid:5) E½rjt(cid:3) þ ^mðtÞ (cid:6) ^
ð23Þ
The variance explained by this approximation is given by the ratio between the first
singular value and the sum of all singular values:
f ðaÞ:
:
s1
∑
isi
ð24Þ
State change triggered response. On each trial, we computed a residual firing
rate by subtracting the average firing rate at each time point during the trial. We
then aligned these residual firing rates to either the model-predicted state changes
or the generative environmental state changes. We masked firing rates before the
preceding state change and after the following state change when applicable. We
computed the mean of these residual firing rates for visualization. To test for
significant discrimination of state, we compared d0 in the real data to a permutation
distribution created by permuting state labels across state changes (2-tailed per-
mutation test, 250 permutations, p < 0.05).
We defined model-predicted state changes as time points where the running
average of the mean of the posterior accumulator value crossed the rat’s best fit
decision boundary B. The running average was computed over 100 bins of 1 ms. To
avoid introducing noisy state changes, we excluded state changes from the first and last
200 ms of the trial. We also excluded state changes that did not meet two change
strength criteria designed to identify state changes that were immediately reversed. The
first, was based on the average value of the posterior mean in the 100 ms before the
change compared to the 100 ms after the change. State changes were excluded if these
strength values were inconsistent with the direction of the identified state change. The
second state change strength was based on the slope of the running average of the
posterior mean at the time of the change. If the sign of the slope was inconsistent with
the sign of the state following the state change, this means that the accumulation value
immediately returned back to the previous state. We excluded these state changes. Our
results were robust to variations in the state change inclusion criteria.
State change triggered tuning. State change triggered tuning maps E[Δr∣a, t −
tc)] were computed using the tuning curve methods described above, but using
time relative to state changes instead of stimulus onset. Firing rates were masked
before and after the preceding and following state changes as described above. Data
was also masked in the 300 ms around the state change where the accumulated
value distribution is too narrow to estimate tuning. Rank 1 approximations and
population tuning maps were computed as above.
Reporting summary. Further information on research design is available in the Nature
Research Reporting Summary linked to this article.
Data availability
The data used in this paper are available at the following url: https://figshare.com/
articles/dataset/Manuscript_Data/16695592. In addition, Source Data are provided with
this paper, which can be used to reproduce figures without rerunning analyses. Source
data are provided with this paper.
Code availability
Analysis code used in this study is in the repository available at https://github.com/
Brody-Lab/dynamic_ephys36.
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
11
ARTICLE
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30736-3
Received: 13 May 2021; Accepted: 13 May 2022;
References
1. Gold, J. I. & Shadlen, M. N. The neural basis of decision making. Ann. Rev.
Neurosci. 30, 535–574 (2007).
2. Krajbich, I., Hare, T., Bartling, B., Morishima, Y. & Fehr, E. A common
mechanism underlying food choice and social decisions. PLoS Comput. Biol.
11, e1004371 (2015).
3. Ratcliff, R. A theory of memory retrieval. Psychol. Rev. 85, 59–108 (1978).
Brody, C. D. & Hanks, T. D. Neural underpinnings of the evidence
4.
accumulator. Curr. Opin. Neurobiol. 37, 149 – 157 (2016).
Siegel, M., Buschman, T. J. & Miller, E. K. Cortical information flow during
flexible sensorimotor decisions. Science 348, 1352–1355 (2015).
5.
6. Katz, L. N., Yates, J. L., Pillow, J. W. & Huk, A. C. Dissociated functional
significance of decision-related activity in the primate dorsal stream. Nature
535, 285–288 (2016).
7. Hanks, T. D. et al. Distinct relationships of parietal and prefrontal cortices to
8.
9.
evidence accumulation. Nature 520, 220–223 (2015).
Brunton, B. W., Botvinick, M. M. & Brody, C. D. Rats and humans can
optimally accumulate evidence for decision-making. Science 340, 95–98
(2013).
Scott, B. B. et al. Fronto-parietal cortical circuits encode accumulated evidence
with a diversity of timescales. Neuron 95, 385–398.e5 (2017).
10. Yartsev, M. M., Hanks, T. D., Yoon, A. M. & Brody, C. D. Causal contribution
and dynamical encoding in the striatum during evidence accumulation. eLife
7, e34929 (2018).
11. Kopec, C. D., Erlich, J. C., Brunton, B. W., Deisseroth, K. & Brody, C. D.
Cortical and subcortical contributions to short-term memory for orienting
movements. Neuron 88, 367 – 377 (2015).
12. Erlich, J. C., Bialek, M. & Brody, C. D. A cortical substrate for memory-guided
orienting in the rat. Neuron 72, 330–343 (2011).
13. Ebbesen, C. L. et al. More than just a “motor”: Recent surprises from the
frontal cortex. J. Neurosci. 38, 9402–9413 (2018).
14. Piet, A., Erlich, J., Kopec, C. & Brody, C. D. Rat prefrontal cortex inactivations
are explained by bistable attractor dynamics. Neural Comput. 29, 2861–2886
(2017).
15. Bogacz, R., Brown, E., Moehlis, J., Holmes, P. & Cohen, J. D. The physics of
optimal decision making: a formal analysis of models of performance in two-
alternative forced-choice tasks. Psychol. Rev. 113, 700–765 (2006).
16. Glaze, C. M., Kable, J. W. & Gold, J. I. Normative evidence accumulation in
unpredictable environments. eLife 4, e08825 (2015).
17. Veliz-Cuba, A., Kilpatrick, Z. P. & Josic, K. Stochastic models of evidence
accumulation in changing environments. SIAM Rev. 58, 264–289 (2016).
18. Piet, A. T., El Hady, A. & Brody, C. D. Rats adopt the optimal timescale for
evidence integration in a dynamic environment. Nat. Commun. 9, 4265
(2018).
19. Radillo, A. E., Veliz-Cuba, A., Josić, K. & Kilpatrick, Z. P. Performance of
normative and approximate evidence accumulation on the dynamic clicks
task. Neurons Behav. Data Anal. Theory 3, 2019. https://doi.org/https://arxiv.
org/pdf/1902.01535v3.pdf.
20. Kilpatrick, Z. P., Holmes, W. R., Eissa, T. L. & Josić, K. Optimal models of
decision-making in dynamic environments. Curr. Opin. Neurobiol. 58, 54–60
(2019).
21. Resulaj, A., Kiani, R., Wolpert, D. M. & Shadlen, M. N. Changes of mind in
decision-making. Nature 461, 263–266 (2009).
22. van den Berg, R. et al. A common mechanism underlies changes of mind
about decisions and confidence. eLife 5, e12192 (2016).
23. Atiya, N. A. A., Rañó, I., Prasad, G. & Wong-Lin, K. A neural circuit model of
decision uncertainty and change-of-mind. Nat. Commun. 10, 2287 (2019).
24. Kiani, R., Cueva, C. J., Reppas, J. B. & Newsome, W. T. Dynamics of neural
population responses in prefrontal cortex indicate changes of mind on single
trials. Curr. Biol. 24, 1542–1547 (2014).
25. Peixoto, D. et al. Decoding and perturbing decision states in real time. Nature
591, 604–609 (2021).
26. Cos, I., Pezzulo, G. & Cisek, P. Changes of mind after movement onset: a
motor-state dependent decision-making process. bioRxiv, 2021. https://doi.
org/10.1101/2021.02.15.431196. URL https://www.biorxiv.org/content/early/
2021/02/16/2021.02.15.431196.
27. Murphy, P. R., Wilming, N., Hernandez-Bocanegra, D. C., Prat-Ortega, G. &
Donner, T. H. Adaptive circuit dynamics across human cortex during
evidence accumulation in changing environments. Nat. Neurosci. 24, 987–997
(2021).
28. Erlich, J. C., Brunton, B. W., Duan, C. A., Hanks, T. D. & Brody, C. D. Distinct
effects of prefrontal and parietal cortex inactivations on an accumulation of
evidence task in the rat. eLife 4, e05457 (2015).
29. Li, N., Daie, K., Svoboda, K. & Druckmann, S. Robust neuronal dynamics in
premotor cortex during motor planning. Nature 532, 459–464 (2016).
30. Finkelstein, A. et al. Attractor dynamics gate cortical information flow during
decision-making. Nat. Neurosci. 24, 843–850 (2021).
31. Prat-Ortega, G., Wimmer, K., Roxin, A. & de la Rocha, J. Flexible
categorization in perceptual decision making. Nat. Commun. 12, 1283 (2021).
32. Daw, N. D. Trial-by-trial data analysis using computational models. In
Decision Making, Affect, and Learning: Attention and Performance XXIII.
Oxford University Press, 5 2011. ISBN 9780191725623. https://doi.org/10.
1093/acprof:oso/9780199600434.003.0001.
33. Revels, J., Lubin, M. & Papamarkou, T. Forward-mode automatic
differentiation in julia. arXiv:1607.07892 [cs.MS], 2016.
34. Stringer, C. et al. Spontaneous behaviors drive multidimensionals, brainwide
activity. Science 364, 255 (2019).
35. Hanley, J. A. & McNeil, B. J. The meaning and use of the area under a receiver
operating characteristic (roc) curve. Radiology 143, 29–36 (1982).
36. Boyd-Meredith, J. T. & Piet, A. T. Stable choice coding in rat frontal orienting
fields across model-predicted changes of mind (analysis code v1.1), https://
doi.org/10.5281/zenodo.6325973 (2022).
Acknowledgements
We thank members of the Brody lab and Zachary Kilpatrick for useful conversations and
feedback. T.B. acknowledges support by NIH grant T32 MH 65214-16. A.E.H.
acknowledges support by NIH grant 1R21MH121889-01. E.J.D. is supported by an
HHMI Hanna H. Gray Fellowship and an HHMI-Helen Hay Whitney Postdoctoral
Fellowship. This work was supported by a grant from the Simons Foundation (Grant #
542953) awarded to C.B., as well as NIH grant R01MH108358 awarded to C.B.
Author contributions
A.P., and A.E.H. designed the study. A.E.H. managed rat training and care. A.E.H., and
E.J.D. recorded the neural data. A.P., and T.B. analyzed the neural data. A.P., A.E.H., and
T.B. wrote the manuscript. A.E.H., and C.B. oversaw all aspects of the project.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information The online version contains supplementary material
available at https://doi.org/10.1038/s41467-022-30736-3.
Correspondence and requests for materials should be addressed to Ahmed El Hady or
Carlos D. Brody.
Peer review information Nature Communications thanks the anonymous reviewer(s) for
their contribution to the peer review of this work. Peer reviewer reports are available.
Reprints and permission information is available at http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
material in this article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not included in the
article’s Creative Commons license and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.
© The Author(s) 2022
12
NATURE COMMUNICATIONS |
(2022) 13:3235 | https://doi.org/10.1038/s41467-022-30736-3 | www.nature.com/naturecommunications
| null |
10.1088_1751-8121_ad0b5c.pdf
|
Data availability statement
No new data were created or analysed in this study.
|
Data availability statement No new data were created or analysed in this study.
|
J. Phys. A: Math. Theor. 56 (2023) 495205 (37pp)
https://doi.org/10.1088/1751-8121/ad0b5c
Journal of Physics A: Mathematical and Theoretical
On Hamiltonian structures of
quasi-Painlevé equations
Galina Filipuk1 and Alexander Stokes2,∗
1 Institute of Mathematics, University of Warsaw, ul. Banacha 2, 02-097 Warsaw,
Poland
2 Graduate School of Mathematical Sciences, The University of Tokyo, 3-8-1
Komaba Meguro-ku, Tokyo 153–8914, Japan
E-mail: [email protected]
Received 5 July 2023; revised 6 November 2023
Accepted for publication 9 November 2023
Published 20 November 2023
Abstract
We describe the quasi-Painlevé property of a system of ordinary differen-
tial equations in terms of a global Hamiltonian structure on an analogue of
Okamoto’s space of initial conditions for the Painlevé equations. In the quasi-
Painlevé case, the Hamiltonian structure is with respect to a two-form which
is allowed to have certain zeroes on the surfaces forming the space of initial
conditions, as opposed to holomorphic symplectic forms in the case of the
Painlevé equations. We provide the spaces and Hamiltonian structures for sev-
eral known quasi-Painlevé equations and also for a new example, which we
prove to have the quasi-Painlevé property via the Hamiltonian structure and
construction of an appropriate auxiliary function which remains bounded on
solutions.
Keywords: Painlevé equations, space of initial conditions, rational surface,
quasi-Painlevé property, non-autonomous Hamiltonian system,
algebraic singularities
(Some figures may appear in colour only in the online journal)
1. Introduction
For each of the Painlevé differential equations PI–PVI, Okamoto [19] constructed an augmented
phase space on which an equivalent Hamiltonian system defines regular initial value problems
everywhere. Later, Takano and collaborators constructed special atlases for Okamoto’s spaces
for PII–PVI providing canonical coordinates for the symplectic structure, in which the extended
∗
Author to whom any correspondence should be addressed.
1751-8121/23/495205+37$33.00 © 2023 IOP Publishing Ltd Printed in the UK
1
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
system is of Hamiltonian form with polynomial Hamiltonian functions [17, 18, 23] (the case of
PI was done later independently by Iwasaki–Okada [10] and Chiba [1]). Further, it was shown
that on each space, the unique non-autonomous Hamiltonian system regular everywhere is that
extended from the Okamoto Hamiltonian form [20] of the relevant Painlevé equation, which
may be interpreted to mean that Okamoto’s spaces encode everything about the equations and
reduce their study completely to geometry—a central idea in the study of Painlevé equations
via rational surfaces, particularly in the discrete case [22].
To construct the space for the Painlevé equation PJ, Okamoto considered an equivalent
Hamiltonian system
dq
dt
=
∂HJ
∂p
,
dp
dt
= (cid:0) ∂HJ
∂q
,
(1.1)
where the Hamiltonian HJ = HJ(q, p, t), due to Okamoto [20], is polynomial in q, p and ana-
(cid:26) C of fixed singularities of PJ. The phase space for the
lytic in t away from the finite set FJ
system (1.1) can be taken initially to be the trivial bundle over the independent variable space
BJ = CnFJ with fibre C2
q,p, where here as well as for the remainder of the paper subscripts
indicate coordinates. Okamoto’s space is constructed by compactifying the fibres, performing
a sequence of blowups to resolve singularities of the system where infinitely many curves pass
through a single point, then finally removing from each fibre certain curves, called inaccess-
ible divisors or vertical leaves, which are vertical with respect to the foliation induced by the
flow of the system. The result is a complex analytic fibre bundle EJ over BJ, where the fibre EJ
t
is a complex rational surface with support of an effective anticanonical divisor removed. The
Painlevé property, namely that all solutions of the system are free of movable branch points,
means that the flow of the system defines a uniform foliation of EJ into disjoint solution curves
transverse to the fibres. Then each fibre is in bijection with the set of all solutions and EJ
t is
called a space of initial conditions for the equation. The system becomes regular everywhere on
EJ, but it is important to note that this regularisation alone is not sufficient to prove the Painlevé
property. To provide a proof via the differential system on EJ one requires, in addition, an aux-
iliary function with certain properties that allow one to show that the divisors removed from
each fibre are indeed inaccessible by solutions—similar auxiliary functions play a role in many
proofs of the Painlevé property for PI–PVI [6–9, 21, 24, 28]. The geometry of EJ can to some
extent provide clues as to the construction of such a function [30], but such a function is not
unique and its construction is far from canonical.
In the Takano framework [17, 18, 23], one requires additionally an atlas for EJ made up of
coordinate charts providing canonical coordinates for the symplectic form extended from dq ^
dp in the initial coordinates, so the system on EJ possesses a global Hamiltonian structure. This
is provided by a collection of Hamiltonian functions, one in each chart, related in a way that
ensures that the differential equation in each chart is of Hamiltonian form. It was shown [1, 10,
17, 18, 23] then that if a system of differential equations on EJ has a holomorphic Hamiltonian
structure (i.e. each Hamiltonian function is holomorphic on the associated coordinate patch)
then it must coincide with that defined by the Okamoto Hamiltonian HJ.
Recently, Kecker and Filipuk [14] showed that a construction similar to Okamoto’s can
be performed for second-order equations with all movable singularities reachable by analytic
continuation along finite length curves being at worst algebraic branch points, a condition
sometimes referred to as the quasi-Painlevé property, or the quasi-Painlevé property along
2
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
rectifiable curves (see also [4] for the case of globally finite branching about movable singu-
larities). The construction proceeds along the same lines as Okamoto’s with compactification
followed by blowups, but the system becomes regular only after transferring the role of inde-
pendent variable from t to a coordinate providing a local equation for one of the exceptional
divisors arising from the sequence of blowups. In this paper we show that this regularisability
can be seen in terms of properties of a global Hamiltonian structure for the systems just as in
the Takano framework in the Painlevé case.
2. Global Hamiltonian structures of Painlevé equations
We begin by recalling and illustrating the theory of global Hamiltonian structures of Painlevé
equations on Okamoto’s spaces as well as the uniqueness results of Takano and collaborat-
ors. We choose the example of PIV to illustrate this, but rather than starting from the usual
Okamoto Hamiltonian form of PIV to construct the space EIV we begin with a Hamiltonian
system obtained by Kecker [11], known to be equivalent to PIV [2, 13, 29], and show how an
appropriate atlas can still be constructed.
Definition 2.1. Consider a complex-analytic fibre bundle E ! B, where B (cid:18) C is some
domain. Suppose that this can be written as a gluing of coordinate patches
(
C2
xi,yi
(cid:2) B
)
,
(2.1)
∪
E =
i
glued by transition functions (xi, yi, t) 7! (xj(xi, yi, t), yj(xi, yi, t), t) which are birational maps
between (xi, yi) and (xj, yj) with coefficients locally analytic in t on B. We say, following [17,
18, 23] that the atlas is symplectic if, for all i, j,
dtxi
^ dtyi = dtxj
^ dtyj,
(2.2)
where dt is the exterior derivative on the fibre so t is treated as a constant.
Definition 2.2. On a bundle E with symplectic atlas as above, a non-autonomous Hamiltonian
system of differential equations on E is defined by a collection of Hamiltonians Hi(xi, yi, t)
such that
dxi
^ dyi + dHi
^ dt = dxj
^ dyj + dHj
^ dt,
(2.3)
where d is the exterior derivative on the total space so t is treated as a variable. We call the
collection of Hamiltonians, modulo functions of t, a Hamiltonian structure for a differential
system on E, or simply a Hamiltonian structure on E.
If a system of differential equations is given in one chart by
dxi
dt
=
∂Hi
∂yi
,
dyi
dt
= (cid:0) ∂Hi
∂xi
,
(2.4)
then under the transformation defined by the gluing it will be of the same Hamiltonian form
in any other chart (xj, yj) with Hamiltonian function Hj, i.e.
dxj
dt
=
∂Hj
∂yj
,
dyj
dt
= (cid:0) ∂Hj
∂xj
.
(2.5)
3
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
With an atlas as above, a Hamiltonian structure is determined uniquely from a Hamiltonian
function in a single chart due to the following standard fact.
Lemma 2.1. If a transformation C3 3 (xi, yi, t) 7! (xj(xi, yi, t), yj(xi, yi, t), t) 2 C3 satisfies
Fi (xi, yi) dtxi
^ dtyi = Fj (xj, yj) dtxj
^ dtyj,
(2.6)
for rational functions Fi, Fj whose coefficients are independent of t, then given Hi(xi, yi, t)
rational in xi, yi and locally analytic in t, there exists Hj(xj, yj, t), unique modulo functions of
only t, such that
Fi (xi, yi) dxi
^ dyi + dHi
^ dt = Fj (xj, yj) dxj
^ dyj + dHj
^ dt.
(2.7)
Further, the system of differential equations
Fi (xi, yi)
dxi
dt
=
∂Hi
∂yi
, Fi (xi, yi)
dyi
dt
= (cid:0) ∂Hi
∂xi
,
is transformed to
Fj (xj, yj)
dxj
dt
=
∂Hj
∂yj
, Fj (xj, yj)
dyj
dt
= (cid:0) ∂Hj
∂xj
.
(2.8)
(2.9)
The collection of Hamiltonians does not define a single function on E, but rather a glob-
ally defined rational two-form according to (2.7), which dictates the corrections between
Hamiltonian functions on the overlaps of coordinate patches arising from t-dependence in the
(cid:0) Hj, the equality (2.7) gives a pair of linear partial
gluing. Letting the correction be Xi,j = Hi
differential equations
(
∂Xi,j
∂xi
= Fj
∂xj
∂xi
∂yj
∂t
(cid:0) ∂yj
∂xi
∂xj
∂t
)
,
(
∂Xi,j
∂yi
= Fj
∂xj
∂yi
∂yj
∂t
(cid:0) ∂yj
∂yi
∂xj
∂t
)
,
(2.10)
which are guaranteed to be compatible by the condition (2.6) (this is the content of lemma 2.1).
It is important to note that if either coefficient Fi or Fj has nontrivial t-dependence, then this
does not hold in general; this is the reason that various local coordinate changes are required
in the construction of the atlas for the space E in both Painlevé and quasi-Painlevé cases in
order for them to possess a global Hamiltonian structure, as we will see below.
Remark 2.1. In order to avoid confusion regarding the condition (2.6) forming the assump-
tion of lemma 2.1, we remark that it should be interpreted as follows. Given (xi, yi, t) 7!
(xj(xi, yi, t), yj(xi, yi, t), t) such that the map (xi, yi) 7! (xj(xi, yi, t), yj(xi, yi, t)) is birational for
^ dtyi is pushed forward under this map to one that
any t, the rational two-form Fi(xi, yi) dtxi
^ dtyj is independent
does not depend on t, i.e. the rational function Fj(xj, yj) in Fj(xj, yj) dtxj
of t. Further, in the partial differential equations (2.10) Fi, Fj and Xi,j are considered purely as
functions on the relevant copies of C3, with no reference to the flow of any differential system.
For a bundle E =
i
a two-form on the fibre Et given in each chart by
∪
(
)
C2
xi,yi
(cid:2) B
glued by transition functions as in definition 2.1 but with
ωt = Fi (xi, yi) dtxi
^ dtyi,
(2.11)
4
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
if we have a collection of Hamiltonians fHi
chart by
g defining a two-form Ω on E given in each
Ω = Fi (xi, yi) dxi
^ dyi + dHi
^ dt,
(2.12)
then we say the Hamiltonian structure is with respect to ωt.
Remark 2.2. While a single Hamiltonian structure does not define a function on E, the differ-
ence of any two Hamiltonian structures does, via an appropriate choice of a function of t. If
g are two Hamiltonian structures on E with respect to ωt = Fi(xi, yi)dtxi
fHi
^ dtyi, so
g, fKi
^ dyj + dHj
^ dt and similarly for Ki, Kj, then we
^ dyi + dHi
Fi(xi, yi)dxi
(cid:24)
(cid:0) Kj, where here and from this point
(cid:0) Ki
(cid:0) Ki) ^ dt = d(Hj
have d(Hi
= Hj
(cid:24)
= means equal modulo addition of functions of only t. Thus if the given Hamiltonians
on
are such that on the overlap of coordinate charts (C2
6=
Hj
(cid:0) Kj, then we can add a function fi,j(t) so that they coincide under the gluing:
^ dt = Fj(xj, yj)dxj
(cid:2) B) we have Hi
(cid:0) Kj) ^ dt, so Hi
(cid:2) B) \ (C2
(cid:0) Ki
xj,yj
xi,yi
Hi
(cid:0) Ki = Hj
(cid:0) Kj + fi,j (t)
and globally give a function G on E such that on each C2
Hi
(cid:0) Ki.
xi,yi
(2.13)
(cid:2) B (cid:26) E, we have G(xi, yi, t) (cid:24)
=
2.1. Kecker’s cubic Hamiltonian for Painlevé-IV
Kecker [12] considered a family of Hamiltonians
H (q, p, t) = qM+1 + pN+1 +
∑
(i,j)∈I
αi,j (t) qip j,
(2.14)
where I (cid:26) [0, M] (cid:2) [0, N] (cid:26) Z2 is an appropriate index set defined such that the terms in qM, pN
turn out to be dominant on the right-hand sides of the corresponding Hamiltonian system
dq
dt
=
∂H
∂p
,
dp
dt
= (cid:0) ∂H
∂q
.
(2.15)
In [12], Kecker showed that with certain differential relations between the coefficients αi,j(t),
all movable singularities of the system (2.15) reachable by analytic continuation along finite
length curves are at worst algebraic branch points, about which solutions are representable
by Puiseux series expansions. These relations are called resonance conditions due to their
relations to recursive relations for coefficients of series solutions of the system.
We consider the (M, N) = (2, 2) case of the Kecker Hamiltonian (2.14) (after a scaling for
convenience), which we write explicitly as
H =
q3 +
1
3
1
3
p3 + γ (t) qp + α (t) q + β (t) p,
(2.16)
where α, β, γ are functions of t analytic in some domain (they will turn out to be entire). With
the resonance conditions
α ′ (t) = β ′ (t) = γ ′ ′ (t) = 0,
(2.17)
5
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
so α and β are constants and γ(t) = γ1t + γ0, for constants γ1, γ0, the corresponding
Hamiltonian system
dq
dt
= p2 + (γ1t + γ0) q + α,
dp
dt
= (cid:0)q2 (cid:0) (γ1t + γ0) p (cid:0) β,
(2.18)
6= 0, is known [29] to be
in fact has only poles as movable singularities and, as long as γ1
related to PIV, and also directly by a birational transformation [2] to the Okamoto Hamiltonian
system.
The space of initial conditions for system (2.18) was con-
2.1.1. Space of initial conditions.
structed in [13] using P2 = P2(C) as the initial compactification, and in [2] using P1 (cid:2) P1 =
P1(C) (cid:2) P1(C). It is known [2] that an isomorphism can be obtained between the surfaces
forming the space of initial conditions for the system (2.18) and those from the Okamoto form
of PIV, so the atlas constructed in [17] can be pulled back under this isomorphism to give the
system (2.18) a global Hamiltonian structure. In what follows we show how to obtain this dir-
ectly from the system (2.18), without reference to the Okamoto Hamiltonian. We use P1 (cid:2) P1
as our compactification of the fibres of the phase space, and after a sequence of blowups with
appropriate local coordinate changes we construct an atlas for the resulting bundle in terms of
which we can obtain the global Hamiltonian structure of system (2.18).
Let B = C
t be the independent variable space for system (2.18) (on which the coefficients
q,p to
are analytic) and from the trivial bundle C2
q,p
P1 (cid:2) P1 via the usual introduction of coordinates to cover P1 (cid:2) P1 by the four charts
(cid:2) B over B, compactify the fibres from C2
P1 (cid:2) P1 = C2
q,p
[ C2
Q,p
[ C2
q,P
[ C2
Q,P,
(2.19)
with gluing defined by Q = 1/q, P = 1/p. Extending the system (2.18) to (P1 (cid:2) P1) (cid:2) B is
done using the gluing as a change of variables, and we find a sequence of ten blowups of the
fibre over t must be performed to resolve indeterminacies of the rational system of differential
equations. In introducing coordinate charts to cover the exceptional divisor arising from each
blowup, we use the following convention: after blowing up a point pi given in some affine
chart (x, y) by
pi : (x, y) = (x∗, y∗) ,
(2.20)
the exceptional divisor Li
and (Ui, Vi) given by
(cid:24)
= P1 replacing pi is covered by two affine coordinate charts (ui, vi)
x = ui vi + x∗,
x (cid:0) x∗
y (cid:0) y∗
ui =
,
y = vi + y∗,
vi = y (cid:0) y∗,
and
x = Vi + x∗,
y (cid:0) y∗
x (cid:0) x∗
Ui =
y = Ui Vi + y∗,
, Vi = x (cid:0) x∗.
(2.21)
In particular the exceptional divisor Li has in these charts local equation vi = 0, respectively
Vi = 0.
6
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
Figure 1. Blowup points for the Kecker cubic Hamiltonian system (2.18).
Figure 2. Surface for the Kecker cubic Hamiltonian system (2.18).
For the system (2.18), we perform blowups of the P1 (cid:2) P1 fibre at points p1, . . . , p10 given
in coordinates in figure 1 below, introduced according to the convention above, where we have
introduced for convenience
p
ζ = e
π i
3 =
1 + i
2
3
,
(2.22)
with here and for the remainder of the paper i being the imaginary unit. We give a schematic
description of the configuration of points and the surface in figure 2.
Letting the surface obtained by blowing up the points as above be X
t, the fact that the
locations of points depend analytically on t means we have a complex analytic fibre bundle
E, with compact fibre over t being Et = X
t. We now remove from each fibre the support of
the inaccessible divisor, which here is the pole divisor of the rational two-form ωt = dtq ^ dtp
extended to the surface X
t, where as in the previous section we use dt to indicate the exterior
derivative on the fibre for a fixed t, rather than on the total space. In the coordinates introduced
above this two-form is given by
7
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
ωt = dtq ^ dtp = (cid:0) dtq ^ dtP
P2
= (cid:0) dtQ ^ dtp
Q2
=
dtQ ^ dtP
Q2P2
=
^ dtv1
dtu1
1v3
u2
1
=
^ dtv2
dtu2
2 ((cid:0)1 + u2v2)2
v2
=
(
dtu3
^ dtv3
(cid:0)1 (cid:0) (γ1t + γ0) v3 + u3v2
3
)
2
v3
=
(
(cid:0)1 (cid:0) (γ1t + γ0) v4 + v2
4
^ dtv4
dtu4
(
α (cid:0) β (cid:0) (γ1t + γ0)2 (cid:0) γ1 + u4v4
))
2
=
^ dtv5
dtu5
5 (ζ + u2v2)2
v2
=
(
dtu6
^ dtv6
ζ (cid:0) (γ1t + γ0) v6 + u6v2
6
)
2
v6
=
(
ζ (cid:0) (γ1t + γ0) v7 + v2
7
(
^ dtv7
dtu7
(cid:0)ζα (cid:0) β + ζ −1
(
)
(γ1t + γ0)2 + γ1
+ u4v4
))
2
=
^ dtv8
dtu8
8 (ζ −1 + u8v8)2
v2
=
(
dtu9
^ dtv9
ζ −1 (cid:0) (γ1t + γ0) v9 + u9v2
9
)
2
v9
=
(
ζ −1 (cid:0) (γ1t + γ0) v10 + v2
10
(
dtu10
^ dtv10
(cid:0)ζ −1α (cid:0) β + ζ
(
)
(γ1t + γ0)2 + γ1
+ u10v10
))
.
2
(2.23)
From the expressions in coordinates as above, we see that the pole divisor of this two-form is
given by
(cid:0) div ωt = 2I1 + 3I1 + 2I3 + 2I4 + 2I5 + 2I6 + I7 + I8 + I9,
(2.24)
where the irreducible components Ii (the inaccessible divisors indicated in blue on figure 2)
are given explicitly by
(cid:0) L8,
(cid:0) L5
(cid:0) L2
I1 = L1
I2 = fq = 1g (cid:0) L1,
I3 = fp = 1g (cid:0) L1,
I4 = L2
I5 = L5
I6 = L8
(cid:0) L3,
(cid:0) L6,
(cid:0) L9,
I7 = L3
I8 = L6
I9 = L9
(cid:0) L4,
(cid:0) L7,
(cid:0) L10.
(2.25)
Here Li is the exceptional divisor from the blowup of pi (or more precisely its pullback/total
transform under any further blowups), fq = 1g and fp = 1g are total transforms of lines
as indicated. We will also recycle this notation for surfaces obtained from other systems in
sections 3–5.
Remark 2.3. It is possible to perform a kind of minimisation by contracting the two excep-
tional curves of the first kind I2 = fq = 1g (cid:0) L1 and I3 = fp = 1g (cid:0) L1 to arrive at a surface
(associated with PIV) in the Sakai classification [22], but these are contained
of the type E
in the support of the anticanonical divisor which we will remove so this minimisation does not
affect the space E and is not important at this stage.
(1)
6
t
n [
We remove the inaccessible curves Ii from each fibre of E to arrive at the space E, with
fibre being Et = X
i Ii, on which the system of differential equations extended from (2.18)
is regular, and ωt provides a holomorphic symplectic form. The only exceptional divisors from
the sequence of blowups which are not contained in the collection of curves that were removed
are L4, L7, and L10, i.e. the final exceptional lines arising from the blowups of the three cascades
of infinitely near points. The coordinates (u4, v4), (u7, v7), (u10, v10) can then be used to cover
8
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
the part of E not visible in the original (q, p) chart, so we have an atlas for the bundle. However,
noting that the symplectic form ωt in these coordinates as in (2.23) has t-dependence in the
coefficients when written in these charts, lemma 2.1 does not apply and indeed the system in
these charts cannot be written in Hamiltonian form with respect to ωt.
In order to obtain appropriate
2.1.2. Symplectic atlas and global Hamiltonian structure.
coordinate charts to cover the parts of the last exceptional divisors contained in E, we note that
in the equations (2.23) we see that in the rational coefficient of ωt in the charts (u2, v2), (u5, v5)
and (u8, v8), the denominator has ceased to be a monomial in the coordinate v1 providing the
local equation for the exceptional line L1, i.e. the coordinates are such that other components
of the pole divisor are visible in these charts. To remedy this, after blowing up p1 we make the
(birational) local coordinate change (u1, v1) 7! (˜u1,˜v1) defined by
u1 =
1
˜u1
,
v1 = ˜v1,
Q =
˜v1
˜u1
, P = ˜v1,
(2.26)
so the exceptional line L1 has local equation ˜v1 = 0, and in these coordinates we have
p2 : (˜u1,˜v1) = ((cid:0)1, 0) ,
p5 : (˜u1,˜v1) =
(
)
,
ζ −1, 0
p8 : (˜u1,˜v1) = (ζ, 0) .
(2.27)
After this coordinate change we proceed with the rest of the blowups, introducing coordinates
according to the same convention as previously but with tildes to distinguish them from those
in (1). We then have expressions for ωt in these new coordinates as follows.
dt˜v1
ωt =
^ dt˜u1
˜v3
1
=
=
=
dt˜v2
dt˜v5
dt˜v8
^ dt˜u2
˜v2
2
^ dt˜u5
˜v2
5
^ dt˜u8
˜v2
8
=
=
=
dt˜v3
dt˜v6
dt˜v9
^ dt˜u3
˜v3
^ dt˜u6
˜v6
^ dt˜u9
˜v9
= dt˜v4
^ dt˜u4
= dt˜v7
^ dt˜u7
= dt˜v10
^ dt˜u10.
(2.28)
Thus we can introduce (x1, y1) = ((cid:0)˜u4,˜v4), (x2, y2) = ((cid:0)˜u7,˜v7), and (x3, y3) = ((cid:0)˜u10,˜v10), to
obtain a symplectic atlas, in terms of which we have a Hamiltonian structure for the sys-
tem (2.18) on E in which all Hamiltonians are polynomial in coordinates and we have the
following.
Proposition 2.1. The space E constructed from the system (2.18) from Kecker’s cubic
Hamiltonian can be described as a gluing of coordinate patches
E = C3
q,p,t
[ C3
x1,y1,t
[ C3
x2,y2,t
[ C3
x3,y3,t,
with gluing defined by
(cid:0)1 + y1 (γ1t + γ0 + y1 ((cid:0)α + β + γ1
(cid:0) x1y1))
,
q =
q =
q =
(
ζ + y2
ζ −1 + y3
y1
(cid:0)ζ −1 (γ1t + γ0) + y2
y2
(
(cid:0)ζ (γ1t + γ0) + y3
y3
(
(
ζα (cid:0) ζ −1β + γ1
(cid:0) x2y2
ζ −1α (cid:0) ζβ + γ1
(cid:0) x3y3
))
))
,
,
9
(2.29)
(2.30)
p =
p =
p =
1
y1
1
y2
1
y3
,
,
,
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
which forms a symplectic atlas for the two-form
ωt = dtq ^ dtp = dtx1
^ dty1 = dtx2
^ dty2 = dtx3
^ dty3.
(2.31)
The Hamiltonian structure on E for the system extended from (2.18) is then given by
Hamiltonians fH(q, p, t), H1(x1, y1, t), H2(x2, y2, t), H3(x3, y3, t)g, which are related modulo
functions of t under the gluing according to
(cid:24)
= H1 + γ1p
(cid:24)
= H2
H
(cid:0) ζ −1γ1p
(cid:24)
= H3
(cid:0) ζγ1p.
(2.32)
Moreover, for each i = 1, 2, 3 the Hamiltonian Hi(xi, yi, t) is polynomial in xi, yi, so the differ-
ential system is regular everywhere on E.
The Hamiltonian Hi in coordinates xi, yi is computed by direct substitution of the gluing
in (2.30) into the expression from (2.32) for Hi in terms of H and corrections. For example H1
is given by
H1 (x1, y1, t) (cid:24)
= (cid:0) 1
3
(
(α (cid:0) β (cid:0) γ1) y5
1
(cid:0) x2
1
x3
1y6
1
(α (cid:0) β (cid:0) γ1)2 y4
1
(
+ x1
(cid:0) (γ1t + γ0) y4
1 + y3
1
)
(cid:0) 2 (γ1t + γ0) (α (cid:0) β (cid:0) γ1) y3
1
)
+
2α (cid:0) β + (γ1t + γ0)2 (cid:0) 2γ1
y2
1
(cid:0) (γ1t + γ0) y1 + 1
(2.33)
(cid:0) (α (cid:0) β (cid:0) γ1)3
1 + (γ1t + γ0) (α (cid:0) β (cid:0) γ1)2 y2
y3
3
(
)
1
+ (α (cid:0) β (cid:0) γ1)
α + (γ1t + γ0)2 (cid:0) γ1
y1.
Remark 2.4. If the resonance conditions (2.17) are not imposed, then the system fails to
become regular even after indeterminacies are resolved. This can also be seen in terms of the
Hamiltonians failing to be polynomial, since the solution of the systems of partial differential
equations (2.10) for the corrections will contain logarithmic terms.
2.2. Uniqueness result for Painlevé-IV
In proving that the unique Hamiltonian structure holomorphic on the Okamoto space EIV and
extending meromorphically to the bundle of compact surfaces is the Okamoto Hamiltonian
form of PIV, Matumiya [18] used the atlas constructed by Takano et al in [17]. The Okamoto
Hamiltonian form of PIV is given by
df
dt
dg
dt
=
∂HIV
∂g
= (cid:0) ∂HIV
∂f
= 4f g (cid:0) f 2 (cid:0) 2tf (cid:0) 2a1,
= (cid:0)2g2 + 2f g + 2tg + a2,
(
HIV = 2f g2 (cid:0)
f 2 + 2tf + 2a1
)
g (cid:0) a2 f, (2.34)
where a1, a2 are free complex parameters corresponding to the root variables in the Sakai the-
ory of rational surfaces associated with Painlevé equations [22], which are related to the para-
meters κ0, κ∞ from the original isomonodromy problem used by Okamoto [20] to construct
this Hamiltonian system by κ0 = a1, κ∞ = (cid:0)a2.
10
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
Lemma 2.2 (Matano et al [17]). The space EIV constructed by Okamoto from the system (2.34)
can be represented as a gluing of coordinate patches
EIV = C3
f,g,t
[ C3
X1,Y1,t
[ C3
X2,Y2,t
[ C3
X3,Y3,t
with gluing defined by
f =
1
X1
,
g = X1 ((cid:0)a2
(cid:0) X1Y1) ,
f = Y2 (a1
f =
1
X3
,
g =
,
g =
(cid:0) X2Y2) ,
1
Y2
1/2 + tX3 + (a1 + a2
X3
(cid:0) 1) X2
3
(cid:0) X3
3Y3
,
(2.35)
(2.36)
which provides a symplectic atlas for the two-form
dtf ^ dtg = dtX1
^ dtY1 = dtX2
^ dtY2 = dtX3
^ dtY3.
(2.37)
Remark 2.5. We have changed notation slightly from that in [17], with our charts being related
to x[(cid:3)(cid:3)], y[(cid:3)(cid:3)] for (cid:3)(cid:3) 2 f00, 10, 01, 11g appearing there by
(f, g) = (x [0, 0] , y [00]) ,
(X2, Y2) = (x [01] , y [01]) ,
(X1, Y1) = (x [10] , y [10]) ,
(X3, Y3) = (x [11] , y [11]) .
(2.38)
The above atlas was obtained after constructing the space EIV using a Hirzebruch surface as
initial compactification, rather than P1 (cid:2) P1.
The coordinate charts in lemma 2.2 also provide an atlas for the space E constructed from
the Kecker system, as a result of the isomorphism between the surfaces Et and E
, which was
obtained in [2]. In using this it is convenient to make a change of coordinate t 7! at + b for the
independent variable space B = C to set γ0 = 0 γ1 = 2i√
3
, after which the transformation
IV
t
(
q =
(cid:0)1 +
ip
3
)
(
p
1 (cid:0) i
t (cid:0) f +
)
(
3
g,
p =
1 +
)
ip
3
t + f (cid:0)
(
)
p
1 + i
3
g,
(2.39)
relates the systems (2.18) and (2.34) with the independent variable t unchanged and parameters
related by
α = (cid:0)1 +
ip
3
(
+ 2a1 +
1 (cid:0) i
p
)
3
a2, β = (cid:0)1 (cid:0) ip
(
)
p
+ 2a1 +
1 + i
3
a2.
(2.40)
3
This transformation provides an identification E = EIV, so the following result also applies to
the Kecker Hamiltonian system on E.
Theorem 2.1 (Matumiya [18]). On the space E = EIV equipped with the symplectic atlas in
lemma 2.2, if a Hamiltonian structure of a differential system is holomorphic on E and extends
meromorphically to E, then it must coincide with that defined by the Okamoto Hamiltonian
form (2.34) of PIV.
11
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
∑
∑
M
i =0
The idea of the method of proof of this result is to first use the assumption that the
Hamiltonian structure extends meromorphically to E to argue that this must be provided by
N
j =0 αi,j(t)f i g j polynomial in the original coordinates f, g
a Hamiltonian H0(f, g, t) =
with coefficients analytic in B. Then one derives linear equations for the coefficients αi,j(t)
which must hold in order for the other Hamiltonians H1, H2 determined from H0 by the gluing
according to lemma 2.1 to be holomorphic in (X1, Y1) and (X2, Y2) respectively. From these,
one can reduce the Hamiltonian H0 to low degrees in f, g, in this case M = N = 2, then finally
solve the system of linear equations for the coefficients to show that H0 must coincide up to
functions of t with the Okamoto Hamiltonian HIV as it appears in (2.34).
3. An equation of Painlevé-II type
In [3], it was shown that each equation in the family
d2y
dt2
=
N∑
n=0
an (t) yn,
(3.1)
where N 2 N, N ⩾ 2, and an(t) are analytic in some domain, has the quasi-Painlevé property
assuming appropriate resonance conditions on the coefficients an, which generalises earlier
results of Shimomura [25, 26]. Through the transformation given in [3], the equation (3.1) can
be taken without loss of generality with aN(t) = 2(N + 1)/(N (cid:0) 1)2 and aN−1(t) = 0. With the
resonance conditions, the resulting equation admits Puiseux series solutions of the form
y (t) =
∞∑
n=−2
Cn (t (cid:0) t∗)n/4 ,
(3.2)
which are convergent in a cut neighbourhood of t∗ in the domain where the coefficients ai are
analytic. The leading coefficient C−2 satisfies C4
−2 = 1, so at first glance there are four possible
leading behaviours of solutions of (3.1). However it was shown in [3] that when N is odd, by
absorbing the choice of C−2 as much as possible into the choice of branch of (t (cid:0) t∗)1/4, there
are essentially only two.
We will consider the N = 5 case, which is written explicitly as
d2y
dt2
=
3
4
y5 + a3 (t) y3 + a2 (t) y2 + a1 (t) y + a0 (t) ,
with the resonance conditions
(
′ ′
3
a
(cid:17) 0,
) ′ (cid:17) 0,
(cid:0) a2
3
4a1
(3.3)
(3.4)
′
is differentiation by t. We call this equation of PII type for two reasons: firstly, the fam-
where
ily (3.1) with odd N to which equation (3.3) belongs contains a family studied by Shimomura
[27] generalising PII, and, secondly, the number of essentially different leading behaviours
of solutions about movable singularities is two, which the same as the case of PII where all
movable singularities are simple poles of residue (cid:6)1. With the resonance conditions (3.4), the
Puiseux series solutions of equation (3.3) are of the form
y (t) =
∞∑
n=−1
Cn (t (cid:0) t∗)n/2 ,
12
(3.5)
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
and the resonance conditions ensure that one of the coefficients (in this case C5) is arbitrary, so
together with the location t∗ of the movable singularity this constitutes a two-parameter family
of solutions for each leading behaviour.
The proof that the only movable singularities of equation (3.3) are algebraic branch points
given by the Puiseux series (3.5) in [3] requires the following:
(1) Variables r(t), s(t) given by rational functions of y(t), y
′(t) such that after interchanging the
role of independent variable from t to s the system defines regular initial value problems
at (r, s, t) = (r∗, 0, t∗)
dr
ds
= F (r, s, t) ,
dt
ds
= G (r, s, t) ,
(3.6)
where F, G are analytic in some neighbourhood of (r, s, t) = (r∗, 0, t∗).
(2) A proof that analytic solutions r(s), t(s) of these regular initial value problems correspond
to algebraic branch points in the original variable y(t).
(3) An auxiliary function W of y, y
′, t which remains bounded on any finite length curve γ on
which y is analytic except at the endpoint where it is singular. This is sometimes interpreted
as an approximate first integral or Lyapunov function for the equation, and allows one
to prove that the only movable singularities of a solution y are algebraic branch points
corresponding to solutions of the initial value problems above.
In [14], a similar method of proof was recast in terms of a bundle of rational surfaces E
iIi which
constructed through compactification and blowups, and a collection of curves I = [
are removed from each fibre to obtain a space E such that the following can be established:
(1) The differential system extended to E defines initial value problems everywhere that are
either regular or regularised by transferring the role of independent variable from t to some
coordinate v providing a local equation for an exceptional divisor. That is, in some coordin-
ate chart (u, v) covering an exceptional divisor L which is not removed as part of I, the
system becomes
du
dv
= F (u, v, t) ,
dt
dv
= G (u, v, t) ,
(3.7)
where v = 0 is the local equation of L, and this defines regular initial value problems at
every point (u, v, t) = (h, 0, t∗) in the fibre over t∗.
(2) Inversion of the power series solutions t(v), u(v) of the regular initial value problems above
gives u(t), v(t) as Puiseux series expansions about a movable singularity t∗. Then the bira-
′) is such that
tional relation between the coordinates (u, v) and the original variables (y, y
the analytic solutions of the regular initial value problem above correspond to algebraic
branch points.
(3) The divisors Ii are inaccessible to the flow of the system so the only movable singularities
are those corresponding to solutions of initial value problems on exceptional divisors as
above. This is done by constructing an auxiliary function W on the bundle of compact
surfaces E which restricts to a rational function on each fibre with coefficients analytic in
t, such that W has poles along all components Ii but is holomorphic on exceptional divisors
not contained in I, and remains bounded on the lift to E of any finite length curve γ ending
at a movable singularity.
13
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
A space E for (3.3) was constructed in [14] using P2 as initial compactification, but here
we use P1 (cid:2) P1 and construct in addition an atlas such that the system on E has a global
Hamiltonian structure with all Hamiltonian functions polynomial in coordinates. We also show
how this Hamiltonian structure can assist in constructing the auxiliary function necessary to
prove the quasi-Painlevé property for the equation (3.3) according to the scheme above.
3.1. Surfaces
Letting y = q, y
′ = p, we consider the equation (3.3) in the form of the Hamiltonian system
dq
dt
=
H =
,
∂H
= (cid:0) ∂H
dp
∂p
∂q
dt
q6 (cid:0) a3 (t)
p2 (cid:0) 1
1
4
8
2
,
q4 (cid:0) a2 (t)
3
q3 (cid:0) a1 (t)
2
q2 + a0 (t) q.
(3.8)
Let the domain on which the coefficients ai(t) are analytic be B, and compactify the fibres of
the initial phase space to P1 (cid:2) P1 by introducing coordinates Q = 1/q, P = 1/p as in section 2,
so we have the system extended to (P1 (cid:2) P1) (cid:2) B. After a sequence of 15 blowups, of points in
the fibre over t given in coordinates in figure 3, all indeterminacies of the system are resolved
and the fibres of the phase space are surfaces X
t as shown in figure 4.
With charts introduced according to the convention established in section 2, without local
coordinate changes at this stage, the system in the charts (u9, v9), (u15, v15) covering the final
exceptional divisors from the sequence of blowups can be computed as
du9
dt
=
b1 (a
′ ′
′
(cid:0) 2a
3 + a3a
3
v2
9P (u9, v9, t)
′
1) + v9F9 (u9, v9, t)
,
dv9
dt
=
c1
v9P (u9, v9, t)
,
(3.9)
where b1, c1 are known nonzero constants, F9(u9, v9, t) is polynomial in u9, v9 with coefficients
analytic in t on B satisfying F9(u9, 0, t) 6= 0, and
P9 (u9, v9, t) = 6 (cid:0) 6a3 (t) v2
9
(cid:0) 8a2v3
+ 8 ((cid:0)3a0 (t) + 3a2 (t) a3 (t) + 2a
(cid:0)12a1 (t) + 9a3 (t)2 + 6a
′
2 (t)) v5
9 + 3u9v6
9,
9 +
)
′
3 (t)
v4
9
(3.10)
and similarly
du15
dt
=
b2 (a
′ ′
3
′
(cid:0) a3a
3 + 2a
v2
15P15 (u15, v15, t)
′
1) + v15F15 (u15, v15, t)
,
dv15
dt
=
c2
v15P15 (u15, v15, t)
,
(3.11)
where again b2, c2 are known nonzero constants, F15(u15, v15, t) is polynomial in u15, v15 with
coefficients analytic in t on B satisfying F15(u15, 0, t) 6= 0, and
(
(
P15 (u15, v15, t) = 6 (cid:0) 6a3 (t) v2
15
(cid:0) 8a2 (t) v3
15 +
(cid:0)12a1 (t) + 9a3 (t)2 (cid:0) 6a
+ 8 ((cid:0)3a0 (t) + 3a2 (t) a3 (t) (cid:0) 2a
′
2 (t)) v5
15
(cid:0) 3u15v6
15.
)
′
3 (t)
v4
15
(3.12)
It is at this stage, as in [14], that we see that for the systems to be regularisable by the inter-
changing of independent variable we need to impose conditions on the coefficients such that a
power of v9 (respectively v15) cancels in the rational function giving du9
in (3.9) (respectively
dt
du15
dt
in (3.11)). This leads to the resonance conditions
(
′ ′
3
a
(cid:17) 0,
4a1
(cid:0) a2
3
) ′ (cid:17) 0,
14
(3.13)
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
Figure 3. Blowup points for the quasi-Painlevé-II system (3.8).
which we solve to set
a3 (t) = λ1t + λ2,
a1 (t) =
λ1t (λ1t + 2λ2)
4
+ λ3,
(3.14)
for constants λ1, λ2, and λ3. After imposing these conditions by substituting (3.14) in sys-
tem (3.8), we perform the blowup calculations again to resolve the indeterminacies of the
system, and find that in the final charts it takes the form
du9
dt
=
F9 (u9, v9, t)
v9P9 (u9, v9, t)
,
dv9
dt
=
c1
v9P9 (u9, v9, t)
,
and
du15
dt
=
F15 (u15, v15, t)
v15P15 (u15, v15, t)
,
dv15
dt
=
c2
v15P15 (u15, v15, t)
.
15
(3.15)
(3.16)
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
Figure 4. Surface for the quasi-Painlevé-II system (3.8).
After transferring the role of independent variable in each of these from t to the coordinate v9,
respectively v15, providing the local equation of the final exceptional divisor, we have
du9
dv9
=
F9 (u9, v9, t)
c1
,
dt
dv9
=
v9P9 (u9, v9, t)
c1
,
and
du15
dv15
=
F15 (u15, v15, t)
c2
,
dt
dv15
=
v15P15 (u15, v15, t)
c2
.
(3.17)
(3.18)
Each of these systems defines regular initial value problems everywhere on the part of the final
exceptional divisor away from the proper transform of the second to last one, and analytic
solutions to these can be shown to correspond to Puiseux series solutions (3.5) in the original
variables as in [14]. However, rather than performing this step in the proof of the quasi-Painlevé
property of the system (3.8) at this point, we will first establish an atlas in which the system
possesses a global Hamiltonian structure. Removing the divisors
(cid:0) L2
I1 = fq = 1g (cid:0) L1,
I2 = fp = 1g (cid:0) L1
(cid:0) L2,
I3 = L1
(cid:0) L3,
I4 = L2
(cid:0) L4
I5 = L3
(cid:0) L10,
(cid:0) L3,
I6 = L4
I7 = L5
I8 = L6
I9 = L7
I10 = L8
(cid:0) L5,
(cid:0) L6,
(cid:0) L7,
(cid:0) L8,
(cid:0) L9,
I11 = L10
I12 = L11
I13 = L12
I14 = L13
I15 = L14
(cid:0) L11,
(cid:0) L12,
(cid:0) L13,
(cid:0) L14,
(cid:0) L15,
(3.19)
indicated in blue on figure 4, which we will later prove are inaccessible to the flow of the sys-
tem, we have a bundle E with fibre Et = X
i Ii, on which the system (3.8) with the resonance
conditions (3.4) defines initial value problems which are regular or regularisable by changes
of variables as above.
n [
t
16
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
3.2. Symplectic atlas and Hamiltonian structure
Similarly to the case of Kecker’s form of PIV with cubic Hamiltonian in section 2, we see
the need for a local coordinate change after the blowup of p3 to obtain an atlas in which the
symplectic form extended from dtq ^ dtp allows for a Hamiltonian structure of the system (3.8)
in which all Hamiltonians are polynomial. This is a local coordinate change (u3, v3) 7! (˜u3,˜v3)
defined by
u3 =
1
˜u3
,
v3 = ˜v3,
Q = ˜v3, P =
˜v3
3
˜u3
,
(3.20)
so the exceptional line L3 has local equation ˜v3 = 0, and in these coordinates we have
p4 : (˜u3,˜v3) = (1/2, 0) ,
p10 : (˜u3,˜v3) = ((cid:0)1/2, 0) .
(3.21)
Proceeding after this coordinate change with the rest of the blowups with tilded coordinates
according to the same convention as previously, we then have expressions for ωt in these new
coordinates as follows:
ωt =
dt˜u3
^ dt˜v3
˜v5
3
=
=
=
=
dt˜u4
dt˜u7
^ dt˜v4
˜v4
4
^ dt˜v7
˜v7
^ dt˜v10
dt˜u10
˜v4
10
^ dt˜v13
˜v13
dt˜u13
dt˜u5
=
= dt˜u8
=
dt˜u6
^ dt˜v6
^ dt˜v5
˜v3
˜v2
6
5
^ dt˜v9
^ dt˜v8 = ˜v9 dt˜u9
^ dt˜v11
^ dt˜v12
˜v2
˜v3
12
11
^ dt˜v14 = ˜v15 dt˜u15
dt˜u12
=
dt˜u11
=
= dt˜u14
^ dt˜v15.
(3.22)
In particular the divisor of ωt on the fibre Et is given in terms of the inaccessible divisors Ii
in (3.19) and the final exceptional divisors L9 and L15 by
(cid:0) div ωt = 2I1 + 2I2 + 3I3 + 4I4 + 5I5 + 4I6 + 3I7 + 2I8 + I9
(cid:0) L9 + 4I11 + 3I12
+ 2I13 + I14
(cid:0) L14.
(3.23)
Letting (x1, y1) = (˜u9,˜v9) and (x2, y2) = (˜u15,˜v15) we have an atlas in which the system takes
^ dtyi with polynomial Hamiltonians. Computing
Hamiltonian form with respect to ωt = yi dtxi
the corrections between Hamiltonians by solving the partial differential equations of the
form (2.10) along the lines of lemma 2.1, we have the following.
Theorem 3.1. The space E constructed from the quasi-Painlevé-II system (3.8), with the coeffi-
cients (3.14) as dictated by the resonance conditions, can be described as a gluing of coordin-
ate patches
(
E =
C2
q,p
(cid:2) B
(
)
[
C2
x1,y1
(cid:2) B
)
(
[
C2
x2,y2
(cid:2) B
)
,
(3.24)
17
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
with gluing defined by
,
1
y1
1
2
q =
p =
q =
p =
+ λ1t+λ2
2
1 + 2a2(t)
y2
3 y3
1 +
−2λ1
−λ2
4
2+4λ3
1 + 6a0(t)−2(λ1t+λ2)a2(t)−4a
y4
y3
1
3
′
2 (t)
1 + x1y6
y5
1
,
,
1
y2
(cid:0) 1
2
(cid:0) λ1t+λ2
2
y2
2
(cid:0) 2a2(t)
3 y3
2 +
−2λ1+λ2
2
4
−4λ3
y4
2
(cid:0) 6a0(t)−2(λ1t+λ2)a2(t)+4a
3
y3
2
′
2 (t)
y5
2 + x2y6
2
,
in which we have the two-form
ωt = dtq ^ dtp = y1dtx1
^ dty1 = y2dtx2
^ dty2.
(3.25)
(3.26)
The Hamiltonian structure on E with respect to ωt for the system (2.18) is then given by
Hamiltonians H(q, p, t), H1(x1, y1, t), H2(x2, y2, t), which are related modulo functions of t
under the gluing (3.25) by
(cid:24)
= H1
H
(cid:24)
= H2 +
(cid:0) λ1
4
λ1
4
q2 (cid:0) 2a
2a
q2 +
′
2 (t)
3
′
2 (t)
3
q (cid:0) 2 (λ1a2 (t) (cid:0) 3a
2 (λ1a2 (t) (cid:0) 3a
q +
′
0 (t) + (tλ1 + λ2) a
3
′
0 (t) + (tλ1 + λ2) a
3
′
2 (t) + 2a
′ ′
2 (t))
′
2 (t) (cid:0) 2a
′ ′
2 (t))
−1
q
−1.
q
(3.27)
Moreover, for each i = 1, 2 the Hamiltonian Hi(xi, yi, t) is polynomial in xi, yi, with coefficients
analytic in t on B.
The system in the charts (x1, y1), (x2, y2) is computed by substitution into system (3.8)
according to the gluing (3.25). In the first chart it takes the form
y1
y1
dx1
dt
dy1
dt
=
∂H1
∂y1
= (cid:0) ∂H1
∂x1
= f (t) + y1
P (x1, y1, t) ,
= (cid:0) 1
2
+ y2
1
Q (x1, y1, t) ,
(3.28)
where P(x1, y1, t) and Q(x1, y1, t) are known polynomials in x1, y1 with coefficients analytic in
t on B, with both P(x1, 0, t), Q(x1, 0, t) 6= 0 as functions of x1 and t, and
f (t) = (λ1t + λ0) a0 (t) (cid:0) 2λ2
1t2 + 3λ2
2 + λ1 (4tλ2
6
(cid:0) 2) (cid:0) 4λ3
a2 (t) (cid:0) 2a
′
0 (t) +
′ ′
2 (t) .
a
4
3
(3.29)
In the system (3.28), we make a transformation changing the role of independent variable from
t to y1 to obtain
dx1
dy1
=
f (t) + y1
(cid:0) 1
+ y2
1
2
P (x1, y1, t)
Q (x1, y1, t)
,
dt
dy1
=
y1
Q (x1, y1, t)
,
(cid:0) 1
2
+ y2
1
(3.30)
18
(3.32)
(3.33)
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
and we have a regular initial value problem at every point (x1, y1, t) = (h1, 0, t∗) on the excep-
tional divisor L9 visible in the (u9, v9) chart for the fibre over t∗ 2 B. The analytic solution
(x1, t) of this initial value problem is of the form
(
)
)
(
t (y1) = t∗ (cid:0) y2
1 +
λ1t∗ + λ2
2
y4
1 +
8a2 (t∗)
15
y5
1 + O
y6
1
,
x1 (y1) = h1
(cid:0) 2f (t∗) y1 + O
y2
1
.
(3.31)
Inverting the power series gives Puiseux series expansions about t = t∗ for x1, y1, which when
mapped under the birational gluing in (3.25) to the original q, p variables gives
)
(
q =
p =
i
(t (cid:0) t∗)1/2
(cid:0)i
2 (t (cid:0) t∗)3/2
+
i (λ1t∗ + λ2)
4
(t (cid:0) t∗)1/2 +
+
i (λ1t∗ + λ2)
8 (t (cid:0) t∗)1/2
+
4a2 (t∗)
15
4a2 (t∗)
15
(
(t (cid:0) t∗)1 + O
(t (cid:0) t∗)3/2
,
)
+ O
(t (cid:0) t∗)1/2
.
Similarly, the system in the (x2, y2) is of the same form
y2
dx2
dt
=
∂H2
∂y2
,
y2
dy2
dt
= (cid:0) ∂H2
∂x2
,
as (3.28), with H2 polynomial in x2, y2 and analytic in t on B. After a similar transformation
exchanging the role of independent variable from t to y2 this defines regular initial value prob-
lems at every point (x2, y2, t) = (h2, 0, t∗) on the part of L15 away from the inaccesible divisors
in the fibre over t∗. The analytic solution is given by
t (y2) = t∗ + y2
2
(cid:0) λ1t∗ + λ2
2
y4
2
(cid:0) 8a2 (t∗)
15
y5
2 + O
(
)
,
y6
2
x2 (y2) = h2 + 2f (t∗) y2 + O
which corresponds to the following Puiseux series in the original variables:
q =
p =
1
(t (cid:0) t∗)1/2
(cid:0)1
2 (t (cid:0) t∗)3/2
(cid:0) λ1t∗ + λ2
4
(cid:0) λ1t∗ + λ2
8 (t (cid:0) t∗)1/2
(t (cid:0) t∗)1/2 (cid:0) 4a2 (t∗)
(
15
(
)
(t (cid:0) t∗)1 + O
(t (cid:0) t∗)3/2
,
)
(cid:0) 4a2 (t∗)
15
+ O
(t (cid:0) t∗)1/2
.
(
)
,
y2
2
(3.34)
(3.35)
Remark 3.1. The parameter hi from the initial value problem in the chart xi, yi enters the cor-
responding Puiseux series solution precisely through the coefficient which is allowed to be
arbitrary due to the resonance conditions.
3.3. Inaccessible divisors and auxiliary function
We now show that the divisors Ii removed from each surface to obtain the space E are indeed
inaccessible to the flow of the system, so the only movable singularities are those given by
the Puiseux series derived above from the solutions to the initial value problems on the last
exceptional divisors. For this we require an auxiliary function W on the bundle of compact
surfaces E which has poles along the divisors Ii in each fibre but remains bounded on the lift
of any finite length curve in B ending at a movable singularity of the system.
19
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
The Hamiltonian structure of the system on E can to some extent provide a way to construct
such a function, as in the case of the Painlevé equations [30]. The Hamiltonians H1(x1, y1, t)
and H2(x2, y2, t) constructed above are holomorphic on the part of the exceptional divisors L9,
L15 respectively contained in E, and both have the required poles on the inaccessible divisors
Ii. Therefore we can aim to construct an auxiliary function with the correct pole structure by
taking an appropriate sum of these Hamiltonians, but H1 has a pole along L15 and similarly
H2 has a pole along L9, so we require extra terms to stitch these together into a single function
with the required properties. Taking cues from the construction of such a function in [3], we
add sufficiently many terms in p/qk, k = 1, 2, . . . and obtain the following, which is verified
by direct calculation in charts.
Proposition 3.1. The function
W = H1 + H2 + λ1
8a
+
p
q
′
2 (t)
3
p
q2
(cid:24)
= 2H +
8a
′ ′
2 (t)
3
1
q
+ λ1
8a
+
p
q
′
2 (t)
3
p
q2
,
(3.36)
extended to the bundle E has the following properties.
(cid:15) Its restriction to the fibre Et = X
t has poles along all Ii, i = 1, . . . 15, but is analytic on the
parts of the exceptional divisors L9, L15 contained in Et,
(cid:15) Under the flow of the system, its logarithmic derivative W
′/W remains bounded on the
divisors Ii, which are given in (3.19) and indicated in blue on figure 4.
This result allows one to prove, using [14, lemma 2], that the divisors Ii are inaccessible by
analytic continuation of solutions along finite length curves.
Remark 3.2. The auxiliary function provided by the above proposition differs slightly from
that constructed using the method described in [3] and used in [14], which is of the form
˜W = H + p
4∑
k=1
ξk (t)
qk
,
(3.37)
with ξk(t) appropriately chosen functions of t. While both choices are sufficient to prove inac-
cessibility of the curves Ii, the method in [3] involves more analysis of the behaviour of the
original Hamiltonian function on the Puiseux series solutions, whereas here our construction
is informed by the Hamiltonian structure.
The system (3.8) with the resonance conditions (3.14) and additionally
a0
(cid:17) a2
(cid:17) 0,
(3.38)
reduces to an equation isolated by Halburd and Kecker [5] as having globally finite branching
about movable singularities. This equation is equivalent to a special case of a system obtained
by Takasaki [31] in the context of the Painlevé–Calogero correspondence [15, 16], which was
studied from a geometric point of view in [4]. This admits an algebraic (but not birational)
transformation to the Okamoto Hamiltonian form of PIV, for which there are several known
auxiliary functions used in proofs of the Painlevé property. For example, Shimomura [24] used
)
2
(
f
′
f
V =
(cid:0) f 3 (cid:0) 4tf 2 (cid:0) 4
(
t2 (cid:0) α
)
f +
2β
f
+ 4
′
f
f + 1
,
(3.39)
20
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
as a function that remains bounded about singular values of PIV, where α, β are the usual
parameters appearing in PIV, related to the root variables from the Okamoto Hamiltonian
form (2.34) by α = 1 (cid:0) a1
1. The relation between the Halburd–Kecker case
of system (3.8) and the Okamoto form of PIV is as follows. If q(t), p(t) solve the system (3.8)
with resonance conditions (3.14) and additionally (3.38), then f(˜t), g(˜t) defined by
(cid:0) 2a2, β = (cid:0)2a2
f (˜t) =
√
2
λ1
q (t)2 ,
g (˜t) =
q (t)3 + (λ1t + λ2) q (t) + 2p (t)
2λ1q (t)
p
2
,
˜t =
λ1t + λ2p
2λ1
,
(3.40)
solve the Okamoto Hamiltonian form (2.34) of PIV with t ! ˜t and the parameter a1 = 0, which
corresponds to β = 0 in PIV. In the β = 0 case of PIV, for Shimomura’s auxiliary function (3.39)
to have the required properties it is sufficient to approximate the term f
f+1 to first order in
1/f, so
′
)
2
(
f
′
f
˜V =
(cid:0) f 3 (cid:0) 4tf 2 (cid:0) 4
)
(
t2 (cid:0) α
f + 4
′
f
f
.
(3.41)
The following shows that the auxiliary function provided in proposition 4.1 corresponds in
the special case to that used by Shimomura for PIV.
Proposition 3.2. Under the transformation (3.40) from the Halburd–Kecker case of the quasi-
Painlevé-II system, i.e. (3.8) with coefficients given by (3.14) and (3.38), to the Okamoto
Hamiltonian form of PIV with parameter β = 0, Shimomura’s auxiliary function is related to
the auxiliary function W in proposition 4.1 by
p
(cid:24)
=
W
˜V,
8
2
λ3/2
1
(3.42)
where again
(cid:24)
= means equal modulo functions of only t analytic in B.
4. Kecker’s quartic Hamiltonian
We next consider the M = N = 3 case of the family of Hamiltonians (2.14) studied by Kecker
[12] which, in contrast to the M = N = 2 case equivalent to PIV studied in section 2, exhib-
its square root-type branching about movable singularities and is a genuine example of a
quasi-Painlevé equation. The Hamiltonian in question, after a rescaling for convenience, is
given by
(
p4 (cid:0) q4
)
+
∑
1
4
{
H =
I =
(i, j) 2 Z2
⩾0
ai,j (t) qi p j,
}
(i,j)∈I
j i + j ⩽ 3
,
(4.1)
where ai,j are analytic on some domain B. It was shown in [12] that the resonance conditions
necessary and sufficient for the system given by the Hamiltonian (4.1) with respect to the
canonical symplectic form to have the quasi-Painlevé property are
′
1,1
a
(cid:17) 0,
(
a2
2,1 + 2a0,2
) ′ (cid:17) 0,
(
a2
1,2
(cid:0) 2a2,0
) ′ (cid:17) 0.
(4.2)
21
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
We solve these conditions and set
a1,1 (t) = λ1,
a0,2 (t) = A2
(cid:0) 1
2
a2,1 (t)2 ,
a2,0 (t) = λ3
(cid:0) 1
2
a1,2 (t)2 ,
(4.3)
where λ1, λ2, λ3 are constants, and study the resulting Hamiltonian system
dq
dt
dp
dt
=
∂H
∂p
= (cid:0) ∂H
∂q
= p3 + a2,1 (t) q2 + 2a1,2 (t) q p + λ1q +
)
(cid:0) a2,1 (t)2
(
2λ2
(
p + a0,1 (t) ,
)
(4.4)
= q3 (cid:0) a1,2 (t) p2 (cid:0) 2a2,1 (t) q p (cid:0) λ1p (cid:0)
2λ3 + a1,2 (t)2
q (cid:0) a1,0 (t) .
The resonance conditions ensure that the system (4.4) admits Puiseux series solutions in the
neighbourhood of a movable singularity t = t∗ 2 B of the form
q (t) =
∞∑
n=−1
Cn (t (cid:0) t∗)n/2 ,
p (t) =
∞∑
n=−1
Dn (t (cid:0) t∗)n/2 ,
where
D−1 = (cid:0)2C3
−1,
C8
−1 =
1
16
,
(4.5)
(4.6)
and the coefficient guaranteed to be free by the resonance conditions can be chosen as C3,
so for a fixed leading coefficient C−1 satisfying (4.6) and a fixed value of C3, the rest of the
coefficients in (4.5) are determined. Taking into account the choice of branch of (t (cid:0) t∗)1/2
there are four distinct leading behaviours, which we will see in terms of the blowups required
to construct the space of initial conditions for the system (4.4) and in particular the number
of coordinate charts required for its global Hamiltonian structure. We remark that a space of
initial conditions for this system was constructed in [14] using P2 as initial compactification,
but in order to obtain a symplectic atlas and global Hamiltonian structure we will use P1 (cid:2) P1
as in the previous sections.
4.1. Surfaces
After extending the system (4.4) to P1 (cid:2) P1, we require a sequence of 17 blowups to resolve all
indeterminacies and arrive at the rational surface X
t which will form the fibre of the bundle E.
We give the locations of points to be blown up, in coordinates introduced according to the same
convention as in the previous sections, in figure 5, and a schematic description of the surface
in figure 6 with the divisors Ii, i = 1, . . . 15, which we will prove to be inaccessible, indicated
in blue. These are given in terms of the exceptional divisors Li arising from the blowups by
I1 = fq = 1g (cid:0) L1
(cid:0) L3,
I4 = L2
(cid:0) L7,
I7 = L6
(cid:0) L11,
I10 = L10
(cid:0) L15,
I13 = L14
I2 = fp = 1g (cid:0) L1,
(cid:0) L4,
I5 = L3
(cid:0) L8,
I8 = L7
I11 = L11
I14 = L15
(cid:0) L12,
(cid:0) L16,
(cid:0) L6
(cid:0) L2
(cid:0) L5,
(cid:0) L9,
I3 = L1
I6 = L4
I9 = L8
I12 = L12
I15 = L16
(cid:0) L13,
(cid:0) L17.
(cid:0) L10
(cid:0) L14,
(4.7)
The system written in the coordinate charts (ui, vi), i = 5, 9, 13, 17 covering the final four
exceptional curves arising from the blowups is regularisable by the same kinds of transforma-
tions transferring the role of independent variable from t to the coordinate vi providing the local
22
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
Figure 5. Blowup points for the Kecker quartic Hamiltonian system (2.18).
Figure 6. Surface for the Kecker quartic Hamiltonian system (2.18).
equation for the exceptional curve. Similarly to the previous section we remove the curves Ii
from each fibre Et = X
t of the bundle of compact surfaces and arrive at the space E with fibre
Et = X
i Ii.
n [
t
23
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
4.2. Symplectic atlas and global Hamiltonian structure
Just as in the previous examples, from inspection of the two-form dtq ^ dtp rewritten in charts
introduced in the standard way we see the need for a local coordinate change after the blowup
of p1 to obtain an appropriate atlas for a global Hamiltonian structure for the system (4.4)
on E to have all Hamiltonians being polynomial. Just as in section 2 this is a local change
(u1, v1) 7! (˜u1,˜v1) defined by
u1 =
1
˜u1
,
v1 = ˜v1,
Q =
˜v1
˜u1
, P = ˜v1,
so the exceptional line L1 has local equation ˜v1 = 0, and in these coordinates we have
p2 : (˜u1,˜v1) = (1, 0) ,
p6 : (˜u1,˜v1) = ((cid:0)1, 0) ,
p10 : (˜u1,˜v1) = ((cid:0)i, 0) ,
p14 : (˜u1,˜v1) = (i, 0) .
(4.8)
(4.9)
Proceeding with the rest of the blowups, introducing coordinates according to the same con-
vention as previously but with tildes to distinguish them from those in (5), we have expressions
for the two-form ωt extended from dtq ^ dtp in these new coordinates as follows:
^ dt˜u1
˜v3
1
dt˜v3
ωt = dtq ^ dtp =
dt˜v1
dt˜v2
= dt˜v4
^ dt˜u4 = ˜v5dt˜v5
^ dt˜u5
=
=
=
=
=
=
^ dt˜u2
˜v2
2
^ dt˜u6
˜v2
6
^ dt˜u10
˜v2
10
^ dt˜u14
˜v2
14
=
=
dt˜v6
dt˜v10
dt˜v14
dt˜v7
^ dt˜u3
˜v3
^ dt˜u7
˜v7
dt˜v11
^ dt˜u11
˜v11
^ dt˜u15
˜v15
dt˜v15
= dt˜v8
^ dt˜u8 = ˜v9dt˜v9
^ dt˜u9
(4.10)
= dt˜v12
^ dt˜u12 = ˜v13dt˜v13
^ dt˜u13
= dt˜v16
^ dt˜u16 = ˜v17dt˜v17
^ dt˜u17.
Therefore by setting (x1, y1) = ((cid:0)˜u5,˜v5), (x2, y2) = ((cid:0)˜u9,˜v9), (x3, y3) = ((cid:0)˜u13,˜v13), and
(x4, y4) = ((cid:0)˜u17,˜v17), we can obtain an atlas in which the two-form on the fibre Et is given
^ dtyi with yi = 0 providing a local equation for the corresponding
in charts by ωt = yi dtxi
exceptional divisor, and we have the following.
Theorem 4.1. The space E constructed from the system (4.4) defined by the M = N = 3 case
of Kecker’s Hamiltonian with the resonance conditions (4.2) can be described as a gluing of
coordinate patches
(
(
)
)
(
)
(
(
E =
C2
q,p
(cid:2) B
[
C2
x1,y1
[
C2
(cid:2) B
[
C2
(cid:2) B
[
C2
x4,y4
x3,y3
x2,y2
,
(4.11)
)
(cid:2) B
)
(cid:2) B
with gluing defined by
q =
1 + y1 (a1,2 + a2,1) + y2
1 (λ1 + λ2 + λ3) + y3
1
R
1
(cid:0) x1y4
1
y1
,
p =
1
y1
,
(4.12)
where
R
1 = a0,1 + a1,0 +
(
(cid:0)λ2 + λ3
(cid:0) a2
1,2
(
)
(cid:0)
a2,1
λ1 + 2λ2
(cid:0) a2
2,1
)
a1,2
(cid:0) a
′
2,1
(cid:0) a
′
2,1,
24
q =
where
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
(cid:0)1 + y2 (a1,2
q =
(cid:0) a2,1) + y2
2 (λ1
y2
(cid:0) λ2
(cid:0) λ3) + y3
2
R
2
(cid:0) x2y4
2
,
p =
1
y2
,
(4.13)
where
R
2 = (cid:0)a0,1 + a1,0 +
(
λ2
(cid:0) λ3
(cid:0) a2
1,2
)
(
a2,1 +
λ1
(cid:0) 2λ2 + a2
2,1
)
a1,2 + a
′
1,2
(cid:0) a
′
2,1,
(cid:0)i + y3 ((cid:0)a1,2 + ia2,1) + y2
3 ((cid:0)λ1
y3
(cid:0) iλ2 + iλ3) + y3
3
R
3
(cid:0) x3y4
3
,
p =
1
y3
,
q =
where
R
3 = (cid:0)ia0,1
(cid:0) a1,0
(cid:0)
(
iλ2 + iλ3
(cid:0) ia2
1,2
)
(
a2,1 +
(cid:0)iλ1 + 2λ2 + a2
2,1
)
a1,2
(cid:0) ia
′
1,2
(cid:0) a
′
2,1,
i + y4 ((cid:0)a1,2
(cid:0) ia2,1) + y2
4 ((cid:0)λ1 + iλ2
y4
(cid:0) iλ3) + y3
4
R
4
(cid:0) x4y4
4
,
p =
1
y4
,
(4.14)
(4.15)
R
4 = i a0,1
(cid:0) a1,0 +
(
iλ2 + iλ3
(cid:0) ia2
1,2
)
(
a2,1 +
iλ1 + 2λ2 + a2
2,1
)
a1,2 + ia
′
1,2
(cid:0) a
′
2,1,
in which we have written ai,j = ai,j(t) and a
given by
′
i,j = a
′
i,j(t). In these coordinates the two-form is
ωt = dtq ^ dtp = y1dtx1
^ dty1 = y2dtx2
^ dty2 = y3dtx3
^ dty3 = y4dtx4
^ dty4.
(4.16)
The Hamiltonian structure on E for the system is then given by the original Hamiltonian
H(q, p, t) as in system (4.4) as well as Hamiltonians Hk(xk, yk, t), k = 1, 2, 3, 4, which are related
modulo functions of t under the gluing (4.12)–(4.15) by
(cid:24)
= H1 + p
H
(cid:24)
= H3 + p
(
(
′
1,2 + a
′
2,1
a
)
(cid:0)
)
(cid:0)a
′
1,2 + ia
′
2,1
R ′
1
p
(cid:0)
R ′
3
p
(cid:24)
= H2 + p
(
a
)
(cid:0)
(cid:0) a
′
2,1
′
1,2
(
R ′
2
p
)
(cid:24)
= H4 + p
(cid:0)a
′
1,2
(cid:0) ia
′
2,1
(cid:0)
(4.17)
R ′
4
p
,
where R ′
Hamiltonian Hi(xi, yi, t) is polynomial in xi, yi, with coefficients analytic in t on B.
i as in (4.12)–(4.15) above. Moreover, for each i = 1, 2, 3, 4 the
i (t) with R
i = R ′
The system in the coordinates (xi, yi) for each case i = 1, 2, 3, 4 can be computed directly
and is of the form
yi
yi
dxi
dt
dyi
dt
=
∂Hi
dyi
= (cid:0) ∂Hi
dxi
= fi (xi, t) + yi
P
i (xi, yi, t) ,
(4.18)
= ci + yi
Q
i (xi, yi, t) ,
where ci is a known nonzero constant, fi(xi, t) is a known polynomial in xi with coefficients
analytic on B, and also P
i are known polynomials in xi, yi with coefficients analytic in
t on B with both P
i(xi, 0, t) 6= 0 as functions of xi and t. Interchanging the role of
independent variable as in section 3, we have
i(xi, 0, t), Q
i and Q
dxi
dyi
=
fi (xi, t) + yi
Q
ci + yi
i (xi, yi, t)
P
i (xi, yi, t)
,
dt
dyi
=
yi
Q
i (xi, yi, t)
,
ci + yi
(4.19)
25
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
so we have a regular initial value problem at every point (xi, yi, t) = (hi, 0, t∗) on the corres-
ponding exceptional divisor in the fibre over t∗ 2 B. We then compute the analytic solution
(xi, t) to this initial value problem as power series in yi, then invert these and transform from
(xi, yi) to (q, p) under the gluing to find solutions given by Puiseux series expansions about
movable algebraic branch points, recovering the results of [12, 14]. For example from the
(x1, y1) chart we find an analytic solution
t (y1) = t∗ (cid:0) 1
2
y2
1 +
2a1,2 (t∗) + a2,1 (t∗)
3
y3
1 + O
(
)
,
y4
1
x1 (y1) = h1 + O (y1) ,
(4.20)
which corresponds to the Puiseux series expansions in the original variables
q (t) =
p (t) =
ip
ip
2
2
(t (cid:0) t∗)
−1/2 +
a1,2 (t∗) + 2a2,1 (t∗)
3
+ O
(t (cid:0) t∗)
−1/2 (cid:0) 2a1,2 (t∗) + a2,1 (t∗)
3
(t (cid:0) t∗)1/2
(
(
)
)
,
,
(4.21)
+ O
(t (cid:0) t∗)1/2
about a movable singularity t = t∗. Performing the calculations for the rest of the charts, we
see that the Puiseux series solutions coming from the regular initial value problems on the
last exceptional divisors exhaust the four leading behaviours of solutions of (4.4) about mov-
able singularities as given by (4.5) with (4.6), and the parameter hi first appears in the same
coefficient that is allowed to be free due to the resonance conditions.
Remark 4.1. Using the birational relations between successive charts introduced during the
blowup process we can rewrite these expansions in ˜ui,˜vi coordinates and see the free parameter
h1 being brought further towards the leading term. For example for the sequence of blowups
culminating in p5 which gives the exceptional line L5 covered by the chart (x1, y1) = ((cid:0)˜u5,˜v5),
we have the following, in which τ = t (cid:0) t∗, d = (cid:0)i
are known functions of only
their arguments:
2, and c
(k)
i
p
)
(
(1)
(t∗) τ 1/2 + (cid:1) (cid:1) (cid:1) + c
4
(t∗; h1) τ 2 + O
τ 5/2
,
˜u1 = c
(1)
0
(1)
(t∗) + c
1
(
)
˜v1 = dτ 1/2 + O
τ 1/2
,
(2)
(2)
˜u2 = c
(t∗) + c
(
0
1
τ 1/2
˜v2 = dτ 1/2 + O
)
,
(t∗) τ 1/2 + (cid:1) (cid:1) (cid:1) + c
(2)
3
(t∗; h1) τ 3/2 + O
(
)
,
τ 2
(
)
τ 3/2
,
(4.22)
˜u3 = c
(3)
0
(3)
(t∗) + c
1
(
(t∗) τ 1/2 + c
)
(3)
2
(t∗; h1) τ 1 + O
˜v3 = dτ 1/2 + O
τ 1/2
,
(4)
1 (t∗; h1)τ 1/2 + O(τ 1),
(4)
˜u4 = c
0 (t∗) + c
˜v4 = dτ 1/2 + O(τ 1/2),
˜u5 = (cid:0)h1 + O(τ 1/2),
˜v5 = dτ 1/2 + O(τ 1/2).
{
{
Just as in the case of the Painlevé equations, the successive blowups bring the free parameter
in the series solutions further towards the leading behaviour.
26
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
4.3. Inaccessible divisors and auxiliary function
For this case we choose the same auxiliary function as in [14], which is of the form
W = H + A1 (t)
p2
q
+ A2 (t)
p3
q2
,
(4.23)
and is obtained using the approach of [12], in which terms of the form pk
teract the divergence of dH
dt
∂t . The following is proved by direct calculation.
= ∂H
ql are added to coun-
Proposition 4.1. The function
W = H (cid:0) a
′
2,1 (t)
p2
q
(cid:0) a
′
1,2 (t)
p3
q2
,
extended to the bundle E has the following properties.
(4.24)
(cid:15) The restriction of W to the fibre Et = X
t has poles along all Ii, i = 1, . . . 15, which are given
in (4.7) and indicated in blue on figure 6, but is analytic on the parts of the exceptional lines
L5, L9, L13, L17 contained in Et,
(cid:15) Under the flow of the system, the logarithmic derivative W
′/W remains bounded on the
divisors Ii, i = 1, . . . 15.
Therefore on the lift to E of any curve ending in a movable singularly the function W
remains bounded, but on the divisors I1, . . . , I15 it diverges, so these are inaccessible by [14,
lemma 2], completing the proof of the quasi-Painlevé property in this case.
5. A new equation of Painlevé-IV type
Motivated by the geometric regularisability [4] of Takasaki’s rational Painlevé–Calogero sys-
tem related to PIV [31], we consider the family of equations
d2y
dt2
=
3
4
y5 + a3 (t) y3 + a2 (t) y2 + a1 (t) y + a0 (t) +
a−2 (t)
y2
+
a−3 (t)
y3
,
(5.1)
where ai(t) are analytic on some B (cid:26) C. In the special case when the coefficients are given by
a−3 (t) =
β
2
,
a1 (t) =
(
)
,
t2 (cid:0) α
a3 (t) = 2t,
a−2 (t) = a0 (t) = a2 (t) = 0,
(5.2)
the equation reduces to a scaled version of Takasaki’s rational Painlevé–Calogero system [31],
so the equation (5.1) with coefficients (5.2) is transformed to the fourth Painlevé equation in
the standard form
(
)
PIV :
d2λ
dt2
=
1
2λ
dλ
dt
by the algebraic transformation
λ = y2.
2
+
3
2
λ3 + 4tλ2 + 2
(
)
λ +
t2 (cid:0) α
β
λ
,
(5.3)
(5.4)
In order to isolate quasi-Painlevé equations from the family (5.1), we will derive resonance
conditions by imposing regularisability on the final exceptional divisors arising in the resolu-
tion of indeterminacies of (5.1) written in Hamiltonian form.
27
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
Figure 7. Surface for the quasi-Painlevé-IV system (5.5).
5.1. Surfaces
Letting y = q, p = dy
system
dt , we write the scalar equation (5.1) as the non-autonomous Hamiltonian
dq
dt
=
H =
,
∂H
= (cid:0) ∂H
dp
∂p
∂q
dt
q6 (cid:0) a3 (t)
p2 (cid:0) 1
1
4
8
2
,
q4 (cid:0) a2 (t)
3
q3 (cid:0) a1 (t)
2
q2 (cid:0) a0 (t) q +
a−2 (t)
q
+
a−3 (t)
2q2
.
(5.5)
We extend this system first to the trivial bundle (P1 (cid:2) P1) (cid:2) B, and perform a sequence of 20
blowups of the fibre over t 2 B, after which all indeterminacies of the system are resolved.
The configuration of points to be blown up is given in figure 7, and is essentially the same as
for the quasi-Painlevé-II system (3.8), but with an added sequence of five blowups required
to resolve the singularity at (q, p) = (0, 1). For the sake of brevity we do not give the precise
locations of points in coordinates at this stage (since these will be given for the system after
imposing the resonance conditions), but there are four final exceptional divisors on which we
derive conditions for the system to be of a form regularisable by similar transformations to
those in the previous sections. In particular the conditions for regularisability on the two final
′
exceptional divisors L9 and L15 over (q, p) = (1, 1) are a
1 +
3 respectively, while on the two final exceptional lines L18 and L20 over (q, p) = (0, 1) the
a3a
condition is a
(cid:17) 0. Thus the resonance conditions reduce to
′
3 and a
(cid:17) (cid:0)2a
(cid:0) a3a
(cid:17) 2a
′ ′
3
′ ′
3
′
1
′
′
−3
′
−3
a
(cid:17) 0,
′ ′
3
a
(cid:17) 0,
2a
′
1
(cid:17) a3a
′
3,
(5.6)
with which it can be shown that the system defines regular initial value problems on these four
exceptional lines after interchanging the role of independent variables along the same lines as
in previous sections.
28
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
Figure 8. Blowup points for the quasi-Painlevé-IV system (5.5) with coefficients (5.7).
In solving the conditions (5.6) we note that through a simple scaling the constant a−3 can
be chosen without loss of generality to be a−3(t) = (cid:0)1/4, which will simplify the numerical
constants appearing in the calculations which will follow. Solving the resonance conditions
and choosing a−3 in this way leads to
a−3 (t) = (cid:0) 1
4
,
a1 (t) = λ1 +
)
2
(
λ2 + λ3t
2
,
a3 (t) = λ2 + λ3t,
(5.7)
where λ1, λ2, λ3 are arbitrary complex constants. After imposing these conditions by substitut-
ing (5.7) into the system (5.5), we perform a sequence of blowups of points given in coordinates
in figure 8 to construct a surface X
t which gives the fibre of the bundle E over B. We remove
the support of the divisors Ii, i = 1, . . . , 19 indicated in blue in figure 7 given by
29
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
(cid:0) L2
I1 = fq = 1g (cid:0) L1,
I2 = fp = 1g (cid:0) L1
(cid:0) L2,
I3 = L1
(cid:0) L3,
I4 = L2
(cid:0) L4
I5 = L3
(cid:0) L19,
(cid:0) L17
I16 = L16
I19 = fq = 0g (cid:0) L16,
(cid:0) L10,
(cid:0) L3
(cid:0) L16,
I6 = L4
I7 = L5
I8 = L6
I9 = L7
I10 = L8
I17 = L17
(cid:0) L5,
(cid:0) L6,
(cid:0) L7,
(cid:0) L8,
(cid:0) L9,
(cid:0) L18,
I11 = L10
I12 = L11
I13 = L12
I14 = L13
I15 = L14
I18 = L19
(cid:0) L11,
(cid:0) L12,
(cid:0) L13,
(cid:0) L14,
(cid:0) L15,
(cid:0) L20,
(5.8)
after which we have the bundle E over B with fibre Et = X
t
n [
i Ii.
5.2. Symplectic atlas and global Hamiltonian structure
We now show how to obtain a description of the space E as a gluing of copies of C2 (cid:2) B
and derive a global holomorphic Hamiltonian structure of the quasi-Painlevé-IV system on E.
While the charts required to cover the parts contained in Et of the final exceptional divisors L9
and L15 coming from the sequence of blowups over p1 : (q, p) = (1, 1) can be constructed
in a similar way to in the case of the quasi-Painlevé-II system (3.8), the exceptional lines L18
and L20 over p16 : (q, p) = (0, 1) require more work.
We begin with the exceptional divisors over p1 : (q, p) = (1, 1), where the need for a local
coordinate change on the exceptional divisor L3 is seen in the same way as in section 3. Without
such a local coordinate change the two-form is given in charts by
=
ωt = dtq ^ dtp =
^ dtv1
dtu1
1v3
u2
1
^ dtv3
= (cid:0) dtu3
3v5
u2
3
dtQ ^ dtP
Q2P2
^ dtV1
= (cid:0) dtU1
1V3
U2
1
^ dtv4
= (cid:0) dtu4
4 (2 + u4v4)2
v4
^ dtv2
= (cid:0) dtu2
u2
2v4
2
= (cid:0) dtu10
10 ((cid:0)2 + u10v10)2
v4
^ dtv10
(5.9)
.
The fact that the denominator of the rational coefficient of ωt in the chart (ui, vi) (respect-
ively (Ui, Vi)) is not just a power of vi (respectively Vi) means that curves other than Li along
which ωt has poles are visible in these coordinates, and proceeding with the introduction of
coordinates in the standard way will not result in the desired form of ωt in the final charts.
To remedy this, we make a local coordinate change on L3 that sends fu3 = 0g to infinity,
so the corresponding problematic components of the pole divisor will no longer be covered by
the subsequent charts. This is the same coordinate change (u3, v3) 7! (˜u3,˜v3) as in section 3,
namely
u3 =
1
˜u3
,
v3 = ˜v3,
Q = ˜v3, P =
˜v3
3
˜u3
,
(5.10)
after which proceeding with the rest of the blowups with tilded coordinates leads to the same
expressions for ωt as in (3.22), so we have the desired coordinates (x1, y1) = (˜u9,˜v9) and
(x2, y2) = (˜u15,˜v15) on L9 and L15 respectively.
30
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
Dealing with the other sequence of blowups over p16 : (q, p) = (0, 1) turns out to be more
subtle. We first note that in charts introduced in the standard way we have
ωt = dtq ^ dtp = (cid:0) dtq ^ dtP
P2
^ dtv17 = (cid:0)v18dtu18
= (cid:0)dtu17
= (cid:0) dtu16
^ dtv16
v16
^ dtv18 = (cid:0)dtu19
= (cid:0) dtU16
U2
^ dtV16
16V16
^ dtv19 = (cid:0)v20dtu20
(5.11)
^ dtv20,
so from the point of view of the two-form these coordinates do not at first glance pose a prob-
lem. However writing the system explicitly in the charts (u18, v18) and (u20, v20) respectively
we see that it is given by
du18
dt
dv18
dt
=
=
f1 (t) + g1 (t) u18 + v18
1 (u18, v18, t)
P
(
)
3
1
v18
2
c1 + v18
(
1
2
− a−2 (t) v18 + u18v2
18
1 (u18, v18, t)
− a−2 (t) v18 + u18v2
18
Q
)
3
,
v18
,
du20
dt
dv20
dt
=
=
f2 (t) + g2 (t) u20 + v20
2 (u20, v20, t)
P
(
v20
− 1
2
c2 + v20
− 1
2
(
v18
− a−2 (t) v20 + u20v2
20
2 (u20, v20, t)
Q
− a−2 (t) v20 + u20v2
20
)
3
,
)
3
,
(5.12)
where fi(t), gi(t) are known functions analytic on B, ci are known constants, and P
i are
known polynomials in their arguments with coefficients analytic on B. While the system in
these charts admits the same kind of regularisation as previous examples (corresponding to
square root-type branching about movable singularities) and the expressions for ωt guarantee
the existence of Hamiltonian functions in these charts via lemma 2.1, the Hamiltonians will
not be analytic everywhere in (u18, v18), respectively (u20, v20).
i, Q
Considering the problematic factors in the denominators of the right-hand-sides of the
equations above, on which the Hamiltonians diverge, we see that these correspond to noth-
ing more than the proper transform of the line q = 1 under the blowups, which is given in
the second chart introduced for L16 by U16 = 0. The fact that inspection of the two-form ωt in
coordinates did not detect the necessity to make a local coordinate change such that this curve
is not visible is related to the fact that a blowdown can be performed to contract the proper
transform of q = 0, on which the system diverges but ωt is holomorphic. We see from (5.11)
that the local coordinate change to ensure the proper transform of p = 1 is not visible can be
performed after the blowup of p16 by taking (U16, V16) 7! ( ˜U16, ˜V16) so that
U16 =
1
˜U16
, V16 = ˜V16,
q = ˜V16, P =
1
p
=
˜V16
˜U16
.
(5.13)
Following this we introduce charts according to the standard convention for the blowups of
the points
p16 : (q, P) = (0, 0)
{
(
(
˜U16, ˜V16
˜U16, ˜V16
)
)
p17 :
p19 :
= (1/2, 0) p18 : (˜u17,˜v17) = ((cid:0)a−2 (t) , 0) ,
= ((cid:0)1/2, 0) p20 : (˜u19,˜v19) = ((cid:0)a−2 (t) , 0) ,
(5.14)
and we see that the two-form on the fibre is given by
ωt = dtq ^ dtp = (cid:0) dtq ^ dtP
= (cid:0) dt
= (cid:0)dt˜u19
P2
^ dt˜v19 = (cid:0)˜v20dt˜u20
˜U16
^ dt
˜V16
^ dt˜v20.
˜V16
= (cid:0)dt˜u17
^ dt˜v17 = (cid:0)˜v18dt˜u18
^ dt˜v18
(5.15)
Thus we can take (x3, y3) = ((cid:0)u18, v18) and (x4, y4) = ((cid:0)u20, v20) and we have the following.
31
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
Theorem 5.1. The space E constructed from the quasi-Painlevé-IV system (5.5) with the coef-
ficients (5.7) can be described as a gluing of coordinate patches
)
)
(
(
)
(
(
(
E =
C2
q,p
(cid:2) B
[
C2
x1,y1
[
C2
(cid:2) B
[
C2
(cid:2) B
[
C2
x4,y4
x3,y3
x2,y2
,
(5.16)
)
(cid:2) B
)
(cid:2) B
with gluing defined by
q =
p =
q =
p =
,
1
y1
1
2
+ λ2+λ3t
2
1 + 2a2(t)
y2
3 y3
1 + 2λ1
−λ3
2
1 + 6a0(t)−2(λ2+λ3t)a2(t)−4a
y4
y3
1
3
′
2 (t)
1 + x1y6
y5
1
,
,
1
y2
(cid:0) 1
2
(cid:0) λ2+λ3t
2
y2
2
(cid:0) 2a2(t)
3 y3
2
(cid:0) 2λ1+λ3
2
y4
2 +
y3
2
−6a0(t)+2(λ2+λ3t)a2(t)+4a
′
2 (t)
3
2 + x2y6
y5
2
,
(5.17)
q = y3,
p =
1
2
(cid:0) 2a−2 (t) y3
(cid:0) x3y2
3
y3
,
q = y4,
p =
(cid:0) 1
2
+ 2a−2 (t) y4
(cid:0) x4y2
4
y4
in which we have the two-form
ωt = dtq ^ dtp = y1dtx1
^ dty1 = y2dtx2
^ dty2 = y3dtx3
^ dty3 = y4dtx4
^ dty4.
(5.18)
The Hamiltonian structure on E with respect to ωt for the system (5.5) is then given by
Hamiltonians H(q, p, t), H1(x1, y1, t), H2(x2, y2, t), H3(x3, y3, t) and H4(x4, y4, t) which are
related modulo functions of t under the gluing (5.17) by
H
(cid:24)
= H1
(cid:0) λ3
4
λ1
(cid:24)
= H2 +
4
(cid:24)
= H3 + 2a
q2 (cid:0) 2a
2a
q2 +
′
−2 (t) q
′
2 (t)
3
′
2 (t)
3
(cid:24)
= H4
q (cid:0) 2 (λ3a2 (t) (cid:0) 3a
2 (λ3a2 (t) (cid:0) 3a
q +
(cid:0) 2a
′
−2 (t) q.
′
0 (t) + (λ2 + λ3t) a
3
′
0 (t) + (λ2 + λ3t) a
3
′
2 (t) + 2a
′ ′
2 (t))
′
2 (t) (cid:0) 2a
′ ′
2 (t))
−1
q
−1
q
(5.19)
Moreover, for each i = 1, 2, 3, 4 the Hamiltonian Hi(xi, yi, t) is polynomial in xi, yi, with coef-
ficients analytic in t on B.
With this atlas and Hamiltonian structure in hand we can proceed to solve the regular initial
value problems on the four final exceptional divisors, invert the power series then map the
result back to (q, p) variables to obtain expansions of solutions about movable singularities.
From the charts (x1, y1) and (x2, y2) corresponding to the exceptional divisors L9 and L15, we
obtain two families of expansions which, with the freedom in choice of branch of (t (cid:0) t∗)1/2,
account for all the solutions of the quasi-Painlevé-IV system given as expansions
q =
C−1
(t − t∗)1/2
− C3
−1 (λ2 + λ3t∗)
4
(t − t∗)1/2 − 4C2
−1a2 (t∗)
15
(t − t∗) + · · · + C5 (t − t∗)5/2 + · · · ,
p = −
C−1
2 (t − t∗)3/2
− C3
−1 (λ2 + λ3t∗)
8 (t − t∗)1/2
− 4C2
−1a2 (t∗)
15
+ · · · +
5C5
2
(t − t∗)3/2 + · · ·
(5.20)
32
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
in which the leading coefficient C−1 satisfies C4
−1 = 1 and the coefficient C5 is free and
depends on the parameter hi from the initial data (xi, yi, t) = (hi, 0, t∗) for the initial value prob-
lem on the corresponding exceptional line.
On the other hand, from the charts (x3, y3) and (x4, y4) corresponding to the exceptional
divisors L18 and L20 over (q, p) = (0, 1) we obtain
q = ˜C1 (t (cid:0) t∗)1/2 (cid:0) 4˜C2
1a−2 (t∗)
3
(cid:0) 4˜C2
1a−2 (t∗)
3
1
2˜C3 (t (cid:0) t∗)1/2
p =
+
(t (cid:0) t∗) + ˜C3 (t (cid:0) t∗)3/2 + (cid:1) (cid:1) (cid:1) ,
3˜C3
2
(t (cid:0) t∗)1/2 + (cid:1) (cid:1) (cid:1) ,
(5.21)
in which ˜C1 satisfies ˜C4
through the coefficient ˜C3.
1 = 1 and the free parameter from the initial value problems enters
Remark 5.1. The fact that the system (5.5) in the special case when the parameters are given
by (5.2) reduces to Takasaki’s rational Painlevé–Calogero system related to PIV can also be
seen in terms of the Hamiltonian structure. The specialisation of coefficients does not affect
the poles and zeroes on Et of the two-form ωt, but it causes the Hamiltonians to depend only
on even powers of yi, i.e. Hi(xi, yi, t) = K(xi, y2
i , t) for some function K polynomial in xi, yi with
coefficients analytic in t on the space where the coefficients (5.2) are analytic, i.e. B = C. Thus
the system in the charts (xi, yi) becomes of the form
dxi
dt
=
1
yi
∂K
∂yi
(
)
,
xi, y2
i , t
= Fi
dyi
dt
= (cid:0) 1
yi
∂K
∂xi
= (cid:0) 1
yi
Gi
(
)
,
xi, y2
i , t
(5.22)
where Fi, and Gi are polynomial in their arguments. Then by the transformation yi
the system becomes
dxi
dt
= Fi (xi, Yi, t) ,
dYi
dt
= (cid:0)Gi (xi, Yi, t) ,
! Yi = y2
i
(5.23)
and the proof of the quasi-Painlevé property in this case becomes a proof that all solutions
of the system are algebraic over the field of meromorphic functions over B = C, sometimes
referred to as the algebro-Painlevé property. For a geometric description of this phenomenon
and the transformation (5.4) to PIV on the level of the space E, see [4].
5.3. Auxiliary function
In order to construct the auxiliary function required to complete the proof of the quasi-Painlevé
property for this new system, since the original Hamiltonian H(q, p, t) is no longer polynomial
and we have a singular value q = 0 we will need to extend the methods of [3, 12, 14] to include
wider classes of correction terms. In particular adding terms of the form p
qk as was done for
the quasi-Painlevé-II equation in section 3 no longer works due to the need to show that the
proper transform of q = 0 is inaccessible. Similarly adding terms qk
p leads to issues with q = 1.
Taking queues from Shimomura’s auxiliary function for PIV, we have the following.
Proposition 5.1. The function
W = H +
(
λ3
2
qp
(t)
′
2
3λ3
q2 (cid:0) 8a
q (cid:0) 1
) ,
33
(5.24)
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
extended to the bundle E has the following properties.
(cid:15) The restriction of W to the fibre Et has poles along all Ii, i = 1, . . . , 19 given in (5.8) and indic-
ated in blue on figure 7, but is analytic on the parts of the exceptional lines L9, L15, L18, L20
contained in Et.
(cid:15) Under the flow of the system, its logarithmic derivative W
′/W remains bounded on the
divisors Ii, i = 1, . . . , 19.
Therefore by [14, lemma 2] the divisors Ii, are inaccessible by analytic continuation of
solutions along finite length curves, and we have the following:
Theorem 5.2. The system given by (5.5) with coefficients (5.7) has the quasi-Painlevé prop-
erty, with all movable singularities reachable by analytic continuation along finite length
curves being square root-type algebraic branch points.
Remark 5.2. In the special case when the coefficients in the quasi-Painlevé-IV system are
given by (5.2) and it is related by the transformation (5.4) to PIV, under this transformation the
auxiliary function in proposition 5.1 recovers that constructed by Shimomura for PIV, so in a
sense our auxiliary function is a generalisation of this.
6. Conclusions and discussion
In summary, we have shown that regularisation of differential equations whose movable singu-
larities reachable by analytic continuation of solutions along finite-length curves are at worst
algebraic branch points can be seen as a property of a global Hamiltonian structure on an
extended phase space similar to Okamoto’s space for the Painlevé equations.
In the case of the Painlevé equations the space E is equipped with a holomorphic symplectic
form on each fibre, with respect to which a global holomorphic Hamiltonian structure of a dif-
ferential system leads to regular initial value problems everywhere on E. In the quasi-Painlevé
case, the global Hamiltonian structure is holomorphic but the two-form on the fibre has zeroes
on the final exceptional divisors arising in the sequence of blowups required to resolve inde-
terminacies of the differential system. The examples of quasi-Painlevé equations considered
in this paper admit only square root-type branching about movable singularities, which corres-
ponds to the fact that the two-form has zeroes of order one along the exceptional lines. More
generally, holomorphic Hamiltonian structures with respect to a rational two-forms with a zero
of order n (cid:0) 1, n ⩾ 2 along a final exceptional line will give expansions in (t (cid:0) t∗)1/n about a
movable singularity t = t∗. In a local coordinate chart (x, y) in which such an exceptional line
has local equation y = 0 and the two-form on the fibre is given by
ωt = yn−1dtx ^ dty,
(6.1)
if a system possesses a holomorphic Hamiltonian structure on E with Hamiltonian H(x, y, t) in
this chart it will take the form
dx
dt
=
1
yn−1
∂H
∂y
,
dy
dt
= (cid:0) 1
yn−1
∂H
∂x
,
so the transfer of the role of independent variable from t to y leads to
dt
dy
=
yn−1
∂H/∂x
,
dx
dy
= (cid:0) ∂H/∂y
∂H/∂x
,
34
(6.2)
(6.3)
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
Therefore for this system to possess analytic solutions to initial value problems at a point
(x, y, t) = (h, 0, t∗) on the exceptional divisor it is sufficient to require that the Hamiltonian
= c + yP(x, y, t), where P is
function is, in addition to being analytic in (x, y), such that ∂H
∂x
analytic in (x, y) everywhere and in t on the base space B of E. Then the initial value problem
above will have solutions of the form
t (y) = t∗ + C
−1
1 yn + O
(
)
,
yn+1
x (y) = h + O (y) ,
(6.4)
where C1
6= 0, so inversion of the power series yields
y (t) = C1 (t (cid:0) t∗)1/n +
∞∑
i =2
Ck (t (cid:0) t∗)k ,
x (t) = h + O
(t (cid:0) t∗)1/n
,
(6.5)
(
)
which under the birational transformation back to the original variables leads to h paramet-
rising a family of Puiseux series expansions of solutions with an nth root-type algebraic branch
point at t = t∗. Therefore as long as one can show that the divisors removed after the blowups
are inaccessible to solutions analytically continued along finite length curves, an atlas and
Hamiltonian structure as above guarantees the quasi-Painlevé property.
We must admit that the construction of such an atlas is by no means canonical or even guar-
anteed to be possible, as is already the case for Okamoto’s spaces for the Painlevé equations. In
the examples studied in this paper the geometry of the surfaces was shown to some extent guide
one in the construction, but it would be interesting to construct the spaces E for other known
quasi-Painlevé equations and see if they admit similar atlases and holomorphic Hamiltonian
structures, with a view to understanding why these atlases exist. Perhaps the most natural and
interesting next step in the development of a theory of quasi-Painlevé equations in terms of
rational surfaces is to prove uniqueness results, which would add weight to the idea that the
study of these equations, just as in the Painlevé case, reduces in principal to geometry.
Data availability statement
No new data were created or analysed in this study.
Acknowledgments
the National Science Center
G F acknowledges the support of
(Poland) via grant
2017/25/B/ST1/00931. The work of G F is also partially supported by the project PID2021-
124472NB-I00 funded by MCIN/AEI/10.13039/501100011033 and by ‘ERDF A way of
making Europe’. A S is supported by a Japan Society for the Promotion of Science (JSPS)
Postdoctoral Fellowship for Research in Japan and also acknowledges the support of JSPS
KAKENHI Grant Numbers 21F21775 and 22KF0073. A S would also like to thank Professor
Ralph Willox and Dr Takafumi Mase for helpful comments and discussions.
ORCID iDs
Galina Filipuk https://orcid.org/0000-0003-2623-5361
Alexander Stokes https://orcid.org/0000-0001-6874-7141
35
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
References
[1] Chiba H 2016 The first, second and fourth Painlevé equations on weighted projective spaces J.
Differ. Equ. 260 1263–313
[2] Dzhamay A, Filipuk G, Lige¸ za A and Stokes A 2021 Different Hamiltonians for differential
Painlevé equations and their identification using a geometric approach p 43
[3] Filipuk G and Halburd R 2009 Movable algebraic singularities of second-order ordinary differential
equations J. Math. Phys. 50 023509
[4] Filipuk G and Stokes A 2023 Takasaki’s rational fourth Painlevé-Calogero system and geometric
regularisability of algebro-Painlevé equations Nonlinearity 36 5661–97
[5] Halburd R and Kecker T 2014 Local and global finite branching of solutions of ordinary differ-
ential equations Proc. Workshop on Complex Analysis and its Applications to Differential and
Functional Equations (Publ. Univ. East. Finl. Rep. Stud. For. Nat. Sci., 14, Univ. East. Finl., Fac.
Sci. For. Joensuu) pp 57–58
[6] Hinkkanen A and Laine I 1999 Solutions of the first and second Painlevé equations are meromorphic
J. Anal. Math. 79 345–77
[7] Hinkkanen A and Laine I 2001 Solutions of a modified third Painlevé equation are meromorphic J.
Anal. Math. 85 323–37
[8] Hinkkanen A and Laine I 2001 Solutions of a modified fifth Painlevé equation are meromorphic
Papers on Analysis (Rep. Univ. JyväSkylä Dep. Math. Stat. vol 83) (Univ. Jyväskylä) pp 133–46
[9] Hinkkanen A and Laine I 2004 The meromorphic nature of the sixth Painlevé transcendents J. Anal.
Math. 94 319–42
[10] Iwasaki K and Okada S 2016 On an orbifold Hamiltonian structure for the first Painlevé equation
J. Math. Soc. Japan 68 961–74
[11] Kecker T 2016 A cubic Hamiltonian system with meromorphic solutions Comput. Methods Funct.
Theory 16 307–17
[12] Kecker T 2016 Polynomial Hamiltonian systems with movable algebraic singularities J. Anal. Math.
129 197–218
[13] Kecker T 2019 Space of initial conditions for a cubic Hamiltonian system Complex Var. Elliptic
Equ. 64 132–42
[14] Kecker T and Filipuk G 2022 Regularising transformations for complex differential equations with
movable algebraic singularities Math. Phys. Anal. Geom. 25 43
[15] Levin A M and Olshanetsky M A 2000 Painlevé-Calogero correspondence Calogero-Moser-
Sutherland Models (CRM Series in Mathematical Physics) (Springer) pp 313–32
[16] Manin Y I 1998 Sixth Painlevé equation, universal elliptic curve and mirror of P2
American Mathematical Society Translation Series 2 (Advances in Mathematical Science, 39
vol 186) (American Mathematical Society)
[17] Matano T, Matumiya A and Takano K 1999 On some Hamiltonian structures of Painlevé systems.
II J. Math. Soc. Japan 51 843–66
[18] Matumiya A 1997 On some Hamiltonian structures of Painlevé systems. III Kumamoto J. Math.
10 45–73
[19] Okamoto K 1979 Sur les feuilletages associés aux équations du second ordre `a points critiques
fixes de P. Painlevé, (French) [On foliations associated with second-order Painlevé equations
with fixed critical points] Japan. J. Math. 5 1–79
[20] Okamoto K 1980 Polynomial Hamiltonians associated with Painlevé equations, I Proc. Japan Acad.
Ser. A Math. Sci. 56 264–8
[21] Okamoto K and Takano K 2001 The proof of the Painlevé property by Masao Hukuhara Funkcial.
Ekvac. 44 201–17
[22] Sakai H 2001 Rational surfaces associated with affine root systems and geometry of the Painlevé
equations Commun. Math. Phys. 220 165–229
[23] Shioda T and Takano K 1997 On some Hamiltonian structures of Painlevé systems. I Funkcial.
Ekvac. 40 271–91
[24] Shimomura S 2003 Proofs of the Painlevé property for all Painlevé equations Japan. J. Math.
29 159–80
[25] Shimomura S 2005 On second order nonlinear differential equations with the quasi-Painlevé prop-
erty II RIMS K¯oky¯uroku 1424 177–83
[26] Shimomura S 2007 A class of differential equations of PI-type with the quasi-Painlevé property
Ann. Mat. Pura Appl. 186 267–80
36
J. Phys. A: Math. Theor. 56 (2023) 495205
G Filipuk and A Stokes
[27] Shimomura S 2008 Nonlinear differential equations of second Painlevé type with the quasi-Painlevé
property Tohoku Math. J. 60 581–95
[28] Steinmetz N 2000 On Painlevé’s equations I, II and IV J. Anal. Math. 82 363–77
[29] Steinmetz N 2018 An old new class of meromorphic functions J. Anal. Math. 134 615–41
[30] Takano K 2000 Defining manifolds for Painlevé equations Toward the Exact WKB Analysis of
Differential Equations, Linear or Non-Linear vol 204 (Kyoto University Press) pp 261–9
[31] Takasaki K 2001 Painlevé–Calogero correspondence revisited J. Math. Phys. 42 1443–73
37
| null |
10.3390_genes14061231.pdf
|
Data Availability Statement: Not applicable.
|
Data Availability Statement: Not applicable. Acknowledgments: The article publishing fees were funded by the University of Oradea.
|
Article
The Use of Xpert MTB/RIF Ultra Testing for Early Diagnosis
of Tuberculosis: A Retrospective Study from a
Single-Center Database
Cristian Sava 1,2
Cristian Phillip Marinău 1,2 and Andreea Bianca Balmos,
, Mihaela Sava 2, Ana-Maria Drăgan 1,2, Alin Iuhas 1,2,*
1,2
, Larisa Niulas,
1,2
,
1
Faculty of Medicine and Pharmacy, University of Oradea, 410087 Oradea, Romania
2 Clinical Emergency Bihor County Hospital, 410167 Oradea, Romania
* Correspondence: [email protected]
Abstract: Tuberculosis (TB) is a multisystemic contagious disease produced by Mycobacterium
tuberculosis complex bacteria (MTBC), with a prevalence of 65:100,000 inhabitants in Romania (six
times higher than the European average). The diagnosis usually relies on the detection of MTBC
in culture. Although this is a sensitive method of detection and remains the “gold standard”, the
results are obtained after several weeks. Nucleic acid amplification tests (NAATs), being a quick and
sensitive method, represent progress in the diagnosis of TB. The aim of this study is to assess the
assumption that NAAT using Xpert MTB/RIF is an efficient method of TB diagnosis and has the
capacity to reduce false-positive results. Pathological samples from 862 patients with TB suspicion
were tested using microscopic examination, molecular testing and bacterial culture. The results show
that the Xpert MTB/RIF Ultra test has a sensitivity of 95% and a specificity of 96.4% compared with
54.8% sensitivity and 99.5% specificity for Ziehl–Neelsen stain microscopy, and an average of 30 days
gained in the diagnosis of TB compared with bacterial culture. The implementation of molecular
testing in TB laboratories leads to an important increase in early diagnostics of the disease and the
prompter isolation and treatment of infected patients.
Keywords: tuberculosis; molecular testing; Xpert MTB/RIF Ultra; Ziehl–Neelsen staining; Lowenstein–
Jensen medium culture
1. Introduction
Tuberculosis (TB) represents a serious global health issue, and is one of the leading
morbidity and mortality factors. Every year, several million people worldwide become
infected with tuberculosis and lose their lives due to the disease. TB is a disease linked with
poverty and economic stress; vulnerability, marginalization, stigma and discrimination are
often problems that people with TB must confront [1].
TB is a multisystemic contagious disease caused by Mycobacterium tuberculosis complex
bacteria (MTBC). It is estimated that over 1.7 billion people (over 25% of world population)
are infected with MTBC. The global incidence had a peak in 2003, and has slowly been
decreasing since then. According to the latest World Health Organization (WHO, Geneva,
Switzerland) report, the estimated number of deaths from TB experienced a decline between
2005 and 2019, with over 10 million people contracting TB and 1.4 million dying in 2019 and
1.5 million in 2020; however, the estimates for 2020 and 2021 indicate that this trend has
been reversed, with an increase in the number of deaths [1,2]. Poverty, HIV infection and
drug resistance are the principal factors that contribute to the re-emergence of the global
TB epidemic [3]. It is projected that in 2020 and 2021, tuberculosis (TB) will be the second
most common cause of death attributed to a single infectious agent, following COVID-
19 [1]. About 95% of the cases are recorded in developing countries; one in every nine
new cases affects HIV-infected people; and 75% of all cases occur in Africa. It is estimated
Citation: Sava, C.; Sava, M.; Dr˘agan,
A.-M.; Iuhas, A.; Niulas, , L.; Marin˘au,
C.P.; Balmos, , A.B. The Use of Xpert
MTB/RIF Ultra Testing for Early
Diagnosis of Tuberculosis: A
Retrospective Study from a
Single-Center Database. Genes 2023,
14, 1231. https://doi.org/10.3390/
genes14061231
Academic Editor: Nathalie
Bissonnette
Received: 28 April 2023
Revised: 2 June 2023
Accepted: 6 June 2023
Published: 7 June 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under
the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Genes 2023, 14, 1231. https://doi.org/10.3390/genes14061231
https://www.mdpi.com/journal/genes
genesG C A TT A C GG C A T Genes 2023, 14, 1231
2 of 10
that 500,000 new multi-drug-resistant TB (MDR-TB) or rifampicin-resistant TB cases occur
annually [1].
TB epidemiology varies substantially around the world. The highest prevalence (over
100:100,000 inhabitants) can be observed in Sub-Saharan Africa, India and South-East
insular Asia and Micronesia. Intermediary rates (25–99 cases per 100,000 inhabitants) are
found in China, Central and South America, Eastern Europe and North Africa. Lower
prevalence (under 25 cases per 100,000 inhabitants) can be observed in North America,
Western Europe, Japan and Australia [1]. In 2018, there were 52,862 cases reported in the
European Union and the European Economic Space (EU/EES), resulting a prevalence of
10.2 cases per 100,000 inhabitants. The prevalence and the incidence in EU/EES countries
had declined over the last five years [4]. Unfortunately, Romania remains the country with
the highest prevalence from the EU/EES—64.6 cases per 100,000 inhabitants in 2017, which
is four times higher than the EU mean, with one of the lowest recovery rates and, at the
same time, an annual increase in the infectious reservoir. Romania has a mortality rate
due to TB of 4.2 per 100,000 inhabitants, more than six times higher than the EU mean and
1.9 times higher than the WHO European Region’s mean, according to the latest report from
the INSP—CNSISP (Romania’s National Public Health Institute, Bucharest, Romania) [5,6].
During the 2014 World Health Summit in Geneva, the WHO proposed a global strategy
and targets for tuberculosis prevention, care and control, aiming to stop the global TB
epidemic [7]. The proposed objectives were a 95%reduction in TB-related death by 2030,
a 90% reduction in disease incidence in the 2015–2035 period and the elimination of
associated catastrophic costs for tuberculosis-affected households. In addition to targets
for 2030, the End TB Strategy defines 2020 and 2025 milestones for reductions in TB
incidence and in the number of TB deaths. The 2020 milestones are a 20% reduction in
TB incidence and a 35% reduction in the number of TB deaths, compared with levels in
2015 [8,9]. Reaching these objectives requires the early diagnosis of TB, including through
the improvement of diagnostic methods, complete treatment of all people with TB, and the
diagnosis and treatment of latent TB infection. The COVID-19 pandemic hugely affected
patients’ access to proper medical services. TB care and prevention were particularly
affected by the redirection of human, financial and other resources to the COVID-19
response. Furthermore, public health measures resulted in reducing access to TB diagnosis
and treatment services [10].
The early diagnosis of tuberculosis enables the prompt initiation of treatment and has
the potential to restrict the transmission of this infectious disease. Its diagnosis usually
relies on the detection of MTBC in culture. Although this is a sensitive method of detection
and remains the “gold standard”, the results are obtained after several weeks. Microscopic
examination is an inexpensive and quick test, but is also a rather insensitive test and cannot
distinguish between non-tuberculosis mycobacteria and MTBC or between susceptible and
resistant strains. Nucleic acid amplification tests (NAATs), being a quick and sensitive
method, represent progress in the diagnosis of TB [11].
The WHO recommends replacing microscopic examinations, as the initial diagnostic
method, with molecular tests capable of identifying MTBC, in certain epidemiological
and geographical settings. The newer, more rapid and more sensitive molecular tests
recommended for the initial detection of MTBC and drug resistance are designated as
mWRDs (molecular WHO-recommended rapid diagnostics tests); these include Xpert
MTB/RIF Ultra and Xpert MTB/RIF (Cepheid, Sunnyvale, CA, USA); Truenat MTB, MTB
Plus and MTB-RIF Dx tests (Molbio Diagnostics, Goa, India); and loop-mediated isothermal
amplification (TB-LAMP; Eiken hemical, Tokyo, Japan) [12].
The Xpert MTB/RIF method is a molecular test that has the capacity to detect the
MTBC and the rpoB gene variant associated with rifampicin resistance [13]. Molecular tests
are becoming increasingly pertinent in the diagnosis of various diseases as their accessibility
and performance capabilities continue to improve [14].
The primary objective of this study was to investigate several hypotheses regarding
the efficiency of nucleic acid amplification tests (NAATs) using the Xpert MTB/RIF Ultra
Genes 2023, 14, 1231
3 of 10
method in the early diagnosis of tuberculosis (TB) and its impact on prompt treatment initi-
ation in positive cases. Additionally, the study aimed to assess the ability of this diagnostic
approach to reduce false-positive results in suspected TB cases and avoid unnecessary
administration of antituberculosis treatment. Moreover, the Xpert MTB/RIF Ultra test was
evaluated for its capacity to identify mutations in the rpoB gene associated with rifampicin
resistance in samples where Mycobacterium tuberculosis complex (MTBC) was detected. The
specific objectives of the study were as follows: (i) evaluating the sensitivity and specificity
of molecular tests compared to microscopic examination and mycobacterial culture for TB
diagnosis, (ii) estimating the time saved in initiating tuberculostatic treatment utilizing
molecular tests, (iii) analyzing molecular tests’ ability to identify non-tuberculosis mycobac-
terial infections and reduce false-positive results, and (iv) detecting rifampicin resistance.
2. Materials and Methods
During the period of 1 January 2018–31 December 2020, in the TB Bacteriology Labora-
tory of the “Dr. Gavril Curteanu” Municipal Clinical Hospital (currently Clinical Emergency
Bihor County Hospital) in Oradea, Bihor County, Romania, 13,916 biological specimens
were analyzed with the purpose of identifying MTBC. All these samples were tested using
microscopic examination and bacterial culture. In 862 cases, the specimens were also tested
using the Xpert MTB/RIF Ultra method.
In this study, 862 patients with a high suspicion of TB infection were included. The
suspicion of the disease was determined in accordance with the guidelines provided by the
Romanian National Guideline for the prevention, surveillance and control of tuberculosis
criteria (epidemiological, clinical and/or imagistic), whose samples were also analyzed
using the Xpert MTB/RIF Ultra method, in the mentioned period [15]. The samples
consisted of sputum obtained via direct matinal sampling, induced sputum, bronchial
aspirate, gastric aspirate, pleural puncture or lumbar puncture (CSF). The quality of the
biological samples was essential in obtaining a trustworthy result. Sputum samples deemed
inadequate (thin, clear sputum; improper sampling) were excluded from the study.
The collected data were analyzed using IBM SPSS Statistics version 26.
2.1. Microscopic Examination Technique
Microscopic examination was performed for all the samples. Sputum was the elective
pathological product.
The sputum smear for the microscopic examination was prepared using a bacteriologic
wire loop, choosing the spots with purulent, opaque sputum and spreading it on the
central portion of the slide, uniformly, in a thin layer, on a surface area of approximately
1 × 2 cm, avoiding the edge of the slide. The slides were left to dry under the hood, at
room temperature, and then, heat-fixated using a Bunsen burner (3 times). Ziehl–Neelsen
staining was used for acid-fast bacilli (AFB) detection. The slide’s surface was flooded with
0.3% Fucsina fenica and heated until steaming. The process was repeated 3–4 times. After
10 min, the slides were rinsed under a gentle flow of water until all free stain was washed
away. Decolorization was performed by flooding the slides with 3% acid-alcohol for 3 min,
and rinsing them thoroughly with water afterwards. Re-colorization was performed by
covering the slides with 0.3% methylene blue for 30 s.
The technique for the other specimens was similar, the only difference being the
processing method of the pathological product (prior centrifugation).
After washing and drying the slides, microscopic examination was performed using
an optic microscope with an immersion lens (100×) and an ocular lens (10×). The slide
was examined over the entire length of the smear. A minimum of 100 fields were examined
before the smear was reported as negative (Table 1).
2.2. Bacterial Culture Technique
A bacterial culture examination was conducted for all the samples using NaOH
method, without centrifugation (dripping method).
Genes 2023, 14, 1231
4 of 10
Table 1. Semiquantitative expression of the microscopic examination results.
Number of AFB under Ziehl–Neelsen Staining
Result
0 AFB
1–9 AFB/100 fields
10–99 AFB/100 fields
1–10 AFB/field
>10 AFB/field
Negative
Positive, scanty (exact value)
Positive 1+
Positive 2+
Positive 3+
From the pathological product, 2–3 mL of purulent particles were extracted using a
Pasteur pipette and put into a sterile tube with a threaded cap. An equal amount of 4%
NaOH with pH indicator was added. The capped tube was put in a mechanical agitator
for 10–15 s. Then, the tube was left at room temperature for 15 min. Neutralization of the
sample is performed using 8% HCl until the color turned greenish yellow (neutral pH).
The culturing was performed using a single-use pipette.
The used culture medium was Lowenstein–Jensen; for every sample, 3 medium tubes
were used. After culturing, the tubes were left in a temperature-controlled room at 37 ◦C,
with the cap half closed, at a 25–30◦ angle, for 2–5 days. The first reading was taken after
48 h, leaving the tubes vertical afterwards, and eliminating the contaminated tubes. The
cultures were monitored weekly until the end of the 8-week period (60 days) of incubation
(Table 2).
Table 2. Semiquantitative expression of the Lowenstein–Jensen solid medium culture results.
Mycobacterium Growth
Result
Absence of colonies
Under 30 colonies
30–100 colonies
Over 100 colonies
Uncountable conflated colonies
3 or 2 tubes contaminated and a tube without Bacterial growth
Negative
Positive, scanty (exact value)
Positive 1+
Positive 2+
Positive 3+
Contaminated
2.3. Xpert MTB/RIF Ultra Test Technique
Xpert MTB/RIF Ultra (Cepheid AB Röntgenvägen 5 171 54, Solna, Sweden) is an
automatized molecular test using nested real-time PCR for the qualitative detection of
M complex and rifampicin resistance, simultaneously. The primers of this test amplify
a region of the rpoB gene containing 81 base-pairs in the core region. The probes are
designed to distinguish between wild-type sequences and mutations in the core region,
which are associated with rifampicin resistance. The tests were performed using Cepheid
GeneXpert® Systems equipment (Cepheid, 904 Caribbean Drive, Sunnyvale, CA, USA),
which automatizes and integrates the sample purification, amplifies the nucleic acids and
detects the targeted sequence using RT-PCR.
The system consisted of apparatus, a computer and dedicated software, and it was
used for the execution of the test and visualization of the results. The system uses single-
use GeneXpert® cartilages which contain the reactive, the RT-PCR process, a sample
processing control (SPC) and a probe check control (PCC). Due to the autonomic nature
of these cartridges, and the automatic processes, the likelihood of cross-contamination
between samples is low. SPC has the role of controlling the bacterial processing and of
monitoring the presence of the inhibitor in the PCR reaction. PCC checks the reactive
rehydration, the PCR tube feeling, the probe integrity and the colorant stability. Xpert
MTB/RIF Ultra simultaneously detects the presence of the M. tuberculosis (MTB) complex
and rifampicin (RIF) resistance by amplifying the specific sequence form the rpoB gene,
which is marked with five signaling molecules (probes A to E) for the mutations of the
rifampicin resistance determining region (RRDR). Each signaling molecule was marked
with a different fluorophore. The cycle threshold (Ct) was set at 39.0 for the A, B and C
probes and at 36.0 for the D and E probes [16].
Genes 2023, 14, 1231
5 of 10
The Xpert MTB/RIF Ultra test was performed for 536 samples during the duration
of the study. For each test, 1 mL of sputum was used, which was sampled with a sterile
pipette and transferred into a sealed sterile tube. A total of 2 mL of reactive was added with
bactericide and mucus lysis properties. After 10 s of vigorous agitation and 10 min rest
at room temperature, followed by further vigorous agitation and 5 min rest, a uniformly
homogenized solution was obtained. The content of the tube was transferred to the reaction
cartilage using the producer-provided pipette.
The test took 90 min, and the results were displayed. GeneXpert® Instrument Systems
generates results using preestablished algorithms. The interpretation of the measurements
is found in Table 3.
Table 3. Possible results of the Xpert MTB/RIF Ultra test.
MTB detected/rifampicin resistance detected
MTB detected/rifampicin resistance not detected
MTB detected/rifampicin resistance indeterminate
MTB not detected
Invalid result
3. Results
In the observed period a total of 862 patients suspected of TB infection were tested with
Xpert MTB/RIF Ultra, Ziehl–Neelsen stain and culture on Lowenstein–Jensen medium.
In 2018, 320 (37.1%) tests were performed, in 2019, 289 (33.5%) tests were performed and
in 2020, 253 (29.4%) tests were performed. From the study sample, 643 (74.6%) were adults
and 219 (25.4%) were pediatric patients.
The collected pathological samples were as follows: 353 (41%)—sputum, 384 (44.5%)—
induced sputum, 24 (2.8%)—bronchial aspirate, 73 (8.5%)—gastric aspirate, 7 (0.8%)—
pleural fluid, 11 (1.3%)—cerebral spinal fluid and 10 (1.2%)—other pathological products
(examples of such fluids include synovial fluid from joints and pus from abscesses located
in various regions).
Out of the 862 tested molecular samples, 306 (35.5%) were positive—MTB detected
(121 positive samples in 2018, 132 in 2019 and 53 in 2020), and 556 (64.5%) were negative. In
the microscopy test, 694 (80.5%) samples were negative and only 168 (19.5%) were positive.
Regarding the culture, 299 (34.7%) had a positive culture, 560 (65%) were negative and
3 samples (0.3%) were contaminated (Table 4).
Table 4. Distribution of negative and positive results in the three TB tests studied.
Test
Positive
Negative
Total
Xpert MTB/RIF Ultra test
Ziehl–Neelsen stain microscopy
Culture on Lowenstein Jensen medium
306 (35.5%)
168 (19.5%)
299 (34.7%)
556 (64.5%)
694 (80.5%)
560 (65%)
862 (100%)
862 (100%)
859 (99.7%) *
* In 3 cases, the culture was contaminated.
Rifampicin resistance was encountered in 27 cases out of the 306 positive tests (8.82%);
in two cases, indeterminate rifampicin resistance was found (0.65%).
Out of the 306 patients with detected MTB in the molecular test, 284 (92.81%) had
a positive result in the bacterial culture, 20 had a negative culture and 2 samples were
contaminated. Of the 556 negative results in the molecular test, 15 had a positive culture.
Based on these data, the sensitivity of the Xpert MTB/RIF Ultra test, when compared to the
“gold standard” culture, was calculated to be 95%, while the specificity was determined to
be 96.4% (Figure 1).
Of the 168 positive result in the Ziehl–Neelsen stain microscopy, 164 (97.6%) had a
positive result in the culture, 3 (1.8%) had a negative culture, and 1 (0.6%) sample was
contaminated. Of the 694 negative microscopy result, 135 (19.5%) had a positive culture
and 2 (0.3%) were contaminated. Based on these data, the microscopy (Ziehl-Neelsen stain)
demonstrates a calculated sensitivity of 54.8% and a specificity of 99.5% (Figure 1).
Genes 2023, 14, 1231
6 of 10
Figure 1. Crosstabulation of molecular and microscopy test results compared with the bacterial
culture results.
Out of the 168 positive results of the microscopy, 166 had a positive molecular test,
and 2 samples were negative. In both cases, the culture was positive for mycobacteria other
than tuberculosis (MOTT).
In the 302 cases where the culture was not negative (299 positive samples and
3 contaminated samples), the median time at which the samples were declared positive
was 30 days (mean: 34.07 days, min: 21 days, max: 60 days). The majority (135, 44.7%) of
the samples were declared positive at the 21-day reading; 53 (17.5%) samples were declared
positive after 30 days; 65 (21.5%) were declared positive after 45 days; and 49 (16.2%) sam-
ples were declared positive at the 60-day reading. There is a statistically relevant correlation
(p < 0.0001), inversely related, between the duration of the positive determination and the
number of colonies isolated in the culture (Figure 2).
Figure 2. Number of days elapsed from the suspicion of TB until the diagnosis established by the
positive culture: indicators of the central tendency (mean: 34.07 days, median: 30 days, min: 21 days,
max: 60 days); these values also represent, as all the patients with positive molecular test were
immediately started on treatment, the days gained in the early treatment of TB using molecular tests
for the diagnosis.
Genes 2023, 14, x FOR PEER REVIEW 6 of 10 Rifampicin resistance was encountered in 27 cases out of the 306 positive tests (8.82%); in two cases, indeterminate rifampicin resistance was found (0.65%). Out of the 306 patients with detected MTB in the molecular test, 284 (92.81%) had a positive result in the bacterial culture, 20 had a negative culture and 2 samples were con-taminated. Of the 556 negative results in the molecular test, 15 had a positive culture. Based on these data, the sensitivity of the Xpert MTB/RIF Ultra test, when compared to the “gold standard” culture, was calculated to be 95%, while the specificity was deter-mined to be 96.4% (Figure 1). Figure 1. Crosstabulation of molecular and microscopy test results compared with the bacterial cul-ture results. Of the 168 positive result in the Ziehl–Neelsen stain microscopy, 164 (97.6%) had a positive result in the culture, 3 (1.8%) had a negative culture, and 1 (0.6%) sample was contaminated. Of the 694 negative microscopy result, 135 (19.5%) had a positive culture and 2 (0.3%) were contaminated. Based on these data, the microscopy (Ziehl-Neelsen stain) demonstrates a calculated sensitivity of 54.8% and a specificity of 99.5% (Figure 1). Out of the 168 positive results of the microscopy, 166 had a positive molecular test, and 2 samples were negative. In both cases, the culture was positive for mycobacteria other than tuberculosis (MOTT). In the 302 cases where the culture was not negative (299 positive samples and 3 con-taminated samples), the median time at which the samples were declared positive was 30 days (mean: 34.07 days, min: 21 days, max: 60 days). The majority (135, 44.7%) of the sam-ples were declared positive at the 21-day reading; 53 (17.5%) samples were declared pos-itive after 30 days; 65 (21.5%) were declared positive after 45 days; and 49 (16.2%) samples were declared positive at the 60-day reading. There is a statistically relevant correlation (p < 0.0001), inversely related, between the duration of the positive determination and the number of colonies isolated in the culture (Figure 2). Molecular positivetestsMolecularnegative testsMicroscopypositive testsMicroscopynegative testsPositive bacterial culture28415164135Negative bacterial culture205413557Contaminated sample20120100200300400500600Number of casesGenes 2023, 14, x FOR PEER REVIEW 7 of 10 Figure 2. Number of days elapsed from the suspicion of TB until the diagnosis established by the positive culture: indicators of the central tendency (mean: 34.07 days, median: 30 days, min: 21 days, max: 60 days); these values also represent, as all the patients with positive molecular test were im-mediately started on treatment, the days gained in the early treatment of TB using molecular tests for the diagnosis. 4. Discussion The early detection and prompt treatment of positive cases are the most effective measures in controlling the spread of tuberculosis [15]. The most reliable method of TB diagnostics is bacteriological culture, which is per-formed, in most cases, using sputum sampled directly, but other pathological products may also be used. The sampling process is essential in ensuring the quality of the result. The microscopic examination of the pathologic product is extremely relevant in the control of tuberculosis, helping to identify the patients with the highest contagion rate. This method aims to identify AFB in the pathologic product; the test is later confirmed via bacteriological culture. However, microscopic examination using the Ziehl–Neelsen stain-ing technique, although it is a fast, cheap method, has a low sensitivity, and it is not able to distinguish between MTBC and other non-tuberculosis mycobacteria [17]. For AFB to be detected, at least 104 CFU/mL must exist in the pathologic product [18]. Culture confir-mation of a TB infection may take 21 to 60 days. Furthermore, neither microscopic exam-ination nor culture can distinguish drug-susceptible TB strains from drug-resistant ones [12]. The testing using nucleic acid amplification tests offered quick and precise diagnosis of tuberculosis, with a sensitivity rate of 95% and a specificity rate of 96.4%. Using this test shortens the isolation period of suspected patients and prevents useless treatment [17,19]. The sputum samples with negative microscopic examination results but with a later positive culture had a lower bacterial load compared with the samples with positive mi-croscopic examination results. With high sensitivity, the NAAT method can detect MTB even in microscopic-negative samples. TB patients coinfected with HIV are known to have a low bacterial load compared with the patients without HIV, even though these patients, untreated, have a more ag-gressive form of the disease [16]. This study cohort did not include any HIV patients. The utilization of NAAT was initially approved in 1995 for patients with positive microscopic examination and clinical signs suggestive of TB [19]. The recent progress in molecular testing for MTBC includes the Xpert MTB/RIF Ultra test, which allows for the simultaneous detection of tuberculous bacilli and rifampicin resistance. Patients with neg-ative result following this test can avoid isolation, and those with positive results may benefit from early treatment [20]. Genes 2023, 14, 1231
7 of 10
4. Discussion
The early detection and prompt treatment of positive cases are the most effective
measures in controlling the spread of tuberculosis [15].
The most reliable method of TB diagnostics is bacteriological culture, which is per-
formed, in most cases, using sputum sampled directly, but other pathological products
may also be used. The sampling process is essential in ensuring the quality of the result.
The microscopic examination of the pathologic product is extremely relevant in the
control of tuberculosis, helping to identify the patients with the highest contagion rate.
This method aims to identify AFB in the pathologic product; the test is later confirmed
via bacteriological culture. However, microscopic examination using the Ziehl–Neelsen
staining technique, although it is a fast, cheap method, has a low sensitivity, and it is
not able to distinguish between MTBC and other non-tuberculosis mycobacteria [17].
For AFB to be detected, at least 104 CFU/mL must exist in the pathologic product [18].
Culture confirmation of a TB infection may take 21 to 60 days. Furthermore, neither
microscopic examination nor culture can distinguish drug-susceptible TB strains from
drug-resistant ones [12].
The testing using nucleic acid amplification tests offered quick and precise diagnosis
of tuberculosis, with a sensitivity rate of 95% and a specificity rate of 96.4%. Using this test
shortens the isolation period of suspected patients and prevents useless treatment [17,19].
The sputum samples with negative microscopic examination results but with a later
positive culture had a lower bacterial load compared with the samples with positive
microscopic examination results. With high sensitivity, the NAAT method can detect MTB
even in microscopic-negative samples.
TB patients coinfected with HIV are known to have a low bacterial load compared with
the patients without HIV, even though these patients, untreated, have a more aggressive
form of the disease [16]. This study cohort did not include any HIV patients.
The utilization of NAAT was initially approved in 1995 for patients with positive
microscopic examination and clinical signs suggestive of TB [19]. The recent progress
in molecular testing for MTBC includes the Xpert MTB/RIF Ultra test, which allows for
the simultaneous detection of tuberculous bacilli and rifampicin resistance. Patients with
negative result following this test can avoid isolation, and those with positive results may
benefit from early treatment [20].
The advantages of NAAT include the possibility of early diagnosis and the prompt
initiation of treatment, resulting a shorter period of contagion. Moreover, the quick dif-
ferentiation of patients with MTBC from those infected with non-tuberculosis mycobac-
teria prevents inadequate and useless treatments and useless investigations of patients’
families [21].
However, there are some limitations to molecular testing interpretations: these meth-
ods can have slightly lower sensitivity than bacterial cultures; a negative molecular result
does not absolutely exclude the diagnosis of tuberculosis. Furthermore, some sporadic
errors in the system may lead to false-positive results in molecular testing [21]. Conven-
tional microscopy and culture remain essential in the evaluation of disease response to
treatment [12].
In this study, we reported several situations in which we had a positive molecular test
using the Xpert MTB/RIF Ultra method that had a negative microscopic examination and
negative culture. We also reported situations with negative molecular testing and negative
microscopy but with positive culture.
As can be seen in Figure 3, molecular testing has superior sensitivity compared with
microscopic examination (95% compared with 54.8%). Regarding specificity, the two
methods (molecular and microscopic) had similar results (96.4% and 99.5%, respectively).
From this, it can be concluded that molecular testing has an important role in the early
diagnosis of TB.
In cases in which the Xpert MTB/RIF Ultra test was positive and the initial microscopy
was negative, the initialization of the treatment would have been delayed by an average
Genes 2023, 14, 1231
8 of 10
of 34.07 days. In these situations, molecular testing enables the prompt initialization
of treatment, which has an impact on the evolution of the disease and the spreading of
the disease. Similar results were reported in previous studies, such as Laraque et al. or
Luetkemeyer et al. [22,23], but they are slightly different from the CDC reports that cite
a 50–80% detection rate in the case of molecular tests performed on negative samples
following microscopic examination [24].
Figure 3. Sensitivity and specificity comparison between molecular testing and microscopic examina-
tion in the diagnosis of TB.
5. Conclusions
The results of the present study validate the recent WHO recommendations; the
implementation of molecular testing in TB laboratories leads to important increases in early
diagnostics, and has superior sensitivity and similar specificity to microscopic examination.
Molecular testing allows for a quicker diagnosis compared with bacterial culture
(90 min vs. weeks), which leads to prompter isolation and treatment of infected patients.
The capacity to distinguish between M. tuberculosis and non-tuberculosis mycobacteria
shortens the isolation period and prevents the unnecessary treatment of suspected patients.
Molecular testing can identify rifampicin-resistant strains (and other resistances),
allowing for a personalized approach to the treatment of TB patients.
Author Contributions: Conceptualization, C.S. and M.S.; methodology, A.B.B.; software, A.I.; val-
idation, C.S., M.S. and A.B.B.; formal analysis, C.P.M.; investigation, M.S. and A.-M.D.; resources,
C.S. and L.N.; data curation, A.I. and A.B.B.; writing—original draft preparation, C.S. and A.I.;
writing—review and editing, A.B.B. and L.N.; visualization, A.I. and C.P.M.; supervision, C.S.; project
administration, A.I. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: This study was conducted in accordance with the Declaration
of Helsinki and approved by the Ethics Committee of the Oradea County Emergency Clinical Hospital
(14613/27 April 2023).
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: Not applicable.
Acknowledgments: The article publishing fees were funded by the University of Oradea.
Conflicts of Interest: The authors declare no conflict of interest.
Genes 2023, 14, x FOR PEER REVIEW 8 of 10 The advantages of NAAT include the possibility of early diagnosis and the prompt initiation of treatment, resulting a shorter period of contagion. Moreover, the quick differ-entiation of patients with MTBC from those infected with non-tuberculosis mycobacteria prevents inadequate and useless treatments and useless investigations of patients’ fami-lies [21]. However, there are some limitations to molecular testing interpretations: these meth-ods can have slightly lower sensitivity than bacterial cultures; a negative molecular result does not absolutely exclude the diagnosis of tuberculosis. Furthermore, some sporadic errors in the system may lead to false-positive results in molecular testing [21]. Conven-tional microscopy and culture remain essential in the evaluation of disease response to treatment [12]. In this study, we reported several situations in which we had a positive molecular test using the Xpert MTB/RIF Ultra method that had a negative microscopic examination and negative culture. We also reported situations with negative molecular testing and negative microscopy but with positive culture. As can be seen in Figure 3, molecular testing has superior sensitivity compared with microscopic examination (95% compared with 54.8%). Regarding specificity, the two methods (molecular and microscopic) had similar results (96.4% and 99.5%, respectively). From this, it can be concluded that molecular testing has an important role in the early diagnosis of TB. Figure 3. Sensitivity and specificity comparison between molecular testing and microscopic exami-nation in the diagnosis of TB. In cases in which the Xpert MTB/RIF Ultra test was positive and the initial micros-copy was negative, the initialization of the treatment would have been delayed by an av-erage of 34.07 days. In these situations, molecular testing enables the prompt initialization of treatment, which has an impact on the evolution of the disease and the spreading of the disease. Similar results were reported in previous studies, such as Laraque et al. or Luet-kemeyer et al. [22,23], but they are slightly different from the CDC reports that cite a 50–80% detection rate in the case of molecular tests performed on negative samples following microscopic examination [24]. Genes 2023, 14, 1231
References
9 of 10
1. World Health Organization. Global Tuberculosis Report 2020. Available online: https://www.who.int/publications/i/item/9789
240013131 (accessed on 4 March 2022).
2. Houben, R.M.; Dodd, P.J. The Global Burden of Latent Tuberculosis Infection: A Re-estimation Using Mathematical Modelling.
PLoS Med. 2016, 13, e1002152. [CrossRef] [PubMed]
4.
5.
3. Wright, A.; Zignol, M.; Van Deun, A.; Falzon, D.; Gerdes, S.R.; Feldman, K.; Hoffner, S.; Drobniewski, F.; Barrera, L.;
van Soolingen, D.; et al. Epidemiology of antituberculosis drug resistance 2002–07: An updated analysis of the Global Project on
Anti-Tuberculosis Drug Resistance Surveillance. Lancet 2009, 373, 1861–1873. [CrossRef] [PubMed]
European Centre for Disease Prevention and Control/WHO Regional Office for Europe. Tuberculosis Surveillance and Monitoring
in Europe 2020–2018 Data. Available online: https://www.ecdc.europa.eu/en/publications-data/tuberculosis-surveillance-and-
monitoring-europe-2020-2018-data (accessed on 4 March 2022).
Analiza de Situatie TB 2019. Available online: https://insp.gov.ro/sites/cnepss/wp-content/uploads/2019/04/Analiza-de-
situatie-tb2019.pdf (accessed on 4 November 2021).
Chiotan, D.; Popa, C.G. Ghidul Pentru Identificarea Suspect,ilor TB s, i Referire a Pacient,ilor cu TB Pentru Tot,i Furnizorii de
Servicii (Inclusiv Identificarea Activă a Cazurilor de TB în Rândul Populat,iilor Vulnerabile. Available online: https://www.raa.
ro/wp-content/uploads/2018/05/Ghid-de-identificare-si-referire-a-persoanelor-suspecte-de-TB-2017-RO.pdf (accessed on
4 November 2021).
Uplekar, M.; Raviglione, M. WHO’s End TB Strategy: From stopping to ending the global TB epidemic. Indian J. Tuberc. 2015,
62, 196–199. [CrossRef] [PubMed]
Lönnroth, K.; Raviglione, M. The WHO’s new End TB Strategy in the post-2015 era of the Sustainable Development Goals. Trans.
R. Soc. Trop. Med. Hyg. 2016, 110, 148–150. [CrossRef] [PubMed]
7.
6.
8.
9. World Health Organization.
Implementing the End TB Strategy: The Essentials; World Health Organization: Geneva,
10.
Switzerland, 2022.
Sava, C.N.; Bodog, T.M.; Niulas, L.R.; Iuhas, A.R.; Marinau, C.P.; Negrut, N.; Balmos, A.B.; Pasca, B.; Roman, N.A.;
Nistor-Cseppento, C.D. Biomarker Changes in Pediatric Patients With COVID-19: A Retrospective Study from a Single Center
Database. Vivo 2022, 36, 2813–2822. [CrossRef] [PubMed]
11. Wu, C.W.; Wu, Y.K.; Lan, C.C.; Yang, M.C.; Dong, T.Q.; Tzeng, I.S.; Hsiao, S.S. Impact of nucleic acid amplification test on
pulmonary tuberculosis notifications and treatments in Taiwan: A 7-year single-center cohort study. BMC Infect. Dis. 2019, 19, 726.
[CrossRef] [PubMed]
12. World Health Organization. Module 3: Diagnosis-rapid diagnostics for tuberculosis detention. In WHO Operational Handbook on
13.
14.
Tuberculosis; World Health Organization: Geneva, Switzerland, 2021.
Steingart, K.R.; Schiller, I.; Horne, D.J.; Pai, M.; Boehme, C.C.; Dendukuri, N. Xpert® MTB/RIF assay for pulmonary tuberculosis
and rifampicin resistance in adults. Cochrane Database Syst. Rev. 2014, 2014, CD009593. [CrossRef] [PubMed]
Iuhas, A.; Jurca, C.; Kozma, K.; Riza, A.-L.; Streat,ă, I.; Petches, i, C.; Dan, A.; Sava, C.; Balmos, , A.; Marinău, C.; et al. PAH
Pathogenic Variants and Clinical Correlations in a Group of Hyperphenylalaninemia Patients from North-Western Romania.
Diagnostics 2023, 13, 1483. [CrossRef] [PubMed]
15. Ministerul Sănătăt,ii din România. Ghidul Metodologic de Implementare a Programului Na¸tional de Prevenire, Supraveghere ¸si
Control al Tuberculozei. Available online: http://old.ms.ro/documente/GHID%20Metodologic%20de%20implementare%20a%
20Programului%20national%20de%20prevenire,%20supraveghere%20si%20control%20al%20tuberculozei%202015_15424_183
33.pdf (accessed on 4 November 2021).
16. Xpert®MTB/RIF Assay. Available online: https://www.cepheid.com/en_US/package-inserts/1608 (accessed on 5 March 2022).
17. Bourgi, K.; Patel, J.; Samuel, L.; Kieca, A.; Johnson, L.; Alangaden, G. Clinical Impact of Nucleic Acid Amplification Testing in the
Diagnosis of Mycobacterium Tuberculosis: A 10-Year Longitudinal Study. Open Forum Infect. Dis. 2017, 4, ofx045. [CrossRef]
[PubMed]
18. Allen, B.W.; Mitchison, D.A. Counts of viable tubercle bacilli in sputum related to smear and culture gradings. Med. Lab. Sci.
1992, 49, 94–98. [PubMed]
19. Dinnes, J.; Deeks, J.; Kunst, H.; Gibson, A.; Cummins, E.; Waugh, N.; Drobniewski, F.; Lalvani, A. A systematic review of rapid
diagnostic tests for the detection of tuberculosis infection. Health Technol. Assess. 2007, 11, 1–196. [CrossRef] [PubMed]
20. World Health Organization. Automated real-time nucleic acid amplification technology for rapid and simultaneous detection of
tuberculosis and rifampicin resistance. In Xpert MTB/RIF Assay for the Diagnosis of Pulmonary and Extrapulmonary TB in Adults
and Children: Policy Update; World Health Organization: Geneva, Swizterland, 2013; Available online: http://apps.who.int/iris/
handle/10665/112472?locale=zh (accessed on 5 March 2022).
21. CDC. Report of an Expert Consultation on the Uses of Nucleic Acid Amplification Tests for the Diagnosis of Tuberculosis.
2012. Available online: https://www.cdc.gov/tb/publications/guidelines/amplification_tests/considerations.htm (accessed on
5 March 2022).
22. Laraque, F.; Griggs, A.; Slopen, M.; Munsiff, S.S. Performance of nucleic acid amplification tests for diagnosis of tuberculosis in a
large urban setting. Clin. Infect. Dis. 2009, 49, 46–54. [CrossRef] [PubMed]
Genes 2023, 14, 1231
10 of 10
23. Luetkemeyer, A.F.; Firnhaber, C.; Kendall, M.A.; Wu, X.; Mazurek, G.H.; Benator, D.A.; Arduino, R.; Fernandez, M.; Guy, E.;
Johnson, P.; et al. Evaluation of Xpert MTB/RIF Versus AFB Smear and Culture to Identify Pulmonary Tuberculosis in Patients
with Suspected Tuberculosis from Low and Higher Prevalence Settings. Clin. Infect. Dis. 2016, 62, 1081–1088. [CrossRef] [PubMed]
24. Centers for Disease Control and Prevention (CDC). Updated guidelines for the use of nucleic acid amplification tests in the
diagnosis of tuberculosis. Morb. Mortal. Wkly. Rep. 2009, 58, 7–10.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.
| null |
10.1088_1361-665x_ad1426.pdf
|
Data availability statement
All data that support the findings of this study are included
within the article (and any supplementary files).
|
Data availability statement All data that support the findings of this study are included within the article (and any supplementary files).
|
Smart Mater. Struct. 33 (2024) 015032 (11pp)
Smart Materials and Structures
https://doi.org/10.1088/1361-665X/ad1426
Research on variable stiffness
asymmetrical resonant linear
piezoelectric actuator based on
multi-modal drive
Liangguo He
and An Qian
∗, Xukang Yue, Haotian Dou, Xinfang Ge, Zhikai Wan
School of Mechanical Engineering, Hefei University of Technology, Hefei, Anhui 230009, People’s
Republic of China
E-mail: [email protected]
Received 24 May 2023, revised 27 November 2023
Accepted for publication 10 December 2023
Published 19 December 2023
Abstract
In this paper, a linear piezoelectric motor with variable stiffness and asymmetric resonance is
proposed, which is driven by a single harmonic signal. Working in the resonant state improve
the output performance of the motor. Motor control is relatively simple and can realize reverse
movement under the driving of second-order single harmonic signal. At the same time, the new
motor can obtain different operating speed and step distance by changing the clamping position
in front and back to meet the requirements of different loads and different working conditions
and has strong applicability. By experiment, the first-order optimal operating frequency of the
motor prototype at three different stiffness adjustment positions is 88 Hz, 90 Hz and 92 Hz
respectively. Under the excitation of 240 Vp–p first-order resonance signal, the corresponding
output speed of the motor prototype is 16.116 mm s
respectively, and the corresponding displacement resolution is 0.18 mm, 0.22 mm and 0.27 mm
respectively. When the stiffness adjustment positions is 2 mm, the maximum load of the motor
prototype reaches 450 g. The second-order optimal operating frequency at the stiffness
adjustment positions 1 mm is 601 Hz. Under the excitation of a 240 Vp–p second-order resonant
−1, and the corresponding
signal, the reverse output speed of the motor prototype is 13.126 mm s
displacement resolution is 0.02 mm.
−1 and 25.015 mm s
−1, 20.457 mm s
−1
Keywords: multi-modal drive, variable stiffness, asymmetrical, bidirectional motion,
resonant-type, inertial impact motor
1. Introduction
The piezoelectric motor uses the converse piezoelectric effect
of piezoelectric ceramics to input specific excitation signals to
piezoelectric ceramics and uses its high-frequency vibration as
the driving source to realize the directional movement of the
∗
Author to whom any correspondence should be addressed.
motor [1–3]. As the piezoelectric motor has the advantages of a
high motion resolution, small size can work in a complex elec-
tromagnetic interference environment, fast response speed,
and others. With these advantages, piezoelectric motors have
been widely used in aerospace, biomedicine [4, 5], robotics
[6–8], semiconductor industry [9], and other fields.
Piezoelectric motors have a variety of structures, but the
mainstream of the academic circle divides them into inchworm
piezoelectric motors [10–12], ultrasonic piezoelectric motors
1
© 2023 IOP Publishing Ltd
Smart Mater. Struct. 33 (2024) 015032
L He et al
[13, 14], and inertial impact piezoelectric motors [15–19]. The
inertia impact piezoelectric motor uses the inertia impact of
the stator vibration to generate micro displacement to realize
the linear motion or rotation of the mover [20, 21]. This motor
has the advantages of a large working frequency band, large
stroke, and simple structure [22, 23].
Inertial impact piezoelectric motors can be divided into
two types according to different driving mechanisms: elec-
tric signal control [24] and mechanism control [25]. Electronic
motors can be mainly divided into ‘stick-slip’ drive type [26]
and smooth impact type. Generally, both use asymmetric elec-
trical signals (sawtooth waves) to excite the inertial impact
block to generate asymmetric periodic motion to drive the
motor to achieve motion [27]. In 2019, Huang and Sun [28]
designed a piezoelectric brake based on the stick-slip prin-
ciple. When the input of the sawtooth wave excitation signal
−1, and
is 150 Hz and 50 Vp–p, the output speed is 1.8 mm s
the minimum step size is 0.875 µm. In 2019, Wei et al [29]
developed a sawtooth wave driven impact motor with an out-
−1. This kind of motor works in the
put speed of 2.41 mm s
quasi-static state, the load and output speed of the motor are
relatively low. In order to improve the output performance of
the motor, researchers use multi-channel harmonic signals to
synthesize resonant sawtooth wave waves based on the prin-
ciple of waveform synthesis. In 2021, Pan et al [30] proposed
a resonant piezoelectric motor used synthetic sawtooth wave,
−1. However, this
with a maximum no-load speed of 17.2 mm s
type of signal needs to be controlled cooperatively by multiple
signals, resulting in very complex drive control.
The other type of mechanism control inertial impact motor
adopts an asymmetric mechanical structure design to realize
the directional movement of the motor driven by symmetric
excitation signals, which is mainly divided into asymmetric
gripper inertia motor [31] and variable friction inertia motor
[32]. The asymmetric clamping inertial motor achieves dir-
ectional drive of the motor by changing the stiffness of the
reciprocating motion of the vibrator. In 2019, Shen et al [33]
proposed an inertial impact motor based on asymmetric mater-
ial clamping. They used two kinds of materials with differ-
ent elastic moduli, namely, copper and steel, to clamp the
vibrator to generate different inertial impact forces. The motor
is driven by a square wave, and under the excitation signal
−1. In
of 13 Hz and 25 Vp–p, the output speed is 0.72 mm s
2019, Zhang et al [34] designed a linear inertial piezoelec-
tric actuator using a single bimorph vibrator. They clamped
the two sides of the vibrator and fixed them with grippers
of different geometric sizes to obtain different stiffness. The
motor is excited by square waves to generate inertial impact
to drive motor movement. Under the excitation of 2 Hz and
−1, and the minimum
50 Vp–p, the output speed is 3.8 µm s
step distance is 1.9 µm. This kind of motor uses different
materials or different geometric dimensions of the clamp-
ing mechanism to produce stiffness differences to form an
inertial impact. The excitation signal does not need wave-
form matching and can make the drive control circuit sim-
pler. However, under the single vibrator, this kind of motor can
only achieve one-way movement. The traditional solution is to
2
arrange two elastic vibrators in a mirror image to achieve bid-
irectional motion, which will make the motor structure more
complex.
Based on the advantages and disadvantages of the above
kinds of motors, the present study designed a linear piezo-
electric motor with variable stiffness and asymmetric struc-
ture which can work in a resonant state. The motor is driven
by a single harmonic signal and operates in a resonant state,
providing better output performance. At the same time, the fre-
quency of the excitation signal can be adjusted to excite the
second-order vibration mode of the elastic vibrator, achiev-
ing reverse motion of the motor using a single vibrator in this
mode. In addition, by adjusting stiffness-regulating mechan-
ism to change the stiffness difference on both sides of the
elastic vibrator, the motor can achieve different displacement
resolutions. Section 2 introduces the working principle of its
bidirectional motion and the overall assembly of the motor,
and conducts modal analysis using finite element software to
obtain the first and second resonant frequencies corresponding
to the elastic vibrator. Established a motor dynamics model,
solved the dynamic equations using MATLAB/Simulink, and
obtained the theoretical displacement curve of the motor. In
section 3, a test bench was built to test the first-order and
second-order motion characteristics of the motor under differ-
ent adjustment stiffness positions, and a comprehensive per-
formance comparison was conducted with the inertial impact
motor developed in recent years. Finally, section 4 summarizes
the research.
2. Structure and working principle
2.1. Working principle
Figure 1 shows the motor movement process after applying
the harmonic excitation signal. Specifically, figure 1(a) shows
a motion period under the excitation of the first-order resonant
signal, which can be divided into the following steps:
Step 1: At time t0, as shown in ‹. The elastic vibrator main-
tains its limit position of swinging to the right under the peak
voltage. At this time, the motor is in the initial position and the
displacement is 0;
Step 2: From t0 to t1, as show in marking (a) ›. The elastic
vibrator rapidly swings from the right limit to the left limit
position. As the effective clamping length of the motor’s front
block for the elastic vibrator is short, its clamping stiffness is
small, so the amplitude of the elastic vibrator swings to the
left is large. Therefore, the inertia impact force generated by
the motor to the right is relatively large, and the motor will
generate a large displacement SR0 to the right;
Step 3: At time t1, as show in marking (a) fi. The elastic
vibrator maintains the leftmost limit position under the negat-
ive peak voltage. The motor is stationary;
Step 4: From t1 to t2, as show in marking (a) fl. The
elastic vibrator rapidly swings from the leftmost to the right-
most limit position under the electric signal excitation. As the
effective clamping length of the motor’s rear block for the
elastic vibrator is large, its clamping stiffness is large, so the
Smart Mater. Struct. 33 (2024) 015032
L He et al
Figure 1. The working processes of the motor: (a) vibration mode
of the piezo-driven vibrator in forwarding motion and (b) vibration
mode of the piezo-driven vibrator in reverse motion.
amplitude of the elastic vibrator swings to the right is small.
Therefore, the inertial impact force generated by the piezo-
electric motor to the left is relatively small, and the motor will
retreat to the left by a small displacement SR1;
At this point, the motor completes the movement of one
working cycle, and the motor generates a tiny displacement
△S (+x), whose length is SR0–SR1. When the continuous peri-
odic excitation signal is input to the motor, the motor can
obtain a macro displacement to the right by continuously accu-
mulating the tiny displacement generated in each cycle, so as
to realize the positive motion of the motor.
Figure 1(b) shows the motor movement diagram of one
cycle under the action of a second-order excitation signal,
which can be divided into the following steps:
Step 1: At the time of t0, as show in marking (b) ‹. The
middle part of elastic vibrator remains at the limit position of
bending to the right under the peak voltage. The mass block is
located at the left extreme position. At this time, the motor is
in the initial position and the displacement is 0;
Step 2: From t0 to t1, as show in marking (b) ›. The middle
part of elastic vibrator of the motor rapidly bends to the left
from the rightmost limit to the leftmost limit position and the
mass block swings to the right extreme position. Due to the
short effective clamping length of the front block on the elastic
vibrator, its clamping stiffness is small, so the curvature of the
elastic vibrator is larger when it bends to the left. The inertia
impact force generated by the motor to the left is relatively
large, and the motor will generate a large displacement SL0 to
the left;
Step 3: At time t1, as show in marking (b) fi. The middle
part of elastic vibrator remains at the limit position of bending
Figure 2. Diagram of the prototype piezoelectric motor.
to the left under the negative peak voltage. The mass block is
located at the right extreme position. The motor is stationary;
Step 4: From t1 to t2, as show in marking (b) fl. The middle
part of elastic vibrator of the motor quickly bends to the right
from the leftmost limit to the rightmost limit position and the
mass block swings to the left extreme position. As the effect-
ive clamping length of the motor’s rear block for the elastic
vibrator is large, its clamping stiffness large, so the curvature
of the elastic vibrator bends to the right is small. The inertia
impact force generated by the motor to the right is relatively
small, and the motor will retreat to the right by a small dis-
placement SL1;
At this point, the motor completes the movement of one
working cycle, and the motor generates a tiny displacement
△SL (−x), whose length is SL0–SL1. When the continuous
periodic excitation signal is input to the motor, the motor can
obtain a macro displacement to the left by continuously accu-
mulating the tiny displacement generated in each cycle, so as
to realize the reverse motion of the motor.
2.2. Structure design
Figure 2 shows the 3D assembly drawing of the motor. The
motor is composed of a composite actuator base, an inertial
impact mechanism, and a stiffness-regulating mechanism. The
inertial impact mechanism is composed of an inertial impact
mass, a 65-Mn elastic vibrator, and a piezoelectric single chip.
Using conductive silver glue and epoxy adhesive, the piezo-
electric ceramic plate is pasted on the surface of the elastic
vibrator, and the inertial impact mass is symmetrically bon-
ded to both sides of the free end of the elastic vibrator.
The inertial impact mechanism is clamped and fixed on
the base of the composite actuator by a clamping mechan-
ism consisting of a front block and a clamping block. The rear
block composed of the left and right blocks is symmetrically
arranged on both sides of the central axis of the composite
movable base. It can move around along the length direction of
the elastic vibrator to realize the role of changing the clamping
stiffness.
The stiffness-regulating mechanism is composed of a
regulating wedge block, a regulating screw, a regulating
3
Smart Mater. Struct. 33 (2024) 015032
L He et al
Figure 3. Motor meshing.
spring, and the rear block. The regulating wedge block is
installed in the base through the dovetail groove, and the two
wedge surfaces are respectively contacted to the rear block,
which is pushed to move the same displacement to both sides
by rotating the regulating screw. As shown in the right sub-
graph of figure 2, the system adjusts the clamping stiffness
by the position of the regulating wedge block. Specifically,
when the regulating wedge block is pushed in, the rear block
is opened and the clamping stiffness difference on both sides
of the elastic vibrator increases; when the regulating wedge
block is pulled out, the rear block contracts and the clamp-
ing stiffness difference on both sides of the elastic vibrator
decreases. When the rear block is adjusted to the appropri-
ate clamping position, the screw between the rear block and
the base for fixing is tightened so that the clamping position
remains unchanged to ensure that the motor work asymmetric
stiffness remains unchanged.
2.3. Motor modal simulation
As the motor works in the resonant state, the modal analysis
of the motor is needed to determine the resonant frequency of
the motor. The modal analysis of the motor is conducted using
Hypermesh combined with ANSYS Workbench. The motor
uses the hexahedral mesh as a whole and controls the mesh
angle at key positions between 45
. The mesh aspect
ratio is less than 1:6 to reduce the influence of mesh errors
on the calculation results. To ensure the mesh quality, some
threaded holes in the motor are simplified. Figure 3 depicts
the mesh division of the motor.
and 135
◦
◦
PZT-4 material is used for piezoelectric ceramic plate, a
65-Mn steel material is used for an elastic vibrator, and the
other parts are 45 structured steels. Figure 4 shows the modal
analysis results under asymmetric clamping when the stiff-
ness adjustment mechanism adjusts the position to 0 mm. The
first- and second-order resonant frequencies of the motor are
89.61 Hz and 603.3 Hz, respectively. Table 1 shows the modal
simulation results of the motor at different stiffness adjustment
positions.
Based on the vibration cloud diagram in figure 4, the first
modal of the piezoelectric actuator is that it oscillates around
the fixed end. At this time, the mass at the free end has a
large displacement, so it can generate a good inertial impact
Figure 4. Motor modal simulation: (a) first-order modal shape of an
elastic motor vibrator and (b) second-order modal shape of elastic
motor vibrator.
Table 1. The modal simulation results of the motor at different
stiffness adjustment positions.
N (mm)
0
1
2
First-order modal (Hz)
Second-order modal (Hz)
89.61
603.3
91.18
616.2
93.12
629.5
Table 2. Structural parameters of main components of piezoelectric
motor.
Name
Size (mm)
Material
Elastic vibrator
Mass block
Composite actuator base
Piezoelectric chip
54
11
52
25
∗
∗
∗
∗
∗
6
∗
∗
11
∗
8
39
11
0.6
12
0.4
65Mn
45#
45#
PZT-4
force and drive the motor to move in one direction. The second
modal of the motor is the bending vibration of the piezoelec-
tric actuator around the fixed end, and the mass block at the
free end will also produce an inertial impact. If the position
of the mass block in the stationary state is taken as a refer-
ence, the vibration direction of the mass block is opposite to
that in the first mode state. Therefore, according to the calcu-
lation results, the motor can realize reverse motion. The finite
element analysis results show that the designed structure con-
forms to the expected working principle. Table 2 presents the
main structure dimensions based on the simulation results.
2.4. Analysis of motor dynamics model
As shown in figure 5(a), the simplified schematic structure of
the elastic vibrator with the front and rear block can be viewed
4
Smart Mater. Struct. 33 (2024) 015032
L He et al
Figure 5. Motor dynamics analysis: (a) dynamic model of the motor (b) calculation parameters of elastic vibrator.
as a spring-damping system. The dynamic equation of an elec-
tric motor can be written as
{
{
m1¨x1 + c˙x1
m2¨x2 + c˙x2
− c˙x2 + kx1
− c˙x1 + kx2
− kx2 = F
− kx1 = 0
(x ⩾ 0)
(1)
m1¨x1 + c
m2¨x2 + c
′˙x1 − c
′˙x2 − c
′˙x2 + k
′˙x1 + k
′x1 − k
′x2 − k
′x2 = F
′x1 = 0
(x ⩾ 0) .
(2)
Among them, F is the driving force of the piezoelectric
ceramic plate, m1 is the total mass of the elastic vibrator, piezo-
electric plate, and mass block, m2 is the total mass of the com-
are equivalent damping
posite base and block, c, k and c
and equivalent stiffness, which can be calculated according to
the following equation:
, k
′
′
Ki =
kik0
ki + k0
c = ζωnm
(3)
(4)
where, k0 is the stiffness of the elastic vibrator, k1 is the stiff-
ness of the front stop block, k2 is the stiffness of the rear stop
block, c is the damping coefficient, ζ is the damping ratio, and
ωn is the natural frequency. k and ki can be further written as,
(
· a ·
k =
1
4
ki =
ab3Em
4l3
i
Est2
s + 6t2
s Eptp + 12tsEpt2
(l − li)3
p + 8Ept3
p
)
(5)
(6)
where a is the section width of the clamping block, b is the
section length of the clamping block, li is the length of the
clamping block, l is the length of the elastic vibrator, ts is the
thickness of the elastic vibrator, tp is the thickness of the piezo-
electric ceramic plate (show in figure 5(b)), Es is the elastic
modulus of the elastic vibrator, and Ep is the elastic modulus of
the piezoelectric ceramic plate. Equivalent external force Fi:
5
Fli =
3a(ts + 2tp)2Ep
8 (l − li)
ts
tp
+ 1
)
·
(
tm
2tp
+ 1
· U
tp
2
· d
(7)
where, d is piezoelectric constant and U is excitation voltage.
Because the equivalent clamping lengths li on both sides
of the elastic vibrator are different, therefore the existence of
the clamping stiffness difference between the two sides of the
elastic vibrator, the impact force generated during the oscil-
lation is different. From the clamping difference, the impact
force difference between the two sides can be deduced:
Fd = Fl1
− Fl2.
(8)
Depending on the impact force difference Fd, the direc-
tional motion of the motor can be realized.
Due to the complexity of the dynamic model, a calculation
module was built in MATLAB/Simulink to calculate and solve
the dynamic differential equations, as shown in figure 6.
Under first-order modal signal excitation, the displacement
output curve of the motor under different stiffness differences
is shown in figure 7. When the stiffness adjustment position
is 0 mm, according to equations (7) and (8), the equival-
ent external force difference generated on both sides of the
elastic vibrator is the smallest. The characteristics reflected on
the displacement curve is that the motor will produce a large
backsliding in each motion cycle, resulting in a lower output
speed and larger vibration of the motor, but at the same time,
a small displacement resolution is obtained. As the differ-
ence in clamping stiffness between the two sides of the elastic
vibrator increases, the difference in inertial impact force Fd
obtained by the motor increases. The characteristics reflec-
ted on the displacement curve is that the backsliding during
motor operation decreases, resulting in higher output speed
and smoother movement of the motor. For example, when the
stiffness adjustment position is 2 mm, the peak step of the
motor during each motion cycle is approximately the same
as that at other stiffness adjustment positions (0 mm, 1 mm).
However, due to the decrease in backsliding, the accumulation
of micro displacement of the motor is faster, therefore the
motor obtains higher speed.
Smart Mater. Struct. 33 (2024) 015032
L He et al
Figure 6. Building the solution module for differential equations in simulink.
Figure 7. Theoretical displacement output curve under the
first-order modal.
3. Experimental testing and analysis
3.1. Experimental setup
Figure 8 shows the experimental device of the piezoelectric
motor prototype. A signal generator (Rigol DG1022) is used
to output the harmonic excitation signal, which is connected
to an oscilloscope to monitor the excitation signal. Moreover,
the harmonic signal is amplified by a power amplifier (200×
amplification), and the amplified signal is connected to the
motor prototype as the excitation signal of the piezoelectric
ceramic plate. During the experiment, the motion state data
of the piezoelectric motor prototype are captured by using a
laser Doppler vibrometer system (Neoark Corp.MLD221D,
Japan). By connecting the laser vibrometer to a computer, the
motor movement data can be displayed in real time on the
computer.
3.2. Admittance characteristics of the motor
Before testing the prototype, LCR impedance analyzer should
be used (LCR-8101, Taiwan Guwei Electronic Industry Co.,
Ltd) to measure the admittance characteristics of the motor
prototype. Perform frequency sweep analysis in the frequency
range of 0–1100 Hz. Figure 9 shows the admittance character-
istic curve of the piezoelectric actuator of the motor prototype.
According to the figure, when the stiffness adjustment position
is 0 mm, 1 mm, and 2 mm, the optimal vibration frequencies of
the motor piezoelectric actuator are approximately 88.63 Hz,
Figure 8. Measurement system of the prototype motor.
90.07 Hz, and 91.96 Hz in the first-order vibration state and
601.15 Hz, 602.52 Hz, and 603.39 Hz in the second-order
vibration state. Table 3 shows the optimal vibration frequency
of the motor at different clamping stiffness positions obtained
from the admittance characteristic curve.
3.3. Motor speed characteristics
Before studying the first-order directional motion speed char-
acteristics of the motor, it is necessary to determine the first-
order optimal operating frequency of the piezoelectric motor
at three different stiffness adjustment positions. In the previ-
ous section the first-order resonant frequency of the motor was
88.63 Hz. Therefore, the frequency range from 83 Hz to 98 Hz
was chosen for testing.
6
Smart Mater. Struct. 33 (2024) 015032
L He et al
Figure 9. Admittance characteristics of the piezoelectric actuator.
Table 3. The optimal vibration frequency of the motor at different
stiffness adjustment positions.
N (mm)
0
1
2
First-order modal (Hz)
Second-order modal (Hz)
88.63
601.15
90.07
602.52
91.96
603.39
Figure 11. Velocity vs frequency curves at different clamping
positions.
Figure 10. Rear block positioning marks of the motor prototype.
During the testing process, it is necessary to rotate the
adjusting bolt to change the stiffness adjustment position. In
order to ensure the accuracy of position adjustment, it is neces-
sary to mark the adjustment positions of ‘0 mm’, ‘1 mm’, and
‘2 mm’ on the piezoelectric bimorph metal elastic vibrator, as
shown in figure 10.
When the input excitation signal voltage is 240 Vp–p, the
relationship between the output speed of the piezoelectric
motor and frequency at different stiffness adjustment positions
is shown in figure 11. At three different stiffness adjustment
positions, the output speed of the piezoelectric motor shows a
trend of first increasing and then decreasing with the increase
of input excitation signal frequency. At the same time, when
the stiffness adjustment position is 0 mm, 1 mm, and 2 mm, the
optimal operating frequencies of the motor are measured to be
88 Hz, 90 Hz, and 92 Hz, respectively, and the corresponding
−1, and
motor output speeds are 16.116 mm s
25.015 mm s
−1, 20.457 mm s
−1, respectively.
From the experimental data, it can be seen that as the dif-
ference in clamping stiffness between the two sides of the
elastic vibrator increases, the effective clamping length of the
elastic vibrator increases, the vibration length l − li of the
7
Figure 12. Speed curve at different positions and voltages under
respective optimal operating frequencies.
beam decreases, and the resonant frequency of the motor also
increases. Similar to the theoretical model calculation results,
the motor achieves its maximum output speed at the stiff-
ness adjustment position of 2 mm, at which point the differ-
ence in inertial impact force Fd generated on both sides of the
elastic vibrator is the largest. The smaller backsliding allows
the motor to accumulate displacement more quickly, exhibit-
ing a higher output speed at the macro level.
The relationship between the output speed and excitation
signal voltage of the motor at different stiffness adjustment
positions with the input excitation signal frequency being the
first-order optimal operating frequency is tested. As shown
in figure 12, under the optimal operating frequency corres-
ponding to different stiffness adjustment positions, the output
speed of the motor increases with the increase of voltage. This
is due to the linear positive correlation between the driving
force generated by the piezoelectric sheet and the excitation
signal voltage. At different positions, the correlation coeffi-
cients between output speed and voltage are R0mm = 0.9847
R1mm = 0.9914 and R1mm = 0.9942, respectively. Prove that
the motor has good voltage controllability. In addition, under
the same voltage, as the difference in clamping stiffness
increases, the output speed of the motor increases. When the
rear block of the motor is located at the 2 mm mark and the
Smart Mater. Struct. 33 (2024) 015032
L He et al
Figure 13. Characteristic curves of motor step at different positions.
excitation signal voltage is 240 Vp–p, the motor speed has
reached 24.982 mm s
−1.
3.4. Motor step characteristics
When the motor prototype is at the first-order optimal oper-
ating frequency, measure the displacement step characterist-
ics of piezoelectric motors corresponding to three different
stiffness adjustment positions by using the laser displacement
sensor. When the voltage is 240 Vp–p, The total displace-
ment of the piezoelectric motor corresponding to three differ-
ent stiffness adjustment positions is 3.96 mm, 4.95 mm and
6.21 mm in 0.25 s, respectively, as shown in figure 13. The
displacement resolution of the motor is 0.18 mm, 0.22 mm
and 0.27 mm.
According to the experimental results, as the difference
in clamping stiffness increases, the micro explanation for the
increase in macroscopic output speed of the motor is the step
size of the motor increased, which is also consistent with
the theoretical calculation results. Due to the neglect of the
frictional effect between the composite base and the moving
plane in theoretical calculations, the step and backsliding in the
experiment are smaller than those in theoretical calculations.
As the stiffness difference increases, the amount of backslid-
ing during each motion cycle of the motor will also decrease,
which can be used as another method to reduce the backslid-
ing during the working cycle of the inertia impact motor in
addition to changing the friction force of the working surface.
3.5. Motor load characteristics
When the motor is under the optimal operating frequency, the
load characteristics are measured by applying external load
to the motor. Figure 14 shows the experimental diagram for
measuring the load characteristics of the motor. The load per-
formance of the motor is evaluated by the maximum load
capacity.
Figure 15 shows the corresponding load characteristic
curves of the piezoelectric motor at three different stiffness
adjustment positions. During the testing process, the input
excitation signal voltage of the piezoelectric motor is 240 Vp–p,
and the input excitation signal frequency is the first-order
Figure 14. Load test working principal diagram.
Figure 15. The speed of the motor under different loads.
optimal operating frequency corresponding to the different
stiffness adjustment positions. From the figure, it can be seen
that as the load increases, the output speed of the corres-
ponding piezoelectric motor at three different stiffness adjust-
ment positions decreases. When the load increases from 0 g
to 450 g, the output speed of the piezoelectric motor gradually
decreases, because as the load increases, the motor produces a
greater backsliding.
When the load is 120 g, the corresponding output speeds
of the piezoelectric motor at three different stiffness adjust-
−1, and
ment positions are 3.432 mm s
−1, respectively, indicating that as the stiffness
20.897 mm s
adjustment position increases, the maximum load capacity of
the piezoelectric motor also increases. When the motor stiff-
ness adjustment position is 2 mm, its maximum load capacity
can reach 450 g.
−1, 13.856 mm s
3.6. Second-order motion characteristics of motors
Select one of the stiffness adjustment positions (the stiffness
adjustment position of 1 mm) to measure the motion charac-
teristics of the motor in reverse motion under second-order
vibration. Refer to the motor admittance characteristic curve
and measure the output speed within the frequency range of
596–610 Hz. When the excitation signal voltage is 240 Vp–p,
the frequency and speed characteristic curves of the motor are
measured, as shown in figure 16. When the excitation signal
8
Smart Mater. Struct. 33 (2024) 015032
L He et al
Figure 16. Velocity vs frequency characteristic curves of second-order state.
Figure 17. Second-order state motor step characteristic curve.
Table 4. Comparison with previous piezoelectric motors.
Motor of Shen et al [33]
Motor of Hu et al [35]
Motor of Li et al [19]
Motor in this paper
−1)
Type
Max. speed (mm s
Max. load (N)
Displacement resolution (µm)
Control signal
Bidirectional
Structural complexity
Non-resonant
0.72
0.882
15.7
Simple
No
Simple
Non-resonant
0.0376
—
0.35
Simple
Yes
Complex
Resonant
125.43
0.5
0.037
Complex
Yes
Complex
Resonant
25.015
4.41
20.0
Simple
Yes
Simple
frequency is 601 Hz, the motor is in the best working condi-
tion. At this time, the maximum reverse speed of the motor
prototype reaches 13.126 mm s
−1.
When the excitation signal voltage is 240 Vp–p and the fre-
quency is 601 Hz, the motor step characteristics are measured.
As shown in figure 17, the displacement generated is 0.24 mm
within 12 movement cycles, and the maximum displacement
resolution of the motor is 0.02 mm.
3.7. Experimental result
Table 4 shows a comparison with other piezoelectric motors of
the same type. Among them are recently published piezoelec-
tric motors driven by the inertial impact principle. Compared
with other motors operating under non-resonant conditions
(Shen et al [33] and Hu et al [35]), the motor in this paper is
driven by resonant signals and operating in resonant state, so
it is used for higher output speed. The motor of Shen et al [33]
has a lower speed, smaller displacement resolution and can
only move in one direction. The motor of Hu et al [35] uses
two driving vibrators to realize bidirectional motor movement,
which will lead to a more complex motor structure and control
circuit. The motor of Li et al [19] is driven by a resonant mech-
anical approximate sawtooth wave. This synthesis of such sig-
nals is difficult, and multichannel harmonic signals are usu-
ally required for synthesis. To realize bidirectional movement,
multichannel signals need to be adjusted simultaneously, so
the signal control circuit will be too complex.
Through the above comprehensive comparison, the motor
prototype in this study adopts the way of multi-modal drive
and adjusting the clamping stiffness, which overcomes the
defects of the traditional inertial impact motor working in
the quasi-static state, such as low output speed, complex bid-
irectional motion structure, and complex control circuit. The
9
Smart Mater. Struct. 33 (2024) 015032
L He et al
motor has the following characteristics: its structure is simple,
its volume is small, and it is driven by a single harmonic sig-
nal. By using a single elastic vibrator for multimodal driv-
ing, the motor can achieve bidirectional operation. In addition,
the stiffness adjustment mechanism can change the speed, dis-
placement resolution, load and other motion characteristics of
the motor to meet the needs of different working conditions,
and has strong applicability.
4. Conclusion
inertial
impact
A variable stiffness asymmetric resonant
linear piezoelectric motor with variable clamping stiffness
and reverse motion excited by second-order vibration was
designed. The first-order optimal operating frequencies of
the piezoelectric motor prototype at three stiffness adjust-
ment positions are 88 Hz, 90 Hz, and 92 Hz, respectively.
Under the excitation of 240 Vp–p first-order resonant signal,
−1,
the output speeds of the motor prototype are 16.116 mm s
−1, respectively, and the cor-
20.457 mm s
responding displacement resolutions are 0.18 mm, 0.22 mm,
and 0.27 mm, respectively. The maximum load of the piezo-
electric motor is 450 g when the stiffness adjustment position
is 2 mm. The optimal second-order resonant frequency of the
motor prototype at one of the stiffness adjustment positions
is 601 Hz. Under the excitation of a 240 Vp–p second-order
resonant signal, the reverse output speed of the motor pro-
−1, and the displacement resolution is
totype is 13.126 mm s
0.02 mm.
−1, and 25.015 mm s
Theoretical calculations and experimental results indicate
that changes in clamping stiffness directly affect the vari-
ous performance of the motor. Overall, as the difference in
clamping stiffness between the two sides of the elastic vibrator
increases, the resonance frequency, output speed, and load
capacity of the motor all increase, meanwhile the backslid-
ing phenomenon of the motor in each motion cycle decreases.
This is due to the motor increases the clamping stiffness dif-
ference by enlarging the effective clamping length, which
reduces the effective vibration length of the elastic vibrator
(l − li) and leads to an increase in its resonant frequency.
Meanwhile, the increase in the clamping stiffness difference
will also increase the inertial impact force difference on both
sides, which is manifested in the specific performance of the
motor as an increase in output speed, an improvement in
load capacity, and a decrease in backsliding. The decrease
in stiffness difference has opposite effects on the above
performance.
In the future, we will further improve and optimize the
structure of the motor and find the optimal friction condition
of the contact surface of the composite actuator base when
the motor works to further improve the motor performance.
Meanwhile, we will further optimize the design of the motor to
improve its displacement resolution, study the impact of high
voltage and wear on the lifespan of the motor, and take cor-
responding measures to improve its output performance and
further enhance its practicality.
10
Data availability statement
All data that support the findings of this study are included
within the article (and any supplementary files).
Acknowledgments
This work was financially supported by the Project of the
Natural Science Foundation of Anhui Province of China (No.
2008085ME154), and the National Natural Science Fund of
China (No. 51405127).
ORCID iDs
Liangguo He https://orcid.org/0000-0001-9739-2407
Xukang Yue https://orcid.org/0000-0002-9365-5502
References
[1] Ma X, Liu Y, Deng J, Zhang S and Liu J 2020 A walker-pusher
inchworm actuator driven by two piezoelectric stacks Mech.
Syst. Signal Process. 169 108636
[2] Liang W, Viktor H, Bai F, Jin J and Jens T 2018 Modeling of
coupled longitudinal and bending vibrations in a sandwich
type piezoelectric transducer utilizing the transfer matrix
method Mech. Syst. Signal Process. 108 216–37
[3] Mohith S, Mohith S, Upadhya A R, Navin K P, Kulkarni S M
and Rao M 2021 Recent trends in piezoelectric actuators for
precision motion and their applications: a review Smart
Mater. Struct. 30 013002
[4] Wang P and Xu Q 2018 Design and testing of a flexure-based
constant-force stage for biological cell micromanipulation
IEEE Trans. Autom. Sci. Eng. 15 1114–26
[5] Wang G and Xu Q 2017 Design and precision position/force
control of a piezo-driven microinjection system
IEEE/ASME Trans. Mechatronics 22 1744–54
[6] Ho T and Lee S 2009 Piezoelectrically actuated biomimetic
self-contained quadruped bounding robot J. Bio. Eng.
6 29–36
[7] He Z, Loh H T and Ong E H 2008 Reliability evaluation of
piezoelectric micro-actuators with application in hard disk
drives IEEE Trans. Magnet 44 3722–5
[8] Dharmawan A G, Hariri H H, Foong S, Soh G S and Wood K L
2017 Steerable miniature legged robot driven by a single
piezoelectric bending unimorph actuator 2017 IEEE Int.
Conf. on Robotics and Automation (ICRA) pp 6008–13
[9] Choi J, Shin M, Rudy R Q, Kao C, Pulskamp J S,
Polcawich R G and Oldham K R 2017 Thin-film
piezoelectric and high-aspect ratio polymer leg mechanisms
for millimeter-scale robotics Int. J. Intell. Robot. Appl.
1 180–94
[10] Gao Y, Wen J, Ma J, Zhang Y, Wang R, Hu Y and Li J 2019 A
self-adapting linear inchworm piezoelectric actuator based
on a permanent magnet clamping structure Mech. Syst.
Signal Process. 132 429–40
[11] Wang R, Hu Y, Shen D, Ma J, Li J and Wen J 2020 A novel
piezoelectric inchworm actuator driven by one channel
direct current signal IEEE Trans. Ind. Electron. 68 2015–23
[12] Sun D, Tang Y, Wang J and Wang X 2020 Design of an
H-shaped linear piezoelectric motor for safety and arming
device Sens. Actuators A 303 111687
[13] Ma X, Liu J, Deng J, Liu Q and Liu Y 2020 A rotary traveling
wave ultrasonic motor with four groups of nested PZT
Smart Mater. Struct. 33 (2024) 015032
L He et al
ceramics: design and performance evaluation IEEE Trans.
Ultrason. Ferroelectr. Freq. 67 1462–9
actuator constructed on four bimorphs Mech. Syst. Signal
Process. 149 107213
[14] Chen N and Fan D 2020 A teeth-discretized electromechanical
model of a traveling-wave ultrasonic motor Mech. Sci.
11 257–66
[15] Zhang B, Chen F, Li H, Du Z, Sun L and Dong W 2018 A
friction-inertial-based rotary motor: design, modelling and
experiments Materials 11 918
[16] Pan Q, Zhao M, Wang Y, Li C, Huang Q, Huang B and Li R
2022 Design, simulation, and motion characteristics of a
novel impact piezoelectric actuator using double stators
Smart Mater. Struct. 31 075012
[17] Wang Y, Xu M, Shao S, Song S and Shao Y 2020 A novel
stick-slip type rotary piezoelectric actuator Adv. Mater. Sci.
Eng. 2020 1–11
[18] Li J, Zhang S, Liu Y, Deng J and Ma X 2022 A 3-DOF inertial
impact locomotion robot constructed on four piezoelectric
bimorph actuators Smart Mater. Struct. 31 095008
[19] Li C, Li R, Zhao M, Wang Y, Wei Y, Huang Q and Pan Q 2022
Development of a high-resolution, high-speed impact
piezoelectric actuator using cross-frequency band method
Smart Mater. Struct. 31 055005
[20] Dalius M and Piotr V 2013 Linear inertial piezoelectric
motor with bimorph disc Mech. Syst. Signal Process.
36 110–7
[25] Hu Y, Ma J, Zhang Y, Li J, Hu Y and Wen J 2021 Performance
comparison of two motion modes of a piezoelectric inertial
linear motor and its potential application in cell
manipulation Mech. Syst. Signal Process. 157 107743
[26] Wang L, Wang H and Cheng T 2022 Design and performance
of a compact stick-slip type piezoelectric actuator based on
right triangle flexible stator Smart Mater. Struct. 31 055013
[27] Dong J, Zhang B, Li X, Xu Z, Wang J, Liu C and Cao Y 2021
A stick-slip piezoelectric actuator with suppressed
backward motion achieved using an active locking
mechanism (ALM) Smart Mater. Struct. 30 095015
[28] Huang W and Sun M 2019 Design, analysis, and experiment
on a novel stick-slip piezoelectric actuator with a lever
mechanism Micromachines 10 863
[29] Wei J, Sergej F, Li H and Zhang X 2019 Design and waveform
assessment of a flexible-structure-based inertia-drive motor
Micromachines 10 771
[30] Pan C, Shi C, Feng A, Hu M, Shu S and Xia H 2021
Resonant-type piezoelectric inertial drive mechanism with
asymmetric inertial masses Rev. Sci. Instrum. 92 015003
[31] Ma J, Wen J, Cheng G and Zeng P 2016 A piezoelectric inertia
rotary actuator based on asymmetric clamping mechanism
Microsyst. Technol. 22 1975–81
[21] Wang S, Rong W, Wang L, Xie H, Sun L and Mills J K 2019 A
[32] Wen J, Shen D, Wang R, Li J and Ma J 2020 A two-fixed-end
survey of piezoelectric actuators with long working stroke
in recent years: classifications, principles, connections
and distinctions Mech. Syst. Signal Process.
123 591–605
[22] Shao Y, Shao S, Xu M, Song S and Tian Z 2019 An inertial
piezoelectric actuator with miniaturized structure
and improved load capacity Smart Mater. Struct.
28 055023
[23] Ding M, Ma J, Hu Y, Li J and Wen J 2020 An inertial
piezoelectric rotary actuator characterized by the motion
without rollback Smart Mater. Struct. 29 095015
[24] Zhang S, Liu Y, Deng J, Tian X and Gao X 2021 Development
of a two-DOF inertial rotary motor using a piezoelectric
beam piezoelectric inertial actuator using electromagnet
controlled magnetorheological fluid (MRF) for friction
regulation Smart Mater. Struct. 29 065011
[33] Shen D, Wen J, Ma J, Hu Y, Wang R and Li J 2020 A novel
linear inertial piezoelectric actuator based on asymmetric
clamping materials Sens. Actuators A 303 111746
[34] Zhang E, Hu Y, Bao H, Li J, Ma J and Wen J 2019 A linear
inertial piezoelectric actuator using a single bimorph
vibrator Smart Mater. Struct. 28 115020
[35] Hu Y, Wang R, Wen J and Liu J 2019 A low-frequency
structure-control-type inertial actuator using miniaturized
bimorph piezoelectric vibrators IEEE Trans. Ind. Electron.
66 6179–88
11
| null |
10.1038/s41467-022-35202-8
|
Data availability
Data supporting the findings of this manuscript are available from the
corresponding author upon request. The source data underlying all
figures are available as a Source Data file provided with this
paper. Source data are provided with this paper.
Code availability
All codes used for data analysis may be requested from the authors.
|
Data availability Data supporting the findings of this manuscript are available from the corresponding author upon request. The source data underlying all figures are available as a Source Data file provided with this paper. Source data are provided with this paper. Code availability All codes used for data analysis may be requested from the authors.
|
Article
https://doi.org/10.1038/s41467-022-35202-8
Membrane-mediated protein interactions
drive membrane protein organization
Received: 6 July 2022
Accepted: 22 November 2022
Check for updates
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
Yining Jiang 1,2, Batiste Thienpont3, Vinay Sapuru4,5, Richard K. Hite 4,
Jeremy S. Dittman 6, James N. Sturgis
3 & Simon Scheuring 2,7,8
The plasma membrane’s main constituents, i.e., phospholipids and membrane
proteins, are known to be organized in lipid-protein functional domains and
supercomplexes. No active membrane-intrinsic process is known to establish
membrane organization. Thus, the interplay of thermal fluctuations and the
biophysical determinants of membrane-mediated protein interactions must
be considered to understand membrane protein organization. Here, we used
high-speed atomic force microscopy and kinetic and membrane elastic theory
to investigate the behavior of a model membrane protein in oligomerization
and assembly in controlled lipid environments. We find that membrane
hydrophobic mismatch modulates oligomerization and assembly energetics,
and 2D organization. Our experimental and theoretical frameworks reveal how
membrane organization can emerge from Brownian diffusion and a minimal
set of physical properties of the membrane constituents.
In an amended version of the fluid mosaic model1,2, the membrane is
not a passive medium but plays an active role modulating membrane
protein function and organization through its physical properties3,4.
Changes in membrane protein function that depend on membrane
properties have been measured experimentally using approaches such
as electrophysiology and fluorescence-based vesicle transport
assays5,6. In contrast, the direct experimental study of membrane-
mediated membrane protein oligomerization and assembly remains
challenging.
The two-dimensional (2D) organization of a biological membrane
would be random if the interaction energies between all components
were of the order of kBT2. In reality, cell membranes and their con-
stituent membrane proteins display a non-random organization.
Fluorescence microscopy and biochemical observations have repor-
ted lipid-protein rafts7,8, functional domains9,10, and membrane protein
supercomplexes11,12, clear signatures of non-randomness of biological
membranes. In eukaryotic cells, both membrane components, e.g.,
phospholipids and cholesterol, and the peripheral environment, e.g.,
cytoskeleton and extracellular matrix, contribute to non-random
membrane organization13. Peripheral interactions tether membrane
molecules and serve as lateral diffusing barriers14, while the membrane
is the medium for molecules to interact and its influence can be stu-
died in a controlled way. To the best of our knowledge, no active
process intrinsic to the membrane is known to steer and place mem-
brane proteins in membranes. Thus, the question is: What drives
membrane organization? First, membrane protein interactions can be
protein-mediated, meaning that the two partner molecules make
direct protein–protein contact, or interact via a third protein that
holds them together. In this case, strong interactions, e.g., hydrogen
bonds, ionic- and dipole-dipole interactions, can be formed between
the partner molecules. Second, membrane protein interactions can be
membrane-mediated, where mainly hydrophobic amino acid residues
1Biochemistry & Structural Biology, Cell & Developmental Biology, and Molecular Biology (BCMB) Program, Weill Cornell Graduate School of Biomedical
Sciences, 1300 York Avenue, New York, NY 10065, USA. 2Weill Cornell Medicine, Department of Anesthesiology, 1300 York Avenue, New York, NY 10065,
USA. 3Laboratoire d’Ingénierie des Systèmes Macromoléculaires (LISM), Unité Mixte de Recherche (UMR) 7255, Centre National de la Recherche Scientifique
(CNRS), Aix Marseille Université, Marseille, France. 4Structural Biology Program, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, New York, NY
10065, USA. 5Physiology, Biophysics, and Systems Biology (PBSB) Program, Weill Cornell Graduate School of Biomedical Sciences, 1300 York Avenue, New
York, NY 10065, USA. 6Weill Cornell Medicine, Department of Biochemistry, 1300 York Avenue, New York, NY 10065, USA. 7Weill Cornell Medicine,
Department of Physiology and Biophysics, 1300 York Avenue, New York, NY 10065, USA. 8Kavli Institute at Cornell for Nanoscale Science, Cornell University,
Ithaca, NY 14853, USA.
e-mail: [email protected]
Nature Communications |
(2022) 13:7373
1
Article
https://doi.org/10.1038/s41467-022-35202-8
on a membrane protein surface are exposed to the hydrophobic
phospholipid bilayer core. As a result, none of the above-mentioned
strong interactions can be formed. In this case, sufficient energy, i.e.,
several kBT, must be generated from weak hydrophobic interactions
between lipids and membrane proteins as well as intrinsic membrane
physical properties. In the past decades, numerous theoretical and
computational studies have predicted a key role of the membrane
mechanics, e.g. thickness, stiffness, curvature, and tension, in these
interactions3,15–22. While strong hydrophilic interactions via cyto-
plasmic and extracellular domains and weak hydrophobic interactions
can coexist, an interesting aspect of membrane-mediated interactions
is their long range. Indeed, membrane proteins sense each other
through the membrane over distances up to ~10 nm15. As a result,
attractive and repulsive membrane-mediated long-range interactions
drive protein positioning and organization before local electrostatic
interactions can form between proteins at short distances on the order
of ~2 nm. Therefore, investigating membrane-mediated interactions is
crucial for understanding membrane organization more generally.
microscopy (HS-AFM)27,28 to directly visualize and quantify membrane-
mediated interactions of unlabeled membrane proteins at high spatial
and temporal resolution: We use the Escherichia coli water channel
Aquaporin-Z (AqpZ) and synthetic lipids of defined hydrocarbon tail
length as an experimental model system to study the oligomerization
and interaction energies of membrane proteins as a function of the
bilayer thickness in which they are embedded. The experimental sys-
tem is well-defined: (i) AqpZ is solved to high-resolution by X-ray
crystallography29, providing details about the AqpZ structure and its
hydrophobic thickness. (ii) An AqpZ-W14A mutant exposes surfaces to
the membrane akin the AqpZ-WT tetramer, but has destabilized pro-
tomer interfaces30, enabling us to study both the protein assembly and
oligomerization processes. (iii) The thickness of the synthetic purified
lipids used here have been resolved by small-angle X-ray diffraction31,
providing precise control and knowledge of the membrane environ-
ment in which the membrane-mediated protein interactions are mea-
sured. Finally, (iv) the HS-AFM movies provide unique direct structural
and dynamic data exploitable for quantitative analysis.
While
circular
dichroism23,
single-molecule
fluorescence
microscopy24, fluorescence correlation spectroscopy (FCS)25, and
Förster resonance energy transfer (FRET)26 have been employed to
study membrane protein interactions and have provided invaluable
observations that informed theory, these approaches are more indir-
ect, make use of labels and/or are resolution limited. Here, we report
an experimental design employing high-speed atomic force
Results
Experimental design to study membrane-mediated protein
interactions
To study membrane-mediated protein interactions, we recon-
stituted AqpZ-W14A into a phospholipid bilayer consisting of 1,2-
1,2-dioleoyl-sn-
dioleoyl-sn-glycero-3-phosphocholine (DOPC),
a
reconstitution
b
c
2D-sheets
proteoliposomes
100nm
d
20nm
E
C
10nm
10nm
E123
T107
E31
P30
D110
e
f
1.
2D-sheets
physisorption
2D-sheets
mica
2.
lipid
addition
liposomes
3.
bilayer spreading
membrane fusion
4.
diffusion
association / dissociation
*
S
P
g
t=44s
t=240s
t=265s
t=318s
D
t=351s
D
t=382s
t=416s
A
M
P
S
D
L
D
D
L
L
L
L
50nm
50nm
50nm
50nm
50nm
50nm
50nm
1
2
3
4
5
AqpZ
membrane
mica
100
200
time (s)
300
400
500
i
*D1
*D3
50nm
*D1
*D2
*D2
*D3
10nm
10nm
10nm
M
2μm
h
%
e
g
a
r
e
v
o
c
100
50
0
0
Fig. 1 | Sample characterization and experimental strategy to study membrane-
mediated protein interactions at the single-molecule level. a Sample: Recon-
stitution at lipid-to-protein ratio (LPR) of 0.1 (w:w) results in 2D-crystalline AqpZ
proteo-liposomes and sheets. b–d Sample characterization. b Negative stain elec-
tron microscopy (EM): Tetragonal packing of AqpZ in a 2D-sheet; Supplementary
Fig. 1. c Cryo-EM 2D-crystallography: Projection map at 4 Å resolution (1 unit cell,
side length: 95 Å; Supplementary Fig. 2). d HS-AFM images at three different
magnifications (Left to right: 0.5 nm/pixel, 0.33 nm/pixel, and 0.17 nm/pixel). E:
Extracellular surface. C: Cytoplasmic surface. Right: LAFM map and surface repre-
sentation of the X-ray structure PDB 3NKC. Surface protruding amino acids are
labeled in the structure (arrowheads in LAFM map). e Experimental strategy to
study membrane-mediated protein interactions: 1. Sample physisorption to the
mica HS-AFM support. 2. Addition of liposomes of lipids with hydro-carbon chain
length C14, C16, C18, C20 (Supplementary Fig. 4). 3. Lipid spreading on the mica
leads to fusion of the free bilayer with AqpZ sheets. 4. Equilibrium membrane
protein interaction dynamics: Diffusion, association, and dissociation. Asterisk:
buffer layer between mica surface and lipid bilayer due to electrostatic shielding of
surface charges on mica and protein allows protein diffusion on the atomically flat
mica surface. f AFM overview. The sample covers <5% of the surface. M: Mica. P:
Proteo-liposomes. S: 2D-sheets. g HS-AFM movie frames (Supplementary Movie 1)
of the membrane fusion experiment (DOPC). M: Mica. A: AqpZ array. L: Lipid
bilayer. D: Diffusing AqpZ. h Analysis of the membrane fusion process in (g). (1)
Lipid addition. (2) Bilayer spreading and membrane fusion. 3: Onset of AqpZ dif-
fusion. 4: 100% membrane coverage. 5: 100% coverage of the membrane by dif-
fusing molecules. i Diffusion in the membrane regions indicated by dashed squares
labeled D1, D2 and D3 (left) (Supplementary Movie 2). Right: Enlarged and contrast
enhanced images at slightly increased imaging force of the diffusion fields D1, D2,
D3. Similar results as in (b), (d), (f), (g), and (i) were observed in all samples/
experiments from all biological replica. Schematics in (a) and (e) were generated
using Biorender.com.
Nature Communications |
(2022) 13:7373
2
Article
https://doi.org/10.1038/s41467-022-35202-8
glycero-3-phosphoethanolamine (DOPE) and 1,2-dioleoyl-sn-gly-
cero-3-phospho-L-serine (DOPS) at ratio 8:1:1 (w:w:w) (Fig. 1a). The
membrane-embedded AqpZ molecules formed 2D crystalline
arrays, either in sheets or in proteo-liposomes, due to recon-
stitution at very low lipid-to-protein ratio (LPR) of 0.1 (w:w; ~20
lipid molecules per AqpZ tetramer) (Fig. 1b, Supplementary Fig. 1).
To get detailed insights into the sample morphology, we solved a
projection map of the AqpZ 2D-arrays to 4 Å resolution using cryo-
electron microscopy (Cryo-EM; Fig. 1c, Supplementary Fig. 2). The
Cryo-EM analysis of the 2D-array revealed p4212 plane symmetry
group, where each AqpZ contacted four AqpZ in the opposite
orientation32,33, and the protein coverage in the 2D-arrays was
~80%. We imaged the AqpZ 2D-arrays using tapping-mode HS-AFM
at various magnifications (Fig. 1d). In HS-AFM, the majority of the
observed AqpZ arrays had a size of ~200 nm in diameter and
allowed the extracellular and the cytoplasmic sides of AqpZ to be
resolved (Fig. 1d, arrowheads E, C). From high-resolution HS-AFM
images, we calculated a localization AFM map (LAFM)34 of the
extracellular AqpZ surface, in which details of surface protruding
residues were resolved (Fig. 1d, right).
The AqpZ arrays serve as model membrane protein assemblies to
study membrane-mediated protein interactions upon controlled lipid
addition. In a typical experiment, membrane-embedded AqpZ is
sparsely distributed on the mica HS-AFM sample support (Fig. 1e, step
1, Fig. 1f). The arrays initially covered <5% of the entire mica (Fig. 1f),
meaning that the lipid coverage provided by the sample represents
<1% of the surface. This is important for the ensuing of the experiment,
in which vesicles of defined lipid composition are supplemented to the
HS-AFM fluid cell, pure DOPC liposomes in the presented experiment
(Fig. 1e, step 2, Fig. 1g, t = 44 s). The added lipids spontaneously dis-
persed across the mica surface and fused with existing membrane
patches to cover the entire imaging area with a lipid bilayer (Fig. 1e,
step 3, Fig. 1g, t = 240 s to t = 351 s). AqpZ dissociated from the edges of
the protein arrays and diffused into the newly formed lipid bilayer,
rapidly reaching a new dynamic equilibrium state comprising a mix-
ture of AqpZ protein arrays and freely diffusing molecules (Fig. 1e, step
4, Fig. 1g, t = 382 s to t = 416 s). The presence of rapidly diffusing AqpZ
was revealed by the increased average height of the bilayer areas as
compared to empty bilayer. These areas (Fig. 1g, labeled D) i.e. the
diffusing molecules, had a height of ~1 nm above the membrane level,
slightly lower than the extracellular face of stable molecules. This is
expected as the average height of transient molecules comprises fre-
quent detections of the edges of fast diffusing molecules, thus
recording a lower height than that of a stable molecule. In the example
experiment (Fig. 1g, Supplementary Movie 1), vesicle addition (Fig. 1h,
step 1) initiated bilayer spreading after ~180 s (Fig. 1h, step 2) which
covered the surface within ~120 s (Fig. 1h, steps 2–4), while membrane
protein diffusion started ~50 s after the first occurrence of membrane
fusion (Fig. 1h, step 3) and equilibrated within ~50 s after complete
membrane formation (Fig. 1h, steps 3–5). At this stage, the entire
surface was covered with a single protein-lipid layer and no vesicular
structures were found (see Fig. 1g, t = 265 s to t = 318 s). Zooming into
membrane regions in HS-AFM imaging mode at slightly increased
imaging force revealed diffusing molecules as transient streaks in scan
lines (Fig. 1i, Supplementary Movie 2). Quantitative characterization
of
the diffusing molecules is possible using HS-AFM height
spectroscopy35, as shown in Fig. 2.
The experimental design described here (Fig. 1) combined with
the high spatio-temporal resolution of HS-AFM, ~0.5 nm/pixel at 1
frame/s in the exemplified experiment, allows us to study membrane-
mediated protein interactions at the single-molecule level (Supple-
mentary Movie 1). Importantly, given that i) the initial sample surface
coverage was <5% (Fig. 1f), ii) the reconstitution was performed at LPR
0.1, and iii) the protein covers ~80% in the 2D-arrays as reported by the
cryo-EM map (Fig. 1c, Supplementary Fig. 2), we know that the
subsequently added vesicles contribute >99% of the lipid surface
coverage in the experiments.
Experimental quantification of membrane-mediated protein
interactions
Under the experimental conditions described here, the AqpZ-
membrane system reached equilibrium after about 6 min: The
bilayer covered the entire sample surface, the bilayer fused with
the pre-existing protein patches, and molecules freely diffused
throughout the bilayer (Fig. 1h, step 5). We waited for another
>15 min and recorded single-molecule membrane-mediated
association-dissociation dynamics to and from the protein array
edges over tens of minutes (Fig. 2a–d dashed outlines, Supple-
mentary Movies 3–6, illustrated experiments are in C18). These
movies were recorded ~40 min (Fig. 2a, Supplementary Movies 3),
~120 min (Fig. 2b, Supplementary Movies 4), ~40 min (Fig. 2c,
Supplementary Movies 5) and ~15 min (Fig. 2d, Supplementary
Movies 6) after continuous bilayer formation. Importantly, the
AqpZ 2D-arrays continued to change shape with local growth and
contraction but without global changes in array size (Fig. 2e, f).
Thus, the association-dissociation events analyzed here were
recorded under equilibrium conditions. We distinguished the
association-dissociation events as either one-bond or two-bond
events, defined by the number of neighbor interactions of a
molecule in the AqpZ array (Fig. 2b–d). We analyzed only com-
plete events, defined as the process in which a diffusing AqpZ
associated to and then dissociated from an array, thus defining
bound-state dwell times (Supplementary Fig. 3). We captured a
large number of complete one-bond (Fig. 2b, asterisk 1) and two-
bond (Fig. 2b, asterisk 2) events over extended experimental
durations (Supplementary Movies 4–6). Association-dissociation
events to and from three-bond locations were rare and were not
analyzed.
HS-AFM imaging of one-bond and two-bond association-dis-
sociation events revealed exponential dwell-time distributions with
distinct time constants (τ1 and τ2). We applied two strategies to analyze
the state dwell times: First, we treated the one-bond (Fig. 2g, left) and
two-bond (Fig. 2g, center) events separately, based on the protein
array images, where the molecular environment was entirely unchan-
ged before and after association-dissociation. Accordingly, the one-
bond dwell-time distribution was well described with a single expo-
nential decay time constant τ1 = 0.81 s (n = 246), and the two-bond
dwell-time distribution decay with a time constant τ2 = 7.4 s (n = 549).
This strategy established that the two interaction morphologies, one-
bond vs two-bond, had significantly different dwell times,
and
assigned the fast time constant to one-bond and the slow time con-
stant to two-bond events. Second, all events were pooled, and the
resulting dwell-time distribution was fit with a double exponential
(Fig. 2g, right):
P Bð Þ = c1e
(cid:2) t
τ
1 + ð1 (cid:2) c1
(cid:2) t
τ
2
Þe
ð1Þ
where P(B) is the normalized cumulative probability of all events (n =
1096). τ1(C18) = 0.77 s and τ2(C18) = 8.2 s were the fast and slow time
constants of the exponential decay, and c1(C18) = 0.55 and c2(C18) = 0.45 (c2
= 1-c1) represented the relative abundance of the fast and slow events,
respectively. The fast and slow time constants, τ1 and τ2, agreed well with
the time constants individually determined for the one-bond and two-
bond events based on imaging knowledge, but the ensemble fitting
quality was better. Therefore, we used ensemble fitting to analyze each
individual array in all experiments to assess detailed statistics with error
estimates between experimental observations (Table 1).
Membrane elastic theory proposes that hydrophobic mis-
match between the bilayer core and the protein transmembrane
domain (TMD) controls membrane protein interactions15,16,36,37. To
Nature Communications |
(2022) 13:7373
3
Article
https://doi.org/10.1038/s41467-022-35202-8
test this, we repeated the above experiments and analyses in
membranes constituted of synthetic purified lipids with different
thicknesses.
In addition to DOPC (C18), we also used
1,2- dimyristoleoyl-sn-glycero-3-phosphocholine (C14), 1,2-dipalmi-
toleoyl-sn-glycero-3-phosphocholine (C16), and 1,2-dieicosenoyl-sn-
glycero-3-phosphocholine (C20) in the lipid addition step (Fig. 1e,
step 2). These lipids have the same degree of saturation but different
hydrocarbon tail
lengths (Supplementary Fig. 4), with a ~1.5 Å
increase in bilayer thickness for each additional carbon atom, ran-
ging from ~24 Å for the C14 bilayer to ~34 Å for the C20 bilayer38. The
phase transition temperatures for all these lipids were far below
room temperature, and thus all bilayers were in the fluid phase in our
experiments. The dynamics of AqpZ, which has a hydrophobic
thickness of ~28.6 Å matching a hypothetical C17 bilayer (Supple-
mentary Fig. 5, “Methods”), were analyzed in C14, C16, C18, and C20
membrane environments, revealing that indeed, the membrane
thickness had an influence on the membrane-mediated protein
interactions (Table 1, columns 1 and 2, Fig. 2h).
Next, we exploited HS-AFM height spectroscopy (HS-AFM-HS, see
“Methods”)35 to characterize the diffusion of unbound molecules that
are not resolved in images (Fig. 2i). HS-AFM-HS captured the height
fluctuations induced by molecules diffusing under the tip with µs
Nature Communications |
(2022) 13:7373
4
Article
https://doi.org/10.1038/s41467-022-35202-8
Fig. 2 | The hydrophobic mismatch between protein and phospholipid bilayer
impacts membrane protein interactions and diffusion. a–d HS-AFM movie
frames (Supplementary Movies 3–6) of single-molecule association-dissociation
dynamics at the edges of AqpZ arrays in a C18 membrane (image parameters: (a)
1.0 nm/pixel, (b) 0.5 nm/pixel, (c) 0.33 nm/pixel, and (d) 0.17 nm/pixel). Dashed
outlines: association-dissociation events. Asterisk 1: one-bond event. Asterisk 2:
two-bond event. These image series have been acquired 40 min (a), 2 h (b), 40 min
(c), and 15 min (d) after lipid vesicle addition and continuous bilayer formation.
e, f Number of molecules vs time of Supplementary movies 3 and 4, respectively
(panels (a) and (b)). g Dwell-time distributions of association-dissociation events in
a C18-membrane. One-bond (left, (n = 246) and two-bond (center, n = 549) events
were fitted separately based on imaging knowledge using one exponential, or
collectively using two exponentials (right, n = 1096). h Normalized fittings of all
events in C14 (n = 308, 3 replicas), C16 (n = 761, 3 replicas), C18 (n = 1096, 3 replicas)
and C20 (n = 288, 3 replicas) membranes (as indicated). Insets: Detail views of the
fast exponential decay. Thick lines: averages. Thin lines: ±s.e. i HS-AFM height
spectroscopy (HS-AFM-HS). Left: Schematic of HS-AFM-HS principle: The tip is at a
temporal resolution, far away, ~100 nm from the border of an AqpZ 2D-
array, in the bilayer membrane. Analysis of height-time traces (Fig. 2i,
middle) revealed the 2D-diffusion coefficient, DU (µm2 s−1), of the freely
D = w2=4DU and the 2D-concentration
diffusing AqpZ molecules using τ
of unbound AqpZ, CU (µm−2) through CU = tðz>HT
Þ=tðtotalÞ (cid:3) 1=AAqpZ,
where τD is the dwell-time of the diffusion events, and w is the detec-
tion area estimated from the area of an AqpZ, AAqpZ, convoluted with
the tip radius (~1 nm)35. t(z>HT) is the total time the tip detects diffusion
events at a height z above HT = 5σ of the height value distribution,
during t(total), the total measurement time35,39. The diffusion events had
a height between 1 nm and 1.5 nm above the membrane, i.e. baseline
(Fig. 2i, middle), agreeing well with the height of the diffusive area
(Fig. 1g, i, labeled D), and with the protrusion height of array-bound
cytoplasmic, ~1 nm, and extracellular, ~1.5 nm, face exposing AqpZ
(Fig. 2b–d). We found that the diffusion coefficient DU of AqpZ also
varied with the bilayer thickness (Table 1, column 3, Fig. 2i).
Both the state dwell times and the diffusion speed were altered by
changes in lipid bilayer thickness (Table 1). Notably, AqpZ in bilayers of
intermediate thickness, closest to the hydrophobic thickness of the
protein, displayed longer τ1 and shorter τ2, as well as faster DU. Since τ1
from the fitting is slightly shorter than the imaging rate, 1 frame/s, we
performed two additional tests. First, we tested the fitting for τ2 while
keeping τ1 fixed at τ1 = 0.7 s (the average τ1 across all lipids). This
constrained fitting strategy confirmed the state dwell-time trend,
where τ2 was prolonged in membranes with increased hydrophobic
mismatch (Table 1, brackets in columns 1 and 2). Second, we imaged
AqpZ 2D-arrays in DOPC at 4 frames/s (Supplementary Movie 7; at
proportionally smaller scan size but identical pixel sampling as the
experiments at 1 frame/s), and estimated τ1(C18) = 0.54 ± 0.09 s and
τ2(C18) = 8.0 ± 1.1 s, thereby supporting the sub-second τ1 derived from
the 1 frame/s data (Table 1, column 1).
A kinetic model of membrane-mediated protein interactions
A first analysis of the interactions can be made by considering the
equilibrium between freely diffusing molecules (U) dissolved in the
Table 1 | Statistics of AqpZ4 association/dissociation kinetics
and diffusion in C14, C16, C18 and C20 lipid bilayers
τ1 (s)
0.5 ± 0.14 (0.7)
τ2 (s)
13 ± 3.3 (16 ± 4.9)
DU (µm2/s)
0.42 ± 0.051
CU (µm−2)
110 ± 39
0.7 ± 0.10 (0.7)
9 ± 3.2 (8 ± 3.4)
0.64 ± 0.039
30 ± 18
0.9 ± 0.10 (0.7)
9 ± 1.1 (8 ± 1.7)
0.59 ± 0.038
30 ± 10
C14
C16
C18
C20
0.6 ± 0.10 (0.7)
15 ± 1.2 (17 ± 1.8)
0.48 ± 0.031
160 ± 93
The time constants τ1 and τ2 were determined using Eq. (1). Brackets: Alternative fitting strategy:
Fixing τ1 to 0.7 s and optimizing τ2. The 2D diffusion coefficients D2D and 2D concentration
CU were determined using HS-AFM-HS. All statistics (mean ± s.e.) were determined from three
biological replica in each condition.
fixed location monitoring molecular diffusion events. Middle: HS-AFM-HS height-
time trace. Light gray: raw data. Dark gray: diffusion events, threshold height HT =
5std above mean of the height distribution next to the trace. The 0 nm height level
was set to the membrane surface. Right: Distribution of event dwell times τD.
j Model of the membrane-mediated protein interactions where a diffusing molecule
U can engage a 1B (one-bond) or 2B (two-bond) interaction with the array.
asso), defined as the energy difference between states U
k Association energy (ΔG0
and B, (l) energy difference between states 1B and 2B (ΔG0
diff), and (m) diffusion
coefficient (DU), as functions of the acyl-chain length (top) and hydrophobic mis-
2) (bottom). lbilayer: Hydrophobic thickness of the
match (u0), or its squared value (u0
membrane. The hydrophobic mismatch is calculated as u0 = 0.5|lbilayer – lAqpZ|,
where lAqpZ is the hydrophobic thickness of AqpZ (~28.6 Å, Supplementary Fig. 5,
dashed red lines in the top panels). Solid curves are quadratic and linear fits to the
data points. Statistics (mean±s.e.) in (k) and (m) are determined from three bio-
logical replica, in each condition. Statistics (mean±s.e.) in (i) is relevant to the
statistics in (h), according to Eq. (4).
lipid membrane, and the bound molecules (B) that form the arrays.
This equilibrium is associated with the reaction:
K U(cid:2)B
eq
U !
B,
ð2Þ
eq = CB
where K U(cid:2)B
=CU is the equilibrium constant for this reaction, and
CB is the concentration of bound and CU the concentration of unbound
molecules. Thus, the association energy (ΔG0
asso) is
ΔG0
asso = (cid:2)kBTln
(cid:1) (cid:3)
CB
CU
ð3Þ
We know CB precisely from cryo-EM (Fig. 1c, Supplementary Fig. 2)
and HS-AFM (Fig. 1d) imaging. In the arrays one AqpZ tetramer occu-
pies 45.125 nm2, and thus CB is 22,161 µm−2. We measured, using HS-
AFM-HS, the concentration of freely diffusing molecules, CU, in all
bilayers (Fig. 2i, Table 1, column 4). From these two measurements,
ΔG0
asso is calculated according to Eq. (3) (Table 2, column 1). The
intuition of Eq. (3) is to relate the association energy (ΔG0
asso) between
unbound and bound molecules to the equilibrium concentrations of
the two species.
To get further insights into the single-molecule behavior in dif-
ferent B states, we consider a simple kinetic three-state model (Fig. 2j).
In this model, the unbound AqpZ (U, red) could bind to an array (A,
gray) in one of two possible modes: either in a one-bond (1B, green) or
in a two-bond (2B, blue) site. The association-dissociation events,
directly observed by HS-AFM at the single-molecule level, were found
to follow first order kinetics from both 1B and 2B states. Accordingly,
the dissociation of a molecule making one contact with the array, state
1B, is fast, i.e., the bond lifetime is short. Making an additional contact
to the array by filling a gap in a corner, state 2B, stabilizes the bound
state and its dissociation is slow, i.e., the bond lifetime is long. Thus,
the energy difference between states 1B and 2B, the energy gain of the
Table 2 | Energies of membrane-mediated AqpZ interactions
in C14, C16, C18 and C20 lipids
ΔG0P-P(asso)
(kBT)
−6.9
−6.9
−6.9
−6.9
C14
C16
C18
C20
ΔG0asso (kBT) ΔG0P-P(diff)
−5.4 ± 0.41
−6.8 ± 0.61
−6.6 ± 0.34
−5.1 ± 0.56
(kBT)
−2.3
−2.3
−2.3
−2.3
ΔG0diff (kBT)
−3.2 ± 0.41
−2.5 ± 0.27
−2.3 ± 0.13
−3.2 ± 0.25
The energies were determined based on the measured state dwell times from HS-AFM imaging
and HS-AFM-HS as described in the text.
Nature Communications |
(2022) 13:7373
5
Article
https://doi.org/10.1038/s41467-022-35202-8
second bond formation (ΔG0
diff), can be estimated as40
(cid:1) (cid:3)
τ
2
τ
diff = (cid:2) kBTln
ΔG0
1
ð4Þ
The intuition of Eq. (4) is that i) the two bound states, 1B and 2B,
interconvert to the same unbound state through the same transition
state, and ii) according to simple rate theory, the logarithm of the
unbinding rate is proportional to the energy barrier height. Therefore,
the logarithm of the ratio of the dwell times is related to the energy
difference of the two bound states.
Based on the measured state dwell times (Table 1) and the kinetic
model (Eqs. (3) and (4)), the association ΔG0
asso of an unbound AqpZ to
others is favored (~−6.7 kBT) in C16 and C18 membranes matching the
protein hydrophobic thickness, while it is less favored (~−5.3 kBT) in
C14 and C20 membranes (Fig. 2k, top, Table 2, column 2). In direct
analogy with an elastic potential energy, ΔG0
asso can be plotted as a
2, and
function of the square value of the hydrophobic mismatch, u0
P-P(asso) of −6.9 kBT
extrapolated to zero mismatch at an energy of ΔG0
(Fig. 2k, bottom), suggesting a favorable membrane-independent
protein–protein interaction energy (Table 2, column 1). In contrast,
diff is lower (~−3.2 kBT) in membranes of the shorter and longer
ΔG0
lipids, C14 and C20, than in C16 and C18 membranes (~−2.4 kBT) that
match the hydrophobic core of the protein (Fig. 2l, top, Table 2, col-
umn 4). Again, the energies can be characterized by fitting a quadratic
curve (Fig. 2l, top), and are therefore plotted as a function of the
2 (Fig. 2l, bottom), and the
square of the hydrophobic mismatch, u0
the membrane mismatch-
intercept provides an estimate of
P-P(diff) = −2.3 kBT
independent protein–protein interaction energy ΔG0
(Table 2, column 3). The difference of the membrane-independent
protein–protein interaction energies calculated from these two inde-
P-P(asso) = −6.9 kBT, characterizing the overall
pendent approaches, ΔG0
P-P(diff) = −2.3 kBT,
association of a free molecule to an array, and ΔG0
characterizing the bond strengthening by one additional protein
partner, is well explained by the fact that an average array-bound AqpZ
molecule has ~3 interactions with neighbors.
Comparing the protomer mobility in different bilayer thicknesses,
the measured diffusion coefficient was found to decrease with
increasing hydrophobic mismatch (Fig. 2m, Table 1, column 3). This
finding agrees with reported deviations from Saffman–Delbrück
diffusion39 due to an increase in the effective membrane viscosity that
scales linearly with membrane mismatch41,42. Extrapolation of these
measurements indicate that AqpZ would attain a maximal value of DU =
0.7 µm2/s in a perfectly matching bilayer in our experimental system
(Fig. 2m, bottom). We consider that the underlying mica may affect
diffusion through interaction with the proteins and/or modulation of
the bilayer physical properties, but note that the atomically flat mica
does not provide diffusion obstacles and a diffusion coefficient of
0.7 µm2/s is a rather typical value for a membrane protein of the size
of AqpZ.
In summary, the interaction energies emerging from hydrophobic
mismatch account for ~1.5 kBT (Table 2), complemented by a larger
direct protein–protein mismatch-independent energy. Importantly,
membrane-mediated membrane protein interactions are long-range –
membrane proteins sense each other through the membrane at dis-
tances far beyond the range where electrostatic and Van der Waals
interactions become important. Thus, membrane-mediated mem-
brane protein interactions represent a key driving force in the orga-
nization of membrane proteins.
Membrane-mediated interactions are long-range and geometry-
sensitive
Past experimental and theoretical work on interactions between inte-
gral membrane proteins and the lipid bilayers in which they are
imbedded has provided physical models that explicitly estimate the
membrane deformation energetics and the impact of hydrophobic
mismatch as a function of distance (d) between membrane proteins
(Fig. 3a). Each membrane protein deforms the membrane at its cir-
cumference to match the hydrophobic core of the membrane with its
hydrophobic membrane exposed residues, so that unfavorable inter-
actions between lipid hydrocarbon tails and hydrophilic residues on
the inner and outer brim of the protein are minimized15. The mem-
brane deformation is approximated as a 2D continuous elastic field,
uxy, representing the deviation of the lipid head-group from its
unperturbed height16. The hydrophobic mismatch of one leaflet u is u0
at the protein-lipid interface and vanishes to zero as the membrane
becomes unperturbed. The deformation energy, Gdef, in this setting
results from membrane compression and bending (Supplementary
Fig. 6, Supplementary Note 1), and both have a form analogous to
Hooke’s law. Thus all components contribute to elastic energy. The
expression of Gdef is
Z Z
"
2
(cid:1) (cid:3)
uxy
l
K A
(cid:4)
+ κ
b
∇2uxy
#
(cid:5)
2
Gdef =
1
2
dxdy,
ð5Þ
∂x2 + ∂2
where KA is the bilayer stretch modulus, l the thickness, κb the bending
modulus, and ∇2 = ∂2
∂y2 the Laplace operator. Since the 2D defor-
mation energy associated with multiple proteins depends on the
complex geometries of the membrane and protein configuration43, i.e.
the shapes of the protein cross-sections as well as the distances and
orientations relative to each other, etc., we first illustrate the intuition
of Gdef generated by two cylindrical proteins as a simple case
(Fig. 3a–e).
We solved the 2D continuous elastic field uxy induced by two
cylindrical proteins of identical membrane mismatch at different edge-
to-edge distances d through numerical simulation (Supplementary
Fig. 7)44. If the protein centers are positioned along the x-axis, and the
closest protein-lipid interfaces are at (x1,0) (Fig. 3a, black square) and
(x2,0) (Fig. 3a, black circle), then ux1,0 = ux2,0 = u0 and d = |x2 – x1|
(Supplementary Note 1). The membranes adopt different profiles
between the proteins as the two molecules approach (Fig. 3b). The
membrane perturbation around each molecule relates to deformation
energy, Gdef, (Fig. 3c), which is the spatial integral of the deformation
energy density dGdef. Thus, the change of the deformation energy in
the approach of two molecules gives the elastic potential, ΔGelas(d)
(Fig. 3d), as
ΔGelas
ðdÞ = Gdef
ðdÞ (cid:2) Gdef
ð + 1Þ
ð6Þ
Due to the physical properties of the membrane, the elastic
potential is ‘felt’ by membrane proteins that are as far as ~7.5 nm apart,
and scales with the hydrophobic mismatch (Fig. 3c, situation 1, Fig. 3d).
At d ~7.5 nm to ~3.5 nm, as the deformed membrane fields overlap, the
potential is repulsive, especially in the case of large hydrophobic
mismatch. This is primarily due to the membrane bending component
that has to accommodate a saddle-shaped membrane topography at
such intermediate distances (Fig. 3c, situation 2, Fig. 3d, e). Decreased
membrane bending and compression at d < ~2 nm produce strongly
attractive potentials at short distances (Fig. 3c, situation 3, Fig. 3d, e).
These results are consistent with previous theoretical studies treating
the approach of two ideal cylindrical inclusions in a membrane16,45,46.
The expected membrane deformation was observed in the space
between 4 AqpZ tetramers. In this region the membrane formed a
saddle point with a height variation of ~1 Å, though this measurement
must be taken with caution because only very sharp tips can poten-
tially probe this region between the proteins (Supplementary Fig. 8).
To relate membrane elastic theory to the experimental observations
described in Fig. 2, we developed a discretized framework that evalu-
ates the changes of the membrane environment associated with
membrane protein assembly configuration changes. In this approach,
Nature Communications |
(2022) 13:7373
6
Article
https://doi.org/10.1038/s41467-022-35202-8
Fig. 3 | Membrane deformation through membrane protein hydrophobic
mismatch provides a theoretical understanding of the experimental results.
a–e Hydrophobic mismatch as an energy source for membrane-mediated mem-
brane protein interactions (Supplementary Fig. 7, Supplementary Note 1). a Sche-
matic of the inclusion-induced membrane deformation with lprotein, protein
hydrophobic thickness, lbilayer, bilayer hydrophobic thickness, u, hydrophobic
mismatch, and d, edge-to-edge distance between proteins (hydrophobic (red) and
hydrophilic (blue) protein surfaces). b Perspective schematic representation of 2D
membrane profiles (one leaflet) when two cylindrical proteins approach. The space
occupied by proteins is not considered part of the deformation field, uxy, and filled
with u0 for illustration: positive, (e.g., C14, left) and negative (e.g., C20, right)
hydrophobic mismatch. Bottom to top: d ~7 nm, ~4 nm, and ~1 nm. c Schematic
representation of the 2D deformation energy density, dGdef, for d ~7 nm, ~4 nm and
~1 nm. d 2D elastic mismatch-dependent energy potential as a function of d for the
four investigated bilayers. e Compression and bending components of the total 2D
elastic mismatch-dependent energy potential, repulsive from ~7 nm to ~3.5 nm
separation, and attractive at separation shorter ~3.5 nm. f, g Changes of 2D mem-
brane configurations in association-dissociation events (Supplementary Fig. 9):
(f) The five basic local-configurations in microscopic array assembly.
g Representative rearrangements of local configurations associated with one-bond
(rearrangement 1) and two-bond (rearrangement 2) interactions. h–j Membrane
protein automata (Supplementary Fig. 10, Supplementary Note 2, and Supple-
mentary Movie 8): (h) Simulated clusters in membranes of no (left), intermediate
(middle), and large (right) hydrophobic mismatch. i Association energy (ΔGasso),
and (j) energy difference between states 1B and 2B (ΔGdiff), as functions of the
deformation energy scale factor ψ/ψnorm, representing the hydrophobic mismatch
square (u0
2), with ψnorm = {1.00 2.06 3.22 4.10} (see “Methods”).
we approximate the membrane as a lattice, each point of which is
either occupied by a molecule or empty (Fig. 3f). A 2 × 2 region of the
membrane lattice has five distinct configurations depending on the
number of molecules that occupy the positions and may be used to
assign deformation energies to the various configurations (see
“Methods”). We denote ψi as the deformation energy of local-
configuration i. Thus, we can write the membrane-dependent energy
change from a membrane configuration rearrangement, e.g., due to a
protein association or dissociation event, Δψ, as
Δψ = δn1
ψ
1 + δn2
ψ
2 + δn3
ψ
3 + δn4
ψ
4,
ð7Þ
where δni is the change, gain or loss, of local-configuration i in the
for the one-bond and two-bond
rearrangement. For example,
association events shown in Fig. 3g, {δn1 δn2 δn3 δn4} equals {−2 −2 2
0} (rearrangement 1) and {−4 0 0 1} (rearrangement 2), respectively,
allowing us to compute the membrane morphological changes (Fig. 3g,
Supplementary Fig. 9). Hence, we developed a simulation, termed
membrane protein automata (Supplementary Note 2, see “Methods”),
with which we simulate distinct membrane protein organizations
through varying the energy term of the direct protein–protein
interaction EP-P, the energy of the relative local-configurations, ψ1, ψ2,
ψ3 and ψ4, and the concentration of the freely diffusing molecules CU
(Supplementary Fig. 10).
To determine {ψ1 ψ2 ψ3 ψ4}, we solved through numerical simu-
lations the 2D continuous elastic field uxy of local-configurations 1 to 4
(Supplementary Note 1, Supplementary Fig. 7)44. Using Eq. (7) to
evaluate the rearrangements in Fig. 3g shows that the rearrangement 2
Nature Communications |
(2022) 13:7373
7
Article
https://doi.org/10.1038/s41467-022-35202-8
Fig. 4 | AqpZ W14A protomer association and dissociation dynamics in a lipid
bilayer that matches the hydrophobic thickness of the AqpZ protomer-
protomer interface. a, b HS-AFM movie frames (Supplementary Movies 9,10) of
non-canonical AqpZ oligomers, AqpZ2 and AqpZ3 in a C20 membrane (image
parameter: 0.33 nm/pixel). Dashed circles highlight AqpZ2 and AqpZ3.
c–e Oligomer transitions: (c) AqpZ4!AqpZ3!AqpZ2, (d) AqpZ4!AqpZ2!AqpZ3,
and (e) AqpZ4!AqpZ3 (arrowheads: molecule of interest; asterisks: neighbor
molecules). Images are averages over 5 consecutive frames (if applicable) with time
stamps corresponding to the first frame of state occurrence. f Occurrence prob-
abilities of AqpZ W14A oligomeric states at the array edge.
is much more favorable than rearrangement 1, since Δψre2-Δψre1 < 0.
Thus, the rectangular shape of the observed arrays with neat borders
and without protruding molecules is favored over more fuzzy protein
assemblies (Figs. 1i, 2a–d, Supplementary Fig. 9).
We performed extensive simulations (Fig. 3h, Supplementary
Fig. 10, Supplementary Note 2), linearly scaling {ψ1 ψ2 ψ3 ψ4} to simulate
2). The membrane
the effect of hydrophobic mismatch square (u0
protein automaton generated protein arrays that displayed similar
morphology as in the experiment (Fig. 3h). Analysis of the simulated
similar membrane-
association/dissociation
dependent energetic trends as in experiments, where ΔGasso scales
positively and ΔGdiff scales negatively with increasing membrane mis-
match square (Fig. 3i, j; compared to Fig. 2k, l).
revealed
events
Lipid thickness matching the protomer interface destabilizes
oligomers
Aquaporins are stable tetramers47, which precluded the analysis of the
oligomerization mechanism. Therefore, we performed these experi-
ments with a W14A mutant, a single residue mutation at the protomer
interface. We reasoned that the exchange of the bulky tryptophan with
the small alanine would allow penetration of membrane lipids into the
interstices between the protomers of tetrameric AqpZ, AqpZ4,
potentially destabilizing the protomer interaction (Supplementary
Fig. 11)30,48. Thus, we expected to observe non-tetrameric AqpZ-W14A,
which we denote AqpZ1 (monomers), AqpZ2 (dimers), and AqpZ3 (tri-
mers). We screened the edges of the AqpZ arrays for these species in
C14, C16, and C18 membranes, but without success. Interestingly, we
reproducibly detected non-tetrameric AqpZ in high-resolution HS-
AFM images in C20 lipids, where ~10% of the array-bound AqpZ had an
oligomeric state that deviated from the tetrameric form (Fig. 4a, b,
outlines, Supplementary Movies 9, 10). As a comparison, non-
tetrameric oligomers were much rarer for WT AqpZ in C20 lipids,
<2%, which suggests that the W14A mutation accounts for at least 2 kBT
in the AqpZ oligomerization (Supplementary Fig. 12). In order to ana-
lyze occurrence probabilities and derive energetics of these states, the
number of interactions with neighboring molecules needed to be
considered: Due to protomer stabilization with the array molecules, an
AqpZn with two neighbors can only dissociate into AqpZ3 or AqpZ2
(Fig. 4c, d), while an AqpZn with three neighbors can only become
AqpZ3 (Fig. 4e). Transition to AqpZ1 would be possible if the array-
bound AqpZ had only one neighbor, but we failed to capture such
events, likely because of the very small size of AqpZ1 combined with
the very short τ1 of the one-bond interaction (see Table 1). We esti-
mated the energy differences between oligomeric states s1 and s2 from
the numbers of observations, Ns1 and Ns2 with neighbors n, in the HS-
AFM imaging period, as:
ΔG0
s2, n
(cid:2) ΔG0
s1,n = (cid:2) kBTln
(cid:3)
(cid:1)
Ns2,n
Ns1,n
ð9Þ
We found that regardless of the neighbor number, AqpZ2 and
AqpZ3 had similar occurrence probabilities, and thus energy difference
compared to AqpZ4, ~2 kBT (Fig. 4f, Table 3), similar in magnitude as
the W14A mutation. These estimated energy differences are hardly
comparable to those between freely diffusing oligomers because i) HS-
AFM experiment requires oligomers having >1 contact with the array
for detection, hence underestimating the real numbers, and ii) the
constrained molecular environment at the array edges reduces the
degree of freedom of movements.
To understand why low-order oligomers occurred in C20 lipids
and not in thinner membranes (C18, C16, and C14), we assessed the
hydrophobic thickness of the AqpZ protomer-protomer interface and
found that it was very different from the hydrophobic thickness that
the tetramer exposes to the membrane.
Indeed, we assessed a
hydrophobic thickness of ~28.6 Å on the membrane exposed surface
Table 3 | Energies of AqpZ-W14A oligomerization in C20 lipid
ΔG0
(C20) (kBT)
AqpZ1-AqpZ2
AqpZ2-AqpZ3
AqpZ2-AqpZ4
AqpZ3-AqpZ4
1 neighbor
2 neighbors
3 neighbors
N/A
N/A
N/A
N/A
N/A
0.0 ± 0.61
−2 ± 1.1
−2 ± 1.1
N/A
N/A
N/A
−2.1 ± 0.51
The energies were determined as described in the text. AqpZ1 is not accessible in both the
2-neighbor and 3-neighbor situations. AqpZ2 is not accessible in the 3-neighbor situation.
Energies in the 1-neighbor situation were not analyzed due to poor statistics.
Nature Communications |
(2022) 13:7373
8
Article
https://doi.org/10.1038/s41467-022-35202-8
Fig. 5 | AqpZ oligomerization and assembly. The AqpZ-W14A oligomerization
energetics was estimated based on observation statistics of non-tetrameric com-
plexes. The protomer interaction ΔG0
match the hydrophobic thickness of the protomer interface. AqpZ diffusion DU is
slowed in lipids with larger hydrophobic mismatch (thicker and thinner).
Membrane-mediated membrane protein interaction ΔG0
olig is weakest (~−2 kBT) in C20 lipids that
asso is most favorable (~
−6.5 kBT) in lipids with thickness close to the hydrophobic thickness of the protein.
Bond formation with two array-bound proteins, filling gaps in the 2D-plane, pro-
diff in lipids with strong mismatch (~−3 kBT). The
vides a maximum energy gain ΔG0
latter driving the assembly towards the formation of membrane protein arrays. The
direct (not membrane-mediated) protein–protein interaction ΔG0
P-P (~−2 kBT) sta-
bilizes these interactions at very short distances.
but estimated a hydrophobic thickness of ~33.0 Å between protomers
(Supplementary Fig. 5c). This much thicker hydrophobic interface
matches roughly the thickness of the C20 lipids. Thus, bilayers with a
hydrophobic core that matches protomer interfaces lower the energy
difference between the oligomeric state and individual protomers,
favoring dissociation.
Strikingly, owing to the experimental design with free membrane
outside the arrays and the time-resolved imaging of HS-AFM, not only
overall statistics of the occurrence of AqpZn could be assessed, but
also real-time transitions AqpZ4 → AqpZ3 → AqpZ2 (Fig. 4c) AqpZ4 →
AqpZ2 → AqpZ3 (Fig. 4d) and AqpZ4 → AqpZ3 (Fig. 4e) could be
observed. Given that these transitions are very slow, tens of seconds
for the dissociation transitions (Fig. 4c, t = 6 s, t = 24 s, t = 55 s, Fig. 4d, t
= 9 s, t = 21 s and Fig. 4e, t = 10 s, t = 80 s), we estimate that the energy
barrier between AqpZ4 and AqpZ3,2 is very high, ~24 kBT in the
experimental conditions (see “Discussion”).
Discussion
Here, we developed an approach to investigate membrane-mediated
protein interactions in a controlled manner and at single-molecule
resolution (Fig. 5). Membrane protein patches that contained very
little lipid (LPR 0.1) and covered only a small portion (~5%) of the
sample surface were supplied with lipids of defined hydrophobic
thickness to form a continuous fluid lipid bilayer in which the
membrane proteins diffused and interacted. Taking HS-AFM movies
of this system, thousands of membrane protein association/dis-
sociation events were recorded in C14, C16, C18, and C20 PC bilayers,
and their dwell times analyzed. Besides, HS-AFM-HS was applied
for the analysis of diffusing molecules, including their 2D con-
centrations and diffusion coefficients. Based on these measures,
together with a mechanical model of the lipid bilayer, we found that
the interaction energies scaled with the hydrophobic mismatch
between protein and the bilayers.
In our model system, the
protein–protein association was more favorable in lipids matching
the protein’s hydrophobic thickness, but the engagement with
multiple neighbors in protein array formation was more favorable in
bilayers with large mismatch. We note that the tested lipids have
similar KA values49,50. In principle κb is proportional to KA and the
bilayer thickness, κb ~KA/l2, thus κb is expected to be somewhat larger
for thinner bilayers. However, we found that C14 and C20 lipids had
very similar energetics and suggest therefore that the apparent
proportionality of κb with thickness is weak.
Based on 1D membrane deformation graphs, one might be led to
think that large hydrophobic mismatch must favor membrane protein
association (Fig. 3a). However, in the 2D membrane, protein associa-
tion in mismatched membranes leads to complex local membrane
deformations, e.g., saddle-points (Fig. 3b, c, Supplementary Fig. 8),
that – as our experiments show – dominate the interactions and are
overall unfavorable. Furthermore, the extrapolation of the mismatch
potential energy dependence of these associations allowed us to
determine the energy of the direct protein–protein interaction. All
interaction energies we found are in the single digit −kBT range (Fig. 2,
Table 2). We reason that such single digit -kBT range energy differences
provide effective biases without locking the molecules in specific
states, and thus leave membrane proteins amenable to rearrange-
ments. We note however that our calculations consider an average
hydrophobic thickness for the protein, but membrane protein surfaces
have local variation of the hydrophobic thickness, and thus the
membrane-mediated protein interaction strength is also expected to
vary locally. While the state energy differences are relatively low, we
note that the dwell times, τ, of the interactions were long, especially for
the molecules engaging in multiple bonds, ~10 s (and for the protomer
interaction, tens of second), suggesting that the bound and unbound
states are separated by high energy barriers. We can estimate the
barrier height ΔEbarrier from the measured dwell times using 1/τ =
Aexp(−ΔEbarrier/kBT), where A is an unknown pre-factor. Using A
~109–1010 s−1, estimated for a bond rupture in a viscous medium51, this
approximation predicts an ΔEbarrier ~23–25 kBT for interactions with τ
~10 s. Thus, the membrane protein tertiary structure and supra-
molecular assembly are kinetically trapped by high energy barriers.
Nature Communications |
(2022) 13:7373
9
Article
https://doi.org/10.1038/s41467-022-35202-8
Using continuous elastic field modeling for the interpretation of
the protein-induced membrane deformation required consideration
of the membrane 2D geometry and the protein configuration. Solving
the deformation fields relevant to the membrane protein arrays of
hundreds of molecules would necessitate significant computational
resources. To this end, we introduced a discretized framework, the
membrane protein automata, to evaluate the morphological changes
and the dynamics of membrane protein assemblies. The complex array
association/dissociation events are considered as rearrangements of a
finite number of local configurations, in which the deformation fields
are readily solved. These automata can, with a fixed set of parameters
obtained from numerical simulations of the local configurations,
reproduce the dependence of protein array self-assembly and
dynamics, as well as its sensitivity to hydrophobic mismatch. Thus, this
framework serves as a simple and complementary approximation to
the elastic continuous model. The automata produced protein
assemblies that matched the experimentally observed protein arrays,
suggesting that the understanding of molecular association/dissocia-
tion kinetics are sufficient to account for the equilibrium large-scale
organization. Membrane protein arrays could potentially be analyzed
from the perspective of the stability of the arrays, rather than from the
perspective of the individual protein component, as has been done for
rafts and nanodomains. The observed dynamics along the array edges
would then resemble the raft formation, size fluctuations and raft
merging processes52, and the array dynamics could be treated using an
entropic analysis, though this might necessitate large-scale imaging
comprising multiple arrays53.
Non-tetrameric AqpZ-W14A (W14A destabilizes the protomer
interface) were found in C20 bilayers, suggesting that the membrane
mechanical properties are also involved in stabilizing membrane pro-
tein oligomerization. In this regard, it is interesting to note that in
eukaryotic cells, membrane proteins are synthesized and assembled
into their native oligomeric state in the endoplasmic reticulum (ER),
which has a particularly thin membrane compared to the plasma
membrane54. Perhaps the ER membrane stabilizes oligomers due to
enhanced hydrophobic mismatch, and this may influence post-
translational modifications, sorting, and trafficking in the secretory
pathway55–57. Dynamic imaging resolves the association/dissociation of
monomers, indicative that monomeric aquaporin protomers are stable
in membranes, shedding light onto the question how membrane
proteins of complex quaternary structure may post-translationally
oligomerize after release from the ribosome-translocon complex58–60.
Here, we show for the case of AqpZ that membrane organization
can emerge from and is modulated by Brownian diffusion and a set of
physical properties of the membrane constituents (Fig. 5). Further
work is needed to test how other membrane proteins with different
oligomeric states and shapes behave in such experiments. HS-AFM of
unlabeled proteins, seeing not only single molecules of interest, but
also their complex molecular environment, and revealing their
to study
dynamics, offers unique experimental possibilities
membrane-mediated protein interactions.
Methods
Plasmid construction
Protein was expressed from a pET22-6His-TEV-Linker-AqpZ-W14A
plasmid derived from a pTrc-10His-AqpZ plasmid61. The AqpZ gene
was amplified by PCR from the pTrc plasmid and inserted in an empty
pET22-6His-TEV plasmid by restriction-ligation cloning using the EcoRI
and XhoI restriction sites62. The W14A mutation and a linker were
introduced using megaprimer based mutagenesis30,63. The linker
(sequence: SGSGSG) was inserted between the glycine of the TEV
cleavage site and the methionine on the N-terminus of the AqpZ gene.
Inserting this linker enabled His-tag cleavage by TEV protease pre-
sumably by bringing the TEV cleavage site into the aqueous
environment64. All constructions were verified by sequencing.
Protein expression
The pET22-6His-TEV-Linker-AqpZ-W14A plasmid was transformed into
E. coli competent cell strain C41 ΔompF ΔacrAB for protein
overexpression65. Cells were grown on Luria broth (LB) plates supple-
mented with 100 μg/mL ampicillin at 37 °C. Cells from a single isolated
colony were inoculated into LB media with 100 μg/mL ampicillin and
incubated at 37 °C for 15 h. The overnight culture was diluted 100-fold
into fresh LB broth and grown to an optical density at 600 nm (OD)
between 1.2 and 1.5. AqpZ expression was induced by adding 1 mM
isopropyl β-D-1-thio-galacto-pyranoside (IPTG), and cells were then
incubated at 30 °C for 3 h at 180 rpm. Cells were harvested by cen-
trifugation at 5000 g for 20 min. The cell pellet was washed with
phosphate-buffered saline (PBS) and resuspended in 1/100 culture
volume of a lysis buffer containing 50 mM Tris-HCl at pH 8.0, 100 mM
NaCl, 10 mM MgCl2, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride
(PMSF) (Merck), 0.1 mg/mL DNase I (Roche), and 0.1 mg/mL Lysozyme
(Merck). Cells were broken by three passages through a French press at
15,000 psi. Unbroken cells and debris were removed from the cell
lysate by centrifugation at 5000 g for 20 min, and then membrane
fragments were collected by centrifugation at 140,000 g for 45 min at
4 °C. The membrane pellet was then solubilized overnight at 4 °C in a
solubilization buffer containing 50 mM Tris-HCl at pH 8.0, 100 mM
NaCl, and 5% n-dodecyl-β-D-maltoside (DDM) (CliniSciences). The
insoluble material was then removed by centrifugation at 210,000 g
for 30 min.
Protein purification
AqpZ was purified from the detergent-solubilized supernatant by
nickel affinity chromatography using a 5 mL His-Trap HP column (GE
Healthcare) attached to an ÄKTA system (GE Healthcare). The column
was equilibrated with 5 column volumes (CV) of a washing buffer (W1)
containing 100 mM Tris-HCl at pH 8.0, 150 mM NaCl, 0.1% DDM, and
100 mM imidazole. After proteins were loaded onto the column, the
nonspecifically bound material was removed by washing with 5 CV of
washing buffer W1. Elution was performed with a 5 CV gradient from 0
to 100% of an elution buffer containing 100 mM Tris-HCl at pH 8.0,
150 mM NaCl, 0.1% DDM, and 500 mM imidazole. Fractions containing
AqpZ were pooled and loaded into dialysis tubing (SnakeSkin Dialysis
Tubing 3.5 kDa, Thermo scientific), and dialyzed against 100 volumes
of a dialysis buffer containing 100 mM Tris-HCl at pH 8.0 and 150 mM
NaCl for 3 h at 4 °C. The dialyzed proteins were concentrated on a 1 mL
His-Trap HP column. The column was equilibrated with 10 CV of a
washing buffer (W2; without imidazole) containing 100 mM Tris-HCl at
pH 8.0, 150 mM NaCl and 0.1% DDM. Proteins were loaded on the
column and washed with 5 CV of washing buffer W2, followed by an
elution step with 100 % elution buffer. Fractions containing the highest
AqpZ concentration were pooled and dialyzed as above. The 6-His-tag
was then removed by digestion with TEV protease. Protease was added
to the purified AqpZ at a ratio of 1:5 (w/w) and the buffer was adjusted
to contain 100 mM Tris-HCl at pH 8.0, 150 mM NaCl, 0.5 mM EDTA,
1 mM DTT, 20 % glycerol, and 0.1 % DDM. Digestion was allowed to
continue overnight at room temperature. The cleaved AqpZ was then
separated from the TEV protease by nickel affinity using a 1 mL His-
Trap HP column. The column was equilibrated with 10 CV of washing
buffer W2. After sample loading, the His-tag free AqpZ was recovered
by washing the column with 5 CV of washing buffer W2. The fractions
containing the protein were pooled and stored with 20% of glycerol
at −80 °C.
Aquaporin-Z W14A reconstitution and physisorption
Purified AqpZ W14A was solubilized in a buffer containing 100 mM
Tris-HCl at pH 7.6, 150 mM NaCl, DDM (>3 critical micelle con-
centration, CMC), and 20% glycerol (protein buffer). The lipid
mixture (1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE),
1,2-dioleoyl-sn-glycero-3-phospho-L-serine (DOPS), 1,2-dioleoyl-sn-
Nature Communications |
(2022) 13:7373
10
Article
https://doi.org/10.1038/s41467-022-35202-8
glycero-3-phosphocholine (DOPC), DOPC:DOPS:DOPE = 8:1:1, www)
was solubilized in DDM too, and supplemented to the protein at a
lipid-to-protein ratio (LPR) of 0.1, and then diluted with the protein
buffer to a final protein concentration of 0.5 mg/ml. The protein-
lipid-detergent mixture was dialyzed in cassettes (NMWL 10 kDa,
ThermoFisher Scientific) at room temperature against 1 L of protein
buffer without DDM (100 mM Tris-HCl at pH 7.6, 150 mM NaCl, and
20% glycerol) for 12 h. The proteo-liposomes were harvested from
the cassettes after dialysis (sample). The reconstitutions were
checked by negative-stain electron microscopy for the presence of
protein-packed vesicles of intermediate size (200~500 nm, Sup-
plementary Fig. 1). For experiments, the samples were diluted with
the physisorption buffer containing 100 mM Tris-HCl at pH 7.6,
150 mM KCl, and 20 mM MgCl2, of which 2 ul was deposited onto
freshly cleaved mica and incubated for 10 min for physisorption.
The excess proteo-liposomes, not physisorbed to the mica, were
rinsed with the imaging buffer containing100 mM Tris-HCl at pH
7.6 and 150 mM KCl. The physisorption was kept short, 10 min, to
assure low sample density on the mica surface.
Cryo-electron microscopy (Cryo-EM) and 2D-crystallographic
analysis
3.5 µl of solution containing 2D crystals were applied to glow-
discharged Quantifoil R1.2/1.3 grids for 1 min, blotted for 3 s and
then vitrified by plunging into liquid nitrogen-cooled liquid ethane in a
FEI Vitrobot Mark IV (FEI). Samples were transferred to an FEI Titan
Krios and 2D crystals were imaged for 2 s in 100 ms frames at a dose of
1 electron per Å per second at 22,500x in super-resolution counting
mode using a Gatan K3 direct electron detector. Images were cor-
rected for drift using whole frame and patch algorithms and Fourier
cropped using MotionCorr266. Images were unbent and the best 8
images were merged using a lattice of a = b = 95 Å and γ = 90˚ using
Focus67. The best 8 images based on merging phase residual were
merged to calculate a projection map in layer group p4212 with a 4 Å
resolution limit.
(C16),
(1,2-dimyristoleoyl-sn-glycero-3-phosphocholine (C14),
Lipid preparation
1,2-
Lipids
dipalmitoleoyl-sn-glycero-3-phosphocholine
1,2-dioleoyl-sn-
glycero-3-phosphocholine (DOPC, C18), and 1,2-dieicosenoyl-sn-gly-
cero-3-phosphocholine (C20)) purchased from Avanti polar lipids
(Supplementary Fig. 4) were solubilized in chloroform. The solubilized
lipids were dried by a nitrogen flow and further dried in a vacuum
chamber for 12 h. The dried lipids were resuspended in the imaging
buffer (100 mM Tris-HCl at pH 7.6 and 150 mM KCl). The resuspended
lipids were tip-sonicated for 2 min to obtain small unilamellar vesicles
(SUVs). SUVs were used during the lipid addition step in the HS-AFM
fluid cell for the membrane extension and fusion experiments (see
main text: Experimental design to study membrane-mediated protein
interactions).
High-speed Atomic force microscopy (HS-AFM)
HS-AFM measurements were performed with a HS-AFM (RIBM) oper-
ated in amplitude modulation mode, using lab built amplitude detec-
tors and force stabilizers68. Igor Pro version 6.37 was used for HS-AFM
data collection. In brief, we used short cantilevers (USC-F1.2-k0.15,
NanoWorld) with a nominal spring constant of 0.15 N m–1, resonance
frequency of ~0.6 MHz and a quality factor of ~1.5 in the imaging buffer
(100 mM Tris-HCl at pH 7.6 and 150 mM KCl). All data was acquired at
room temperature.
HS-AFM height spectroscopy (HS-AFM-HS)
HS-AFM-HS data was taken by disabling x- and y-scanning directly after
HS-AFM imaging, as previous described35. In this mode, the tip is
positioned at the center of the previous image with the z-feedback
loop remaining active, monitoring the molecules diffusing in the
membrane under the tip ~100 nm away from the closest AqpZ array. All
measurements were taken with a free amplitude ~3 nm and a set-point
amplitude of >90% of the free amplitude. Z-piezo data, 0 nm height
was set to the membrane surface baseline (Fig. 2f, middle), was cap-
tured with home written software and a data acquisition board with a
maximum acquisition rate of 2,000,000 samples s−1 (LabView pro-
gramming, NI-USB-6366 card, National Instruments). All data was
acquired at room temperature.
Bilayer extension
SUVs of interest (C14, C16, C18 and C20, Supplementary Fig. 4)
were diluted with the imaging buffer (100 mM Tris-HCl at pH 7.6
and 150 mM KCl) to a final lipid concentration of 1 mg/ml, of which
10 ul was added to the HS-AFM fluid chamber during HS-AFM
imaging. Continuous HS-AFM imaging directly reported bilayer
formation, extension and fusion with the membrane protein pat-
ches. Based on the low surface density, ~5% (Fig. 1f) and the low
LPR = 0.1 of the reconstituted sample, we estimated that >99% of
the lipids in each experiment are supplemented C14, C16,
C18 or C20.
Protein hydrophobic thickness determination
Protein hydrophobic thickness was determined with home written
MATLAB scripts (MatLab, Mathworks). In brief, atom/residue coordi-
nate data of the protein structure was obtained from the Protein Data
Bank (PDB). First, the coordinates were normalized so that the center
of the protein is at the origin of the coordinate system and the sym-
metry axis accords with the z-axis. Then the x and y coordinates were
converted to their polar equivalents, r and θ, so that an atom can be
characterized with three values: r, θ and z. Each pixel, p, of the 360°
‘unrolled’ structure surface plot can be characterized with two values
in space: the polar angle θ, i.e. the position in a row, and z, i.e. the
position in a column. All atoms within a region of defined size
(height: 10 Å, angle: 10°) around each pixel were considered to score
and determine its relative abundance in hydrophobic (red), hydro-
philic (blue) and aromatic (green) surface exposed residues. The
Þexp
ðδ
hydrophobic score, Rp, is calculated as: Rp =
ð(cid:2)ðri
iR= 1 if atom i belongs to a hydro-
phobic residue and 0 otherwise, and rmax is the radius of the most
exposed residue in the region. This equation gives higher scores to the
the hydrophilic score, Bp,
surface exposed residues. Similarly,
and aromatic score, Gp, are calculated by substituting δ
iB and
iG, respectively. The highest score among the three defines the pixel’s
δ
property, e.g. the pixel is a hydrophobic pixel and colored red when
the hydrophobic score is highest. For the calculation of the hydro-
phobic thickness, we only consider the hydrophilicity and hydro-
phobicity. In particular cases, e.g. OmpF used here as test protein,
aromatic residues form girdles around membrane proteins separating
rather well defined hydrophobic and hydrophilic regions. The hydro-
phobic thickness l is determined as l = Ahydrophobic/csurface, where
Ahydrophobic represents the area of the hydrophobic pixels on the
‘unrolled’ surface and csurface represents the surface width.
(cid:2)θ
Þexpð(cid:2)ðθ
i
p
σθ
iRexpð(cid:2) ðzi
ÞÞ, where δ
iR with δ
(cid:2)rmax
σ
2
r
(cid:2)zp
σ
2
z
P
Þ2
Þ2
Þ2
2
HS-AFM data analysis
HS-AFM movies were aligned, flattened, and calibrated using home
written ImageJ plugins (ImageJ, NIH). HS-AFM-HS data were analyzed
with home written MATLAB scripts, as described35. For one-bond and
two-bond events analysis, we picked and identified particles, i.e.,
cytoplasmic and extracellular proteins, in the protein array using a
home written ImageJ plugin to obtain the coordinates of the array-
bound proteins in each HS-AFM frame (time-resolved coordinates, see
Supplementary Fig. 3). The time-resolved coordinates were then ana-
lyzed with home written MATLAB scripts for event sorting, dwell time
counting, and fittings.
Nature Communications |
(2022) 13:7373
11
Article
https://doi.org/10.1038/s41467-022-35202-8
Membrane protein automata
The membrane protein automata simulations were implemented as a
custom written Python program, modified from the cellpylib
package69. The simulations were analyzed using custom written
MATLAB scripts, akin the experimental data analysis. See Supple-
mentary Note 2 for the state-update rules and other details.
We write the membrane-dependent energy change of a mem-
brane configuration rearrangement, Δψ, as (Eq. (7) in the main text):
Δψ = δn1
ψ
1 + δn2
ψ
2 + δn3
ψ
3 + δn4
ψ
4,
where δni is the change, gain or loss, of local-configuration i in the
rearrangement. We solved through numerical simulations the 2D
continuous elastic field uxy of local-configurations 1 to 4 in Fig. 3f
and determined ψ1 to ψ4. In the automata, we consider {ψ1 ψ2 ψ3 ψ4} =
ψ1ψnorm, where ψnorm = {ψ1 ψ2 ψ3 ψ4}/ψ1 relates the relative energies
of the local configurations to each other. We first used a cylindrical
protein model in the numerical simulation, which gave, on average,
ψnorm = {1.00 1.81 3.01 3.50} (Supplementary Fig. 7). Intuitively, gain
of one local-configuration 2 costs two local-configuration 1 s, and in
this scenario, configuration 2 with ψnorm = 1.81 is favored, because
ψ2−2ψ1 < 0. Accordingly, configuration 3 (3.01) is slightly unfavored,
while configuration 4 (3.50) is strongly favored. Because the protein
cross-section shape and orientation in the configurations matter for
the 2D deformation field, we modeled AqpZ using a clover-leaf-like
cross-section (Supplementary Fig. 7), based on the cryo-EM data
(Supplementary Fig. 2). The numerical simulation gave, on average,
ψnorm = {1.00 2.06 3.22 4.10}. Thus, in both models, configuration 3 is
unfavored relative to configurations 2 and 4, and therefore will be
rare at equilibrium. For example, using Eq.
the
rearrangements in Fig. 3g are Δψre1 = 0.4 and Δψre2 = −0.5 using
the cylindrical model, and Δψre1 = 0.32 and Δψre2 = 0.1 using the
clover-leaf model. In both cases, the difference between the two
rearrangements is Δψre2-Δψre1 < 0, explaining why arrays tend to
have square shape.
(7), Δψ of
For all simulations shown in Fig. 3h and supplementary Fig. 10, we
used the shape-realistic ψnorm = {1.00 2.06 3.22 4.10} and scaled the
energies by ψ/ψnorm to approximate the experimentally determined
2). The mem-
energy gain due to hydrophobic mismatch square (u0
brane protein automaton generated protein arrays that displayed
similar morphology as in the experiment (Fig. 3h). We also performed
simulations using different ψnorm favoring individual local configura-
tions and found that the assembly morphology changed dramatically
from fuzzy to square arrays (Supplementary Fig. 10d–f), or using dif-
ferent ψ/ψnorm mimicking different hydrophobic membrane mismatch
(Supplementary Fig. 10g–i).
Reporting summary
Further information on research design is available in the Nature
Portfolio Reporting Summary linked to this article.
Data availability
Data supporting the findings of this manuscript are available from the
corresponding author upon request. The source data underlying all
figures are available as a Source Data file provided with this
paper. Source data are provided with this paper.
Code availability
All codes used for data analysis may be requested from the authors.
References
1.
Singer, S. J. & Nicolson, G. L. The fluid mosaic model of the struc-
ture of cell membranes. Science 175, 720–731 (1972).
Engelman, D. M. Membranes are more mosaic than fluid. Nature
438, 578–580 (2005).
2.
3. Mouritsen, O. G. & Bloom, M. Models of lipid-protein interactions in
membranes. Annu. Rev. Biophys. Biomol. Struct. 22, 145–171 (1993).
4. Sharpe, H. J., Stevens, T. J. & Munro, S. A comprehensive compar-
5.
ison of transmembrane domains reveals organelle-specific prop-
erties. Cell 142, 158–169 (2010).
Lewis, A. H. & Grandl, J. Mechanical sensitivity of Piezo1 ion chan-
nels can be tuned by cellular membrane tension. Elife 4,
e12088 (2015).
6. Startek, J. B., Talavera, K., Voets, T. & Alpizar, Y. A. Differential
interactions of bacterial lipopolysaccharides with lipid membranes:
implications for TRPA1-mediated chemosensation. Sci. Rep. 8,
12010 (2018).
Simons, K. & Gerl, M. J. Revitalizing membrane rafts: new tools and
insights. Nat. Rev. Mol. Cell Biol. 11, 688–699 (2010).
Komura, N. et al. Raft-based interactions of gangliosides with a GPI-
anchored receptor. Nat. Chem. Biol. 12, 402–410 (2016).
Ziółkowska, N. E., Christiano, R. & Walther, T. C. Organized living:
formation mechanisms and functions of plasma membrane
domains in yeast. Trends Cell Biol. 22, 151–158 (2012).
7.
8.
9.
10. Mukherjee, S. & Maxfield, F. R. Membrane domains. Annu. Rev. Cell
Dev. Biol. 20, 839–866 (2004).
11. Schägger, H. & Pfeiffer, K. Supercomplexes in the respiratory
chains of yeast and mammalian mitochondria. EMBO J. 19,
1777–1783 (2000).
12. Letts, J. A. & Sazanov, L. A. Clarifying the supercomplex: the higher-
13.
14.
order organization of the mitochondrial electron transport chain.
Nat. Struct. Mol. Biol. 24, 800–808 (2017).
Jacobson, K., Liu, P. & Lagerholm, B. C. The lateral organization and
mobility of plasma membrane components. Cell 177,
806–819 (2019).
Jaqaman, K. & Grinstein, S. Regulation from within: the cytoskeleton
in transmembrane signaling. Trends Cell Biol. 22, 515–526 (2012).
15. Phillips, R., Ursell, T., Wiggins, P. & Sens, P. Emerging roles for lipids in
shaping membrane-protein function. Nature 459, 379–385 (2009).
16. Ursell, T., Huang, K. C., Peterson, E. & Phillips, R. Cooperative gating
and spatial organization of membrane proteins through elastic
interactions. PLoS Comput. Biol. 3, e81 (2007).
17. Reynwar, B. J. et al. Aggregation and vesiculation of membrane
proteins by curvature-mediated interactions. Nature 447,
461–464 (2007).
18. Harroun, T. A., Heller, W. T., Weiss, T. M., Yang, L. & Huang, H. W.
Theoretical analysis of hydrophobic matching and membrane-
mediated interactions in lipid bilayers containing gramicidin. Bio-
phys. J. 76, 3176–3185 (1999).
19. Aranda-Espinoza, H., Berman, A., Dan, N., Pincus, P. & Safran, S.
Interaction between inclusions embedded in membranes. Biophys.
J. 71, 648–656 (1996).
20. Derényi, I., Jülicher, F. & Prost, J. Formation and interaction of
membrane tubes. Phys. Rev. Lett. 88, 238101 (2002).
21. Grabe, M., Neu, J., Oster, G. & Nollert, P. Protein interactions and
membrane geometry. Biophys. J. 84, 854–868 (2003).
22. Kim, K. S., Neu, J. & Oster, G. Curvature-mediated interactions
between membrane proteins. Biophys. J. 75, 2274–2291 (1998).
23. Harroun, T. A., Heller, W. T., Weiss, T. M., Yang, L. & Huang, H. W.
Experimental evidence for hydrophobic matching and membrane-
mediated interactions in lipid bilayers containing gramicidin. Bio-
phys. J. 76, 937–945 (1999).
24. Chadda, R. et al. The dimerization equilibrium of a ClC Cl(-)/H(+)
antiporter in lipid bilayers. Elife 5, e17438 (2016).
25. Betaneli, V., Mücksch, J. & Schwille, P. Fluorescence correlation
spectroscopy to examine protein-lipid interactions in membranes.
Methods Mol. Biol. 2003, 415–447 (2019).
26. Gutzeit, V. A. et al. Conformational dynamics between transmem-
brane domains and allosteric modulation of a metabotropic gluta-
mate receptor. Elife 8, e45116 (2019).
Nature Communications |
(2022) 13:7373
12
Article
https://doi.org/10.1038/s41467-022-35202-8
27. Ando, T. et al. A high-speed atomic force microscope for studying
biological macromolecules. Proc. Natl Acad. Sci. USA 98,
12468–12472 (2001).
28. Ando, T., Uchihashi, T. & Scheuring, S. Filming biomolecular pro-
cesses by high-speed atomic force microscopy. Chem. Rev. 114,
3120–3188 (2014).
29. Savage, D. F., Egea, P. F., Robles-Colmenares, Y., O’Connell, J. D. &
Stroud, R. M. Architecture and selectivity in aquaporins: 2.5 a X-ray
structure of aquaporin Z. PLoS Biol. 1, E72 (2003).
30. Schmidt, V. & Sturgis, J. N. Making monomeric Aquaporin Z by
disrupting the hydrophobic tetramer interface. ACS Omega 2,
3017–3027 (2017).
31. Nagle, J. F. & Tristram-Nagle, S. Structure of lipid bilayers. Biochim
Biophys. Acta 1469, 159–195 (2000).
32. Ringler, P. et al. Structure of the water channel AqpZ from
Escherichia coli revealed by electron crystallography. J. Mol. Biol.
291, 1181–1190 (1999).
33. Scheuring, S. et al. High resolution AFM topographs of the
Escherichia coli water channel aquaporin Z. EMBO J. 18,
4981–4987 (1999).
34. Heath, G. R. et al. Localization atomic force microscopy. Nature
594, 385–390 (2021).
35. Heath, G. R. & Scheuring, S. High-speed AFM height spectroscopy
reveals µs-dynamics of unlabeled biomolecules. Nat. Commun. 9,
4983 (2018).
51. Evans, E. Probing the relation between force-lifetime-and chemistry
in single molecular bonds. Annu. Rev. Biophys. Biomol. Struct. 30,
105–128 (2001).
52. Kuzmin, P. I., Akimov, S. A., Chizmadzhev, Y. A., Zimmerberg, J. &
Cohen, F. S. Line tension and interaction energies of membrane
rafts calculated from lipid splay and tilt. Biophys. J. 88,
1120–1133 (2005).
53. Frolov, V. A., Chizmadzhev, Y. A., Cohen, F. S. & Zimmerberg, J.
“Entropic traps” in the kinetics of phase separation in multi-
component membranes stabilize nanodomains. Biophys. J. 91,
189–205 (2006).
54. Mitra, K., Ubarretxena-Belandia, I., Taguchi, T., Warren, G. & Engel-
man, D. M. Modulation of the bilayer thickness of exocytic pathway
membranes by membrane proteins rather than cholesterol. Proc.
Natl Acad. Sci. USA 101, 4083–4088 (2004).
55. Levitz, J. et al. Mechanism of assembly and cooperativity of
homomeric and heteromeric metabotropic glutamate receptors.
Neuron 92, 143–159 (2016).
56. Petaja-Repo, U. E., Hogue, M., Laperriere, A., Walker, P. & Bouvier, M.
Export from the endoplasmic reticulum represents the limiting step
in the maturation and cell surface expression of the human delta
opioid receptor. J. Biol. Chem. 275, 13727–13736 (2000).
57. Burmakina, S., Geng, Y., Chen, Y. & Fan, Q. R. Heterodimeric coiled-
coil interactions of human GABAB receptor. Proc. Natl Acad. Sci.
USA 111, 6958–6963 (2014).
36. Andersen, O. S. & Koeppe, R. E. Bilayer thickness and membrane
58. Pitonzo, D. & Skach, W. R. Molecular mechanisms of aquaporin
protein function: an energetic perspective. Annu. Rev. Biophys.
Biomol. Struct. 36, 107–130 (2007).
biogenesis by the endoplasmic reticulum Sec61 translocon. Bio-
chim. Biophys. Acta 1758, 976–988 (2006).
37. Helfrich, W. Elastic properties of lipid bilayers: theory and possible
59. Singh, S. & Mittal, A. Transmembrane domain lengths serve as
experiments. Z. Naturforsch. C 28, 693–703 (1973).
38. Ridder, A. N. et al. Importance of hydrophobic matching for spon-
taneous insertion of a single-spanning membrane protein. Bio-
chemistry 41, 4946–4952 (2002).
39. Saffman, P. G. & Delbrück, M. Brownian motion in biological
membranes. Proc. Natl Acad. Sci. USA 72, 3111–3113 (1975).
40. Yamamoto, D., Uchihashi, T., Kodera, N. & Ando, T. Anisotropic
diffusion of point defects in a two-dimensional crystal of strepta-
vidin observed by high-speed atomic force microscopy. Nano-
technology 19, 384009 (2008).
41. Ramadurai, S. et al. Influence of hydrophobic mismatch and amino
acid composition on the lateral diffusion of transmembrane pep-
tides. Biophys. J. 99, 1447–1454 (2010).
42. Guigas, G. & Weiss, M. Influence of hydrophobic mismatching on
membrane protein diffusion. Biophys. J. 95, L25–L27 (2008).
43. Haselwandter, C. A. & Phillips, R. Directional interactions and
cooperativity between mechanosensitive membrane proteins.
Europhys. Lett. 101, 68002p68001–68002p68006 (2013).
44. Kahraman, O., Koch, P. D., Klug, W. S. & Haselwandter, C. A. Bilayer-
thickness-mediated interactions between integral membrane pro-
teins. Phys. Rev. E 93, 042410 (2016).
45. Gao, J., Hou, R., Li, L. & Hu, J. Membrane-mediated interactions
signatures of organismal complexity and viral transport mechan-
isms. Sci. Rep. 6, 22352 (2016).
60. Lang, S. et al. Functions and mechanisms of the human ribosome-
translocon complex. Subcell. Biochem. 93, 83–141 (2019).
61. Borgnia, M. J., Kozono, D., Calamita, G., Maloney, P. C. & Agre, P.
Functional reconstitution and characterization of AqpZ, the E. coli
water channel protein. J. Mol. Biol. 291, 1169–1179 (1999).
62. Wahl, A., My, L., Dumoulin, R., Sturgis, J. N. & Bouveret, E. Antag-
onistic regulation of dgkA and plsB genes of phospholipid synthesis
by multiple stress responses in Escherichia coli. Mol. Microbiol 80,
1260–1275 (2011).
63. Barik, S. Site-directed mutagenesis by double polymerase chain
reaction. Mol. Biotechnol. 3, 1–7 (1995).
64. Chen, Y., Li, Q., Yang, J. & Xie, H. Promoting tag removal of a MBP-
fused integral membrane protein by TEV protease. Appl Biochem
Biotechnol. 181, 939–947 (2017).
65. Kanonenberg, K., Smits, S. H. J. & Schmitt, L. Functional recon-
stitution of HlyB, a type I secretion ABC transporter, in Saposin-A
nanoparticles. Sci. Rep. 9, 8436 (2019).
66. Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-
induced motion for improved cryo-electron microscopy. Nat.
Methods 14, 331–332 (2017).
between protein inclusions. Front Mol. Biosci. 8, 811711 (2021).
67. Biyani, N. et al. Focus: the interface between data collection and
46. klemen, B., Veronika, K.-I. & Sylvio, M. Interaction between two
cylindrical inclusions in a symmetric lipid bilayer. J. Chem. Phys.
119, 7435–7444 (2003).
47. Engel, A., Fujiyoshi, Y. & Agre, P. The importance of aquaporin water
channel protein structures. EMBO J. 19, 800–806 (2000).
48. Schmidt, V., Sidore, M., Bechara, C., Duneau, J. P. & Sturgis, J. N. The
lipid environment of Escherichia coli Aquaporin Z. Biochim. Bio-
phys. Acta Biomembr. 1861, 431–440 (2019).
49. Rawicz, W., Olbrich, K. C., McIntosh, T., Needham, D. & Evans, E.
Effect of chain length and unsaturation on elasticity of lipid bilayers.
Biophys. J. 79, 328–339 (2000).
50. Deserno, M. Fluid lipid membranes: from differential geometry to
curvature stresses. Chem. Phys. Lipids 185, 11–45 (2015).
data processing in cryo-EM. J. Struct. Biol. 198, 124–133 (2017).
68. Miyagi, A. & Scheuring, S. A novel phase-shift-based amplitude
detector for a high-speed atomic force microscope. Rev. Sci.
Instrum. 89, 083704 (2018).
69. Antunes, L. CellPyLib: A Python library for working with cellular
automata. J. Open Source Softw. 6, 3608 (2021).
Acknowledgements
The authors thank M. de la Cruz at The MSKCC Richard Rifkind Center for
assistance with data collection and the MSKCC HPC group for assis-
tance with data processing. Funding: Work in the Scheuring laboratory
was supported by grants from the National Institute of Health (NIH),
National Center for Complementary and Integrative Health (NCCIH),
Nature Communications |
(2022) 13:7373
13
Article
https://doi.org/10.1038/s41467-022-35202-8
DP1AT010874 and National Institute of Neurological Disorders and
Stroke (NINDS), R01NS110790, and by the Kavli Institute at Cornell. Work
in the Dittman laboratory was supported by a grant from the NIH, NINDS,
R01NS116747. Work in the Hite laboratory was supported in part by NIH-
NCI Cancer Center Support Grant (P30 CA008748), the Josie Robertson
Investigators Program (to R.K.H.) and the Searle Scholars Program
(to R.K.H.).
Author contributions
Y.J. and S.S. designed the study; Y.J. performed all HS-AFM experiments;
Y.J., J.D., and S.S. performed HS-AFM image processing and data ana-
lysis; Y.J. performed numerical simulation and analysis; B.T. and J.S.
performed protein expression and purification; V.S. and R.K.H. per-
formed cryo-EM imaging and data processing; Y.J., J.D., and S.S. wrote
the manuscript.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information The online version contains
supplementary material available at
https://doi.org/10.1038/s41467-022-35202-8.
Peer review information Nature Communications thanks Thomas Weikl
and the other, anonymous, reviewer(s) for their contribution to the peer
review of this work.
Reprints and permissions information is available at
http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to jur-
isdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as
long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate if
changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons license and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.
Correspondence and requests for materials should be addressed to
Simon Scheuring.
© The Author(s) 2022
Nature Communications |
(2022) 13:7373
14
| null |
10.1371_journal.pgen.1011201.pdf
| null |
The consensus sequence of Spoink as well as the sequences of the six PCR amplicons are available at https://github. com/rpianezza/Dmel-Spoink/tree/main/
|
RESEARCH ARTICLE
Spoink, a LTR retrotransposon, invaded D.
melanogaster populations in the 1990s
Riccardo PianezzaID
Robert KoflerID
1*
1,2☯, Almorò Scarpa1,2☯, Prakash Narayanan3, Sarah Signor3*,
1 Institut fu¨ r Populationsgenetik, Vetmeduni Vienna, Vienna, Austria, 2 Vienna Graduate School of Population
Genetics, Vetmeduni Vienna, Vienna, Austria, 3 Biological Sciences, North Dakota State University, Fargo,
North Dakota, United States of America
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
☯ These authors contributed equally to this work.
* [email protected] (SS); [email protected] (RK)
Abstract
OPEN ACCESS
Citation: Pianezza R, Scarpa A, Narayanan P,
Signor S, Kofler R (2024) Spoink, a LTR
retrotransposon, invaded D. melanogaster
populations in the 1990s. PLoS Genet 20(3):
e1011201. https://doi.org/10.1371/journal.
pgen.1011201
Editor: Ce´dric Feschotte, Cornell University,
UNITED STATES
Received: November 28, 2023
Accepted: February 27, 2024
Published: March 26, 2024
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pgen.1011201
Copyright: © 2024 Pianezza et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The consensus
sequence of Spoink as well as the sequences of the
six PCR amplicons are available at https://github.
com/rpianezza/Dmel-Spoink/tree/main/
During the last few centuries D. melanogaster populations were invaded by several trans-
posable elements, the most recent of which was thought to be the P-element between 1950
and 1980. Here we describe a novel TE, which we named Spoink, that has invaded D. mela-
nogaster. It is a 5216nt LTR retrotransposon of the Ty3/gypsy superfamily. Relying on
strains sampled at different times during the last century we show that Spoink invaded
worldwide D. melanogaster populations after the P-element between 1983 and 1993. This
invasion was likely triggered by a horizontal transfer from the D. willistoni group, much as
the P-element. Spoink is probably silenced by the piRNA pathway in natural populations
and about 1/3 of the examined strains have an insertion into a canonical piRNA cluster such
as 42AB. Given the degree of genetic investigation of D. melanogaster it is perhaps surpris-
ing that Spoink was able to invade unnoticed.
Author summary
Horizontal transfer of transposable elements (TE) is a major factor driving genome evolu-
tion. Yet well documented cases of such horizontal transfer events are rare. Most evidence
is indirect, relying on sequence similarity of TEs between species. Based on strains sam-
pled during the last decades we provide direct evidence that the retrotransposon Spoink
was absent in worldwide D. melanogaster populations before 1983 but present in popula-
tions after 1993. We suggest that the Spoink invasion was triggered by a horizontal transfer
from a Drosophila species of the willistoni group.
Introduction
Transposable elements (TEs) are short genetic elements that can increase in copy number
within the host genome. They are abundant in most organisms and can make up the majority
of some genomes, i.e. maize where TEs constitute 83% of the genome [1]. There are two classes
of TEs which transpose by different mechanisms—DNA transposons which replicate by
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
1 / 25
PLOS GENETICS releasedseqs. The tool LTRtoTE is available on
GitHub (https://github.com/Almo96/LTRtoTE). The
analysis performed in this work have been
documented with RMarkdown and have been
made publicly available, together with the resulting
figures, at GitHub (https://github.com/rpianezza/
Dmel-Spoink; see *.md files).
Funding: This work was supported by the National
Science Foundation Established Program to
Stimulate Competitive Research grants NSF-
EPSCoR-1826834 and NSF-EPSCoR-2032756 to
SS, and by the Austrian Science Fund (FWF) grants
P35093 and P34965 to RK. The funders had no
role in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
directly moving to a new genomic location in a ‘cut and paste’ method, and retrotransposons
which replicate through an RNA intermediate in a ‘copy and paste’ method [2–4]. From
humans to flies, more genetic variation (in bp) is due to repetitive sequences such as transpos-
able elements than all single nucleotide variants combined [5]. Although some TEs, such as R1
and R2 elements, may benefit hosts [6, 7] most TE insertions are thought to be deleterious [8,
9]. Host genomes have therefore evolved an elaborate system of suppression frequently involv-
ing small RNAs [10]. Suppression of TEs in Drosophila relies upon small RNAs termed
piRNA, which are cognate to TE sequences [11–13]. These small RNAs bind to PIWI clade
proteins and mediate the degradation of TE transcripts and the formation of heterochromatin
silencing the TE [11, 14–19]. However, while host defenses quickly adapt to new transposon
invasions, TEs can escape silencing through horizontal transfer to new, defenseless, genomes
[20–23]. This horizontal transfer allows TEs to colonize the genomes of novel species [20, 23–
26]. The first well-documented instance of horizontal transfer of a TE was the P-element,
which spread from D. willistoni to D. melanogaster [27]. Following this horizontal transfer the
P-element invaded natural D. melanogaster populations between 1950 and 1980 [28, 29]. It
was further realized that the I-element, Hobo and Tirant spread in D. melanogaster populations
earlier than the P-element, between 1930 and 1960 [29–31]. The genomes from historical D.
melanogaster specimens collected about two hundred years ago, recently revealed that Opus,
Blood, and 412 spread in D. melanogaster populations between 1850 and 1933 [21]. In total, it
was suggested that seven TEs invaded D. melanogaster populations during the last two hun-
dred years where one invasion (the P-element) was triggered by horizontal transfer from a spe-
cies of the willistoni group and six invasions by horizontal transfer from the simulans complex
[21, 27, 31–34].
It was, however, widely assumed until now that the P-element invasion, which occurred
between 1950–1980, was the last and most recent TE invasion in D. melanogaster [21, 29, 31,
35, 36]. Here we report the discovery of Spoink, a novel TE which invaded worldwide D. mela-
nogaster populations between 1983 and 1993, i.e. after the invasion of the P-element. Spoink is
a LTR retrotransposon of the Ty3/gypsy group. We suggest that the Spoink invasion in D. mel-
anogaster was triggered by horizontal transfer from a species of the willistoni group, similarly
to the P-element invasion in D. melanogaster. In a model species as heavily investigated as D.
melanogaster it is perhaps surprising that Spoink was able to invade undetected.
Materials and methods
Discovery of the recent Spoink invasion
We identified TE insertions in different long-read assemblies using RepeatMasker [37] and the
TE library from [5]. When comparing the TE composition between strains collected in the
1950’s and 1960’s [38, 39] and more recently collected strains (� 2003 [40] we noticed an ele-
ment labeled ‘gypsy-7_DEl’ which was only present in short degraded copies in the older
genomes but was present in full length copies in the more recent genomes (S1 Table).
Structure and classification of Spoink
To generate a consensus sequence of Spoink we extracted the sequence of full-length matches
of ‘gypsy-7_DEl’ plus some flanking sequences from long-read assemblies [Ten-15, RAL91,
RAL176, RAL732, RAL737, Sto-22; [40]] and made a consensus sequence by performing mul-
tiple sequence alignment (MSA) with MUSCLE (v3.8.1551) [41] and then choosing the most
abundant nucleotide in each position of the MSA with a custom Python script
(MSA2consensus).
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
2 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
The consensus sequence of the LTR was used to identify the TSD with our new tool
LTRtoTE (https://github.com/Almo96/LTRtoTE). We used LTRdigest to identify the PBS of
Spoink [42].
We picked several sequences from each of the known LTR superfamily/groups using the
consensus sequences of known TEs [2, 43] (v9.44). We performed a blastx search against the
NCBI database to identify the RT domain in the consensus sequences of the TE [44]. We then
performed a multiple sequence alignment of the amino-acid sequences of the RT domain
using MUSCLE (v3.8.1551) [41]. We obtained the xml file using BEAUti2 [45] (v2.7.5) and
generated the trees with BEAST (v2.7.5) [45]. The maximum credibility tree was built using
TreeAnnotator (v2.7.5) [45] and visualized with FigTree (v1.4.4, http://tree.bio.ed.ac.uk/
software/figtree/).
Distribution of Spoink insertions
Genes were annotated in each of the 31 genomes from [40] using the annotation of the refer-
ence genome of D. melanogaster (6.49; Flybase) and liftoff 1.6.3 [46, 47]. The 1kb regions
upstream of each gene were classified as putative promotors. The location of canonical D. mel-
anogaster piRNA clusters was determined using CUSCO, which lifts over the flanks of known
clusters in a reference genome to locate the homologous region in a novel genome [48]. The
location of Spoink insertions within genes or clusters was determined with bedtools intersect
[49]. To determine if genic insertions were shared or independent, the sequence of the inser-
tion was extracted from each genome along with an extra 1 kb of flanking sequence on each
end. Insertions purportedly in the same gene were then aligned, and if the flanks aligned they
were considered shared insertions. To determine if cluster insertions were shared the flanking
TE regions were aligned using Manna, which aligns TE annotations rather than sequences, to
determine if there was any shared synteny in the surrounding TEs [50].
Abundance of Spoink insertions in different D. melanogaster strains
We investigated the abundance of Spoink in multiple publicly available short-read data sets
[31, 40, 51–53]. These data include genomic DNA from 183 D. melanogaster strains sampled
at different geographic locations during the last centuries. For an overview of all analysed
short-read data see S5 Table. We mapped the short reads to a database consisting of the con-
sensus sequences of TEs [43] (v9.44), the sequence of Spoink and three single copy genes (rhi,
tj, RpL32) with bwa bwasw (version 0.7.17-r1188) [54]. We used DeviaTE (v0.3.8) [55] to esti-
mate the abundance of Spoink. DeviaTE estimates the copy number of a TE (e.g. Spoink) by
normalizing the coverage of the TE by the coverage of the single copy genes. We also used
DeviaTE to visualize the abundance and diversity of Spoink as well as to compute the fre-
quency of SNPs in Spoink (see below).
To identify Spoink insertions in 49 long-read assemblies of D. melanogaster strains collected
during the last 100 years we used RepeatMasker [37] (open-4.0.7; -no-is -s -nolow). For an
overview of all analysed assemblies see S6 Table [39, 40, 48, 56]. For estimating the abundance
of Spoink in the long-read assemblies we solely considered canonical Spoink insertions (> 80%
of length, < 5% sequence divergence).
Population frequency of Spoink insertions
For every putative Spoink insertion (including degraded ones) in the eight long-read assem-
blies of individuals from Raleigh [40], we extracted the sequence of the insertion plus 1 kb of
flanking sequence with bedtools [49]. The sequence of the Spoink insertion was removed with
seqkit [57] and the flanking sequences were mapped to the AKA017 genome (i.e. the common
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
3 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
coordinate system) with minimap2 allowing for spliced mappings [40, 57, 58]. The mapping
location of each read was extracted and if they overlapped between strains they were consid-
ered putative shared sites. Regions with overlapping reads were visually inspected in IGV
(v2.4.14) and if the mapping location was shared they were considered shared insertions sites
[59, 60].
PCR
To validate whether Spoink is absent in old D. melanogaster strains but present in recent strains
we used PCR. We designed two primers pairs for Spoink and one for vasa as a control. We
extracted DNA from different strains of D. melanogaster (Lausanne-S, Hikone-R, iso-1, RAL59,
RAL176, RAL737) using a high salt extraction protocol [61]. We designed two primers pairs
for Spoink (P1,P2) and one for the gene vasa (P1 FWD TCAGAAGTGGGATCGGGCTCGG,
P1 REV CAGTAGAGCACCATGCCGACGC, P2 FWD ATGGACCGTAATGGCAGCAGCG,
P2 REV ACACTCCGCGCCAGAGTCAAAC, Vasa FWD AACGAGGCGAGGAAGTTTGC,
Vasa REV GCGATCACTACATGGCAGCC). We used the following PCR conditions: 1 cycle
of 95˚C for 3 minutes; 33 cycles of 95˚C for 30 seconds, 58˚C for 30 seconds and 72˚C for 20
seconds; 1 cycle of 72˚C for 6 minutes.
Small RNAs
To identify piRNAs complementary to Spoink we analysed the small-RNA data from 10 GDL
strains [62]. The adaptor sequence GAATTCTCGGGTGCCAAGG was removed using cuta-
dapt (v4.4 [63]). We filtered for reads having a length between 18 and 36nt and aligned the
reads to a database consisting of D. melanogaster miRNAs, mRNAs, rRNAs, snRNAs, snoR-
NAs, tRNAs [64], and TE sequences [43] with novoalign (v3.09.04). We used previously devel-
oped Python scripts [65] to compute ping-pong signatures and to visualize the piRNA
abundance along the sequence of Spoink.
UMAP
We used the frequencies of SNPs in the sequence of Spoink to compute the UMAP. This fre-
quencies reflect the Spoink composition in a given sample. For example if a specimen has 20
Spoink insertions and a biallelic SNP with a frequency of 0.8 at a given site in Spoink than
about 16 Spoink insertions will have the SNP and 4 will not have it. The frequency of the
Spoink SNPs was estimated with DeviaTE [55]. Solely bi-allelic SNPs were used and SNPs only
found in few samples were removed (�3 samples). UMAPs were created in R (umap package;
v0.2.10.0 [66]).
Origin of horizontal transfer
To identify the origin of the horizontal transfer of Spoink we used RepeatMasker [37] (open-
4.0.7; -no-is -s -nolow) to identify sequences with similarity to Spoink in the long-read assem-
blies of 101 drosophilid species and in 99 different insect species [67, 68] (S8 Table). We
included the long-read assembly of the D. melanogaster strain RAL737 and of the D. simulans
strain SZ129 in the analysis [23, 40]. We used a Python script to identify in each assembly the
best hit with Spoink (i.e. the highest alignment score) and than estimated the similarity
between this best hit and Spoink. The similarity was computed as s = rmsbest/rmsmax, where
rmsbest is the highest RepeatMasker score (rms) in a given assembly and rmsmax the highest
score in any of the analysed assemblies. A s = 0 indicates no similarity to the consensus
sequence of Spoink whereas s = 1 represent the highest possible similarity. To generate a
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
4 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
phylogenetic tree we identified Spoink insertions in the assemblies of the 101 drosophilid spe-
cies and RAL737 using RepeatMasker. We extracted the sequences of full-length insertions
(> 80% of the length) from species having at least one full-length insertion using bedtools [49]
(v2.30.0). A multiple sequence alignment of the Spoink insertions was generated with MUS-
CLE (v3.8.1551) [41] and a tree was generated with BEAST (v2.7.5) [45].
Results
Previous work showed that at least seven TE families invaded D. melanogaster populations
during the last two hundred years [21, 29, 31]. To explore whether additional, hitherto poorly
characterised TEs could have invaded D. melanogaster, we investigated long-read assemblies
of recently collected D. melanogaster strains [40] using a newly assembled repeat library [5].
Interestingly we found differences in the abundance of “gypsy_7_DEl” between the reference
strain Iso-1 and more recently collected D. melanogaster strains (S1 Table). To better charac-
terize this TE, we generated a consensus sequence based on the novel insertions and checked if
this consensus sequence matches any of the repeats described in repeat libraries generated for
D. melanogaster and related species [5, 40, 43, 69, 70]. A fragmented copy of this TE, with just
one of the two LTRs being present, was reported by [40] (0.13% divergence; “con41_-
UnFmcl001_RLX-incomp”; S2 Table). The next best hits were gypsy7 Del, gypsy2 DSim, micro-
pia and Invader6 (18–30% divergence; S2 Table). Given this high sequence divergence from
previously described TE families and the fact that this novel TE belongs to an entirely different
superfamily/group than gypsy7 (see below), we decided to give this TE a new name. We call
this novel TE “Spoink” inspired by a Poke´mon that needs to continue jumping to stay alive.
Spoink is an LTR retrotransposon with a length of 5216 bp and LTRs of 349 bp (Fig 1A; for
coordinates of the analysed insertions see S3 Table). At positions 4639–4700 Spoink contains a
poly-A tract, which length may differ by a few bases between insertions. Spoink encodes a 695
aa putative gag-pol polyprotein. Ordered from the N- to the C-terminus, the conserved
domains of the polyprotein are: reverse transcriptase of LTR (e-value = 2.2e − 59; CDD v3.20
[71]), RNase HI of Ty3/gypsy elements (e-value = 1.65e − 48;) and integrase zinc binding
domain (e-value = 4.81e − 16). Spoink lacks an env. The order of these domains, with the inte-
grase downstream of the reverse transcriptase, is typical for Ty3/gypsy transposons [72].
A phylogeny based on the reverse transcriptase domain of different TE families suggests
that Spoink is a member of the gypsy/mdg3 superfamily/group of LTR retrotransposons (Fig
1B; [2]). As expected for members of the Ty3/gypsy superfamily, Spoink generates a target site
duplication of 4 bp and it has an insertion motif enriched for ATAT (Fig 1A; [2, 73]). A gag-
pol polyprotein as encoded by Spoink was observed for some Ty3/gypsy transposons [74, 75]
but not for others [72]. However, Spoink differs from what is expected for the Ty3/gypsy super-
family in two ways. First, the predicted primer binding site of Spoink directly follows the LTR,
whereas typically for Ty3/gypsy there is a shift of 5–8nt (Fig 1A; [2]). Second, the LTR motif is
TG. . .TA which is different from the TG. . .CA motif usually reported for gypsy TEs [2].
Finally we investigated the genomic distribution of Spoink insertions in long-read assem-
blies of D. melanogaster strains collected � 2003 [40]. In total, these assemblies contains 481
full-length (> 80% length with at least one LTR) insertions of Spoink (on the average 16 per
genome). Unlike the P-element which has a strong insertion bias into promoters, Spoink inser-
tions are mostly found in introns and intergenic regions (S1 Fig). 54% of the Spoink insertions
are in 201 different genes. Interestingly we found 7 independent Spoink insertions in Myo83F.
To summarize we characterized a novel LTR-retrotransposon of the Ty3/gypsy superfamily
in the genome of D. melanogaster that we call Spoink.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
5 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
Fig 1. Spoink is a novel TE of the Ty3/gypsy superfamily. A) Overview of the composition of Spoink. Features are shown in color and the alignments show the
sequences around the main features of Spoink for two insertions in each of three different long-read assemblies of D. melanogaster. B) Phylogenetic tree based on the
reverse-transcriptase domain of pol for Spoink and several other LTR retrotransposons. Multiple families have been picked for each of the main superfamilies/groups
of LTR transposons [2]. Our data suggest that Spoink is a member of the gypsy/mdg3 group.
https://doi.org/10.1371/journal.pgen.1011201.g001
Spoink recently invaded worldwide D. melanogaster populations
To substantiate our hypothesis that Spoink recently invaded D. melanogaster we used three
independent approaches: Illumina short read data, long-read assemblies, and PCR/Sanger
sequencing. First we aligned short reads from a strain collected in 1958 (Hikone-R) and a strain
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
6 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
Fig 2. Spoink invaded D. melanogaster. A) DeviaTE plots of Spoink for a strain collected in 1954 (Hikone-R) and a strain collected in 2015
(Ten-15). Short reads were aligned to the consensus sequence of Spoink and the coverage was normalized to the coverage of single-copy genes.
The coverage based on uniquely mapped reads is shown in dark grey and light grey is used for ambiguously mapped reads. Single-nucleotide
polymorphisms (SNPs) and small internal deletions (indels) are shown as colored lines. The coverage was manually curbed at the poly-A track
(between dashed lines). B) Insertions with a similarity to the consensus sequence of Spoink in the long-read assemblies of Oregon-R (collected
around 1925) and the more recently collected strain RAL737 (2003). C) PCR results for two Spoink primer pairs (for location of primers see
sketch at bottom) and one primer pair for the gene vasa. Spoink is absent in old strains (Lausanne-S, Hikone-R and Iso-1) and present in more
recently collected strains (RAL59, RAL176, RAL737). D) Population frequency of Spoink insertions in long-read assemblies of strains collected
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
7 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
in 2003 from Raleigh [40]. Note that highly diverged insertions are largely segregating at a high frequency while canonical Spoink insertions
mostly segregate at a low frequency.
https://doi.org/10.1371/journal.pgen.1011201.g002
collected in 2015 (Ten-15) [31, 40] to the consensus sequence of Spoink using DeviaTE [55].
DeviaTE estimates the abundance of Spoink insertions by normalizing the coverage of Spoink
to the coverage of a sample of single-copy genes. Furthermore, DeviaTE is useful for generat-
ing an intuitive visualization of the abundance and composition (i.e. SNPs, indels, truncations)
of Spoink in samples. We found that only a few degraded reads aligned to Spoink in the 1950’s
strain (Hikone-R) whereas many reads covered the sequence of Spoink in the more recently
collected strain Ten-15 (Fig 2A). There were also very few SNPs or indels in the recently col-
lected strain suggesting that most insertions have a very similar sequence (Fig 2A). This obser-
vation holds true when multiple old and young D. melanogaster strains are analysed (S2 Fig).
Next we investigated the abundance of Spoink in long-read assemblies of a strain collected
in 1925 (Oregon-R) and a strain collected in 2003 (RAL737). We found solely highly diverged
and fragmented copies of sequences with similarity to Spoink in Oregon-R (Fig 2B). These
degraded fragments were mostly found near the centromeres of Oregon-R. Investigating the
identity of these degraded fragments of Spoink in more detail we found that they largely match
with short and highly diverged fragments of Invader6, micropia and the Max-element (S4
Table). In addition to these degraded fragments, the more recently collected strain RAL737
also carries a large number of full-length insertions with a high similarity to the consensus
sequence of Spoink (henceforth canonical Spoink insertions; Fig 2B). The canonical Spoink
insertions are distributed all over the chromosomes of RAL737 (Fig 2B). This observation is
again consistent when several long-read assemblies of old and young D. melanogaster strains
are analysed (S3 Fig).
Finally we used PCR to test whether Spoink recently spread in D. melanogaster. We
designed two PCR primer pairs for Spoink and, as a control, one primer pair for vasa (Fig 2C;
bottom panel). The Spoink primers amplified a clear band in three strains collected 2003 in
Raleigh but no band was found in earlier collected strains, including the reference strain of D.
melanogaster, Iso-1 (Fig 2C). We sequenced the fragments amplified by the Spoink primers
using Sanger sequencing and found that the sequence of the six amplicons matches with the
consensus sequence of Spoink (S4 Fig).
Finally we investigated the population frequency of canonical and degraded Spoink inser-
tions. Using the long-read assemblies of eight strains collected in 2003 in Raleigh we computed
the population frequency of different Spoink insertions. We found that canonical Spoink inser-
tions (< 5% divergence) are largely segregating at a low population frequency, as expected for
recently active TEs (Fig 2D). While several degraded fragments that were annotated as Spoink
are private, there were many at a higher population frequency as expected for older sequences
(Fig 2D).
In summary our data suggest that Spoink recently spread in D. melanogaster and that
degraded fragments with some similarity to Spoink are present in heterochromatic regions of
the centromeres of all investigated D. melanogaster strains. These degraded fragments may be
the remnants of more ancient invasions of TEs sharing some sequence similarity with Spoink.
Timing the Spoink invasion
Next we sought to provide a more accurate estimate of the time when Spoink spread in D. mel-
anogaster. First we generated a rough timeline of the Spoink invasion using D. melanogaster
strains sampled during the last two hundred years. We estimated the abundance of Spoink in
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
8 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
Fig 3. Spoink invaded D. melanogaster between 1983 and 1993 after the invasion of the P-element. A) Rough
timeline of the Spoink and P-element invasion based on different strains sampled during the last two hundred years.
The numbers represent the estimated copy number of Spoink and P-element based on DeviaTE. B) Timeline of the
Spoink and P-element invasion based on 183 strains sampled between 1960 and 2015. The intensity of the color varies
due to overlapping dots C) Abundance of canonical Spoink insertions (> 80% length and < 5% divergence) in long-
read assemblies of D. melanogaster strains collected between 1925 and 2018.
https://doi.org/10.1371/journal.pgen.1011201.g003
these strains using DeviaTE [55]. As reference we also estimated the abundance of the P-ele-
ment, which is widely assumed as to be the most recent TE that invaded D. melanogaster popu-
lations [28, 31]. Spoink was absent from all strains collected �1983 but present in strains
collected �1993 (Fig 3A). By contrast our data suggest that the P-element was absent in the
strains collected � 1962 but present in strains collected �1967 (Fig 3A). This is consistent
with previous works suggesting that the P-element invaded D. melanogaster between 1950 and
1980 [21, 29, 35, 36]. Our data thus suggest that Spoink invaded D. melanogaster after the P-ele-
ment invasion. To investigate the timing of the invasion in more detail we estimated the abun-
dance of Spoink in short-read data of 183 strains collected between 1960 and 2015 from
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
9 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
different geographic regions using DeviaTE (S5 Table; data from [31, 40, 51–53]). The analysis
of these 183 strains supports the view that Spoink was largely absent in strains collected � 1983
but present in strains collected � 1993 (Fig 3B). However there are two outliers. Spoink is pres-
ent in one strain collected in 1979 in Providence (USA), which could be due to a contamina-
tion of the strain. On the other hand Spoink is absent in one strain collected in 1993 in
Zimbabwe (Fig 3B). As Spoink was present in six other strains collected in 1993 from Zimba-
bwe, it is feasible that Spoink was still spreading in populations from Zimbabwe around 1993.
The strains supporting the absence of Spoink prior to 1983 were collected from Europe, Amer-
ica, Asia and Africa while the strains supporting the presence of Spoink after 1993 were col-
lected from all five continents (S5 Table).
Finally we estimated the abundance of Spoink in 49 long-read assemblies of strains collected
during the last 100 years (S6 Table; [39, 40, 48, 56]). We used RepeatMasker [37] to estimate the
abundance of canonical Spoink insertions (> 80% length and < 5% divergence) in these strains.
Canonical Spoink insertions were absent in strains collected before 1975 but present in all long-
read assemblies of strains collected after 2003 (Fig 3C). The strains of the assemblies supporting
the absence of canonical Spoink insertions were collected from America, Europe, Asia, and
Africa whereas the strains showing the presence of Spoink were largely collected from Europe,
though genomes from North America and Africa are also represented (S6 Table).
In summary we conclude that Spoink invaded worldwide populations of D. melanogaster
approximately between 1983 and 1993. Moreover, the Spoink invasion is more recent than the
P-element invasion.
Geographic heterogeneity in the Spoink sequence variation
Previous work showed that the composition of TEs within a species may differ among geo-
graphic regions [21, 31]. Such geographic heterogeneity could result from founder effects
occurring during the geographic spread of a TE. For example, a TE spreading in a species with
a cosmopolitan distribution such as D. melanogaster may need to overcome geographic obsta-
cles such as oceans and deserts. The few individuals that overcome these obstacles, thereby
spreading the TE into hitherto naive populations, may carry slightly different variants of the
TE than the source populations. These distinct variants will then spread in the new population.
Such founder effects during the invasion may lead to a geographically heterogeneous composi-
tion of a TE within a species. For example, for the retrotransposon Tirant, individuals sampled
from Tasmania carry distinct variants [31], while for 412 and Opus individuals from Zimba-
bwe are distinct from the other populations [21]. To investigate whether we find such geo-
graphic heterogeneity we analysed the Spoink composition in the Global Diversity Lines
(GDL), which comprise 85 D. melanogaster strains sampled after 1988 from five different con-
tinents (Africa—Zimbabwe, Asia—Beijing, Australia—Tasmania, Europe—Netherlands,
America—Ithaca; [51]). Except for a single strain from Zimbabwe all GDL strains harbour
Spoink insertions (S5 Fig). We estimated the allele frequency of SNPs in Spoink, where a SNP
refers to a variant among dispersed copies of Spoink. The allele frequency estimate thus reflects
the composition of Spoink within a particular strain. To summarize differences in the compo-
sition among the GDL strains we used UMAP [76]. We found that the composition of Spoink
varies among regions where three distinct groups can be distinguished: Tasmania, Bejing/Ith-
aca and Netherlands/Zimbabwe (S5 Fig). It is interesting that clusters are formed by geograph-
ically distant populations such as Bejing (Asia) and Ithaca (America). We speculate that
human activity, where flies might for example hitchhike with merchandise, could be responsi-
ble for this pattern. In summary, we found a geographically heterogeneous composition of
Spoink which is likely due to founder effects occurring during the spread of this TE.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
10 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
Fig 4. A piRNA based defence against Spoink emerged in D. melanogaster A) piRNAs mapping to Spoink in a strain sampled 1938 (Lausanne-S) and
2004 (I17). The transposon HMS Beagle is included as reference. Solely the 5’ positions of piRNAs are shown and the piRNA abundance is normalized
to one million piRNAs. Sense piRNAs are shown on the positive y-axis and antisense piRNAs on the negative y-axis. B) Ping-pong signature for the
piRNAs mapping to Spoink and HMS Beagle in the D. melanogaster strain I17 (2004).
https://doi.org/10.1371/journal.pgen.1011201.g004
Spoink is silenced by the piRNA pathway in natural populations
The host defence against TEs in Drosophila is based on small RNAs termed piRNAs. These
piRNAs bind to PIWI clade proteins and silence a TE at the transcriptional as well as the post-
transcriptional level [11, 12, 14, 77]. To test whether Spoink is silenced in D. melanogaster pop-
ulations we investigated small RNA data from the GDL lines [62]. Small RNA were sequenced
for 10 out of the 84 GDL lines such that two strains were picked from each of the five conti-
nents [62].
We find piRNAs mapping along the sequence of Spoink in the GDL strain I17 which was
collected in 2004 but not in the strain Lausanne-S which was sampled around 1938 (Fig 4A;
[78]). piRNAs mapping to Spoink were further found for all 10 GDL strains (S6 Fig).
An important feature of germline piRNA activity in D. melanogaster is the ping-pong cycle
[11, 12]. An active ping-pong cycle generates a characteristic overlap between the 5’ positions
of sense and antisense piRNAs, i.e. the ping-pong signature. Computing a ping-pong signature
thus requires several overlapping sense and antisense piRNAs. Since the amount of piRNAs
was too low we could not compute a ping-pong signature for the strain Lausanne-S (collected
in 1938; see above). However we found a pronounced ping-pong signature in all 10 GDL sam-
ples (Fig 4B and S6 Fig).
It is an important open question as to which events trigger the emergence of piRNA based
host defence. The prevailing view, the trap model, holds that the piRNA based host defence is
initiated by a copy of the TE jumping into a piRNA cluster [17, 25, 79–81]. If this is true we
expect Spoink insertions in piRNA clusters in each of the long-read assemblies of the recently
collected D. melanogaster strains [40]. We identified the position of piRNA clusters in these
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
11 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
long-read assemblies based on unique sequences flanking the piRNA clusters [48]. Interest-
ingly, we found an extremely heterogeneous abundance of Spoink insertions in piRNA clus-
ters, where some strains (e.g. RAL176) have up to 14 cluster insertions whereas 18 out of 31
strains did not have a single cluster insertion (S7 Table). Three of the cluster insertions were
into 42AB, which usually generates the most piRNAs [11, 69]. It is an important open question
whether such a heterogeneous distribution of Spoink insertions in piRNA clusters is compati-
ble with the trap model [82, 83]. In summary we found evidence that Spoink is silenced by the
piRNA pathway but the number of Spoink insertions in piRNA clusters is very heterogeneous
among strains.
Origin of Spoink
The invasion of Spoink in D. melanogaster was likely triggered by horizontal transfer from a
different species. To identify the source of the horizontal transfer we investigated the long-
read assemblies of 101 Drosophila species [67] (and D. simulans strain SZ129) and of 99 insect
species [23, 67, 68] (S8 Table). We did not consider short-read assemblies, as TEs may be
incompletely represented in them [48]. Apart from D. melanogaster we found insertions with a
high similarity to Spoink in D. sechellia, in one out of two D. simulans assemblies (in SZ129 but
not in 006), and species of the willistoni group, in particular D. willistoni (Fig 5A). In agree-
ment with this, a sequence from D. willistoni with a high similarity to Spoink can be found in
RepBase (Gypsy-78_DWil; I: 99.73% similarity, LTR: 93.54% similarity [84]). Spoink insertions
with a somewhat smaller similarity were found in D. cardini and D. repleta. No sequences sim-
ilar to Spoink were found in the 99 insect species (S7 Fig). To further shed light on the origin
of the Spoink invasion we constructed a phylogenetic tree with full-length insertions of Spoink
in D. melanogaster, D. sechellia, D. simulans (SZ129) D. cardini and species of the willistoni
group (Fig 5B and for a star phylogeny see S8 Fig). We did not find a full-length insertion of
Spoink in D. repleta. This tree reveals that Spoink insertions in D. sechellia and D. simulans
have very short branches. Furthermore, in D. simulans just one out of the two analysed assem-
blies has Spoink insertions. We thus suggest that the Spoink invasion in these two species is
also of recent origin (manuscript in preparation).
However, Spoink insertions in D. melanogaster are nested within insertions from species of
the willistoni group (Fig 5B). Our data thus suggest that, similar to the P-element invasion in
D. melanogaster [27], the Spoink invasion in D. melanogaster was also triggered by horizontal
transfer from a species of the willistoni group. The synonymous divergence of Spoink is lower
than for any of 140 single copy orthologous genes shared between D. melanogaster and D. will-
istoni, further supporting the recent horizontal transfer of Spoink (S9 Fig) [20, 85, 86]. Species
of the willistoni group are Neotropical, occurring throughout Central and South America [87–
89]. Therefore horizontal transfer of Spoink only became feasible after D. melanogaster
extended its habitat into the Americas approximately 200 years ago [90–92]. Insertions of D.
cardini are next to species of the willistoni group, suggesting that D. cardini also acquired
Spoink by horizontal transfer from the willistoni group, likely independent of D. melanogaster
(Fig 5B). D. cardini is also a Neotropical species and its range overlaps many species of the will-
istoni group, thus horizontal transfer between the species is physically feasible [93, 94].
In summary, similarly to the P-element, horizontal transfer from a species of the willistoni
group likely triggered the Spoink invasion in D. melanogaster.
Discussion
Here we suggest that the LTR-retrotransposon Spoink invaded D. melanogaster populations
between 1983 and 1993, after the spread of the P-element. Similarly to the P-element, the
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
12 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
Fig 5. The Spoink invasion in D. melanogaster was likely triggered by a horizontal transfer from a species of the willistoni group. A) Similarity of
TE insertions in long-read assemblies of diverse drosophilid species to Spoink. The barplots show for each species the similarity between Spoink and the
best match in the assembly. For example, a value of 0.9 indicates that at least one insertion in an assembly has a high similarity (� 90%) to the
consensus sequence of Spoink. B) Bayesian tree of Spoink insertions in the different drosophilid species. Only full-length insertions of Spoink (> 80% of
the length) were considered. Node support values are posterior probabilities estimated by BEAST [45]. Note that Spoink insertions of D. melanogaster
are nested in insertions from the willistoni group (blue shades).
https://doi.org/10.1371/journal.pgen.1011201.g005
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
13 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
Spoink invasion was likely triggered by horizontal transfer from a species in the willistoni
group. Horizontal transfer of a TE is usually inferred from three lines of evidence: i) a patchy
distribution of the TE among closely related species, ii) a phylogenetic discrepancy between
the TE and the host species and iii) a high similarity between the TE of the donor and recipient
species, which is frequently quantified by the synonymous divergence of the TE [95, 96]. All of
these three lines of arguments support a horizontal transfer of Spoink in D. melanogaster, with
a species of the willistoni group being the likely donor. First we found a patchy distribution
among species of the melanogaster group (for D. simulans we even have a patchy distribution
among different strains; Fig 5A). Second Spoink insertions of D. melanogaster (and other spe-
cies that may have gotten Spoink recently) are nested within species of the willistoni group (Fig
5B), a clear phylogenetic discrepancy. Third we found that the synonymous divergence of
Spoink is lower than for all orthologous genes in D. melanogaster and D. willistoni (S9 Fig). In
addition to this classical but indirect lines of evidence, we have however more direct and thus
more compelling evidence for the horizontal transfer of Spoink. Based on strains collected dur-
ing the last hundred years from all major geographic regions we showed that Spoink insertions
were absent in all strains collected before 1983 but present in all strains collected after 1993
(using Illumina short read data, long-read assemblies, and PCR/Sanger sequencing). This
makes Spoink one of the best documented cases of a recent horizontal transfer of a TE, simi-
larly to the P-element where also strains collected during the last 100 years support the recent
horizontal transfer [28, 29].
The abundance of sequencing data from strains collected at different time points during the
last century allowed us to pinpoint the timing of the invasion in a way that would not have
been previously possible. Spoink appears to have rapidly spread throughout global populations
of D. melanogaster between 1983 and 1993. The narrow time-window of 10 years is plausible
as studies monitoring P-element invasions in experimental populations showed that the P-ele-
ment can invade populations within 20–60 generations [65, 97, 98]. Assuming that natural D.
melanogaster populations have about 15 generations per year [99], a TE could penetrate a nat-
ural D. melanogaster population within 1–3 years. Given this potential rapidness of TE inva-
sions it is likely that Spoink spread quickly between 1983 and 1993. Since there is a gap
between strains sampled at 1983 and 1993 we cannot further narrow down the timing of the
invasion. Furthermore, the strains used for timing the invasions were sampled from diverse
geographic regions and Spoink likely spread at different times in different geographic regions.
If horizontal transfer from a willistoni species triggered the invasion, as suggested by our data,
then Spoink will have first spread in D. melanogaster populations from South America (the
habitat of willistoni species), followed by populations from North America and the other conti-
nents. It is also feasible that Spoink invaded D. melanogaster indirectly, for example using D.
simulans as intermediate host, in which case the Spoink invasion in D. melanogaster may have
been triggered in almost any geographic region (both, D. simulans and D. melanogaster, are
cosmopolitan species [100]). Unfortunately, we cannot infer the timing of the geographic
spread of the Spoink invasion in different continents as D. melanogaster strains were not sam-
pled sufficiently densely from different regions. Our work thus highlights the importance of
efforts such as DrosEU, GDL and DrosRTEC to densely sample Drosophila strains in time and
space [51, 101, 102]. It is also interesting to ask as to which extent human activity (e.g. traffick-
ing of goods) contributed to the rapid spread of Spoink. Given that our analysis of the Spoink
composition shows that geographically distant populations (Bejing/Ithaca or Netherlands/
Zimbabwe) cluster together, human activity may have played a role. Increasing human activity
could also explain why Spoink (invasion 1983–1993) seems to have spread faster than the P-
element (1950–1980).
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
14 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
Our investigation of Spoink insertions in different drosophilid species suggests that the
Spoink invasion in D. melanogaster was triggered by horizontal transfer from a species of the
willistoni group. Although it is possible that we did not analyse the true donor species, we con-
sider it unlikely to be a species outside of the willistoni group given the wide distribution of
Spoink in all species in the willistoni group. In addition, the phylogenetic tree of Spoink has
deep branches within the willistoni group, suggesting that Spoink is ancestral in this group
(S10 Fig).
A related open question is when Spoink first entered D. melanogaster populations. Since a
TE may initially solely spread in some isolated subpopulations there could be a considerable
lag time between the horizontal transfer of a TE and its spread in worldwide population. The
presence of Spoink in a strain collected around 1979 in Providence (USA; Fig 3B) could be due
to this lag time (or contamination). Nevertheless, the horizontal transfer of Spoink must have
happened between the spread of D. melanogaster into the habitat of the willistoni group, about
200 years ago, and the invasion of Spoink in worldwide populations between 1983 and 1993. In
addition to the P-element, Spoink is the second TE that invaded D. melanogaster populations
following horizontal transfer from a species of the willistoni group. Species from the willistoni
group are very distantly related with D. melanogaster (about 100my [103]) and we were thus
wondering whether it is a coincidence that a species of the willistoni group is again acting as
donor of a TE invasion in D. melanogaster. The recent habitat expansion of D. melanogaster
into the Americas resulted in novel contacts with many species, in addition to species of the
willistoni group, that might have acted as donors of novel TEs such as D. pseudoobscura or D.
persimilis [104]. Why is again a species of the willistoni group and not one of these other spe-
cies acting as donor of a novel TE? Apart from mere chance, there are several, not mutually
exclusive, hypotheses for this observation. First, it is feasible TEs of the willistoni group are
exceptionally compatible with D. melanogaster at a molecular level. Second, some parasites tar-
geting both D. melanogaster and species of the willistoni group could be efficient vectors for
horizontal transfer of TEs. Third, the physical contact between D. melanogaster and some spe-
cies of the willistoni group might be unusually tight, facilitating horizontal transfer of TEs by
an unknown vector. D. willistoni is a common drosophilid in South American forests [105].
Habitat fragmentation caused by human deforestation may thus generate intensive contacts
between human commensal species, such as D. melanogaster, and abundant forest species like
D. willistoni. Fourth, species of the willistoni group might be exceptionally numerous resulting
in elevated probability for horizontal transfer of a TE.
The Spoink invasion is the eighth TE invasion in D. melanogaster that has occurred during
the last 200 year. As we argued previously, such a high rate of TE invasions is likely unusual
during the evolution of the D. melanogaster lineage since the number of TE families in D. mel-
anogaster is much smaller than what would be expected if this rate of invasions would persist
[21]. It is possible that the high rate of TE invasions continues beyond the past 200 years since
many LTR transposons in D. melanogaster are likely of very recent origin
(possibly < 16.000years [85, 106]). One possible explanation for this high rate of recent TE
invasions is that human activity contributed to the habitat expansion of D. melanogaster. Due
to this habitat expansion D. melanogaster spread into the habitat of D. willistoni which enabled
the horizontal transfer of Spoink. This raises the possibility that other species with recent habi-
tat expansions also experienced unusually high rates of TE invasions. It is also interesting to
ask whether the rate of TE invasions differs among species. For example cosmopolitan species,
such as D. melanogaster, may generally experience higher rates of horizontal transfer than
more locally confined species. The cosmopolitan distribution will bring species into contact
with many diverse species, thereby increasing the opportunities for horizontal transfer of a TE.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
15 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
The Spoink invasions also opens up several novel opportunities for research. First, the
broad availability of strains with and without Spoink will enable testing whether Spoink activity
induces phenotypic effects, similarly to hybrid dysgenesis described for the P-element, I-ele-
ment and hobo, but not for Tirant [31, 107–109]. Second, it will be interesting to investigate
whether some Spoink insertions participated in rapid adaptation of D. melanogaster popula-
tions, similar to a P-element insertion which contribute to insecticide resistance [110]. Third,
it will enable studying Spoink invasions in experimental populations, shedding light on the
dynamics of TE invasions, much as other recent studies investigating the invasion dynamics of
the P-element [97, 98, 111]. Fourth, investigation into the distribution of species that have
been infected with Spoink will shed light on the networks of horizontal transfer in drosophilid
species. Fifth, the Spoink invasion provides an opportunity to study the establishment of the
piRNA-based host defence [similar to [24, 65]]. For example we found that none of the piRNA
cluster insertions are shared between individuals, suggesting there is no or solely weak selec-
tion for piRNA cluster insertions. Furthermore we found an extremely heterogeneous abun-
dance of Spoink insertions in piRNA clusters where we could not find a single cluster
insertions of Spoink in several strains. It is an important open question whether such a hetero-
geneous distribution is compatible with the trap model [83]. One possibility is that a few clus-
ter insertions in populations are sufficient to trigger the paramutation of regular (non-
paramutated) Spoink insertions into piRNA producing loci [16, 112, 113]. These paramutated
Spoink insertions may then compensate for the low number of Spoink insertions in piRNA-
clusters [112]. Paramutations could thus explain why several studies found that stand-alone
insertions of TEs can nucleate their own piRNA production [69, 83, 114, 115].
The war between transposons and their hosts is constantly raging, with potentially large fit-
ness effects for the individuals in populations. Over the last two hundred years there have been
at least eight invasions of TEs into D. melanogaster, each of which could disrupt fertility for
example by inducing some form of hybrid dysgenesis. TEs are responsible for > 80% of visible
spontaneous mutations in D. melanogaster, and produce more variation than all SNPs com-
bined [116–118]. In the long read assemblies considered here, more than half of insertions of
Spoink were into genes [40]. The recent Spoink invasion could thus have a significant impact
on the evolution of D. melanogaster lineage.
Supporting information
S1 Fig. Abundance of Spoink and P-element insertions in different genomic features. TE
insertions were identified in 31 long-read assemblies of D. melanogaster [40] and the reference
annotation was lifted to each assembly with liftoff [46, 47]. Note that the P-element has a pro-
nounced insertion bias in promoters (defined as 1000bp upstream of the first exon) whereas
Spoink insertions are largely found in introns and intergenic regions.
(AI)
S2 Fig. DeviaTE plots of six D. melanogaster strains collected during the last century. The
short reads were aligned to the consensus sequence of Spoink and the coverage was normalized
to the the coverage of single-copy genes. The coverage was manually curbed at the poly-A
track (indicated by dashed lines). Note that very few reads of old strains (� 1975) align to
Spoink whereas a contiguous coverage of reads along Spoink is observed for more recently col-
lected strains (� 1993).
(AI)
S3 Fig. Abundance of Spoink insertions in six long-read assemblies of D. melanogaster
strains collected during the last century. Note that all strains contain fragmented and
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
16 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
diverged insertions of Spoink, while solely recently collected strains (�2003) contain canonical
Spoink insertions (i.e. full-length insertions with little divergence from the consensus
sequence).
(AI)
S4 Fig. The Sanger sequence of the six PCR amplicons matches with the consensus
sequence of Spoink. The Sanger sequences of the amplicons of P1 (red) and P2 (green) have
been aligned to the consensus sequence of Spoink (blue, top) and the coordinates of the align-
ments are indicated. The D. melanogaster strain and the sequence similarity between the
Sanger sequence and the consensus sequence of Spoink are provided next to each matching
region.
(SVG)
S5 Fig. Abundance and composition of Spoink insertions in the GDL. A) Abundance of
Spoink in the GDL. Note that one strain from Zimbabwe does not have any Spoink insertion.
B) UMAP summarizing the composition of Spoink among the GDL. Note that Spoink shows a
pronounced population structure, where three main clusters can be discerned: Tasmania, Bej-
ing/Ithaca and Netherlands/Zimbabwe.
(SVG)
S6 Fig. A piRNA based defence against Spoink is active in the 10 GDL strains. Two strains
are analysed for each continent (Bxx Beijing/Asia, Ixx Ithaca/America, Nxx Netherlands/
Europe, Txx Tasmania/Australia, ZWxx Zimbabwe/Africa; the second strain from Ithaca (I17)
is shown in the main manuscript). A) piRNAs mapping to the sequence of Spoink. Solely the 5’
positions of piRNAs are shown and the piRNA abundance is normalized to one million piR-
NAs. Sense piRNAs are shown on the positive y-axis and antisense piRNAs on the negative y-
axis. B) ping-pong signature of Spoink.
(SVG)
S7 Fig. Barplots show the similarity between the consensus sequence of Spoink and the best
match in each of 99 long-read assemblies of diverse insect species. As a reference, two D.
melanogaster assemblies (red) were included, where D.mel.RAL176 has canonical Spoink
insertions while D.mel.Iso1 solely has degraded fragments of sequences having some similarity
with Spoink.
(AI)
S8 Fig. Star phylogeny of Spoink insertions in the different drosophilid species. Only full-
length insertions of Spoink (> 80% of the length) were considered.
(SVG)
S9 Fig. Distribution of synonymous divergence for Spoink and 140 single copy orthologous
genes shared between D. melanogaster and D. willistoni (red). For Spoink we used the shared
part of the longest ORF (green). The red dashed line is the 2.5% quantile of nuclear genes [85].
Note that the dS of Spoink is lower than the dS of any of the orthologous genes shared between
D. melanogaster and D. willistoni, consistent with a horizontal transfer of Spoink between the
two species. The genes were obtained with the software BUSCO [119]. The predicted proteins
were aligned using Clustal Omega [120]. The codons information from the protein alignment
was used for the nucleotide alignment using PAL2NAL [121]. The dS was calculated using the
software PAML.
(PNG)
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
17 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
S10 Fig. Average distance between 100 pairs of Spoink insertions randomly sampled within
either the melanogaster group (i.e D. melanogaster, D. simulans, D. sechellia) or the willis-
toni group. Distances within the willistoni group are significantly longer than the distances in
the melanogaster group (t = −6.31, df = 193.88, p = 1.762e − 09). Note that this test accounts
for the phylogenetic information of the tree using the distances of the insertions within the
two groups.
(PNG)
S1 Table. Differences in the abundance of Gypsy_7_DEl between the reference genome
Iso1 and a long-read assemblies from a more recently collected strain. The best ten matches
for Gypsy_7_DEl and the consensus sequence of Spoink are shown for both assemblies.
Matches were identified with RepeatMasker [37]. Note that the discrepancy between Iso1 and
TOM007 is more pronounced when the consensus sequence of Spoink is considered.
(XLSX)
S2 Table. Similarity between Spoink and other TEs in the different repeat libraries gener-
ated for D. melanogaster. For each repeat library the best five hits are shown. Solely matches
with a minimum overlap with Spoink of at least 30% are considered. subst. substitions in per-
cent between Spoink and the TE, len. fraction of the length of a TE aligning with Spoink; a [40],
b [43], c [5], d [69], e [70].
(XLSX)
S3 Table. Coordinates of Spoink insertions in the strains RAL091, RAL176 and RAL732
used for Fig 1A of the main manuscript.
(XLSX)
S4 Table. Identity of sequences in Oregon-R having some similarity with the consensus
sequence of Spoink. Solely sequences having a divergence of �25% and minimum overlap of
at least 10% with Spoink are considered. The sequences were extracted from the assembly of
Oregon-R (chromosome:start-end) and aligned against the TE library of D. melanogaster
using blastn [43, 122]. Most of these sequences match TARTC and DMDM11.
(XLSX)
S5 Table. Overview of the short-read data analysed in this work. Data are from [31, 40, 51–
53]).
(XLSX)
S6 Table. Overview of the long-read assemblies of D. melanogaster strains analysed in this
work. For each strain we show the assembly ID, the strain, the sampling location and the sam-
pling date. a [38, 39], b [48], c [56], d [40], e [123].
(XLSX)
S7 Table. Spoink insertions in piRNA clusters of long-read assemblies of different D. mela-
nogaster strains [40]. Note that for several strains we could not find a single Spoink insertion
in a piRNA cluster. On the other hand, some strains, like RAL176, have multiple Spoink inser-
tions in piRNA clusters.
(XLSX)
S8 Table. Overview of the long-read assemblies of diverse insect species analysed in this
work.
(XLSX)
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
18 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
Acknowledgments
We thank Matthew Beaumont for the idea to call the here described transposon Spoink. We
thank Silke Jensen for comments. We thank Neda Barghi and Claudia Ramirez Lanzas for pro-
viding fly strains used for PCR. SS would like to thank J. B. Signor for helpful comments on
the manuscript. RK, RP, and AS thank all members of the Institute of Population Genetics for
feedback and support.
Author Contributions
Conceptualization: Sarah Signor, Robert Kofler.
Data curation: Riccardo Pianezza, Almorò Scarpa, Robert Kofler.
Formal analysis: Riccardo Pianezza, Almorò Scarpa, Sarah Signor, Robert Kofler.
Funding acquisition: Sarah Signor, Robert Kofler.
Investigation: Almorò Scarpa, Prakash Narayanan, Sarah Signor, Robert Kofler.
Methodology: Riccardo Pianezza.
Project administration: Sarah Signor, Robert Kofler.
Resources: Sarah Signor, Robert Kofler.
Software: Riccardo Pianezza, Robert Kofler.
Supervision: Sarah Signor, Robert Kofler.
Visualization: Riccardo Pianezza, Almorò Scarpa, Sarah Signor, Robert Kofler.
Writing – original draft: Riccardo Pianezza, Almorò Scarpa, Sarah Signor, Robert Kofler.
Writing – review & editing: Riccardo Pianezza, Almorò Scarpa, Sarah Signor, Robert Kofler.
References
1. Schnable PS, Pasternak S, Liang C, Zhang J, Fulton L, Graves TA, et al. The B73 Maize Genome:
Complexity, Diversity, and Dynamics. Science. 2009; 326(5956):1112–1115. https://doi.org/10.1126/
science.1178534 PMID: 19965430
2. Kapitonov VV, Jurka J. Molecular paleontology of transposable elements in the Drosophila melanoga-
ster genome. Proceedings of the National Academy of Sciences of the United States of America.
2003; 100(11):6569–74. https://doi.org/10.1073/pnas.0732024100 PMID: 12743378
3.
Finnegan DJ. Eukaryotic transposable elements and genome evolution. Trends in Genetics. 1989; 5
(4):103–107. https://doi.org/10.1016/0168-9525(89)90039-5 PMID: 2543105
4. Wicker T, Sabot F, Hua-Van A, Bennetzen JL, Capy P, Chalhoub B, et al. A unified classification sys-
tem for eukaryotic transposable elements. Nature Reviews Genetics. 2007; 8(12):973–982. https://
doi.org/10.1038/nrg2165 PMID: 17984973
5. Chakraborty M, Chang C, Khost D, Vedanayagam J A J, Liao Y, Montooth K, et al. Evolution of
genome structure in the Drosophila simulans species complex. Genome Research. 2021; 31:380–
396. https://doi.org/10.1101/gr.263442.120 PMID: 33563718
6. Eickbush DG, Eickbush TH. Vertical transmission of the retrotransposable elements R1 and R2 during
the evolution of the Drosophila melanogaster species subgroup. Genetics. 1995; 139(2):671–684.
https://doi.org/10.1093/genetics/139.2.671 PMID: 7713424
7. Nelson J, Slicko A, Yamashita Y. The retrotransposon R2 maintains Drosophila ribosomal DNA
repeats. PNAS. 2023; 120(23):e2221613120. https://doi.org/10.1073/pnas.2221613120 PMID:
37252996
8. Elena SF, Ekunwe L, Hajela N, Oden SA, Lenski RE. Distribution of fitness effects caused by random
insertion mutations in Escherichia coli. Genetica. 1998; 102-103:349–358. https://doi.org/10.1023/
A:1017031008316 PMID: 9720287
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
19 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
9. Pasyukova E, Nuzhdin S, Morozova T, Mackay T. Accumulation of transposable elements in the
genome of Drosophila melanogaster is associated with a decrease in fitness. Journal of Heredity.
2004; 95(4):284–290. https://doi.org/10.1093/jhered/esh050 PMID: 15247307
10. Sarkies P, Selkirk ME, Jones JT, Blok V, Boothby T, Goldstein B, et al. Ancient and Novel Small RNA
Pathways Compensate for the Loss of piRNAs in Multiple Independent Nematode Lineages. PLoS
Biol. 2015; 13(2):1–20. https://doi.org/10.1371/journal.pbio.1002061 PMID: 25668728
11. Brennecke J, Aravin AA, Stark A, Dus M, Kellis M, Sachidanandam R, et al. Discrete small RNA-gen-
erating loci as master regulators of transposon activity in Drosophila. Cell. 2007; 128(6):1089–1103.
https://doi.org/10.1016/j.cell.2007.01.043 PMID: 17346786
12. Gunawardane LS, Saito K, Nishida KM, Miyoshi K, Kawamura Y, Nagami T, et al. A slicer-mediated
mechanism for repeat-associated siRNA 5’ end formation in Drosophila. Science. 2007; 315
(5818):1587–1590. https://doi.org/10.1126/science.1140494 PMID: 17322028
13. Brennecke J, Malone CD, Aravin AA, Sachidanandam R, Stark A, Hannon GJ. An epigenetic role for
maternally inherited piRNAs in transposon silencing. Science. 2008; 322(5906):1387–1392. https://
doi.org/10.1126/science.1165171 PMID: 19039138
14.
15.
16.
Le Thomas A, Rogers AK, Webster A, Marinov GK, Liao SE, Perkins EM, et al. Piwi induces
piRNA-guided transcriptional silencing and establishment of a repressive chromatin state. Genes
and Development. 2013; 27(4):390–399. https://doi.org/10.1101/gad.209841.112 PMID:
23392610
Le Thomas A, Marinov GK, Aravin AA. A transgenerational process defines piRNA biogenesis in Dro-
sophila virilis. Cell Reports. 2014; 8(6):1617–1623. https://doi.org/10.1016/j.celrep.2014.08.013
PMID: 25199836
Le Thomas A, Stuwe E, Li S, Du J, Marinov G, Rozhkov N, et al. Transgenerationally inherited piRNAs
trigger piRNA biogenesis by changing the chromatin of piRNA clusters and inducing precursor pro-
cessing. Genes and Development. 2014; 28(15):1667–1680. https://doi.org/10.1101/gad.245514.114
PMID: 25085419
17. Yamanaka S, Siomi MC, Siomi H. piRNA clusters and open chromatin structure. Mobile DNA. 2014; 5
(1):22. https://doi.org/10.1186/1759-8753-5-22 PMID: 25126116
18. Andreev VI, Yu C, Wang ea J. Panoramix SUMOylation on chromatin connects the piRNA pathway to
the cellular heterochromatin machinery. Nat Struct Mol Biol. 2022; 29:130–142. https://doi.org/10.
1038/s41594-022-00721-x PMID: 35173350
19. Rangan P, Malone CD, Navarro C, Newbold SP, Hayes PS, Sachidanandam R, et al. piRNA produc-
tion requires heterochromatin formation in Drosophila. Current Biology. 2011; 21(16):1373–1379.
https://doi.org/10.1016/j.cub.2011.06.057 PMID: 21820311
20. Peccoud J, Loiseau V, Cordaux, Gilbert C. Massive horizontal transfer of transposable elements in
insects. Proc Natl Acad Sci U S A. 2017; 114(18):4721–26. https://doi.org/10.1073/pnas.1621178114
PMID: 28416702
21. Scarpa A, Pianezza R, Wierzbicki F, Kofler R. Genomes of historical specimens reveal multiple inva-
sions of LTR retrotransposons in Drosophila melanogaster populations during the 19th century. PNAS
(in press). 2023.
22. Kofler R, Hill T, Nolte V, Betancourt A, Schlo¨ tterer C. The recent invasion of natural Drosophila simu-
lans populations by the P-element. PNAS. 2015; 112(21):6659–6663. https://doi.org/10.1073/pnas.
1500758112 PMID: 25964349
23. Signor S, Vedanayagam J, Kim BY, Wierzbicki F, Kofler R, Lai EC. Rapid evolutionary diversification
of the flamenco locus across simulans clade Drosophila species. PloS Genetics. 2023; 19(8):
e1010914. https://doi.org/10.1371/journal.pgen.1010914 PMID: 37643184
24.
25.
Zhang S, Pointer B, Kelleher ES. Rapid evolution of piRNA-mediated silencing of an invading trans-
posable element was driven by abundant de novo mutations. Genome Research. 2020; 30(4):566–
575. https://doi.org/10.1101/gr.251546.119 PMID: 32238416
Zanni V, Eymery A, Coiffet M, Zytnicki M, Luyten I, Quesneville H, et al. Distribution, evolution, and
diversity of retrotransposons at the flamenco locus reflect the regulatory properties of piRNA clusters.
Proceedings of the National Academy of Sciences. 2013; 110(49):19842–19847. https://doi.org/10.
1073/pnas.1313677110 PMID: 24248389
26. Schaack S, Gilbert C, Feschotte C. Promiscuous DNA: horizontal transfer of transposable elements
and why it matters for eukaryotic evolution. Trends in ecology & evolution. 2010; 25(9):537–46. https://
doi.org/10.1016/j.tree.2010.06.001 PMID: 20591532
27. Daniels SB, Peterson KR, Strausbaugh LD, Kidwell MG, Chovnick A. Evidence for horizontal transmis-
sion of the P transposable element between Drosophila species. Genetics. 1990; 124(2):339–55.
https://doi.org/10.1093/genetics/124.2.339 PMID: 2155157
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
20 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
28. Anxolabe´hère D, Kidwell MG, Periquet G. Molecular characteristics of diverse populations are consis-
tent with the hypothesis of a recent invasion of Drosophila melanogaster by mobile P elements. Molec-
ular Biology and Evolution. 1988; 5(3):252–269. PMID: 2838720
29. Kidwell MG. Evolution of hybrid dysgenesis determinants in Drosophila melanogaster. Proceedings of
the National Academy of Sciences. 1983; 80(6):1655–1659. https://doi.org/10.1073/pnas.80.6.1655
PMID: 6300863
30. Periquet G, Hamelin MH, Bigot Y, Lepissier A. Geographical and historical patterns of distribution of
hobo elements in Drosophila melanogaster populations. Journal of Evolutionary Biology. 1989; 2
(3):223–229. https://doi.org/10.1046/j.1420-9101.1989.2030223.x
31. Schwarz F, Wierzbicki F, Senti KA, Kofler R. Tirant Stealthily Invaded Natural Drosophila melanoga-
ster Populations during the Last Century. Molecular Biology and Evolution. 2021; 38(4):1482–1497.
https://doi.org/10.1093/molbev/msaa308 PMID: 33247725
32.
Loreto ELS, Carareto CMA, Capy P. Revisiting horizontal transfer of transposable elements in Dro-
sophila. Heredity. 2008; 100(6):545–54. https://doi.org/10.1038/sj.hdy.6801094 PMID: 18431403
33. Simmons G. Horizontal transfer of hobo transposable elements within the Drosophila melanogaster
species complex: evidence from DNA sequencing. Molecular biology and evolution. 1992; 9(6):1050–
1060. PMID: 1331701
34. Blumenstiel JP. Birth, school, work, death, and resurrection: The life stages and dynamics of transpos-
able element proliferation. Genes. 2019; 10(5):336. https://doi.org/10.3390/genes10050336 PMID:
31058854
35. Anxolabe´hère D, Nouaud D, Pe´ riquet G, Tchen P. P-element distribution in Eurasian populations of
Drosophila melanogaster: a genetic and molecular analysis. Proceedings of the National Academy of
Sciences. 1985; 82(16):5418–5422. https://doi.org/10.1073/pnas.82.16.5418 PMID: 16593591
36. Bonnivard E, Higuet D. Stability of European natural populations of Drosophila melanogaster with
regard to the P-M system: a buffer zone made up of Q populations. Journal of Evolutionary Biology.
1999; 12(4):633–647. https://doi.org/10.1046/j.1420-9101.1999.00063.x
37. Smit AFA, Hubley R, Green P. RepeatMasker Open-3.0; 1996-2010. Available from: http://www.
repeatmasker.org.
38. King EG, Merkes CM, McNeil CL, Hoofer SR, Sen S, Broman KW, et al. Genetic dissection of a model
complex trait using the Drosophila Synthetic Population Resource. Genome Research. 2012; 22
(8):1558–1566. https://doi.org/10.1101/gr.134031.111 PMID: 22496517
39. Chakraborty M, Emerson JJ, Macdonald SJ, Long AD. Structural variants exhibit widespread allelic
heterogeneity and shape variation in complex traits. Nature Communications. 2019; 10(1):4872.
https://doi.org/10.1038/s41467-019-12884-1 PMID: 31653862
40. Rech GE, Radı´o S, Guirao-Rico S, Aguilera L, Horvath V, Green L, et al. Population-scale long-read
sequencing uncovers transposable elements associated with gene expression variation and adaptive
signatures in Drosophila. Nature Communications. 2022; 13(1):1948. https://doi.org/10.1038/s41467-
022-29518-8 PMID: 35413957
41. Edgar RC. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic
acids research. 2004; 32(5):1792–1797. https://doi.org/10.1093/nar/gkh340 PMID: 15034147
42. Steinbiss S, Willhoeft U, Gremme G, Kurtz S. Fine-grained annotation and classification of de novo
predicted LTR retrotransposons. Nucleic acids research. 2009; 37(21):7002–7013. https://doi.org/10.
1093/nar/gkp759 PMID: 19786494
43. Quesneville H, Bergman CM, Andrieu O, Autard D, Nouaud D, Ashburner M, et al. Combined evidence
annotation of transposable elements in genome sequences. PLoS Computational Biology. 2005; 1
(2):166–175. https://doi.org/10.1371/journal.pcbi.0010022 PMID: 16110336
44. Wheeler DL, Barrett T, Benson DA, Bryant SH, Canese K, Chetvernin V, et al. Database resources of
the national center for biotechnology information. Nucleic acids research. 2007; 35(suppl_1):D5–D12.
https://doi.org/10.1093/nar/gkl1031 PMID: 17170002
45. Bouckaert R, Vaughan TG, Barido-Sottani J, Duchêne S, Fourment M, Gavryushkina A, et al. BEAST
2.5: An advanced software platform for Bayesian evolutionary analysis. PLOS Computational Biology.
2019; 15(4):e1006650. https://doi.org/10.1371/journal.pcbi.1006650 PMID: 30958812
46. Shumate A, Salzberg S. Liftoff: accurate mapping of gene annotations. Bioinformatics. 2021; 37
(12):1639–1643. https://doi.org/10.1093/bioinformatics/btaa1016 PMID: 33320174
47. Gramates LS, Agapite J, Attrill H, Calvi BR, Crosby MA, Dos Santos G, et al. FlyBase: a guided tour of
highlighted features. Genetics. 2022; 220(4):iyac035. https://doi.org/10.1093/genetics/iyac035 PMID:
35266522
48. Wierzbicki F, Schwarz F, Cannalonga O, Kofler R. Novel quality metrics allow identifying and generat-
ing high-quality assemblies of piRNA clusters. Molecular Ecology Resources. 2021. PMID: 34181811
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
21 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
49. Quinlan AR, Hall IM. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformat-
ics (Oxford, England). 2010; 26(6):841–842. https://doi.org/10.1093/bioinformatics/btq033 PMID:
20110278
50. Wierzbicki F, Kofler R, Signor S. Evolutionary dynamics of piRNA clusters in Drosophila. Molecular
Ecology. 2023; 32(6):1306–1322. https://doi.org/10.1111/mec.16311 PMID: 34878692
51. Grenier JK, Arguello JR, Moreira MC, Gottipati S, Mohammed J, Hackett SR, et al. Global diversity
lines–a five-continent reference panel of sequenced Drosophila melanogaster strains. G3: Genes,
Genomes, Genetics. 2015; 5(4):593–603. https://doi.org/10.1534/g3.114.015883 PMID: 25673134
52.
53.
54.
Long Q, Rabanal FA, Meng D, Huber CD, Farlow A, Platzer A, et al. Massive genomic variation and
strong selection in Arabidopsis thaliana lines from Sweden. Nature genetics. 2013; 45(8):884–890.
https://doi.org/10.1038/ng.2678 PMID: 23793030
Lange JD, Bastide H, Lack JB, Pool JE. A Population Genomic Assessment of Three Decades of Evo-
lution in a Natural Drosophila Population. Molecular Biology and Evolution. 2021; 39(2). https://doi.org/
10.1093/molbev/msab368
Li H, Durbin R. Fast and accurate short read alignment with Burrows–Wheeler transform. Bioinformat-
ics. 2009; 25(14):1754–1760. https://doi.org/10.1093/bioinformatics/btp324 PMID: 19451168
55. Weilguny L, Kofler R. DeviaTE: Assembly-free analysis and visualization of mobile genetic element
composition. Molecular Ecology Resources. 2019; 19(5):1346–1354. https://doi.org/10.1111/1755-
0998.13030 PMID: 31056858
56. Hoskins RA, Carlson JW, Wan KH, Park S, Mendez I, Galle SE, et al. The Release 6 reference
sequence of the Drosophila melanogaster genome. Genome Research. 2015; 25(3):445–458. https://
doi.org/10.1101/gr.185579.114 PMID: 25589440
57. Shen W, Le S, Li Y, Hu F. SeqKit: A Cross-Platform and Ultrafast Toolkit for FASTA/Q File Manipula-
tion. PLOS ONE. 2016; 11(10):1–10. https://doi.org/10.1371/journal.pone.0163962 PMID: 27706213
58.
Li H. Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics. 2018; 34(18):3094–
3100. https://doi.org/10.1093/bioinformatics/bty191 PMID: 29750242
59. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression
analysis of digital gene expression data. Bioinformatics. 2010; 26(1):139–140. https://doi.org/10.1093/
bioinformatics/btp616 PMID: 19910308
60.
Thorvaldsdo´ ttir H, Robinson JT, Mesirov JP. Integrative Genomics Viewer (IGV): high-performance
genomics data visualization and exploration. Briefings in bioinformatics. 2012;. https://doi.org/10.
1093/bib/bbs017 PMID: 22517427
61. Miller SA, Dykes DD, Polesky HF. A simple salting out procedure for extracting DNA from human
nucleated cells. Nucleic acids research. 1988; 16(3):1215. https://doi.org/10.1093/nar/16.3.1215
PMID: 3344216
62.
Luo S, Zhang H, Duan Y, Yao X, Clark AG, Lu J. The evolutionary arms race between transposable
elements and piRNAs in Drosophila melanogaster. BMC Evolutionary Biology. 2020; 20(1):1–18.
https://doi.org/10.1186/s12862-020-1580-3 PMID: 31992188
63. Martin M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet
journal. 2011; 17(1):10–12. https://doi.org/10.14806/ej.17.1.200
64.
Thurmond J, Goodman JL, Strelets VB, Attrill H, Gramates LS, Marygold SJ, et al. FlyBase 2.0: the
next generation. Nucleic acids research. 2019; 47(D1):D759–D765. https://doi.org/10.1093/nar/
gky1003 PMID: 30364959
65. Selvaraju D, Wierzbicki F, Kofler R. P-element invasions in Drosophila erecta shed light on the estab-
lishment of host control over a transposable element. bioRxiv. 2022;.
66. McInnes L, Healy J, Melville J. Umap: Uniform manifold approximation and projection for dimension
reduction. arXiv preprint arXiv:180203426. 2018;.
67. Kim BY, Wang JR, Miller DE, Barmina O, Delaney E, Thompson A, et al. Highly contiguous assemblies
of 101 drosophilid genomes. eLife. 2021; 10:e66405. https://doi.org/10.7554/eLife.66405 PMID:
34279216
68. Hotaling S, Sproul JS, Heckenhauer J, Powell A, Larracuente AM, Pauls SU, et al. Long Reads Are
Revolutionizing 20 Years of Insect Genome Sequencing. Genome Biology and Evolution. 2021; 13(8):
evab138. https://doi.org/10.1093/gbe/evab138 PMID: 34152413
69. Srivastav S, Feschotte C, Clark AG. Rapid evolution of piRNA clusters in the Drosophila melanogaster
ovary. bioRxiv. 2023;. https://doi.org/10.1101/2023.05.08.539910 PMID: 37214865
70. Ellison CE, Cao W. Nanopore sequencing and Hi-C scaffolding provide insight into the evolutionary
dynamics of transposable elements and piRNA production in wild strains of Drosophila melanogaster.
Nucleic Acids Research. 2020; 48(1):1–14. https://doi.org/10.1093/nar/gkz1080 PMID: 31754714
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
22 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
71. Wang J, Chitsaz F, Derbyshire MK, Gonzales NR, Gwadz M, Lu S, et al. The conserved domain data-
base in 2023. Nucleic Acids Research. 2023; 51(D1):D384–D388. https://doi.org/10.1093/nar/
gkac1096 PMID: 36477806
72. Eickbush TH, Malik HS. Origins and evolution of retrotransposons. vol. 93. ASM Press; 2002.
73.
Linheiro RS, Bergman CM. Whole genome resequencing reveals natural target site preferences of
transposable elements in Drosophila melanogaster. PloS one. 2012; 7(2):e30008. https://doi.org/10.
1371/journal.pone.0030008 PMID: 22347367
74. Neumann P, Nova´k P, Hosˇta´ kova´ N, Macas J. Systematic survey of plant LTR-retrotransposons eluci-
dates phylogenetic relationships of their polyprotein domains and provides a reference for element
classification. Mobile DNA. 2019; 10:1–17. https://doi.org/10.1186/s13100-018-0144-1 PMID:
30622655
75. Wells JN, Feschotte C. A field guide to eukaryotic transposable elements. Annual review of genetics.
2020; 54:539–561. https://doi.org/10.1146/annurev-genet-040620-022145 PMID: 32955944
76. Diaz-Papkovich A, Anderson-Trocme´ L, Gravel S. A review of UMAP in population genetics. Journal of
Human Genetics. 2021; 66(1):85–91. https://doi.org/10.1038/s10038-020-00851-4 PMID: 33057159
77. Sienski G, Do¨nertas D, Brennecke J. Transcriptional silencing of transposons by Piwi and maelstrom
and its impact on chromatin state and gene expression. Cell. 2012; 151(5):964–980. https://doi.org/10.
1016/j.cell.2012.10.040 PMID: 23159368
78.
Lindsley DH, Grell EH. Genetic variations of Drosophila melanogaster. Carnegie Institute of Washing-
ton Publication; 1968.
79. Bergman CM, Quesneville H, Anxolabe´hère D, Ashburner M. Recurrent insertion and duplication gen-
erate networks of transposable element sequences in the Drosophila melanogaster genome. Genome
Biology. 2006; 7(11):R112. https://doi.org/10.1186/gb-2006-7-11-r112 PMID: 17134480
80. Malone CD, Hannon GJ. Small RNAs as Guardians of the Genome. Cell. 2009; 136(4):656–668.
https://doi.org/10.1016/j.cell.2009.01.045 PMID: 19239887
81. Goriaux C, The´ron E, Brasset E, Vaury C. History of the discovery of a master locus producing piR-
NAs: The flamenco/COM locus in Drosophila melanogaster. Frontiers in Genetics. 2014; 5:257.
https://doi.org/10.3389/fgene.2014.00257 PMID: 25136352
82. Kofler R. Dynamics of Transposable Element Invasions with piRNA Clusters. Molecular Biology and
Evolution. 2019; 36(7):1457–1472. https://doi.org/10.1093/molbev/msz079 PMID: 30968135
83. Wierzbicki F, Kofler R. The composition of piRNA clusters in Drosophila melanogaster deviates from
expectations under the trap model. BMC Biology. 2023;. https://doi.org/10.1186/s12915-023-01727-7
PMID: 37858221
84. Bao W, Kojima KK, Kohany O. RepBase Update, a database of repetitive elements in eukaryotic
genomes. Mobile Dna. 2015; 6:1–6. https://doi.org/10.1186/s13100-015-0041-9 PMID: 26045719
85. Bartolome´ C, Bello X, Maside X. Widespread evidence for horizontal transfer of transposable elements
across Drosophila genomes. Genome biology. 2009; 10(2):R22. https://doi.org/10.1186/gb-2009-10-
2-r22 PMID: 19226459
86. Wallau GL, Ortiz MF, Loreto ELS. Horizontal transposon transfer in eukarya: detection, bias, and per-
spectives. Genome Biology and Evolution. 2012; 4(8):801–811. https://doi.org/10.1093/gbe/evs055
87. Burla H, da Cunha AB, Cordeiro AR, Dobzhansky T, Malogolowkin C, Pavan C. The Willistoni Group
of Sibling Species of Drosophila. Evolution. 1949; 3(4):300–314. https://doi.org/10.2307/2405716
PMID: 15396748
88. Spassky B, Richmond RC, Perez-Salas S, Pavlovsky O, Mourao CA, Hunter AS, et al. Geography of
the Sibling Species Related to Drosophila willistoni, and of the Semispecies of the Drosophila Paulis-
torum Complex. Evolution. 1971; 25(1):129–143. https://doi.org/10.1111/j.1558-5646.1971.tb01866.x
PMID: 28562939
89.
Zanini R, Depra´ M, Valante V. On the geographic distribution of the Drosophila willistoni group (Dip-
tera, Drosophilidae)—updated geographic distribution of the Neotropical willistoni subgroup. Drosoph-
ila Information Service. 2015; 98:39–43.
90. Sturtevant AH. The North American species of Drosophila. Nature. 1921; 107:1476–4687.
91.
Johnson CW. The distribution of some species of Drosophila. Psyche. 1913; 20:202–205. https://doi.
org/10.1155/1913/41505
92. Bock IR, Parsons PA. Species of Australia and New Zealand. In: Ashburner M, Carson LH, Thompson
JHJ, editors. The genetics and biology of Drosophila. vol. 3a. Oxford: Academic Press; 1981. p. 349–
393.
93. Heed WB, Russell JS. Phylogeny and population structure in island and continental species of the car-
dini group of Drosophila studied by inversion analysis. Stud Genet. 1971;.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
23 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
94. Brisson JA, Wilder J, Hollocher H. Phylogenetic analysis of the cardini group of Drosophila with respect
to changes in pigmentation. Evolution. 2006; 60(6):1228–1241. https://doi.org/10.1554/05-552.1
PMID: 16892973
95. Peccoud J, Cordaux R, Gilbert C. Analyzing Horizontal Transfer of Transposable Elements on a Large
Scale: Challenges and Prospects. Bioessays. 2018; 40(2). https://doi.org/10.1002/bies.201700177
PMID: 29283188
96. Wallau GL, Ortiz MF, Loreto ELS. Horizontal transposon transfer in eukarya: detection, bias, and per-
spectives. Genome Biol Evol. 2012; 4(8):689–699. https://doi.org/10.1093/gbe/evs055 PMID: 22798449
97. Kofler R, Senti KA, Nolte V, Tobler R, Schlo¨tterer C. Molecular dissection of a natural transposable ele-
ment invasion. Genome Research. 2018; 28(6):824–835. https://doi.org/10.1101/gr.228627.117
PMID: 29712752
98. Kofler R, Nolte V, Schlo¨ tterer C. The Transposition Rate Has Little Influence on the Plateauing Level of
the P-element. Molecular Biology and Evolution. 2022; 39(7):msac141. https://doi.org/10.1093/
molbev/msac141 PMID: 35731857
99. Pool JE. The Mosaic Ancestry of the Drosophila Genetic Reference Panel and the D. melanogaster
Reference Genome Reveals a Network of Epistatic Fitness Interactions. Molecular biology and evolu-
tion. 2015; p. msv194. https://doi.org/10.1093/molbev/msv194
100. Capy P, Gibert P. Drosophila melanogaster, Drosophila simulans: so similar yet so different. Genetica.
2004; 120(1-3):5–16. https://doi.org/10.1023/B:GENE.0000017626.41548.97 PMID: 15088643
101. Kapun M, Barro´ n MG, Staubach F, Obbard DJ, Wiberg RAW, Vieira J, et al. Genomic analysis of Euro-
pean Drosophila melanogaster populations reveals longitudinal structure, continent-wide selection,
and previously unknown DNA viruses. Molecular Biology and Evolution. 2020; 37(9):2661–2678.
https://doi.org/10.1093/molbev/msaa120 PMID: 32413142
102. Machado HE, Bergland AO, Taylor R, Tilk S, Behrman E, Dyer K, et al. Broad geographic sampling
reveals the shared basis and environmental correlates of seasonal adaptation in Drosophila. Elife.
2021; 10:e67577. https://doi.org/10.7554/eLife.67577 PMID: 34155971
103. Obbard DJ, Maclennan J, Kim KW, Rambaut A, O’Grady PM, Jiggins FM. Estimating Divergence
Dates and Substitution Rates in the Drosophila Phylogeny. Molecular Biology and Evolution. 2012; 29
(11):3459–3473. https://doi.org/10.1093/molbev/mss150 PMID: 22683811
104. Prakash S. Genetic divergence in closely related sibling species Drosophila pseudoobscura, Drosoph-
ila persimilis and Drosophila miranda. Evolution. 1977; p. 14–23. https://doi.org/10.2307/2407540
PMID: 28567734
105. Regner L, Pereira M, Alonso C, Abdelhay E, Valente V. Genomic distribution of P elements in Dro-
sophila willistoni and a search for their relationship with chromosomal inversions. Journal of Heredity.
1996; 87(3):191–198. https://doi.org/10.1093/oxfordjournals.jhered.a022984 PMID: 8683096
106. Bergman CM, Bensasson D. Recent LTR retrotransposon insertion contrasts with waves of non-LTR
insertion since speciation in Drosophila melanogaster. Proceedings of the National Academy of Sci-
ences of the United States of America. 2007; 104(27):11340–5. https://doi.org/10.1073/pnas.
0702552104 PMID: 17592135
107. Bucheton A, Lavige J, Picard G, L’heritier P. Non-mendelian female sterility in Drosophila melanoga-
ster: quantitative variations in the efficiency of inducer and reactive strains. Heredity. 1976; 36(3):305–
314. https://doi.org/10.1038/hdy.1976.38 PMID: 819401
108. Kidwell MG, Kidwell JF, Sved JA. Hybrid dysgenesis in Drosophila melanogaster: A syndrome of aber-
rant traits including mutations, sterility and male recombination. Genetics. 1977; 86(4):813–833.
https://doi.org/10.1093/genetics/86.4.813 PMID: 17248751
109. Blackman RK, Grimaila R, Macy M, Koehler D, Gelbart WM. Mobilization of hobo elements residing within
the decapentaplegic gene complex: Suggestion of a new hybrid dysgenesis system in Drosophila mela-
nogaster. Cell. 1987; 49(4):497–505. https://doi.org/10.1016/0092-8674(87)90452-1 PMID: 3032458
110. Schmidt JM, Good RT, Appleton B, Sherrard J, Raymant GC, Bogwitz MR, et al. Copy number varia-
tion and transposable elements feature in recent, ongoing adaptation at the Cyp6g1 locus. PLoS
genetics. 2010; 6(6):e1000998. https://doi.org/10.1371/journal.pgen.1000998 PMID: 20585622
111. Wang L, Zhang S, Hadjipanteli S, Saiz L, Nguyen L, Silva E, et al. P-element invasion fuels molecular
adaptation in laboratory populations of Drosophila melanogaster. Evolution. 2023; 77(4):980–994.
https://doi.org/10.1093/evolut/qpad017 PMID: 36749648
112. Scarpa A, Kofler R. The impact of paramutations on the invasion dynamics of transposable elements.
Genetics. 2023; p. iyad181. https://doi.org/10.1093/genetics/iyad181 PMID: 37819004
113. Hermant C, Boivin A, Teysset L, Delmarre V, Asif-Laidin A, Van Den Beek M, et al. Paramutation in
Drosophila requires both nuclear and cytoplasmic actors of the piRNA pathway and induces cis-
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
24 / 25
PLOS GENETICS Spoink, a LTR retrotransposon, invaded D. melanogaster populations in the 1990s
spreading of piRNA production. Genetics. 2015; 201(4):1381–1396. https://doi.org/10.1534/genetics.
115.180307 PMID: 26482790
114. Shpiz S, Ryazansky S, Olovnikov I, Abramov Y, Kalmykova A. Euchromatic transposon insertions trig-
ger production of novel pi-and endo-siRNAs at the target sites in the Drosophila germline. PLoS
Genetics. 2014; 10(2):e1004138. https://doi.org/10.1371/journal.pgen.1004138 PMID: 24516406
115. Mohn F, Sienski G, Handler D, Brennecke J. The rhino-deadlock-cutoff complex licenses noncanoni-
cal transcription of dual-strand piRNA clusters in Drosophila. Cell. 2014; 157(6):1364–1379. https://
doi.org/10.1016/j.cell.2014.04.031 PMID: 24906153
116. Ashburner M, Bergman CM. Drosophila melanogaster: A case study of a model genomic sequence
and its consequences. Cold Spring Harbor perspectives in biology. 2005; 15:1661–1667.
117. Green MM. Mobile DNA elements and spontaneous gene mutation. Banbury Rep. 1988; 30:41–50.
118. Sankaranarayanan K. Mobile genetic elements, spontaneous mutations, and the assessment of
genetic radiation hazards in man. Cold Spring Harbor; 1988.
119. Manni M, Berkeley MR, Seppey M, Simão FA, Zdobnov EM. BUSCO update: novel and streamlined
workflows along with broader and deeper phylogenetic coverage for scoring of eukaryotic, prokaryotic,
and viral genomes. Molecular biology and evolution. 2021; 38(10):4647–4654. https://doi.org/10.1093/
molbev/msab199 PMID: 34320186
120. Sievers F, Wilm A, Dineen D, Gibson TJ, Karplus K, Li W, et al. Fast, scalable generation of high-qual-
ity protein multiple sequence alignments using Clustal Omega. Molecular systems biology. 2011; 7
(1):539. https://doi.org/10.1038/msb.2011.75 PMID: 21988835
121. Suyama M, Torrents D, Bork P. PAL2NAL: robust conversion of protein sequence alignments into the
corresponding codon alignments. Nucleic acids research. 2006; 34(suppl_2):W609–W612. https://doi.
org/10.1093/nar/gkl315 PMID: 16845082
122. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. Journal of
Molecular Biology. 1990; 215(3):403–410. https://doi.org/10.1016/S0022-2836(05)80360-2 PMID:
2231712
123. Mackay TF, Richards S, Stone EA, Barbadilla A, Ayroles JF, Zhu D, et al. The Drosophila melanoga-
ster genetic reference panel. Nature. 2012; 482(7384):173–178. https://doi.org/10.1038/nature10811
PMID: 22318601
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1011201 March 26, 2024
25 / 25
PLOS GENETICS
|
Could not heal snippet
|
10.1371_journal.pone.0222013.pdf
|
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
|
All relevant data are within the paper and its Supporting Information files.
|
RESEARCH ARTICLE
MicroRNA regulation in colorectal cancer
tissue and serum
Lukasz Gmerek1,2☯, Kari Martyniak1☯, Karolina Horbacka2, Piotr Krokowicz2,
Wojciech Scierski3, Pawel Golusinski4,5, Wojciech Golusinski5, Augusto Schneider6*,
Michal M. MasternakID
1,5*
1 College of Medicine, Burnett School of Biomedical Sciences, University of Central Florida, Orlando, FL,
United States of America, 2 Department of General and Colorectal Surgery, Poznan University of Medical
Sciences, Poznan, Poland, 3 Department of Otorhinolaryngology and Laryngological Oncology in Zabrze,
Medical University of Silesia, Katowice, Poland, 4 Department of Otolaryngology and Maxillofacial Surgery,
University of Zielona Gora, Zielona Gora, Poland, 5 Department of Head and Neck Surgery, Poznan
University of Medical Sciences, The Greater Poland Cancer Centre, Poznan, Poland, 6 Faculdade de
Nutric¸ão, Universidade Federal de Pelotas, Pelotas, RS, Brazil
☯ These authors contributed equally to this work.
* [email protected] (AS); [email protected] (MMM)
Abstract
Colorectal cancer is recognized as the fourth leading cause of cancer-related deaths world-
wide. Thus, there is ongoing search for potential new biomarkers allowing quicker and less
invasive detection of the disease and prediction of the treatment outcome. Therefore, the
aim of our study was to identify colorectal cancer specific miRNAs expressed in cancerous
and healthy tissue from the same patient and to further correlate the presence of the same
miRNAs in the circulation as potential biomarkers for diagnosis. In the current study we
detected a set of 40 miRNAs differentially regulated in tumor tissue when comparing with
healthy tissue. Additionally, we found 8 miRNAs differentially regulated in serum of colorec-
tal cancer patients. Interestingly, there was no overlap in miRNAs regulated in tissue and
serum, suggesting that serum regulated miRNAs may be not actively secreted from colorec-
tal tumor cells. However, four of differentially expressed miRNAs, including miR-21, miR-17,
miR-20a and miR-32 represent the miRNAs characteristic for different tumor types, includ-
ing breast, colon, lung, pancreas, prostate and stomach cancer. This finding suggests
important groups of miRNAs which can be further validated as markers for diagnosis of
tumor tissue and regulation of carcinogenesis.
Introduction
Cancer development encompasses alterations in cell growth, differentiation and regulation of
apoptosis. Over a decades of cancer research many oncogenes and tumor suppressor genes
have been identified and extensively studied for its role in the pathogenesis and malignancy of
different types of cancer [1, 2]. In this scenario, the discovery of short small non-coding RNAs
(sncRNAs) unveiled new potential molecular regulators of tumorigenesis [3]. MicroRNAs
(miRNAs) are a class of sncRNAs that interact with the RNA Induced Silencing Complex
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Gmerek L, Martyniak K, Horbacka K,
Krokowicz P, Scierski W, Golusinski P, et al. (2019)
MicroRNA regulation in colorectal cancer tissue
and serum. PLoS ONE 14(8): e0222013. https://
doi.org/10.1371/journal.pone.0222013
Editor: Klaus Roemer, Universitat des Saarlandes,
GERMANY
Received: July 26, 2019
Accepted: August 20, 2019
Published: August 30, 2019
Copyright: © 2019 Gmerek et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: The study was funded by the Florida
Legislative Crohn’s grant (MMM). The funders had
no role in study design, data collection and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
1 / 12
miRNAs in colorectal cancer
(RISC) to bind to the 3’ untranslated region (UTR) of mRNA molecules and regulate tran-
scription and mRNA stability [4, 5]. miRNAs have been shown to have an active role in cell
growth and proliferation, being also implicated in tumorigenesis by regulating oncogenes and
tumor suppressor genes expression [6, 7]. Tumor produced miRNAs are also regarded as pre-
dictors of malignancy and response to chemotherapy [3, 6].
miRNAs are produced in the nucleus and regulate gene expression in the cytoplasm of the
cell [4]. However, miRNAs can be also found in the extracellular environment, including in
serum, suggesting that it does not have an exclusively intracellular role [8–10]. The origin of
extracellular miRNAs may include passive leakage from apoptotic or damaged cells and/or
through secretory activity mainly within extracellular vesicles which includes exosomes [11].
Circulating miRNAs can have a role in intercellular communication, affecting gene expression
in distant or adjacent target cells [11], or serve as biomarkers for pathological conditions [8].
Therefore, it is hypothesized that the signature of circulating miRNAs provide high sensitivity,
success and reproducibility in the diagnostics of different types of cancer using a non-invasive
approaches [8, 12, 13]. Despite previous work on miRNA signatures in serum or tissue of vari-
ous types of cancer, including colorectal, very few studies approach tissue and serum variations
of miRNAs simultaneously in the same patients. This paired method can suggest if the changes
in circulating miRNA signatures are derived from the main tumoral tissue or are due second-
ary causes.
Due to high rate of colorectal cancer-related deaths worldwide [14, 15], the miRNA profile
in biopsies and serum has been extensively studied for this condition [16–20] but the lack of
more comprehensive studies in both tissue and serum from the same patients and the repeat-
ability for the identified miRNAs in different conditions is needed. Therefore, the goal of our
study was to investigate the populations of miRNAs expressed in colorectal cancerous tissue
when compared with a healthy adjacent tissue and serum from the same patients, to determine
potential new biomarkers for early detection, prediction of patient recovery and future more
personalized therapeutic approaches.
Results
After sequencing and processing, 12,540,784 adapter cleaned reads/sample with a 64.6% align-
ment rate to the human genome (hg19) for tissues was obtained in average. In the serum sam-
ples, 1,341,762 adapter cleaned reads/sample resulted in a 43.7% alignment rate to the human
genome (hg19) in average. Principal component analysis (PCA) from the 500 miRNAs with
the most variation in tissue and serum samples indicates a different and very clear pattern of
expression between healthy and cancer tissue and serum samples (Fig 1).
Following the initial analysis, the samples with < 3 reads per million (rpm) in more than
half of tested samples were removed, which resulted in identification of final 388 different
miRNAs expressed in tissue (S1 Table) and 110 miRNAs in the serum samples (S2 Table).
Comparison of the expression patterns of miRNAs in tumor and healthy tissue identified 40
differentially expressed miRNAs. Out of these 40 miRNAs, 20 were downregulated, while 20
indicated increased expression (False Discovery rate—FDR<0.05 and Fold Change–FC<0.5
or >2.0; Table 1). For serum samples 8 miRNAs were differentially expressed (4 down- and 4
up-regulated; FDR<0.05 and FC<0.5 or >2.0; Table 2). There was no overlap in the differen-
tially expressed miRNAs between tissue and serum. Only one miRNA regulated in serum was
not found as overall expressed in tissue samples (hsa-miR-486-3p), the other seven serum reg-
ulated miRNAs were also found in tissue samples, although not differentially regulated.
Pathway and GO term enrichment analysis was performed using the miRNAs differentially
regulated in serum (40 miRNAs–see Table 1) and tissue (8 miRNAs–see Table 2) allowed us to
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
2 / 12
miRNAs in colorectal cancer
Fig 1. Principal component analysis of the 500 most variable miRNAs in the tissue and serum samples (healthy tissue—H and tumor tissue—T) from patients
diagnosed with colorectal cancer.
https://doi.org/10.1371/journal.pone.0222013.g001
identify several known cellular processes regulated by these differentially expressed tissue and
serum specific miRNAs. Importantly, the analysis indicated that cancer related pathways are
among the top miRNA-regulated pathways in analyzed tissue (Table 3) and serum (Table 4).
Additionally, several pathways involving well known oncogenes were significantly targeted by
the regulated miRNAs in biopsies samples, as TGF and Foxo signaling pathways (Table 3 and
Figs 2 and 3, respectively). GO Terms for biological process and molecular function are pre-
sented in S3 and S4 Tables.
Discussion
In the current study we detected a set of 40 miRNAs differentially regulated in tissue and 8
miRNAs differentially regulated in serum of colorectal cancer patients. There was no overlap
in miRNAs regulated in tissue and serum, suggesting that serum regulated miRNAs may be
not actively secreted from colorectal tumor cells. However, the differential regulated miRNAs
in serum may be leaking passively from damaged cells into circulation [11]. Additionally, this
suggests that other cancer driven conditions, i.e. systemic inflammation, oxidative stress, may
be driven changes in serum miRNAs to be used as biomarkers.
Some miRNAs are consistently differentially regulated in a myriad of solid cancers (i.e.,
breast, colon, lung, pancreas, prostate and stomach cancer), with 21 miRNAs identified as reg-
ulated in at least three different types of cancer [6]. Interestingly, four of these miRNAs over-
lapped with miRNAs we currently identified as regulated in colorectal cancer tissue samples,
including miR-21, miR-17, miR-20a and miR-32. All these four miRNAs were also identified
as differentially expressed in colorectal cancer tissue [6], and miR-21 and miR-17 were identi-
fied as regulated in at least five different types of cancer, including breast, lung, prostate, pan-
creas and stomach [6], suggesting a consistent marker for diagnosis of tumor tissue and
involved in carcinogenesis. Additionally, a recent review paper identified several tissue
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
3 / 12
Table 1. MicroRNAs differentially expressed between tumor and healthy adjacent tissue in six patients diagnosed with colorectal cancer.
miRNAs in colorectal cancer
miRNA1
Down-regulated
Healthy
Tumor
hsa-miR-133b
hsa-miR-1-3p
hsa-miR-133a-3p
hsa-miR-363-3p
hsa-miR-143-3p
hsa-miR-145-5p
hsa-miR-129-5p
hsa-miR-135a-5p
hsa-miR-504-5p
hsa-miR-145-3p
hsa-miR-139-3p
hsa-miR-139-5p
hsa-miR-143-5p
hsa-miR-30c-2-3p
hsa-miR-30a-3p
hsa-miR-195-3p
hsa-miR-9-5p
hsa-miR-378i
hsa-miR-138-5p
hsa-miR-378d
Up-regulated
hsa-miR-135b-5p
hsa-miR-592
hsa-miR-503-5p
hsa-miR-424-5p
hsa-miR-514a-3p
hsa-miR-584-5p
hsa-miR-20a-5p
hsa-miR-708-5p
hsa-miR-1277-3p
hsa-miR-18a-5p
hsa-miR-625-3p
hsa-miR-224-5p
hsa-miR-21-5p
hsa-miR-450b-5p
hsa-miR-17-5p
hsa-miR-32-5p
hsa-miR-32-3p
hsa-miR-148a-3p
hsa-miR-19a-3p
hsa-miR-941
214 ± 309433
49721 ± 309318
4053 ± 309299
1440 ± 19226
2341934 ± 309469
44128 ± 309313
190 ± 19225
60 ± 36501
234 ± 309469
9463 ± 309435
60 ± 19225
949 ± 309434
10052 ± 19253
160 ± 19246
1127 ± 19232
241 ± 36519
951 ± 19245
36 ± 36720
40 ± 36749
900 ± 36720
149 ± 19248
39 ± 36519
13 ± 36500
57 ± 36501
12 ± 36766
37 ± 36520
1869 ± 19237
68 ± 36750
2 ± 36769
25 ± 19248
115 ± 36769
602 ± 36517
198267 ± 19264
61 ± 7568
988 ± 19248
473 ± 36518
13 ± 36764
374325 ± 36725
362 ± 36747
357 ± 36749
13 ± 6
4077 ± 702
405 ± 151
170 ± 52
294834 ± 51415
5646 ± 1232
25 ± 12
10 ± 4
42 ± 12
1771 ± 328
12 ± 4
200 ± 50
2317 ± 547
43 ± 7
309 ± 76
72 ± 10
293 ± 46
11 ± 2
15 ± 4
368 ± 69
1101 ± 348
263 ± 107
64 ± 25
284 ± 147
58 ± 24
163 ± 54
7150 ± 1639
257 ± 136
10 ± 2
87 ± 20
403 ± 201
2117 ± 537
660219 ± 187444
200 ± 67
2971 ± 572
1322 ± 180
35 ± 4
942026 ± 152194
879 ± 148
859 ± 191
FC2
0.06
0.08
0.10
0.12
0.13
0.13
0.13
0.16
0.18
0.19
0.20
0.21
0.23
0.27
0.27
0.30
0.31
0.32
0.37
0.41
7.38
6.68
5.03
4.94
4.67
4.37
3.83
3.76
3.65
3.54
3.52
3.51
3.33
3.30
3.01
2.80
2.62
2.52
2.43
2.41
P Value
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0012
<0.0001
<0.0001
<0.0001
<0.0001
0.0003
0.0001
<0.0001
0.0009
0.0001
0.0038
0.0030
0.0040
0.0002
0.0004
0.0009
0.0011
0.0021
0.0004
0.0001
0.0029
0.0017
0.0001
0.0013
0.0005
0.0004
0.0044
0.0003
0.0004
0.0022
0.0042
0.0026
0.0028
FDR3
0.0007
0.0003
0.0003
0.0013
0.0008
0.0003
0.0013
0.0168
0.0007
0.0007
0.0012
0.0007
0.0065
0.0017
0.0015
0.0135
0.0020
0.0396
0.0325
0.0406
0.0048
0.0071
0.0135
0.0159
0.0261
0.0065
0.0017
0.0318
0.0226
0.0029
0.0179
0.0081
0.0065
0.0428
0.0058
0.0065
0.0263
0.0415
0.0306
0.0314
1miRNAs are expressed as reads per million (rpm). miRNA with less than 3 rpm in more than 50% of the samples were removed from analysis.
2Fold change in Tumor compared to Healthy tissue
3False discovery rate. Only miRNAs with FDR lower than 0.05 were considered as significantly regulated.
https://doi.org/10.1371/journal.pone.0222013.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
4 / 12
miRNAs in colorectal cancer
Table 2. MicroRNAs differentially expressed in serum of tumor and healthy patients diagnosed with colorectal cancer.
miRNA1
Down-regulated
hsa-miR-375
hsa-miR-486-3p
hsa-miR-486-5p
hsa-miR-1180-3p
Up-regulated
hsa-let-7d-5p
hsa-let-7a-5p
hsa-miR-30e-3p
hsa-let-7f-5p
Healthy
Tumor
120 ± 22
97 ± 11
13664 ± 1286
20 ± 4
87 ± 15
956 ± 146
24 ± 2
642 ± 128
15 ± 4
27 ± 9
3995 ± 1097
7 ± 1
266 ± 79
2569 ± 606
63 ± 13
1620 ± 415
FC2
0.13
0.27
0.29
0.34
3.03
2.69
2.66
2.53
PValue
<0.0001
0.0002
<0.0001
0.0035
0.0010
0.0006
0.0019
0.0034
FDR3
<0.0001
0.0056
0.0010
0.0477
0.0225
0.0161
0.0342
0.0477
1miRNAs are expressed as reads per million (rpm). miRNA with less than 3 rpm in more than 50% of the samples were removed from analysis.
2Fold change in Tumor compared to Healthy tissue
3False discovery rate. Only miRNAs with FDR lower than 0.05 were considered as significantly regulated.
https://doi.org/10.1371/journal.pone.0222013.t002
expressed miRNAs associated to poor prognosis in colorectal cancer patients [17]. Our study
overlapped with 7 of these identified miRNAs, including miR-21, miR-195, miR-17, miR-20a,
miR-145, miR-224 and miR-139. It is interesting that overlapping our study with the previous
mentioned studies [6, 17], we can observe that miR-21, miR20a and miR-17 are both predic-
tors of cancer occurrence and poor prognosis in colorectal cancer patients, further indicating
their central role in cancer pathogenesis.
Previous studies indicated significant role of miR-21 regulation in colorectal cancer [6, 17].
In present study miR-21-5p was among the highest expressed miRNAs, and more importantly
Table 3. Pathways of target genes from the 40 miRNAs differentially expressed between tumor and healthy tissue
of colorectal cancer patients.
KEGG pathway
Prion diseases
Morphine addiction
Mucin type O-Glycan biosynthesis
ECM-receptor interaction
Fatty acid biosynthesis
Signaling pathways regulating pluripotency of stem cells
TGF-beta signaling pathway
GABAergic synapse
Axon guidance
Thyroid hormone signaling pathway
Proteoglycans in cancer
Glioma
FoxO signaling pathway
Prolactin signaling pathway
Estrogen signaling pathway
Renal cell carcinoma
P value1
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0001
0.0001
0.0006
0.0007
0.0050
0.0131
0.01325
0.02071
0.0211
Genes2
1
44
13
26
5
70
38
28
68
34
64
28
56
42
24
28
miRNAs3
2
10
7
8
1
8
8
8
6
7
5
7
5
7
4
5
1Only pathways with P values lower than 0.05 were considered as significant
2Number of genes affected in the pathway by the regulated miRNAs
3Number of miRNAs differentially expressed that have a target gene in the pathway
https://doi.org/10.1371/journal.pone.0222013.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
5 / 12
miRNAs in colorectal cancer
Table 4. Pathways of target genes from the 8 miRNAs differentially expressed between tumor and healthy serum of colorectal cancer patients.
KEGG pathway
Prion diseases
ECM-receptor interaction
Mucin type O-Glycan biosynthesis
Signaling pathways regulating pluripotency of stem cells
Thyroid hormone signaling pathway
Biotin metabolism
Amoebiasis
Glycosaminoglycan biosynthesis
P value1
<0.0001
<0.0001
<0.0001
0.0004
0.0005
0.0013
0.0070
0.0279
Genes2
1
10
4
17
16
1
11
3
miRNAs3
1
3
3
3
4
1
2
2
1Only pathways with P values lower than 0.05 were considered as significant
2Number of genes affected in the pathway by the regulated miRNAs
3Number of miRNAs differentially expressed that have a target gene in the pathway
https://doi.org/10.1371/journal.pone.0222013.t004
it was significantly upregulated in tumor tissue when comparing with healthy tissue. miR-21
was identified as overexpressed in six different types of cancer [6], and we have previously
detected miR-21 as highly abundant and overexpressed in a similar fold change in oral squa-
mous cell carcinoma samples [21]. miR-21 is associated with prognosis of colorectal cancer
patients [17], as overexpression of miR-21 shows negative correlation with patients responses
to chemotherapy as well as progression-free survival [22]. The central role of miR-21 may be
Fig 2. Schematic representation of the FOXO signaling pathway and the target genes of the microRNAs differentially regulated between tumor tissue and healthy
tissue from patients diagnosed with colorectal cancer. Yellow box–target gene of one down-regulated miRNA; Orange box–target gene of two or more down-
regulated miRNA.
https://doi.org/10.1371/journal.pone.0222013.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
6 / 12
miRNAs in colorectal cancer
Fig 3. Schematic representation of the TGF-β signaling pathway and the target genes of the microRNAs differentially regulated between tumor tissue and healthy
tissue from patients diagnosed with colorectal cancer. Yellow box–target gene of one down-regulated miRNA; Orange box–target gene of two or more down-
regulated miRNA.
https://doi.org/10.1371/journal.pone.0222013.g003
explained by its target genes which include cell growth and proliferation regulating PTEN, a
negative regulator of the Pi3k/Akt pathway [23]. Therefore, our study further confirms the
central role of miR-21 in cancer development in colorectal patients.
Previous studies have identified 32 miRNAs in serum as regulated in colorectal cancer
patients [17]. Comparing to our current study only one miRNA overlapped, miR-375. Others
have identified miR-375 as down-regulated in serum of cancer patients, and predictor of can-
cer recurrence [24], further suggesting its role in diagnosis. In our study miR-375 was ten-fold
down-regulated in the serum of cancer patients. A previous paper from our group with oral
squamous cell carcinoma patients also identified miR-375 as strongly down-regulated in tissue
samples [21]. The hsa-miR-375 is known to target MMP13, which is associated to increased
metastatic behavior and cancer aggressiveness [25]. Therefore, it is important to focus more in
depth on the role of serum miR-375 in the diagnosis of different types of cancer as well as in
the metastatic process, given its target genes and its systemic presence.
We identified miR-143 as strongly down-regulated in serum samples, as others have
observed in osteosarcoma, breast cancer and esophageal squamous cell carcinoma [26–28].
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
7 / 12
miRNAs in colorectal cancer
miR-143 targets the FOSL2 gene, promoting cell proliferation and metastasis and inhibiting
apoptosis [26]. miR-143 constitute a functional cluster along miR-145 [27], which we also
identified as down-regulated in our current study, further consolidating both as serum mark-
ers for diagnosis. miR-486-5p also was highly expressed and strongly down-regulated in serum
of colorectal cancer patients in the current study. miR-485-5p was identified as biomarker of
colorectal cancer and malignancy when locally expressed in tumorous tissue [29, 30]. How-
ever, serum miR-486-5p was not identified as regulated in a recent review paper on many stud-
ies with colorectal cancer patients [17]. One recent study identified both miR-486-3p and -5p
as down-regulated in late stage colorectal cancer patients serum but not in early stages patients
[31]. This suggests that miR-486 it is not a good marker, as it is not an indicator of early stage
cancer, which would constitute a better diagnostic tool for intervention.
miR-148 was strongly up-regulated in serum of colorectal cancer patients in our study. This
is controversial, as others have found that miR-148 overexpression inhibited colon cancer cell
proliferation and migration [32]. miR-148 expression in tissue samples was down-regulated in
a cohort of colorectal cancer patients [33]. More studies are necessary to better understand the
role of miR-148, and the effects of cancer type and stage in its regulation to better understand
its role in cancer pathogenesis. We also observed that members of the let-7 family were up-reg-
ulated in serum of colorectal cancer patients. This is controversial as a previous study has
found let-7 to be down-regulated and negatively correlated with metastasis in serum of breast
cancer patients [34]. Interestingly, it is suggested that a metastatic gastric cancer line actively
secrets members of the let-7 family in the extracellular environment via exosomes to maintain
their oncogenesis [35]. Therefore, although let-7 is a tumor suppressor miRNA, its presence in
serum may be an indication of increased tumorigenesis and metastatic activity in cancerous
tissue, providing a new approach to understand regulation of these biomarkers.
In sum, we detected a set of 40 miRNAs differentially regulated in tissue and 8 miRNAs dif-
ferentially regulated in serum of colorectal cancer patients. There was no overlap in miRNAs
regulated in tissue and serum. Therefore, our study further validates previous miRNAs
observed as important in colorectal cancer and other types of cancer and suggests that serum
regulated miRNAs may not be the same locally regulated in tissue samples.
Materials and methods
Sample and tissue collection
Tissue and serum samples were obtained during surgical procedure from six patients diag-
nosed with colorectal cancer (4 men and 2 women) with average age of 67.3 years (from 44 to
76 years old). All samples included in the study consisted of tumors in stage G2 (adenocarci-
noma tubulare invasivum coli, G2). Recurrences and patients initially treated with radiother-
apy were excluded from the study. The details including TNM, Dukes and Astler-Coller
classification are presented in Table 5. Additionally, blood samples from six healthy patients
were collected for RNA extraction.
Table 5. Characteristics of the samples used in the study.
Sample
1
2
3
4
5
6
TNM
pT3, pN2b
pT3, pN1b
pT1, pN0
pT3, pN1b
pT3, pN0
pT4a, pN1a
https://doi.org/10.1371/journal.pone.0222013.t005
Dukes
Astler-Coller
C
C
A
C
B
C
C2
C2
B1
C2
B2
C2
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
8 / 12
miRNAs in colorectal cancer
Blood samples (n = 12, six colorectal cancer patients and six healthy subjects, never diag-
nosed with any type of tumor, with the average age of 66.6 years) were collected approximate-
ly24hours prior to any surgical intervention, in BD Vacutainer Serum Separation Tubes,
incubated 15 minutes in room temperature, centrifuged for serum separation and then stored
in -80o C. Additionally, from every colorectal patient two separate tissue specimens were
obtained during surgical resection. Core biopsy from the tumor and healthy adjacent tissue
within the range of 15–20 cm distal from tumor tissue were collected to allow comparison of
tumor site versus non-tumor healthy tissue, in the same cancer patient. Specimens were imme-
diately frozen in liquid nitrogen and then stored in -80 o C.
This study was carried out in accordance with the recommendations and approval by Insti-
tutional Review Board of the University of Medical Sciences in Poznan. All subjects gave writ-
ten informed consent in accordance with the Declaration of Helsinki.
RNA extraction and miRNA library preparation
Previously frozen tissues samples (n = 12) were homogenized with Qiazol (Qiagen, Valencia,
CA, USA) using zirconium oxide beads (0.5 mm) in the Bullet Blender 24 (Next Advance,
Averill Park, NY, USA). Total RNA was extracted from tissue samples using a commercial col-
umn purification system (miRNeasy Mini Kit, Qiagen) and on-column DNase treatment
(RNase-free DNase Set, Qiagen) following manufacturer’s instructions. RNA extraction from
serum samples (n = 12) was performed with the miRNEasy Serum/Plasma kit (Qiagen) also
following manufacturers instructions.
TruSeq Small RNA Sample Prep Kit (Illumina Inc., San Diego, CA, USA) following the
manufacturer’s instructions as adjusted by Matkovich, Hu [36] was used to prepare the miR-
NAs libraries. Briefly, small RNAs from serum and tissue samples total RNA were ligated with
30 and 50 adapters, followed by reverse transcription to produce single stranded cDNAs. Adap-
tor-ligated miRNAs were then amplified by 14 cycles PCR using indexes to allow individual
libraries to be processed together in a single flowcell lane during the sequencing step (12 tissue
and 12 serum samples). Samples were mixed and a 6% acrylamide gel was used to size-select
and purify the amplified libraries.
BioAnalyzer and RNA Nano Lab Chip Kit (Agilent Technologies, Santa Clara, CA, USA)
was used to determine the quality and quantity of the libraries. Following the quality check all
samples were pooled into one tube and sent for sequencing on a HiSeq 2500 instrument (Illu-
mina Inc.).
miRNAs libraries analysis and statistical analyses
Alignment and quantification of miRNA libraries was performed using sRNAtoolbox as
described before [37]. Statistical analyses of differentially expressed miRNAs was performed
using EdgeR [38] on the R software (3.2.2) and miRNAs with a FDR<0.05 and FC>2.0
were considered as up-regulated; and FDR<0.05 and FC<0.50 were considered as down-
regulated.
miRNAs target prediction and enriched pathways and GO Terms
Target genes of the differentially regulated miRNAs were predicted using the mirPath tool
(version 3.0) and the microT-CDS v. 5.0 database [39]. Gene ontology (GO) terms (biological
processes) and KEGG molecular pathways [40, 41] were also retrieved using the same tool.
Pathways and processes regulated with P values lower than 0.05 were considered as significant.
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
9 / 12
miRNAs in colorectal cancer
Supporting information
S1 Table. MicroRNAs expressed in serum from tumor and healthy patients diagnosed with
colorectal cancer.
(DOCX)
S2 Table. MicroRNAs expressed in tumor and healthy adjacent tissue in patients diag-
nosed with colorectal cancer.
(DOCX)
S3 Table. Gene ontology terms for biological processes, molecular function and cellular
compartment of target genes from the 40 miRNAs differentially expressed between tumor
and healthy tissue of colorectal cancer patients.
(DOCX)
S4 Table. Gene ontology terms for biological processes, molecular function and cellular
compartment of target genes from the 8 miRNAs differentially expressed between tumor
and healthy serum of colorectal cancer patients.
(DOCX)
Acknowledgments
The authors are thankful to the Kegg Database Project team from Kanehisa Laboratories for
providing permission to use the pathway images.
Author Contributions
Conceptualization: Lukasz Gmerek, Karolina Horbacka, Piotr Krokowicz, Pawel Golusinski,
Wojciech Golusinski, Augusto Schneider, Michal M. Masternak.
Data curation: Kari Martyniak, Pawel Golusinski, Augusto Schneider, Michal M. Masternak.
Formal analysis: Lukasz Gmerek, Kari Martyniak, Wojciech Scierski, Pawel Golusinski, Woj-
ciech Golusinski, Augusto Schneider, Michal M. Masternak.
Investigation: Kari Martyniak.
Resources: Lukasz Gmerek, Karolina Horbacka, Piotr Krokowicz, Wojciech Scierski.
Supervision: Augusto Schneider, Michal M. Masternak.
Validation: Wojciech Scierski, Augusto Schneider.
Writing – original draft: Pawel Golusinski, Wojciech Golusinski, Augusto Schneider, Michal
M. Masternak.
Writing – review & editing: Karolina Horbacka, Piotr Krokowicz, Wojciech Scierski.
References
1. Yang J, Chen L, Kong X, Huang T, Cai YD. Analysis of tumor suppressor genes based on gene ontol-
ogy and the KEGG pathway. PloS one. 2014; 9(9):e107202. Epub 2014/09/11. doi: 10.1371/journal.
pone.0107202. PMID: 25207935; PubMed Central PMCID: PMC4160198.
2. Bell DW. Our changing view of the genomic landscape of cancer. J Pathol. 2010; 220(2):231–43. Epub
2009/11/18. doi: 10.1002/path.2645. PMID: 19918804; PubMed Central PMCID: PMC3195356.
3. Calin GA, Croce CM. MicroRNA signatures in human cancers. Nature reviews Cancer. 2006; 6
(11):857–66. doi: 10.1038/nrc1997. PMID: 17060945.
4. Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell. 2004; 116(2):281–97.
https://doi.org/10.1016/s0092-8674(04)00045-5 PMID: 14744438.
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
10 / 12
miRNAs in colorectal cancer
5. Cuellar TL, McManus MT. MicroRNAs and endocrine biology. The Journal of endocrinology. 2005; 187
(3):327–32. doi: 10.1677/joe.1.06426. PMID: 16423811.
6. Volinia S, Calin GA, Liu CG, Ambs S, Cimmino A, Petrocca F, et al. A microRNA expression signature
of human solid tumors defines cancer gene targets. Proceedings of the National Academy of Sciences
of the United States of America. 2006; 103(7):2257–61. doi: 10.1073/pnas.0510565103. PMID:
16461460; PubMed Central PMCID: PMC1413718.
7. Calin GA, Croce CM. MicroRNA-cancer connection: the beginning of a new tale. Cancer Res. 2006; 66
(15):7390–4. doi: 10.1158/0008-5472.CAN-06-0800. PMID: 16885332.
8. Cortez MA, Bueso-Ramos C, Ferdin J, Lopez-Berestein G, Sood AK, Calin GA. MicroRNAs in body flu-
ids—the mix of hormones and biomarkers. Nature reviews Clinical oncology. 2011; 8(8):467–77. doi:
10.1038/nrclinonc.2011.76. PMID: 21647195; PubMed Central PMCID: PMC3423224.
9. Nunez Lopez YO, Victoria B, Golusinski P, Golusinski W, Masternak MM. Characteristic miRNA expres-
sion signature and random forest survival analysis identify potential cancer-driving miRNAs in a broad
range of head and neck squamous cell carcinoma subtypes. Rep Pract Oncol Radiother. 2018; 23
(1):6–20. doi: 10.1016/j.rpor.2017.10.003. PMID: 29187807; PubMed Central PMCID: PMC5698002.
10. Kolenda T, Guglas K, Rys M, Bogaczynska M, Teresiak A, Blizniak R, et al. Biological role of long non-
coding RNA in head and neck cancers. Rep Pract Oncol Radiother. 2017; 22(5):378–88. Epub 2017/
08/11. doi: 10.1016/j.rpor.2017.07.001. PMID: 28794691; PubMed Central PMCID: PMC5542372.
11. Chen X, Liang H, Zhang J, Zen K, Zhang CY. Secreted microRNAs: a new form of intercellular commu-
nication. Trends in cell biology. 2012; 22(3):125–32. doi: 10.1016/j.tcb.2011.12.001. PMID: 22260888.
12. Vychytilova-Faltejskova P, Radova L, Sachlova M, Kosarova Z, Slaba K, Fabian P, et al. Serum-based
microRNA signatures in early diagnosis and prognosis prediction of colon cancer. Carcinogenesis.
2016; 37(10):941–50. doi: 10.1093/carcin/bgw078. PMID: 27485599.
13.
Zheng H, Zhang L, Zhao Y, Yang D, Song F, Wen Y, et al. Plasma miRNAs as diagnostic and prognos-
tic biomarkers for ovarian cancer. PloS one. 2013; 8(11):e77853. doi: 10.1371/journal.pone.0077853.
PMID: 24223734; PubMed Central PMCID: PMC3815222.
14. McGuire S. World Cancer Report 2014. Geneva, Switzerland: World Health Organization, International
Agency for Research on Cancer, WHO Press, 2015. Adv Nutr. 2016; 7(2):418–9. Epub 2016/03/17. doi:
10.3945/an.116.012211. PMID: 26980827; PubMed Central PMCID: PMC4785485.
15. Witold K, Anna K, Maciej T, Jakub J. Adenomas—Genetic factors in colorectal cancer prevention. Rep
Pract Oncol Radiother. 2018; 23(2):75–83. Epub 2018/02/22. doi: 10.1016/j.rpor.2017.12.003. PMID:
29463957; PubMed Central PMCID: PMC5814382.
16. Danese E, Minicozzi AM, Benati M, Paviati E, Lima-Oliveira G, Gusella M, et al. Reference miRNAs for
colorectal cancer: analysis and verification of current data. Scientific reports. 2017; 7(1):8413. Epub
2017/08/23. doi: 10.1038/s41598-017-08784-3. PMID: 28827728; PubMed Central PMCID:
PMC5567181.
17. Masuda T, Hayashi N, Kuroda Y, Ito S, Eguchi H, Mimori K. MicroRNAs as Biomarkers in Colorectal
Cancer. Cancers (Basel). 2017; 9(9). Epub 2017/09/14. doi: 10.3390/cancers9090124. PMID:
28902152; PubMed Central PMCID: PMC5615339.
18.
Falzone L, Scola L, Zanghi A, Biondi A, Di Cataldo A, Libra M, et al. Integrated analysis of colorectal
cancer microRNA datasets: identification of microRNAs associated with tumor development. Aging.
2018; 10(5):1000–14. Epub 2018/05/21. doi: 10.18632/aging.101444. PMID: 29779016; PubMed Cen-
tral PMCID: PMC5990389.
19. Cirillo F, Catellani C, Sartori C, Lazzeroni P, Amarri S, Street ME. Obesity, Insulin Resistance, and
Colorectal Cancer: Could miRNA Dysregulation Play A Role? International journal of molecular sci-
ences. 2019; 20(12). Epub 2019/06/19. doi: 10.3390/ijms20122922. PMID: 31207998; PubMed Central
PMCID: PMC6628223.
20. Xie B, Zhao R, Bai B, Wu Y, Xu Y, Lu S, et al. Identification of key tumorigenesisrelated genes and their
microRNAs in colon cancer. Oncol Rep. 2018; 40(6):3551–60. Epub 2018/10/03. doi: 10.3892/or.2018.
6726. PMID: 30272358.
21. Schneider A, Victoria B, Lopez YN, Suchorska W, Barczak W, Sobecka A, et al. Tissue and serum
microRNA profile of oral squamous cell carcinoma patients. Scientific reports. 2018; 8(1):675. Epub
2018/01/14. doi: 10.1038/s41598-017-18945-z. PMID: 29330429.
22. Chan JK, Blansit K, Kiet T, Sherman A, Wong G, Earle C, et al. The inhibition of miR-21 promotes apo-
ptosis and chemosensitivity in ovarian cancer. Gynecologic oncology. 2014; 132(3):739–44. doi: 10.
1016/j.ygyno.2014.01.034. PMID: 24472409.
23. Yan LX, Liu YH, Xiang JW, Wu QN, Xu LB, Luo XL, et al. PIK3R1 targeting by miR-21 suppresses
tumor cell migration and invasion by reducing PI3K/AKT signaling and reversing EMT, and predicts clin-
ical outcome of breast cancer. International journal of oncology. 2016; 48(2):471–84. doi: 10.3892/ijo.
2015.3287. PMID: 26676464.
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
11 / 12
miRNAs in colorectal cancer
24. Yan Y, Wang X, Veno MT, Bakholdt V, Sorensen JA, Krogdahl A, et al. Circulating miRNAs as biomark-
ers for oral squamous cell carcinoma recurrence in operated patients. Oncotarget. 2017; 8(5):8206–14.
doi: 10.18632/oncotarget.14143. PMID: 28030794.
25. Osako Y, Seki N, Kita Y, Yonemori K, Koshizuka K, Kurozumi A, et al. Regulation of MMP13 by antitu-
mor microRNA-375 markedly inhibits cancer cell migration and invasion in esophageal squamous cell
carcinoma. Int J Oncol. 2016; 49(6):2255–64. doi: 10.3892/ijo.2016.3745. PMID: 27779648; PubMed
Central PMCID: PMC5117997.
26. Sun X, Dai G, Yu L, Hu Q, Chen J, Guo W. miR-143-3p inhibits the proliferation, migration and invasion
in osteosarcoma by targeting FOSL2. Scientific reports. 2018; 8(1):606. Epub 2018/01/14. doi: 10.
1038/s41598-017-18739-3. PMID: 29330462; PubMed Central PMCID: PMC5766605.
27.
Johannessen C, Moi L, Kiselev Y, Pedersen MI, Dalen SM, Braaten T, et al. Expression and function of
the miR-143/145 cluster in vitro and in vivo in human breast cancer. PloS one. 2017; 12(10):e0186658.
Epub 2017/10/27. doi: 10.1371/journal.pone.0186658. PMID: 29073169; PubMed Central PMCID:
PMC5657998.
28. He Z, Yi J, Liu X, Chen J, Han S, Jin L, et al. MiR-143-3p functions as a tumor suppressor by regulating
cell proliferation, invasion and epithelial-mesenchymal transition by targeting QKI-5 in esophageal
squamous cell carcinoma. Molecular cancer. 2016; 15(1):51. Epub 2016/07/01. doi: 10.1186/s12943-
016-0533-3. PMID: 27358073; PubMed Central PMCID: PMC4928305.
29.
Liu C, Li M, Hu Y, Shi N, Yu H, Liu H, et al. miR-486-5p attenuates tumor growth and lymphangiogenesis
by targeting neuropilin-2 in colorectal carcinoma. Onco Targets Ther. 2016; 9:2865–71. Epub 2016/06/
11. doi: 10.2147/OTT.S103460. PMID: 27284245; PubMed Central PMCID: PMC4881731.
30. Kelley KA, Wieghard N, Chin Y, Potter A, Mori M, Wong MH, et al. MiR-486-5p Downregulation Marks
an Early Event in Colorectal Carcinogenesis. Dis Colon Rectum. 2018; 61(11):1290–6. Epub 2018/09/
22. doi: 10.1097/DCR.0000000000001192. PMID: 30239392; PubMed Central PMCID: PMC6172144.
31. Aherne ST, Madden SF, Hughes DJ, Pardini B, Naccarati A, Levy M, et al. Circulating miRNAs miR-34a
and miR-150 associated with colorectal cancer progression. BMC Cancer. 2015; 15:329. Epub 2015/
05/01. doi: 10.1186/s12885-015-1327-5. PMID: 25924769; PubMed Central PMCID: PMC4417244.
32.
33.
Tsai HL, Yang IP, Huang CW, Ma CJ, Kuo CH, Lu CY, et al. Clinical significance of microRNA-148a in
patients with early relapse of stage II stage and III colorectal cancer after curative resection. Transl Res.
2013; 162(4):258–68. Epub 2013/08/13. doi: 10.1016/j.trsl.2013.07.009. PMID: 23933284.
Takahashi M, Cuatrecasas M, Balaguer F, Hur K, Toiyama Y, Castells A, et al. The clinical significance
of MiR-148a as a predictive biomarker in patients with advanced colorectal cancer. PloS one. 2012; 7
(10):e46684. Epub 2012/10/12. doi: 10.1371/journal.pone.0046684. PMID: 23056401; PubMed Central
PMCID: PMC3463512.
34. Elghoroury EA, ElDine HG, Kamel SA, Abdelrahman AH, Mohammed A, Kamel MM, et al. Evaluation of
miRNA-21 and miRNA Let-7 as Prognostic Markers in Patients With Breast Cancer. Clin Breast Cancer.
2018; 18(4):e721–e6. Epub 2018/01/03. doi: 10.1016/j.clbc.2017.11.022. PMID: 29292183.
35. Ohshima K, Inoue K, Fujiwara A, Hatakeyama K, Kanto K, Watanabe Y, et al. Let-7 microRNA family is
selectively secreted into the extracellular environment via exosomes in a metastatic gastric cancer cell
line. PloS one. 2010; 5(10):e13247. Epub 2010/10/16. doi: 10.1371/journal.pone.0013247. PMID:
20949044; PubMed Central PMCID: PMC2951912.
36. Matkovich SJ, Hu Y, Dorn GW 2nd., Regulation of cardiac microRNAs by cardiac microRNAs. Circula-
tion research. 2013; 113(1):62–71. doi: 10.1161/CIRCRESAHA.113.300975. PMID: 23625950;
PubMed Central PMCID: PMC3769967.
37. Rueda A, Barturen G, Lebron R, Gomez-Martin C, Alganza A, Oliver JL, et al. sRNAtoolbox: an inte-
grated collection of small RNA research tools. Nucleic acids research. 2015; 43(W1):W467–73. doi: 10.
1093/nar/gkv555. PMID: 26019179; PubMed Central PMCID: PMC4489306.
38. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression
analysis of digital gene expression data. Bioinformatics. 2010; 26(1):139–40. doi: 10.1093/
bioinformatics/btp616. PMID: 19910308; PubMed Central PMCID: PMC2796818.
39. Vlachos IS, Zagganas K, Paraskevopoulou MD, Georgakilas G, Karagkouni D, Vergoulis T, et al.
DIANA-miRPath v3.0: deciphering microRNA function with experimental support. Nucleic acids
research. 2015; 43(W1):W460–6. doi: 10.1093/nar/gkv403. PMID: 25977294; PubMed Central PMCID:
PMC4489228.
40. Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic acids research. 2000;
28(1):27–30. https://doi.org/10.1093/nar/28.1.27 PMID: 10592173; PubMed Central PMCID:
PMC102409.
41. Kanehisa M, Sato Y, Kawashima M, Furumichi M, Tanabe M. KEGG as a reference resource for gene
and protein annotation. Nucleic acids research. 2016; 44(D1):D457–62. doi: 10.1093/nar/gkv1070.
PMID: 26476454; PubMed Central PMCID: PMC4702792.
PLOS ONE | https://doi.org/10.1371/journal.pone.0222013 August 30, 2019
12 / 12
| null |
10.1007_s00208-022-02489-3.pdf
|
Data Availability Data sharing not applicable to this article as no data sets were generated or analysed
during the current study.
|
Data Availability Data sharing not applicable to this article as no data sets were generated or analysed during the current study.
|
Mathematische Annalen
https://doi.org/10.1007/s00208-022-02489-3
Mathematische Annalen
Time periodic motion of temperature driven compressible
fluids
Eduard Feireisl1
· Piotr Gwiazda2
· Agnieszka ´Swierczewska-Gwiazda3
Received: 11 April 2022 / Revised: 7 September 2022 / Accepted: 3 October 2022
© The Author(s) 2022
Abstract
We consider the Navier–Stokes–Fourier system describing the motion of a compress-
ible viscous fluid in a container with impermeable boundary subject to time periodic
heating and under the action of a time periodic potential force. We show the existence
of a time periodic weak solution for arbitrarily large physically admissible data.
Contents
.
.
.
.
.
.
.
.
.
.
1 Introduction .
.
2 Main result .
.
.
2.1 Constitutive theory .
.
2.2 Weak solutions
.
.
2.3 Main result
.
.
.
3 Approximate problem .
.
. .
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
. .
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
The work of E.F. was partially supported by the Czech Sciences Foundation (GA ˇCR), Grant Agreement
21-02411S. The Institute of Mathematics of the Academy of Sciences of the Czech Republic is supported
by RVO:67985840. This work is partially supported by the Simons Foundation Award No 663281 granted
to the Institute of Mathematics of the Polish Academy of Sciences for the years 2021-2023. The work of
A. ´S-G. and P.G. was partially supported by National Science Centre (Poland), agreement no
2021/43/B/ST1/02851.
B Agnieszka ´Swierczewska-Gwiazda
[email protected]
Eduard Feireisl
[email protected]
Piotr Gwiazda
[email protected]
1
2
3
Institute of Mathematics of the Academy of Sciences of the Czech Republic,
Žitná 25, 115 67 Praha 1, Czech Republic
Institute of Mathematics of Polish Academy of Sciences, ´Sniadeckich 8, 00-956 Warsaw, Poland
Institute of Applied Mathematics and Mechanics, University of Warsaw,
Banacha 2, 02-097 Warsaw, Poland
123
4 Uniform bounds .
.
3.1 Existence of approximate solutions .
.
.
3.2 Approximate ballistic energy balance .
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
4.1 Mass conservation .
.
4.2 Energy estimates
.
.
.
4.3 Pressure estimates .
4.4 Uniform bounds for ε → 0 .
.
.
.
.
5 Convergence .
.
6 Concluding remarks .
.
.
References .
.
.
. .
. .
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
E. Feireisl et al.
.
.
.
.
. .
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 Introduction
There are numerous examples of turbulent fluid motion excited by changes of the
boundary temperature, among which is the well studied problem of Rayleigh–Bénard
convection, see e.g. Davidson [8]. Motivated by similar problems in astrophysics of
gaseous stars, we consider a general compressible viscous possibly rotating fluid, occu-
pying a bounded domain (cid:3) ⊂ Rd , d = 2, 3, driven by periodic changes of boundary
temperature. The relevant system of field equations for the standard variables: the
mass density (cid:4) = (cid:4)(t, x), the velocity u = u(t, x), and the (absolute) temperature
ϑ = ϑ(t, x) reads:
∂t (cid:4) + divx ((cid:4)u) = 0,
(1.1)
∂t ((cid:4)u) + divx ((cid:4)u ⊗ u) + (cid:4)(ω × u) + ∇x p((cid:4), ϑ) = divx S + (cid:4)∇x G,
(1.2)
∂t ((cid:4)e((cid:4), ϑ)) + divx ((cid:4)e((cid:4), ϑ)u) + ∇x q = S : Dx u − p((cid:4), ϑ)divx u,
(1.3)
where S is the viscous stress given by Newton’s rheological law
(cid:2)
(cid:3)
S(ϑ, Dx u) = μ(ϑ)
∇x u + ∇t
divx uI
+ η(ϑ)divx uI,
(1.4)
x u − 2
3
and q is the heat flux given by Fourier’s law
q(ϑ, ∇x ϑ) = −κ(ϑ)∇x ϑ.
(1.5)
The momentum equation is augmented by the Coriolis force with the rotation constant
vector ω, the associated centrifugal force as well as the gravitation and other possible
inertial time-periodic forces are regrouped in the potential G. The fluid occupies a
bounded smooth domain (cid:3) ⊂ Rd , d = 2, 3 endowed with the Dirichlet boundary
conditions
u|∂(cid:3) = 0,
ϑ|∂(cid:3) = ϑB.
(1.6)
123
Time periodic motion of temperature...
The functions ϑB = ϑB(t, x) and G = G(t, x) are smooth and T -periodic in the
time variable,
ϑB(t + T , x) = ϑB(t, x),
G(t + T , x) = G(t, x).
(1.7)
Hereafter, the problem (1.1)–(1.6) is referred to as Navier–Stokes–Fourier system.
Our goal is to show the existence of a time–periodic solution to problem (1.1)–(1.7).
There is a substantial number of references, where such a result is proven under some
smallness and smoothness assumption on the data. Valli and Zajaczkowski [24, 25]
observe that the distance of two smooth global in time solutions decays in time for the
system close to a stable equilibrium, and, as a by product, they deduce the existence of a
time periodic solution. Similar ideas have been followed by many authors, see Bˇrezina
and Kagei [4], [5], Jin and Yang [17] , Kagei and Oomachi [18], Kagei and Tsuda
[19], Tsuda [23] to name only a few. Turbulent fluid flows given by large forces out of
equilibrium are mostly considered in the framework of weak solutions. Based on the
mathematical theory of compressible fluids developed by Lions [20, 21], the existence
of large time periodic solutions for the simplified isentropic system was proved in [9]
for the isentropic pressure–density equation of state p((cid:4)) = a(cid:4)γ , γ ≥ 9
5 . The later
development of the theory in [12] enabled to extend the result to the case γ > 5
3 , see
Cai and Tan [6].
The situation is more delicate for the complete fluid systems including thermal
effects. As a direct consequence of the Second law of thermodynamics, the existence
of (forced) time periodic solutions is ruled out for problems with purely conservative
boundary conditions, see [13]. In [10], the heat flux was controlled by means of a
Robin type boundary condition
q · n = d(ϑ − (cid:10)0) on ∂(cid:3),
(1.8)
with a given “mean” temperature (cid:10)0. Accordingly, the internal energy is transferred
out of the fluid domain in the high temperature regime and the time periodic motion
is possible, see [10, Theorem 1]. Our goal is to show a similar result for the Dirichlet
boundary conditions (1.6). Note that the problem is much more delicate than in [10]
as the heat flux through the boundary is a priori not controlled. Additional novelty
is that the function ϑB in (3.1) is time dependent whereas its counterpart (cid:10)0 in [10],
cf. (1.8), depends only on x. Finally we note that the presence of the Coriolis force
in the momentum equation, though physically relevant in some situations, does not
represent any extra analytical difficulties.
Our approach is based on several rather new ideas that appeared only recently in
the mathematical theory of open fluid systems.
• The concept of weak solution for the Navier–Stokes–Fourier system based on a
combination of the entropy inequality and the ballistic energy balance developed
in [7].
• Uniform bounds and large time asymptotics of the weak solutions in the spirit of
[14].
123
E. Feireisl et al.
• An approximation scheme based on a penalization of the Dirichlet boundary con-
ditions via (1.8).
The concept of weak solution developed in the monograph [11] and used in [10] is
based on the total energy balance as an integral part of the definition of weak solution
to the Navier–Stokes–Fourier system. This approach applies solely to problems with
conservative boundary conditions, where the energy flux vanishes on the boundary of
the physical space or it is at least controlled as in (1.8). The problems with inhomoge-
neous Dirichlet boundary conditions require an alternative approach developed in [7],
where the energy is replaced by the ballistic energy, for which the boundary flux is
again controllable. This approach has been used recently in [14], where the existence
of bounded absorbing sets and asymptotic compactness of bounded trajectories was
established.
The constitutive restrictions imposed on the equations of state as well as the trans-
port coefficients are the same as in the existence theory [7]. In particular, the general
equation of state of real monoatomic gases proposed in [11, Chapters 1,2] is included.
From this point of view, the result is apparently better than in the isentropic case stud-
ied in [9], and later revisited by Cai and Tan [6], where the condition γ > 5
3 is needed.
The price to pay is the potential form of the driving force f = ∇x G that, however,
includes the physically relevant centrifugal as well as gravitational forces.
The paper is organized as follows. In Sect. 2, we introduce the basic hypotheses
concerning the constitutive relations and state the main result. In Sect. 3, we introduce
an approximation scheme inspired by [10]. Section 4 is the heart of the paper. Here
we establish the necessary uniform bounds to perform the limit in the sequence of
approximate solutions. Finally, in Sect. 5, we obtain the desired solution as a limit of
the approximate sequence.
2 Main result
Before stating the main result, we recall the form of the constitutive equations proposed
in [11, Chapters 1,2]. To comply with the Second law of thermodynamics, we postulate
the existence of entropy s, related to the internal energy e and the pressure p through
Gibbs’ equation
ϑ Ds((cid:4), ϑ) = De((cid:4), ϑ) + p((cid:4), ϑ)D
(cid:2)
(cid:3)
.
1
(cid:4)
(2.1)
2.1 Constitutive theory
Similarly to [11, Chapters 1,2] we consider the pressure equation of state in the form
p((cid:4), ϑ) = pm((cid:4), ϑ) + prad(ϑ),
123
Time periodic motion of temperature...
where pm is the pressure of a general monoatomic gas related to the internal energy
through
pm((cid:4), ϑ) = 2
3
(cid:4)em((cid:4), ϑ),
(2.2)
augmented by the radiation pressure
prad(ϑ) = a
3
ϑ 4, a > 0.
Similarly, the internal energy reads
e((cid:4), ϑ) = em((cid:4), ϑ) + erad((cid:4), ϑ), erad((cid:4), ϑ) = a
(cid:4)
ϑ 4.
Now, Gibbs’ equ. (2.1) gives rise to a specific form of pm,
pm((cid:4), ϑ) = ϑ 5
2 P
(cid:3)
(cid:2)
(cid:4)
ϑ 3
2
for a certain P ∈ C 1[0, ∞). Consequently,
p((cid:4), ϑ) = ϑ 5
2 P
(cid:3)
(cid:2)
(cid:4)
ϑ 3
2
+ a
3
ϑ 4, e((cid:4), ϑ) = 3
2
(cid:3)
ϑ 5
2
(cid:4)
P
(cid:2)
(cid:4)
ϑ 3
2
+ a
(cid:4)
ϑ 4, a > 0.
In addition, we suppose
P(0) = 0, P (cid:9)(Z ) > 0 for Z ≥ 0, 0 <
5
3 P(Z ) − P (cid:9)(Z )Z
Z
≤ c for Z > 0,
(2.3)
(2.4)
that may be seen as a direct consequence of hypothesis of thermodynamic stability,
5
see [11, Chapter 1], and Bechtel et al. [1]. It follows that the function Z (cid:11)→ P(Z )/Z
3
is decreasing, and we suppose
lim
Z→∞
P(Z )
5
3
Z
= p∞ > 0.
The associated entropy takes the form
s((cid:4), ϑ) = S
(cid:3)
(cid:2)
(cid:4)
ϑ 3
2
+ 4a
3
ϑ 3
(cid:4)
,
(2.5)
(2.6)
123
E. Feireisl et al.
where
S(cid:9)(Z ) = − 3
2
5
3 P(Z ) − P (cid:9)(Z )Z
Z 2
< 0.
(2.7)
Finally, we impose the Third law of thermodynamics, see e.g. Belgiorno [2, 3], requir-
ing the total entropy to vanish as soon as the absolute temperature approaches zero,
S(Z ) = 0.
lim
Z→∞
(2.8)
It is easy to check that (2.4)–(2.8) imply
0 ≤ (cid:4)S
(cid:3)
(cid:2)
(cid:4)
ϑ 3
2
(cid:4)
1 + (cid:4) log
≤ c
+((cid:4)) + (cid:4) log
+(ϑ)
(cid:5)
.
(2.9)
As for the transport coefficients, we suppose that they are continuously differen-
tiable functions of the absolute temperature satisfying
0 < μ(1 + ϑ) ≤ μ(ϑ), |μ(cid:9)(ϑ)| ≤ μ,
0 ≤ η(ϑ) ≤ η(1 + ϑ),
0 < κ(1 + ϑ β ) ≤ κ(ϑ) ≤ κ(1 + ϑ β ),
(2.10)
where, in accordance with the existence theory developed in [7], we require
β > 6.
(2.11)
2.2 Weak solutions
It is convenient to identify the time periodic functions (distributions) with objects
defined on a periodic “flat sphere”
ST = [0, T ]|{0,T }.
We are ready to introduce the concept of time periodic solution to the Navier–Stokes–
Fourier system (1.1)–(1.7).
123
Time periodic motion of temperature...
Definition 2.1 (weak solution) We say that a trio ((cid:4), ϑ, u) is a weak time–periodic solution to
the problem (1.1)–(1.7) if the following holds:
• Regularity class:
(cid:4) ∈ Cweak(ST ; L γ ((cid:3))) for γ = 5
3
((cid:3); Rd )), (cid:4)u ∈ Cweak(ST , L
u ∈ L 2(ST ; W 1,2
,
0
2γ
γ +1 ((cid:3); Rd )),
(2.12)
ϑ β/2, log(ϑ) ∈ L 2(ST ; W 1,2((cid:3))),
(ϑ − ϑB ) ∈ L 2(ST ; W 1,2
((cid:3))).
0
• Equation of continuity:
(cid:6)
(cid:6)
(cid:7)
(cid:6)
(cid:6)
(cid:9)
ST
(cid:3)
b((cid:4))∂t ϕ + b((cid:4))u · ∇x ϕ +
(cid:8)
(cid:4)∂t ϕ + (cid:4)u · ∇x ϕ
(cid:12)
(cid:11)
(cid:3)
ST
(cid:10)
b((cid:4)) − b(cid:9)((cid:4))(cid:4)
dx dt = 0,
(2.13)
divx uϕ
dx dt = 0
(2.14)
for any ϕ ∈ C 1(ST × (cid:3)), and any b ∈ C 1(R), b(cid:9) ∈ Cc(R).
• Momentum equation:
(cid:6)
(cid:6)
(cid:9)
(cid:4)u · ∂t ϕ + (cid:4)u ⊗ u : ∇x ϕ − (cid:4)(ω × u) · ϕ + pdivx ϕ
(cid:6)
(cid:12)
(cid:12)
dx dt
(cid:9)
S : ∇x ϕ − (cid:4)∇x G · ϕ
dx dt
(2.15)
ST
(cid:3)
(cid:6)
=
ST
(cid:3)
for any ϕ ∈ C 1
c
(ST × (cid:3); Rd ).
• Entropy inequality:
(cid:6)
(cid:6)
(cid:9)
−
(cid:4)s∂t ϕ + (cid:4)su · ∇x ϕ + q
ϑ
S : Dx u − q · ∇x ϑ
(cid:13)
ϑ
ST
(cid:6)
(cid:3)
(cid:6)
≥
ST
(cid:3)
ϕ
ϑ
for any ϕ ∈ C 1
c
(ST × (cid:3)), ϕ ≥ 0;
• Ballistic energy balance:
(cid:13)
(cid:6)
−
(cid:6)
∂t ψ
ST
(cid:6)
(cid:3)
(cid:6)
≤
ψ
ST
(cid:3)
(cid:14)
(cid:6)
(cid:6)
(cid:4)|u|2 + (cid:4)e − ˜ϑ(cid:4)s
1
2
(cid:9)
(cid:4)u · ∇x G − (cid:4)su · ∇x
dx dt +
ST
˜ϑ
· ∇x
˜ϑ − q
ϑ
ψ
(cid:12)
dx dt
for any ψ ∈ C 1(ST ), ψ ≥ 0, and any ˜ϑ ∈ C 1(ST × (cid:3)),
˜ϑ > 0, ˜ϑ|∂(cid:3) = ϑB .
(cid:12)
· ∇x ϕ
(cid:14)
dx dt
dx dt
(cid:13)
˜ϑ
ϑ
(cid:3)
S : Dx u − q · ∇x ϑ
ϑ
(cid:14)
(2.16)
dx dt
(2.17)
(2.18)
123
The weak time–periodic solutions are therefore the weak solutions in the sense of [7]
that are T -periodic in the time variable. The instantaneous values of the conservative
variables (cid:4)(τ, ·), ((cid:4)u)(τ, ·) are well defined as well as the right and left-hand limits
of the total entropy S = (cid:4)s((cid:4), ϑ),
E. Feireisl et al.
(cid:12)S(τ −, ·); φ(cid:13) ≡ lim
δ→0+
1
δ
(cid:12)S(τ +, ·); φ(cid:13) ≡ lim
δ→0+
1
δ
2.3 Main result
(cid:6) τ
(cid:6)
(cid:4)s(t, ·)φ dx dt,
(cid:6)
(cid:3)
τ −δ
(cid:6) τ +δ
(cid:4)s(t, ·)φ dx dt.
τ
(cid:3)
Having collected the necessary preliminary material we are ready to state our main
result.
Theorem 2.2 (existence of time periodic solutions) Let (cid:3) ⊂ Rd , d = 2, 3
be a bounded domain of class C 2+ν. Suppose that the pressure p, the internal
energy e, the entropy s, as well as the transport coefficients μ, η, and κ satisfy
the hypotheses (2.2)–(2.11). Finally, let the data G ∈ W 1,∞(ST × (cid:3)), ϑB ∈
C 3(ST × Rd ) be time periodic as stated in (1.7), and
ϑB = ϑ > 0.
inf
ST ×(cid:3)
Then for any M0 there exists at least one time periodic solution ((cid:4), ϑ, u) of
the problem (1.1)–(1.7) in the sense specified in Definition 2.1 satisfying
(cid:6)
(cid:3)
(cid:4)(t, ·) dx = M0 for any t ∈ ST .
Remark 2.3 In the hypotheses of Theorem 2.2, we assume that ϑB|∂(cid:3) is a restriction
of a (smooth) function defined on the whole space Rd .
The rest of the paper is devoted to the proof of Theorem 2.2.
3 Approximate problem
The most efficient way of constructing suitable approximate solutions seems adapting
the result of [10] to the present setting. Specifically, the approximation scheme is based
on penalization of the Dirichlet boundary condition for the temperature via the Robin
boundary conditions
q · n = 1
ε
|ϑ − ϑB|k(ϑ − ϑB), k ≥ 0, on ∂(cid:3),
(3.1)
where ε > 0 is a small parameter.
123
Time periodic motion of temperature...
The approximate solutions ((cid:4)ε, ϑε, uε) are defined similarly to Definition 2.1:
• Regularity class:
(cid:4)ε ∈ Cweak(ST ; L γ ((cid:3))) for γ = 5
3
uε ∈ L 2(ST ; W 1,2
((cid:3); Rd )), (cid:4)εuε ∈ Cweak(ST , L
, log(ϑε) ∈ L 2(ST ; W 1,2((cid:3))).
,
0
ϑ β/2
ε
2γ
γ +1 ((cid:3); Rd ))
(3.2)
• Equation of continuity:
(cid:6)
(cid:6)
(cid:9)
ST
(cid:3)
b((cid:4)ε)∂t ϕ + b((cid:4)ε)uε · ∇x ϕ +
(cid:6)
(cid:6)
ST
(cid:3)
(cid:7)
(cid:4)ε∂t ϕ + (cid:4)εuε · ∇x ϕ
(cid:8)
dx dt = 0,
(cid:10)
b((cid:4)ε) − b(cid:9)((cid:4)ε)(cid:4)ε
(cid:11)
(3.3)
(cid:12)
divx uεϕ
dx dt = 0
for any ϕ ∈ C 1(ST × (cid:3)), and any b ∈ C 1(R), b(cid:9) ∈ Cc(R).
• Momentum equation:
(cid:9)
(cid:4)εuε · ∂t ϕ + (cid:4)εuε ⊗ uε : ∇x ϕ − (cid:4)ε(ω × uε) · ϕ + pdivx ϕ
(cid:6)
(cid:6)
ST
(cid:6)
(cid:3)
(cid:6)
=
ST
(cid:3)
(cid:9)
S : ∇x ϕ − (cid:4)ε∇x G · ϕ
(cid:12)
dx dt
(3.4)
(cid:12)
dx dt
(3.5)
for any ϕ ∈ C 1
c
(ST × (cid:3); Rd ).
• Entropy inequality:
(cid:6)
(cid:6)
(cid:13)
−
(cid:3)
(cid:6)
ST
+ 1
ε
(cid:4)εs∂t ϕ + (cid:4)εsuε · ∇x ϕ+ q
ϑε
(cid:6)
ϕ
ST
∂(cid:3)
|ϑB − ϑε|k (ϑB − ϑε)
ϑε
dσx dt
(cid:14)
(cid:6)
(cid:6)
· ∇x ϕ
dx dt≥
ST
(cid:3)
ϕ
ϑε
(cid:13)
S : Dx uε− q · ∇x ϑε
ϑε
(cid:14)
dx dt
(3.6)
for any ϕ ∈ C 1(ST × (cid:3)), ϕ ≥ 0.
• Energy balance:
(cid:6)
−
(cid:6)
(cid:13)
∂t ψ
ST
(cid:6)
(cid:3)
(cid:6)
1
2
(cid:4)ε|uε|2 + (cid:4)εe
(cid:14)
dx dt + 1
ε
(cid:6)
(cid:6)
ψ
ST
∂(cid:3)
=
ψ
ST
(cid:3)
(cid:4)εuε · ∇x G dx dt
for any ψ ∈ C 1(ST ),
cf. [10, Section 2.2].
|ϑε − ϑB |k (ϑε − ϑB )dσx dt
(3.7)
123
E. Feireisl et al.
3.1 Existence of approximate solutions
Our aim is to use the existence result proved in [10, Theorem 1] to obtain the approx-
imate solutions ((cid:4)ε, ϑε, uε)ε>0. To perform this step some comments are in order. In
comparison with [10], the present problem features the following new ingredients:
• The action of the Coriolis force in the momentum equation (3.5).
• The function ϑB in (3.1) is time dependent whereas its counterpart (cid:10)0 in [10]
depends only on x.
• The exponent k in (3.1) equals zero in [10].
It is easy to check that the existence proof in [10] can be modified to accommodate
the above changes as soon as suitable a priori bounds similar to those in [10, Section
2.4] are established. To see this, we start with the energy balance (3.7) with ψ ≡ 1
yielding
1
ε
(cid:6)
(cid:6)
(cid:6)
(cid:6)
∂(cid:3)
|ϑε − ϑB |k (ϑε − ϑB )dσx dt =
(cid:6)
ST
≤ M0(cid:15)∂t G(cid:15)L∞(ST ×(cid:3)), where M0 =
(cid:3)
ST
(cid:3)
(cid:4)ε dx.
(cid:4)εuε · ∇x G dx dt = −
(cid:6)
(cid:6)
ST
(cid:3)
(cid:4)ε∂t G dx dt
As ϑε > 0 a.a., (3.8) yields the bound
(cid:15)ϑε(cid:15)
Lk+1(ST ×∂(cid:3))
<
∼ 1
in terms of the data and uniform for ε → 0. Consequently, the entropy inequality (3.6)
gives rise to the bound on the entropy production rate
(cid:13)
(cid:6)
(cid:6)
ST
(cid:3)
1
ϑε
S : Dx uε − q · ∇x ϑε
ϑε
(cid:14)
dx dt ≤ c(ε, G, ϑB , M0)
(3.10)
and the remaining estimates are obtained exactly as in [10, Section 2.4]. Note that the
right–hand side of (3.10) may blow up for ε → 0.
With the necessary a priori bounds at hand, we obtain a family of approximate
solutions ((cid:4)ε, ϑε, uε)ε>0 exactly as in [10].
Proposition 3.1 (Approximate solutions) In addition to the hypotheses of Theorem 2.2,
let
ε > 0, 6 < k + 1 = β, M0 > 0
(3.11)
be given.
Then the approximate problem (3.2)–(3.7) admits a solution ((cid:4)ε, ϑε, uε).
123
(3.8)
(3.9)
Time periodic motion of temperature...
3.2 Approximate ballistic energy balance
Let ˜ϑ ∈ C 1(ST × (cid:3)) satisfy (2.18). Choosing ϕ(t, x) = ψ(t) ˜ϑ(t, x), where ψ ∈
C 1(ST ), ψ ≥ 0, as a test function in the approximate entropy inequality (3.6) and
adding the resulting integral to the energy balance (3.7), we deduce
(cid:4)ε|uε|2 + (cid:4)εe − ˜ϑ(cid:4)εs
dx dt +
(cid:14)
(cid:6)
(cid:6)
ψ
ST
(cid:3)
˜ϑ
ϑε
(cid:13)
S : Dx uε − q · ∇x ϑε
ϑε
(cid:14)
dx dt
(cid:6)
−
(cid:6)
(cid:3)
(cid:13)
1
2
(cid:6)
ψ
ST
(cid:6)
∂(cid:3)
(cid:13)
∂t ψ
(cid:6)
ST
+ 1
ε
(cid:6)
|ϑε − ϑB |k+2
ϑε
dσx dt
≤
ψ
ST
(cid:3)
(cid:4)εuε · ∇x G − (cid:4)εsuε · ∇x
˜ϑ − q
ϑε
(cid:14)
· ∇x
˜ϑ − ∂t
˜ϑ(cid:4)εs
dx dt.
(3.12)
Inequality (3.12) is obviously a counter part of the ballistic energy balance (2.17) and
will be used in the forthcoming part to deduce the necessary bounds on the family of
approximate solutions.
4 Uniform bounds
In order to perform the limit ε → 0 in the family of approximate solutions obtained
in Proposition 3.1, we need uniform bounds independent of ε.
4.1 Mass conservation
Obviously, as the total mass of the fluid is conserved, we get
(cid:6)
M0 =
(cid:4)ε(t, ·) dx for all t ∈ ST ⇒ sup
t∈ST
(cid:3)
(cid:15)(cid:4)ε(t, ·)(cid:15)
L1((cid:3))
<
∼ 1.
(4.1)
4.2 Energy estimates
As both ∂(cid:3) and the boundary data ϑB are smooth, we may suppose that
(cid:19)x ϑB(t, ·) = 0 in (cid:3) for any t ∈ ST .
(4.2)
Choosing ψ = 1, ˜ϑ = ϑB in the ballistic energy inequality (3.12) we get
(cid:15)
(cid:6)
(cid:6)
ST
(cid:3)
(cid:6)
≤
ϑB
ϑε
(cid:6)
ST
(cid:3)
(cid:16)
(cid:6)
(cid:6)
κ(ϑε)|∇x ϑε|2
ϑε
S(Dx uε) : Dx uε +
(cid:13)
(cid:4)εuε · ∇x G − (cid:4)εs((cid:4)ε, ϑε)uε · ∇x ϑB +
dx dt + 1
ε
ST
κ(ϑε)∇x ϑε
ϑε
|ϑε − ϑB |k+2
ϑε
∂(cid:3)
dσx dt
(cid:14)
˜ϑ(cid:4)εs((cid:4)ε, ϑε)
· ∇x ϑB − ∂t
dx dt.
(4.3)
123
E. Feireisl et al.
By virtue of hypothesis (2.10) and Korn’s inequality, we obtain
(cid:15)uε(cid:15)2
W 1,2
0
((cid:3);Rd )
(cid:6)
<
∼
ϑB
ϑε
(cid:3)
S(Dx uε) : Dx uε dx.
Moreover, again by virtue of (2.10),
(cid:6)
(cid:3)
(cid:13)
|∇x ϑ
β
2ε |2 + |∇x log(ϑε)|2
(cid:14)
(cid:6)
<
∼
dx
ϑB
ϑε
κ(ϑε)|∇x ϑε|2
ϑε
(cid:3)
dx.
By Poincarè inequality (see e.g. Theorem 4.4.6 in [27]) we obtain that
(cid:6)
<
∼
∂(cid:3)
β
|ϑ
2ε |2 dσx +
(cid:17)
(cid:17)
(cid:17)
(cid:17)ϑ
β
2ε −
(cid:6)
(cid:3)
(cid:6)
∂(cid:3)
β
2ε dσx
ϑ
2
(cid:17)
(cid:17)
(cid:17)
(cid:17)
<
∼
dx
(cid:6)
∂(cid:3)
β
|ϑ
2ε |2 dσx
(cid:6)
(cid:3)
β
2ε |2 dx
|ϑ
(cid:6)
β
|∇x ϑ
2ε |2 dx,
+
(cid:3)
as well as
(cid:6)
| log(ϑε)|2 dx
(cid:3)
(cid:6)
<
∼
∂(cid:3)
| log(ϑε)|2 dσx +
(cid:6)
(cid:3)
|∇x log(ϑε)|2 dx.
Collecting the last three inequalities, hypothesis (3.11) and estimating the boundary
terms
β
|ϑ
2ε |2 + | log(ϑε)|2
<
∼
|ϑε − ϑB|k+2
ϑε
∼ 1
<
2ε
|ϑε − ϑB|k+2
ϑε
gives
(cid:18)
(cid:18)
β
(cid:18)
(cid:18)ϑ
2ε
<
∼
(cid:18)
(cid:18)
2
(cid:18)
(cid:18)
W 1,2((cid:3))
(cid:2)
1 + 1
2ε
+ (cid:15)log(ϑε)(cid:15)2
(cid:6)
W 1,2((cid:3))
|ϑε − ϑB|k+2
ϑε
∂(cid:3)
dσx +
(cid:6)
(cid:3)
ϑB
ϑε
κ(ϑε)|∇x ϑε|2
ϑε
(cid:3)
dx
.
Gathering the previous observations, we may infer that
(cid:15)
(cid:6)
(cid:18)
(cid:18)
(cid:18)
(cid:18)ϑ
β
2ε
(cid:18)
(cid:18)
2
(cid:18)
(cid:18)
+
(cid:2)
(cid:15)uε(cid:15)2
W 1,2
0
(cid:17)
(cid:6)
(cid:17)
(cid:17)
(cid:17)
1 +
ST
<
∼
+ (cid:15)log(ϑε)(cid:15)2
((cid:3);Rd )
(cid:13)
(cid:6)
(cid:4)εuε · ∇x G − (cid:4)εs((cid:4)ε, ϑε)uε · ∇x ϑB +
W 1,2((cid:3))
W 1,2((cid:3))
ST
(cid:3)
(cid:16)
(cid:6)
(cid:6)
dt + 1
ε
ST
κ(ϑε)∇x ϑε
ϑε
|ϑε − ϑB |β+1
ϑε
dσx dt
(cid:14)
∂(cid:3)
· ∇x ϑB − ∂t
˜ϑ(cid:4)εs((cid:4)ε, ϑε)
dx dt
.
(cid:3)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(4.4)
Now, as (cid:4)ε, uε solve the equation of continuity (2.13),
(cid:6)
(cid:6)
ST
(cid:3)
(cid:4)εuε · ∇x G dx dt = −
(cid:6)
ST
(cid:4)ε∂t G dt ≤ c(M0, G).
123
Time periodic motion of temperature...
In addition, denoting
K(ϑ) =
(cid:6) ϑ
1
κ(z)
z
dz,
we obtain, by virtue of (4.2),
(cid:6)
(cid:3)
κ(ϑε)∇x ϑε
ϑε
· ∇x ϑB dx =
(cid:6)
(cid:3)
∇x K(ϑε) · ∇x ϑB dx =
(cid:6)
∂(cid:3)
K(ϑε)∇x ϑB · ndσx .
Consequently, as κ satisfies hypothesis (2.10), we conclude
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:6)
(cid:3)
κ(ϑε)∇x ϑε
ϑε
· ∇x ϑB dx
(cid:17)
(cid:17)
(cid:17)
(cid:17) =
<
∼
Thus inequality (4.4) reduces to
(cid:6)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:2)
∂(cid:3)
K(ϑε)∇x ϑB · ndσx
(cid:6)
|ϑε − ϑB|β+1
ϑε
1 +
∂(cid:3)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:3)
dσx
.
(cid:15)
(cid:6)
ST
(cid:15)uε(cid:15)2
W 1,2
0
(cid:6)
(cid:6)
+ 1
ε
(cid:2)
<
∼
ST
(cid:6)
∂(cid:3)
(cid:6)
1 +
ST
(cid:3)
(cid:18)
(cid:18)
β
(cid:18)
(cid:18)ϑ
2ε
+
(cid:18)
(cid:18)
2
(cid:18)
(cid:18)
W 1,2((cid:3))
((cid:3);Rd )
(cid:16)
+ (cid:15)log(ϑε)(cid:15)2
W 1,2((cid:3))
dt
|ϑε − ϑB|β+1
ϑε
dσx dt
(cid:10)
|(cid:4)εs((cid:4)ε, ϑε)uε · ∇x ϑB| + |∂t
˜ϑ(cid:4)εs((cid:4)ε, ϑε)|
(cid:11)
(cid:3)
dx dt
.
(4.5)
In accordance with hypothesis (2.6), we decompose the entropy as
(cid:4)εs((cid:4)ε, ϑε) = (cid:4)εS
(cid:19)
(cid:20)
(cid:4)ε
3
2ε
ϑ
+ 4a
3
ϑ 3
ε .
Consequently, the radiation component may be handled as
(cid:6)
(cid:3)
|ϑ 3
ε uε · ∇x ϑB| dx ≤ δ(cid:15)uε(cid:15)2
L2((cid:3);Rd )
+ c(δ, ϑB)
(cid:6)
(cid:3)
ϑ 6
ε dx
123
for any δ > 0. Consequently, as β > 6, this term can be absorbed by the left–hand
side of (4.5) yielding
E. Feireisl et al.
(cid:15)
(cid:6)
(cid:15)uε(cid:15)2
W 1,2
0
(cid:6)
ST
(cid:6)
+ 1
ε
(cid:19)
ST
(cid:6)
∂(cid:3)
(cid:6)
<
∼
1 +
ST
(cid:3)
(cid:18)
(cid:18)
β
(cid:18)
(cid:18)ϑ
2ε
+
(cid:18)
(cid:18)
2
(cid:18)
(cid:18)
W 1,2((cid:3))
((cid:3);Rd )
(cid:16)
+ (cid:15)log(ϑε)(cid:15)2
W 1,2((cid:3))
dt
|ϑε − ϑB|β+1
ϑε
(cid:19)
(cid:20)
(cid:17)
(cid:17)
(cid:17)
(cid:4)εS
(cid:17)
(cid:17)
(cid:4)ε
3
2ε
ϑ
dσx dt
uε · ∇x ϑB
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
+
(cid:4)εS
(cid:19)
(cid:20)
(cid:4)ε
3
2ε
ϑ
(cid:20)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:17) dx dt
.
˜ϑ
∂t
(4.6)
Finally, following the arguments of [14, Section 4.4], we make use of the Third law
of thermodynamics enforced through hypothesis (2.8). Specifically, if
(cid:4)
ϑ 3
2
< r meaning (cid:4) < r ϑ 3
2 ,
we get, by virtue of (2.9),
0 ≤ (cid:4)S
(cid:3)
(cid:2)
(cid:4)
ϑ 3
2
(cid:10)
1 + r ϑ 3
2
(cid:9)
log
<
∼
+(r ϑ 3
2 ) + log
+(ϑ)
Consequently, we deduce from (4.6),
(cid:12)(cid:11)
.
(cid:16)
(cid:15)
(cid:6)
(cid:18)
(cid:18)
β
(cid:18)
(cid:18)ϑ
2ε
+
(cid:18)
(cid:18)
2
(cid:18)
(cid:18)
W 1,2((cid:3))
((cid:3);Rd )
|ϑε − ϑB|β+1
ϑε
dσx dt
(cid:15)uε(cid:15)2
W 1,2
0
(cid:6)
ST
(cid:6)
+ 1
ε
⎛
ST
∂(cid:3)
+ (cid:15)log(ϑε)(cid:15)2
W 1,2((cid:3))
dt
<
∼
⎜
⎜
⎝(cid:20)(r ) +
(cid:6)
(cid:6)
ST
(cid:3)
1⎧
⎨
⎩
≥r
(cid:4)ε
3
2ε
ϑ
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
⎫
⎬
⎭
(cid:4)εS
(cid:19)
(cid:20)
(cid:4)ε
3
2ε
ϑ
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:17) dx dt
uε · ∇x ϑB
(cid:6)
(cid:6)
+
ST
(cid:3)
1⎧
⎨
⎩
≥r
(cid:4)ε
3
2ε
ϑ
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
⎫
⎬
⎭
(cid:4)εS
(cid:19)
(cid:20)
(cid:4)ε
3
2ε
ϑ
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:17) dx dt
˜ϑ
∂t
⎞
⎟
⎟
⎠ ,
where (cid:20)(r ) → ∞ as r → ∞.
Now, again by hypothesis (2.8),
(cid:19)
0 ≤ 1⎧
⎨
⎩
⎫
⎬
S
≥r
⎭
(cid:4)ε
3
2ε
ϑ
123
(cid:20)
(cid:4)ε
3
2ε
ϑ
≤ S(r ) → 0 as r → ∞.
(4.7)
(4.8)
Time periodic motion of temperature...
In an analogous way we treat the term
conclude
!
!
ST
(cid:3) ∂t
˜ϑ(cid:4)εs dx dt. Going back to (4.8) we
(cid:15)
(cid:6)
(cid:15)uε(cid:15)2
ST
+ 1
ε
(cid:2)
(cid:6)
ST
<
∼
(cid:18)
(cid:18)
β
(cid:18)
(cid:18)ϑ
2ε
(cid:18)
(cid:18)
2
(cid:18)
(cid:18)
+
W 1,2
0
(cid:6)
∂(cid:3)
((cid:3);Rd )
|ϑε − ϑB|β+1
ϑε
(cid:6)
(cid:6)
(cid:16)
+ (cid:15)log(ϑε)(cid:15)2
W 1,2((cid:3))
dt
W 1,2((cid:3))
dσx dt
(cid:3)
(cid:20)(r ) + S(r )
ST
(cid:3)
(|(cid:4)εuε| + (cid:4)ε) dx dt
,
(cid:20)(r ) → ∞, S(r ) → 0 as r → ∞.
(4.9)
4.3 Pressure estimates
To close the estimates we have to control the density in terms of the integrals on the
right–hand side of (4.9). To this end, we use the nowadays standard pressure estimates
obtained via Bogovskii operator. Specifically, we use the quantity
ϕ(t, x) = B
(cid:6)
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:3)
(cid:14)
(cid:4)ω
ε dx
, ω > 0,
as a test function in the momentum equation (3.5). Here B denotes the operator enjoying
the following properties:
•
•
"
B : Lq
0
((cid:3)) ≡
v ∈ Lq ((cid:3))
(cid:6)
(cid:17)
(cid:17)
(cid:17)
(cid:3)
#
v dx = 0
→ W 1,q
0
((cid:3); Rd ), 1 < q < ∞;
(4.10)
divx B[v] = v;
• if v = divx g, with g ∈ Lq ((cid:3); Rd ), divx g ∈ Lr ((cid:3)), g · n|∂(cid:3) = 0, then
(cid:15)B[divx g](cid:15)
Lr ((cid:3);Rd )
<
∼ (cid:15)g(cid:15)
Lr ((cid:3);Rd ),
(4.11)
see Galdi [15, Chapter 3] or Geißert et al. [16]. Boundedness of the operator B stated
in (4.10), (4.11) will be systematically used in the estimates below.
123
After a straightforward manipulation (see e.g. [9]), we obtain
E. Feireisl et al.
p((cid:4)ε, ϑε)(cid:4)ω
(cid:2)(cid:6)
ε dx dt
(cid:3) (cid:2)(cid:6)
(cid:3)
(cid:4)ω
ε dx
(cid:3)
(cid:3)
(cid:4)ε(uε ⊗ uε) : ∇x B
(cid:4)ε(ω × uε) · B
S(ϑε, Dx uε) : ∇x B
(cid:4)ε∇x G · B
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
p((cid:4)ε, ϑε) dx
(cid:13)
(cid:6)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:6)
(cid:3)
dt
(cid:14)
(cid:4)ω
ε dx
(cid:14)
dx dt
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:6)
(cid:3)
(cid:4)ω
ε dx
(cid:6)
dx dt
(cid:14)
(cid:4)ω
ε dx
dx dt
(cid:3)
(cid:14)
(cid:4)ω
ε dx
dx dt
(cid:3)
(cid:6)
(cid:6)
ST
(cid:3)
(cid:6)
=
1
|(cid:3)|
(cid:6)
ST
(cid:6)
−
+
+
−
+
(cid:3)
ST
(cid:6)
(cid:6)
(cid:3)
ST
(cid:6)
(cid:6)
(cid:3)
ST
(cid:6)
(cid:6)
ST
(cid:6)
(cid:3)
(cid:6)
ST
(cid:3)
+ (ω − 1)
(cid:4)εuε · B[divx ((cid:4)ω
ε uε)] dx dt
(cid:13)
(cid:6)
ε divx uε − 1
(cid:4)ω
|(cid:3)|
(cid:4)εuε · B
(cid:6)
ST
(cid:3)
(cid:14)
(cid:4)ω
ε divx uε dx
dx dt. (4.12)
(cid:6)
(cid:3)
Since the total mass M0 is constant, the smoothing properties of B yield
(cid:18)
(cid:18)
(cid:18)
(cid:18)B
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:6)
(cid:3)
(cid:14)(cid:18)
(cid:18)
(cid:18)
(cid:18)
(cid:4)ω
ε dx
≤ c(M0) as soon as ω < 1
d
.
L∞(ST ×(cid:3);Rd )
Moreover, in accordance with hypotheses (2.3)–(2.5),
(cid:4) 5
3 + ϑ 4
<
∼ p((cid:4), ϑ)
<
∼ (cid:4) 5
3 + ϑ 4 + 1.
In view of these facts, inequality (4.12) gives rise to
dx dt ≤ c(M0)
ε dx dt
(cid:10)
(cid:6)
(cid:6)
ϑ 4
(cid:3)
1 +
ST
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:6)
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:3)
(cid:4)ε(uε ⊗ uε) : ∇x B
(cid:4)ε(ω × uε) · B
S(ϑε, Dx uε) : ∇x B
(cid:6)
(cid:14)
(cid:4)ω
ε dx
(cid:14)
(cid:3)
dx dt
(cid:4)ω
ε dx
(cid:6)
dx dt
(cid:14)
(cid:4)ω
ε dx
dx dt
(cid:3)
(cid:6)
(cid:6)
ST
(cid:4)
(cid:3)
(cid:6)
+ω
5
3
ε
(cid:6)
−
+
+
(cid:3)
ST
(cid:6)
(cid:6)
(cid:3)
ST
(cid:6)
(cid:6)
ST
(cid:3)
123
Time periodic motion of temperature...
(cid:6)
(cid:6)
+
ST
(cid:3)
+ (ω − 1)
(cid:4)εuε · B[divx ((cid:4)ω
ε uε)] dx dt
(cid:13)
(cid:6)
ε divx uε − 1
(cid:4)ω
|(cid:3)|
(cid:4)εuε · B
(cid:6)
ST
(cid:3)
(cid:14)
(cid:4)ω
ε divx uε dx
dx dt
(cid:11)
.
(cid:6)
(cid:3)
(4.13)
The following steps will be performed for d = 3. Obviously even better estimates
can be obtained if d = 2. First,
(cid:6)
(cid:3)
(cid:6)
(cid:17)
(cid:6)
(cid:17)
(cid:17)
(cid:17)
ST
<
∼
ST
<
∼ sup
t∈ST
where
Fixing
(cid:4)ε(uε ⊗ uε) : ∇x B
(cid:6)
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:3)
(cid:14)
(cid:4)ω
ε dx
dx dt
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:15)(cid:4)ε(cid:15)Lγ ((cid:3))(cid:15)uε(cid:15)2
L6((cid:3);R3)
(cid:6)
(cid:15)(cid:4)ω
ε (cid:15)Lq ((cid:3)) dt
(cid:15)(cid:4)ε(cid:15)Lγ ((cid:3))
(cid:15)uε(cid:15)2
W 1,2
0
((cid:3);R3) dt sup
t∈ST
ST
(cid:15)(cid:4)ω(cid:15)Lq ((cid:3)) dt,
q = 3γ
2γ − 3
> 1 provided γ > 3
2
.
γ = 5
3
, ω = 3γ
2γ − 3
= 1
15
(4.14)
and using the fact that the total mass M0 is conserved, we get
(cid:6)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:6)
(cid:13)
(cid:4)ε(uε ⊗ uε) : ∇x B
(cid:6)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:6)
(cid:3)
(cid:14)
(cid:4)ω
ε dx
dx dt
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:3)
ST
≤ c(M0) sup
t∈ST
(cid:15)(cid:4)ε(cid:15)Lγ ((cid:3))
(cid:15)uε(cid:15)2
W 1,2
0
((cid:3);R3) dt.
ST
Seeing that the integral containing the Coriolis force can be controlled in a similar
way we may rewrite (4.13) in the form
(cid:6)
(cid:6)
ST
(cid:3)
+ω
5
3
ε
(cid:4)
dx dt ≤ c(M0)
(cid:6)
(cid:10)
1 +
(cid:6)
(cid:6)
ST
(cid:3)
ϑ 4
ε dx dt
+ sup
t∈ST
(cid:6)
(cid:15)(cid:4)ε(cid:15)Lγ ((cid:3))
(cid:6)
ST
(cid:15)uε(cid:15)2
+
ST
(cid:3)
S(ϑε, Dx uε) : ∇x B
((cid:3);R3) dt
W 1,2
0
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:14)
(cid:4)ω
ε dx
dx dt
(cid:6)
(cid:3)
123
E. Feireisl et al.
(cid:14)
(cid:4)ω
ε divx uε dx
dx dt
(cid:11)
.
(4.15)
(cid:6)
(cid:3)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:6)
(cid:6)
+
ST
(cid:3)
+ (ω − 1)
(cid:4)εuε · B[divx ((cid:4)ω
ε uε)] dx dt
(cid:13)
(cid:6)
ε divx uε − 1
(cid:4)ω
|(cid:3)|
(cid:4)εuε · B
(cid:6)
ST
(cid:3)
In a similar way, we get
(cid:6)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
ST
<
∼
(cid:6)
(cid:3)
(cid:6)
ST
where
In addition,
(cid:4)εuε · B[divx ((cid:4)ω
ε uε)] dx dt
(cid:15)(cid:4)ε(cid:15)Lγ ((cid:3))(cid:15)uε(cid:15)
L6((cid:3);R3)(cid:15)(cid:4)ω
ε uε(cid:15)
Lq ((cid:3);R3) dt,
1
γ
+ 1
6
+ 1
q
= 1.
(cid:15)(cid:4)ω
ε uε(cid:15)
Lq ((cid:3);R3) ≤ (cid:15)uε(cid:15)
L6((cid:3);R3)(cid:15)(cid:4)ω
ε (cid:15)L p((cid:3)), where
1
q
= 1
6
+ 1
p
;
whence
(cid:17)
(cid:6) τ +1
(cid:6)
(cid:17)
(cid:17)
(cid:17)
τ
(cid:3)
(cid:4)u · B[divx ((cid:4)ωu)] dx dt
(cid:17)
(cid:17)
(cid:17)
(cid:17) ≤ c(M)
(cid:15)(cid:4)(cid:15)
sup
t∈(τ,τ +1)
(cid:6) τ +1
5
3 ((cid:3))
L
τ
(cid:15)u(cid:15)2
W 1,2
0
((cid:3);R3) dt
as soon as (4.14) holds.
Finally,
(cid:17)
(cid:6)
(cid:17)
(cid:17)
(cid:17)
(cid:6)
(cid:13)
(cid:4)εuε · B
(cid:6)
ε divx uε − 1
(cid:4)ω
|(cid:3)|
(cid:18)
(cid:18)
(cid:18)
(cid:18)B
L6((cid:3);R3)
(cid:3)
(cid:13)
(cid:14)
(cid:4)ω
ε divx uε dx
ε divx uε − 1
(cid:4)ω
|(cid:3)|
(cid:17)
(cid:17)
(cid:17)
(cid:17)
dx dt
(cid:6)
(cid:4)ω
ε divx uε dx
(cid:3)
(cid:14)(cid:18)
(cid:18)
(cid:18)
(cid:18)
dt,
Lq ((cid:3);R3)
(cid:15)(cid:4)ε(cid:15)Lγ ((cid:3))(cid:15)uε(cid:15)
(cid:3)
(cid:6)
ST
≤
ST
where
Here,
1
γ
+ 1
6
+ 1
q
= 1.
(cid:13)
(cid:18)
(cid:18)
(cid:18)
(cid:18)B
ε divx uε − 1
(cid:4)ω
|(cid:3)|
(cid:4)ω
ε divx uε dx
(cid:14)(cid:18)
(cid:18)
(cid:18)
(cid:18)
(cid:6)
(cid:3)
<
∼ (cid:15)(cid:4)ω
ε divx uε(cid:15)
Lr ((cid:3);R3), q = 3r
3 − r
,
Lq ((cid:3);R3)
123
Time periodic motion of temperature...
and
(cid:15)(cid:4)ω
ε divx uε(cid:15)
Lr ((cid:3);R3) ≤ (cid:15)uε(cid:15)
W 1,2
0
((cid:3);R3)
(cid:15)(cid:4)ω
ε (cid:15)L p((cid:3)), with
1
2
+ 1
p
= 1
r
.
Thus using (4.14) we may infer that
(cid:6)
(cid:6)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:4)εuε · B
(cid:3)
ST
≤ c(M0) sup
t∈ST
(cid:13)
ε divx uε − 1
(cid:4)ω
|(cid:3)|
(cid:6)
(cid:14)
(cid:4)ω
ε divx uε dx
dx dt
(cid:17)
(cid:17)
(cid:17)
(cid:17)
(cid:6)
(cid:3)
(cid:15)(cid:4)ε(cid:15)
5
3 ((cid:3))
L
ST
(cid:15)uε(cid:15)2
W 1,2
0
((cid:3);R3) dt.
Going back to (4.15) and summarizing the previous estimates we conclude
(cid:6)
(cid:6)
(cid:2)
(cid:6)
(cid:6)
+ω
5
3
ε
(cid:4)
ST
(cid:3)
dx dt ≤ c(M0)
(cid:6)
1 +
ST
(cid:3)
ϑ 4
ε dx dt
+ sup
t∈ST
(cid:6)
(cid:15)(cid:4)ε(cid:15)
(cid:6)
5
3 ((cid:3))
L
ST
(cid:15)uε(cid:15)2
+
ST
(cid:3)
S(ϑε, Dx uε) : ∇x B
((cid:3);R3) dt
W 1,2
0
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
The last step is estimating
(cid:6)
(cid:3)
S(ϑε, Dx uε) : ∇x B
(cid:6)
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
≤ (1 + (cid:15)ϑε(cid:15)
L4((cid:3)))(cid:15)uε(cid:15)
W 1,2
0
((cid:3);R3)
(cid:14)
(cid:3)
(cid:4)ω
ε dx
dx dt
(cid:6)
(cid:3)
, where ω = 1
15
(4.16)
.
(cid:14)
(cid:4)ω
(cid:3)
(cid:18)
(cid:18)
(cid:18)
(cid:18)∇x B
dx
dx
(cid:13)
ε − 1
(cid:4)ω
|(cid:3)|
(cid:14)(cid:18)
(cid:18)
(cid:18)
(cid:18)
(cid:4)ω
ε dx
(cid:6)
(cid:3)
L4((cid:3);R3)
≤ c(M)(1 + (cid:15)ϑε(cid:15)
L4((cid:3)))(cid:15)uε(cid:15)
W 1,2
0
((cid:3);R3)
.
We therefore conclude the pressure estimates:
(cid:6)
(cid:6)
ST
(cid:3)
+ω
5
3
ε
(cid:4)
(cid:19)
dx dt ≤ c(M0)
(cid:13)
1 +
(cid:20) (cid:6)
(cid:6)
(cid:6)
ST
(cid:3)
ϑ 4
ε dx dt
(cid:16)
+
1 + sup
t∈ST
(cid:15)(cid:4)ε(cid:15)
5
3 ((cid:3))
L
(cid:15)uε(cid:15)2
W 1,2
0
((cid:3);R3) dt
ST
, ω = 1
15
.
(4.17)
4.4 Uniform bounds for " → 0
As β > 6, we deduce from the inequalities (4.9), (4.17) that
(cid:6)
(cid:6)
(cid:2)
(cid:2)
(cid:3)
(cid:6)
ϑ 4
ε dx dt
ST
(cid:3)
<
∼
1 +
ST
β
2ε (cid:15)2
(cid:15)ϑ
W 1,2((cid:3)) dt
<
∼
1 +
(cid:6)
(cid:6)
ST
(cid:3)
(cid:3)
(cid:4)ε|uε| dx dt
123
provided we fix r = 1 in (4.9). Furthermore,
(cid:6)
(cid:6)
ST
(cid:3)
(cid:4)ε|uε| dx dt ≤ 1
2
≤ 1
2
(cid:6)
(cid:6)
ST
(cid:3)
(cid:4)ε dx dt + 1
2
T M0 + 1
2
(cid:19)
sup
t∈ST
(cid:15)(cid:4)ε(cid:15)
5
3 ((cid:3))
L
ST
(cid:6)
(cid:15)uε(cid:15)2
L5((cid:3);R3) dt
(cid:20)
E. Feireisl et al.
(cid:6)
(cid:6)
(cid:3)
ST
(cid:6)
(cid:4)ε|uε|2 dx dt
≤ c(M0)
1 + sup
t∈ST
(cid:15)(cid:4)ε(cid:15)
5
3 ((cid:3))
L
ST
(cid:15)uε(cid:15)2
W 1,2
0
((cid:3);R3) dt
.
Consequently, inequality (4.17) reduces to
(cid:6)
(cid:6)
(cid:3)
ST
ω = 1
15
+ω
5
3
ε
(cid:4)
dx dt ≤ c(M0)
1 +
.
(cid:15)
(cid:19)
(cid:20) (cid:6)
(cid:16)
1 + sup
t∈ST
(cid:15)(cid:4)ε(cid:15)
5
3 ((cid:3))
L
ST
(cid:15)uε(cid:15)2
W 1,2((cid:3);R3) dt
,
Next, going back to (4.9) we get
(cid:6)
(cid:15)uε(cid:15)2
W 1,2
0
((cid:3);R3) dt
ST
<
∼
(cid:2)
(cid:6)
(cid:6)
S(r )
ST
(cid:3)
((cid:4)ε|uε| + (cid:4)ε) dx dt + (cid:20)(r )
where, by means of the standard Sobolev embedding theorem,
(cid:6)
(cid:4)ε|uε| dx ≤ (cid:15)
(cid:3)
≤ c(M0)(cid:15)
√
√
(cid:4)ε(cid:15)
√
L2((cid:3))(cid:15)
(cid:4)ε(cid:15)
L3((cid:3))(cid:15)uε(cid:15)
L6((cid:3);R3)
(cid:4)ε(cid:15)
L3((cid:3))(cid:15)uε(cid:15)
W 1,2
0
((cid:3);R3)
.
(4.18)
(cid:3)
Consequently,
(cid:6)
(cid:15)uε(cid:15)2
W 1,2
0
((cid:3);R3) dt
ST
<
∼
(cid:2)
(cid:6)
S(r )
(cid:15)(cid:4)ε(cid:15)
ST
3
2 ((cid:3))
L
(cid:3)
dt + (cid:20)(r )
.
(4.19)
Now, introducing the total energy of the system,
E((cid:4), ϑ, u) = 1
2
(cid:4)|u|2 + (cid:4)e((cid:4), ϑ)
we first observe that
(cid:6)
sup
t∈ST
(cid:3)
E((cid:4)ε, ϑε, uε) dx
<
∼
(cid:2)
(cid:6)
(cid:6)
1 +
ST
(cid:3)
(cid:3)
E((cid:4)ε, ϑε, uε) dx dt
.
(4.20)
The estimate (4.20) follows from the mean value theorem and the ballistic energy
inequality (3.12). Indeed, in view of the uniform bounds established in Sect. 4.2, we
123
Time periodic motion of temperature...
first deduce (4.20) for the ballistic energy
E((cid:4)ε, ϑε, uε) − ϑB(cid:4)εs((cid:4)ε, ϑε),
and then use (2.9) to observe that the entropy part ϑB(cid:4)εs((cid:4)ε, ϑε) is a lower order
perturbation.
Now, we estimate the kinetic energy using (4.19),
(cid:6)
(cid:6)
(cid:4)ε|uε|2 dx dt
(cid:6)
ST
(cid:3)
(cid:15)(cid:4)ε(cid:15)
≤ sup
t∈ST
3
2 ((cid:3))
L
(cid:15)uε(cid:15)2
L6((cid:3);R3) dt
ST
(cid:6) τ +1
≤ c sup
t∈ST
(cid:15)(cid:4)ε(cid:15)
3
2 ((cid:3))
L
τ
(cid:15)uε(cid:15)2
W 1,2
0
((cid:3);R3) dt
(cid:6)
<
∼ (cid:20)(r ) sup
t∈ST
(cid:15)(cid:4)ε(cid:15)
3
2 ((cid:3))
L
+ S(r ) sup
t∈ST
(cid:15)(cid:4)ε(cid:15)
3
2 ((cid:3))
L
ST
(cid:15)(cid:4)ε(cid:15)
3
2 ((cid:3))
L
dt.
In addition, by interpolation,
(cid:15)(cid:4)ε(cid:15)
3
2 ((cid:3))
L
≤ (cid:15)(cid:4)ε(cid:15)
5
6
L
5
3 ((cid:3))
(cid:15)(cid:4)ε(cid:15)
1
6
L1((cid:3))
.
(4.21)
Consequently,
(cid:6)
(cid:6)
ST
(cid:3)
(cid:4)ε|uε|2 dx dt ≤
<
∼ c(M0)(cid:20)(r ) sup
t∈ST
(cid:15)(cid:4)ε(cid:15)
5
6
L
5
3 ((cid:3))
+ S(r ) sup
t∈ST
(cid:15)(cid:4)ε(cid:15)
5
6
L
(cid:6)
5
3 ((cid:3))
ST
(cid:15)(cid:4)ε(cid:15)
5
6
L
5
3 ((cid:3))
dt.
(4.22)
Combining (4.18), (4.19), (4.21) we get
(cid:6)
(cid:6)
(cid:4)
ST
(cid:3)
+ω
5
3
ε
dx dt
(cid:19)
(cid:15)
≤ c(M0)
1 +
(cid:15)
≤ c(M0)
1 +
(cid:15)
1 + sup
t∈ST
(cid:19)
1 + sup
t∈ST
(cid:19)
(cid:15)(cid:4)ε(cid:15)
5
3 ((cid:3))
L
(cid:20) (cid:6)
ST
(cid:20) (cid:2)
(cid:15)uε(cid:15)2
W 1,2
0
((cid:3);R3) dt
(cid:6)
(cid:16)
(cid:3)(cid:16)
(cid:15)(cid:4)ε(cid:15)
5
3 ((cid:3))
L
S(r )
(cid:15)(cid:4)ε(cid:15)
(cid:20) (cid:2)
ST
(cid:6)
≤ c(M0)
1 +
1 + sup
t∈ST
(cid:15)(cid:4)ε(cid:15)
5
3 ((cid:3))
L
S(r )
(cid:15)(cid:4)ε(cid:15)
ST
dt + (cid:20)(r )
3
2 ((cid:3))
(cid:3)(cid:16)
dt + (cid:20)(r )
.
5
3 ((cid:3))
L
5
6
L
(4.23)
123
E. Feireisl et al.
Interpolating L 1 and L
5
3
+ω and using boundedness of the total mass we have
(cid:6)
(cid:6)
ST
(cid:3)
(cid:4)
5
3ε dx dt ≤ c(M0)
(cid:2)(cid:6)
(cid:6)
ST
(cid:3)
(cid:3) 10
11
+ω
5
3
ε
(cid:4)
dx dt
provided ω = 1
15
.
(4.24)
Thus summing up (4.20)–(4.24) we may infer that
(4.20) ⇒ sup
t∈ST
(cid:15)
(cid:6)
(cid:6)
(cid:3)
(cid:6)
E((cid:4)ε, ϑε, uε) dx
(cid:19)
(cid:2)
(cid:6)
(cid:6)
(cid:3)
ST
(cid:3)
E((cid:4)ε, ϑε, uε) dx dt
(cid:20)
<
∼
1 +
(cid:18)
(cid:18)
β
(cid:18)
(cid:18)ϑ
2ε
(cid:18)
(cid:18)
2
(cid:18)
(cid:18)
+ (cid:15)log(ϑε)(cid:15)2
W 1,2((cid:3))
dx dt
<
∼
1 +
(cid:6)
ST
+
(4.9) ⇒
(cid:3)
ST
(cid:6)
(cid:15)uε(cid:15)2
W 1,2
0
((cid:3);Rd )
(cid:6)
+
(cid:6)
W 1,2((cid:3))
(cid:14)
(cid:4)
ST
(cid:3)
5
3ε dx dt
(cid:6)
(cid:6)
(cid:4)ε|uε|2 dx dt +
ST
(cid:3)
5
3ε dx dt
(cid:3)
<
∼
(cid:6)
(cid:6)
(cid:4)ε|uε|2 dx dt +
(cid:13)
1 +
(cid:13)
1 +
ST
(cid:6)
(cid:3)
(cid:6)
(cid:4)
ST
(cid:3)
(4.22) ⇒
<
∼
+ (cid:20)(r )
(cid:19)
(cid:6)
sup
t∈ST
(cid:3)
E((cid:4)ε, ϑε, uε) dx
+ S(r )
(cid:15)
(cid:2)(cid:6)
(cid:6)
(cid:3) 10
11
+ω
5
3
ε
(cid:4)
dx dt
ST
(cid:3)
(4.24) ⇒≤ c(M0)
1 +
(cid:20) 1
2
(cid:20) 1
2
(cid:14)
5
3ε dx dt
(cid:4)
(cid:19)
(cid:6)
sup
t∈ST
(cid:3)
E((cid:4)ε, ϑε, uε) dx
+ (cid:20)(r )
(cid:19)
(cid:6)
sup
t∈ST
(cid:3)
(4.23) ⇒≤ c(M0)
+ (cid:20)(r )
(cid:19)
(cid:6)
sup
t∈ST
(cid:3)
E((cid:4)ε, ϑε, uε) dx
+ S(r )
(cid:15)
(cid:20)(r ) + S 10
11 (r )
(cid:6)
(cid:3)
(cid:19)
sup
t∈ST
(cid:20) 1
2
E((cid:4)ε, ϑε, uε) dx
+ S(r )
(cid:19)
(cid:6)
sup
t∈ST
(cid:3)
E((cid:4)ε, ϑε, uε) dx
(cid:20)
E((cid:4)ε, ϑε, uε) dx
(cid:19)
(cid:6)
sup
t∈ST
(cid:3)
E((cid:4)ε, ϑε, uε) dx
(4.25)
As S(r ) → 0 as r → ∞, we fix r > 0 large enough to deduce from (4.25) the desired
energy bound
(cid:6)
sup
t∈ST
(cid:3)
E((cid:4)ε, ϑε, uε) dx ≤ c(M0).
(4.26)
123
(cid:20)⎤
⎦
(cid:20)⎤
⎦
(cid:20)⎤
⎦ .
Time periodic motion of temperature...
5 Convergence
Our ultimate goal is to perform the limit in the sequence of approximate solutions
((cid:4)ε, ϑε, uε)ε>0 to obtain the existence of the time–periodic solution claimed in The-
orem 2.2. With the energy estimate (4.26) at hand, this is nowadays well understood
routine matter. Indeed the test functions used in the entropy inequality (2.16) are
compactly supported thus unaffected by the boundary integral in its approximate coun-
terpart (3.6). Similarly, the approximate ballistic energy (3.12) is in fact stronger than
(2.17) due to the penalization
(cid:6)
(cid:6)
ψ
ST
∂(cid:3)
1
ε
|ϑε − ϑB|k+2
ϑε
dσx dt
<
∼ 1, ψ ≥ 0.
(5.1)
In particular, for ψ = 1, the above inequality together with the (3.12) yield
ϑε → ϑ weakly in L 2(0, T ; W 1,2((cid:3); Rd ))
with the limit trace ϑ|∂(cid:3) = ϑB as required in Theorem 2.2.
Consequently, the proof of convergence is exactly the same as in the existence
theory elaborated in [7] with the exception of the strong convergence of the density,
the “initial” value of which is unspecified in the periodic setting. Fortunately, the
compactness arguments based on Lions’ identity and boundedness of the oscillation
defect measure can be modified to accommodate the time periodic setting exactly as
in [10, Section 9.3]. Thus the proof of Theorem 2.2 can be completed.
6 Concluding remarks
In comparison with [10], the available a priori bounds do not allow to handle a general
driving force (cid:4)g in the momentum equation. Although the potential case g = ∇x G is
physically relevant, more general (non–potential) forces occur when the fluid is stirred
up by the motion of the container. A detailed inspection of the arguments in Sect. 4.3
reveals that they could be considerably improved in the case d = 2 due to the Sobolev
embedding W 1,2 ⊂ Lq for any finite q. Similar improvement may also be expected
in the case the total mass M0 is small, cf. Wang and Wang [26]. We therefore strongly
conjecture that the present result can be extended to a general driving force g provided
• either d = 2,
• or
(cid:6)
M0 =
(cid:4) dx
(cid:3)
is small enough with respect to the amplitude of g.
As potentiality of g was also used in the estimate (3.8) crucial for boundedness of
the approximate sequence, the proof of the above conjecture would require a different
kind of approximation scheme.
123
E. Feireisl et al.
Finally, let us discuss briefly the possibility of extending the results to non–smooth
spatial domains. In view of the Sobolev space theory, notably various embedding
theorems, one is tempted to say that everything works well for domains with Lipschitz
boundary. Indeed we believe that such an extension is possible, however, there are some
technical difficulties to overcome in the construction of the approximate solutions, see
e.g. Poul [22].
Data Availability Data sharing not applicable to this article as no data sets were generated or analysed
during the current study.
Declarations
Conflict of interest On behalf of all authors, the corresponding author states that there is no conflict of
interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence,
and indicate if changes were made. The images or other third party material in this article are included
in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If
material is not included in the article’s Creative Commons licence and your intended use is not permitted
by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
References
1. Bechtel, S.E., Rooney, F.J., Forest, M.G.: Connection between stability, convexity of internal energy,
and the second law for compressible Newtonian fuids. J. Appl. Mech. 72, 299–300 (2005)
2. Belgiorno, F.: Notes on the third law of thermodynamics. I. J. Phys. A 36, 8165–8193 (2003)
3. Belgiorno, F.: Notes on the third law of thermodynamics. II. J. Phys. A 36, 8195–8221 (2003)
4. Bˇrezina, J., Kagei, Y.: Decay properties of solutions to the linearized compressible Navier–Stokes
equation around time-periodic parallel flow. Math. Models Methods Appl. Sci. 22(7), 1250007 (2012).
(53)
5. Bˇrezina, J., Kagei, Y.: Spectral properties of the linearized compressible Navier–Stokes equation around
time-periodic parallel flow. J. Differ. Equ. 255(6), 1132–1195 (2013)
6. Cai, H., Tan, Z.: Weak time-periodic solutions to the compressible Navier–Stokes equations. Acta
Math. Sci. Ser. B (Engl. Ed.) 36(2), 499–513 (2016)
7. Chaudhuri, N., Feireisl, E.: Navier–Stokes–Fourier system with Dirichlet boundary conditions. arxiv
preprint No. arXiv:2106.05315 (2021)
8. Davidson, P.A.: Turbulence: An Introduction for Scientists and Engineers. Oxford University Press,
Oxford (2004)
9. Feireisl, E., Matušu-Neˇcasová, Š, Petzeltová, H., Straškraba, I.: On the motion of a viscous compressible
flow driven by a time-periodic external flow. Arch. Rational Mech. Anal. 149, 69–96 (1999)
10. Feireisl, E., Mucha, P., Novotný, A., Pokorný, M.: Time periodic solutions to the full Navier–Stokes–
Fourier system. Arch. Rational. Mech. Anal. 204, 745–786 (2012)
11. Feireisl, E., Novotný, A.: Singular limits in thermodynamics of viscous fluids. In: Advances in Math-
ematical Fluid Mechanics, 2nd edn. Birkhäuser/Springer, Cham (2017)
12. Feireisl, E., Novotný, A., Petzeltová, H.: On the existence of globally defined weak solutions to the
Navier–Stokes equations of compressible isentropic fluids. J. Math. Fluid Mech. 3, 358–392 (2001)
13. Feireisl, E., Petzeltová, H.: On the long-time behaviour of solutions to the Navier–Stokes–Fourier
system with a time-dependent driving force. J. Dyn. Differ. Equ. 19(3), 685–707 (2007)
14. Feireisl, E., ´Swierczewska-Gwiazda, A.: The Rayleigh–Bénard problem for compressible fluid flows.
arxiv preprint No. arXiv:2110.10198 (2021)
123
Time periodic motion of temperature...
15. Galdi, G.P.: An Introduction to the Mathematical Theory of the Navier–Stokes Equations. I. Springer,
New York (1994)
16. Geißert, M., Heck, H., Hieber, M.: On the equation div u = g and Bogovski˘ı’s operator in Sobolev
spaces of negative order. In: Partial Differential Equations and Functional Analysis, Volume 168 of
Oper. Theory Adv. Appl., pp. 113–121. Birkhäuser, Basel (2006)
17. Jin, Ch., Yang, T.: Time periodic solution to the compressible Navier–Stokes equations in a periodic
domain. Acta Math. Sci. Ser. B (Engl. Ed.) 36(4), 1015–1029 (2016)
18. Kagei, Y., Oomachi, R.: Stability of time periodic solution of the Navier–Stokes equation on the
half-space under oscillatory moving boundary condition. J. Differ. Equ. 261(6), 3366–3413 (2016)
19. Kagei, Y., Tsuda, K.: Existence and stability of time periodic solution to the compressible Navier–
Stokes equation for time periodic external force with symmetry. J. Differ. Equ. 258(2), 399–444 (2015)
20. Lions, P.-L.: Mathematical Topics in Fluid Dynamics, Vol. 1, Incompressible Models. Oxford Science
Publication, Oxford (1996)
21. Lions, P.-L.: Mathematical Topics in Fluid Dynamics, Vol. 2, Compressible Models. Oxford Science
Publication, Oxford (1998)
22. Poul, L.: Existence of weak solutions to the Navier–Stokes–Fourier system on Lipschitz domains.
In: Discrete Contin. Dyn. Syst., Proceedings of the 6th AIMS International Conference, suppl., pp.
834–843 (2007)
23. Tsuda, K.: Existence and stability of time periodic solution to the compressible Navier–Stokes–
Korteweg system on R3. J. Math. Fluid Mech. 18(1), 157–185 (2016)
24. Valli, A.: Navier-Stokes equations for compressible fluids: global estimates and periodic solutions.
In: Nonlinear Functional Analysis and its Applications, Part 2 (Berkeley, Calif., 1983), Volume 45 of
Proc. Sympos. Pure Math., pp. 467–476. Amer. Math. Soc., Providence (1986)
25. Valli, A., Zajaczkowski, M.: Navier–Stokes equations for compressible fluids: global existence and
qualitative properties of the solutions in the general case. Commun. Math. Phys. 103, 259–296 (1986)
26. Wang, X., Wang, W.: On global behavior of weak solutions to the Navier–Stokes equations of com-
pressible fluid for γ = 5/3. Bound. Value Probl. 2015, 176 (2015). (13)
27. Ziemer, W.: Weakly Differentiable Functions. Springer, Berlin (1989)
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.
123
| null |
10.1017_s0260210522000614.pdf
| null | null |
Review of International Studies (2023), 49: 4, 763–779
doi:10.1017/S0260210522000614
R E S E A R C H A R T I C L E
From savages to snowflakes: Race and the enemies of
free speech
Darcy Leigh*
Sussex Law School, School of Law, Politics and Sociology, Freeman Centre, University of Sussex, Brighton, United Kingdom
*Corresponding author. Email: [email protected]
(Received 5 July 2021; revised 28 July 2022; accepted 1 September 2022)
Abstract
Right-wing free speech advocacy is increasingly shaping global politics. In IR, free speech has generally
been viewed within human rights and international legal frameworks. However, this article shows that
contemporary free speech advocates often ignore or oppose human rights and international law, focusing
instead on (what they describe as) a defence of the nation state against the enemies of free speech. This
article examines this articulation of free speech’s enemies: first historically as the ‘savage’ in John Stuart
Mill’s influential formulation of free speech; and then contemporarily as the ‘snowflake’, ‘mob’, and ‘cul-
tural Marxist’ by elected officials and lobbyists in the UK and US. The article argues that John Stuart Mill’s
savage is figured within a racialised civilisational hierarchy of degrees of humanity. Today, right-wing free
speech advocates extend and reconfigure this hierarchy, imagining the ‘snowflake’, ‘mob’, and ‘cultural
Marxist’ as lesser human, subhuman, and extra-human, respectively. Thus, in contrast to rights-based
analyses of free speech advocacy – which assume or assess the promotion of rights as a ‘public good’ –
the article argues that narratives of free speech’s enemies are deployed by right-wing free speech advocates
to underwrite racialised policy responses and global hierarchies.
Keywords: Free Speech; Far Right; Race; The Human; White Supremacy
Introduction
On 3 July 2020, on the eve of US Independence Day, former US President Donald Trump spoke
at Mount Rushmore in defence of free speech.1 According to Trump, the censorious enemies of
free speech were engaged in a ‘merciless campaign to wipe out our history … erase our values,
and indoctrinate our children.’2 These enemies had, in Trump’s narrative, taken over state and
societal institutions, instituting ‘extreme indoctrination and bias’ in which left-wing domination
was enforced through the threat of being ‘censored, banished, blacklisted, persecuted, and pun-
ished’.3 Trump described Black Lives Matter (BLM) protests, then taking place globally, as a par-
ticular threat: these ‘angry mobs’ were attacking the free expression of American nationalism and
global civilisation. The ‘mob’ was variously criminalised (‘unleash[ing] a wave of violent crime in
our cities’), lacking rationality (having ‘no idea why they are doing this’), and/or highly inten-
tional (‘some know why they are doing this’).4 Throughout, Trump used the language of war.
US citizens had ‘fought’, ‘struggled’, and ‘bled’ to secure freedom of speech, which was now
1Donald Trump, ‘Speech at Mount Rushmore’, South Dakota, 3 July 2020, available at: {rev.com/blog/transcripts/donald-
trump-speech-transcript-at-mount-rushmore-4th-of-july-event} accessed 20 February 2021.
2Ibid
3Ibid
4Ibid
© The Author(s), 2023. Published by Cambridge University Press on behalf of the British International Studies Association. This is an Open
Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which
permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
764
Darcy Leigh
under ‘attack’ and ‘radical assault’ from the dangerous ‘weapon’ of ‘cancel culture’.5 The
American people were not ‘weak’ but ‘strong’, and ready to fight in defence of ‘the nation’s chil-
dren’.6 Trump closed by announcing the creation of the ‘National Guard of American Heroes’: a
‘vast outdoor park’ in which statues of ‘the greatest Americans who have ever lived’ would defend
America and civilisation against the enemies of free speech.7
Trump’s speech embodies the concerns of a right-wing free speech movement that has become
increasingly voluble and influential in the Global North during the last decade.8 The speech also
illustrates the failure of IR to address this development or its significance in global politics. Free
speech in IR is usually viewed, as by some Constructivist IR scholars, as a human right located within
international legal frameworks.9 These scholars join a rich literature beyond IR, in Philosophy, Law
and Media Studies, which explores the legal or practical scope of a right to free speech.10 Yet con-
temporary right-wing free speech advocates tend not to reference or act on – or are actively opposed
to – international law and/or human rights.11 Further, while Constructivist IR research tends to focus
on less powerful actors using rights frameworks to challenge power inequities,12 right-wing free
speech advocates often have disproportionally large public platforms, which they use to consolidate
existing hierarchies.13 In this light, a focus on human rights and international law is ill equipped to
grasp the nature of contemporary right-wing free speech advocacy, which, as illustrated by Trump’s
speech, is more often concerned with securing the nation against its enemies.
If contemporary right-wing free speech advocacy does not uphold (or even address) human
rights, international law and/or a defence of the voiceless, what is its function? To answer this
question, this article examines the articulation of free speech’s enemies as a central feature of con-
temporary free speech advocacy. The article argues that free speech advocates locate their enemies
on a hierarchy of development, via an account of their proximity to whiteness, statehood, and
humanity. Historically, this civilisational rationality was made integral to free speech in ‘the
most famous liberal defence of free speech’,14 John Stuart Mill’s On Liberty, which also figured
‘the savage’ as a proto-enemy of free speech.15 Today, the enemies of free speech are figured
5Ibid
6Ibid
7Ibid
8See overviews of this movement in Gavin Titley, Is Free Speech Racist? (Cambridge, UK: Polity, 2020); P. Moskowitz, The
Case against Free Speech: The First Amendment, Fascism, and the Future of Dissent (New York, NY: Bold Type Books, 2019).
9D. C. Thomas, ‘The Helsinki effect’, in Thomas Risse, Stephen Ropp, and Kathryn Sikkink (eds), The Power of Human
Rights: International Norms and Domestic Change (Cambridge, UK: Cambridge University Press, 1999); D. C. Thomas, The
Helsinki Effect (Princeton, NJ: Princeton University Press, 2001); A. Callamard and L. Bollinger (eds), Regardless of Frontiers
(New York, NY: Columbia University Press, 2021). For broader Constructivist analyses of human rights norms, see Risse,
Ropp, and Sikkink (eds), The Power of Human Rights; Kathryn Sikkink, ‘Transnational politics, International Relations the-
ory, and human rights’, Political Science and Politics, 31:3 (1998), pp. 516–23; Martha Finnemore and Kathryn Sikkink,
‘Taking stock: The constructivist research program in International Relations and comparative politics’, Annual Review of
Political Science, 4 (2001), pp. 391–416.
10Eric Barendt, Freedom of Speech (Oxford, UK: Oxford University Press, 2007); Ivan Hare and James Weinstein (eds),
Extreme Speech and Democracy (Oxford, UK: Oxford University Press, 2011).
11For example, some free speech advocates who supported the UK exit from the EU oppose European human rights legis-
lation and promote ‘British liberties’ as a replacement for human rights. C. R. G., ‘Murray, Magna Carta’s tainted legacy:
Historic justifications for a British Bill of Rights and the case against the Human Rights Act’, in F. Cowell (ed.), The Case
Against the 1998 Human Rights Act: A Critical Assessment (London, UK: Routledge, 2017).
12Finnemore and Sikkink, ‘Taking stock’.
13This is illustrated, as Will Davies argues, by professors and journalists writing about their own censorship in major news
outlets. William Davies, ‘The free speech panic: How the right concocted a crisis’, The Guardian (26 July 2018), available at:
{https://www.theguardian.com/news/2018/jul/26/the-free-speech-panic-censorship-how-the-right-concocted-a-crisis}
accessed 6 December 2022.
14David van Mill, ‘Freedom of Speech’, Stanford Encyclopaedia of Philosophy (2017), available at: {plato.stanford.edu/
entries/freedom-speech} accessed 20 February 2021.
15John Stuart Mill and Elizabeth Rapaport, On Liberty (Cambridge, MA: Hackett Publishing, 1869). Hereafter ‘Mill, On
Liberty’.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
Review of International Studies
765
as ‘generation snowflake’, ‘the mob’, and the ‘cultural Marxist’. These figures – which variously
repeat, extend, and refigure a Millian civilisational racial hierarchy – are deployed, the article
shows, to enact and/or underwrite (especially racialised and/or colonial) statecraft and global
hierarchies.
The article proceeds in four sections. The first section locates right-wing free speech advocacy
in IR and, empirically, in global politics. The second section develops an analytic framework
based in Critical and Queer IR (on Cynthia Weber’s ‘figuration’ specifically), as well as Black
Studies scholarship.16 The third section reads John Stuart Mill’s account of free speech through
this framework, showing how statehood, whiteness, and free speech are connected, in the figure of
‘the savage’, through Mill’s civilisational rationality. The fourth section situates imagined contem-
porary enemies of free speech – ‘generation snowflake’, ‘the mob’, and ‘cultural Marxism’ – as
differently located within, informed by and/or revising Mill’s framework. The article concludes
with a discussion of the implications of its analysis for the populations who these figurations
are claimed to represent and for future IR research on free speech.
Locating right-wing free speech advocacy: In global politics and in IR
Research on free speech is relatively absent from IR. This section discusses three exceptions –
regarding human rights,17 right-wing populism18 and the securitising regulation of speech19 –
where IR scholarship directly or indirectly addresses an aspect of contemporary free speech advocacy.
In my reading, scholarship in these fields situates free speech as a human rights discourse poten-
tially open to co-optation or distortion, relating to a rising global populist movement, and
entangled with narratives of defence, sovereignty, and exceptionalism. Ultimately, however, the
section argues that these approaches fail to capture the significance of free speech advocacy as:
part of the white supremacist histories of the US and UK, as well as global imperialism more
broadly; undermining divisions between ‘moderate’ and ‘fringe’ right-wing politics; and deploy-
ing ‘freedom’ in racially stratifying ways (making a turn to ‘freedom’ a problematic response to
the racialised securitisation of regulation).
From the 1960s to the 1980s, free speech was a central demand of left wing, Black, women’s
and LGBT rights movements.20 Today, in Western liberal democracies, free speech advocacy is
16Cynthia Weber, Queer International Relations: Sovereignty, Sexuality and the Will to Knowledge (New York, NY: Oxford
University Press, 2016), pp. 28–33; see also Donna Haraway, Modest Witness@Second Millennium.FemaleMan Meets
OncoMouse: Feminism and technoscience (New York, NY: Routledge, 1997).
17Thomas, ‘The Helsinki effect’; Callamard and Bollinger (eds), Regardless of Frontiers; Risse, Ropp, and Sikkink (eds), The
Power of Human Rights; Sikkink, ‘Transnational politics, International Relations theory, and human rights’; Finnemore and
Sikkink, ‘Taking stock’.
18Sandra Destradi and Johannes Plagemann, ‘Populism and International Relations: (Un)predictability, personalisation,
and the reinforcement of existing trends in world politics’, Review of International Studies, 45:5 (2019,) pp. 711–30; Bice
Maiguashca, ‘Resisting the “populist hype”: A feminist critique of a globalising concept’, Review of International Studies,
45:5 (2019), pp. 768–85; Vedi Hadiz and Angelos Chryssogelos, ‘Populism in world politics: A comparative cross-regional
perspective’, International Political Science Review, 38:4 (2017), pp. 399–411; Pablo de Orellana and Nicholas Michelsen,
‘Reactionary internationalism: The philosophy of the New Right’, Review of International Studies, 45:5 (2019), pp. 748–67;
Jean-Francois Drolet and Michael C. Williams, ‘The radical Right, realism, and the politics of conservatism in postwar inter-
national thought’, Review of International Studies, 47:3 (2021), pp. 273–93.
19Nadya Ali, ‘Seeing and unseeing prevent’s racialised borders’, Security Dialogue, 51:6 (2020), pp. 579–96; Andrew Neal,
‘University free speech as a space of exception in Prevent?’, in Ian Cram (ed.), Extremism, Free Speech and Counter-Terrorism
Law and Policy (London, UK: Routledge, 2019); Randy Borum, ‘Rethinking radicalization’, Journal of Strategic Security, 4:4
(2011), pp. 1–6; P. R. Neumann, ‘The trouble with radicalization’, International Affairs, 89:4 (2013), pp. 873–93; Mark
Sedgwick, ‘The concept of radicalization as a source of confusion’, Terrorism and Political Violence, 22:4 (2010), pp. 479–
94; see also Rita Floyd, ‘Parallels with the hate speech debate: The pros and cons of criminalising harmful securitising
requests’, Review of International Studies, 44:1 (2017), pp. 43–6.
20Cynthia Enloe and Review of International Studies, ‘Interview with Professor Cynthia Enloe’, Review of International
Studies, 27:4 (2001), pp. 649–66.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
766
Darcy Leigh
more often associated with a range of right-wing movements, including those identified as centre-
right, right-wing populist, libertarian, and/or conservative. Constructivist scholarship is one of
the few fields in IR where free speech has been addressed, either as an explicit and central object
of analysis or, more often, within a broader package of international human rights norms or legal
frameworks.21 For example, Daniel Thomas examines how norms surrounding the right to free
speech circulate internationally, as well as how shared ideas, identities, or information contribute
to (or inhibit) the implementation of international law.22 Often focusing on authoritarian or post-
authoritarian states, such analyses tend to view free speech and its advocacy, along with rights
more broadly, as a public good and challenge to the powerful by the powerless.23 This leaves con-
structivist approaches ill-equipped to account for contemporary right-wing free speech advocacy
in Western liberal democracies, which often opposes human rights and international law, or
consolidates rather than challenging existing hierarchies.
Nonetheless, a rights-based approach illuminates some aspects of the landscape of contempor-
ary free speech politics. Assessed against the ‘successful’ diffusion or implementation of the right
to free speech, contemporary right-wing free speech advocates can be viewed as claiming but
failing to protect free speech as a right.24 Or, contemporary free speech advocates might be
viewed as deploying free speech rhetoric to legitimise right-wing political activities and/or to have
‘co-opted’ free speech from ‘the left’ and/or from international human rights advocates. This
argument is made in recent longform journalism by William Davies and Nesrine Malik.25
Yet this narrative alone misapprehends the history of free speech activism, which, as I show
elsewhere26 and illustrate in the discussion of Mill below, has been co-constituted with racialised
state formation and empire since the 1800s.27 That is, the racial stratification of modern state
formation was expressed and extended through free speech advocacy long before its recent uptake
by right-wing advocates. The implications of this history are obscured if we assume that
right-wing free speech advocacy can be fully explained as a ‘recent’ ‘co-optation’ of human rights
discourse. In this way, the article situates free speech within the co-constitution of liberalism,
modern statehood, and empire, observed by Critical IR scholars.28
For Mill, however, free speech is not simply one of many rights constituting state citizenship
but the principle upon which both statehood and international order are based.29 This article
argues that this state-forming role is taken up and rearticulated in contemporary right-wing
free speech advocates’ accounts of the enemies of free speech: in their accounts of their enemies
free speech advocates are not simply failing or dishonest in their claims to promote rights, but are
engaged in a long-running project of colonial and racialised statecraft enacted in the name of free
speech.30 This chronology undermines any straightforward narrative that the ‘public good’ of free
speech has been appropriated for harmful ends. In fact, this chronology suggests that even 1960s
21For example, Risse, Ropp, and Sikkink (eds), The Power of Human Rights; Sikkink, ‘Transnational politics, International
Relations theory, and human rights’; Finnemore and Sikkink, ‘Taking stock’.
22Thomas, ‘The Helsinki effect’.
23Finnemore and Sikkink describe this as a trend in Constructivist research in general. Finnemore and Sikkink, ‘Taking
stock’.
24Moskowitz shows that right-wing free speech advocates are often more invested in controlling or constraining speech
than ‘freeing’ it. Moskowitz, The Case against Free Speech.
25Davies, ‘The free speech panic’; Nesrine Malik, ‘The myth of the free speech crisis’, The Guardian (3 September 2019),
available at: {https://www.theguardian.Com/world/2019/sep/03/the-myth-of-the-free-speech-crisis} accessed 6 December
2022.
26Darcy Leigh, ‘The settler coloniality of free speech’, International Political Sociology, 16:3 (2022), pp. 1–16.
27I argue elsewhere that this is true from the emergence of modern free speech as a concept in the 1700s, but say 1800s
here because this is the time period addressed in this article. Leigh, ‘The settler coloniality of free speech’.
28Jens Bartelson, A Genealogy of Sovereignty (Cambridge, UK: Cambridge University Press, 1995); Jens Bartelson, The
Critique of the State (Cambridge, UK: Cambridge University Press, 2001).
29Mill, On Liberty.
30Leigh, ‘The settler coloniality of free speech’.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
Review of International Studies
767
left-wing free speech activism might – in a similar vein to Critical and Queer IR analyses of other
human rights movements31– be revisited and resituated in light of the racialised history of advo-
cacy for the right to free speech. This is not to say that all free speech advocacy is determined by
or reducible to the civilisational rationality embodied in Mill’s ‘savage’, or to foreclose how a range
of movements might be situated within the Mill’s legacy (resistance or alternative to that legacy
may be possible). Rather, this suggestion underscores the potential implications of interrupting
the chronology implied by a narrative of free speech as recently co-opted by the right, and refuses
to assume that left-wing expressions of free speech are unshaped by a racialising heritage.
A second field in IR that addresses an aspect of right-wing free speech advocacy is the growing
body of scholarship on the rise of the neofascist populist far-right and right-wing extremism.32
Although this scholarship does not address free speech itself, free speech is a central component
of the emergent far-right populist ‘reactionary internationalism’,33 which IR scholars show is
reshaping international politics. Free speech advocacy should be viewed, like far-right populist,
neofascist, and extremist movements, as international: even when free speech advocacy is
expressed as a concern with the decline of the nation,34 or an intrusion into the expression of
nationalism,35 these concerns are taken up and deployed internationally on both practical and
ideological levels.36 As such, despite this article’s focus on the UK and US, it addresses a move-
ment that spans Western Europe, North America, Australia, and Aotearoa/New Zealand.
However, not only are ‘fringe’, ‘extremist’, neofascist, far right, or populist politics not the pri-
mary object of this article, but the article calls into question an exceptionalist delineation of those
politics. The article shows that the figuration of free speech’s enemies is one way in which the
neofascist, extremist, and/or populist far right and more ‘moderate’ free speech advocates are
connected and collaborate: the enemies of free speech are figured similarly or jointly across a
wide spectrum of right-wing politics. In this way, right-wing free speech advocacy is entangled
with populist far-right politics via the figuration of the enemy of free speech. As such, rather
than addressing the populist far right directly, by centring the imagined enemies of free speech,
this article undermines any clear lines or exceptionalism surrounding far right populism.
A final field of IR scholarship relating to free speech addresses the regulation or constraint of
speech in the name of ‘counter-terror’37 and ‘deradicalisation’.38 In these cases, some speech is
designated as threatening to the security of the nation-state and in need of (often exceptional
or violent) constraint. This securitisation of the regulation of speech, some Critical IR scholars
31These arguments are often focused on the roles of women’s and LGBT rights in military intervention, border policies,
and neocolonialism, see, for example, Weber, Queer International Relations and Jasbir Puar, Terrorist Assemblages:
Homonationalism in Queer Times (Durham, NC: Duke University Press, 2007).
32Destradi and Plagemann, ‘Populism and International Relations’; Maiguashca, ‘Resisting the “populist hype”: A feminist
critique of a globalising concept’; de Orellana and Michelsen, ‘Reactionary internationalism’; Drolet and Williams, ‘The rad-
ical right’.
33This term is borrowed from de Orellana and Michelsen, ‘Reactionary internationalism’.
34As in Greg Lukianoff and Jonathan Haidt, The Coddling of the American Mind: How Good Intentions and Bad Ideas are
Setting up a Generation for Failure (London, UK: Penguin, 2018); Hara Estroff Marano, A Nation of Wimps: The High Cost of
Invasive Parenting (New York, NY: Broadway Books, 2008).
35As in Trump, ‘Speech at Mount Rushmore’.
36This was evidenced in March 2018, when Martin Sellner, the Austrian leader of far-right European group Generation
Identity, was denied entry to the United Kingdom. UK-based far-right leader Tommy Robinson then delivered Sellner’s
speech in his stead, citing the refused entry as censorship. Later it was revealed that both activists collaborate to circulate
funds internationally. James Poulter, ‘The far right are uniting around their right to free speech’, Vice (20 March 2018), avail-
able at:
{https://www.vice.com/en/article/j5ax9d/the-far-right-are-uniting-around-their-right-to-free-speech} accessed 20
February 2021; Ben Quinn, ‘Far-right fundraising not taken seriously by UK, report finds’, The Guardian (31 May 2019),
available at:
{https://www.theguardian.com/world/2019/may/31/far-right-fundraising-not-taken-seriously-uk-government-
extremists} accessed 20 February 2021.
37Ali, ‘Seeing and unseeing Prevent’s racialised borders’; Neal, ‘University free speech as a space of exception in Prevent?’.
38Borum, ‘Rethinking radicalization’; Neumann, ‘The trouble with radicalization’; Sedgwick, ‘The concept of radicalization
as a source of confusion’; Floyd, ‘Parallels with the hate speech debate’.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
768
Darcy Leigh
argue, underwrites white supremacy and other racial hierarchies. For example, analyses by
Nadia Ali39 and Andrew Neal40 show how defence of the state against terrorism via regimes of
speech is racialised, whether by assigning whiteness to narratives of the state41 or targeting com-
munities of colour in practice.42 While this scholarship addresses specific policy contexts (for
example, Prevent in the UK), and does not consider free speech or its imagined enemies expli-
citly, it does reflect the concerns of contemporary free speech advocates when it comes to figuring
the enemies of free speech, as well as the securitising and racially stratifying effects of this
figuration.
Yet focusing solely on the regulation of speech implies that the ‘unfreedom’ of regulation is in
some way tied to the ‘unfreedom’ of racialised state suppression43 – or, to put it another way, that the
racialised constraint of speech is an affront to free speech and/or could be corrected with freer speech.
Without disputing the observation that speech is restricted along racial lines, the current article com-
plicates any simple turn to ‘free speech’ or its advocacy as a response to the racialised constraint of
speech: the article shows that, through the racialised figuration of free speech’s enemies, calls for free
speech can restrict freedoms and enact white supremacy as much as calls for restriction do.
Overall, when free speech has been considered in IR, it has been primarily addressed within a
framework of rights as a ‘social good’ or international legal norm. This not only fails to account
for the contemporary right-wing expression of free speech, but risks obscuring a history in which
free speech is articulated through state-formation and racialised state violence. While free speech
is a concern of right-wing populist, extremist, or neofascist movements, centring the figuration of
the enemies of free speech shows that these movements are not exceptional nor fully distinct from
more ‘moderate’ politics. Finally, while calls for the regulation of speech highlight speech as a site
of racialised securitisation, they fail to address the ways in which, through references to an
imagined enemy, calls for free speech do not necessarily oppose, but rather extend, racially
hierarchical state formation. The following section further situates the current article within IR
scholarship, developing a methodology grounded in Critical, Queer, and Decolonial IR.
Analytic framework: Figuration, developmental temporality, and racialised degrees
of humanity
Since Richard Ashley’s 1989 account of ‘statecraft as mancraft’,44 which shows how sovereign state
formation is underwritten by the articulation of ‘sovereign man’, Critical, Feminist, and Queer IR
scholars have identified a range of figures through which modern statehood is constituted.
Echoing Ashley’s identification of both ‘man’ and ‘his others’ as constitutive of sovereign state
formation,45 IR scholarship on figures has focused both on those that stand in for the modern
state, and on the others, outsiders and threats, against which statehood is articulated. Such figures
include, for example, soldiers and statesmen,46 ‘mothers, monsters and whores’,47 diplomats,48
39Ali, ‘Seeing and unseeing’.
40Neal, ‘University free speech as a space of exception in Prevent?’.
41Ali, ‘Seeing and unseeing’.
42Neal, ‘University free speech as a space of exception in Prevent?’.
43This is illustrated by Neal’s discussion of whether or not Prevent unfairly targets or constrains people of colour in uni-
versities. Neal, ‘University free speech as a space of exception in Prevent?’.
44Richard Ashley ‘Living on border lines: Man, poststructuralism, and war’, in James Der Derian and Michael Shapiro
(eds), International/Intertextual Relations (New York, NY: Lexington Books, 1989), pp. 260–313.
45Ibid.
46Christine Sylvester, Feminist Theory and International Relations in a Postmodern Era (Cambridge, UK: Cambridge
University Press, 1994).
47Laura Sjoberg and Caron E. Gentry, Mothers, Monsters, Whores: Women’s Violence in Global Politics (London, UK: Zed
Books, 2007).
(2020), pp. 573–93.
48Ann Towns, ‘“Diplomacy is a feminine art”: Feminised figurations of the diplomat’, Review of International Studies, 46:5
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
Review of International Studies
769
and, beyond the discipline of IR, the ‘monster, terrorist [and/or] fag’,49 and ‘the soldier and the
terrorist’.50 More recently, in a study of figures of ‘the homosexual’, Cynthia Weber labels the pro-
cess through which figures are articulated in global politics ‘figuration’, setting out a framework
for analysing figuration in IR.51 This section draws on and adapts Weber’s framework, centring
Weber’s focus on developmental temporality. It draws on Black Studies scholarship to add an
emphasis on the racialisation of ‘the human’ (or humanisation and dehumanisation). The article
subsequently locates the enemies of free speech among the many figures identified by IR scholars
as sites of global politics.
Weber describes how figures come to be seen as extant and stable through the process of
figuration, which occurs in practices, policies, ideas, and rhetoric.52 Figures do not correspond
to the groups they are claimed to represent, but are instead mobilised as statecraft to underwrite
policies and/or global hierarchies. For example, Weber shows how the figure of the ‘normal LGBT
rights holder’53 marks Western states as developed nations, legitimises their dominance in the
international sphere, and obscures inaction on issues affecting queer populations not represented
as normal (for example, on queer migration or homelessness). In contrast, the figure of the
‘perverse’ homosexual immigrant or terrorist justifies border and deportation policies aimed at
securing Western states against a ‘racially darkened’ dangerous threat, as well as international
intervention in the name of ‘development’.54
Weber’s analysis provides a framework for analysing free speech advocates’ focus on the
developmental status of their enemies. Weber argues that figuration relies on and reproduces a
developmental temporality, which subsequently underpins the policies and hierarchies enacted
by figuration.55 In doing so, Weber echoes Critical IR scholarship on temporality, which
shows that a developmental temporality is constitutive of liberal statehood and modern colonial
global order.56 Weber’s analysis shows that the relationship of figures to this temporality is com-
plex, eschewing binaries of ‘developed’ vs ‘underdeveloped’ or ‘past’ vs ‘present’. For example, the
‘normal’ LGBT rights holder is located as both advanced in comparison with the underdeveloped
‘perverse’ homosexual, and temporally universal in contrast to the provincial ‘perverse’ homosex-
ual.57 At the same time, some ‘perverse’ homosexuals are located as less developed within linear-
progressive time (as ‘underdeveloped’), or as stuck in the past or prior-to developmental time (as
‘undevelopable’).58 In the case of Weber’s homosexual, it is this developmental temporality that
informs, for example, the interventionist or anti-immigration policies and other statecraft justified
by these figures.
Given free speech advocates’ emphasis on the humanity (or lack thereof) of the enemies of free
speech, it is worth noting how ‘the human’ is situated within Weber’s developmental temporality.
Weber argues that ‘the human’ of human rights is situated within the universal, which is equated
49Jasbir Puar and Amit Rai, ‘Monster, terrorist, fag: The war on terrorism and the production of docile patriots’, Social
Text, 20:3 (2002) pp. 117–48.
50Adi Kuntsman, ‘The soldier and the terrorist: Sexy nationalism, queer violence’, Sexualities, 11:1–2 (2008), pp. 142–70.
51Weber, Queer International Relations; Weber borrows this term and concept from Haraway, Modest Witness@Second
Millennium.FemaleMan Meets OncoMouse
52Weber’s use of the term ‘figuration’ as both a verb and a noun emphasises the ongoing-ness of any figure that appears as
stable. Here, however, I use both ‘figure’ and ‘figuration’ for ease of reading: the term ‘figure’ should be read as expressing the
same unfolding process as ‘figuration’. Weber, Queer International Relations.
53Weber, Queer International Relations, p. 29.
54Ibid., pp. 31–5.
55Ibid., pp. 29–31; drawing on Donna Haraway, Modest Witness@Second Millennium.FemaleMan Meets OncoMouse.
56See, for example, Kimberly Hutchings, ‘Happy Anniversary! Time and critique in International Relations theory’, Review
of International Studies, 33:S1 (2007), pp. 71–89; Anna Agathangelou and Kyle Killian (eds), Time, Temporality and Violence
in International Relations: (De)Fatalizing the Present, Forging Radical Alternatives (New York, NY: Routledge, 2016).
57Weber, Queer International Relations, quotations from p. 32, argument made throughout book.
58Ibid.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
770
Darcy Leigh
with progress and development.59 This is underscored by poststructuralist,60 posthuman,61 and
decolonial IR62 scholars, who show that ‘the human’ more broadly is often articulated as a
white, non-disabled, heterosexual, Christian and male citizen-subject. This scholarship shows
that the figuration of this human – standing in for progress, citizenship, security, and sovereign
statehood – is integral to developmental and colonising global politics.
The racialisation of ‘the human’ – and the implications of this figuration for global politics –
is underscored in Black Studies scholarship on the dehumanisation of blackened figures.63 This
scholarship shows that blackness is often figured as animal, object, and/or otherwise sub-
human.64 As Zakiyyah Iman Jackson describes, blackness has been repeatedly dehumanised,
bestialised, or objectified, with a lack of (perceived) development or civilisation cited as evi-
dence of a lack of full humanity.65 This blackened subhumanity has legitimised and informed
anti-Black state formation, not least the transatlantic slave trade and imperialism. Especially
relevant to figurations of the enemy of free speech – who is often viewed as lacking the capacity
for rationality – Jackson draws attention to the ways that lack of development or humanity is
articulated through an assessment of Black minds and rationality as lacking self-conscious
rationality, or ‘the clarity of self-knowledge’.66 Both blackness and irrationality have also,
Jackson argues, been feminised and/or articulated in relation to deviant or ‘uncivilised’ femin-
inity. As I describe below, this blackened dehumanisation is especially, but not exclusively, res-
onant with right-wing free speech advocates’ narratives surrounding the ‘uncivilised’ ‘threat’
posed by anti-racist or Black activism.
Methodologically, then, the current article follows an adapted version of Weber’s approach
to figuration. It analyses books, articles, and speeches by right-wing free speech advocates –
specifically elected politicians and lobbyists – as sites of the figuration of free speech’s
enemies. The selection of these texts is not comprehensive, but each captures or circulates
a particularly central or influential narrative among free speech advocates (e.g., they coined
a term, informed a political response and/or are by high ranking politicians). The article does
not treat ‘snowflakes’, ‘the mob’, or ‘cultural Marxists’ as existent subjects, but rather inves-
tigates how their figuration in free speech advocacy informs policy and hierarchies. Like
Weber, the article emphasises temporality, situating free speech advocates’ own emphasis
on temporality within the developmental temporality of state formation and international
relations. Finally, following Jackson, the article considers the degrees of humanity attributed
to the enemies of free speech, especially when these are racialised and/or signalled by a
perceived lack of rationality.
59Weber, Queer International Relations.
60See, for example, Ashley, ‘Living on border lines’.
61Audra Mitchell,
‘Only human? A worldly approach to security’, Security Dialogue, 45:1 (2014), pp. 5–22; Erika
Cudworth, Stephen Hobden, and Emilian Kavalski (eds), Posthuman Dialogues in International Relations (London, UK:
Routledge, 2018); Erika Cudworth, and Stephen Hobden, Posthuman International Relations: Complexity, Ecologism and
Global Politics (London, UK: Zed, 2011).
62Vicki Squire,
‘Migration and the politics of “the human”: Confronting the privileged subjects of IR’, International
Relations, 34:3 (2020), pp. 290–308; Louisa Odysseos, ‘Prolegomena to any future decolonial ethics: Coloniality, poetics
and “being human as praxis”’, Millennium, 45:3 (2017), pp. 447–72; Audra Mitchell, International Intervention in a
Secular Age: Re-Enchanting Humanity? (London, UK: Routledge, 2014).
63Sylvia Wynter,
‘Unsettling the coloniality of being/power/truth/freedom: Towards the human, after man, its over-
representation – an argument’, The New Centennial Review, 3:3 (2003), pp. 257–337; Bénédicte Boisseron, Afro-Dog:
Blackness and the Animal Question (New York, NY: Columbia University Press, 2018); Zakiyyah Iman Jackson, Becoming
Human: Matter and Meaning in an Antiblack World (New York, NY: New York University Press, 2020).
64Ibid.
65Jackson’s discussion dehumanisation takes place in the introduction to Becoming Human, which subsequently seeks to
displace this analysis as the sole register in which blackness and humanity are analysed together. Jackson, Becoming Human,
p. 7.
66Ibid., p. 5.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
Review of International Studies
771
John Stuart Mill’s civilisational free speech and its ‘savage’ other
Working in the East India Company for thirty years, Mill was a colonial official in the mid-1800s
whose work shaped European empire and state-formation.67 Today, Mill is widely recognised as
‘the most influential liberal thinker’68 on free speech. His well-known defence of free speech in
On Liberty posits free speech as the most important principle in liberal states, with free expression
driving societal progress.69 This section shows how Mill’s theory of free speech operates through
the developmental temporality described by Weber, as well as the (connected) whitened version
of the human and rationality described by Jackson. I argue that Mill’s ‘savage’ other to free
speech, while not always viewed as a ‘threat’ as such, is nonetheless a proto-enemy of contempor-
ary figurations of free speech’s enemies. While Mill is not the only nor even the original free
speech theorist (John Locke before him advocated for greater ‘toleration’),70 he is exceptionally
influential. The analysis of his work offered here is deployed later in the article to illuminate
the civilisational logics that continue to underpin – or are otherwise taken up and rearticulated
by – contemporary right-wing free speech advocacy.
That a colonial framework underpins Mill’s work in general is well established.71 Yet the rela-
tionship between this civilisational framework and Mill’s account of free speech – not least as
expressed in Mill’s figure of ‘the savage’ – remains largely unexamined. One exception is my
own work on the settler colonial dimension of the genealogy of free speech, where I detail
how Mill articulates free speech through his colonising civilisational framework and vice versa,
making the two inseparable.72 In my reading of On Liberty, Mill makes the following set of
(somewhat circular) arguments: because statehood is the most rational and civilised form of gov-
ernance, state formation indicates that a society is civilised and rational, while the absence of state
formation indicates an absence of civilisation or rationality; because sovereign statehood is the
most civilised and rational form of governance, and free speech drives towards rationalism and
progressive civilisation, free speech should lead organically to state formation; only those societies
that are civilised and rational already (again, signalled by the occurrence of state formation),
should be granted free speech, and with it other citizenship rights and sovereign statehood.73
These are not abstract arguments, nor accounts of why colonial subjects did not speak (freely
or otherwise). Rather, these arguments legitimised ‘despotism’74 over colonial subjects, including
exclusion from participation in colonial states, and repression of Indigenous and Black cultures,
languages, and political systems. They also authorised colonial expansion and governance in the
1800s more broadly.75
Departing from this analysis, I deploy Weber’s framework of figuration here to situate Mill’s
‘savage’ as central to his account of civilisational free speech. Mill’s ‘savage’ or ‘barbarian’ is
figured as living in ‘… those backward states of society in which the race itself may be considered
as in its [infancy].’76 Mill describes the ‘savage’ as ‘wandering or thinly scattered over a vast tract
of country’), lacking ‘commerce’, ‘manufactures’, ‘agriculture’, ‘law’, ‘administration of justice’,
‘property’, or ‘intelligence’.77 For Mill, these forms of life define savagery as well as constituting
67Lynn Zastoupil, John Stuart Mill and India (Stanford, CA: Stanford University Press, 1994).
68van Mill, ‘Freedom of speech’.
69Mill, On Liberty; Barendt, Freedom of Speech.
70John Locke, An Essay Concerning Toleration (Indianapolis: Liberty Fund, 1685); for a reading of Locke’s work on free
speech in relation to Mill’s, see Leigh, The Settler Coloniality of Free Speech.
71Jahn, ‘Barbarian thoughts’; Zastoupil, John Stuart Mill and India; Mehta, Liberalism and Empire.
72Leigh, ‘The settler coloniality of free speech’, pp. 8–11.
73This reading of the first chapter of Mill, On Liberty, is given in Leigh, ‘The settler coloniality of free speech’, pp. 8–11.
74Mill, On Liberty, pp. 9–10.
75Uday Singh Mehta, Liberalism and Empire: A Study in Nineteenth-Century British Liberal Thought (Chicago, IL:
University of Chicago Press, 1999); Zastoupil, John Stuart Mill and India.
76Mill, On Liberty, pp. 9–10.
77John Stuart Mill, On Civilization (1836), p. 120.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
772
Darcy Leigh
a failure to form states or capitalist agricultural arrangements. Mill also figures the ‘savage’ with
direct reference to their unreadiness for free expression, as living in a ‘… state of things anterior
to the time when mankind have become capable of being improved by free and equal discus-
sion.’78 To reiterate, ‘being improved by free and equal discussion’ would, for Mill, mean state-
formation. Here we see how the figure of the ‘savage’ embodies Mill’s civilisational colonial
framework of free speech described above.
We also see both Weber’s developmental temporality and Jackson’s dehumanisation. The
terms ‘infancy’ and ‘anterior to’ signal the developmental temporal relations between the ‘savage’
or ‘barbarian’ and what Mill describes as ‘human beings in the maturity of their faculties’.79 The
emphasis on ‘the maturity of their faculties’ ties (what Mill sees as) the development of the
human mind to both the practice of and right to sovereign state formation.80 Significantly for
today’s free speech advocates, this infantilisation places ‘the savage’ outside the realm of legitimate
political participation. The circularity of the argument means that colonised peoples are only
entitled to ‘freedom’ of speech so long as that freedom is not expressed outside or against
European state formation or colonial governance. Otherwise, in the name of rationality and
civilisation, they are figured as unready for such freedom.
However, in the same way that today’s free speech advocates imagine a varied set of enemies of
free speech, so too Mill differentiated free speech’s others within a civilisational hierarchy.
Different colonial subjects were, for Mill, located at different points within the temporality of
development, with correlate rationales for varied regimes of British colonial governance in the
name of development and civilisation.81 In some cases, Mill deemed figures as more capable
of or susceptible to assimilation into rationality, civilisation, and statehood (this made Mill’s
work ‘progressive’ – and Mill a ‘radical’ – in contrast to his predecessors in colonial governance).
For example, Mill argued that Indian religious elites should be recruited by colonial officials to
assist in governing or civilising other Indians.82 In contrast, Indigenous peoples in Europe’s settler
colonies were figured as more lacking in modern human individuality, rationality, and civilised
political organisation, justifying violent tactics of colonial occupation.
In these ways, Mill establishes the tradition of free speech advocacy within a developmental
temporality and in relation to racialised degrees of humanity. He figures the ‘savage’ as the
‘other’ to free speech and is concerned with the savage’s lack of rationality and/or inability to self-
govern (and thus exclusion from the realm of the political). The following section turns to the
contemporary figuration of free speech’s enemies and shows how each is figured within, extends
or departs from a Millian hierarchy of civilisation. Contemporary right-wing free speech advo-
cates, it argues, follow Mill in promoting or enacting (often racialised) state policies based on
the civilisational status assigned to its enemies. The civilisational status accorded speech’s enemies
today not only echoes and repeats, but also refigures and reworks Mill’s framework, not least by
extending it through the hyper- or extra- human ‘cultural Marxist’.
Contemporary figurations of free speech’s enemies: The lesser-human infantile
‘snowflake’, subhuman animalistic ‘mob’, and extra-human puppeteer ‘cultural Marxist’
This section argues that today the enemies of free speech are figured as infantile (‘the snowflake’),
subhuman and animalistic (‘the mob’), and extra-human (‘the cultural Marxist’) in relation to
Mill’s civilisational hierarchy. Overall, the section argues that the enemies of free speech function
to inform policies and politics that ‘defend’ a whitened state against a racially darkened ‘enemy’ –
78Mill, On Liberty, pp. 9–10.
79Ibid.
80Ibid.
81Mehta, Liberalism and Empire.
82Zastoupil, John Stuart Mill and India, pp. 28–50.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
Review of International Studies
773
not least by placing anti-racist and other activism outside the realm of legitimate participation in
state politics. Mill’s ‘savage’ or civilisational framework are not uniformly reproduced in later
iterations of free speech advocacy – these latter iterations not only reproduce and extend, but
also – especially through the figure of the ‘cultural Marxist’ – rearticulate the racialised rationality
of free speech in new ways.
Infantile generation snowflake
The trope of ‘generation snowflake’ – now in wide public circulation – centres on the figure of the
young as weak, infantile, overly emotional, irrational, feminised, racialised, and/or deindividua-
lised.83 Generation snowflake is figured as a censorious threat to free speech but also a victim
of infantilisation by policymakers, educators, and parents (and, in turn, as a threat to and/or
marker of threatened national character).84 In this way, the snowflake is a lesser and undeveloped
human, but not always inhuman, and sometimes recoverable or developable.
In 2016, Claire Fox, a peer in the UK House of Lords and former Member of European
Parliament, as well as director of the think tank Academy of Ideas, offered an early public articu-
lation of ‘generation snowflake’. Fox says younger generations are weaker than previous genera-
tions (here she introduces the temporality of decline) and lacking in the robustness required for
free debate. Fox describes ‘generation snowflake’ as ‘thin-skinned’,85 ‘febrile’,86 ‘fragile’,87 and ‘too
mollycoddled and infantilised for the rough and tumble of real life’.88 According to Fox, weakness
is joined with emotionality to cloud the judgement of generation snowflake and makes it unable
to confront ideas or arguments as such (or as ‘just words’). Instead, as Fox argues elsewhere, when
faced with ideas and arguments they disagree with, generation snowflake becomes ‘hysterical’ and
‘can’t cope’.89 Describing the reaction of some school students who objected to her views on sex-
ual violence, Fox says, ‘Some of the girls were sobbing and hugging each other … while others
shrieked.’90 Similarly, describing a group of Muslim girls approaching her after another speech
to express their disagreement with her views on Islam, Fox says that their emotional reactions
prevented them from receiving her rational argument rationally.91
Here, Weber’s developmental temporality is visible in the figuration of generation snowflake.
Fox argues that members of ‘generation snowflake’ are underdeveloped, or wrongly developed, at
the level of their individual life experiences. At the same time, by articulating this as generational
and a departure from the trajectory of previous generations, Fox suggests this is a societal or
national developmental problem. Concerns with ‘the human’ embodied in an individual rational
mind are also present. Figuring the threat to free speech as generational deindividualises members
of generation snowflake. When a younger person objects to Fox’s speech, this objection is framed
as part of a generational ‘trend’, rather than political expression by an individual with the capacity
for thought or political agency.
Fox also racialises and genders the irrational ‘snowflake’ enemy of free speech by repeatedly
associating it with Islam. Even when talking about non-Muslims, Fox uses the term ‘offense
83As in Fox, I Find That Offensive!.
84As in Lukianoff and Haidt, The Coddling of the American Mind; Marano, A Nation of Wimps.
85Fox, I Find That Offensive!, p. 7.
86Ibid., p. 17.
87Ibid., p. 37.
88Ibid., p. 9.
89Claire Fox, ‘Why today’s young women are just so FEEBLE’, Mail Online (9 June 2016), available at: {https://www.daily-
mail.co.uk/femail/article-3632119/Why-today-s-young-women-just-FEEBLE-t-cope-ideas-challenge-right-view-world-says-
academic.html} accessed 20 February 2021.
90Ibid.
91Fox, I Find That Offensive!, pp. 6–7.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
774
Darcy Leigh
fatwas’ to associate what she sees as over-emotional irrationality with Islam more broadly.92
In the story above, Fox also draws on misogynist tropes of shrieking and hysteria. She combines
these with racialisation and deindividualisation into the ultimate ’snowflakes’: a group of emo-
tional and irrational Muslim girls.
The figuration of ‘generation snowflake’ informs a particular political response, as illustrated
by Greg Lukianoff and Jonathan Haidt’s influential The Coddling of the American Mind. Drawing
on a Cognitive Behavioural Therapy based psychological approach, Lukianoff and Haidt not only
analyse generation snowflake, but set out a programme to address the threat posed by ‘snowflakes’
to free speech. The programme draws on Cognitive Behavioural Therapy techniques, along with
metaphors of free debate as a ‘mental gymnasium’ or boxing ring.93 They argue that young people
need to participate in debate as they would a gym or sparring session, in order to develop their
strength for debate and disagreement, and to stop seeing themselves as weak. In the spirit of this
argument, Haidt
founded and now codirects the impactful US free speech organisation
Foundation for Individual Rights in Education, which supports legal action against US univer-
sities for perceived free speech violations (among other activities).
In contrast to Fox, Lukianoff and Haidt reindividualise generation snowflake. Yet the effects
are equally depoliticising. By suggesting the maldevelopment of generation snowflake can be cor-
rected through individual psychological redevelopment, Lukianoff and Haidt further deny the
rational thought and political agency of generation snowflake: they do not see collective youth
organising as political expression, instead figuring it as an individualised psychological problem.
They thus legitimise an interventionist, individualised, and pathologised response to opposition
to right-wing politics.94
In all these ways, the figuration of ‘generation snowflake’ echoes Mill’s account of the ‘savage’
and those who ‘lack the maturity of their faculties’.95 Unlike Mill’s savage, however, ‘generation
snowflake’ is also sometimes a victim of indoctrination. Yet like Mill’s ‘savage’, ‘snowflakes’ are
often seen as developable. This may be because ‘the snowflake’ is associated with universities,
which are, in turn, associated with whiteness, proximity to the state and access to institutions.
Overall, however, in the absence of such development or assimilation, ‘generation snowflake’ is
infantilised and depoliticised.
The criminal, animalistic, and subhuman ‘mob’
The trope of ‘the mob’ figures the enemies of free speech as animalistic, criminal, and often black-
ened. Here, I discuss the blackened animality, criminality, and threat to security of ‘the mob’,
before showing how, as with the snowflake, opponents of right-wing free speech advocates are
articulated as irrational, deindividualised, depoliticised. Unlike the snowflake, however, I suggest
that the mob appears as entirely subhuman, threatening and unassimilable within the terms of
free speech. I begin by discussing the blackened BLM ‘mob’, then consider the more generic
‘social justice mob’.
As illustrated by the Trump speech with which this article opened, ‘the mob’ is often asso-
ciated with anti-racist protesters, especially BLM and the removal or destruction of statues.
When BLM protests and statue removal took place in mid-2020, UK and US governments framed
their responses not as related to the politics of racism or antiracism, but with the rhetoric of free
speech. BLM protestors were figured as a censorious ‘mob’. The ‘mob’ figured by UK and US gov-
ernments in response to BLM was dehumanised and depoliticised through two key figurative
moves.
92Ibid., p. 18.
93As in Lukianoff and Haidt, The Coddling of the American Mind, p. 18.
94Ibid.
95Mill, On Liberty, pp. 9–10.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
Review of International Studies
775
First, ‘the mob’ was repeatedly articulated as animalistic and irrational. For example, then UK
Secretary of State for Housing, Communities and Local Government, Robert Jenrick, called pro-
testors ‘a baying mob’,96 equating BLM protestors with animals (‘baying’ is a noise made by a
pack of dogs). This directly echoes the white supremacist articulation of blackness as animalistic
described by Jackson. Jenrick’s bestialisation of BLM also figures the political expression of
opposition to racism – including the toppling of statues – as a noise unintelligible to humans.
Dehumanisation and animalisation were further expressed through claims that BLM protestors
were unable or unwilling to express their dissent through rational and civilised state channels.
For example, Jenrick argued that ‘what has stood for generations should be considered thought-
fully, not removed on a whim’,97 as if BLM protestors had not ‘thought’ or ‘considered’ their
actions but instead acted on some animalistic urge.
Second, the mob was repeatedly figured as criminal. UK Secretary of State Priti Patel and
Trump both reduced the protests to criminal acts, rarely mentioning BLM by name or even
using the words ‘race’ or ‘protest’. Trump (2020) variously called BLM protestors a ‘mob’, ‘van-
dals’, ‘violent extremists’, and arsonists, advocating ‘the full force of the law’ in response.98 Patel
similarly called the BLM protests ‘hooliganism and thuggery’.99 Criminalising the protests in this
way not only evoked stereotypes of working class and Black criminality, but also places interac-
tions between the state and BLM within the realms of criminal justice or exceptional security,
rather than politics.
The enemy of free speech is not figured as ‘the mob’ solely in response to BLM protests. The
term is also applied to left-wing activists or ‘social justice warriors’ more broadly.100 For example,
students protesting right-wing free speech advocates visiting campuses across the UK and US are
often figured as ‘mobs’ threatening free speech.101 Here, the racialisation of the enemy of free
speech by free speech activists functions in complex ways. While these mobs may not be black-
ened or otherwise racialised in the same way as BLM protestors, they may be implicitly racialised
via their articulation as animalistic, irrational, and uncivilised. At the same time, the naming of
these ‘social justice mobs’ as such avoids naming the politics of the groups the figure of ‘the mob’
is claimed to represent, which are often anti-racist or Black politics. In this way, race is evoked to
further criminalise the mob, or goes unnamed in order to depoliticise opposition to racism.
However, this does not mean the joining of blackness and animality in the trope of ‘the mob’
affects all those targeted by free speech activists equally. For example, while a majority white stu-
dent anti-racist group may be described as an animalistic mob by free speech activists, they may
also be figured as ‘snowflakes’, and it is unlikely that they will be responded to with the same state
violence as, for example, the majority black participants in a BLM protest. Images of the white
‘mob’ – from KKK lynching to the ‘storming’ of the US Capitol building in 2021 – further com-
plicate and extend this picture. Perhaps the ‘mob’ must be blackened to be fully criminalised and
securitised. It is also possible that applying the language of the ‘mob’ to white supremacist violent
risks naming animality or incivility rather than white supremacy as ‘the problem’.
96Cited in ‘Statues to get protection from "baying mobs"’, BBC News (17 January 2021), available at: {https://www.bbc.co.
uk/news/uk-55693020} accessed 20 March 2021.
97Ibid.
98Trump, ‘Speech at Mount Rushmore’.
99Speech to UK Conservative Party Conference 2020, cited in Patrick Daly, ‘Priti Patel slams XR and BLM activists for
“hooliganism and thuggery” during protests’, The Scotsman (4 October 2020), available at: {https://www.scotsman.com/
news/politics/priti-patel-slams-xr-and-blm-activists-hooliganism-and-thuggery-during-protests-2992424} accessed 20 April
2021.
100See, for example, by Stella Morabito, ‘What to learn from the social justice warrior who was eaten by his own mob’, The
Federalist (18 July 2018), available at: {https://thefederalist.com/2018/07/18/learn-social-justice-warrior-eaten-mob/} accessed
20 April 2021.
101See, for example, by Mathew Goodwin, ‘Mob rule is crushing free speech on campus’, The Times (30 June 2019), avail-
able at: {https://www.thetimes.co.uk/article/mob-rule-is-crushing-free-speech-on-campus-30269p6q9} accessed 20 March
2021.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
776
Darcy Leigh
Finally, the figuration of the ‘social justice mob’ as emerging in universities illustrates the over-
lapping of different figurations of free speech’s enemies – in this case ‘the mob’ and ‘the snow-
flake’. Often both tropes are mobilised simultaneously and in interconnected ways. Both
deindividualise and depoliticise the political opponents of right-wing free speech activists.
Both deny some degree of humanity, civilisation, and development among those opponents,
with a focus on their lack of capacity for rational thought, rational discussion, or political subject-
hood. However, while generation snowflake is brought into the realm of psychology (articulated
as over-emotional), the mob is situated in the realm of criminality and security (articulated as
violent and threatening). While the snowflake is articulated as vulnerable, the mob is articulated
as threatening. In these ways, while both the snowflake and the mob can be understood in relation
to Mill’s civilisational hierarchy, they are located differently within this hierarchy. Generation
snowflake is articulated as a lesser human threat to national character or progress and in need
of rescue or development (in need of CBT); the mob is articulated as subhuman and undevelop-
able threats to the rule of law (in need of incarceration or a military response).
The extra-human ‘cultural Marxist’
The trope of ‘cultural Marxism’ articulates a behind-the-scenes international conspiracy of Jewish
intellectuals who are taking over liberal institutions and replacing free speech with indoctrin-
ation.102 This section shows how figurations of the enemy of free speech as a ‘cultural Marxist’
rely on pre-existing antisemitic tropes of Jews as scheming, rich, and power-hungry. I argue
that the ‘cultural Marxist’ is figured as extra-human and hyper-modern in its organisation and
power, and as such as a threat to national sovereignty and state institutions.
To understand the figuration of ‘cultural Marxism’ it is necessary to understand how this
figure is deployed across ‘fringe’ neo-Nazi and alt-right groups (e.g.,
formed part of
Norwegian mass shooter Anders Breivik’s manifesto),103 as well as ‘mainstream’ party politics
(described below). The term originates with an explicit naming of cultural Marxists as Jews,
Jews as dangerous intellectuals or
and builds on an antisemitic tradition that paints
Bolsheviks, wandering and thus disloyal to states, and/or controlling or taking over world polit-
ics.104 Elected officials and lobbyists, however, tend to omit mentioning this heritage of the term
or explicitly naming Jews, even while all other elements of the far right conspiracy theory remain
intact. In this way, ‘cultural Marxism’ functions as a ‘dog whistle’ through which antisemitism is
expressed in state politics in a plausibly deniable way.105
it
A 2019 speech by Member of the UK Parliament and free speech advocate Suella Braverman
captures the way that ‘cultural Marxists’ are figured as enemies of free speech.106 Braverman
argues that, as a result of the overwhelming aims and power of ‘cultural Marxists’, ‘banning things
is becoming de rigueur’, ‘freedom of speech is becoming a taboo’ and ‘our universities … are
being shrouded in censorship and a culture of no-platforming’.107 This cultural Marxist takeover
102Tanner Mirrlees, ‘The Alt-Right’s discourse on “cultural Marxism”, Atlantis, 39:1 (2018), pp. 49–69.
103Andrew Berwick, A European Declaration of Independence (2011). This is searchable online but, following Sarah
Ahmed’s politics of citation, I decline to link to it here. See Sara Ahmed, Living a Feminist Life (Durham, NC: Duke
University Press, 2017). A survey of white supremacist texts deploying the trope of
including
Berwick’s manifesto, can be found in Mirrlees, ‘The Alt-Right’s discourse on “cultural Marxism”’.
‘cultural Marxism’,
104Bill Berkowitz, ‘Cultural Marixsm Catching On’, Southern Poverty Law Centre (15 August 2003), available at: {https://
www.splcenter.org/fighting-hate/intelligence-report/2003/cultural-marxism-catching} accessed 20 April 2021.
105For an analysis of this process, illustrated by a case study of the Australian far right, see Rachel Busbridge, Benjamin
‘Cultural Marxism: Far-right conspiracy theory in Australia’s culture wars’, Social
106Cited in Peter Walker, ‘Tory MP criticised for using antisemitic term “cultural Marxism”’, The Guardian (26 March
{https://www.theguardian.com/news/2019/mar/26/tory-mp-criticised-for-using-antisemitic-term-cul-
Moffitt, and Joshua Thorburn,
Identities, 26:6 (2020), pp. 722–38.
2019), available at:
tural-marxism} accessed 20 March 2021.
107Ibid.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
Review of International Studies
777
was, for Braverman, ‘absolutely damaging for our spirit as British people, and our genius, whether
it’s for innovation and science, or culture and civilisation … for statecraft’.108 As such, Braverman
argues, ‘Conservatives are engaged in a battle’ against these enemies.109 A similar enemy of free
speech was also figured by Trump at Mount Rushmore, as taking over ‘our schools, our news-
rooms, even our corporate boardrooms’.
Here, ‘cultural Marxists’ are viewed not simply as the political opponents of right-wing free
speech advocates, but rather – via their imagined threat to free speech – as the enemies of the
British nation and civilisation. In addition to being seen as disloyal threats to nationhood, and
as power-hungry or scheming, they are attributed the power and coordination necessary to
take over state institutions (rather than, for example, being seen as relatively limited and disem-
powered student, left wing, or Jewish groups).110
Once again, the relationships between different figurations of free speech’s enemies are blurry.
Is the cultural Marxist preying on vulnerable ‘snowflake’ youth, or creating them through a cen-
sorious orthodoxy? Are the same ‘coddled’ university students also predatory ‘cultural Marxists’?
For example, Braverman accused cultural Marxists of ‘putting everyone in cotton wool’, arguing
that ‘a risk-averse mentality is now taking over’.111 ‘Cotton wool’ is often, as it is for Fox, a sig-
nifier of ‘generation snowflake’.112 There is no one specific manifestation of the relationship of
‘cultural Marxism’ to other enemies of free speech: a range of narratives attendant to each circu-
late between and are combined multiply by right-wing free speech advocates. This echoes Weber’s
account of the complex interrelated developmental temporalities of figuration.
In all these ways, like the ‘snowflake’ and ‘mob’, the ‘Cultural Marxist’ is deindividualised, fig-
ured not as a human individual but a mass conspiracy. However, unlike the ‘snowflake’ and
‘mob’, the ‘Cultural Marxist’ is represented as hyper-rational and over-intelligent, rather than
irrational or incapable of thought. The cultural Marxist is not a ‘normal’ rational human citizen-
subject, but nor is this enemy a vulnerable infant or subhuman (despite sometimes overlapping or
connecting with vulnerable youth and ‘snowflakes’). Instead, this enemy of free speech is figured
as extra-human, hyper-strategic, and hyper-influential. The location of the ‘cultural Marxist’ does
not appear within Weber’s analysis of developmental temporality or Jackson’s analysis of the
human. Nor is it discussed by Mill in relation to civilisation. Instead, contemporary figurations
of ‘cultural Marxism’ extend the developmental temporality with which racialised degrees of
humanity are articulated into a distorted and threatening futurity.
Conclusion
This article has shown that Mill’s civilisational framework for free speech – embodied in his fig-
uration of ‘the savage’ – is reproduced and rearticulated in contemporary free speech advocates’
articulation of their enemies. The ‘snowflake’, ‘mob’, and ‘cultural Marxist’ are all figured through
and/or extend this framework. The article has further argued that the figuration of the enemies of
free speech as ‘generation snowflake’, ‘the mob’, and ‘cultural Marxism’ authorise right-wing free
speech advocates’ policymaking, depoliticise their opponents, and/or underwrite racialised hier-
archies. Before closing, I now consider some possible implications of this analysis. First, for the
populations which figured enemies are claimed to represent. Second, for researching free speech
advocacy beyond right-wing electoral expressions in the UK and US.
As Weber (2016) describes, figures do not correspond to the lived experience of subjects. In
fact, this article has observed how right-wing free speech advocates often apply ‘generation
108Ibid.
109Ibid.
110Berkowitz, ‘Cultural Marixsm Catching On’; Mirrlees, ‘The Alt-right’s discourse on “cultural Marxism”’; Moffitt and
Thorburn, ‘Cultural Marxism’.
111Moffitt and Thorburn, ‘Cultural Marxism’.
112Fox, I Find That Offensive!, p. 31.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
778
Darcy Leigh
snowflake’, ‘the mob’, and ‘the cultural Marxist’ (or aspects of these figures) to the very same
populations. This is clear in free speech advocates’ opposition to BLM protestors, who are ima-
gined both as ‘the mob’ and as a ‘cultural Marxist’ takeover. Similarly, university students are
framed as both sensitive ‘snowflake’ victims, and a ‘censorious Marxist mob’ stifling free expres-
sion. Given that each figure comes with its own political logic and implications – for example,
rescue, development or incarceration/securitisation – it is possible that how and when popula-
tions are figured as a particular ‘enemy’ reflects the broader (often racialised) politics of free
speech advocates in relation to those populations. This would account for the shifting and multi-
ply applied figurations of free speech’s enemies by free speech advocates depending on the
context.
While figurations do not correspond to the lived lives of subjects, the populations that figures
are claimed to represent may engage – or be forced to engage – the process of figuration.
According to Weber, particular figurations may be inhabited performatively and intentionally
or forcibly. For example, Weber suggests that some ‘“homosexuals” welcome the opportunity
to inhabit the image of the “LGBT rights holder’”, while others may find this figure constraining
and/or inaccessible. In a very different context, some Black Studies scholars argue that wilfully
embracing uncivility, the non-human and animality may be an opportunity for political solidar-
ity, agency, and organising.113 They note, however, that this comes with risks in a context where
the figuration of black people as subhuman is enforced, and might be co-opted, as a core function
of white supremacist violence. With regards to the enemies of free speech, it is likely that the loca-
tion of a figure within a civilisational framework determines, to some degree, the costs and oppor-
tunities embracing that figure represents: a Black activist embracing the criminality of ‘the mob’
may find themselves at greater risk than, for example, a white activist embracing that same figure,
or of either embracing the (potentially whitened) category of ‘generation snowflake’. At the same
time, perhaps the same outsider status of ‘the mob’, which legitimises violence may also make it a
politically potent and disruptive category. The question of whether or how the figures of ‘gener-
ation snowflake’, ‘the mob’, and/or ‘cultural Marxism’ might be embraced or inhabited remains
open.
Finally, what does this article’s analysis of free speech’s enemies mean for how we understand
free speech advocacy more broadly? The article has focused on right-wing conservative, libertar-
ian, and populist elected politicians and lobbyists in the UK and US. This focus reflects the
increasing dominance and influence of right-wing free speech politics in the Global North
today, which has not been accounted for by research in IR that tends to view free speech as solely
a public good, human right, and/or matter of international law. This leaves a wide range of con-
temporary free speech advocacy unexamined. In the US and UK, this includes both those who
identify as neo-Nazis or overt white supremacists and as left wing (notable examples of the latter
in the US are academics facing university censure for criticism of the state of Israel or use of
‘Critical Race Theory’). In other countries, it includes movements countering state censorship,
such as journalists and academics in Turkey, or religious minorities in China. In contrast to
the right-wing free speech advocates examined here, who often have disproportionately large
public platforms despite their claims to being victims of free speech’s enemies, some of these
other free speech advocates face severe, even carceral or lethal, penalties for advocating free
speech.
While the specifics of these varied cases put them beyond the scope of this article, and it is
absolutely not my intention to homogenise or dismiss all free speech advocacy, the article none-
theless raises questions free speech advocacy beyond its right-wing electoral expression in the US
and UK. At the very least, the article calls into question the framework of human rights, inter-
national law, and norm diffusion as the de facto sole lens through which all free speech advocacy
must be viewed. As I describe above, though such a lens might usefully assess free speech
113Jackson, Becoming Human; Bénédicte Boisseron, Afro-Dog.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
Review of International Studies
779
advocacy as more or less successful or disingenuous, it fails to capture the potentially productive
function of such advocacy within global racial hierarchies. More specifically, without foreclosing
the answer, the article raises the question of whether and how free speech advocates beyond UK
and US right-wing advocacy figure, racialise and/or (de)humanise their enemies. For those work-
ing within Mill’s legacy – which includes not only right-wing advocacy but also liberal multicul-
turalism and ‘equality and diversity’ agendas114 – the question is raised as to whether and how
Mill’s ‘savage’ and civilisational rationality persist or, perhaps, can be resisted.
In these ways, this article expands and updates the small IR literature on free speech that has
focused primarily on human rights diffusion, international law, and/or ‘progressive’ advocacy for
free speech. It does so empirically, by examining recent right-wing free speech advocacy in the US
and UK that often explicitly opposes human rights and international law. It does so methodo-
logically, by addressing how free speech advocates figure the enemies of free speech, including
how those enemies are racialised as human, subhuman, or extra-human. This shifts the analysis
of free speech away from instrumental questions about rights implementation towards discursive
and political ones. Free speech becomes visible as integral to a range of core IR concerns, not least
(in Mill’s account) sovereignty and (in Trump’s account) national security. Free speech’s enemies
become located among the constitutive figures of international politics.
Acknowledgements. This article has been improved by comments and/or support from Daniel Bulley, Harry Josephine
Giles, Laura Jung, Louiza Odysseos, Maddie Breeze, Matthew Evans, attendees of the Pan-European Conference on
International Relations 2019, members of the Centre for Rights and Anti-Colonial Justice at the University of Sussex, the
Editors of Review of International Studies, and anonymous reviewers.
Dr Darcy Leigh is a Lecturer in Law at the University of Sussex, where she researches the history and ongoing present of the
British Empire, with a focus on its settler colonial dimension and/or expression in gender and sexuality. Dr Leigh also teaches
about colonialism, gender, and sexuality in university, activist, and community contexts, using democratic and creative
pedagogies.
s
s
e
r
P
y
t
i
s
r
e
v
i
n
U
e
g
d
i
r
b
m
a
C
y
b
e
n
i
l
n
o
d
e
h
s
i
l
b
u
P
4
1
6
0
0
0
2
2
5
0
1
2
0
6
2
0
S
/
7
1
0
1
0
1
/
g
r
o
.
.
i
o
d
/
/
:
s
p
t
t
h
114I explore the question of multicultural policies and note its relevance to safer spaces activism elsewhere. Leigh, ‘The
settler coloniality of free speech’.
Cite this article: Leigh, D. 2023. From savages to snowflakes: Race and the enemies of free speech. Review of International
Studies
49, 763–779. https://doi.org/10.1017/S0260210522000614
| null |
10.1158_1078-0432.ccr-21-3817.pdf
|
t.
The somatic variant calls and normalized RNA-seq intensity data,
code, and deidentified clinical data are available here: https://github.
com/kbolton-lab/Bolton_OCCC. This will enable all the figures and
tables to be re-generated and also provide data for others for future
analyses. We will also make the BAMs/FASTQs available to research-
ers through contacting Kelly Bolton ([email protected]).
|
Data availability The somatic variant calls and normalized RNA-seq intensity data, code, and deidentified clinical data are available here: https://github. com/kbolton-lab/Bolton_OCCC . This will enable all the figures and tables to be re-generated and also provide data for others for future analyses. We will also make the BAMs/FASTQs available to researchers through contacting Kelly Bolton ([email protected]).
|
Washington University School of Medicine
Washington University School of Medicine
Digital Commons@Becker
Digital Commons@Becker
2020-Current year OA Pubs
Open Access Publications
11-14-2022
Molecular subclasses of clear cell ovarian carcinoma and their
Molecular subclasses of clear cell ovarian carcinoma and their
impact on disease behavior and outcomes
impact on disease behavior and outcomes
Kelly L Bolton
Washington University School of Medicine in St. Louis
Irenaeus C C Chan
Washington University School of Medicine in St. Louis
Brian J Wiley
Washington University School of Medicine in St. Louis
et al.
Follow this and additional works at: https://digitalcommons.wustl.edu/oa_4
Part of the Medicine and Health Sciences Commons
Please let us know how this document benefits you.
Recommended Citation
Recommended Citation
Bolton, Kelly L; Chan, Irenaeus C C; Wiley, Brian J; and et al., "Molecular subclasses of clear cell ovarian
carcinoma and their impact on disease behavior and outcomes." Clinical cancer research. 28, 22. 4947 -
4956. (2022).
https://digitalcommons.wustl.edu/oa_4/980
This Open Access Publication is brought to you for free and open access by the Open Access Publications at
Digital Commons@Becker. It has been accepted for inclusion in 2020-Current year OA Pubs by an authorized
administrator of Digital Commons@Becker. For more information, please contact [email protected].
CLINICAL CANCER RESEARCH | TRANSLATIONAL CANCER MECHANISMS AND THERAPY
Molecular Subclasses of Clear Cell Ovarian Carcinoma
and Their Impact on Disease Behavior and Outcomes
Kelly L. Bolton1, Denise Chen2, Rosario Corona de la Fuente3, Zhuxuan Fu4, Rajmohan Murali5,
Martin K€obel6, Yanis Tazi5, Julie M. Cunningham7, Irenaeus C.C. Chan1, Brian J. Wiley1, Lea A. Moukarzel5,
Stacey J. Winham7, Sebastian M. Armasu7, Jenny Lester8, Esther Elishaev4, Angela Laslavic4,
Catherine J. Kennedy9,10, Anna Piskorz11, Magdalena Sekowska11, Alison H. Brand9,12, Yoke-Eng Chiew9,10,
Paul Pharoah11, Kevin M. Elias13, Ronny Drapkin14, Michael Churchman15, Charlie Gourley15,
Anna DeFazio9,10,12,16, Beth Karlan8, James D. Brenton11, Britta Weigelt5, Michael S. Anglesio17,
David Huntsman17, Simon Gayther3, Jason Konner5, Francesmary Modugno4, Kate Lawrenson3,
Ellen L. Goode7, and Elli Papaemmanuil5
ABSTRACT
◥
Purpose: To identify molecular subclasses of clear cell ovarian
carcinoma (CCOC) and assess their impact on clinical presentation
and outcomes.
Experimental Design: We profiled 421 primary CCOCs that
passed quality control using a targeted deep sequencing panel of 163
putative CCOC driver genes and whole transcriptome sequencing
of 211 of these tumors. Molecularly defined subgroups were iden-
tified and tested for association with clinical characteristics and
overall survival.
Results: We detected a putative somatic driver mutation in at
least one candidate gene in 95% (401/421) of CCOC tumors
including ARID1A (in 49% of tumors), PIK3CA (49%), TERT
(20%), and TP53 (16%). Clustering of cancer driver mutations and
RNA expression converged upon two distinct subclasses of CCOC.
The first was dominated by ARID1A-mutated tumors with
enriched expression of canonical CCOC genes and markers of
platinum resistance; the second was largely comprised of tumors
with TP53 mutations and enriched for the expression of genes
involved in extracellular matrix organization and mesenchymal
differentiation. Compared with the ARID1A-mutated group,
women with TP53-mutated tumors were more likely to have
advanced-stage disease, no antecedent history of endometriosis,
and poorer survival, driven by their advanced stage at presentation.
In women with ARID1A-mutated tumors, there was a trend toward
a lower rate of response to first-line platinum-based therapy.
Conclusions: Our study suggests that CCOC consists of two
distinct molecular subclasses with distinct clinical presentation and
outcomes, with potential relevance to both traditional and exper-
imental therapy responsiveness.
See related commentary by Lheureux, p. 4838
Introduction
Historically, tumor treatment approaches have been dictated by
tissue site, but large-scale molecular profiling efforts have shown that
remarkable heterogeneity exists in the landscape of cancer driver genes
and pathways within tumor types and even within histologic subtypes.
This has been well characterized for many common tumors through
multi-omic profiling (1) and characterization of the genetic determi-
nants of tumor behavior and outcome has led to the development of
personalized therapeutic approaches. Indeed, for some cancers, prog-
nosis and therapeutic strategies are based primarily on their presence
of genetic driver mutations identified in the tumor (2–7). For several
rare cancer types such as ovarian clear cell carcinoma (CCOC), no
strong associations between molecular profiles and clinical presenta-
tion or outcomes are known and broad-acting platinum-based che-
motherapy remains the standard of care.
When diagnosed at an advanced stage, CCOC has a worse out-
come than other invasive ovarian cancers including the more
common high-grade serous ovarian carcinoma (HGSOC; median
overall survival of 10 months; refs. 8, 9) presents at a younger age
(10), and is less responsive to platinum-based therapy (11). Relatively
small studies suggest that CCOC possesses several driver events that
1Washington University School of Medicine, St. Louis, Missouri. 2Philadelphia
College of Osteopathic Medicine, Philadelphia, Pennsylvania. 3Cedars-Sinai
Medical Center, Los Angeles, California. 4University of Pittsburgh Graduate
School of Public Health, Pittsburgh, Pennsylvania. 5Memorial Sloan-Kettering
Cancer Center, New York, New York. 6The University of Calgary, Calgary,
Alberta, Canada. 7Mayo Clinic, Rochester, Minnesota. 8David Geffen School of
Medicine, Department of Obstetrics and Gynecology, University of California
at Los Angeles, Los Angeles, California. 9Department of Gynaecological
Oncology, Westmead Hospital, Sydney, New South Wales, Australia.
10Centre for Cancer Research, The Westmead Institute for Medical Research,
Sydney, New South Wales, Australia. 11University of Cambridge, Cambridge,
United Kingdom. 12The University of Sydney, Sydney, New South Wales,
13Brigham and Women’s Hospital, Boston, Massachusetts.
Australia.
14University of Pennsylvania, Philadelphia, Pennsylvania.
15University of
Edinburgh, Edinburgh, United Kingdom. 16The Daffodil Centre, The Univer-
sity of Sydney, a joint venture with Cancer Council NSW, Sydney, New
South Wales, Australia. 17University of British Columbia, Vancouver, British
Columbia, Canada.
K.L. Bolton, D. Chen, R. Corona de la Fuente, Z. Fu, R. Murali, M. K€obel, Y. Tazi,
J.M. Cunningham, L.A. Moukarzel, S. Gayther, J. Konner, F. Modugno, K. Lawrenson,
E.L. Goode, and E. Papaemmanuil contributed equally to this article.
Corresponding Author: Kelly L. Bolton, Medicine, Washington University, 4444
Forest Park, St. Louis, MO 63108. E-mail: [email protected]
Clin Cancer Res 2022;28:4947–56
doi: 10.1158/1078-0432.CCR-21-3817
This open access article is distributed under the Creative Commons Attribution-
NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.
(cid:1)2022 The Authors; Published by the American Association for Cancer Research
AACRJournals.org | 4947
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
a
a
c
r
j
o
u
r
n
a
s
.
o
r
g
/
c
l
l
i
n
c
a
n
c
e
r
r
e
s
/
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
2
8
2
2
4
9
4
7
3
2
3
3
0
6
5
4
9
4
7
p
d
/
.
f
i
t
b
y
W
a
s
h
n
g
o
n
U
n
v
e
r
s
i
t
y
S
i
t
i
L
o
u
s
u
s
e
r
o
n
1
2
J
a
n
u
a
r
y
2
0
2
3
Bolton et al.
Translational Relevance
Clear cell ovarian cancer (CCOC) is the second most common
subtype of epithelial ovarian cancer and when diagnosed at an
advanced stage, has a poor prognosis. The relationship between
molecular profiles and clinical presentation or outcomes is still
unknown but could help guide the development of personalized
therapeutic approaches for CCOC. Here, we profiled 421 primary
CCOCs using deep targeted sequencing and whole transcriptome
sequencing on a subset of 211. Clustering of cancer driver muta-
tions and RNA expression converged upon two distinct subclasses
of CCOC. The first was dominated by ARID1A-mutated tumors
with enriched expression of canonical CCOC genes and markers of
platinum resistance; the second was largely comprised of tumors
with TP53 mutations and enriched for the expression of genes
involved in extracellular matrix organization and mesenchymal
differentiation. These two distinct molecular subclasses showed
distinct clinical presentation and outcomes, with potential rele-
vance to therapeutic responsiveness.
are distinct from HGSOC. CCOC is thought to arise from endome-
triotic lesions with recurrent somatic mutations in PIK3CA and
ARID1A, which are rare in HGSOC (12–15). In addition, the
existing data suggests that CCOCs are commonly TP53-wild-type
(whereas HGSOC ubiquitously harbors TP53 mutations) and exhi-
bits fewer structural rearrangements than HGSOC (13). However, it
is not known whether clinically meaningful molecular subtypes of
CCOC exist.
In the current study, we performed comprehensive targeted
sequencing and transcriptomic profiling of a large, multi-ethnic cohort
of 421 primary CCOCs to identify disease subclasses with distinct
biology and clinical behavior, which in turn may provide avenues for
personalized therapeutic approaches.
Materials and Methods
Study participants
Clinical data and therapy-na€(cid:2)ve fresh frozen tumor material were
utilized from women diagnosed with invasive CCOC and enrolled
into research studies from the following sites: Memorial Sloan Ketter-
ing Cancer Center Gynecology Tissue Bank (MSK), Mayo Clinic
(MAY), Addenbrooks Hospital (ADD), Cedars-Sinai Medical Center
(WCP; Los Angeles, CA), University of Pittsburgh (PIT; Pittsburgh,
PA), Gynaecological Oncology Biobank (GynBiobank) at Westmead
Hospital (WMH, Sydney, Australia), University of Edinburgh (SCOT;
Scotland), Canadian Ovarian Experimental Unified Resource
(COEUR; multiple sites), Brigham and Women’s Hospital (BWH;
Boston, MA), and University of Pennsylvania (UPA; Philadelphia,
PA). Participants provided written informed consent. The studies were
conducted in accordance with recognized ethical guidelines (e.g.,
Declaration of Helsinki, CIOMS, Belmont Report, U.S. Common
Rule), and approved by local institutional review boards. Extraction
of DNA/RNA was performed centrally at MSK (for cases from MSK,
WCP, PIT, BWH, and UPA) or locally (for cases from MAY, ADD,
WMH, and COEUR). For the cases which were extracted centrally at
MSK, slides from frozen tissue sections were reviewed by a pathologist
(R. Murali) and extraction of DNA/RNA was performed from tumor
sections, selected based on high content (>80%) of clear cell carcinoma.
In total, tumors from 447 women diagnosed with CCOC were
analyzed. Race and menstruation status (pre vs. postmenopausal) was
obtained through participant self-report. History of endometriosis was
also obtained through self-report except at MSK where endometriosis
was only available if mentioned on the pathology report. Tumor
characteristics and clinical outcomes were obtained through medical
record review.
Targeted DNA sequencing and analysis
We performed targeted sequencing of 163 putative CCOC driver
genes (Supplementary Table S1) in DNA samples from the 447 tumor
and blood-derived DNA from 16 unmatched controls using a custom
Nimblegen capture-based panel. Genes were selected based on a
combined analysis of 105 clear cell somatic sequencing studies includ-
ing: (i) whole genome sequencing of 31 CCOCs from Wang and
colleagues (13); (ii) whole-exome sequencing of eight cases from Jones
and colleagues (12); (iii) targeted sequencing of 26 CCOCs using a
panel of 465 known cancer drivers (MSK-IMPACT; ref. 16); and
targeted or whole exome sequencing of 40 CCOCs from project
GENIE (17). Included in our panel were 119 genes where somatic
mutations have been identified in two or more CCOCs; 41 established
cancer driver genes based on the COSMIC Cancer Gene Census (18)
mutated in one CCOC and three genes in the SWI/SNF complex
(SMARCB1, SMARCC1, SMARCC2) (14) that have been implicated in
CCOC biology (Supplementary Table S1; ref. 19). We also included on
the sequencing panel highly polymorphic single nucleotide variants
distributed every 3 MB throughout the genome to capture large copy
number deletions/amplifications.
Of 447 tumor samples, 421 (94%) passed quality control. As a
technical set of normal samples (panel of normals), we included DNA
extracted from the blood of 10 healthy, cancer-free individuals. Two
tumor samples failed due to low coverage, 12 due to sample contam-
ination and 12 due to duplication. The median sequencing coverage
per sample was 539x. Raw sequence data were aligned to the human
genome (NCBI build 37) using BWA (20). Variant calling for single
nucleotide variants was performed using Mutect2 (21), Strelka (22),
and CaVEMan (23) and for insertions/deletions using Pindel (24),
Mutect2 (21), and Strelka (22). We considered mutations to be true if
they: (i) passed at least two variant callers; (ii) were present at a variant
allele fraction of greater than 2%; (iii) were present in gNOMAD (25)
whole-exome sequencing data with a maximum population frequency
of less than 0.001; (iv) had a variant allele frequency (VAF) at least two
times greater than the median VAF in a panel of normal samples; and
(v) were present in none of the panel of normal samples at a VAF of 2%
or greater. We further excluded mutations in low complexity regions
[DUST (26) score >7]. Mutations in known cancer hotspots that met
all other requirements but failed due to low complexity or to only being
passed by one variant caller were retained for consideration. We
calculated a microsatellite instability score for each tumor using MSI
sensor (27)
We used Bayesian Dirichlet processes to establish classification rules
that partitioned tumors into subgroups, minimizing overlap between
categories. The Dirichlet process defines an infinite prior distribution
for the number and proportions of clusters in a mixture model,
fitted with the use of the Markov chain Monte Carlo method (28).
Our method was based on an implementation of the Dirichlet process
mixture model available at https://github.com/nicolaroberts/hdp
using a non-hierarchical Dirichlet process. We used 5,000 burnin
iterations and subsequently sampled 10,000 realizations at intervals of
20 iterations. From this collection of data, we computed the optimal
number of clusters, requiring that 90% of the samples were assigned
a cluster.
4948
Clin Cancer Res; 28(22) November 15, 2022
CLINICAL CANCER RESEARCH
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
a
a
c
r
j
o
u
r
n
a
s
.
o
r
g
/
c
l
l
i
n
c
a
n
c
e
r
r
e
s
/
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
2
8
2
2
4
9
4
7
3
2
3
3
0
6
5
4
9
4
7
p
d
/
.
f
i
t
b
y
W
a
s
h
n
g
o
n
U
n
v
e
r
s
i
t
y
S
i
t
i
L
o
u
s
u
s
e
r
o
n
1
2
J
a
n
u
a
r
y
2
0
2
3
Molecular Subclasses of Clear Cell Ovarian Carcinoma
Whole transcriptome sequencing and analysis
Data availability
RNA sequencing (RNA-Seq) libraries were prepared for 211 cases
from total RNA derived from the same tumor section using poly(A)
enrichment of the mRNA. One hundred bp paired-end libraries were
sequenced on Illumina’s HiSeq at a targeted depth of 40 million reads
per
sample. We performed alignment using STAR (version
STAR_2.5.1b; ref. 29) against the reference genome hg38 (GENCODE
v26). Reads were summarized using featureCounts (version 1.5.0-p1;
ref. 30). RNA clusters were defined using hierarchical clustering using
the top 500 most variable protein coding genes (clustering parameters:
method ¼ ward. D2, distance ¼ canberra). Differentially expressed
genes between RNA cluster 1 and RNA cluster 2 samples were obtained
using the R package DESeq2 (version 1.28.1; ref. 31) with collection site
and RNA cluster as part of the design formula. Pathway enrichment
analysis was performed using Metascape (version 3.5; ref. 4), looking
for enrichment of GO and KEGG terms, Hallmark, Reactome and
BioCarta Gene Sets, and Canonical Pathways. The top 500 most
overexpressed genes in RNA cluster 1 (log2 fold change <1 and FDR
<0.05) and the top 500 most overexpressed genes in RNA cluster 2 were
used as input for Metascape (32).
Outcome analyses
Survival data was available for 350 cases. Survival time was calcu-
lated from the date of diagnosis to last follow-up and allowed for left
truncation for cases who were consented following diagnosis. We right
censored at five years from diagnosis to reduce non-ovarian cancer
related deaths. Race, age at diagnosis (continuous and quadratic,
assigned as site median for three cases), tumor stage, extent of residual
disease, and study site were considered as covariates using a Cox
proportional hazards model. Proportionality of hazards was examined
using Schoenfeld residuals. In addition, contingency analysis was done
on tumor mutational status and tumor cluster with primary treatment
response (complete response or partial response compared to stable or
progressive disease) stratified by tumor stage and vital status up to five
years using a c2 test.
The somatic variant calls and normalized RNA-seq intensity data,
code, and deidentified clinical data are available here: https://github.
com/kbolton-lab/Bolton_OCCC. This will enable all the figures and
tables to be re-generated and also provide data for others for future
analyses. We will also make the BAMs/FASTQs available to research-
ers through contacting Kelly Bolton ([email protected]).
Results
Clinical characteristics
Key characteristics, other than race, of the 421 participants included
in the study did not vary between study sites (Table 1). Compared with
clinical characteristics reported in the literature for women with
HGSOC (10, 33), women with CCOC in this cohort were more likely
to be of Asian ancestry (12% of individuals with non-missing race),
have a history of endometriosis (13%), and present with early-stage
disease (69%).
Targeted DNA sequencing of candidate CCOC driver genes
In 163 candidate CCOC driver genes we identified 6,361 mutations.
Of these, 1,488 mutations were classified as potentially pathogenic
based upon annotation in OncoKB (34), frequency in COSMIC,
frequency in previously published CCOC sequencing data (12, 13, 16),
predicted pathogenicity based on PolyPhen (35) and SIFT (36), and
prior evidence in the literature (Supplementary Table S2). At least one
putative driver mutation was identified in 401 of 421 tumors (95%)
(mean number of mutations 3, range 1–25; Fig. 1A and C). The most
commonly mutated genes were ARID1A (49%, N ¼ 205), PIK3CA
(45%, N ¼ 188), and the TERT promoter (20%, N ¼ 84). The most
frequently recurrent mutations were clonally dominant with a VAF
>35% (e.g., ARID1A and TP53) suggesting that they represented early
events while others (e.g., CREBBP) were more often subclonal, possibly
representing secondary events (Fig. 1B). We detected a higher pro-
portion (16%, N ¼ 71) of tumors with TP53 mutations than has been
Table 1. Clinical characteristics of CCOC cases sequenced using targeted panel.
ADD
(N ¼ 28)
BWH
(N ¼ 9)
COEUR
(N ¼ 181)
MAY
(N ¼ 38)
MSK
(N ¼ 60)
PIT
(N ¼ 24)
SCOT
(N ¼ 22)
UPA
(N ¼ 7)
WCP
(N ¼ 28)
WMH
(N ¼ 24)
Overall
(N ¼ 421)
Age (y)
<40
40–50
50–60
60–70
70–80
≥80
Missing
Race
0 (0%)
1 (3.6%)
8 (28.6%)
13 (46.4)
6 (21.4%)
0 (0%)
0 (0%)
0 (0%)
0 (0%)
37 (20.4%)
0 (0%)
2 (22.2%)
81 (44.8%)
7 (77.8%) 48 (26.5%)
0 (0%)
0 (0%)
0 (0%)
10 (5.5%)
1 (0.6%)
4 (2.2%)
0 (0%)
3 (7.9%)
16 (42.1)
9 (23.7%)
7 (18.4%)
2 (5.3%)
1 (2.6%)
0 (0%)
0 (0%)
6 (10.0%) 4 (16.7%)
9 (37.5%)
28 (46.7)
19 (31.7)
5 (20.8%)
6 (10.0%) 4 (16.7%)
2 (8.3%)
0 (0%)
0 (0%)
1 (1.7%)
0 (0%)
4 (18.2%)
7 (31.8%)
9 (40.9%)
1 (4.5%)
1 (4.5%)
0 (0%)
0 (0%)
0 (0%)
7 (25.0%)
2 (28.6%)
2 (28.6%)
14 (50.0)
2 (28.6%) 4 (14.3%)
1 (14.3%)
0 (0%)
0 (0%)
1 (3.6%)
0 (0%)
2 (7.1%)
0 (0%)
0 (0%)
5 (20.8%) 69 (16.4%)
177 (42.0)
10 (41.7)
121 (28.7)
5 (20.8%)
39 (9.3%)
3 (12.5%)
6 (1.4%)
0 (0%)
9 (2.1%)
1 (4.2%)
White
Asian
Black
Other
Unknown
Endometriosis
16 (57.1)
2 (7.1%)
0 (0%)
0 (0%)
10 (35.7)
9 (100%)
0 (0%)
0 (0%)
0 (0%)
0 (0%)
0 (0%)
0 (0%)
0 (0%)
0 (0%)
181 (100%)
38 (100%) 44 (73.3)
0 (0%)
0 (0%)
0 (0%)
0 (0%)
13 (21.7)
1 (1.7%)
2 (3.3%)
0 (0%)
23 (95.8)
0 (0%)
1 (4.2%)
0 (0%)
0 (0%)
6 (85.7%)
0 (0%)
0 (0%)
0 (0%)
1 (14.3%)
0 (0%)
0 (0%)
0 (0%)
22 (100%) 0 (0%)
23 (82.1)
4 (14.3%)
1 (3.6%)
0 (0%)
0 (0%)
10 (41.7)
4 (16.7%)
0 (0%)
0 (0%)
10 (41.7)
169 (40.1)
23 (5.5%)
4 (1.0%)
2 (0.5%)
223 (53.0)
Yes
No
Unknown
FIGO stage
0 (0%)
0 (0%)
28 (100%) 0 (0%)
0 (0%)
9 (100%)
13 (7.2%)
168 (92.8)
0 (0%)
10 (26.3)
26 (68.4)
2 (5.3%)
6 (10.0%) 0 (0%)
0 (0%)
49 (81.7)
24 (100%) 0 (0%)
5 (8.3%)
2 (9.1%)
20 (90.9)
2 (28.6%)
5 (71.4%)
0 (0%)
7 (25.0%)
0 (0%)
21 (75.0)
3 (12.5%)
0 (0%)
21 (87.5)
43 (10.2%)
277 (65.8)
101 (24.0)
I/II
III/IV
Missing
17 (60.7)
5 (17.9%)
6 (21.4%)
7 (77.8%)
128 (70.7)
2 (22.2%) 46 (25.4%)
0 (0%)
7 (3.9%)
25 (65.8)
12 (31.6)
1 (2.6%)
42 (70.0)
17 (28.3)
1 (1.7%)
16 (66.7)
8 (33.3%)
0 (0%)
14 (63.6)
7 (31.8%)
1 (4.5%)
2 (28.6%)
5 (71.4%)
0 (0%)
15 (53.6)
13 (46.4)
0 (0%)
16 (66.7)
7 (29.2%)
1 (4.2%)
282 (67.0)
122 (29.0)
17 (4.0%)
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
a
a
c
r
j
o
u
r
n
a
s
.
o
r
g
/
c
l
l
i
n
c
a
n
c
e
r
r
e
s
/
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
2
8
2
2
4
9
4
7
3
2
3
3
0
6
5
4
9
4
7
p
d
.
/
f
i
t
b
y
W
a
s
h
n
g
o
n
U
n
v
e
r
s
i
t
y
S
i
t
i
L
o
u
s
u
s
e
r
o
n
1
2
J
a
n
u
a
r
y
2
0
2
3
AACRJournals.org
Clin Cancer Res; 28(22) November 15, 2022
4949
Bolton et al.
A
C
n
o
i
t
a
t
u
m
h
t
i
l
w
s
e
p
m
a
s
f
o
n
o
i
t
r
o
p
o
r
P
s
e
s
a
c
f
o
n
o
i
t
r
o
p
o
r
P
0.5
0.4
0.3
0.2
0.1
0.0
TERT
ARID1A
PIK3CA
TP53
KRAS
ATM
KMT2D
BRCA1
SPOP
KMT2C
ARID1B
CHD4
CTNNB1
PTEN
MSH6
SMARCA4
PIK3R1
PPP2R1A
CREBBP
ERBB3
NF1
BRCA2
NFE2L2
PBRM1
FBXW7
TET2
PMS2
MAP3K1
Gene
0.25
98
96
0.20
81
61
0.15
0.10
0.05
0.00
28
13
7
3
1
2
2
2
2
1
1
1
1
1
2
3
4
5
6
7
9
Number of mutated genes
11
10
8
12
13
14
18
19
25
B
1.00
y
c
n
e
u
q
e
r
f
e
e
l
l
l
a
t
n
a
i
r
a
V
0.75
0.50
0.25
0.00
ARID1A
D
TP53
PIK3CA
ARID1B
CHD4
SPOP
KRAS
PPP2R1A
ATM
PTEN
PIK3R1
Gene
TERT
SMARCA4
KMT2C
CREBBP
398
46
5
53 68
Variant effect
5_prime_UTR_variant
148
36
130
Nucleotide substitution
287
643
Frameshift indel
Inframe indel
Missense
Nonsense
Promoter
Splice or other
99
38
563
C>A
C>G
C>T
T>A
T>C
T>G
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
a
a
c
r
j
o
u
r
n
a
s
.
o
r
g
/
c
l
l
i
n
c
a
n
c
e
r
r
e
s
/
a
r
t
i
c
e
-
p
d
l
f
/
Figure 1.
Mutational landscape of 401 clear cell ovarian carcinomas with a detectable mutation. A, Proportion of patients with mutations in commonly mutated genes.
B, Mutation variant allele frequency (VAF) by genes mutated in at least 10% of individuals. C, Number of mutated genes per individual. D, Variant effect and nucleotide
substitution change for single nucleotide variants.
described by some (9%–15%; refs. 13, 37) but not all NGS studies (18%;
ref. 38). This raises the possibility that some of the CCOCs in this
cohort were misdiagnosed high-grade serous or endometrioid ovarian
cancers. We explored this possibility in detail. First, we noted that 10 of
71 TP53 mutations (14%) were deeply subclonal (VAF<10%); previous
studies may not have detected these mutations as they used lower-
depth sequencing (Fig. 1B). Second, we performed additional path-
ologic review to verify clear cell histology for a subset of the cases
where formalin-fixed paraffin-embedded (FFPE) tissue sections were
available. This included 14 (20%) of the TP53-mutated cases and 4
(15%) of the BRCA1/2-mutated cases where FFPE tissue sections were
available. On the basis of morphology combined with and immuno-
histochemical staining of Napsin A, p53, and WT1 (markers of
HGSOC and not CCOC; ref. 39), it was determined that four of 14
TP53-mutant cases (28%; three endometrioid carcinomas and one
HGSOC) were misclassified as CCOC. None of the BRCA1/2-mutated
cases were misclassified. Thus, by extrapolation we estimate that
approximately 19 of our 71 TP53-mutant tumors in this cohort were
misclassified.
A subset of tumors (N ¼ 20) bore mutations in SMARCA4, a gene
that is the sole driver mutation in ovarian small cell carcinoma
refs. 40–42). However, unlike
hypercalcemic type (OSCCHT;
OSCCHT, in our CCOC cases we observed SMARCA4 to be most
commonly comutated with either ARID1A (50%) or PIK3CA (35%).
Similar to our analysis of TP53 mutated cases, we performed central
pathology review of a subset (N ¼ 8) of the SMARCA4 mutated
cases. All of these cases showed typical CCOC morphology and were
positive for clear cell markers such as PAX8 (8/8 diffuse), and Napsin
A (5/8 diffuse, 2/8 focal), or HNF1B (5/5 diffuse). We conclude that
there was no evidence for these cases being misclassified OSCCHT.
Whether SMARCA4 has a similar driver capacity in CCOC compared
with OSCCHT requires further study.
Most cases (75%) had at least one large-scale copy number event
with the most frequently recurrent events reflecting common cancer-
driver aneuploidies including 8q amplification (Supplementary Fig. S1;
ref. 19). Cases with TP53 mutations had more whole chromosome
or arm-level aneuploidies (mean ¼ 12) compared with wild-type
tumors (mean ¼ 8; Supplementary Fig. S2). TP53-mutant/ARID1A-
mutant tumors showed less genomic instability (mean number of
aneuploidies ¼ 7) compared with TP53-mutant/ARID1A-wild type
tumors (mean number of aneuploidies ¼ 13). We detected recurrent
fusions in TGM7 (N ¼ 5) as previously shown by Earp and collea-
gues (43). In addition, recurrent fusions involving BCAR4 (N ¼ 6),
ITCH (N ¼ 6), and DCAF12 (N ¼ 5) were observed. These are known
4950
Clin Cancer Res; 28(22) November 15, 2022
CLINICAL CANCER RESEARCH
/
/
/
2
8
2
2
4
9
4
7
3
2
3
3
0
6
5
4
9
4
7
p
d
/
.
f
i
t
b
y
W
a
s
h
n
g
o
n
U
n
v
e
r
s
i
t
y
S
i
t
i
L
o
u
s
u
s
e
r
o
n
1
2
J
a
n
u
a
r
y
2
0
2
3
Molecular Subclasses of Clear Cell Ovarian Carcinoma
cancer fusion partners but have not been reported in CCOC before
(Supplementary Fig. S3).
We evaluated mutation status with respect
to clinical and
epidemiological factors including age, race, tumor, and history
of endometriosis. Compared with ARID1A-mutated tumors,
patients with KRAS mutations were older at presentation (median
age 53 vs. 67, P ¼ 0.03; Fig. 2A). Individuals with a history of
endometriosis were more likely to have ARID1A-mutated tumors
(72% and 47% of patients with and without endometriosis respec-
–4; Fig. 2B). Advanced stage tumors were more
tively, P ¼ 2 (cid:2) 10
likely to harbor TP53 mutations than early-stage tumors (27% vs.
–4; Fig. 2C). Among TP53 mutant
11% respectively, P ¼ 2 (cid:2) 10
tumors, a similar proportion (50% and 51%, respectively) were
advanced stage with or without co-occurring ARID1A mutations.
There was a trend toward a higher frequency of ARID1A-mutated
tumors in women of east Asian descent but this was not significant
(Fig. 2D).
We next examined the relationship between mutational burden,
cancer driver genes, and patterns of genetic cooccurrence. Several
genes harbored recurrent mutations within the same tumor (Sup-
plementary Fig. S4). This seen for both tumor suppressor genes
(e.g., ARID1A) and specific oncogenes including PIK3R1 and
PIK3CA. Among tumors with multiple PIK3CA mutations, variants
were more likely to occur in nonhotspot locations within the gene
(Supplementary Fig. S5; ref. 44). MSIsensor score was higher
among individuals more than 10 driver mutations (N ¼ 12, 3%)
and among those with MSH2 and MSH6 mutations (Supplementary
Fig. S6). We observed a statistically significant co-occurrence
between mutations in ARID1A, PIK3CA, TP53 and BRCA1/BRCA2
Mutual exclusivity between somatic mutations of ARID1A, TP53,
PIK3CA and PIK3R1 (Supplementary Fig. S7) suggests that these
may represent distinct pathways to oncogenesis. The exclusivity
between TP53 and ARID1A mutation was stronger in the setting of
multiple ARID1A mutations (OR ¼ 0.21; 95% CI, 0.07–0.54; P ¼ 2
–4) compared with a single ARID1A mutations (OR ¼ 0.68;
(cid:2) 10
95% CI, 0.32–1.34; P ¼ 0.28). “We observed 54 mutations in genes
known to be relevant to high penetrance genetic predisposition to
ovarian cancer including PMS2, MSH6, MSH2, BRCA1, and
BRCA2. Overall, 52% of these mutations were present at a VAF
in the tumor of ≥35%. In the absence of matched normal tissue
sequencing, we were not able to distinguish these from germline
variants. Thus, it is possible that up to 26 cases (6% of the cohort)
harbored a germline pathogenic variant in a known cancer sus-
ceptibility gene.”
Because we observed clear patterns of exclusivity and cooccurrence
between gene drivers, we used unsupervised clustering approaches to
define nonoverlapping subgroups of CCOC based on their mutational
spectrum. We defined seven subgroups (Supplementary Fig. S8) and
compared the frequency of mutations between subgroups. Four
clusters were characterized by having an ARID1A mutation; the first
cluster (cluster A) was characterized by a single ARID1A mutation
in combination with another disease defining mutation (e.g., PIK3CA,
TERT, TP53, KRAS, PTEN, PPP2R1A, PIK3R1, CREBBP, or SPOP; N ¼
86); the second (cluster B) with a single ARID1A mutation alone or in
combination with non-disease defining mutation (N ¼ 19); the third
(cluster C) with multiple ARID1A mutations combined with a PIK3CA
mutation (N ¼ 81); and a forth (cluster D) with multiple ARID1A
mutations and PIK3CA wild-type (N ¼ 25). Two clusters were
ARID1A wildtype: Cluster E was defined by a TP53 mutation
(N ¼ 50); and cluster F by other non-TP53 disease-defining muta-
tions (N ¼ 104). A final cluster (cluster G) was characterized by
mutations in SMARCA4 (N ¼ 13); a mutation typically observed in
small cell ovarian carcinoma (23). The remaining tumors were
undefined (N ¼ 57).
A
0.03
i
i
s
s
o
n
g
a
d
t
a
e
g
A
90
60
30
ARID1A
ARID1B
C
ARID1A
PIK3CA
T E RT
PPP2R1A
KRAS
TP53 **
SPOP
CREBBP
PIK3R1
PTEN
KMT2C
A T M
ARID1B
SMARCA4
CHD4
CTNNB1
ATM
CREBBP
CHD4
CTNNB1
KMT2C
KRAS
PIK3CA
PPP2R1A
PIK3R1
SMARCA4
PTEN
SPOP
TERT
TP53
FIGO Tumor stage
I/II
III/IV
B
ARID1A **
PIK3CA
T E RT
TP53
KRAS
PPP2R1A
CREBBP
SPOP
PIK3R1
KMT2C
PTEN
A T M
ARID1B
SMARCA4
CHD4
History of endometriosis
No
Yes
–50
–25
0
25
50
75
Percentage with mutation
D
ARID1A
PIK3CA
T E RT
TP53
KRAS
SPOP
PPP2R1A
PTEN
PIK3R1
Asian ancestry
No
Yes
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
a
a
c
r
j
o
u
r
n
a
s
.
o
r
g
/
c
l
l
i
n
c
a
n
c
e
r
r
e
s
/
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
2
8
2
2
4
9
4
7
3
2
3
3
0
6
5
4
9
4
7
p
d
.
/
f
i
t
b
y
W
a
s
h
n
g
o
n
U
n
v
e
r
s
i
t
y
S
i
t
i
L
o
u
s
u
s
e
r
o
n
1
2
J
a
n
u
a
r
y
2
0
2
3
−50
−25
0
25
50
–50
–25
0
25
50
75
Percentage with mutation
Percentage with mutation
Figure 2.
Frequency of somatic mutations by clinical characteristics including age at diagnosis (A), endometriosis (B), stage (C), and race (D). Genes that were mutated
in at least 20 individuals with nonmissing values for the clinical characteristic were included. Shown are q-values (FDR corrected P values) based on Fisher exact test.
(cid:3), q < 0.05; (cid:3) (cid:3), q < 0.01.
AACRJournals.org
Clin Cancer Res; 28(22) November 15, 2022
4951
Bolton et al.
Similar to the patterns we observed when studying the association
between individual mutations and clinical features, the TP53-mutated,
ARID1A wild-type cluster showed an enrichment of advanced stage
disease while tumors belonging to the ARID1A-mutant clusters were
more likely in individuals of Asian ancestry and those with a history of
endometriosis (Supplementary Fig. S9). Individuals in cluster G
(SMARCA4-mutant tumors) had a nonsignificant trend towards a
younger age at diagnosis (P ¼ 0.32).
Transcriptomic profiling of CCOC
Transcriptomic profiles were generated for 212 CCOC tumors in
which targeted sequencing was also performed. Using unsupervised
clustering informed by expression of the 500 most variable genes,
we identified two main RNA clusters (Supplementary Fig. S10):
Expression cluster 1 showed higher expression of genes previously
reported as highly expressed in CCOC including ANXA4 and GPX3,
both of which are linked to platinum resistance (45, 46). Among
the most highly expressed genes in cluster 1 compared with 2 also
included GPX3 (47), which is known to be overexpressed in endo-
metriosis compared to normal endometrial tissue, and EEF1A2,
known to be overexpressed in CCOC associated endometriosis but
not benign endometriosis (48). Genes that characterized this cluster
were enriched in metabolic pathways including flavonoid glucuroni-
–13).
dation (P ¼ 10
Expression cluster 2 showed enriched expression of genes involved in
–22) and mesenchy-
extracellular matrix (ECM) organization (P ¼ 10
mal differentiation, including genes such as ADGR2 and PDCH19
(Supplementary Fig. S10 and Fig. 3B). Compared to cluster 1, expres-
sion cluster 2 also showed higher expression of WT1 and lower
expression of CCOC marker HNF1B, which are features classically
associated with high-grade serous ovarian cancer (Fig. 3B; ref. 9).
Expression cluster 2 was enriched with TP53-mutant tumors (55% of
cases in cluster 2 compared with 10% in cluster 1). When comparing
RNA expression and mutation clusters, cluster 2 was largely comprised
of tumors belonging to mutation cluster E, that is, TP53-mutant
ARID1A-wild type tumors (45% of cluster 2) and the undefined
mutation cluster (33% of cluster 2; Fig. 3A).
–15) and monocarboxylic acid metabolism (P ¼ 10
Clinical outcomes
There was no statistically significant association between overall
survival and CCOC mutations when examined on a per-gene level in
Cox proportional hazards models stratified by study site (Supplemen-
tary Table S3). We observed a nonsignificant trend toward improved
survival for patients with ARID1A (HR ¼ 0.82; 95% CI, 0.58–1.15; P ¼
0.24) and PTEN (HR ¼ 0.52; 95% CI, 0.24–1.12; P ¼ 0.10) mutant
tumors. Because of the similarity of the ARID1A-mutant clusters in
regards to clinical presentation and outcome, we combined these
clusters for the purpose of survival analysis. Women with TP53-
mutant, ARID1A-wild type tumors had worse overall survival com-
pared to those with ARID1A-mutant tumors (HR ¼ 1.72; 95% CI,
1.06–2.81; P ¼ 0.03; Fig. 4A). Similarly, RNA-seq cluster 2 showed
an increased risk of death compared with RNA-seq cluster 1
(Fig. 4B, Tumor Cluster 2 vs. Tumor Cluster 1 HR 2.8; 95% CI,
–4). Covariate adjustment for age, race, stage, and
1.66–4.84; P ¼ 1 (cid:2) 10
residual disease attenuated the estimated mutation and cluster-
associated risk (Supplementary Table S4). To explore how these
subgroups might influence therapy outcome, we studied the relation-
ship between mutation status and response to first line therapy with
platinum/taxane combination therapy. We limited this to women with
advanced stage disease who successfully underwent debulking surgery
followed by combination platinum/taxol therapy (N ¼ 36). Women
with ARID1A wild-type, TP53-mutant tumors were more likely to have
a complete response 75% (N ¼ 11) compared to ARID1A-mutant
tumors (55%), although this was not statistically significant (P ¼ 0.33)
in this small sample size.
Discussion
Our results have several clinical implications. First, the results of
both genomic and transcriptomic cluster associations with clinical
presentation and outcome converged, suggesting two main subgroups
of CCOC: The first subtype included ARID1A-mutant tumors (par-
ticularly double-mutant tumors) and other common CCOC mutations
(e.g., PIK3CA, TERT, etc.) that showed enriched expression of met-
abolic pathways, presented with early-stage disease and were more
likely to have a history of endometriosis. We denote this group as
“classic-CCOC”, which represented 83% of our cohort. The second
CCOC subtype was dominated by TP53-mutant tumors that showed
enriched expression of genes involved in extracellular matrix organi-
zation, mesenchymal differentiation and immune-related pathways.
These cases presented with advanced disease and had worse survival.
Interestingly, TP53 mutations either in the presence or absence of
cooccurring ARID1A mutations were associated with a higher degree
of genomic instability and aggressive, advanced stage tumors. The
worse survival for tumors in this “HGSOC-like” subgroup was largely
explained by advanced stage and higher burdens of residual disease.
Within both the “classic-CCOC” and “HGSOC-like” subgroups we
noted a subset of individuals had tumor with mutations in genes
known to be both somatic drivers of ovarian cancer and germline
susceptibility genes including PMS2, MSH6, MSH2, BRCA1, and
BRCA2. Due to the absence of matched normal samples, we were
unable to fully distinguish whether these represented somatic or
germline events and is a limitation of our study. Future studies
estimating the frequency of CCOC cases that arise in women with
strong hereditary predisposition and who may be considered for risk
reducing bilateral salpingo-oophorectomy should be prioritized (49).
There is increasing recognition that other histologic types of ovarian
including HGSOC and endometrioid carcinoma, can
carcinoma,
contain areas with clear cell change complicating the histologic
diagnosis (50). While a subset of cases in the “HGSOC-like” cluster
are misclassified HGSOC, and is a weakness of our study, it is unlikely
that this alone explains our findings. Firstly, all of our cases were
morphologically diagnosed by expert gynecological pathologists and at
some centers, this morphologic review was supplemented by immu-
nohistochemistry for histotype-specific markers. Secondly, in a subset
of TP53-mutant cases, we reconfirmed the diagnosis of CCOC using a
combination of morphologic and immunohistochemical features.
Thus, our results suggest that a subset of bona fide CCOCs with
HGSOC-like features exist. Our results also emphasize that expert
histologic review of CCOC cases, particularly those who present with
TP53-mutant, ARID1A-wild type tumors, is warranted given similar-
ities to the biology and behavior of HGSOC.
Gene expression profiles of the “classic-CCOC” and “HGSOC-like”
CCOC subtypes we observed are similar to those reported by Tan and
colleagues (51) which also reported two clusters, the first enriched for
genes in metabolic pathways and the second, a less common mesen-
chymal-like subgroup associated with late-stage disease. However,
unlikely Tan and colleagues, we observed differences in the frequency
of TP53-mutated tumors across clusters. The source of this discrep-
ancy is unclear and may include differences in sequencing technology
(Tan and colleagues performed targeted sequencing using Ion Tor-
rent) and patient characteristics (Tan and colleagues, included only
4952
Clin Cancer Res; 28(22) November 15, 2022
CLINICAL CANCER RESEARCH
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
a
a
c
r
j
o
u
r
n
a
s
.
o
r
g
/
c
l
l
i
n
c
a
n
c
e
r
r
e
s
/
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
2
8
2
2
4
9
4
7
3
2
3
3
0
6
5
4
9
4
7
p
d
/
.
f
i
t
b
y
W
a
s
h
n
g
o
n
U
n
v
e
r
s
i
t
y
S
i
t
i
L
o
u
s
u
s
e
r
o
n
1
2
J
a
n
u
a
r
y
2
0
2
3
A
RNA cluster
DNA cluster
Molecular Subclasses of Clear Cell Ovarian Carcinoma
Cluster C (Multi ARID1Am/PIK3CAm)
Cluster D (Multi ARID1Am/PIK3CAwt)
Cluster A (Single ARID1Am/Other DD mutant)
Cluster B (Single ARID1Amt)
Cluster F (ARID1Awt/Other DD mutant)
Cluster G (SMARCA4m)
Cluster E (ARID1Awt/TP53m)
Undefined
Cluster 1
Cluster 2
B
RNA cluster
DNA cluster
Collection site
WT1
KLK5
LGR6
DAPL1
CYP4B1
PCDH19
ADGRG2
CYP4X1
LRRC19
OPN5
CYP2C19
MOGAT1
PRRT1B
HAVCR1
PYY
MIOX
OLIG3
TGM5
TGM7
DLX6
DLX5
SLC15A1
HGD
GGT1
SGK2
SLC6A12
PGBD5
NAPSA
XDH
DNER
GPX3
RIMKLB
ANXA4
AOC1
EEF1A2
TMEM101
ATP11A
AP2A2
PROM1
RXFP1
C2CD4A
GDA
KIF12
HNF1B
GLRX
IGSF3
SLC3A1
DDX52
LRATD1
CEP44
Normalized
gene expression
20
15
10
5
RNA cluster
2
1
DNA cluster
C (Multi ARID1Am/PIK3CAm)
D (Multi ARID1Am/PIK3CAwt)
A (Single ARID1Am/Other DD mutant)
B (Single ARID1Amt)
F (ARID1Awt/Other DD mutant)
G (SMARCA4m)
E (ARID1Awt/TP53m)
Undefined
Collection site
BWH
COEUR
MSK
PIT
SCOT
UPA
WCP
WMH
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
a
a
c
r
j
o
u
r
n
a
s
.
o
r
g
/
c
l
l
i
n
c
a
n
c
e
r
r
e
s
/
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
2
8
2
2
4
9
4
7
3
2
3
3
0
6
5
4
9
4
7
p
d
/
.
f
i
t
b
y
W
a
s
h
n
g
o
n
U
n
v
e
r
s
i
t
y
S
i
t
i
L
o
u
s
u
s
e
r
o
n
1
2
J
a
n
u
a
r
y
2
0
2
3
Figure 3.
The transcriptome of clear cell ovarian cancer samples. A, Sankey plot showing the correspondence of the sample annotations RNA clusters and DNA clusters.
B, Heatmap showing the normalized gene expression of the top 50 most differentially expressed genes between RNA cluster 1 and RNA cluster 2.
women of Asian ancestry which trend towards lower frequencies of
TP53-mutated tumors in our analysis and which are known to have
lower frequencies of endometrial ovarian cancer). The overlap between
genes highly expressed in our “classic-CCOC” subgroup and those
enriched in endometriosis provide further support for the likely
transition from endometriosis to carcinoma in CCOC.
The greatest translational impact from these molecular CCOC
therapeutic
subtypes is expected to lie in the development of
AACRJournals.org
Clin Cancer Res; 28(22) November 15, 2022
4953
+
5
65
8
5
Bolton et al.
A
y
t
i
l
i
b
a
b
o
r
p
l
i
a
v
v
r
u
S
1.00
0.75
0.50
0.25
0.00
Strata +
ARID1Am
+
ARID1Awt/TP53m
++ ++++ +++++++
+
+
+ + ++
+ ++ +++++++++++++ +++++ + +++ + ++ + ++++++++++++++
+
+ +
+
0
1
2
3
4
Years
Number at risk
+
5
++ +++
+
B
y
t
i
l
i
b
a
b
o
r
p
l
i
a
v
v
r
u
S
1.00
0.75
0.50
0.25
0.00
Strata +
RNA Cluster 1
+
RNA Cluster 2
+++++++ ++ ++++++ ++++++++++++++++ +++++ +++
+
++
+ + ++++ +++++++
0
1
2
3
4
Years
Number at risk
Strata
ARID1Am
ARID1Awt/TP53m
Strata
RNA Cluster 1
RNA Cluster 2
ARID1Am
ARID1Awt/TP53m
178
46
0
168
37
1
141
28
2
116
24
3
Years
103
17
4
81
15
5
RNA Cluster 1
RNA Cluster 2
141
33
0
137
24
1
112
19
2
90
16
3
78
10
4
Years
Figure 4.
Association between CCOC molecular subgroups and all-cause mortality. Shown are the Kaplan–Meier plots for the survival probability over 5 years following CCOC
diagnosis stratified by mutational clusters defined by ARID1A/TP53 mutation status (A) and RNA-seq expression clusters (B).
approaches tailored to the vulnerabilities of each group. Interestingly,
despite being aggressive on presentation, a trend was seen towards the
“HGSOC-like” CCOC subgroup having higher response rates to first
line platinum-based chemotherapy. Future studies are warranted to
further explore whether genomic subtypes of CCOC predict response
to platinum-based and other therapies as treatment data were limited
here. The “classic-CCOC” subgroup dominated by mutations in the
SWI/SNF pathway and markers linked to chemo-resistance may be of
particular relevance to target for investigational first-line therapies.
Recent data suggests that the SWI/SNF pathway plays a novel role in
the regulation of antitumor immunity, and that SWI/SNF deficiency
can be therapeutically targeted by immune checkpoint blockade (19).
Several studies are currently evaluating the role of immune check point
inhibitors in CCOC including NCT03405454, NCT03425565. While a
limitation of our study was that we were unable to assess MMR
functional status, we did note a rare subset of tumors (3%) with
higher mutational burden (>10 drivers) and MSIsensor score. The
extent to which the subset of CCOCs with higher total mutation
and with MMR deficiency show improved responsiveness to
immune checkpoint blockade in ongoing clinical trials will be an
important avenue of investigation. Additional targeted therapeutic
strategies have been explored in preclinical settings including
epigenetic synthetic lethality, some of which are entering into
clinical trials. The PI3K inhibitor, alpelisib, is now FDA approved
for HR-positive breast cancer and ongoing trials in additional
PIK3CA-mutated cancers including CCOC are underway. Double
PIK3CA mutations appear to hyperactivate PI3K signaling and
enhance tumor growth and may confer increased responsiveness
to PI3K inhibitors than those with a single mutation (52). Thus, for
CCOC cases harboring multiple PIK3CA mutations, PI3K inhibi-
tors either alone or in combination with other agents may represent
a promising approach.
The strengths of this study include the large sample size, use of
multiple study sites, inclusion of women of European and non-
European ancestry, and integration of genetic and transcriptomic
markers of disease behavior and outcome. While this is the most
extensive genomic study of CCOC to date, greater sample size with
additional follow-up data will allow improved assessment and vali-
dation of these clinically relevant subtypes. Although future analyses
would benefit from larger patient collections, our current results
suggest that genomic classification may inform the future development
of targeted therapeutics in CCOC.
Authors’ Disclosures
C.J. Kennedy reports grants from National Health and Medical Research
Council and Cancer Institute New South Wales during the conduct of the study.
Y.-E. Chiew reports grants from National Health and Medical Research Council
of Australia and The Cancer Institute New South Wales during the conduct of
the study. P. Pharoah reports grants from Cancer Research UK during the conduct
of the study. R. Drapkin reports personal fees from Repare Therapeutics and
Cedilla Therapeutics and other support from VOC Health outside the submitted
work. C. Gourley reports grants and personal fees from AstraZeneca, MSD,
GSK, and Nucana; personal fees from Clovis, Foundation One, Chugai, Cor2Ed,
and Takeda; and grants from Novartis, Aprea, BerGenBio, and Medannexin
outside the submitted work. A. DeFazio reports grants from National Health and
Medical Research Council of Australia and The Cancer Institute NSW during
the conduct of the study, as well as grants and other support from AstraZeneca
outside the submitted work. B. Karlan reports grants from American Cancer
Society during the conduct of the study as well as relationships with AstraZeneca
(investigational therapeutic), Merck (investigational therapeutic), and Amgen
(investigational therapeutic). J.D. Brenton reports grants from Cancer Research
UK during the conduct of the study as well as personal fees from GSK and
AstraZeneca and other support from Tailor Bio outside the submitted work.
B. Weigelt reports personal fees from Repare Therapeutics outside the submitted
work. D. Huntsman reports being Founder and CMO for Canexia Health.
J. Konner reports personal fees from AstraZeneca, Clovis, and Tesaro outside the
submitted work. F. Modugno reports grants from University of Pittsburgh during the
conduct of the study. E. Papaemmanuil reports other support from Isabl Inc outside
the submitted work. No disclosures were reported by the other authors.
Authors’ Contributions
K.L. Bolton: Conceptualization, resources, data curation, formal analysis, funding
acquisition, writing–original draft, writing–review and editing. D. Chen: Resources,
data curation, investigation, writing–original draft, project administration, writing–
review and editing. R. Corona de la Fuente: Formal analysis, writing–original draft,
writing–review and editing. Z. Fu: Data curation, writing–original draft, writing–
review and editing. R. Murali: Data curation, validation, investigation, writing–
4954
Clin Cancer Res; 28(22) November 15, 2022
CLINICAL CANCER RESEARCH
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
a
a
c
r
j
o
u
r
n
a
s
.
o
r
g
/
c
l
l
i
n
c
a
n
c
e
r
r
e
s
/
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
2
8
2
2
4
9
4
7
3
2
3
3
0
6
5
4
9
4
7
p
d
/
.
f
i
t
b
y
W
a
s
h
n
g
o
n
U
n
v
e
r
s
i
t
y
S
i
t
i
L
o
u
s
u
s
e
r
o
n
1
2
J
a
n
u
a
r
y
2
0
2
3
Molecular Subclasses of Clear Cell Ovarian Carcinoma
original draft, writing–review and editing. M. K€obel: Data curation, investigation,
methodology, writing–original draft, writing–review and editing. Y. Tazi: Formal
analysis, writing–original draft, writing–review and editing. J.M. Cunningham:
Writing–original draft, writing–review and editing. I.C.C. Chan: Formal analysis,
writing–original draft, writing–review and editing. B.J. Wiley: Formal analysis,
writing–original draft, writing–review and editing. L.A. Moukarzel: Data curation,
writing–original draft, writing–review and editing. S.J. Winham: Formal analysis,
writing–original draft. S.M. Armasu: Formal analysis, writing–original draft.
J. Lester: Resources, data curation, writing–original draft. E. Elishaev: Resources,
data curation, writing–original draft. A. Laslavic: Resources, writing–original draft.
C.J. Kennedy: Resources, writing–original draft. A. Piskorz: Resources, writing–
original draft. M. Sekowska: Data curation, writing–original draft. A.H. Brand: Data
curation, writing–original draft. Y.-E. Chiew: Data curation, writing–original draft.
P. Pharoah: Conceptualization, data curation, writing–original draft, writing–review
and editing. K.M. Elias: Data curation, writing–original draft. R. Drapkin: Data
curation, writing–original draft. M. Churchman: Data curation, writing–original
draft. C. Gourley: Resources, data curation, writing–original draft. A. DeFazio:
Resources, data curation, writing–original draft. B. Karlan: Resources, data curation,
supervision, writing–original draft. J.D. Brenton: Resources, data curation, super-
vision, writing–original draft. B. Weigelt: Resources, data curation, supervision,
investigation, writing–original draft. M.S. Anglesio: Resources, data curation, super-
vision, writing–original draft. D. Huntsman: Resources, data curation, writing–
original draft. S. Gayther: Conceptualization, resources, data curation, supervision,
investigation, writing–original draft, writing–review and editing. J. Konner: Con-
ceptualization, resources, supervision, funding acquisition, writing–original draft,
project administration, writing–review and editing. F. Modugno: Conceptualization,
resources, data curation, supervision, writing–original draft, project administration,
writing–review and editing. K. Lawrenson: Conceptualization, resources, data
curation, supervision, visualization, methodology, writing–original draft, project
administration, writing–review and editing. E.L. Goode: Conceptualization,
resources, data curation, supervision, investigation, writing–original draft, project
administration, writing–review and editing. E. Papaemmanuil: Conceptualization,
resources, supervision, funding acquisition, methodology, writing–original draft,
writing–review and editing.
Acknowledgments
Survival including the Fatma Fund. B. Weigelt is funded in part by Breast Cancer
Research Foundation and NIH/NCI (P50 CA247749 01) grants. K.L. Bolton is funded
by the Damon Runyon Cancer Research Foundation, the American Society of
Hematology, the Evans MDS Foundation and the NCI (Grant 5K08CA241318).
Additional support was provided by R21CA222867, R01CA248288, P30CA015083,
and P50CA136393. M.S. Anglesio was funded through a Michael Smith Health
Research BC Scholar Program award and the Janet D. Cottrelle Foundation Scholars
Program (managed by the BC Cancer Foundation). This study used resources
provided by the Canadian Ovarian Cancer Research Consortium’s COEUR biobank
funded by the Terry Fox Research Institute and managed and supervised by the
Centre hospitalier de l’Universit(cid:3)e de Montr(cid:3)eal. The Consortium acknowledges
contributions of its COEUR biobank from Institutions across Canada (for a full
list see https://www.tfri.ca/coeur). This work was supported by the Westmead
Hospital Department of Gynaecological Oncology, Sydney Australia. The
Gynaecological Oncology Biobank at Westmead (GynBiobank), a member of
the Australasian Biospecimen Network-Oncology group, was funded by the
National Health and Medical Research Council of Australia (Enabling Grants
ID 310670 & ID 628903) and the Cancer Institute NSW (Grants 12/RIG/1–17 &
15/RIG/1–16). The Westmead GynBiobank acknowledges financial support from
the Sydney West Translational Cancer Research Centre, funded by the Cancer
Institute NSW. A. Piskorz, M. Sekowska, and J.D. Brenton were supported by
Cancer Research UK grant 22905. Additional support was also provided by the
National Institute of Health Research (NIHR) Cambridge Biomedical Research
Centre (BRC-1215–20014). The views expressed are those of the authors and not
necessarily those of the NHS, the NIHR, or the Department of Health and Social
Care.
The publication costs of this article were defrayed in part by the payment of
publication fees. Therefore, and solely to indicate this fact, this article is hereby
marked “advertisement” in accordance with 18 USC section 1734.
Note
Supplementary data for this article are available at Clinical Cancer Research Online
(http://clincancerres.aacrjournals.org/).
Research reported in this publication was supported in part by a Cancer Center
Support Grant of the NIH/NCI (Grant No. P30CA008748, MSK) and the Cycle for
Received October 25, 2021; revised February 24, 2022; accepted July 7, 2022;
published first July 11, 2022.
References
1. Bailey MH, Tokheim C, Porta-Pardo E, Sengupta S, Bertrand D, Weerasinghe A,
et al. Comprehensive characterization of cancer driver genes and mutations. Cell
2018;173:371–85.
2. Dong J, Li B, Lin D, Zhou Q, Huang D. Advances in targeted therapy and
immunotherapy for non-small cell lung cancer based on accurate molecular
typing. Front Pharmacol 2019;10:230.
3. Quint(cid:3)as-Cardama A, Kantarjian H, Cortes J.
Imatinib and beyond—
exploring the full potential of targeted therapy for CML. Nat Rev Clin Oncol
2009;6:535–43.
4. Ortega J, Vigil CE, Chodkiewicz C. Current progress in targeted therapy for
colorectal cancer. Cancer Control 2010;17:7–15.
5. Kayser S, Levis MJ. Advances in targeted therapy for acute myeloid leukaemia.
Br J Haematol 2018;180:484–500.
6. Tangutoori S, Baldwin P, Sridhar S. A new era of targeted therapy. Maturitas
2015;81:5–9.
7. Vaishnavi A, Le AT, Doebele RC. TRKing down an old oncogene in a new era of
targeted therapy. Cancer Discov 2015;5:25–34.
9.
8. Lee Y-Y, Kim T-J, Kim M-J, Kim H-J, Song T, Kim MK, et al. Prognosis of ovarian
clear cell carcinoma compared to other histological subtypes: a meta-analysis.
Gynecol Oncol 2011;122:541–7.
Irodi A, Rye T, Herbert K, Churchman M, Bartos C, Mackean M, et al.
Patterns of clinicopathological features and outcome in epithelial ovarian
cancer patients: 35 years of prospectively collected data. BJOG 2020;127:
1409–20.
11.
12.
Iida Y, Okamoto A, Hollis RL, Gourley C, Herrington CS. Clear cell carcinoma of
the ovary: a clinical and molecular perspective. Int J Gynecol Cancer 2021;31:
605–16.
Jones S, Wang T-L, Shih I-M, Mao T-L, Nakayama K, Roden R, et al. Frequent
mutations of chromatin remodeling gene ARID1A in ovarian clear cell carci-
noma. Science 2010;330:228–31.
13. Wang YK, Bashashati A, Anglesio MS, Cochrane DR, Grewal DS, Ha G, et al.
Genomic consequences of aberrant DNA repair mechanisms stratify ovarian
cancer histotypes. Nat Genet 2017;49:856–65.
14. Anglesio MS, Bashashati A, Wang YK, Senz J, Ha G, Yang W, et al. Multifocal
endometriotic lesions associated with cancer are clonal and carry a high
mutation burden. J Pathol 2015;236:201–9.
15. Murakami R, Matsumura N, Brown JB, Higasa K, Tsutsumi T, Kamada M, et al.
Exome sequencing landscape analysis in ovarian clear cell carcinoma shed light
on key chromosomal regions and mutation gene networks. Am J Pathol 2017;
187:2246–58.
16. Cheng DT, Mitchell TN, Zehir A, Shah RH, Benayed R, Syed A, et al.
Memorial sloan kettering-integrated mutation profiling of actionable cancer
targets (MSK-IMPACT): A hybridization capture-based next-generation
sequencing clinical assay for solid tumor molecular oncology. J Mol Diagn
2015;17:251–64.
17. Andr(cid:3)e F, Arnedos M, Baras AS, Baselga J, Bedard PL, Berger MF, et al. AACR
Project GENIE: powering precision medicine through an international consor-
tium. Cancer Discov 2017;7:818–31.
10. Wentzensen N, Poole EM, Trabert B, White E, Arslan AA, Patel AV, et al.
Ovarian cancer risk factors by histologic subtype: an analysis from the ovarian
cancer cohort consortium. J Clin Oncol 2016;34:2888–98.
18. Sondka Z, Bamford S, Cole CG, Ward SA, Dunham I, Forbes SA. The COSMIC
cancer gene census: describing genetic dysfunction across all human cancers.
Nat Rev Cancer 2018;18:696–705.
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
a
a
c
r
j
o
u
r
n
a
s
.
o
r
g
/
c
l
l
i
n
c
a
n
c
e
r
r
e
s
/
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
2
8
2
2
4
9
4
7
3
2
3
3
0
6
5
4
9
4
7
p
d
/
.
f
i
t
b
y
W
a
s
h
n
g
o
n
U
n
v
e
r
s
i
t
y
S
i
t
i
L
o
u
s
u
s
e
r
o
n
1
2
J
a
n
u
a
r
y
2
0
2
3
AACRJournals.org
Clin Cancer Res; 28(22) November 15, 2022
4955
Bolton et al.
19. Fukumoto T, Magno E, Zhang R. SWI/SNF complexes in ovarian cancer:
mechanistic insights and therapeutic implications. Mol Cancer Res 2018;16:
1819–25.
37. Kuo K-T, Mao T-L, Jones S, Veras E, Ayhan A, Wang T-L, et al. Frequent
activating mutations of PIK3CA in ovarian clear cell carcinoma. Am J Pathol
2009;174:1597–601.
20. Li H, Durbin R. Fast and accurate short read alignment with burrows-wheeler
transform. Bioinformatics 2009;25:1754–60.
21. Benjamin D. Calling somatic SNVs and indels with Mutect2. bioRxiv 2019;
861054.
23.
22. Saunders CT, Wong WSW, Swamy S, Becq J, Murray LJ, Cheetham RK. Strelka:
accurate somatic small-variant calling from sequenced tumor-normal sample
pairs. Bioinformatics 2012;28:1811–7.
Jones D, Raine KM, Davies H, Tarpey PS, Butler AP, Teague JW, et al.
cgpCaVEManWrapper: simple execution of CaVEMan in order to detect
somatic single nucleotide variants in NGS data. Curr Protoc Bioinformatics
2016;56:15.10.1–15.10.18.
24. Ye K, Schulz MH, Long Q, Apweiler R, Ning Z. Pindel: a pattern growth approach
to detect break points of large deletions and medium sized insertions from
paired-end short reads. Bioinformatics 2009;25:2865–71.
25. Karczewski KJ, Francioli LC, Tiao G, Cummings BB, Alf€oldi J, Wang Q, et al. The
mutational constraint spectrum quantified from variation in 141,456 humans.
Nature 2020;581:434–43.
26. Morgulis A, Gertz EM, Sch€affer AA, Agarwala R. A fast and symmetric DUST
implementation to mask low-complexity DNA sequences. J Comput Biol 2006;
13:1028–40.
27. Niu B, Ye K, Zhang Q, Lu C, Xie M, Mclellan MD, et al. MSIsensor: microsatellite
instability detection using paired tumor-normal sequence data. Bioinformatics
2014;30:1015–6.
28. Teh YW, Jordan MI, Beal MJ, Blei DM. Hierarchical Dirichlet Processes. J Am
Stat Assoc 2006;101:1566–81.
29. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. STAR:
ultrafast universal RNA-seq aligner. Bioinformatics 2013;29:15–21.
30. Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose
program for assigning sequence reads to genomic features. Bioinformatics
2014;30:923–30.
31. Love MI, Huber W, Anders S. Moderated estimation of fold change and
dispersion for RNA-seq data with DESeq2. Genome Biol 2014;15:550.
32. Zhou Y, Zhou B, Pache L, Chang M, Khodabakhshi AH, Tanaseichuk O, et al.
Metascape provides a biologist-oriented resource for the analysis of systems-level
datasets. Nat Commun 2019;10:1523.
33. Lisio M-A, Fu L, Goyeneche A, Gao Z-H, Telleria C. High-grade serous
ovarian cancer: basic sciences, clinical and therapeutic standpoints. Int J Mol
Sci 2019;20:952.
34. Chakravarty D, Gao J, Phillips S, Kundra R, Zhang H, Wang J, et al. OncoKB: a
precision oncology knowledge base. JCO Precis Oncol 2017;2017:PO.17.00011.
35. Adzhubei I, Jordan DM, Sunyaev SR. Predicting functional effect of human
missense mutations using PolyPhen-2. Curr Protoc Hum Genet 2013;Chapter 7:
Unit7.20.
36. Ng PC. SIFT: predicting amino acid changes that affect protein function.
Nucleic Acids Res 2003;31:3812–4.
38. Friedlander ML, Russell K, Millis S, Gatalica Z, Bender R, Voss A. Molecular
profiling of clear cell ovarian cancers. Int J Gynecol Cancer 2016;26:648–54.
39. K€obel M, Rahimi K, Rambau PF, Naugler C, Le Page C, Meunier L, et al. An
immunohistochemical algorithm for ovarian carcinoma typing. Int J Gynecol
Pathol 2016;35:430–41.
Jelinic P, Mueller JJ, Olvera N, Dao F, Scott SN, Shah R, et al. Recurrent
SMARCA4 mutations in small cell carcinoma of the ovary. Nat Genet 2014;
46:424–6.
40.
41. Ramos P, Karnezis AN, Craig DW, Sekulic A, Russell ML, Hendricks WPD, et al.
Small cell carcinoma of the ovary, hypercalcemic type, displays frequent
inactivating germline and somatic mutations in SMARCA4. Nat Genet 2014;
46:427–9.
42. Witkowski L, Carrot-Zhang J, Albrecht S, Fahiminiya S, Hamel N, Tomiak E,
et al. Germline and somatic SMARCA4 mutations characterize small cell
carcinoma of the ovary, hypercalcemic type. Nat Genet 2014;46:438–43.
43. Earp MA, Raghavan R, Li Q, Dai J, Winham SJ, Cunningham JM, et al.
Characterization of fusion genes in common and rare epithelial ovarian cancer
histologic subtypes. Oncotarget 2017;8:46891–9.
44. Saito Y, Koya J, Araki M, Kogure Y, Shingaki S, Tabata M, et al. Landscape and
function of multiple mutations within individual oncogenes. Nature 2020;582:
95–99.
45. Morimoto A, Serada S, Enomoto T, Kim A, Matsuzaki S, Takahashi T, et al.
Annexin A4 induces platinum resistance in a chloride-and calcium-dependent
manner. Oncotarget 2014;5:7776–87.
46. Saga Y, Ohwada M, Suzuki M, Konno R, Kigawa J, Ueno S, et al. Glutathione
peroxidase 3 is a candidate mechanism of anticancer drug resistance of ovarian
clear cell adenocarcinoma. Oncol Rep 2008;20:1299–303.
47. Tamaresis JS, Irwin JC, Goldfien GA, Rabban JT, Burney RO, Nezhat C, et al.
Molecular classification of endometriosis and disease stage using high-
dimensional genomic data. Endocrinology 2014;155:4986–99.
48. Worley MJ, Liu S, Hua Y, Kwok JS-L, Samuel A, Hou L, et al. Molecular changes
in endometriosis-associated ovarian clear cell carcinoma. Eur J Cancer 2015;51:
1831–42.
49. Berek JS, Chalas E, Edelson M, Moore DH, Burke WM, Cliby WA, et al.
Prophylactic and risk-reducing bilateral salpingo-oophorectomy: recommenda-
tions based on risk of ovarian cancer. Obstet Gynecol 2010;116:733–43.
50. Delair D, Han G, Irving JA, Leung S, Ewanowich CA, Longacre TA, et al. HNF-1b
in ovarian carcinomas with serous and clear cell change. Int J Gynecol Pathol
2013;32:541–6.
51. Tan TZ, Ye J, Yee CV, Lim D, Ngoi NYL, Tan DSP, et al. Analysis of gene
expression signatures identifies prognostic and functionally distinct ovarian
clear cell carcinoma subtypes. EBioMedicine 2019;50:203–10.
52. Vasan N, Razavi P, Johnson JL, Shao H, Shah H, Antoine A, et al. Double PIK3CA
mutations in cis increase oncogenicity and sensitivity to PI3Ka inhibitors.
Science 2019;366:714–23.
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
a
a
c
r
j
o
u
r
n
a
s
.
o
r
g
/
c
l
l
i
n
c
a
n
c
e
r
r
e
s
/
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
2
8
2
2
4
9
4
7
3
2
3
3
0
6
5
4
9
4
7
p
d
.
/
f
i
t
b
y
W
a
s
h
n
g
o
n
U
n
v
e
r
s
i
t
y
S
i
t
i
L
o
u
s
u
s
e
r
o
n
1
2
J
a
n
u
a
r
y
2
0
2
3
4956
Clin Cancer Res; 28(22) November 15, 2022
CLINICAL CANCER RESEARCH
| null |
10.1371_journal.pdig.0000221.pdf
|
rnal.pdig.0000221 May 15, 2023
1 / 21
PLOS DIGITAL HEALTH protocol approved by the Danish Data Protection
Agency. In the paper, we have also had to limit the
number of indirect identifiers in cases where there
might be a risk of compromising patient privacy
(e.g. Tables 1 and 2). Reviewers and others may
obtain access to the data by request, and after a
data processing agreement has been signed. There
will be no limitation to data sharing as long as a
data sharing agreement is signed. As per Danish
research code, we are required to store research
data for a minimum of 5 years starting from the
time of publication. We will ensure this by standard
data management and storage at our institution.
Please address any correspondence to Dr Thomas
Bandholm; [email protected]
(corresponding author) or the Department of
Clinical Research, Hvidovre Hospital, Kettegaard
Alle 30, DK-2650 Hvidovre, Copenhagen, Denmark.
Phone: +45 3862 3862.
| null |
RESEARCH ARTICLE
Using the app “Injurymap” to provide exercise
rehabilitation for people with acute lateral
ankle sprains seen at the Hospital Emergency
Department–A mixed-method pilot study
Jonas Bak1, Kristian Thorborg2,3,4, Mikkel Bek Clausen2,5, Finn Elkjær Johannsen6,7,
Jeanette Wassar Kirk1,8, Thomas BandholmID
1,2,3,4*
1 Department of Clinical Research, Copenhagen University Hospital, Amager and Hvidovre, Copenhagen,
Denmark, 2 Department of Orthopedic Surgery, Copenhagen University Hospital, Amager and Hvidovre,
Copenhagen, Denmark, 3 Department of Clinical Medicine, University of Copenhagen, Copenhagen,
Denmark, 4 Physical Medicine & Rehabilitation Research–Copenhagen (PMR-C), Department of Physical
and Occupational Therapy, Copenhagen University Hospital, Amager and Hvidovre, Denmark, 5 Department
of Midwifery, Physiotherapy, Occupational Therapy and Psychomotor Therapy, Faculty of Health, University
College Copenhagen, Copenhagen N, Denmark, 6 Institute of Sports Medicine Copenhagen, Copenhagen
University Hospital, Bispebjerg and Frederiksberg, Copenhagen, Denmark, 7 Injurymap Aps, Copenhagen N,
Denmark, 8 Department of Health and Social Context, National Institute of Public Health, University of
Southern Denmark, Odense, Denmark
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
* [email protected]
Citation: Bak J, Thorborg K, Clausen MB,
Johannsen FE, Kirk JW, Bandholm T (2023) Using
the app “Injurymap” to provide exercise
rehabilitation for people with acute lateral ankle
sprains seen at the Hospital Emergency
Department–A mixed-method pilot study. PLOS
Digit Health 2(5): e0000221. https://doi.org/
10.1371/journal.pdig.0000221
Editor: Jasmit Shah, Aga Khan University - Kenya,
KENYA
Received: November 2, 2022
Accepted: February 27, 2023
Published: May 15, 2023
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pdig.0000221
Copyright: © 2023 Bak et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: Public deposition of
raw data points would breach compliance with the
Abstract
Background
Acute lateral ankle sprains (LAS) account for 4–5% of all Emergency Department (ED) vis-
its. Few patients receive the recommended care of exercise rehabilitation. A simple solution
is an exercise app for mobile devices, which can deliver tailored and real-time adaptive exer-
cise programs.
Purpose
The purpose of this pilot study was to investigate the use and preliminary effect of an app-
based exercise program in patients with LAS seen in the Emergency Department at a public
hospital.
Materials and methods
We used an app that delivers evidence-based exercise rehabilitation for LAS using algo-
rithm-controlled progression. Participants were recruited from the ED and followed for four
months. Data on app-use and preliminary effect were collected continuously through the
exercise app and weekly text-messages. Baseline and follow-up data were collected though
an online questionnaire. Semi-structured interviews were performed after participants
stopped using the app. Results: Health care professionals provided 485 patients with study
information and exercise equipment. Of those, 60 participants chose to enroll in the study
and 43 became active users. The active users completed a median of 7 exercise sessions.
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
1 / 21
PLOS DIGITAL HEALTH protocol approved by the Danish Data Protection
Agency. In the paper, we have also had to limit the
number of indirect identifiers in cases where there
might be a risk of compromising patient privacy
(e.g. Tables 1 and 2). Reviewers and others may
obtain access to the data by request, and after a
data processing agreement has been signed. There
will be no limitation to data sharing as long as a
data sharing agreement is signed. As per Danish
research code, we are required to store research
data for a minimum of 5 years starting from the
time of publication. We will ensure this by standard
data management and storage at our institution.
Please address any correspondence to Dr Thomas
Bandholm; [email protected]
(corresponding author) or the Department of
Clinical Research, Hvidovre Hospital, Kettegaard
Alle 30, DK-2650 Hvidovre, Copenhagen, Denmark.
Phone: +45 3862 3862.
Funding: The study was funded partly by the
European Regional Development Fund (https://ec.
europa.eu/regional_policy/funding/erdf_en), which
was administered by the Copenhagen Center for
Health Technologies (CACHET, https://www.cachet.
dk/) and given to TB. Funding was for salary
support for a research assistant (JB). The funders
had no role in study design, data collection and
analysis, decision to publish, or preparation of the
manuscript (please see details in the conflict-of-
interest statement).
Competing interests: I have read the journal’s
policy and the authors of this manuscript have the
following competing interests: FEJ is a co-founder
of Injurymap. This conflict was accommodated by
restricting FEJ from any deciding role in terms of
study design, study management, data
interpretation, report writing and submission. TB
has received speaker’s honoraria for talks or expert
testimony on the efficacy of exercise therapy to
enhance recovery after surgery at meetings or
symposia held by biomedical companies (Zimmer
Biomet and Novartis). He is an editorial board
member with Br J Sports Med. KT is a deputy
editor with Br J Sports Med and receives an annual
honoraria. JWK, JB and MBC have declared that no
competing interests exist.
App-based exercise rehabilitation and ankle sprains
Most of the active users were very satisfied or satisfied (79%-93%) with the app and 95.7%
would recommend it to others. The interviews showed that ankle sprains were considered
an innocuous injury that would recover by itself. Several app users expressed they felt insuf-
ficiently informed from the ED health care professionals. Only 39% felt recovered when they
stopped exercising, and 33% experienced a recurrent sprain in the study period. Conclu-
sion: In this study, only few patients with LAS became active app users after receiving infor-
mation in the ED about a free app-based rehabilitation program. We speculate the reason
for this could be the perception that LAS is an innocuous injury. Most of the patients starting
training were satisfied with the app, although few completed enough exercise sessions to
realistically impact clinical recovery. Interestingly more than half of the participants did not
feel fully recovered when they stopped exercising and one third experienced a recurrent
sprain.
Trial-identifiers
https://clinicaltrials.gov/ct2/show/NCT03550274, preprint (open access): https://www.
medrxiv.org/content/10.1101/2022.01.31.22269313v1.
Introduction
An ankle sprain is one of the most common musculoskeletal injuries with a comprehensive
burden for individuals as well as society [1]. They account for 4–5% of all Emergency Depart-
ment visits in Denmark [2] which is consistent with data from other countries [1]. This might
just be the tip of the iceberg since less than half of the people who sustain an ankle sprain seek
health care [1,3]. A lateral ankle sprain (LAS) is often regarded as an innocuous injury [4] and
especially health care professionals tend to overestimate the recovery [5]. However, 32–74% of
people who sustain a LAS have prolonged symptoms such as pain, decreased function and sub-
jective instability for several years after their initial injury [1,6]. In sport, up to 34% will sustain
a recurrent sprain in the following years after their initial injury [6]. However, exercise therapy
is a well-documented cost-effective rehabilitation modality to treat LAS [7–10] and prevent
recurrent sprains [8,11,12]. It is unfortunate that few patients are prescribed exercise programs
or physiotherapy after a LAS, and that most expenses relate to diagnostic procedures rather
than exercise-based rehabilitation [4]. Technologically supported self-management may be a
solution to this problem but requires investigation.
Applications for smart devices (apps) have the potential to be powerful tools in providing
easily accessible exercise therapy programs and in attaining important information about exer-
cise behavior that have been practically unobtainable previously [13]. They represent a flexible
telehealth solution that most often does not require the online presence of a health care profes-
sional. Apps have the ability through interaction between users and smartphone to tailor spe-
cific information and exercise programs. Furthermore, apps can give real-time, real-life
feedback during an exercise session without waiting for a health specialist to be available [13].
A serious challenge in the use of apps, however, is that there is a general lack of evidence-based
solutions, and that health apps often wrongly claim to be evidence founded [14–16]. Another
challenge is that the effectiveness of the majority of health apps are fairly unknown and their
ability to make actual behavioral change is often poorly reported [17,18]. The high availability
of health apps on a unregulated market poses a major concern since it may have a major influ-
ence on rehabilitation success or at worst cause harm [14].
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
2 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Injurymap is an exercise app designed for treating different musculoskeletal problems
including LAS. The Injurymap exercise program has been developed by health care profession-
als and has the potential to provide an easy-accessible management of rehabilitation. However,
the app has currently not been tested in a clinical study. Before undertaking a large-scale
study, we wish to pilot test the app to assess the use and preliminary effect of the app-based
exercise program. We are particularly interested in the proportion of patients who become
active app users (and for how long) when provided with an option of a free app-based rehabili-
tation program in an acute care clinical setting.
Purpose statement
The purpose of this pilot study was to investigate the use and preliminary effect of an app-
based exercise program in patients with LAS seen in the Emergency Department at a public
hospital. Consistent with the mixed-method study design outlined below, both quantitative
and qualitative outcomes were collected to fully explore factors related to uptake and adher-
ence to the exercise program.
Method
Study design
The study used an explanatory sequential mixed method cohort study design (QUAN!qual)
[19]. The study was conducted in subsequent phases, with the development of the quantitative
outcomes leading the following development of qualitative outcomes, and the quantitative
app-use data guiding a purposive sampling for semi-structured interviews [20]. Quantitative
and qualitative outcomes were analyzed separately and integration was done in the interpreta-
tion process by triangulation (sections with integration: Methods, Results, and Discussion)
[21]. The study process is outlined in Fig 1. We consider the study exploratory and, hence, it
was designed with a flat outcome structure using multiple evenly valued outcome measures.
Outcomes related to “app-use” were quantitative data collected directly from the app during
the exercise period and after by a follow-up questionnaire. Outcomes meant to provide explan-
atory insight of the app-use data were qualitative and collected by semi-structured interviews
after the exercise period. Outcomes related to “preliminary effect” were quantitative clinical
recovery-data collected using weekly TEXT-MESSAGES for four months after the initial
injury.
The SPIRIT checklist [22] and PREPARE Trial guide [23] were used to develop the study
protocol. The Danish National Committee on Health Ethics approved this study (ID:
17041467). All personal data were handled according to the Danish act concerning processing
of personal data. Prior to the study, Injurymap had been approved for handling personal data
by the Danish Data Protection Agency (file no. 2016-42-3535). The study followed the princi-
ples of the Helsinki declaration and it was pre-registered at ClinicalTrials.gov (https://
clinicaltrials.gov/ct2/show/NCT03550274). We report the study using the Good Reporting of
A Mixed Methods Study (GRAMMS) checklist [24] (S1 Table).
Study setting
The study was performed at a Danish public hospital where patients are covered by the Danish
healthcare system. At the hospital, the current practice for non-surgical management of LAS
patients is RICE (Rest, Ice, Compression, Elevation), mobility exercises and recommendations
of slowly returning to activity. The current practice does not involve any on-site systematic
instruction in evidence-based rehabilitation programs or referral to such elsewhere. Therefore,
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
3 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Fig 1. Mixed-method study process.
https://doi.org/10.1371/journal.pdig.0000221.g001
we wanted to explore the use of a simple app-based solution in this setting. Informed written
consent from the participants were registered before participants started the app-based exercise
program. Participants could withdraw from the study at any time, without any consequences.
Participants
Participants were recruited from the Emergency Department at Copenhagen University Hos-
pital, Hvidovre. They were asked to participate in a home-based rehabilitation delivered by an
app (Injurymap) available on any smart device. The inclusion criteria was; patients with an
acute lateral ankle sprain (< 48 hours from injury) diagnosed by a relevant health care profes-
sional at the hospital ED. Gradings of ankle sprain severity was not performed, since this is not
standard procedure in the ED. The exclusion criteria were; concurrent fracture of the leg or
foot (Ottawa rules and/or x-ray), previous surgery in the ankle or surgery as a consequence of
the current ankle sprain, serious illness (terminal patient, rheumatoid arthritis, fibromyalgia
etc.), not owning a smart device (phone or tablet) or unable to understand and read Danish.
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
4 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Procedures
Health care professionals associated with the ED and responsible for ankle examinations
recruited participants. When a health care professional identified a patient with ankle sprain,
they delivered a recruitment bag containing several rubber bands of different thickness and a
description of the free app-based exercise opportunity in this study. This approach was chosen
to resemble a delivery method applicable in clinical practice. Besides the written information,
the health care professionals were encouraged to recommend the exercise program to the
patients. If a patient was willing to participate in the project, they contacted the research assis-
tant (JB) by the contact information in the written material. When contact was established,
and participants were deemed eligible, they received a voucher for free access to the app pro-
gram, informed consent, and a baseline questionnaire.
The project was implemented at the ED by the primary investigator (JB). Health care pro-
fessionals in the ED were informed about the study at staff meetings, by the weekly newsletter,
and by the primary investigator who participated in the daily routines prior to recruitment for
the project. Two large boxes containing the recruitment bags were placed strategically in the
ED office and the primary examination room. The boxes had a large picture of an ankle sprain
at the front and a text asking to give patients an exercise opportunity. This recruitment proce-
dure was chosen to reflect a normal clinical care setting and to encompass both the health care
professionals’ willingness to promote the app solution to patients with LAS, as well as partici-
pants’ willingness to accept the offer. For the same reason, we tried to inform about the study
as being a rehabilitation exercise opportunity more than a research study.
Intervention
InjuryMap offers exercise programs for LAS and other musculoskeletal conditions. The app
requires user-registrations and a monthly paid subscription fee to access the exercise program.
In this study, the app company provided free access for participants to the LAS exercise pro-
gram. Examples from the app content can be seen in Fig 2.
The exercise program was available on any mobile device and/or tablet using Android or
iOS operating systems. Participants could perform the exercises at any preferred location and
were not restricted from seeking additional care. After four months of training or if a partici-
pant was inactive in the app for more than two consecutive weeks, they were considered to
have stopped the exercise intervention.
The exercise program for LAS consisted of three phases with increasing difficulty. Each
phase consisted of four categories of exercise. The categories were 1) mobility, 2) stability/bal-
ance, 3) strength and 4) stretching. The app could adjust the difficulty for each individual exer-
cise depending on user-feedback. The user is asked to rate each exercise on a pain scale and
difficulty scale after completing it. Their rating is used in an algorithm to calculate progres-
sion/regression for each exercise and to catch red flags. A comprehensive description of the
exercises can be found in S1 Text and S2 Table.
The exercise program was set up so that several exercises were completed after each other
to successfully complete a training session. Each exercise was accompanied by a video with an
explanatory voice-over and the number of required repetitions written on the display. After
completing an exercise, the participants registered pain level and difficulty of performing the
exercise. If the participants registered no or low pain and low exercise difficulty, the app chose
a progression of the exercise for the next exercise session. The app recommended participants
to complete exercise sessions three times a week after app registration. If the participants fol-
lowed the recommended three sessions a week, they would be able to reach the highest diffi-
culty level in two to four months depending on their pain and difficulty answers. There were
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
5 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Fig 2. Example of app content.
https://doi.org/10.1371/journal.pdig.0000221.g002
no limitations in how many exercise sessions could be performed per week or a maximum
number of weeks they could exercise except for the strengthening exercises which could only
be performed once per day. Participants were able to activate a reminder function so that the
app would remind them to exercise on a daily timepoint of their choosing. The spoken and
written language in the application was Danish.
In the exercise program, circular rubber bands were used in several exercises. The rubber
bands are common and cheap products available in most sports stores. In this study, the
recruitment bag contained a selection of rubber bands with varying resistance.
Outcomes
In this study, the outcome data were divided into two categories reflecting the study objective.
The first category (“App-use”) consists of quantitative data on uptake, retention, and adher-
ence. Furthermore, it contained user-experience which is comprised of quantitative data on
satisfaction with the app and qualitative data on the factors that influenced the app use. The
second category (“Preliminary effect”) consists of quantitative data on clinical recovery and
recurrent injuries. A baseline questionnaire was completed after enrollment to gather descrip-
tive data.
App-use. The overall rationale for the app-use outcomes below was to investigate how
many participants exposed to the app that started using it; how much they used it; when they
stopped using it; and how the user experience was. As a part of the app evaluation, we assessed
uptake of the app-based exercise program. For app uptake, we calculated the following: num-
ber of participants diagnosed with an ankle sprain at the ED in the study period; number of
participants who received a recruitment bag; number of participants willing to participate
(contacted by the principal investigator); number of participants who became active users
(defined as having downloaded the app and initiated the exercise program). By counting how
many recruitment-bags the health care professionals delivered, the number of participants
with ankle sprain who had been informed about the exercise opportunity could be estimated.
From the Danish National Patient Register it was possible to obtain the number of ankle
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
6 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
sprains diagnosed at the ED. For retention, following were calculated: Number of participants
completing baseline and follow-up questionnaire; and number of text-messages responded
through the study period.
The app had mandatory user-registration so all exercise activity in the app was registered at
the individual participant level. From these data, we calculated the following adherence out-
comes: Number of exercise sessions completed per participant; and completed exercise ses-
sions per week. If a participant did not commence the exercise program within two weeks,
when active users was inactive for two consecutive weeks, or if they were active in the app for
four months from their initial injury, they were considered finalized and received the follow-
up questionnaire containing user-experience with the app-based exercise solution. Satisfaction
on a five-point Likert scale were assessed for active users for the following items: the difficulty
and the progression of the exercise program (Difficulty), the content of the exercise program
(Content), the results from the exercise program (Results), and the usability of the app (User-
friendly). Furthermore, all participants were asked if they would recommend the app to others
(yes/no) and how much they would be willing to pay for the app (DKK).
After ending the exercise intervention, a group of participants were contacted for semi-
structured interviews. The interviews focused on understanding and explaining motivational
factors or barriers that may influence the use of the app “Injurymap”. The study used a pur-
poseful sampling for the interviews as recommended for explorative mixed method studies
[20]. The sampling of participants was based on different number of completed exercise ses-
sions, different age groups, both men and women. We did this to capture get a broad perspec-
tive on the motivational factors or barriers from both those with many completed exercise
sessions and those who dropped out early. Based on the sampling criteria, a pragmatic number
of ten participants were selected (both men and women at different age groups).
The interviews were performed by phone by the principal investigator (JB). Participants for
the semi-structured interviews were contacted by mail and phone at the same time as the fol-
low-up questionnaire. An interview guide was developed by the principal investigator (JB)
with supervision from a senior researcher experienced in qualitative research (JWK). The
guide was pilot tested on a person with experience in the exercise app, but otherwise not
involved in the study. Recordings from the pilot testing were examined by two researchers
(JWK and JB) to improve the interview technique and evaluate coherence of the questions in
the interview guide. After the first interview, the recordings were examined again by the senior
researcher (JWK) the recordings were compared to the purpose statement to ensure it was ade-
quately covered in the interviews. Changes in the interview guide from the first interview con-
sisted primarily of merging separate themes, based on how the participants associated and
described their experiences. We also added questions about the experience at the ED and its
impact on app-use since this factor was mentioned as an influential factor.
Interviews were recorded and transcribed verbatim. The data were analyzed using a the-
matic approach as described by Castleberry [25]. The data were coded, and recurring phrases
or words were grouped into basic themes by the principal investigator (JB). Themes and codes
were compared to the transcribed interview by two researchers (JB and JWK) to ensure that
coding was performed with the same consistency and true to the original statements. This was
performed in several processes until agreement was achieved. In this process, basic themes
were clustered into global themes. The initial interpretation was performed by the principal
investigator (JB) and reviewed by a researcher (JWK). Finally, the qualitative results were dis-
cussed with the whole research group.
Preliminary effect. The overall rationale for the preliminary effect-outcomes outlined
below was to investigate if exercise adherence, was related to clinical recovery. Clinical recov-
ery was evaluated by self-reported evaluation of symptoms using a weekly string of text-
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
7 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
messages for four months after their initial injury. The following Clinical Recovery items were
collected by text-messages: Not able to fully participate in work/study because of the ankle
sprain (days); Return to sport (RTS), defined as not able to fully participate in sport because of
the ankle sprain for participants who registered as being “sports active” in the baseline ques-
tionnaire (weeks); Recurrent lateral ankle sprains in the same ankle (number); Subjective feel-
ing of ankle stability (0–10 points). From the follow-up questionnaire, the clinical recovery
item: Subjective feeling of recovery (yes/no) was also collected.
A recurrent sprain was defined as an inversion episode on the same ankle as assessed in the
ED. Recurrent sprains were divided into two groups; 1) Recurrent sprain with time-loss,
defined as being unable to continue current activity and/or unable to participate in work/
sports activities the next day because of the ankle; 2) Recurrent sprain with no time loss was
defined as able to continue with current activities and able to participate in sports/work activi-
ties the next day.
As the exercise program in the app is built with similar component as other evidence-based
exercise programs [8], we expected no harms from the intervention. Nonetheless, participants
received a text-message in the weekly string of text-messages where they could register any dis-
comforts or injuries related to performing the exercise program.
Materiel and outcome assessment
The baseline and the follow-up questionnaires were collected without assessor involvement
through RedCap (Research Electronic Data Capture)—a browser-based software developed by
Vanderbilt University. Participants received emails containing a link to their personal online
questionnaire. The research team had access to app use data through a log-in to the Injurymap
online database. The text-messages were collected using SMS-track—an online system used to
send and receive standardized text-messages. Each week the SMS-track system sent the first of
six questions to the participants and waited for an answer before the next question was sent. If
a participant reported an ankle sprain, she or he received a phone call from the principal inves-
tigator (JB) to clarify the type of sprain.
The principal investigator (JB) performed the outcome assessment, follow-up assessment,
data extraction and data analysis. Participants active in the app were contacted as little as possi-
ble to minimize any potential influence on their exercise behavior.
Statistical analysis
Baseline characteristics are summarized using suitable descriptive statistics. Normal distribu-
tion was assessed using the Shapiro-Wilk test and Q-Q plots. For the app-use outcomes,
recruitment rates, retention rates are presented in suitable descriptive tables. Adherence is
summarized in total sessions per participant and exercise sessions per week during the inter-
vention period. Registered Harms were addressed individually. Quantitative user-experience
are summarized using descriptive statistics.
We planned to determine the preliminary effect of the app by examining the relationship
between the exercise dose (adherence) and clinical recovery outcomes using linear or logistic
regression models, depending on type of outcome. The models would include the clinical
recovery outcome as the dependent variable and exercise dose as independent variable. How-
ever, as the weekly text-messages that contained questions on clinical recovery were answered
more frequently by participants who were very active in the app, we chose not to conduct the
analyses of how exercise adherence related the clinical-outcomes as they would be biased.
Instead, we report the clinical recovery data using descriptive statistics. Analyses were done
using MS Excel, R, and MS Word.
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
8 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Sample size
Approaches to sample size justification for studies that investigate preliminary effectiveness of
interventions, such as pilot and feasibility trials, vary. One rule of thumb-approach is 12 per
group for a pilot RCT [26]. We used a pragmatic sample size 60 participants for this study. It
was based on the rationale that 30 out of 60 would download the app, start using the app, and
start using the exercise program, based on previous experience with the app on low back pain
patients. We figured a sample size of 30 app users would equate to two groups of 12 partici-
pants each, if pooled [26], plus 6 to account for some attrition.
Results
60 participants were recruited during the period from July 3, 2018, to April 3, 2019. 43 of these
stated that they participated in weekly sports activities (Sport active group). Accounting for
half of the sprains, sports were the most frequently reported cause of injury. One third (n = 20)
reported a previous ankle sprain in the same ankle. The sprains were equally divided between
left and right site sprains. Baseline characteristics are presented in Table 1.
Table 1. Baseline characteristics of participants with acute lateral ankle sprains (N = 60).
Item
Age (yr)
Weight (kg)
Height (cm)
Women
Injury site (Right)
Previous (same) ankle sprain (yes)
Sports active, (yes)
Education level:
Primary education
Upper secondary education
Vocational Education
Short higher education
Bachelors program
Master program or higher
Other
Physical demands on work:
Mostly sitting
Equal sitting and walking
Mostly Walking
Activity when injured:
Work
Sports
Leisure
Other
Mean(SD) N
In their 30s 60*
76.60(19.85) 56
174.04(10.75) 56
n(%) N
36(64.3%) 56
29(51.8%) 56
20(35.7%) 56
43(78.2%) 55
56
11(19.6%)
10(17.9%)
4(7.1%)
1(1.8%)
23(41.1%)
6(10.7%)
1(1.8%)
22(40%)
19(34.5%)
14(25.5%)
8(14.3%)
26(46.4%)
22(39.3%)
0 (0%)
55
56
One or more element of the RICE-principle (yes)
52(92.9%) 56
*Data on age were collected via the written informed consent. All other data were collected via the baseline
questionnaire. Wording chosen to limit the number of indirect identifiers.
https://doi.org/10.1371/journal.pdig.0000221.t001
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
9 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Table 2. Characteristics of the interviewed participants with acute lateral ankle sprains (N = 10).
M/W
Age (yr)
Weight (Kg)
Height (cm)
Educationa
Employment
No sessions completed
1
W
*
*
*
2
W
*
*
*
PostG
PostG
Out
26
Job
21
3
M
*
*
*
Skill
Job
4
3
M
*
*
*
Grad
Job
16
5
M
*
*
*
Grad
job
8
6
W
*
*
*
Grad
Job
3
7
M
*
*
*
Grad
Out
5
8
M
*
*
*
Grad
Job
1
9
M
*
*
*
10
W
*
*
*
<High
job
7
<High
Stud<
3
aEducation: < High School (<high), High School (High), Skilled (Skill), Graduate (Grad), Post Graduate (PostG)
bEmployment: Employees or self-employed (job), Unemployed (No), Not in the labor force (out), Student in high school or lower education level (Stud<), graduate
student or higher education level (Stud>).
* Removed to limit the number of indirect identifiers.
https://doi.org/10.1371/journal.pdig.0000221.t002
Characteristics of the 10 participants interviewed can be seen in Table 2. They had com-
pleted between 1 and 26 exercise sessions. 19 participants were contacted to achieve 10 inter-
views. One participant refused to participate in the interview due to lack of time. The
remaining eight participants did not respond to the request.
From the coding process [25], three themes were deducted (see Table 3). Each theme was
divided by several sub-themes. Theme I-II are characterized by a substantial amount of data
and a clear link to the purpose of the interviews. Theme III is also characterized by a substan-
tial amount of data and includes factors that were not directly linked to the app experience but
with a potential impact on the use of the app. Quantitative adherence data and qualitative data
are presented through joint display table as recommended by Guetterman et al. [27].
App-use
Uptake. According to the Danish National Patient Register [28], a total of 1110 people
were diagnosed with an ankle sprain in the ED during the study period (see Table 4). It should
be noted that this does not only include isolated sprains or lateral sprains, which is why the
number of people is considered an absolute maximum of potentially eligible participants. 485
people received the recruitment bags containing rubber bands and study information. Of
Table 3. Themes and subthemes.
Themes
I: Motivational factors
II: Technology assisted exercise behavior
III: Factors of importance for start-up
https://doi.org/10.1371/journal.pdig.0000221.t003
Subtheme
Usability
The app’s exercise level and ability to adapt
Ankle symptoms and expectations
Influence of work and leisure time
Process statistics
Reminder function
Exercise Comprehension
Views on the visual expression
Diagnostic and prognostic expectations
Treatment and preventive expectations
Provider integrity
The user as independent searcher
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
10 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Table 4. Uptake of the app intervention.
Project stage
No of people
Percentage of total diagnosed
Percentage of those who received info
Diagnosed with ankle sprain in the ED
Received study information
Enrolled in the study
Active users
1110
485
60
48
https://doi.org/10.1371/journal.pdig.0000221.t004
100%
44.7%
5.4%
4.3%
-
100%
12.4%
9.9%
these, 60 contacted the principal investigator and were included. 48 became active users of the
app, according to our definition.
Retention. Of the 60 participants enrolled, 54 (92%) answered the baseline questionnaire
(one participant had a partial completion) and 46 (77%) answered the follow-up questionnaire.
Two participants dropped out, one because he did not become an active app user and, hence,
did not want to answer the SMS-string, one due to pregnancy.
For the 60 participants in the 17 weeks follow-up period, a total of 6120 SMSs could poten-
tially have been answered. A total of 4387 answers were received (72%), with the highest fre-
quency in week 1 (85.0%) and the lowest in week 13 (61.9%) (see Fig 3).
Exercise adherence. 48 participants became active users because they completed a mini-
mum of 1 exercise session (see Fig 4). The median number of completed exercise sessions in
the four months period was 5.5 and ranged from 0 to 68 completed sessions with the majority
of the participants completing few or very few sessions (see S1 Fig for completed sessions at
the individual level). An exploratory analysis of the adherence by education level, age, work,
and sports active can be found in S2 Fig.
The interviews sought to gain an explanatory insight regarding factors that may have influ-
enced uptake and adherence to the exercise program. Theme I “Motivational factors” describes
factors directly linked to adherence by the participants. Theme II “Technology assisted exercise
behavior” describes how participants viewed technological features that may have influenced
adherence but was not directly connected by the participants. Theme III “Factors of impor-
tance for start-up” describes factors that may have influenced participants to become active
users. See Table 3 for subthemes related to the themes and Table 5 for a joint display of the
quantitative and qualitative findings.
Fig 3. Weekly SMS response rate. Calculated as a mean percentage of the response percentages to the 6 SMS
questions per week.
https://doi.org/10.1371/journal.pdig.0000221.g003
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
11 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Fig 4. Weekly distribution of completed exercise sessions.
https://doi.org/10.1371/journal.pdig.0000221.g004
Satisfaction. When asked at follow-up, 95.7% of the participants would recommend the
app to other people with an ankle sprain. 71% were willing to pay for the exercise program
with an average cost of 46 DKK equivalent to 6.16 EUR. Satisfaction scores can be seen in
Fig 5.
Preliminary effect
A total of 36 recurrent sprains were reported in the follow-up period. Of these, 32 were time-
loss injuries and 4 were not. 20 participants had minimum 1 recurrent sprain and 11 of the 20
had 2 or 3 recurrent sprains. No participants had more than 3 recurrent sprains in the period.
At follow-up, 39.1% (18 participants) felt that the ankle was able to perform at the pre-injury
level. The absent from work/study ranged from 0 to 100 days, the median being 1 (IQR = 6)
day. In week 1 the average ankle stability was 4.5 (SD = 2.5) on the 0–10 scale and increased to
8.7 (SD = 1.5) in week 17. Weekly changes can be seen in Table 6. The sports active group
(n = 43) had in average 9 weeks (SD = 4.9) where they could not participate in sports activities
without restrictions from their sprained ankle. After 17 weeks, 30.2% still reported that they
were restricted by their ankle in sports activities.
Harms
No harms were registered.
Discussion
The present study investigated the use of an app-based, rehabilitation exercise program for lat-
eral ankle sprains by collecting data on use of the app “Injurymap”.
From the 1110 patients who were diagnosed with an ankle sprain at the ED during the
study period, 45% received the information about the free app-based exercise program. Of
those who received the information, 10% became active users. In this study, we were able to
determine exactly how many received study information and how many became active users.
To our knowledge, this is the first study to provide a precise estimate of how likely people are
to use a rehabilitation app when presented with the opportunity in a clinical setting. This, in
turn, allows for evaluation of app uptake and if changes in recruiting method or the app design
affect use. We consider such data important in understanding the expanding and unregulated
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
12 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Table 5. Joint display of quantitative adherence data and qualitative explanatory findings.
QUAN outcome
Theme Sub-theme
Qual outcome (participant)
Interpretation
Uptake: 79% (38 of 48)
completed at least one
session in week 1.
III
Diagnostic and
prognostic action
I definitely had an expectation that they
would take x-rays which they also did and
when they then determined that nothing was
broken, I was given a bandage dressing and
instructions about how to treat this sort of
swelling, so that was strictly by the book. (ID
7)
Patients expect a diagnostic focus when
visiting the ED, and do not expect or
demand a focus on rehabilitation.
Impact
Decreased
uptake
Adherence: Minimum 50%
(24 of 48) completed at least
1 exercise per week in the
first 6 weeks
III
III
Diagnostic and
prognostic action
Treatment and
preventive action
III
Provider integrity
III
The user as
independent
searcher
I
I
Usability
Usability
II
Process statistics
I
The app’s exercise
level and ability to
adapt
I thought that they [the Emergency
Department] should be able to see on the x-
ray if it would take a month or two months.
(ID 12)
Patients believes that time to full
recovery can be predicted from the
diagnostics but did not consider their
involvement as part this prognostic.
Decreased
uptake
I don’t know if it was a health professional, it
was of course, probably, a doctor, but there
was also an intern on the side, but I still think
the level of information were fairly low you
could say. I had expected some more advice
and guidance and a more extensive
explanation on what the injury was.
(ID 12)
It seemed sort of verified. Like it wasn’t some
kind of scam-app, who would be like “try this”
and then there would be all kind of
commercials and premium stuff and whatever
that you end up spending money on. This felt
trustworthy and verified. (ID 13)
I don’t use apps much so I would never have
figured to go and search for an app that could
help me. (ID 5)
I thought it was good because I could do it at
work if I had 5 minutes to spare. First of all, it
didn’t take very long, and you were able to do
it everywhere. That was a major plus, that
you were able to do it everywhere. (ID 17)
I didn’t fancy those exercises where you had to
lay on a madras, because then you have lie
down and then you have to find the madras
and where should it lie? . . . I liked those
exercises where you just use a chair in front of
you, and then either you have to go stretch on
it, or you have to hold your balance, or sit on
it and do an exercise, I think that’s simple.
(ID 2)
I think it’s fine, then I can see that now I have
done 10% now I have done 20%, which I
really like. That is if I progress you know. I
actually really like It . . . It’s like running on a
treadmill where you can see how far you’ve
run. I really like it. (ID 4)
I actually think it was fine to begin with. So,
to begin with, it was actually fine, there was
nothing that bothered me, it wasn’t until I
tried it, yeah, I think maybe I had been doing
it for about a week and had been doing it
those three, four sessions that I felt “okay this,
doesn’t challenge my ankle enough for it to
help with my rehabilitation. (ID 13)
Patients felt insufficiently informed
about their ankle sprain and did not feel
guided in the following clinical course
Decreased
uptake
Health personnel is seen as important
influencers when patients are exposed to
the app.
Increased
uptake
The ED may be an important setting to
present an app solution, since patients
may not independently search
themselves.
Increased
uptake
Short duration made the program easy to
commence
Increased
adherence
Even basic requirements may decrease
the usability.
Decreased
adherence
Simple statistics gave a feeling of being
part of a progress
Increased
adherence
Inappropriate starting level and/or
progression gave frustrations when
performing exercises
Decreased
adherence
(Continued )
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
13 / 21
PLOS DIGITAL HEALTH Table 5. (Continued)
QUAN outcome
Theme Sub-theme
Qual outcome (participant)
Interpretation
Impact
App-based exercise rehabilitation and ankle sprains
I
I
I
I
I
The app’s exercise
level and ability to
adapt
The app’s exercise
level and ability to
adapt
The app’s exercise
level and ability to
adapt
Ankle symptoms
and expectations.
Ankle symptoms
and expectations.
I
II
Influence of work
and leisure time
Process statistics
For over a month I’ve been in phase two and I
simply don’t understand it and I actually get
pretty mad when it says “you have completed”
and then it’s still at 90%, I should have
finished phase three by now but I still can’t get
to it.” (ID 4)
I don’t know why it didn’t progress . . . I
answered that the exercise was hard, but that
doesn’t make it bad. It is like doing strength
training then you also must push yourself to
increase you level and that is hard but doesn’t
make it bad. (ID 2)
It seemed that you had to go through the first
step and then the second and the third. You
couldn’t just skip to step three if you wanted
something different or more challenging. (ID
13)
I thought it would take a month. Especially
because I heal very well (laughing), maybe not
as well as previously but I still thought that it
probably would take about a month before I
was completely healthy again. (ID 12)
The problem was that it started to go well
with my ankle . . . I don’t think that I was
injured enough to keep the motivation. It’s
like people with back pain. As soon as the
pain is gone, they don’t do their exercises and
I think it was the same that happened to me.
(ID 17)
It was really just returning to work and the
chores of daily living. It didn’t have anything
to do with the exercises or the app. (ID 7)
I think it’s fine, then I can see that now I have
done 10% now I have done 20%, which I
really like. That is if I progress you know. I
actually really like It . . . It’s like running on a
treadmill where you can see how far you’ve
run. I really like it. (ID 4)
II
Reminder function Well I basically turned it off, because I
II
II
Exercise
comprehension
Views on visual
expression
decided not to receive the push-notifications
which is what I generally do with all apps or
programs I install, so I don’t get bombarded
with various pointless information. Here, it
would actually have been fine for me. (ID 7)
I was surprised how easy it was to follow the
instructions in the exercise videos. I really
think this is relevant because it was so easy to
just start the program. (ID 5)
I actually think it was really good with a
naked room without anything but the things
that were being used, a chair, a table, a
mattress. You know a room with nothing, not
even on the wall. I think that was really good.
(ID 2)
Patients felt stagnated when they did not
understand progression in the app
Decreased
adherence
If participants felt, they could not
communicate around key concepts with
the app they became frustrated.
Decreased
adherence
Lack of autonomy in progression may
negatively affect adherence
Decreased
adherence
There was a general belief that an ankle
sprain would recover spontaneously
within few months.
Decreased
adherence
When symptoms from the ankle sprain
decreased, they felt less motivated to
exercise
Decreased
adherence
When participants became able to work,
perform daily chores or travel they lost
motivation for exercising.
Decreased
adherence
Simple statistics gave a feeling of being
part of a progress
Increased
adherence
Reminders did not seem to be an
important function in the app, but many
had turned it off already from the
beginning.
No influence
on adherence
The videos were a strong contributor to
exercise comprehension.
Increased
adherence
A plain video expression seemed to give
integrity to the exercises.
Increased
adherence
https://doi.org/10.1371/journal.pdig.0000221.t005
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
14 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Fig 5. Satisfaction scores for different app use items.
https://doi.org/10.1371/journal.pdig.0000221.g005
field of health apps [14]. Since this is the first study to evaluate the delivery of an exercise app
for ankle sprains in an ED setting, it is difficult to compare the uptake data to many other stud-
ies. Vriend et al. [29] did evaluate the uptake of an exercise app for people with previous ankle
sprains when advertised in national media and on sports facilities. They estimated that the app
was downloaded by less than 2.6% of their targeted population despite intensive marketing,
and that only 62% of those who downloaded the app became active users. Though the authors
were not able to determine the percentage of the targeted population that became aware of the
app´s existence, they concluded that a “marketing” type of strategy may not be the optimal
method of implementing an evidence-based app. Compared to the estimated 2.6% referenced
above, the 10% active users in our study was better, however, it is difficult to consider 10%
uptake a success. The higher uptake-level in our study indicates that the direct delivery of the
app from a health care professional in the ED can encourage more people to download the
exercise program. This was further supported by our qualitative data showing that the app,
when given by a health care professional, seemed trustworthy and this had influenced partici-
pants to download the app; a finding that aligns with a previous study where people stated they
were more likely to use an app if it was endorsed by a health care professional [30]. It was,
however, beyond the scope of this study to investigate how the health care professionals deliv-
ered the app information and their beliefs towards it, and whether this affects uptake to a high
degree. Interestingly, the interviews also revealed that several participants felt insufficiently
informed about their injury when leaving the ED. This might indicate that more adequate
information from health care professionals about the consequences of LAS and recommended
exercise rehabilitation could prompt more patients to use the app and health care professionals
in the ED could have a great opportunity to influence people’s behavior by providing such
information in a clinical setting.
The active users completed a median of 5.5 exercise sessions. Most participants became
active in the first week, but 20% started after the first week. In general, adherence declined
through the study period and after 2 weeks less than half of the users performed 2 or more
weekly sessions. After 9 weeks around 20% continued to use the app through the 8 months
study period. Because this study was exploratory, we did not pre-specify a threshold for accept-
able adherence. This is important because studies have found that a home-based exercise
Table 6. Weekly ankle stability scores.
Week
Ankle stability
1
4.5
2
5.7
3
6.5
4
6.8
5
6.9
6
7.4
7
7.5
8
7.8
9
7.9
10
8.2
11
7.9
12
8.1
13
8.1
14
8.6
15
8.6
16
8.6
17
8.7
The mean scores for subjective ankle stability (0–10 points) for each week. 0 = Very unstable, 10 = Completely stable.
https://doi.org/10.1371/journal.pdig.0000221.t006
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
15 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
program with 24 exercise sessions could reduce the risk of recurrent sprains by 35% [12,31]. If
we consider 24 exercise sessions to be an acceptable adherence threshold (not considering
time per session or time intervals between completed sessions) only 15% of the participants in
the present study were adherent.
Despite the low adherence, the participants reported high satisfaction with the app although
they did not use it much. Almost all participants would recommend the app to others, which is
consistent with data from Vriend et al. [29]. Their app was also given high appraisal by its
users even though they only completed, on average, 3.3 exercise sessions out of the recom-
mended 24 in the app. One would think that they stopped exercising because they no longer
felt restricted by their ankle sprain. However, when asked about this in the present study (if
they felt recovered when they stopped using the app) the majority responded “No”. Partici-
pants stated in the interviews that they had lost motivation when the symptoms declined to a
level where they were able to manage daily tasks. From the clinical recovery data, we can see
that crutches are predominantly used in the first two weeks. So, even though 80% of the partic-
ipants after two weeks still suffered symptoms, it is likely that they could have started to work
and participate in leisure activities and thus feel it less necessary to continue using the app.
This may have contributed to why they stopped exercising with the app. The high satisfaction
with the app–despite limited use–could be related to social desirability bias, that is, patients
reporting what they expect would please the investigators.
The interviews revealed that participants in general expected a complete spontaneous
recovery from the ankle sprain regardless of their actions. That people with ankle sprain
believe the injury to be innocuous is anecdotally supported by several studies [1,4,10,32–34]
but to our knowledge, this is the first study that has interviewed people on this perception. The
perception that an ankle sprain is an innocuous injury may also be reflected in the reason for
seeking medical attention at the ED, as participants primarily went to the ED because they
were worried that the ankle had a fracture, not because they were worried about the conse-
quences of a sprained–not fractured–ankle. How the perception of LAS as being an innocuous
injury influences adherence to an app and an exercise program is largely unknown. But from
the joint display of the qualitative and quantitative data, it seems that the focus of their ER visit
was diagnostic and that when symptoms decreased, they lost motivation for exercise. This
likely resulted in low overall exercise adherence but with paradoxically high satisfaction across
the different app use items. The interviews revealed that several participants experienced the
starting level to be either too difficult or too easy. They found it discouraging if the app did not
match their expectations regarding exercise difficulty within a few completed sessions. The
app was designed with a fixed starting level and a hierarchical development of exercises. The
advantage of this design is that the progression can be matched to exercise guidelines and min-
imize the risk that users are presented with exercises that may cause them harm. The disadvan-
tage is that some participants may need to complete several sessions before they reach a
desired exercise difficulty, which may negatively impact adherence.
The app did provide the participants with an exercise solution they found easy to access
and the short programs were perceived as easy to include in the daily routines. Both time
restrictions and access to exercise opportunities have been found to be major barriers for phys-
ical activity among patients with musculoskeletal disorders [35]. It is interesting that despite
the app-based exercise program resolved these two major barriers and it was highly appraised,
it was not enough to substantially motivate our participants to exercise. Whether the low
adherence was primarily due to a general opinion that ankle sprains are an innocuous condi-
tion, and/or the app-based exercise program–especially the starting level–is currently
unknown. Further research is needed to evaluate how different recruitment methods, program
designs and conditions affect adherence for app-based exercise interventions.
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
16 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Future app optimizing
The app-based exercise program enables users to perform exercises wherever and whenever.
However, data from the interviews point out that usability of an exercise app may be more than
just being ever-present in your pocket. Time to complete sessions, exercise materials, the need
for changing into gym clothes or just getting down on the floor are all factors that may influence
and limit the usability. Some demands can be necessary for the exercise program to ensure cor-
rect exercise form; however, it may enhance adherence if users could customize their program to
fit not only their injury but also their exercise behavior. Giving users more control of their exer-
cise program may also be a solution for users who feel that the starting level is far from what they
want it to be. The exercise app seemed to have one or two sessions to match people´s expecta-
tions before they would quit the program. Since the app can make real-time changes dependent
on user feedback, an improvement could be to include initial questions on people’s functional
disabilities (e.g. ability to stand or walk), so that it may guide the app to a more motivational
starting difficulty. Furthermore, giving users the ability to skip difficultly levels would make the
app more adaptable to both user expectations and day to day changes in symptoms.
It seems that the explanatory exercise videos are important, and they seem to make people
feel secure in their exercise execution. The visual expression in the videos seems to influence
the integrity of the app. Most of our participants seemed to prefer what they called the “clinical
expression” used in this app with a bare room and a regular looking person, compared to
more fitness-focused app with highly trained athletes and fast paced music. They wanted a per-
son they could relate to. It is likely that different age groups, gender etc. prefer different video
expressions and a personalized video could enhance motivation to perform exercises with the
app as a partner.
With regards to the statistics in the app, we were surprised that participants did not con-
sider them important for their motivation. It was surprising because many thought the process
statistics presented after each completed session were a measure of their clinical recovery
despite the fact it only reflected the program process. This is, however, similar to the finding
from Liao et al. [36] who found that visual demonstration was perceived as the only important
motivational factor among 52 app design features including reminder and statistical functions.
A reason that the statistics are not found to be motivating could be that they are generic and
not user tailored. A patient-specific goal setting may enhance users work towards those goals
[37,38] and a scale like the Patient Specific Functional Scale [39] would be easy to include as an
app feature. Finally–and also related to goal setting—the app could potentially benefit from
some educational information initially about what an ankle sprain is, and how much rehabili-
tation exercise is needed for recovery. This should also include the importance of continuing
exercise when symptoms decrease so that the risk of recurrent sprains is decreased. A standard
user experience questionnaire could then be used for evaluation of any app changes made.
Strengths and limitations
The adherence data in this study do not rely on feedback from participants and are therefore
not affected by potential recall or reporting bias. This is a major study strength. Only exercises
that are registered in the app are recorded, however. One participant described that after learn-
ing several of the exercises, she had performed them without opening the app.
The app collects feedback on individual exercises, which is much more detailed than just
session collecting completion, which is a commonly used proxy measures for adherence [40].
From the data, it would be possible to identify possible exercises that participants experience
too difficult or pain provoking at a certain stage. A limitation, however, it that the app does
not monitor patients during the exercises and performance quality is not assessed.
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
17 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
In this study it was possible to elaborate the quantitative app-use data with qualitative data
from the interviews and suggest possible explanations. This would not have been possible if
only one methodological approach has been chosen. The mixed method approach in this
study is described with regards to “Good Reporting for a mixed method study” (GRAMMS)
[24] to increase transparency and the author group comprised of both quantitative and quali-
tative experts to obtain quality of each approach which is advocated for mixed method design
[24].
Conclusion
In this study, only few of the patients seen for an ankle sprain in the ED became active app
users after they received information about a free app-based rehabilitation exercise program.
Those who did, liked the app very much, but few completed enough exercise sessions to realis-
tically impact clinical recovery. The ankle sprain was generally considered an innocuous injury
that would spontaneously recover even though more than half the participants did not feel
fully recovered when they stopped exercising, and a third experienced a recurrent sprain. To
improve the care of these patients in the ED, we suggest that health-care personnel who asses
acute ankle sprains should be aware of their importance in informing patients about the risk of
prolonged symptoms and recurrent sprains, so that patients may have more realistic expecta-
tions on the clinical course. Recommendations from a health care professional in the ED to
use an ankle sprain exercise program seems to carry more weight than similar recommenda-
tions given elsewhere, making the ED setting interesting from an implementation point of
view.
Supporting information
S1 Table. Good Reporting of A Mixed Methods Study (GRAMMS) checklist.
(DOCX)
S2 Table. Exercise program.
(DOCX)
S1 Fig. Number of completed exercise sessions per participant.
(DOCX)
S2 Fig. Adherence by different grouping variables.
(DOCX)
S1 Text. Description of the exercise program.
(DOCX)
S2 Text. ICMJE disclosure forms for all authors.
(PDF)
S3 Text. Study protocol.
(PDF)
Author Contributions
Conceptualization: Kristian Thorborg, Jeanette Wassar Kirk, Thomas Bandholm.
Data curation: Jonas Bak, Mikkel Bek Clausen.
Formal analysis: Jonas Bak, Mikkel Bek Clausen, Jeanette Wassar Kirk, Thomas Bandholm.
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
18 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
Funding acquisition: Kristian Thorborg, Thomas Bandholm.
Investigation: Jonas Bak, Kristian Thorborg, Mikkel Bek Clausen, Finn Elkjær Johannsen,
Jeanette Wassar Kirk, Thomas Bandholm.
Methodology: Jonas Bak, Kristian Thorborg, Mikkel Bek Clausen, Finn Elkjær Johannsen,
Jeanette Wassar Kirk, Thomas Bandholm.
Project administration: Jonas Bak.
Resources: Kristian Thorborg, Mikkel Bek Clausen, Finn Elkjær Johannsen, Thomas
Bandholm.
Software: Mikkel Bek Clausen.
Supervision: Kristian Thorborg, Jeanette Wassar Kirk, Thomas Bandholm.
Validation: Jonas Bak, Jeanette Wassar Kirk.
Visualization: Jonas Bak, Kristian Thorborg, Jeanette Wassar Kirk, Thomas Bandholm.
Writing – original draft: Jonas Bak, Kristian Thorborg, Mikkel Bek Clausen, Jeanette Wassar
Kirk, Thomas Bandholm.
Writing – review & editing: Jonas Bak, Kristian Thorborg, Mikkel Bek Clausen, Finn Elkjær
Johannsen, Jeanette Wassar Kirk, Thomas Bandholm.
References
1. Gribble PA, Bleakley CM, Caulfield BM, Docherty CL, Fourchet F, Fong DT-P, et al. Evidence review
for the 2016 International Ankle Consortium consensus statement on the prevalence, impact and long-
term consequences of lateral ankle sprains. Br J Sports Med. 2016; 50: 1496–1505. https://doi.org/10.
1136/bjsports-2016-096189 PMID: 27259753
2. DS93: Luksation og distorsion af led og ligamenter i ankel og fod [Internet]. The Danish National Patient
Registry. 2017. Available from: https://www.esundhed.dk/Registre/Landspatientsregisteret
3. McGovern RP, Martin RL. Managing ankle ligament sprains and tears: current opinion. Open Access J
Sports Med. 2016; 7: 33–42. https://doi.org/10.2147/OAJSM.S72334 PMID: 27042147
4.
5.
6.
Feger MA, Glaviano NR, Donovan L, Hart JM, Saliba SA, Park JS, et al. Current Trends in the Manage-
ment of Lateral Ankle Sprain in the United States. Clin J Sport Med Off J Can Acad Sport Med. 2017;
27: 145–152. https://doi.org/10.1097/JSM.0000000000000321 PMID: 27347860
Larmer PJ, McNair PJ, Smythe L, Williams M. Ankle sprains: patient perceptions of function and perfor-
mance of physical tasks. A mixed methods approach. Disabil Rehabil. 2011; 33: 2299–2304. https://doi.
org/10.3109/09638288.2011.568668 PMID: 21470050
van Rijn RM, van Os AG, Bernsen RMD, Luijsterburg PA, Koes BW, Bierma-Zeinstra SMA. What is the
clinical course of acute ankle sprains? A systematic literature review. Am J Med. 2008; 121: 324–331.
e6. https://doi.org/10.1016/j.amjmed.2007.11.018 PMID: 18374692
7. Bleakley CM, O’Connor SR, Tully MA, Rocke LG, Macauley DC, Bradbury I, et al. Effect of accelerated
rehabilitation on function after ankle sprain: randomised controlled trial. BMJ. 10. 2010; 340: c1964.
https://doi.org/10.1136/bmj.c1964 PMID: 20457737
8. Doherty C, Bleakley C, Delahunt E, Holden S. Treatment and prevention of acute and recurrent ankle
sprain: an overview of systematic reviews with meta-analysis. Br J Sports Med. 2017; 51: 113–125.
https://doi.org/10.1136/bjsports-2016-096178 PMID: 28053200
9. Hupperets MDW, Verhagen EALM, Heymans MW, Bosmans JE, van Tulder MW, van Mechelen W.
Potential savings of a program to prevent ankle sprain recurrence: economic evaluation of a random-
ized controlled trial. Am J Sports Med. 2010; 38: 2194–2200. https://doi.org/10.1177/
0363546510373470 PMID: 20699429
10. Ardèvol J, Bolı´bar I, Belda V, Argilaga S. Treatment of complete rupture of the lateral ligaments of the
ankle: a randomized clinical trial comparing cast immobilization with functional treatment. Knee Surg
Sports Traumatol Arthrosc Off J ESSKA. 2002; 10: 371–377. https://doi.org/10.1007/s00167-002-0308-
9 PMID: 12444517
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
19 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
11. Calatayud J, Borreani S, Colado JC, Flandez J, Page P, Andersen LL. Exercise and ankle sprain inju-
ries: a comprehensive review. Phys Sportsmed. 2014; 42: 88–93. https://doi.org/10.3810/psm.2014.02.
2051 PMID: 24565825
12. Hupperets MDW, Verhagen EALM, van Mechelen W. Effect of unsupervised home based propriocep-
tive training on recurrences of ankle sprain: randomised controlled trial. BMJ. 9. 2009; 339: b2684.
https://doi.org/10.1136/bmj.b2684 PMID: 19589822
13. Verhagen E, Bolling C. Protecting the health of the @hlete: how online technology may aid our common
goal to prevent injury and illness in sport. Br J Sports Med. 2015; 49: 1174–1178. https://doi.org/10.
1136/bjsports-2014-094322 PMID: 25614537
14. Wong SJ, Robertson GA, Connor KL, Brady RR, Wood AM. Smartphone apps for orthopaedic sports
medicine—a smart move? BMC Sports Sci Med Rehabil. 2015; 7: 23. https://doi.org/10.1186/s13102-
015-0017-6 PMID: 26464800
15. Subhi Y, Bube SH, Rolskov Bojsen S, Skou Thomsen AS, Konge L. Expert Involvement and Adherence
to Medical Evidence in Medical Mobile Phone Apps: A Systematic Review. JMIR MHealth UHealth.
2015; 3: e79. https://doi.org/10.2196/mhealth.4169 PMID: 26215371
16.
van Mechelen DM, van Mechelen W, Verhagen EALM. Sports injury prevention in your pocket?! Pre-
vention apps assessed against the available scientific evidence: a review. Br J Sports Med. 2014; 48:
878–882. https://doi.org/10.1136/bjsports-2012-092136 PMID: 23511697
17. Bondaronek P, Alkhaldi G, Slee A, Hamilton FL, Murray E. Quality of Publicly Available Physical Activity
Apps: Review and Content Analysis. JMIR MHealth UHealth. 2018; 6: e53. https://doi.org/10.2196/
mhealth.9069 PMID: 29563080
18. Badawy SM, Kuhns LM. Texting and Mobile Phone App Interventions for Improving Adherence to Pre-
ventive Behavior in Adolescents: A Systematic Review. JMIR MHealth UHealth. 2017; 5: e50. https://
doi.org/10.2196/mhealth.6837 PMID: 28428157
19. Halcomb E, Hickman L. Mixed methods research. Nurs Stand R Coll Nurs. 2015; 29: 41–47. https://doi.
org/10.7748/ns.29.32.41.e8858 PMID: 25850508
20. Palinkas LA, Horwitz SM, Green CA, Wisdom JP, Duan N, Hoagwood K. Purposeful sampling for quali-
tative data collection and analysis in mixed method implementation research. Adm Policy Ment Health.
2015; 42: 533–544. https://doi.org/10.1007/s10488-013-0528-y PMID: 24193818
21. O’Cathain A, Murphy E, Nicholl J. Three techniques for integrating data in mixed methods studies. BMJ.
2010; 341: c4587. https://doi.org/10.1136/bmj.c4587 PMID: 20851841
22. Chan A-W, Tetzlaff JM, Altman DG, Laupacis A, Gøtzsche PC, Krlezˇa-Jerić K, et al. SPIRIT 2013 state-
ment: defining standard protocol items for clinical trials. Ann Intern Med. 2013; 158: 200–207. https://
doi.org/10.7326/0003-4819-158-3-201302050-00583 PMID: 23295957
23. Bandholm T, Christensen R, Thorborg K, Treweek S, Henriksen M. Preparing for what the reporting
checklists will not tell you: the PREPARE Trial guide for planning clinical research to avoid research
waste. Br J Sports Med. 2017; 51: 1494–1501. https://doi.org/10.1136/bjsports-2017-097527 PMID:
28882839
24. O’Cathain A, Murphy E, Nicholl J. The quality of mixed methods studies in health services research. J
Health Serv Res Policy. 2008; 13: 92–98. https://doi.org/10.1258/jhsrp.2007.007074 PMID: 18416914
25. Castleberry A, Nolen A. Thematic analysis of qualitative research data: Is it as easy as it sounds? Curr
Pharm Teach Learn. 2018; 10: 807–815. https://doi.org/10.1016/j.cptl.2018.03.019 PMID: 30025784
26.
Julious SA. Issues with number needed to treat. Stat Med. 2005; 24: 3233–3235. https://doi.org/10.
1002/sim.2150 PMID: 16189812
27. Guetterman TC, Fetters MD, Creswell JW. Integrating Quantitative and Qualitative Results in Health
Science Mixed Methods Research Through Joint Displays. Ann Fam Med. 2015; 13: 554–561. https://
doi.org/10.1370/afm.1865 PMID: 26553895
28. DS93: Luksation og distorsion af led og ligamenter i ankel og fod. Search period: July 3rd, 2018 to April
3rd, 2019. [Internet]. The Danish National Patient Registry. 2017. Available from: https://www.
esundhed.dk/Registre/Landspatientsregisteret
29. Vriend I, Coehoorn I, Verhagen E. Implementation of an app-based neuromuscular training programme
to prevent ankle sprains: a process evaluation using the RE-AIM Framework. Br J Sports Med. 2015;
49: 484–488. https://doi.org/10.1136/bjsports-2013-092896 PMID: 24470587
30.
Zheng P, Leff G, Ith MA, Smuck M. App Development for Therapeutic Exercise. PM R. 2017; 9: S116–
S117. https://doi.org/10.1016/j.pmrj.2017.04.004 PMID: 28527496
31. Van Reijen M, Vriend I, van Mechelen W, Verhagen EA. Preventing recurrent ankle sprains: Is the use
of an App more cost-effective than a printed Booklet? Results of a RCT. Scand J Med Sci Sports. 2018;
28: 641–648. https://doi.org/10.1111/sms.12915 PMID: 28543566
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
20 / 21
PLOS DIGITAL HEALTH App-based exercise rehabilitation and ankle sprains
32. Bleakley C. Supervised physiotherapy for mild or moderate ankle sprain. BMJ. 2016; 355: i5984.
https://doi.org/10.1136/bmj.i5984 PMID: 27852567
33. Brison RJ, Day AG, Pelland L, Pickett W, Johnson AP, Aiken A, et al. Effect of early supervised physio-
therapy on recovery from acute ankle sprain: randomised controlled trial. BMJ. 2016; 355: i5650.
https://doi.org/10.1136/bmj.i5650 PMID: 27852621
34. Malliaropoulos N, Ntessalen M, Papacostas E, Longo UG, Maffulli N. Reinjury after acute lateral ankle
sprains in elite track and field athletes. Am J Sports Med. 2009; 37: 1755–1761. https://doi.org/10.1177/
0363546509338107 PMID: 19617530
35. McPhail SM, Schippers M, Marshall AL, Waite M, Kuipers P. Perceived barriers and facilitators to
increasing physical activity among people with musculoskeletal disorders: a qualitative investigation to
inform intervention development. Clin Interv Aging. 2014; 9: 2113–2122. https://doi.org/10.2147/CIA.
S72731 PMID: 25584023
36.
Liao G-Y, Chien Y-T, Chen Y-J, Hsiung H-F, Chen H-J, Hsieh M-H, et al. What to Build for Middle-Agers
to Come? Attractive and Necessary Functions of Exercise-Promotion Mobile Phone Apps: A Cross-
Sectional Study. JMIR MHealth UHealth. 2017; 5: e65. https://doi.org/10.2196/mhealth.6233 PMID:
28546140
37. Stevens A, Beurskens A, Ko¨ ke A, van der Weijden T. The use of patient-specific measurement instru-
ments in the process of goal-setting: a systematic review of available instruments and their feasibility.
Clin Rehabil. 2013; 27: 1005–1019. https://doi.org/10.1177/0269215513490178 PMID: 23881336
38. Wade DT. Goal setting in rehabilitation: an overview of what, why and how. Clin Rehabil. 2009; 23:
291–295. https://doi.org/10.1177/0269215509103551 PMID: 19293289
39. Horn KK, Jennings S, Richardson G, Vliet DV, Hefford C, Abbott JH. The patient-specific functional
scale: psychometrics, clinimetrics, and application as a clinical outcome measure. J Orthop Sports
Phys Ther. 2012; 42: 30–42. https://doi.org/10.2519/jospt.2012.3727 PMID: 22031594
40. Picorelli AMA, Pereira LSM, Pereira DS, Felı´cio D, Sherrington C. Adherence to exercise programs for
older people is influenced by program characteristics and personal factors: a systematic review. J Phy-
siother. 2014; 60: 151–156. https://doi.org/10.1016/j.jphys.2014.06.012 PMID: 25092418
PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0000221 May 15, 2023
21 / 21
PLOS DIGITAL HEALTH
| null |
10.1103_physrevd.107.063005.pdf
| null | null |
PHYSICAL REVIEW D 107, 063005 (2023)
Flavor changing interactions confronted with meson
mixing and hadron colliders
A. E. Cárcamo Hernández ,1,2,3,* L. Duarte ,4,†
A. S. de Jesus ,4,5,‡
S. Kovalenko,2,3,6,§ F. S. Queiroz,3,4,5,∥
C. Siqueira ,7,¶ Y. M. Oviedo-Torres ,8,** and Y. Villamizar
4,5,††
1Universidad T´ecnica Federico Santa María, Casilla 110-V, Valparaiso, Chile
2Centro Científico Tecnológico de Valparaíso-CCTVal, Universidad T´ecnica Federico Santa María,
Casilla 110-V, Valparaíso, Chile
3Millennium Institute for Subatomic Physics at the High-Energy Frontier (SAPHIR) of ANID,
Fernández Concha 700, Santiago, Chile
4International Institute of Physics, Universidade Federal do Rio Grande do Norte,
Campus Universitario, Lagoa Nova, Natal-RN 59078-970, Brazil
5Departamento de Física, Universidade Federal do Rio Grande do Norte,
59078-970, Natal, Rio Grande do Norte, Brazil
6Departamento de Ciencias Físicas, Universidad Andres Bello, Sazi´e 2212, Piso 7, Santiago, Chile
7Instituto de Física de São Carlos, Universidade de São Paulo,
Av. Trabalhador São-carlense 400, São Carlos-SP, 13566-590, Brazil
8Departamento de Fisica, Universidade Federal da Paraiba,
Caixa Postal 5008, 58051-970, Joao Pessoa, Paraíba, Brazil
(Received 19 September 2022; accepted 17 February 2023; published 10 March 2023)
We have witnessed some flavor anomalies appearing in the past years, and explanations based on
extended gauge sectors are among the most popular solutions. These beyond the Standard Model (SM)
theories often assume flavor-changing interactions mediated by new vector bosons. Still, at the same
time, they could yield deviations from the SM in the K0 − ¯K0, D0 − ¯D0, B0
s meson
systems. Using up-to-date data on the mass difference of these meson systems, we derive lower mass
bounds on vector mediators for two different parametrizations of the quark mixing matrices. Focusing on a
well-motivated model based on the fundamental representation of the weak SU(3) gauge group, we put our
findings into perspective with current and future hadron colliders to conclude that meson mass systems can
give rise to bounds much more stringent than those from high-energy colliders and that recent new physics
interpretations of the b → s and RðD(cid:2)Þ anomalies are disfavored.
d, and B0
s − ¯B 0
− ¯B 0
d
DOI: 10.1103/PhysRevD.107.063005
I. INTRODUCTION
Since flavor-changing neutral current (FCNC) processes
are forbidden at tree level in the Standard Model (SM),
they are very sensitive to new physics. For this reason,
‡
*[email protected]
†
[email protected]
[email protected]
§[email protected]
∥
[email protected]
¶[email protected]
**[email protected]
††
Corresponding author.
[email protected]
Published by the American Physical Society under the terms of
license.
the Creative Commons Attribution 4.0 International
Further distribution of this work must maintain attribution to
the author(s) and the published article’s title, journal citation,
and DOI. Funded by SCOAP3.
meson–antimeson mixing that belongs to the class of flavor-
changing neutral current processes are great laboratories for
flavor-changing interactions. Meson systems are key to our
understanding of the fundamental interactions and contin-
uously give rise to important results such as the recent
measurement of mixing and CP violation in neutral charm
mesons collected by the LHCb experiment [1].
FCNCs have historically been important to the develop-
ment of the SM. From the considerations of FCNC, the
charm quark was predicted to accommodate the data
that ruled out
larger FCNC effects [2]. Analyzing the
neutral kaon meson system, the value of charm mass was
estimated [3]. Charged kaon decays revealed that weak
interactions do not conserve parity and charge operators.
Moreover, the KL decay into pions has shown that CP is
not preserved [4]. The SM with two fermion generations
could not reproduce this decay because CP-violating
interactions of quarks necessarily involve complex cou-
plings. Those complex couplings, if introduced in a 2 × 2
2470-0010=2023=107(6)=063005(11)
063005-1
Published by the American Physical Society
A. E. CÁRCAMO HERNÁNDEZ et al.
PHYS. REV. D 107, 063005 (2023)
on the mass difference of these mesons to constrain any
new physics contribution to the mass differences. These
mesons are comprised of different quark flavors and,
consequently, are sensitive to different entries of
the
CKM matrix, Therefore, the new physics reach for meson
mixing systems relies on the parametrization used for the
quark mixing matrices. In summary, a robust assessment of
the new physics potential of flavor probes requires control
over the systematic errors [11–13].
FIG. 1. An illustration of how flavor eigenstates of mesons
lead to mass eigenstates of mesons with different masses. The
mass difference in such systems of mixed mesons is at the core of
our study.
mixing matrix, are eliminated after rotation, leaving, in
the end, a real 2 × 2 Cabibbo matrix. Kobayashi and
Maskawa concluded in 1973 that such complex terms
would survive in the quark mixing matrix if there were at
least three generations. This fact was ignored for quite
some time until the discovery of the bottom quark in 1977
by Ledermann [5], which later hinted at the existence of the
top quark [6,7]. Therefore, it is clear that mesons have
played a crucial role in our understanding of fundamental
interactions. As they are made up of one quark and one
antimatter quark. The antimatter state of a given meson is
also comprised of a quark and one antimatter quark. For
instance, the D0 meson consists of a charm quark and an up
antiquark, whereas its antiparticle, the ¯D0, is made of a
charm antiquark and an up quark. In the quantum physics
world, the meson D0 particle can be itself and its anti-
particle at once, leading to a quantum superposition of
states, say D1 and D2, each with their own mass and their
decay width Γ1 and Γ2 (see Fig. 1). This superposition
allows a continuous oscillation between the D0 particle and
its antiparticle. In other words, the Hamiltonian is not
diagonal in the flavor basis, and thus flavor changing
interactions are present. The mass difference, mD1 − mD2,
determines the frequency of oscillations, which is measured
[8–10] and reported in terms of the dimensionless param-
eter x ¼ ðmD1 − mD2Þ=Γ, where Γ is the average width,
ðΓ1 þ Γ2Þ=2.
d
− ¯B0
Such an oscillation pattern is present in four well-known
meson systems, namely, K0 − ¯K0, D0 − ¯D0, B0
d, and
s − ¯B0
B0
s. The SM FCNC occurs at a one-loop level via a W
boson exchange in a box diagram, involving the Cabibbo-
Kobayashi-Maskawa (CKM) matrix; precisely for that
reason, any new physics-inducing flavor-changing inter-
actions are tightly constrained by the mesons systems
aforementioned. The relevant quantity for our reasoning
is the mass difference between these mesons, where an
excellent agreement between theory and measurement is
found. In other words, one can use precise measurements
In our work, we focus on the FCNC effects stemming
from neutral vector bosons. A wealth of Abelian and non-
Abelian extended gauge symmetries predict the existence
of extra neutral gauge bosons [14]. One can parametrize
these new physics contributions in terms of gauge cou-
plings and the mediator mass [15], but we will concentrate
our phenomenology on vector bosons arising from the
SUð3Þ
N gauge group, referred to as
c
3-3-1 models [16–20] because models based on this gauge
symmetry have been considered as a plausible explanation
to the b → s and RðD(cid:2)Þ anomalies [21–26].1 FCNC studies
in the context of 3-3-1 models have been carried out in the
past [25,29–49], but our work differs from previous studies
for the following reasons:
⊗ SUð3Þ
⊗ Uð1Þ
L
(i) We take into account
the four
relevant meson
systems, including updated measurements;
(ii) We consider two different parametrizations to assess
the impact of systematic errors;
(iii) As the SM prediction agrees well with the data, we
enforce the new physics contribution to be within the
reported experimental error bar;
(iv) We put our results into perspective with future
hadron colliders; and
(v) We investigate whether recent proposals based on
the 3-3-1 symmetry are consistent with meson
mixings and collider bounds.
Our goal is to find lower mass bounds on the vector
mediator, a Z0, which mediates flavor-changing inter-
actions. Consequently, our findings are relevant to 3-3-1
constructions that feature a similar neutral current with SM
quarks [50–66].
Our work is structured as follows: in Sec. II we revise the
key ingredients of the 3-3-1 model under study; in Sec. III
we derive the 3-3-1 contribution to the mass difference of
these mesons; in Sec. IV we discuss the current and future
hadron collider bounds; we draw our conclusions in Sec. V.
II. THE MODEL
Our FCNC investigation is dedicated to models that
are based on the SUð3Þ
⊗ SUð3Þ
N symmetry,
L
c
which promotes the SM SUð2Þ
L gauge group to a SUð3Þ
L
one. There are several ways to arrange fermions in a
SUð3Þ
L triplet, and these multiple possibilities give rise to
⊗ Uð1Þ
1See other flavor studies [27,28].
063005-2
FLAVOR CHANGING INTERACTIONS CONFRONTED WITH …
PHYS. REV. D 107, 063005 (2023)
different 3-3-1 models [18,33,34,67–71]. In this work, we
will focus on two of the most popular models based on
the 3-3-1 symmetry, namely, the 3-3-1 model with right-
handed neutrinos (RHN) and the 3-3-1 model with heavy
neutral fermion (LHN). These two particular versions of
the 3-3-1 symmetry can accommodate dark matter and
neutrino masses, which are the most convincing evidence
for physics beyond the SM. We will focus on the 3-3-1
model with right-handed neutrinos, but we emphasize that
those two models feature the same neutral current involving
the Z0 gauge boson and SM quarks. Therefore, our
conclusions are valid for both models. That said, under
the SUð3Þ
⊗ SUð3Þ
N gauge group the lepton
L
c
sector is arranged as
⊗ Uð1Þ
¼
fa
L
0
B
@
νa
l
ea
l
ðνc
R
Þa
1
C
A ∼ ð1; 3; −1=3Þ;
ea
R
∼ ð1; 1; −1Þ;
ð1Þ
where a ¼ 1, 2, 3, indicate the three fermion generations.
Regarding the hadronic sector, gauge anomaly cancella-
tion requires that the quark generations transform differ-
ently under the SUð3Þ
L group. The most simple way to
accomplish that without
invoking several exotic new
fermions is by assuming that the first generation transforms
as triplets under SUð3Þ
L, whereas the second and third ones
as antitriplets as follows:
1
0
Q3L ¼
B
@
∼ ð3; 3; 1=3Þ;
C
A
L
QiL ¼
B
@
C
A
∼ ð3; ¯3; 0Þ;
L
u3
d3
u0
3
di
−ui
d0
i
u3R ∼ ð3; 1; 2=3Þ; d3R ∼ ð3; 1; −1=3Þ; u0
3R
∼ ð3; 1; 2=3Þ;
0
1
uiR ∼ ð3; 1; 2=3Þ; diR ∼ ð3; 1; −1=3Þ; d0
iR
∼ ð3; 1; −1=3Þ;
ð2Þ
1;2Þ ¼ −1=3.
where i ¼ 1, 2, with q0 being heavy exotic quarks with
3Þ ¼ 2=3 and Qðd0
electric charges Qðu0
We highlight that in the 3-3-1 LHN, a new heavy neutral
lepton Na
L replaces the left-handed neutrino in the lepton
∼
triplet. In addition, a right-handed neutral fermion Na
R
ð1; 1; 0Þ is introduced, which transforms as a singlet under
SUð3Þ
L. The quark sector remains the same though. Hence,
as we stressed before, our reasoning for flavor-changing
interactions involving quarks applies to both 3-3-1 models.
Fermion masses are generated through the spontaneous
symmetry-breaking mechanism governed by three scalar
triplets. From a top-down approach, the scalar triplet χ
acquires a vacuum expectation value (vev) in the scale of
the TeVs with,
hχi ¼
1
C
A;
0
B
@
0
0
vχ
ð3Þ
⊗ Uð1Þ
breaking SUð3Þ
Y,
L
thus generating masses for the additional gauge bosons
and new fermions, namely,
the exotic quarks via the
Yukawa Lagrangian,
N down to SUð2Þ
L
⊗ Uð1Þ
Lχ
Yuk
¼ λ1
¯Q1Lu0
1R
χ þ λ2ij
¯QiLd0
jR
χ(cid:2) þ H:c:;
ð4Þ
where χ ∼ ð1; 3; −1=3Þ.
Then the SUð2Þ ⊗ Uð1Þ
when two scalar triplets ρ, η get a vev as follows:
1
0
0
1
Y breaks into electromagnetism
hρi ¼
B
@
C
A;
hηi ¼
B
@
C
A;
ð5Þ
0
vρ
0
vη
0
0
yielding masses for the SM quarks and charged lepton
masses through
L
Yuk
¼ λ1a
þ G0
ab
¯Q1LdaRρ þ λ2ia
¯fa
ρ þ λ3a
Leb
R
¯QiLuaRρ(cid:2) þ Gab
¯Q1LuaRηþ λ4ia
L
ðfb
L
¯fa
Þcρ(cid:2)
¯QiLdaRη(cid:2) þ H:c:
ð6Þ
Notice that the scalar triplets transform as ρ ∼ ð1; 3; 2=3Þ
and η ∼ ð1; 3; −1=3Þ. Furthermore, the third term in Eq. (6)
gives rise to two mass degenerate neutrinos and a massless
one. It is well known that this neutrino mass pattern cannot
reproduce the three mass differences observed in the
neutrino oscillation data [72–74]. However, one can nicely
solve this problem by adding a scalar sextet and realizing a
type II seesaw mechanism, or adding three right-handed
Majorana neutrinos to incorporate an inverse or linear
seesaw [75,76]. We emphasize that either way neutrino
masses are generated, our reasoning concerning FCNC is
left unchanged.
Besides the usual bilinear and quartic terms in the scalar
potential, these scalars give rise to the term − fffiffi
p ϵijkηiρjχk,
2
where f is in principle a free parameter which has energy
dimension. The main energy scale in our work is the energy
scale at which the 3-3-1 symmetry is broken down to the
SM one. Hence, it is natural to assume that f ∼ vχ.
We highlight this fact, because there is often the question
of the importance of FCNC mediated by scalar fields in
3-3-1 constructions. However, if f ∼ vχ, the new scalars in
the model are heavier than the Z0, and consequently the
FCNC effects induced by them are relatively smaller than
those rising from the Z0. For concreteness, taking f ¼ vχ,
the scalars that induce FCNC have masses larger than vχ.
In contrast, the mass of the Z0 boson would be 0.45vχ.
063005-3
A. E. CÁRCAMO HERNÁNDEZ et al.
PHYS. REV. D 107, 063005 (2023)
However, one can assume different values for the f
parameter, allowing scalars to be lighter than the Z0, as
has been explored in [49]. Notice that even if they are
indeed lighter than the Z0, this does not warrant a larger
FCNC effect because the magnitude of the FCNC induced
by the scalar fields will be subject to arbitrary choices of the
couplings in the scalar potential, see Appendix B of [35].
Thus,
fields usually give rise to
relatively meager FCNC effects. Moreover, FCNC arising
from scalar fields are necessarily less predictive than the
ones stemming from gauge interactions mediated by the Z0
field. Albeit, one can in principle overlook all these facts
and tune the couplings in the scalar potential in such a way
as to enhance the FCNC effects coming from scalar fields
and potentially make them the dominant contribution.
in summary, scalar
Now that we have reviewed the key aspects of the model,
we will concentrate on the main source of flavor-changing
neutral current, namely, the Z0 gauge boson.
III. FCNC IN THE 3-3-1
Flavor-changing neutral current is a common feature in
3-3-1 models because gauge anomaly cancellation requires
one of the fermion generations to transform differently than
the others. This requirement naturally induces a flavor-
changing neutral current once one rotates the quark flavors
and introduces the CKM matrix. In other words, the Z0 does
not have universal couplings to quarks, and thus flavor
changing interactions arise. This is key because the Z boson
does not induce flavor changing interactions in the SM,
conversely to the charged current mediated by the W boson.
Flavor-changing interactions in the SM model occurs
through the charged current. Thus, flavor changing inter-
actions induced by a W0 would be swamped by numerous
W boson interactions. Therefore, it is wise to investigate
flavor-changing interactions mediated by a neutral gauge
boson, as they are not masked by a large SM effect. We
remark that in the 3-3-1 models, we have additional neutral
gauge bosons, namely, the W0(cid:3)
, U0, and U0†. Nevertheless,
they do not generate FCNC. Thus, we focus on the Z0 field.
As we have explained earlier, mesons mass systems are
great
laboratories to probe such flavor-changing inter-
actions because Z0 fields can induce sizable flavor tran-
sitions, impacting the mass difference of meson systems
[77–79], see Fig. 2. We would like to stress again that
FCNC seeded by scalar fields are typically suppressed
than the Z0
compared to those generated by Z0 bosons because these
scalars are typically heavier
field; see
Refs. [35,51] and references therein. In fact, it has been
shown in [35] that two neutral scalars can induce sizable
FCNC, but their masses go as m ∼ vχ, rendering them
relatively heavier than the Z0. Besides, their contribution to
FCNC adds an extra systematic effect to the 3-3-1 prediction,
which are the Yukawa couplings and the couplings in the
scalar potential. Therefore, there is no predictivity regarding
neutral scalar contributions to FCNC. Anyway, this aspect
has been explored in [49]. Lastly, the scalars in the 3-3-1
models do not offer clean collider signals and their couplings
to SM fermions are proportional
to Yukawa couplings,
which result in suppressed production rates at colliders.
Consequently, the interplay between FCNC and collider
physics is lost. Albeit, in principle, one can certainly fine-
tune the couplings in the scalar potential and generate a
scalar lighter than the Z0 boson making the reasoning in [49]
valid, but thus far, this has not been explicitly proven. For all
these reasons, we focus on the Z0 field.
In this way, after developing the covariant derivative, we
find the following currents,
(cid:3)
u ¼ g
LZ0
2CW
− g
2CW
(cid:3)
LZ0
d
¼ g
2CW
− g
2CW
(cid:4)
½ ¯u3Lγμu3L(cid:4)Z0
μ;
½ ¯uiLγμuiL(cid:4)Z0
μ
(cid:4)
ð3 − 4S2
Þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p
W
3 − 4S2
3
W
(cid:3)
6ð1 − S2
Þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p
W
3 − 4S2
3
W
(cid:4)
½ ¯diLγμdiL(cid:4)Z0
(cid:4)
½ ¯d3Lγμd3L(cid:4)Z0
μ;
ð3 − 4S2
Þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p
W
3 − 4S2
3
W
(cid:3)
6ð1 − S2
Þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p
W
3 − 4S2
3
W
μ
ð7Þ
ð8Þ
with i ¼ 1, 2,
indicating the generation indices, and
CW ≡ cos θW, SW ≡ sin θW, with θW being the Weinberg
angle. Note that Eqs. (7) and (8) are in the mass eigenstate
basis, and once we rotate to the flavor basis, FCNC arises.
The mass eigenstate and flavor bases are connected as
follows:
0
1
0
1
0
1
1
0
B
@
u
c
t
C
A
¼ VU
L;R
B
@
L;R
B
@
C
A
;
L;R
C
A
¼ VD
L;R
B
@
L;R
d
s
b
u0
c0
t0
C
A;
d0
s0
b0
FIG. 2. Feynman diagrams illustrating how the Z0 gauge boson changes the mass difference of the four meson systems under
investigation, namely, K0 − ¯K0, D0 − ¯D0, B0
− ¯B0d , and B0
s − ¯B0s .
d
063005-4
FLAVOR CHANGING INTERACTIONS CONFRONTED WITH …
PHYS. REV. D 107, 063005 (2023)
where VU
VCKM
L;R and VD
L VD
¼ VU†
L , known to be [80]
L;R are 3 × 3 unitary matrices, which determine the Cabibbo-Kobayashi-Maskawa (CKM) matrix
0
B
@
VCKM
¼
0.97435 (cid:3) 0.00016 0.22500 (cid:3) 0.00067 0.00369 (cid:3) 0.00011
0.22486 (cid:3) 0.00067 0.97349 (cid:3) 0.00016
1
C
A:
ð9Þ
0.00857þ0.00020
−0.00018
0.04110þ0.00083
−0.00072
0.04182þ0.00085
−0.000074
0.999118þ0.000031
−0.000036
After rotation, we get the tree level Z0 mediated neutral
current interactions,
LK0− ¯K0
Z0eff
LD0− ¯D0
Z0eff
− ¯B0
d
LB0
d
Z0eff
LB0
s − ¯B0
Z0eff
s
¼ G0 M2
Z
M2
Z0
¼ G0 M2
Z
M2
Z0
¼ G0 M2
Z
M2
Z0
¼ G0 M2
Z
M2
Z0
jðVD
L
Þ(cid:2)
31ðVD
L
Þ32j2j ¯d0
1L
γμd0
2L
j2;
jðVU
L
Þ(cid:2)
31ðVU
L
Þ32j2j ¯u0
1L
γμu0
2L
j2;
jðVD
L
Þ(cid:2)
31ðVD
L
Þ33j2j ¯d0
1L
γμd0
3L
j2;
jðVD
L
Þ(cid:2)
32ðVD
L
Þ33j2j ¯d0
2L
γμd0
3L
j2;
jðVU
L
jðVD
L
ðΔmKÞ
ðΔmDÞ
and, consequently [81–83],
Z0 ¼ G0 M2
Z
M2
Z0
Z0 ¼ G0 M2
Z
M2
Z0
Z0 ¼ G0 M2
Z
M2
Z0
Z0 ¼ G0 M2
Z
M2
Z0
ffiffi
p
2
3−4S2
W
ðΔmBd
ðΔmBs
jðVD
L
jðVD
L
GFC4
W
Þ
Þ
Þ(cid:2)
31ðVD
L
Þ(cid:2)
31ðVU
L
Þ(cid:2)
31ðVD
L
Þ(cid:2)
32ðVD
L
Þ32j2f2
KBKηKmK;
Þ32j2f2
DBDηDmD;
Þ33j2f2
Bd
BBd
ηBdmBd;
Þ33j2f2
Bs
BBs
ηBsmBs;
ð10Þ
where G0 ¼ 4
, with GF being the Fermi constant,
BK, BD, BB the bag parameters, fK, fD, fB the decay
constants, and ηK, ηD, ηB the QCD leading order correction
obtained in [25,77–79,84–86], and mK, mD, mB the masses
of the mesons. In Table I we summarize the values of these
parameters.
Our reasoning to constrain new physics contributions to
the mass mixing systems goes as follows:
(i) The experimental mass difference of the K0 − ¯K0
system is given by ðΔmKÞ
(ii) The SM prediction ðΔmKÞ
exp,
SM (see Table I) has good
agreement with the experimental, but errors are not
included in the SM prediction. We find different
values for the SM contribution in the literature,
of
þ
(iii) Therefore,
ðΔmKÞ
exp as done in previous works
[38,93–97], we enforce the Z0 contribution to be
smaller than the statistical error bar Table I. In this
instead
Z0 < ðΔmKÞ
ðΔmKÞ
imposing
SM
063005-5
way, our conclusions are less sensitive to theoretical
uncertainties and are driven by experimental mea-
surements.
(iv) We follow the same strategy for all four meson
systems.
(v) In summary, we impose,
Z0 < 0.006 × 10−12 MeV;
ðΔmKÞ
Z0 < 2.69 × 10−12 MeV;
ðΔmDÞ
Z0 < 0.013 × 10−10 MeV;
Þ
ðΔmBd
Z0 < 0.0013 × 10−8 MeV:
Þ
ðΔmBs
ð11Þ
We remind the reader that the Z0 boson mediates FCNC
at tree level through Eq. (10) and for this reason, we will
be able to severely constrain the mass of this particle. An
advantage of working in the scope of a 3-3-1 model is
that the Z0 boson couples to SM fields proportional to the
TABLE I. Meson masses [10,87–92] and the values of the bag
parameters [80,92].
Input parameters
ðΔmKÞ
¼ ð3.484 (cid:3) 0.006Þ × 10−12 MeV
exp
¼ 3.483 × 10−12 MeV
ðΔmKÞ
SM
mK ¼ ð497.611 (cid:3) 0.013Þ MeV
p
ffiffiffiffiffiffi
BK
fK ¼ 131 MeV
ηK ¼ 0.57
ðΔmDÞ
exp
−2.8962 Þ × 10−12 MeV
¼ 10−14 MeV
mD ¼ ð1865 (cid:3) 0.005Þ MeV
¼ ð6.25316þ2.69873
ðΔmDÞ
ffiffiffiffiffiffi
p
BD
SM
fD ¼ 187 MeV
ηD ¼ 0.57
ðΔmBd
Þ
SM
ðΔmBd
¼ ð3.334 (cid:3) 0.013Þ × 10−10 MeV
Þ
exp
¼ ð3.653 (cid:3) 0.037 (cid:3) 0.019Þ × 10−10 MeV
mBd
p
¼ ð5279.65 (cid:3) 0.12Þ MeV
ffiffiffiffiffiffiffiffi
fBd ¼ 210.6 MeV
BBd
ηBd
¼ 0.55
ðΔmBs
Þ
SM
ðΔmBs
¼ ð1.1683 (cid:3) 0.0013Þ × 10−8 MeV
Þ
exp
¼ ð1.1577 (cid:3) 0.022 (cid:3) 0.051Þ × 10−8 MeV
mBs
p
¼ ð5366.9 (cid:3) 0.12Þ MeV
ffiffiffiffiffiffiffi
BBs
fBs
ηBs
¼ 256.1 MeV
¼ 0.55
A. E. CÁRCAMO HERNÁNDEZ et al.
PHYS. REV. D 107, 063005 (2023)
SUð2Þ
the mixing matrices and the Z0 mass.
L gauge coupling. The only unknown quantities are
We will assume two different parametrizations of the
mixing matrices that yield significant changes in the new
physics contribution to the mass difference systems. In this
way, we can assess the impact of such parametrizations. We
adopt parametrization 1,
VD
L
¼ VD
R
¼
0
B
@
0.972
0.45
0.1
1
C
A
0.5
0.46
1.00 0.88
1.01
0.1
ð12Þ
and
VU
L
¼ VU
R
0
B
@
¼
1.18622007 −0.22070355 −0.09032872
1.17174168 −0.01837301
−0.34446205
1.04637372
−0.23647983 −0.87899906
1
C
A;
and parametrization 2,
VD
L
¼ VD
R
¼
0
B
@
0.972
0.45
0.1
1
C
A
0.46
0.5
0.88
1.00
0.0001 1.01
ð13Þ
and
VU
L
¼ VU
R
0
B
@
¼
1.19772759 −0.17792992 −0.1412471
0.11162792
0.37384218
1.06253529
0.95629613
−0.21612235 −0.80332999
1
C
A:
L and Vd
Knowing the entries of the up-quark and down-quark
L, we determine the Z0 con-
mixing matrices Vu
tribution to the mass difference of the meson systems and
consequently place a lower mass bound. We adopt these
parametrizations because they yield very strong and very
conservative 3-3-1 contributions to the FCNC processes,
respectively, while keeping the CKM matrix in agreement
with the data. With this information at hand, we use
Eq. (10) combined with Eq. (11) to plot our findings in
Figs. 3–6. Using these parametrizations, one can assess the
systematic uncertainty embedded in FCNC studies. In other
words, FCNC alone is not robust enough.
Before discussing our results, it is important to put them
into context with current and future collider bounds. To do
so, we address those limits below.
IV. DILEPTON RESONANCE SEARCHES
AT THE LHC
Z0 gauge bosons are often targets of experimental
searches going from low to the multi-TeV mass range
[98–101]. In the TeV range, which is the focus of our study,
Z0 gauge bosons that feature sizable couplings to fermions
can leave a clear signature at the LHC in the form of dijet
and dilepton events. In the 3-3-1 model, the Z0 has similar
FIG. 3. The Z0 contribution to the ðΔmKÞ
Z0 as a function of it mass [see Eq. (10)] for the parametrization 1, Eq. (12) (blue solid curve),
and parametrization 2, Eq. (13). The silver region corresponds to the FCNC exclusion region. We overlaid current and projected
colliders bounds. Note that parametrization 2 is not shown in the plot because the lower bound of the Z0 boson mass in both
parametrizations is substantially different. See text for details.
063005-6
FLAVOR CHANGING INTERACTIONS CONFRONTED WITH …
PHYS. REV. D 107, 063005 (2023)
FIG. 4. The blue curves correspond to Z0 contribution to the ðΔmDÞ
Z0 as a function of its mass [see Eq. (10)], for parametrization 1,
Eq. (12), and parametrization 2, Eq. (13). The silver region corresponds to the FCNC exclusion region. The lower mass bound for
parametrization 2 is mZ0 > 256 TeV. We overlaid current and projected colliders’ bounds. See text for details.
FIG. 5. The solid blue line corresponds to Z0 contribution to the ðΔmBs
Z0 as a function of it mass [see Eq. (10)], for parametrization 1,
Eq. (12), and parametrization 2, Eq. (13). The silver region corresponds to the FCNC exclusion region. The lower mass bound for
parametrization 1 is mZ0 > 154 TeV. We overlaid current and projected colliders’ bounds. See text for details.
Þ
couplings to quarks and leptons. As dilepton events have
relatively good signal efficiencies and acceptance and a
well-controlled background originating primarily from
Drell-Yann processes [100,102,103],
tighter constraints
on the Z0 mass are found compared to dijet events.
There have been experimental searches for Z0 gauge bosons
belonging to the 3-3-1 symmetry in the past [104,105]. The
most recent analysis taking advantage of the full dataset
from LHC was carried out in [106]. We consider the most
conservative bounds, which is the third benchmark scenario
presented in Table IV of [106]. The LHC bound was
based on an integrated luminosity of L ¼ 139 fb−1 with
p
ffiffiffi
¼ 13 TeV, whereas for the high-luminosity LHC setup
s
¼ 14 TeV. For the high-energy
luminosity was adopted, but using
L ¼ 3000 fb−1 with
LHC the latter
p
ffiffiffi
s
¼ 27 TeV. In summary, we used
ffiffiffi
s
p
063005-7
A. E. CÁRCAMO HERNÁNDEZ et al.
PHYS. REV. D 107, 063005 (2023)
FIG. 6. The solid blue and dotted red lines correspond to Z0 contribution to the ðΔmBd
two parametrizations of the VD
current and projected colliders bounds. See text for details.
Z0 as a function of it mass (see Eq. (10), for the
L matrix, see Eqs. (12) and (13). The silver region corresponds to the FCNC exclusion region. We overlaid
Þ
(i) MZ0 ≥ 4 TeV, LHC 13 TeV
(ii) MZ0 ≥ 5.6 TeV, HL-LHC 14 TeV
(iii) MZ0 ≥ 9.6 TeV, HE-LHC 27 TeV
(iv) MZ0 ≥ 27 TeV, FCC-hh 100 TeV
These limits are exhibited in Figs. 3–6. We have gathered
enough information to discuss our results.
V. DISCUSSION
For the K0 − ¯K0 system, the results are summarized in
Fig. 3. The silver region corresponds to the region in
which the Z0 contribution exceeds the experimental error
the parametrizations
[see Eq. (11)]. One can see that
one and two give rise to distinct bounds on the Z0 mass.
Adopting parametrization 1 we find mZ0 > 113 TeV,
whereas using parametrization 2 we get mZ0 >112GeV.
We superimposed the LHC 13 TeV bound as well as
projections for
the HL-LHC, HE-LHC, and FCC-hh
collider.
Regarding the D0 − ¯D0 system, Fig. 4, we get mZ0 >
307 TeV for parametrization 1, and for parametrization 2
we get mZ0 > 256 TeV. We superimposed the LHC 13 TeV
bound as well as projections for the HL-LHC, HE-LHC,
0 system, Fig. 5, we
and FCC-hh collider. As for the B0
obtain mZ0 > 154 TeV for parametrization 1, and for
parametrization 2 we get mZ0 > 154 GeV. Lastly, for the
B0
0 system, Fig. 6, we find mZ0 > 400 TeV for both
d
parametrizations.
s − ¯Bs
− ¯Bd
We highlight that in Figs. 3 and 5 the lower bound on the
Z0 boson mass rising from parametrization 2 is too weak,
falling out of the plot range. Thus, it does not appear in
the figures. It is clear from our findings that one ought to
consider all four meson systems at the same time because
one can randomly pick a parametrization designed to
suppress the new physics contribution for a given meson
system. Without a general approach over FCNC no solid
conclusions can be drawn. Moreover, for the parametriza-
tion explored in this work, the B0
0 system is the most
d
constraining. We remind the reader that our lower mass
bounds are driven by experimental errors, as discussed in
Eq. (11). It is exciting to see the interplay between future
colliders and FCNC because depending on the parametri-
zation used, FCNC can offer a most restrictive probe than
future colliders.
− ¯Bd
We highlight that our conclusions are also applicable to
the 3-3-1 model with heavy neutral leptons because the
neutral current is identical [51,54,55,107]. One should have
in mind, depending on the parametrization adopted, FCNC
does lead to a lower mass bound much stronger than the
LHC and even future colliders. Hence, one cannot overlook
the Z0 contributions to FCNC processes.
Having in mind the complementary aspect between
flavor physics and colliders, we discuss recent flavor
anomalies in the context of 3-3-1 models.
VI. FLAVOR ANOMALIES
A. b → s transitions
b → s transitions not consistent with the SM predictions
have been observed in the LHCb data [108–112], which has
triggered a multitude of new physics studies in the context
of Z0 models. Some of them that are of interest to us reside
on the SUð3Þ
L × Uð1Þ
N gauge group. It is true
that there are several ways to arrange the fermion content
C × SUð3Þ
063005-8
FLAVOR CHANGING INTERACTIONS CONFRONTED WITH …
PHYS. REV. D 107, 063005 (2023)
under this gauge symmetry, and these arrangements have
an impact on the precise neutral current mediated by the Z0
boson. However, the impact is minimal as far as collider
physics goes. If there are new exotic fermions that couple to
the Z0 boson and are sufficiently light, the collider limits
based on dilepton searches will be weakened due to the
presence of a new and significant decay mode. Besides
collider physics, the mass difference of the four meson
systems also places a bound on the Z0 mass. That said, we
will assess whether these interpretations to explain the
b → s anomaly are indeed viable. In [23],
the authors
considered a model similar to ours but with five lepton
generations. The SM quarks possess the same quantum
numbers as ours. If the exotic leptons in [23] are sufficiently
heavy to not contribute to the Z0 decay width, the afore-
mentioned collider limits are also applicable. In order to fit
the b → sll anomaly, according to the recent global fits one
μ
μ
μ
μ
10 ≃ −0.6. Being C
9 ¼ −C
needs C
9 and C
10 the Wilson
coefficients that contribute to new physics present in the
[23].
effective Hamiltonian described in Eq.
In [23] however, two quantities are important rBs and
μ
C
9, with the former controlling the bound from the Bs
mixing and the latter the b → sll anomaly. The 3-3-1
model could explain the LHCb anomaly without being
μ
μ
≃ 0.1.
9 ¼ −C
excluded by Bs mixing if C
μ
¼ 347 × 103ðmW=mZ0Þ2d2, and C
9 ¼ 11.3×
However, rBs
103ðmW=mZ0Þ2d, where d ¼ −0.005 is a parameter that
depends on the entries of the quark mixing matrices
relevant for Bs mixing (d ¼ ðVD
Þ33). This value
L
was assigned to obey the current bound. When we use our
the parameter d takes the following
parametrizations,
values: 0.101 and 0.000101 for the parametrizations 1
and 2, respectively.
10 ≃ −0.6 and rBs
Þ(cid:2)
32ðVD
L
(30) of
Given the current LHC bound on the Z0 mass, ∼4 TeV,
one cannot explain simultaneously address the LHCb
anomaly and respect the LHC lower mass bound. We
emphasize that this 4 TeV bound relies on the assumption
that there are no extra decay modes besides the usual
3-3-1 field content. Hence, a way to circumvent our
conclusion is allowing the extra leptons added in [23] to
be sufficiently light to decrease the Z0 branching ratio into
charged leptons and consequently weaken the LHC bound.
This is a nontrival task knowing that these leptons are chiral
leptons, thus can be produced via SM gauge bosons at
colliders, and consequently are subject to strong collider
bounds [113–120].
In [26], we investigated a similar 3-3-1 model and
advocated that existing collider bounds on the Z0 gauge
boson belonging to the 3-3-1 model could be significantly
lowered if all Z0 decay channel modes are included. The
possible 3-3-1 decay channels have already been included
in [106]. Once more, a weakening of the LHC bound would
require the chiral leptons introduced in [26] to be suffi-
ciently light. Our reason to disfavor this possibility was
mentioned above.
B. RðD(cid:2)Þ Anomaly
In [24], the authors considered an exotic field content based
on the 3-3-1 symmetry, and focused on the charged Higgs
contribution to the semileptonic B-meson decay, particularly
on RðD(cid:2)Þ the anomaly reported by BABAR, Belle, and LHCb.
However, the authors argue that they can take the charged
Higgs mass below 1 TeV while keeping the gauge boson
masses at sufficiently high scales. It has been shown that
despite being a scalar, its mass is naturally predicted to be
around the energy scale at which the 3-3-1 symmetry is
spontaneously broken, unless ones invoke a fine-tuning in the
quartic scalar couplings [49]. In other words, the mass of the
charged scalar is around vχ. Therefore, given the collider
bounds, and the FCNC bounds we derived, the proposed
3-3-1 explanation to the RðD(cid:2)Þ anomaly is disfavored.
VII. CONCLUSIONS
d
− ¯B0
We have studied FCNC in a 3-3-1 model using the four
meson systems, namely, K0 − ¯K0, D0 − ¯D0, B0
d, and
s − ¯B0
B0
s. We derived lower mass bounds that range from
112 GeV up to 400 TeV using different parametrizations of
the quark mixing matrices to solidly show that constraints
to large systematic
stemming from FCNC are subject
uncertainties. We have shown that a robust assessment
of FCNC should consider the four meson systems because
specific parametrizations of the quark mixing matrices can
suppress new physics effects at one of the meson systems.
However, as the CKM matrix should be preserved, these
parametrizations tend to enhance FCNC effects on the other
mesons. We carried out a study based on the Z0 contribu-
tions to FCNC, as the scalars are typically much heavier
than the Z0 field, their corrections to FCNC are subdomi-
nant. Considering only gauge interactions, the systematic
effects already drive the new physics sensitivity, let alone
the scalar fields whose contributions depend on arbitrary
choices of the Yukawa couplings and scalar potential
parameters. In summary, a broader view of FCNC is
needed before drawing conclusions. Lastly, we argued that
recent anomalies in b → s and RðD(cid:2)Þ transitions are
disfavored in light of recent collider bounds.
ACKNOWLEDGMENTS
ICTP-SAIFR
2021/01089-1,
We thank Diego Cogollo and Carlos Pires for discussions.
This work was financially supported by the Simons
Foundation (Grant No. 884966, A. F.), FAPESP Grant
No.
FAPESP Grant
No. 2021/14335-0, Coordenação de Aperfeiçoamento de
Pessoal de Nível Superior—Brasil
(CAPES)—Finance
Code 001, CNPq Grant No. 408295/2021-0, Serrapilheira
Foundation (Grant No. Serra-1912–31613), ANID-Chile
FONDECYT 1210378 and 1190845, ANID PIA/APOYO
SAPHIR-Programa Milenio—code
AFB180002
ICN2019 044. C. S.
is supported by Grant No. 2020/
00320-9, São Paulo Research Foundation (FAPESP).
and
063005-9
A. E. CÁRCAMO HERNÁNDEZ et al.
PHYS. REV. D 107, 063005 (2023)
[1] R. Aaij et al. (LHCb Collaboration), Phys. Rev. Lett. 127,
[34] P. V. Dong, H. N. Long, C. H. Nam, and V. V. Vien, Phys.
111801 (2021).
[2] H. D. Politzer, Phys. Rev. Lett. 30, 1346 (1973).
[3] S. L. Glashow, J. Iliopoulos, and L. Maiani, Phys. Rev. D
2, 1285 (1970).
Rev. D 85, 053001 (2012).
[35] D. Cogollo, A. V. de Andrade, F. S. Queiroz, and P.
Rebello Teles, Eur. Phys. J. C 72, 2029 (2012).
[36] P. T. Giang, L. T. Hue, D. T. Huong, and H. N. Long, Nucl.
[4] M. K. Gaillard and B. W. Lee, Phys. Rev. D 10, 897
Phys. B864, 85 (2012).
(1974).
[37] L. T. Hue, D. T. Huong, and H. N. Long, Nucl. Phys. B873,
[5] S. W. Herb et al., Phys. Rev. Lett. 39, 252 (1977).
[6] F. Abe et al. (CDF Collaboration), Phys. Rev. Lett. 74,
2626 (1995).
207 (2013).
[38] A. C. B. Machado, J. C. Montero, and V. Pleitez, Phys.
Rev. D 88, 113002 (2013).
[7] S. Abachi et al. (D0 Collaboration), Phys. Rev. Lett. 74,
[39] A. E. Cárcamo Hernández, R. Martinez, and F. Ochoa, Eur.
2632 (1995).
Phys. J. C 76, 634 (2016).
[8] M. Staric et al. (Belle Collaboration), Phys. Rev. Lett. 98,
[40] V. V. Vien and H. N. Long, J. Korean Phys. Soc. 66, 1809
211803 (2007).
(2015).
[9] B. Aubert et al. (BABAR Collaboration), Phys. Rev. Lett.
[41] V. V. Vien and H. N. Long, Int. J. Mod. Phys. A 30,
98, 211802 (2007).
1550117 (2015).
[10] Y. S. Amhis et al. (HFLAV Collaboration), Eur. Phys. J. C
[42] L. T. Hue, H. N. Long, T. T. Thuc, and T. Phong Nguyen,
81, 226 (2021).
Nucl. Phys. B907, 37 (2016).
[11] A. J. Buras, P. Gambino, M. Gorbahn, S. Jager, and L.
[43] A. E. Cárcamo Hernández and R. Martinez, Nucl. Phys.
Silvestrini, Phys. Lett. B 500, 161 (2001).
[12] A. J. Buras, Acta Phys. Pol. B 34, 5615 (2003).
[13] G. D’Ambrosio, G. F. Giudice, G. Isidori, and A. Strumia,
B905, 337 (2016).
[44] V. Van Vien, Braz. J. Phys. 45, 467 (2015); 45, 807(E)
(2015).
Nucl. Phys. B645, 155 (2002).
[45] A. C. B. Machado, J. Montaño, and V. Pleitez, J. Phys. G
[14] A. J. Buras, F. De Fazio, and J. Girrbach-Noe, J. High
46, 115005 (2019).
Energy Phys. 08 (2014) 039.
[46] R. M. Fonseca and M. Hirsch, J. High Energy Phys. 08
[15] J. Aebischer, A. J. Buras, M. Cerd`a-Sevilla, and F. De
(2016) 003.
Fazio, J. High Energy Phys. 02 (2020) 183.
[47] D. T. Huong, D. N. Dinh, L. D. Thien, and P. Van Dong,
[16] M. Singer, J. W. F. Valle, and J. Schechter, Phys. Rev. D
J. High Energy Phys. 08 (2019) 051.
22, 738 (1980).
[17] F. Pisano and V. Pleitez, Phys. Rev. D 46, 410 (1992).
[18] R. Foot, O. F. Hernandez, F. Pisano, and V. Pleitez, Phys.
Rev. D 47, 4158 (1993).
[48] T. T. Hong, H. T. Hung, H. H. Phuong, L. T. T. Phuong, and
L. T. Hue, Prog. Theor. Exp. Phys. 2020, 043B03 (2020).
[49] V. Oliveira and C. A. d. S. Pires, arXiv:2208.00420.
[50] C. A. de S. Pires and P. S. Rodrigues da Silva, J. Cosmol.
[19] J. C. Montero, F. Pisano, and V. Pleitez, Phys. Rev. D 47,
Astropart. Phys. 12 (2007) 012.
2918 (1993).
[20] A. G. Dias and V. Pleitez, Phys. Rev. D 80, 056007 (2009).
[21] A. J. Buras, F. De Fazio, and J. Girrbach, J. High Energy
Phys. 02 (2014) 112.
[22] R. Gauld, F. Goertz, and U. Haisch, J. High Energy Phys.
01 (2014) 069.
[23] S. Descotes-Genon, M. Moscati, and G. Ricciardi, Phys.
Rev. D 98, 115030 (2018).
[24] M. Wei and Y. Chong-Xing, Phys. Rev. D 95, 035040
[51] J. K. Mizukoshi, C. A. de S. Pires, F. S. Queiroz, and P. S.
Rodrigues da Silva, Phys. Rev. D 83, 065024 (2011).
[52] J. D. Ruiz-Alvarez, C. A. de S. Pires, F. S. Queiroz, D.
Restrepo, and P. S. Rodrigues da Silva, Phys. Rev. D 86,
075011 (2012).
[53] F. S. Queiroz, AIP Conf. Proc. 1604, 83 (2014).
[54] C. Kelso, C. A. de S. Pires, S. Profumo, F. S. Queiroz, and
P. S. Rodrigues da Silva, Eur. Phys. J. C 74, 2797 (2014).
[55] S. Profumo and F. S. Queiroz, Eur. Phys. J. C 74, 2960
(2017).
(2014).
[25] D. Nguyen Tuan, T. Inami, and H. Do Thi, Eur. Phys. J. C
[56] C. Kelso, H. N. Long, R. Martinez, and F. S. Queiroz,
81, 813 (2021).
Phys. Rev. D 90, 113011 (2014).
[26] A. Addazi, G. Ricciardi, S. Scarlatella, R. Srivastava, and
[57] P. V. Dong, N. T. K. Ngan, and D. V. Soa, Phys. Rev. D 90,
J. W. F. Valle, Phys. Rev. D 106, 035030 (2022).
075019 (2014).
[27] J. M. Cabarcas, J. Duarte, and J. A. Rodriguez, Int. J. Mod.
[58] P. V. Dong, D. T. Huong, F. S. Queiroz, and N. T. Thuy,
Phys. A 29, 1450015 (2014).
Phys. Rev. D 90, 075021 (2014).
[28] D. L. Anderson and M. Sher, Phys. Rev. D 72, 095014
[59] D. Cogollo, A. X. Gonzalez-Morales, F. S. Queiroz, and
(2005).
P. R. Teles, J. Cosmol. Astropart. Phys. 11 (2014) 002.
[29] J. A. Rodriguez and M. Sher, Phys. Rev. D 70, 117702
[60] R. Martínez, J. Nisperuza, F. Ochoa, and J. P. Rubio, Phys.
(2004).
[30] A. Palcu, Mod. Phys. Lett. A 22, 939 (2007).
[31] C. Promberger, S. Schatt, and F. Schwab, Phys. Rev. D 75,
115007 (2007).
[32] P. V. Dong and H. N. Long, Phys. Rev. D 77, 057302
(2008).
Rev. D 90, 095004 (2014).
[61] R. Martinez, J. Nisperuza, F. Ochoa, J. P. Rubio, and C. F.
Sierra, Phys. Rev. D 92, 035016 (2015).
[62] P. S. Rodrigues da Silva, Phys. Int. 7, 15 (2016).
[63] P. V. Dong, C. S. Kim, D. V. Soa, and N. T. Thuy, Phys.
Rev. D 91, 115019 (2015).
[33] P. V. Dong, H. N. Long, D. V. Soa, and V. V. Vien, Eur.
[64] R. Martinez and F. Ochoa, J. High Energy Phys. 05 (2016)
Phys. J. C 71, 1544 (2011).
113.
063005-10
FLAVOR CHANGING INTERACTIONS CONFRONTED WITH …
PHYS. REV. D 107, 063005 (2023)
[65] D. T. Huong and P. V. Dong, Eur. Phys. J. C 77, 204
(2017).
[66] C. A. de S. Pires, P. S. Rodrigues da Silva, A. C. O. Santos,
and C. Siqueira, Phys. Rev. D 94, 055014 (2016).
[67] F. Pisano and V. Pleitez, Phys. Rev. D 46, 410 (1992).
[68] R. Foot, H. N. Long, and T. A. Tran, Phys. Rev. D 50, R34
(1994).
[94] A. E. Cárcamo Hernández, I. de Medeiros Varzielas, M. L.
López-Ibáñez, and A. Melis, J. High Energy Phys. 05
(2021) 215.
[95] A. E. Cárcamo Hernández, S. F. King, and H. Lee, Phys.
Rev. D 105, 015021 (2022).
[96] A. E. C. Hernández and I. Schmidt, Nucl. Phys. B976,
115696 (2022).
[69] H. N. Long, Phys. Rev. D 54, 4691 (1996).
[70] H. N. Long, Phys. Rev. D 53, 437 (1996).
[71] W. A. Ponce, Y. Giraldo, and L. A. Sanchez, Phys. Rev. D
67, 075001 (2003).
[97] A. E. C. Hernández, D. T. Huong, and I. Schmidt, Eur.
Phys. J. C 82, 63 (2022).
[98] A. Alves, A. Berlin, S. Profumo, and F. S. Queiroz, J. High
Energy Phys. 10 (2015) 076.
[72] J. Kopp, P. A. N. Machado, M. Maltoni, and T. Schwetz,
[99] A. Alves, A. Berlin, S. Profumo, and F. S. Queiroz, Phys.
J. High Energy Phys. 05 (2013) 050.
Rev. D 92, 083004 (2015).
[73] M. C. Gonzalez-Garcia, M. Maltoni, and T. Schwetz, Nucl.
[100] M. Klasen, F. Lyonnet, and F. S. Queiroz, Eur. Phys. J. C
Phys. B908, 199 (2016).
77, 348 (2017).
[74] J. Bergstrom, M. C. Gonzalez-Garcia, M. Maltoni, and T.
[101] F. Kahlhoefer, A. Mück, S. Schulte, and P. Tunney, J. High
Schwetz, J. High Energy Phys. 09 (2015) 200.
Energy Phys. 03 (2020) 104.
[75] A. G. Dias, C. A. de S. Pires, P. S. Rodrigues da Silva, and
A. Sampieri, Phys. Rev. D 86, 035007 (2012).
[76] S. M. Boucenna, J. W. F. Valle, and A. Vicente, Phys. Rev.
[102] T. Jezo, M. Klasen, D. R. Lamprea, F. Lyonnet, and I.
Schienbein, J. High Energy Phys. 12 (2014) 092.
[103] T. Ježo, M. Klasen, D. R. Lamprea, F. Lyonnet, and I.
D 92, 053001 (2015).
Schienbein, arXiv:1508.03539.
[77] M. Misiak, S. Pokorski, and J. Rosiek, Adv. Ser. Dir. High
[104] Y. A. Coutinho, V. Salustino Guimarães, and A. A.
Energy Phys. 15, 795 (1998).
Nepomuceno, Phys. Rev. D 87, 115014 (2013).
[78] G. Barenboim, F. J. Botella, and O. Vives, Phys. Rev. D 64,
[105] C. Salazar, R. H. Benavides, W. A. Ponce, and E. Rojas,
015007 (2001).
[79] A. Datta, Phys. Rev. D 78, 095004 (2008).
[80] R. L. Workman and Others (Particle Data Group), Prog.
Theor. Exp. Phys. 2022, 083C01 (2022).
[81] F. Pisano, D. Gomez Dumm, F. Pisano, and V. Pleitez,
Mod. Phys. Lett. A 09, 1609 (1994).
[82] H. N. Long and V. T. Van, J. Phys. G 25, 2319 (1999).
[83] R. H. Benavides, Y. Giraldo, and W. A. Ponce, Phys. Rev.
J. High Energy Phys. 07 (2015) 096.
[106] A. Alves, L. Duarte, S. Kovalenko, Y. M. Oviedo-Torres,
F. S. Queiroz, and Y. S. Villamizar, Phys. Rev. D 106,
055027 (2022).
[107] J. D. Ruiz-Alvarez, C. A. de S. Pires, F. S. Queiroz, D.
Restrepo, and P. S. Rodrigues da Silva, Phys. Rev. D 86,
075011 (2012).
[108] R. Aaij et al. (LHCb Collaboration), Phys. Rev. Lett. 113,
D 80, 113009 (2009).
151601 (2014).
[84] R. J. Dowdall, C. T. H. Davies, R. R. Horgan, G. P. Lepage,
C. J. Monahan, J. Shigemitsu, and M. Wingate, Phys. Rev.
D 100, 094508 (2019).
[85] P. Zyla et al. (Particle Data Group), Prog. Theor. Exp.
Phys. 2020, 083C01 (2020). and 2021 update.
[86] G. C. Branco, J. T. Penedo, P. M. F. Pereira, M. N. Rebelo,
and J. I. Silva-Marcos, J. High Energy Phys. 07 (2021)
099.
[87] A. Angelopoulos et al. (CPLEAR Collaboration), Phys.
[109] R. Aaij et al. (LHCb Collaboration), J. High Energy Phys.
02 (2016) 104.
[110] S. Descotes-Genon, J. Matias, and J. Virto, Phys. Rev. D
88, 074002 (2013).
[111] G. Hiller and M. Schmaltz, Phys. Rev. D 90, 054014 (2014).
[112] S. Descotes-Genon, L. Hofer, J. Matias, and J. Virto,
J. High Energy Phys. 06 (2016) 092.
[113] I. Gogoladze, B. He, and Q. Shafi, Phys. Lett. B 690, 495
(2010).
Lett. B 444, 38 (1998).
[114] W. Altmannshofer, M. Bauer, and M. Carena, J. High
[88] M. Artuso, G. Borissov, and A. Lenz, Rev. Mod. Phys. 88,
Energy Phys. 01 (2014) 060.
045002 (2016); 91, 049901(A) (2019).
[115] T. Ma, B. Zhang, and G. Cacciapaglia, Phys. Rev. D 89,
[89] T. Jubb, M. Kirk, A. Lenz, and G. Tetlalmatzi-Xolocotzi,
093022 (2014).
Nucl. Phys. B915, 431 (2017).
[90] B. Wang, Proc. Sci. LATTICE2018 (2019) 286 [arXiv:
1812.05302].
[116] D. Cogollo, Int. J. Mod. Phys. A 30, 1550187 (2015).
[117] S. Gopalakrishna, T. S. Mukherjee, and S. Sadhukhan,
Phys. Rev. D 93, 055004 (2016).
[91] A. Lenz and G. Tetlalmatzi-Xolocotzi, J. High Energy
[118] J. C. Helo, M. Hirsch, and Z. S. Wang, J. High Energy
Phys. 07 (2020) 177.
Phys. 07 (2018) 056.
[92] Y. Aoki et al., Eur. Phys. J. C 82, 869 (2022).
[93] F. C. Correia and V. Pleitez, Phys. Rev. D 92, 113006
[119] S. Rappoccio, Rev. Phys. 4, 100027 (2019).
[120] G. Guedes and J. Santiago, J. High Energy Phys. 01 (2022)
(2015).
111.
063005-11
| null |
10.1016_j.envres.2023.115368.pdf
|
Data availability
Data will be made available on request.
|
Data availability Data will be made available on request.
|
Contents lists available at ScienceDirect
Environmental Research
journal homepage: www.elsevier.com/locate/envres
Effects of pesticide exposure on oxidative stress and DNA methylation
urinary biomarkers in Czech adults and children from the
CELSPAC-SPECIMEn cohort
Tom´aˇs Janoˇs a, Ilse Ottenbros b,c, Lucie Bl´ahov´a a, Petr
ˇ
Jessica Sheardov´a a, Jelle Vlaanderen b, Pavel
Cupr a, *
a RECETOX, Faculty of Science, Masaryk University, Kotlarska 2, Brno, Czech Republic
b Institute for Risk Assessment Sciences, Utrecht University, Utrecht, the Netherlands
c Center for Sustainability, Environment and Health, National Institute for Public Health and the Environment (RIVM), Bilthoven, Netherlands
ˇ
Senk a, Libor
ˇ
Sulc a, Nina P´aleˇsov´a a,
A R T I C L E I N F O
A B S T R A C T
Handling Editor: Jose L Domingo
Keywords:
Epigenetics
DNA methylation
Oxidative stress
Current-use pesticides (CUPs)
Urine
Current-use pesticide (CUP) exposure occurs mainly through diet and environmental application in both agri-
cultural and urban settings. While pesticide exposure has been associated with many adverse health outcomes,
the intermediary molecular mechanisms are still not completely elucidated. Among others, their roles in epi-
genetics (DNA methylation) and DNA damage due to oxidative stress are presumed. Scientific evidence on
urinary biomarkers of such body response in general population is limited, especially in children.
A total of 440 urine samples (n = 110 parent-child pairs) were collected during the winter and summer seasons
in order to describe levels of overall DNA methylation (5-mC, 5-mdC, 5-hmdC, 7-mG, 3-mA) and oxidative stress
(8-OHdG) biomarkers and investigate their possible associations with metabolites of pyrethroids (3-PBA, t/c-
DCCA), chlorpyrifos (TCPY), and tebuconazole (TEB-OH). Linear mixed-effects models accounting for intra-
individual and intrahousehold correlations were utilized. We applied false discovery rate procedure to account
for multiplicity and adjusted for potential confounding variables.
Higher urinary levels of most biological response biomarkers were measured in winter samples. In adjusted
repeated measures models, interquartile range (IQR) increases in pyrethroid metabolites were associated with
higher oxidative stress. t/c-DCCA and TCPY were associated with higher urinary levels of cytosine methylation
biomarkers (5-mC and/or 5-mdC). The most robust association was observed for tebuconazole metabolite with 3-
mA ((cid:0) 15.1% change per IQR increase, 95% CI = (cid:0) 23.6, (cid:0) 5.69) suggesting a role of this pesticide in reduced
demethylation processes through possible DNA glycosylase inhibition.
Our results indicate an urgent need to extend the range of analyzed environmental chemicals such as azole
pesticides (e.g. prothioconazole) in human biomonitoring studies. This is the first study to report urinary DNA
methylation biomarkers in children and associations between CUP metabolites and a comprehensive set of
biomarkers including methylated and oxidized DNA alterations. Observed associations warrant further large-
scale research of these biomarkers and environmental pollutants including CUPs.
1. Introduction
Pesticides are agrochemicals used worldwide for the protection of
crops from various types of pests, as well as to control detrimental or-
ganisms (e.g., rodents) or vector-borne diseases. Their role is key in
sufficient food production and management of human diseases. How-
ever, their overproduction and overuse are problematic not only in
agricultural areas but in urban environment as well (Md Meftaul et al.,
2020). Although some progress in recent years toward safer use of
pesticides was achieved, current-use pesticides (CUPs) still represent a
potential risk for human health (K. H. Kim et al., 2017). Exposure to CUP
mixtures can occur through several routes and pathways. While diet has
been identified as the main exposure route (Becker et al., 2006;
Nougad`ere et al., 2012), non-dietary routes such as direct skin contact,
exposure via house dust, providing a long-term residential exposure
route or airborne pesticides inhalation play additionally a significant
* Corresponding author. RECETOX Centre, Faculty of Science, Masaryk University, Kamenice 753/5, pavilion A29, 625 00 Brno, Czech Republic.
E-mail address: [email protected] (P.
ˇ
Cupr).
https://doi.org/10.1016/j.envres.2023.115368
Received 13 September 2022; Received in revised form 22 January 2023; Accepted 24 January 2023
EnvironmentalResearch222(2023)115368Availableonline28January20230013-9351/©2023TheAuthors.PublishedbyElsevierInc.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/). T. Janoˇs et al.
role, especially when considering population living close to agricultural
lands (K. H. Kim et al., 2017; Dereumeaux et al., 2020).
To better understand health consequences and provide risk assess-
ment, CUPs or their metabolites are monitored in human fluids such as
serum (Chang et al., 2017) or urine (Dereumeaux et al., 2018). As
collection of urine samples is less invasive, urine is available in adequate
quantity and easily collected by the study participants themselves, urine
collection is often the preferred method over blood samples (Oerlemans
et al., 2021).
Several previously conducted human biomonitoring studies sug-
gested an association between CUP exposure and potential adverse
health effects, for instance male reproductive effects (sperm quality and
sperm DNA damage, reproductive hormone disorders), neurobehavioral
development problems, endocrine disrupting effects, or even cancer
(Koureas et al., 2012; Gonz´alez-Alzaga et al., 2014; Saillenfait et al.,
2015). Moreover, specific population sub-groups (e.g. children, preg-
nant women) could be more sensitive to the pesticide exposure than
others (K. H. Kim et al., 2017).
While exposure has been associated with many above mentioned
health outcomes, a majority of the intermediary molecular mechanisms
by which CUPs could exert their harmful effects are still not completely
elucidated. Among various mechanisms linked to pesticide-induced
chronic diseases, their roles in epigenetics (DNA methylation) and
oxidative DNA damage are presumed (Banerjee et al., 2001; Collotta
et al., 2013; Sabarwal et al., 2018). Both methylated and oxidized DNA
lesions have been previously associated with various health outcomes,
such as cancer (Robertson, 2005; Guo et al., 2017), pulmonary (Neo-
fytou et al., 2012) and cardiovascular (di Minno et al., 2016) diseases,
male reproductive issues (Pan et al., 2016) or neurodegenerative dis-
orders (Guo et al., 2017; Gątarek et al., 2020). DNA methylation is one of
the most important epigenetic modifications with the ability to regulate
gene expression, cellular differentiation and genetic imprinting (Gehr-
ing et al., 2009). It comprises adding a methyl group enzymatically,
typically but not exclusively to the cytosine nucleotide (DNA methyl-
ation) as well as nonenzymatically to the adenine and guanine nucleo-
tide (methylated DNA lesions) (Hu et al., 2012). Abnormal methylation
levels could be caused by both methylation and demethylation processes
(Chen and Riggs, 2011). Furthermore, DNA may be altered by the hy-
droxyl radical, hazardous reactive oxygen species (ROS) with subse-
quent methylated and oxidized DNA lesions creation. ROS also play a
role in the oxidation of methionine which could contribute to the for-
mation of methyl radicals, leading to potential chemical DNA hyper-
methylation (Hu et al., 2012). Under normal circumstances, there is a
balance between ROS production and antioxidant activity or accumu-
lation. If a disbalance occurs, ROS overproduction can cause oxidative
damage to nucleic acids, including both nuclear and mitochondrial DNA
and RNA, with
adduct 8-hydrox-
and
y-2
-deoxyguanosine (8-OHdG) creation (Valavanidis et al., 2009). Both
methylated and oxidized DNA alterations could be removed by several
pathways, especially base excision repair (BER), nucleotide excision
repair (NER), oxidation or hydrolysis and their resulting products
appear in the bloodstream and are excreted and present in urine (Hu
et al., 2011, 2012; Fleming et al., 2015). These biomarkers then reflect
the DNA repair processes in the whole body and have been previously
detected in urine and proposed as possible response biomarkers to
exogenous exposures (Valavanidis et al., 2009; Hu et al., 2011, 2012;
Pan et al., 2016; Graille et al., 2020) or promising early biomarkers of
several diseases (Pan et al., 2016; Onishi et al., 2019). Furthermore,
some of them were associated with exposure to other environmental
pollutants such as phthalates, benzene or organophosphate flame re-
tardants (Pan et al., 2016; Ait Bamai et al., 2019; Lovreglio et al., 2020).
Nevertheless, there is no study investigating associations of urinary
response biomarkers of both oxidized and methylated DNA alterations in
relation to CUP exposure in general population. Moreover, there is a
complete knowledge gap in urinary DNA methylation biomarkers in
children.
abundant
stable
′
Therefore, in the present study we (I.) determined and described
urinary levels of DNA methylation and oxidative stress biomarkers in
samples from winter and summer seasons among Czech adults and
children from the CELSPAC-SPECIMEn cohort and (II.) investigated
possible associations between response biomarkers (DNA methylation
and oxidative stress) and urinary levels of CUP metabolites or parental
compounds.
2. Materials and methods
2.1. Study population and sample collection
The present study is part of the SPECIMEn (Survey on PEstiCIde
Mixtures in Europe) study with the initial aim to assess exposure to
pesticide mixtures in the general population (Ottenbros et al., 2023).
This work is focused on the Czech cohort of the SPECIMEn study:
CELSPAC-SPECIMEn (Central European Longitudinal Studies of Parents
and Children). The CELSPAC-SPECIMEn study in the Czech Republic
received ethical approval under ref. no. ELSPAC/EK/3/2019. A detailed
ˇ
description of the study protocol has been published previously (
Sulc
et al., 2022). Briefly, adult-child pairs were recruited during 2019 and
2020. Only adults older than 20 years with school-age (5–12 years)
children were accepted into the study. Farmers and other professionals
with potential occupational exposure to CUPs were excluded. Urine
sample collection took place in two rounds, from mid-January 2020 to
mid-March 2020 (hereinafter “winter season”) and from the end of May
2020 to the end of July 2020 (hereinafter “summer season”). Samples
were not collected on weekends and Mondays due to possible differences
in the participant behavior during the weekend. Each participant
received the materials needed for urine collection, including urine
containers, collection cups, storage bags, informed consent, and a
questionnaire. Urine samples (first-morning void) were self-collected by
participants then stored in the fridge until the arrival of the field worker.
Samples were transported to the laboratory under refrigeration, ali-
quoted, and stored at (cid:0) 80
C until analysis. The whole process from
sample collection to sample storage took no longer than 24 h. One
adult-child pair was excluded from further analyses because of the
dropout during the study course. The final number of participants was
110 adults and 110 children sampled in two seasons (total n = 440).
◦
2.2. Urinary biomarkers
Collected urine samples (n = 440) were analyzed for twelve bio-
markers of exposure to CUPs and six response biomarkers. The selection
of CUP biomarkers was based on the recommendation of HBM4EU (The
European Human Biomonitoring Initiative) (Prioritized substance
group: Pesticides) (Ougier et al., 2021), the annual reports of Plant
Protection Products in the Czech Republic (CISTA, 2022), and on the
European Food Safety Authority (EFSA) report (EFSA, 2021). Urinary
CUP metabolite concentrations were measured by means of
high-performance liquid chromatography (HPLC) in tandem with a mass
spectrometer-mass spectrometer system (MS-MS). Detailed description
of the method, quality assurance and quality control, coupled with the
ˇ
list of exposure biomarkers have been described elsewhere (
Sulc et al.,
2022). Only biomarkers with detection frequency at least 40% of all the
samples were included in the current study: 3-phenoxybenzoic acid
(3-PBA) and trans/cis-3-(2,2-dichlorovinyl)-2,2-dimethylcyclopro-pane
carboxylic acid (t/c-DCCA) as pyrethroid metabolites, chlorpyrifos
metabolite 3,5,6-trichloro-2- pyridinol (TCPY) and tebuconazole
metabolite hydroxy-1-tebuconazole (TEB-OH).
Response biomarkers (biomarker of oxidative stress and DNA
methylation) were selected based on prior literature research and
consist of (I.) urinary biomarker of oxidative damage, specifically 8-
hydroxydeoxyguanosine (8-OHdG) and (II.) biomarkers of potential
nucleic acid methylation, namely 5-methylcytosine (5-mC), its deoxy-
nucleoside and oxidized modification: 5-methyl-2
-deoxycytidine (5-
′
EnvironmentalResearch222(2023)1153682 T. Janoˇs et al.
′
mdC) and 5-hydroxymethyl-2
-deoxycytidine (5-hmdC); 7-methylgua-
nine (7-mG) and 3-methyladenine (3-mA). Extraction of selected
epigenetic biomarkers and biomarker of oxidative stress was performed
according to the previously published study (Bl´ahov´a et al., 2023 –
manuscript submitted). In short, the urine sample were thawed and 10
μL of internal standards mixture was added to 0.5 mL of each sample and
calibration solutions (0, 0.05, 0.5, 5, 50, 500 μg/L in 0.1% v/v formic
acid). Samples were freeze-dried and then extracted with isopropanol.
Insoluble particles were removed by centrifugation, supernatants were
evaporated to dryness and further redissolved in 0.1% formic acid (v/v).
Possible residual particles were removed using microspin filters (0.2 μm;
cellulose acetate; Fisher Scientific). Filtrates were stored in glass inserts
◦
at (cid:0) 20
until the analyses.
The analysis of selected response biomarkers was done by ultra-
performance liquid chromatograph Acquity UPLC (Waters, Ireland)
followed by tandem mass spectrometer Xevo TQ-S (Waters, Ireland).
The mobile phase consisted of 0.1% formic acid in water (A) and
acetonitrile acidified by 0.1% formic acid (B). The binary pump gradient
was linear (3% B to 80% B at 5 min). The flow rate was 0.2 mL/min, and
10 μL of the individual sample was injected for the analysis. Analytes
were detected in ESI positive ion mode and the ionization parameters
were as follows: capillary voltage, 2.5 kV; the source temperature and
◦
the desolvation temperature, 150 and 750
C, respectively; the cone gas
flow, 150 (L/h); the cone voltages, 30 V; the desolvation gas flow, 750
(L/h); and the collision gas flow, 0.15 mL/min. The concentrations of
response biomarkers in extracts were determined from a calibration
curve with the use of an internal standard (software Mass Lynx, Man-
chester, UK). Concentrations of 5-mC, 5-mdC and 5-hmdC were cor-
rected for the content of internal standard 5-mdC d3; the concentration
of 3-mA was corrected for the content of internal standard 3-mA d3 and
concentrations of 7-mG and 8-OHdG were corrected for the content of
internal standard 8-OHdG 15N5. Quality assurance and quality control
samples, including blanks, spiked samples (5.0 ng/mL of all analytes in
0.1% formic acid) and model urine samples (in house reference material
with known level of biomarkers) were repeatedly extracted and included
in the analysis. Quality control samples were analyzed after every 25
urine samples and found repeatability was acceptable (relative standard
deviation (RSD) ≤ 15%). Five procedural blanks were analyzed in each
analytical run with concentration below LOD. Mass spectrometer and
other validation parameters are listed in SI Table 1.
2.3. Data analysis
Data were analyzed and visualized in the R programming language,
version 4.1.1 (R Core Team, 2021). All urinary biomarkers were cor-
rected for urine dilution using specific gravity (SG). SG was measured at
the time of response biomarkers analysis using handheld refractometer
Atago PAL-10 S, Japan. SG-corrected concentrations were created using
following formula:
(
)
Bc = B ×
SGavg. (cid:0) 1
SG (cid:0) 1
where BC is the SG corrected concentration of a biomarker, B is the
measured concentration of a biomarker, SGavg. is the average specific
gravity of all adult (1.017) or child (1.021) samples and SG is the specific
gravity of the respective urine sample (Sauv´e et al., 2015). Values below
limit of quantification (LOQ) and/or limit of detection (LOD) were
imputed on the basis of maximum likelihood multiple estimation
dependent on observed values, which were expected to have a
lognormal distribution (Lubin et al., 2004). The imputation was done
only for compounds detected in at least 40% of all the samples. Before
statistical analyses, we used natural log (ln) transformation to achieve
normal distribution of measured biomarkers. The Pearson coefficient (r)
was used to determine correlations between response biomarkers.
To examine the associations of response biomarkers with CUP
metabolites, the linear mixed effect (LME) model was utilized. Random
intercepts for participant ID and specific household were used to ac-
count for intraindividual and intrahousehold correlations. For modeling,
uncorrected concentrations were used and models were adjusted for
specific gravity of urine as proposed by (Barr et al., 2005). We first
constructed basic model with specific gravity (continuous), age (in
years, continuous), sex (male, female) and body mass index (BMI) (in
kg/m2, continuous) adjustment and then potential variables were added
to examine the associations. Minimal sufficient adjustment set of cova-
riates included in LME models were selected based on prior knowledge
and direct acyclic graphs (DAGs) approach (Shrier and Platt, 2008).
Therefore, single exposure mixed effect models were additionally
adjusted for the following characteristics: season (winter, summer), area
of agricultural fields in 250 m radius around the households (in m2,
continuous) (adjusted model 1), frequency of organic food consumption
(<1 per month, 1–3 per month, 1 per week, 2–6 per week, daily),
amount of fruit consumed in 3 days before sampling (number of pieces),
amount of vegetable consumed in 3 days before sampling (number of
pieces) (adjusted model 2). Variables adult smoking status (never
smoker, former smoker, current smoker), household income (<25%,
25–50%, 50–75%, >75% of South-Moravian region average), adult ed-
ucation (primary, secondary, tertiary, university) were identified as
redundant by DAG and were not included in the models to avoid over
adjustment (see SI Fig. 1). To account for multiplicity (72 comparisons),
false discovery rate (FDR) procedure was applied and
false
coverage-statement rate adjusted 95% confidence intervals (FCR-ad-
justed 95% CIs) were constructed (Benjamini and Yekutieli, 2005).
Several sensitivity analyses were performed. First, we changed the urine
dilution adjustment approach by constructing models of SG-corrected
biomarkers of exposure and response instead of including SG as a co-
variate. Second, a 90% winsorizing transformation method was applied
to reduce the effect of possible outliers. Third, we estimated effects using
multiple exposure mixed effects model additionally adjusted for multi-
ple urinary CUP metabolites.
3. Results
Demographic and behavioral characteristics are presented in
Table 1. Median age of study participants was 41 and 9 years for adults
and children, respectively. Among adults prevail females (65%) with
boys as an offspring (57%). At the beginning of the study, the majority of
participants (64% of adults and 89% of children) had their BMI in
normal range. Most adults had a university education (78%) and were
predominantly non-smokers (88%). Households were both in urban and
agricultural areas with 34% of them lacking agricultural fields within a
radius of 250 m around the household.
Descriptive statistics of SG adjusted and non-adjusted urinary levels
of response biomarkers and CUP metabolites coupled with the LOQ, LOD
and detection frequency for each measured compound are summarized
in Table 2 and SI Tables 2–3. In our study, response biomarkers were
present in all urine samples (n = 440). Out of CUP metabolites, the most
frequently detected metabolite was TEB-OH, with a detection frequency
varying from 94.6% to 99.1% across the seasons and subgroups, fol-
lowed by 3-PBA (51.8%–88.2%), t/c-DCCA (50%-86.4), and TCPY
(40.9%-87.3). Concentrations of all biomarkers (both response bio-
markers and CUP metabolites) were higher in children than in adults in
both seasons except of TEB-OH in the winter season. When comparing
urinary biomarker levels in the winter and summer seasons, we observed
a few significant differences (p < 0.05). Most of them were characterized
by higher levels in winter: 8-OHdG, 5-mC, 3-PBA, t/c-DCCA, TCPY in
children and 5-mC, 3-mA, t/c-DCCA, TCPY in adults. The only exception
was the 7-mG biomarker in adults, which was detected in statistically
significant higher concentrations in summer samples.
Correlation analysis among the response biomarkers across the sea-
sons and subgroups separately (children in winter, children in summer,
adults in winter, adults in summer) showed some statistically significant
EnvironmentalResearch222(2023)1153683 T. Janoˇs et al.
Table 1
Demographic characteristics of the study population at baseline.
Characteristic
Age (years)
Parent
Child
Sex
Adults
Female (%)
Male (%)
Children
Girls (%)
Boys (%)
BMI
Adults
Underweight or normal (<25) (%)
Overweight (25–30) (%)
Obese (>30) (%)
Children
Underweight or normal (<25) (%)
Overweight (25–30) (%)
Obese (>30) (%)
Adult education
Primary and secondary (%)
Tertiary (%)
University (%)
Adult smoking status
Never smoker (%)
Former smoker (%)
Current smoker (%)
Household income1
1st quartile (%)
2nd quartile (%)
3rd quartile (%)
4th quartile (%)
Any agricultural area around the household2
No (%)
Yes (%)
Median (Min – Max)
41 (31–54)
9 (4–15)
Percent
65
35
43
57
Percent
64
31
5
89
8
3
Percent
6
16
78
Percent
68
20
12
Percent
17
46
20
17
Percent
34
66
1 Total household income (% of South-Moravian region average).
2 Presence of agricultural fields within a radius of 250 m around the
household.
patterns. The highest Pearson correlation coefficients were observed
between 5-mdC and 8-OHdG (r ranging from 0.37 to 0.55 among the
seasons and subgroups, p < 0.0001) and between 5-mdC and 5-mC (r =
0.34–0.58, p < 0.001). Weaker correlations were observed between 5-
mdC and 5-hmdC (r = 0.25–0.42, p < 0.01) and between 8-OHdG and
5-mC (r = 0.19–0.39, p < 0.05). The remaining correlations were
observed only in some season and/or subgroup or were insignificant,
showing no conclusive pattern (see SI Tables 4–7). Significant positive
correlations were also found between some CUP metabolites and are
published and discussed in detail elsewhere (
ˇ
Sulc et al., 2022).
Estimates of effects from LME models showed some robust associa-
tions across all diversly adjusted models and results are given in Table 3.
Results are presented as a percentage change in response biomarker
concentrations associated with inter-quartile range (IQR) change in the
urinary concentration of a CUP metabolite. Both pyrethroid metabolites
(3-PBA and t/c-DCCA) were associated with an increase in the concen-
tration of oxidative stress biomarker 8-OHdG in the final fully adjusted
model 2. The percentage change in 8-OHdG associated with IQR change
was 10.2% (95% CI: 2.85, 18.1) in the case of 3-PBA and 11.6% (95% CI:
2.47, 21.5) in the case of t/c-DCCA. Furthermore, IQR change in t/c-
DCCA concentration was also associated with 13.6% change (95% CI:
2.50, 25.8) in 5-mdC concentration in adjusted model 2. Similarly,
higher concentrations of 5-mdC and 5-mC were also associated with
chlorpyrifos metabolite TCPY (% change = 11.6%; 95% CI: 0.48, 23.9
and 14.6%; 95% CI: 1.58, 29.2, respectively). The only negative estimate
(increase in CUP metabolite associated with a decrease in response
biomarker) was found between TEB-OH and 3-mA ((cid:0) 15.1%; 95% CI:
23.6, (cid:0) 5.69) which is also the most robust effect observed across all
models. Significant results of adjusted model 2 were similar to adjusted
Table 2
Specific gravity adjusted urinary levels of CUP metabolites and biological
response biomarkers with an overall detection frequency higher than 40%.
Biomarker
(ng/mL)
LOD/
LOQ
DF
(%)
GM
(GSD)
P95
DF
(%)
GM
(GSD)
P95
Summer season (n = 110)
10.2
100
Adults
8-OHdG
5-mC***
5-mdC
5-hmdC
7-mG***
3-mA***
3-PBA
t/c-DCCA*
TCPY***
TEB-OH
Children
8-OHdG***
5-mC***
5-mdC
5-hmdC
7-mG
3-mA
3-PBA*
t/c-DCCA***
TCPY**
TEB-OH
0.05/
0.17
0.05/
0.17
0.10/
0.33
0.05/
0.17
1.00/
3.33
0.10/
0.33
0.04/
0.14
0.03/
0.11
0.03/
0.09
0.02/
0.05
0.05/
0.17
0.05/
0.17
0.10/
0.33
0.05/
0.17
1.00/
3.33
0.10/
0.33
0.04/
0.14
0.03/
0.11
0.03/
0.09
0.02/
0.05
100
100
100
100
29.6
3.90
48.7
Winter season (n = 110)
9.42
5.43
100
(1.42)
22.0
(1.62)
16.9
(1.49)
1.73
(1.59)
1778
(1.85)
9.20
(2.45)
0.121
(4.03)
0.300
(6.10)
2.29
(3.93)
0.459
(2.45)
98.2
7.37
1.75
51.8
60.9
3.16
87.3
36.1
100
5881
0.905
100
100
65.8
Winter season (n = 110)
10.8
6.72
100
(1.38)
31.7
(1.63)
24.5
(1.54)
2.74
(1.47)
3377
(1.70)
11.4
(2.64)
6.00
54.3
43.7
100
100
100
7407
100
100
100
100
100
51.8
50
40.9
94.6
100
100
100
100
100
5.62
(1.44)
17.5
(1.72)
16.8
(1.58)
1.71
(1.70)
2223
(1.65)
7.29
(2.35)
0.123
(4.62)
0.195
(5.02)
0.243
(6.13)
0.494
(2.98)
5.70
(1.45)
26.3
(1.85)
23.9
(1.49)
2.71
(1.51)
3390
(1.61)
9.18
(2.64)
88.2
86.4
83.6
99.1
0.465
(3.65)
1.08
(5.25)
2.53
(5.37)
0.459
(2.26)
2.26
82.7
6.66
76.4
9.73
83.6
1.77
97.3
0.317
(3.72)
0.534
(4.14)
1.17
(4.67)
0.558
(3.29)
36.5
33.4
3.977
4962
29.7
1.18
1.77
3.07
4.05
59.7
44.5
6.08
6984
42.3
1.57
3.23
4.44
9.86
Summer season (n = 110)
10.3
100
Abbreviations: 8-OHdG: 8-hydroxydeoxyguanosine, 5-mC: 5-methylcytosine, 5-
′
′
-deoxycytidine,
-deoxycytidine, 5-hmdC: 5-hydroxymethyl-2
mdC: 5-Methyl-2
7-mG: 7-methylguanine, 3-mA: 3-methyladenine, 3-PBA: 3-phenoxybenzoic
acid, t/c-DCCA: trans/cis-3-(2,2-dichlorovinyl)-2,2-dimethylcyclopro-pane car-
boxylic acid, TCPY: 3,5,6-trichloro-2-pyridinol, TEBOH: hydroxy-1-tebucona-
zole.
DF = detection frequency.
GM = geometric mean.
GSD = geometric standard deviation.
P95 = 95th percentile.
*p < 0.05, **p < 0.01, ***p < 0.001 for significant difference in biomarker
concentration between seasons, estimated from LME model with random in-
tercepts for participant ID and specific household, adjusted for age, BMI, sex,
specific gravity, agricultural area, fruit consumption, vegetable consumption,
organic food consumption.
model 1 and base model. The only exception was decrease in strength of
the association between TCPY and 5-mC which was observed when
comparing base model and adjusted model 1. The remaining associa-
tions were not significant or conclusive. Significant results of the final
fully adjusted model 2 were robust and observed also in sensitivity
analysis. We observed a few increases/decreases in effect estimates,
however still significant, when using SG-corrected biomarker levels,
applying winsorizing and constructing multiple exposure mixed effects
EnvironmentalResearch222(2023)1153684 T. Janoˇs et al.
Table 3
Percentage change and FCR-adjusted 95% confidence interval in urinary
response biomarkers associated with IQR increase in urinary CUP metabolite
concentrations.
Base model a
Adjusted model 1 b
Adjusted model 2 c
% change (95% CI)* % change (95% CI)*
% change (95% CI)*
8-OHdG
3-PBA
t/c-
DCCA
TCPY
TEB-OH
5-mC
3-PBA
t/c-
DCCA
TCPY
TEB-OH
5-mdC
3-PBA
t/c-
DCCA
TCPY
TEB-OH
5-hmdC
3-PBA
t/c-
DCCA
TCPY
TEB-OH
7-mG
3-PBA
t/c-
DCCA
TCPY
TEB-OH
3-mA
3-PBA
t/c-
DCCA
TCPY
TEB-OH
10.5 (3.27, 18.3)
14.3 (5.26, 24.2)
9.82 (2.61, 17.5)
12.7 (3.56, 22.7)
10.2 (2.85, 18.1)
11.6 (2.47, 21.5)
8.92 (0.80, 17.7)
3.09 ((cid:0) 2.06, 8.51)
6.07 ((cid:0) 2.83, 15.8)
3.53 ((cid:0) 1.62, 8.96)
5.37 ((cid:0) 3.48, 15.0)
3.94 ((cid:0) 1.3, 9.46)
(cid:0) 0.44 ((cid:0) 10.0, 10.1)
10.7 ((cid:0) 1.97, 25.0)
(cid:0) 2.48 ((cid:0) 11.5, 7.49)
4.82 ((cid:0) 7.12, 18.3)
(cid:0) 4.59 ((cid:0) 13.6, 5.37)
4.65 ((cid:0) 7.3, 18.1)
24.6 (11.9, 38.7)
(cid:0) 1.19 ((cid:0) 8.18, 6.34)
14.6 (1.61, 29.2)
(cid:0) 0.01 ((cid:0) 6.85, 7.32)
14.6 (1.58, 29.2)
(cid:0) 0.39 ((cid:0) 7.31, 7.05)
3.87 ((cid:0) 4.38, 12.8)
14.0 (3.29, 25.9)
3.65 ((cid:0) 4.63, 12.6)
14.0 (2.97, 26.2)
3.22 ((cid:0) 5.21, 12.4)
13.6 (2.50, 25.8)
11.2 (1.39, 21.9)
0.58 ((cid:0) 5.36, 6.9)
12.5 (1.31, 24.8)
0.74 ((cid:0) 5.25, 7.11)
11.6 (0.48, 23.9)
0.90 ((cid:0) 5.20, 7.39)
4.41 ((cid:0) 4.25, 13.9)
3.99 ((cid:0) 6.37, 15.5)
4.14 ((cid:0) 4.50, 13.6)
3.80 ((cid:0) 6.75, 15.5)
5.02 ((cid:0) 3.92, 14.8)
3.26 ((cid:0) 7.37, 15.1)
7.76 ((cid:0) 2.20, 18.7)
0.37 ((cid:0) 5.85, 6.99)
8.34 ((cid:0) 2.91, 20.9)
0.99 ((cid:0) 5.28, 7.68)
8.74 ((cid:0) 2.64, 21.4)
1.83 ((cid:0) 4.64, 8.73)
6.46 ((cid:0) 3.76, 17.8)
3.40 ((cid:0) 8.43, 16.8)
7.34 ((cid:0) 2.96, 18.7)
6.22 ((cid:0) 6.13, 20.2)
7.91 ((cid:0) 2.56, 19.5)
5.18 ((cid:0) 7.08, 19.1)
4.15 ((cid:0) 7.21, 16.9)
6.94 ((cid:0) 0.71, 15.2)
11.8 ((cid:0) 1.63, 27.1)
6.69 ((cid:0) 0.94, 14.9)
12.2 ((cid:0) 1.21, 27.4)
6.75 ((cid:0) 0.96, 15.1)
7.86 ((cid:0) 6.57, 24.5)
(cid:0) 5.91 ((cid:0) 21.0, 12.1)
5.52 ((cid:0) 8.43, 21.6)
(cid:0) 12.1 ((cid:0) 26.4, 4.88)
5.92 ((cid:0) 8.40, 22.5)
(cid:0) 13.4 ((cid:0) 27.5, 3.47)
16.0 ((cid:0) 1.03, 36.0)
4.49 ((cid:0) 12.7, 25.0)
¡16.1 (-24.4, -6.84) ¡15.0 (-23.3, -5.78) ¡15.1 (-23.6, -5.69)
4.80 ((cid:0) 12.5, 25.5)
′
Abbreviations: 8-OHdG: 8-hydroxydeoxyguanosine, 5-mC: 5-methylcytosine, 5-
′
mdC: 5-Methyl-2
-deoxycytidine,
-deoxycytidine, 5-hmdC: 5-hydroxymethyl-2
7-mG: 7-methylguanine, 3-mA: 3-methyladenine, 3-PBA: 3-phenoxybenzoic
acid, t/c-DCCA: trans/cis-3-(2,2-dichlorovinyl)-2,2-dimethylcyclopro-pane car-
acid, TCPY: 3,5,6-trichloro-2-pyridinol, TEBOH: hydroxy-1-
boxylic
tebuconazoleEstimates from linear mixed effects models with random in-
tercepts for participant ID and households (n = 440, 220 subjects). Levels of
biomarkers were ln transformed.
a Base model, adjusted for age, BMI, sex, specific gravity.
b Adjusted model 1, adjusted for the same variables as Base model + season,
agricultural area.
c Adjusted mode 2, adjusted for the same variables as Adjusted model 1 + fruit
consumption, vegetable consumption, organic food consumption.
* 95% confidence intervals were false coverage-statement adjusted to account
for multiple testingIQR 3-PBA = 0.560 ng/mL; IQR t/c-DCCA = 1.46 ng/mL; IQR
TCPY = 4.15 ng/mL; IQR TEB-OH = 0.567 ng/mL.
model additionally adjusted for multiple urinary CUP metabolites (SI
Table 8).
4. Discussion
In the population of Czech adults and children from the CELSPAC-
SPECIMEn cohort, we examined urinary levels of DNA methylation
and oxidative stress biomarkers in repeatedly collected samples from the
winter and summer seasons. Urinary levels of CUP metabolites were also
examined and higher levels in children’s urine were found in compari-
son to adult samples. In adults, urinary CUP levels were often similar
among seasons, in children higher in winter. Detailed discussion of the
ˇ
Sulc et al., 2022). We observed that levels
results is provided elsewhere (
of all response biomarkers were higher in children’s urine samples in
both seasons. It is reasonable to expect, as age-related DNA methylation
patterns were reported to have regulatory roles on gene activity and
developmental processes. Hence the increased levels of response bio-
markers in children are mainly caused due to extensive development
during childhood (Gervin et al., 2016). The shortened period for DNA
repair and the multiple changes that are occurring within DNA, together
with different toxicokinetics of many environmental pollutants, could
also lead to increased susceptibility and vulnerability to environmental
pollutants in children, subsequently leading to increased DNA alter-
ations (Bearer, 1995).
The children’s 8-OHdG levels found in this study were slightly lower
than those reported among children from Japan (Ait Bamai et al., 2019),
Uruguay (Kordas et al., 2018) and the US and Canada (Jacobson et al.,
2020). However, similar levels to our study were observed in Chinese
young children (Wei et al., 2022). Such slight deviations in urinary
8-OHdG levels could be explained by the usage of different methods with
higher levels reported for immunological techniques compared to
chemical methods (Graille et al., 2020). As the immunochemical
enzyme-linked immunosorbent assays (ELISA) has several analytical
limitations, chromatographic methods are considered to be the gold
standard (Bl´ahov´a et al., 2023 – manuscript submitted). The same is true
when comparing the adult 8-OHdG levels, which are mostly in line with
previous studies when considering chemical analytical techniques and
slightly lower compared to immunological ones (Graille et al., 2020).
Data on methylated DNA bases in the urine of general population are
sparse so far. Although their usage could be very useful. By measuring
urinary levels of these biomarkers, changes of the DNA methylation
status in the whole body could be assessed and the DNA demethylation
mechanisms could be investigated in vivo (Hu et al., 2012). In addition,
such a non-invasive measurement could serve as a useful biomarker of
exposure to methylating agents or other xenobiotics (Hu et al., 2011;
Lovreglio et al., 2020) and as a promising early biomarker of several
disorders (Pan et al., 2016; Onishi et al., 2019). There are few studies in
occupationally exposed populations or in population sub-groups. In the
occupational study of Lovreglio et al. (2020), control group (n = 93)
from Italy showed the same levels of 7-mG and 5-mC, but lower levels of
5-mdC (median 2.77 μg/g crea.). Contrary, urinary levels of 5-mdC and
5-hmdC of male partners (n = 562) of subfertile couples from China
were in line with our findings (Pan et al., 2016). Higher urinary levels of
5-mC, 7-mG, and lower levels of 5-mdC, 3-mA were reported in the study
of 376 healthy male subjects from Taiwan, but still within the same
order of magnitude (Hu et al., 2012). These slight discrepancies might be
linked to the study population, as it has already been proposed that
global DNA methylation patterns differ with subject and lifestyle char-
acteristics, such as age, gender, alcohol drinking (Zhu et al., 2012), or
subject health status (Robertson, 2005). Considering the health status, in
these specific cases, urinary levels of 5-mdC significantly differed with
progression of chronic kidney disease (Onishi et al., 2019) and those of
3-mA and 7 mG between Parkinson’s Disease and Parkinsonian Syn-
dromes Patients compared to control group (Gątarek et al., 2020). To the
best of our knowledge, there is no previous study in children population
investigating urinary biomarkers of potential nucleic acid methylation.
The repeated sampling design of our study allows us to do the first
comparison of levels of oxidative stress and nucleic acid methylation
biomarkers of general population in two different seasons (winter vs.
summer). Increased concentrations could be seen particularly in urine
samples from the winter season. There are several possible explanations
for the observed pattern. Many seasonal factors such as temperature and
light could affect transcriptional mechanisms via DNA methylation
(Alvarado et al., 2014). The total antioxidant capacity of a human sys-
tem was also proven to vary seasonally with significantly greater ca-
pacity in the summer season (Morera-Fumero et al., 2018). This could be
the consequence of different dietary patterns between summer and
winter seasons, especially considering fruit and vegetable consumption
as a principal source of an antioxidative potential in a diet (Capita and
EnvironmentalResearch222(2023)1153685 T. Janoˇs et al.
Alonso-Calleja, 2005; Człapka-Matyasik and Ast, 2014). Last, but not
least, exposure to some environmental chemicals is expected to be dis-
similar between seasons due to distinct behavioral habits. As demon-
strated in the case of exposure to CUPs in the present study and discussed
ˇ
in detail previously (
Sulc et al., 2022) or in the case of multiple volatile
organic compounds due to increased time spent indoors and reduced
ventilation during winter season (Paciˆencia et al., 2016).
The significant, robust (across both seasons in both adults and chil-
dren) correlations were found among some biomarkers. The positive
correlations of urinary 5-mdC with its oxidized form 5-hmdC and
nucleobase 5-mC are consistent among other studies (Hu et al., 2012;
Pan et al., 2016) because all of them may be urinary products of
methylation within the cytosine nucleotide. Urinary 5-mdC was, along
with 5-mC, further correlated with urinary biomarker of oxidative stress
8-OHdG. This suggests that increased oxidative stress may exhaust
antioxidant activity which is biochemically linked to biosynthesis of
S-adenosylmethionine (SAM), an important methyl donor for DNA
methylation and thus induce increased methylation (de Prins et al.,
2013). Whereas methylation of the C-5 position of cytosine is predom-
inantly catalyzed enzymatically, 3-mA and 7-mG are products of
nonenzymatic DNA methylation. Therefore, no conclusive pattern was
found when considering correlations with 7-mG and 3-mA biomarkers.
However, Hu et al., 2012 found significant associations of 5-mC with
3-mA and 7-mG biomarkers, confirming that SAM, as a methyl donor for
enzymatic methylation, may play an important role in nonenzymatic
methylation as well.
The effects of CUPs on oxidative stress and DNA methylation in
Czech adults and children were examined using LME model. An
important observation in our study was the positive association of py-
rethroid metabolites (3-PBA and t/c-DCCA) with the urinary level of the
oxidative stress biomarker. The effects of pyrethroid exposure on
oxidative DNA damage have already been explored by animal exposure
studies and proposed as one of the mechanisms linked to pesticide-
induced chronic diseases (Banerjee et al., 2001). In such cases,
increased levels of oxidative stress biomarkers and/or enzymes and
decreased levels of antioxidants suggest the involvement of pyrethroid
pesticides in oxidative stress generation (Kale et al., 1999; Aouey et al.,
2017). Similar changes in antioxidant enzymes and biomarkers of
oxidative stress were observed in studies from occupational setting
(Sharma et al., 2013; Zepeda-Arce et al., 2017). Nevertheless, results are
inconclusive despite the fact that agricultural workers may by constantly
exposed to remarkably higher levels of pyrethroids. In general popula-
tion, studies of CUPs in relation to oxidative stress biomarkers are
limited. The only general population study among primary school chil-
dren in Cyprus brings consistent results with our study (Makris et al.,
2022). Using creatinine-adjusted biomarkers of exposure and effect,
urinary levels of 8-OHdG were significantly associated with 3-PBA
metabolite (β = 0.19, 95% CI: 0.02, 0.37) but at the edge of signifi-
cance with t/c-DCCA (β = 0.12, 95% CI: 0.02, 0.27 and β = 0.12, 95% CI:
0.02, 0.25 for cis- and trans-respectively). A significant association was
also observed for the chlorpyrifos metabolite TCPY (β = 0.42, 95% CI:
0.16, 0.68) which is in agreement with other studies exploring effects of
organophosphate pesticides on oxidative stress. Increased urinary levels
of 8-OHdG were reported on the first day after chlorpyrifos spraying in
the case of farmers (Wang et al., 2016) and decreased levels of gluta-
thione, which is part of an antioxidant system, were found among
children in the agricultural community compared to the urban com-
munity (Sapbamrer et al., 2020). On the contrary, urinary levels of TCPY
were not associated with urinary levels of 8-OHdG in our study which
may be related to relatively strict parameters of our models (multiple
adjustment variables, multiple testing correction) compared with above
mentioned studies.
In addition to indicating effects of pyrethroids on oxidative stress,
our results suggest that CUP exposure might induce changes in DNA
methylation patterns (either hyper- or hypo-methylation). The proposed
mechanism of environmental chemicals action consists mainly of an
altered function of methyl donor SAM and enzyme DNA methyl-
transferases, which catalyzes the transfer of the methyl group (Ruiz--
Hernandez et al., 2015). This is supported by few epidemiological
studies which have reported altered DNA methylation levels of specific
gene promoters in response to pesticide mixture exposure (Rusiecki
et al., 2017; Declerck et al., 2017, 2017van der Plaat et al., 2018;
Benitez-Trinidad et al., 2018). Considering global DNA methylation
changes, Benedetti et al. (2018) carried out the study on soybean
farmers who were actively engaged in the preparation and application of
the complex mixture of pesticides. The results showed a significant
difference in the percentage of global DNA methylation in individuals
exposed to pesticides compared to the control group. Whereas most of
the studies are usually conducted in an occupational setting and are not
focused on specific CUPs, general population evidence is limited. In the
present investigation, the associations between urinary biomarkers of
DNA methylation and CUP exposure were studied for the first time. The
robust associations of 5-mC and its deoxynucleoside 5-mdC with
chlorpyrifos and pyrethroid metabolites were observed across all
models. Nevertheless, proportion of imputed data in some study strata
(particularly TCPY in adults in summer) was high which could poten-
tially lead to bias. Furthermore, higher urinary levels of TEB-OH were
importantly associated with lower levels of urinary 3-mA indicating a
potential role of tebuconazole in hypomethylation or decreased deme-
thylation of DNA. Tebuconazole is often used in a mixture with pro-
thioconazole which are triazole and triazole-thionine based fungicides,
respectively. Both of them are used to control fungal plant diseases of
major crops like cereals and canola or for seed treatment (Jørgensen and
Heick, 2021) and are the most frequently used azole pesticides in the
Czech Republic (CISTA, 2022). The possible mechanism of their effect
may consist of alkyladenine-DNA glycosylase (AAG) inhibition. It is well
known that excision repair of 3-mA in DNA is initiated by AAG enzyme
with subsequential generation of an apurinic/apyrimidinic site (Fu
et al., 2012). Simultaneously, potential inhibition of AAG by
triazole-thione-based compounds is proposed (Al and Ba, 2017) which
could lead to decreased demethylation processes and therefore to
increased methylation of adenine in DNA and decreased presence of
3-mA in urine.
When comparing effect estimates from all diversly adjusted models,
only notable change was observed in the case of association between
chlorpyrifos metabolite and 5-mC. Decrease of the effect estimate is
expected to be caused by season adjustment as significant seasonal dif-
ferences were observed in both biomarkers (see Table 2). Robustness of
the results when replicating LME models in sensitivity analyses showed
that associations are not influenced by method used for urine dilution
adjustment, by outliers neither by multiple urinary CUP metabolites.
There are some limitations to this study. Firstly, as measurement of
urinary response biomarkers reflects the results of DNA repair processes
in the “whole body”, we are not able to interpret the results as an
epigenetic change within regional or individual genes (e.g. DNA
hypermethylation of tumor suppressor genes), which could be useful
when associating with specific health outcomes. Secondly, both expo-
sure and response biomarkers were measured in a first morning void
urine samples and thus may not be fully representative of a long-term
temporal variability in the given season. Thirdly, considering environ-
mental complexity, observed associations could be confounded by other
unmeasured environmental chemical exposures or other factors, despite
the fact that potential confounding variables were carefully selected.
Finally, the data on CUP biomarkers were imputed. Although imputa-
tion is common scientific practice, imputation cutoff (40%) could
potentially influence the results. This should be considered especially in
the case of TCPY. As we are primarily interested in the detection of
possible new associations, rather than confirming with certainty hy-
pothesized associations, we accepted the risk of potential false-positive
results. On the other hand, these limitations are countered by a num-
ber of strengths. Mainly, repeated measurements in winter and summer
in both adults and children allowed us to cover variability in both
EnvironmentalResearch222(2023)1153686 T. Janoˇs et al.
exposure and response biomarkers across the seasons and population
subgroups. Furthermore, the wide scope of response biomarkers,
assessed using a more precise mass spectrometry method, enhanced our
ability to reveal more possible effects in the human body.
5. Conclusion
In conclusion, to the best of our knowledge, this is the first study to
measure and describe biomarkers of DNA methylation and oxidative
stress in urine samples of Czech adult population and the first in children
overall. In addition, it is the first epidemiological study to assess the
associations of urinary biomarkers of response with CUP exposure. We
observed significant, robust associations across all assessed models.
Pyrethroid metabolites were associated with higher levels of both
oxidative stress and DNA methylation biomarkers. Moreover, urinary
levels of the chlorpyrifos metabolite were also associated with urinary
products of methylation within the cytosine nucleotide. Finally, the
most robust, negative association was observed between the tebucona-
zole metabolite and 3-methyladenine indicating a possible role of azole
pesticides in demethylation processes. These findings suggest an urgent
need to extend the range of analyzed environmental chemicals such as
azole pesticides (for instance prothioconazole) in human biomonitoring
studies to responsibly evaluate associated health risks. In addition,
observed associations warrant further large-scale research of these bio-
markers and environmental pollutants including CUPs.
Credit author statement
Tom´aˇs Janoˇs: Conceptualization, Methodology, Formal analysis,
Investigation, Data curation, Writing – original draft, Visualization, Ilse
Ottenbros: Conceptualization, Methodology, Writing – review & edit-
ing, Lucie Bl´ahov´a: Methodology, Validation, Investigation, Writing –
ˇ
review & editing, Petr
Senk: Methodology, Validation, Investigation,
ˇ
Writing – review & editing, Libor
Sulc: Methodology, Writing – review
& editing, Nina P´aleˇsov´a: Investigation, Writing – review & editing,
Jessica Sheardov´a: Writing – review & editing, Visualization, Jelle
Vlaanderen: Conceptualization, Methodology, Resources, Writing –
ˇ
review & editing, Supervision, Pavel
Cupr: Conceptualization, Meth-
odology, Resources, Writing – review & editing, Supervision, Project
administration, Funding acquisition.
Declaration of competing interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Data availability
Data will be made available on request.
Acknowledgment
This project received funding from the European Union’s Horizon
2020 research and innovation program under grant agreement No
733032, grant agreement No 857340, grant agreement No 874627 and
grant agreement No 857560. The authors thank Research Infrastructure
RECETOX RI (No LM2018121) financed by the Ministry of Education,
Youth and Sports, and Operational Programme Research, Development
and Innovation – project CETOCOEN EXCELLENCE (No CZ.02.1.01/
0.0/0.0/17_043/0009632) for supportive background. This publication
reflects only the author’s view and the European Commission is not
responsible for any use that may be made of the information it contains.
We would like to thank Richard Hůlek, Mazen Ismael, Zuzana Luhov´a
and Jiˇrí Bilík from RECETOX Information systems and data services for
the preparation of a data warehouse and infastructure to store and
manage the CELSPAC-SPECIMEn study data. We thank Lenka
Andrýskov´a for her help when addressing the ethical aspects of the
study. The authors thank Ondˇrej Mikeˇs for his help with the preparation
ˇ
Sebejov´a,
of exposure questionnaires. The authors thank Ludmila
Zuzana Jaˇskov´a and Lenka Koci´anov´a from CELSPAC Biobank for sup-
port with the preparation of urine aliquots and for storing samples in the
CELSPAC biobank facility. In addition, we are grateful to Roman Prokeˇs
and Jakub Vinkler for collecting the samples.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.envres.2023.115368.
References
Ait Bamai, Y., Bastiaensen, M., Araki, A., Goudarzi, H., Konno, S., Ito, S., Miyashita, C.,
Yao, Y., Covaci, A., Kishi, R., 2019. Multiple exposures to organophosphate flame
retardants alter urinary oxidative stress biomarkers among children: the hokkaido
study, vol. 131. Environ Int. https://doi.org/10.1016/j.envint.2019.105003
Al, Yahyaei, Ba, 2017. Design and Synthesis of Triazole-Based Inhibitors of the DNA
Repair Enzyme Alkyladenine Glycosylase (AAG). [Doctoral dissertation, University
of Surrey. British Library. https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos
.751834.
Alvarado, S., Fernald, R.D., Storey, K.B., Szyf, M., 2014. The dynamic nature of DNA
methylation. Comparative Biology 54 (1), 68–76. https://doi.org/10.2307/
26369481.
Aouey, B., Derbali, M., Chtourou, Y., Bouchard, M., Khabir, A., Fetoui, H., 2017.
Pyrethroid insecticide lambda-cyhalothrin and its metabolites induce liver injury
through the activation of oxidative stress and proinflammatory gene expression in
rats following acute and subchronic exposure. Environ. Sci. Pollut. Control Ser. 24
(6), 5841–5856. https://doi.org/10.1007/s11356-016-8323-4.
Banerjee, B.D., Seth, V., Ahmed, R.S., 2001. Pesticide-induced oxidative stress:
perspectives and trends. Rev. Environ. Health 16 (1), 1–40. https://doi.org/
10.1515/reveh.2001.16.1.1.
Barr, D.B., Wilder, L.C., Caudill, S.P., Gonzalez, A.J., Needham, L.L., Pirkle, J.L., 2005.
Urinary creatinine concentrations in the U.S. population: implications for urinary
biologic monitoring measurements. EHP (Environ. Health Perspect.) 113 (2),
192–200. https://doi.org/10.1289/ehp.7337.
Bearer, C.F., 1995. How are children different from adults? Environ. Health Perspect.
103 (Suppl. 6), 7–12.
Becker, K., Seiwert, M., Angerer, J., Kolossa-Gehring, M., Hoppe, H.W., Ball, M.,
Schulz, C., Thumulla, J., Seifert, B., 2006. GerES IV Pilot Study: assessment of the
exposure of German children to organophosphorus and pyrethroid pesticides. Int. J.
Hyg Environ. Health 209 (3), 221–233. https://doi.org/10.1016/j.
ijheh.2005.12.002.
Benedetti, D., Lopes Alderete, B., de Souza, C.T., Ferraz Dias, J., Niekraszewicz, L.,
Cappetta, M., Martínez-L´opez, W., da Silva, J., 2018. DNA damage and epigenetic
alteration in soybean farmers exposed to complex mixture of pesticides. Mutagenesis
33 (1), 87–95. https://doi.org/10.1093/mutage/gex035.
Benitez-Trinidad, A.B., Medina-Díaz, I.M., Bernal-Hern´andez, Y.Y., Barr´on-Vivanco, B.S.,
Gonz´alez-Arias, C.A., Herrera-Moreno, J.F., Alvarado-Cruz, I., Quintanilla-Vega, B.,
Rojas-García, A.E., 2018. Relationship between LINE-1 methylation pattern and
pesticide exposure in urban sprayers. Food Chem. Toxicol. 113, 125–133. https://
doi.org/10.1016/j.fct.2018.01.035.
Benjamini, Y., Yekutieli, D., 2005. False discovery rate-adjusted multiple confidence
intervals for selected parameters. J. Am. Stat. Assoc. 100 (469), 71–81. https://doi.
org/10.1198/016214504000001907.
Bl´ahov´a, L., Janoˇs, T., Mustieles, V., Rodríguez-Carrillo, A., Fern´andez, M.F., Bl´aha, L.,
2023. Rapid extraction and analysis of oxidative stress and DNA damage biomarker
8- hydroxy-2’-deoxyguanosine (8-OHdG) in urine: application to a study with
pregnant women. Int. J. Hyg. Environ. Health. Under review.
Capita, R., Alonso-Calleja, C., 2005. Differences in reported winter and summer dietary
intakes in young adults in Spain. Int. J. Food Sci. Nutr. 56 (6), 431–443. https://doi.
org/10.1080/09637480500407875.
EFSA (European Food Safety Authority), Carrasco Cabrera, L., Medina Pastor, P., 2021.
The 2019 European Union report on pesticide residues in food. EFSA J. 19 (4)
https://doi.org/10.2903/j.efsa.2021.6491, 6491, 89.
Chang, C., Chen, M., Gao, J., Luo, J., Wu, K., Dong, T., Zhou, K., He, X., Hu, W., Wu, W.,
Lu, C., Hang, B., Meeker, J.D., Wang, X., Xia, Y., 2017. Current pesticide profiles in
blood serum of adults in Jiangsu Province of China and a comparison with other
countries. Environ. Int. 102, 213–222. https://doi.org/10.1016/j.
envint.2017.03.004.
Chen, Z.X., Riggs, A.D., 2011. DNA methylation and demethylation in mammal. J. Biol.
Chem. 286 (21), 18347–18353. https://doi.org/10.1074/jbc.R110.205286.
Cista, 2022. Pesticide Use in Czech Republic [WWW Document]. Czech Repub. – Usage
Act. Subst. 2020. URL. https://eagri.cz/public/web/ukzuz/portal/pripravky-na-or
/ucinne-latky-v-por-statistika-spotreba/spotreba-pripravku-na-or/spotreba-v-jednot
livych-letech/. accessed 7.14.22.
Collotta, M., Bertazzi, P.A., Bollati, V., 2013. Epigenetics and pesticides. Toxicology 307,
35–41. https://doi.org/10.1016/j.tox.2013.01.017.
EnvironmentalResearch222(2023)1153687 T. Janoˇs et al.
Człapka-Matyasik, M., Ast, K., 2014. Total antioxidant capacity and its dietary sources
and seasonal variability in diets of women with different physical activity levels. Pol.
J. Food Nutr. Sci. 64 (4), 267–276. https://doi.org/10.2478/pjfns-2013-0026.
de Prins, S., Koppen, G., Jacobs, G., Dons, E., van de Mieroop, E., Nelen, V., Fierens, F.,
Int Panis, L., de Boever, P., Cox, B., Nawrot, T.S., Schoeters, G., 2013. Influence of
ambient air pollution on global DNA methylation in healthy adults: a seasonal
follow-up. Environ. Int. 59, 418–424. https://doi.org/10.1016/j.
envint.2013.07.007.
Declerck, K., Remy, S., Wohlfahrt-Veje, C., Main, K.M., van Camp, G., Schoeters, G.,
vanden Berghe, W., Andersen, H.R., 2017. Interaction between prenatal pesticide
exposure and a common polymorphism in the PON1 gene on DNA methylation in
genes associated with cardio-metabolic disease risk—an exploratory study. Clin.
Epigenet. 9 (1) https://doi.org/10.1186/s13148-017-0336-4.
Dereumeaux, C., Saoudi, A., Goria, S., Wagner, V., de Crouy-Chanel, P., Pecheux, M.,
Berat, B., Zaros, C., Guldner, L., 2018. Urinary levels of pyrethroid pesticides and
determinants in pregnant French women from the Elfe cohort. Environ. Int. 119,
89–99. https://doi.org/10.1016/J.ENVINT.2018.04.042.
Dereumeaux, C., Fillol, C., Quenel, P., Denys, S., 2020. Pesticide exposures for residents
living close to agricultural lands: a review. Environ. Int. 134, 105210 https://doi.
org/10.1016/j.envint.2019.105210.
di Minno, A., Turnu, L., Porro, B., Squellerio, I., Cavalca, V., Tremoli, E., di Minno, M.N.
D., 2016. 8-Hydroxy-2-Deoxyguanosine levels and cardiovascular disease: a
systematic review and meta-analysis of the literature. Antioxidants Redox Signal. 24
(Issue 10), 548–555. https://doi.org/10.1089/ars.2015.6508. Mary Ann Liebert Inc.
Fleming, A.M., Alshykhly, O., Zhu, J., Muller, J.G., Burrows, C.J., 2015. Rates of
chemical cleavage of DNA and RNA oligomers containing guanine oxidation
products. Chem. Res. Toxicol. 28 (6), 1292–1300. https://doi.org/10.1021/acs.
chemrestox.5b00096.
Fu, D., Calvo, J.A., Samson, L.D., 2012. Balancing repair and tolerance of DNA damage
caused by alkylating agents. Nat. Rev. Cancer 12 (Issue 2), 104–120. https://doi.org/
10.1038/nrc3185.
Gątarek, P., Kału˙zna-Czapli´nska, J., Pawełczyk, M., Jastrzębski, K., Giebułtowicz, J.,
Głąbi´nski, A., Bobrowska-Korczak, B., 2020. LC-MS/MS determination of modified
nucleosides in the urine of Parkinson’s disease and parkinsonian Syndromes
Patients. Molecules 25 (21). https://doi.org/10.3390/molecules25214959.
Gehring, M., Reik, W., Henikoff, S., 2009. DNA demethylation by DNA repair. Trends
Genet. 25 (Issue 2), 82–90. https://doi.org/10.1016/j.tig.2008.12.001.
Gervin, K., Andreassen, B.K., Hjorthaug, H.S., Carlsen, K.C.L., Carlsen, K.H., Undlien, D.
E., Lyle, R., Munthe-Kaas, M.C., 2016. Intra-individual changes in DNA methylation
not mediated by cell-type composition are correlated with aging during childhood.
Clin. Epigenet. 8 (1) https://doi.org/10.1186/s13148-016-0277-3.
Gonz´alez-Alzaga, B., Lacasa˜na, M., Aguilar-Gardu˜no, C., Rodríguez-Barranco, M.,
Ballester, F., Rebagliato, M., Hern´andez, A.F., 2014. A systematic review of
neurodevelopmental effects of prenatal and postnatal organophosphate pesticide
exposure. Toxicol. Lett. 230 (2), 104–121. https://doi.org/10.1016/j.
toxlet.2013.11.019.
Graille, M., Wild, P., Sauvain, J.J., Hemmendinger, M., Canu, I.G., Hopf, N.B., 2020.
Urinary 8-OHDG as a biomarker for oxidative stress: a systematic literature review
and meta-analysis. Int. J. Mol. Sci. 21 (11), 1–25. https://doi.org/10.3390/
ijms21113743.
Guo, C., Ding, P., Xie, C., Ye, C., Ye, M., Pan, C., Cao, X., Zhang, S., Zheng, S., 2017.
Potential application of the oxidative nucleic acid damage biomarkers in detection of
diseases. Oncotarget 8 (Issue 43). Oncotarget 75767. www.impactjournals.com/o
ncotarget. www.impactjournals.com/oncotarget/.
Hu, C.W., Lin, B.H., Chao, M.R., 2011. Quantitative determination of urinary N3-
methyladenine by isotope-dilution LC-MS/MS with automated solid-phase
extraction. Int. J. Mass Spectrom. 304 (2–3), 68–73. https://doi.org/10.1016/j.
ijms.2010.04.003.
′
Hu, C.W., Liu, H.H., Li, Y.J., Chao, M.R., 2012. Direct analysis of 5-methylcytosine and 5-
-deoxycytidine in human urine by isotope dilution LC-MS/MS: correlations
methyl-2
with N-methylated purines and oxidized DNA lesions. Chem. Res. Toxicol. 25 (2),
462–470. https://doi.org/10.1021/tx2004954.
Jacobson, M.H., Wu, Y., Liu, M., Attina, T.M., Naidu, M., Karthikraj, R., Kannan, K.,
Warady, B.A., Furth, S., Vento, S., Trachtman, H., Trasande, L., 2020. Serially
assessed bisphenol A and phthalate exposure and association with kidney function in
children with chronic kidney disease in the US and Canada: a longitudinal cohort
study. PLoS Med. 17 (10) https://doi.org/10.1371/journal.pmed.1003384.
Jørgensen, L.N., Heick, T.M., 2021. Azole use in agriculture, horticulture, and wood
preservation – is it indispensable?. In: Frontiers in Cellular and Infection
Microbiology, vol. 11 Frontiers Media S.A. https://doi.org/10.3389/
fcimb.2021.730297.
Kale, M., Rathore, N., John, S., Bhatnagar, D., 1999. Lipid peroxidative damage on
pyrethroid exposure and alterations in antioxidant status in rat erythrocytes: a
possible involvement of reactive oxygen species. Toxicol. Lett. 105.
Kim, K.H., Kabir, E., Jahan, S.A., 2017. Exposure to pesticides and the associated human
health effects. Sci. Total Environ. 575, 525–535. https://doi.org/10.1016/j.
scitotenv.2016.09.009.
Kordas, K., Roy, A., Vahter, M., Ravenscroft, J., Ma˜nay, N., Peregalli, F., Martínez, G.,
Queirolo, E.I., 2018. Multiple-metal exposure, diet, and oxidative stress in
Uruguayan school children. Environ. Res. 166, 507–515. https://doi.org/10.1016/j.
envres.2018.06.028.
Koureas, M., Tsakalof, A., Tsatsakis, A., Hadjichristodoulou, C., 2012. Systematic review
of biomonitoring studies to determine the association between exposure to
organophosphorus and pyrethroid insecticides and human health outcomes. Toxicol.
Lett. 210 (2), 155–168. https://doi.org/10.1016/j.toxlet.2011.10.007.
Lovreglio, P., Stufano, A., Andreoli, R., Tomasi, C., Cagnazzi, P., Barbieri, A., Soleo, L., de
Palma, G., 2020. Urinary biomarkers of nucleic acid oxidation and methylation in
workers exposed to low concentrations of benzene. Toxicol. Lett. 331, 235–241.
https://doi.org/10.1016/j.toxlet.2020.06.010.
Lubin, J.H., Colt, J.S., Camann, D., Davis, S., Cerhan, J.R., Severson, R.K., Bernstein, L.,
Hartge, P., 2004. Epidemiologic evaluation of measurement data in the presence of
detection limits. EHP (Environ. Health Perspect.) 112 (17), 1691–1696. https://doi.
org/10.1289/ehp.7199.
Makris, K.C., Efthymiou, N., Konstantinou, C., Anastasi, E., Schoeters, G., Kolossa-
Gehring, M., Katsonouri, A., 2022. Oxidative stress of glyphosate, AMPA and
metabolites of pyrethroids and chlorpyrifos pesticides among primary school
children in Cyprus. Environ. Res. 212 https://doi.org/10.1016/j.
envres.2022.113316.
Md Meftaul, I., Venkateswarlu, K., Dharmarajan, R., Annamalai, P., Megharaj, M., 2020.
Pesticides in the urban environment: a potential threat that knocks at the door. Sci.
Total Environ. 711, 134612 https://doi.org/10.1016/j.scitotenv.2019.134612.
Morera-Fumero, A.L., Abreu-Gonzalez, P., Henry-Benitez, M., Fernandez-Lopez, L., Diaz-
Mesa, E., del Rosario Cejas-Mendez, M., Guillen-Pino, F., 2018. Day/night and
summer/winter changes in serum total antioxidant capacity. Med. Chem. 14 (3),
225–229. https://doi.org/10.2174/1573406413666171002123158.
Neofytou, E., Tzortzaki, E.G., Chatziantoniou, A., Siafakas, N.M., 2012. DNA damage due
to oxidative stress in chronic obstructive pulmonary disease (COPD). Int. J. Mol. Sci.
13 (12), 16853–16864. https://doi.org/10.3390/ijms131216853.
Nougad`ere, A., Sirot, V., Kadar, A., Fastier, A., Truchot, E., Vergnet, C., Hommet, F.,
Bayl´e, J., Gros, P., Leblanc, J.C., 2012. Total diet study on pesticide residues in
France: levels in food as consumed and chronic dietary risk to consumers. Environ.
Int. 45 (1), 135–150. https://doi.org/10.1016/j.envint.2012.02.001.
Oerlemans, A., Figueiredo, D.M., Mol, J.G.J., Nijssen, R., Anzion, R.B.M., van Dael, M.F.
P., Duyzer, J., Roeleveld, N., Russel, F.G.M., Vermeulen, R.C.H., Scheepers, P.T.J.,
2021. Personal exposure assessment of pesticides in residents: the association
between hand wipes and urinary biomarkers. Environ. Res. 199, 111282 https://doi.
org/10.1016/J.ENVRES.2021.111282.
Onishi, A., Sugiyama, H., Kitagawa, M., Yamanari, T., Tanaka, K., Ogawa-Akiyama, A.,
Kano, Y., Mise, K., Tanabe, K., Morinaga, H., Kinomura, M., Uchida, H.A., Wada, J.,
2019. Urine 5MedC, a Marker of DNA Methylation, in the Progression of Chronic
Kidney Disease. Disease Markers. https://doi.org/10.1155/2019/5432453, 2019.
Ottenbros, I., Lebret, E., Huber, C., Lommen, A., Antignac, J.P.,
ˇ
Sulc, L.,
Mikeˇs, O., Szigeti, T., K¨oz´epesy, S., Martinsone, I., Martinsone, Z., Akulova, L.,
Pardo, O., Fern´andez, S.F., Coscoll´a, C., Pedraza-Diaz, S., Krauss, M., Debrauwer, L.,
et al., 2023. Assessment of exposure to pesticide mixtures in five European countries
by a harmonized urinary suspect screening approach. Int. J. Hyg Environ. Health vol.
248. https://doi.org/10.1016/j.ijheh.2022.114105. Elsevier GmbH.
ˇ
Cupr, P.,
Ougier, E., Ganzleben, C., Lecoq, P., Bessems, J., David, M., Schoeters, G., Lange, R.,
Meslin, M., Uhl, M., Kolossa-Gehring, M., Rousselle, C., Vicente, J.L., 2021. Chemical
prioritisation strategy in the European human biomonitoring initiative (HBM4EU) –
development and results. Int. J. Hyg Environ. Health 236, 113778. https://doi.org/
10.1016/j.ijheh.2021.113778.
Paciˆencia, I., Madureira, J., Rufo, J., Moreira, A., Fernandes, E. de O., 2016. A systematic
review of evidence and implications of spatial and seasonal variations of volatile
organic compounds (VOC) in indoor human environments. J. Toxicol. Environ.
Health B Crit. Rev. 19 (Issue 2), 47–64. https://doi.org/10.1080/
10937404.2015.1134371. Taylor and Francis Inc.
Pan, Y., Jing, J., Yeung, L.W.Y., Sheng, N., Zhang, H., Yao, B., Dai, J., 2016. Associations
′
′
-deoxycytidine with
-deoxycytidine and 5-hydroxymethyl-2
of urinary 5-methyl-2
phthalate exposure and semen quality in 562 Chinese adult men. Environ. Int. 94,
583–590. https://doi.org/10.1016/j.envint.2016.06.020.
R Core Team, 2021. R: A Language and Environment for Statistical Computing, 4.1.1.
Robertson, K.D., 2005. DNA methylation and human disease. Nat. Rev. Genet. 6 (8),
597–610. https://doi.org/10.1038/nrg1655.
Ruiz-Hernandez, A., Kuo, C.C., Rentero-Garrido, P., Tang, W.Y., Redon, J., Ordovas, J.M.,
Navas-Acien, A., Tellez-Plaza, M., 2015. Environmental chemicals and DNA
methylation in adults: a systematic review of the epidemiologic evidence. In Clinical
Epigenetics. Springer Verlag 7 (Issue 1). https://doi.org/10.1186/s13148-015-0055-
7.
Rusiecki, J.A., Beane Freeman, L.E., Bonner, M.R., Alexander, M., Chen, L., Andreotti, G.,
Barry, K.H., Moore, L.E., Byun, H.M., Kamel, F., Alavanja, M., Hoppin, J.A.,
Baccarelli, A., 2017. High pesticide exposure events and DNA methylation among
pesticide applicators in the agricultural health study. Environ. Mol. Mutagen. 58 (1),
19–29. https://doi.org/10.1002/em.22067.
Sabarwal, A., Kumar, K., Singh, R.P., 2018. Hazardous effects of chemical pesticides on
human health–Cancer and other associated disorders. Environ. Toxicol. Pharmacol.
63 (September), 103–114. https://doi.org/10.1016/j.etap.2018.08.018.
Saillenfait, A.M., Ndiaye, D., Sabat´e, J.P., 2015. Pyrethroids: exposure and health effects
- an update. Int. J. Hyg Environ. Health 218 (3), 281–292. https://doi.org/10.1016/
j.ijheh.2015.01.002.
Sapbamrer, R., Hongsibsong, S., Khacha-Ananda, S., 2020. Urinary organophosphate
metabolites and oxidative stress in children living in agricultural and urban
communities. Environ. Sci. Pollut. Control Ser. 27 (20), 25715–25726. https://doi.
org/10.1007/s11356-020-09037-z.
Sauv´e, J.F., L´evesque, M., Huard, M., Drolet, D., Lavou´e, J., Tardif, R., Truchon, G., 2015.
Creatinine and specific gravity normalization in biological monitoring of
occupational exposures. J. Occup. Environ. Hyg. 12 (2), 123–129. https://doi.org/
10.1080/15459624.2014.955179.
Sharma, R.K., Upadhyay, G., Siddiqi, N.J., Sharma, B., 2013. Pesticides-induced
biochemical alterations in occupational North Indian suburban population. Hum.
Exp. Toxicol. 32 (11), 1213–1227. https://doi.org/10.1177/0960327112474835.
EnvironmentalResearch222(2023)1153688 T. Janoˇs et al.
Shrier, I., Platt, R.W., 2008. Reducing bias through directed acyclic graphs. BMC Med.
Res. Methodol. 8, 1–15. https://doi.org/10.1186/1471-2288-8-70.
ˇ
Sulc, L., Janoˇs, T., Figueiredo, D., Ottenbros, I.,
ˇ
Senk, P., Mikeˇs, O., Huss, A.,
ˇ
Cupr, P.,
2022. Pesticide exposure among Czech adults and children from the CELSPAC-
SPECIMEn cohort: urinary biomarker levels and associated health risks. Environ.
Res. 214, 114002 https://doi.org/10.1016/j.envres.2022.114002.
′
Valavanidis, A., Vlachogianni, T., Fiotakis, C., 2009. 8-Hydroxy-2
-deoxyguanosine (8-
OHdG): a critical biomarker of oxidative stress and carcinogenesis. J. Environ. Sci.
Health C Environ. Carcinog. Ecotoxicol. Rev. 27 (2), 120–139. https://doi.org/
10.1080/10590500902885684.
van der Plaat, D.A., de Jong, K., de Vries, M., van Diemen, C.C., Nedeljkovic, I., Amin, N.,
Kromhout, H., Vermeulen, R., Postma, D.S., van Duijn, C.M., Boezen, H.M., Vonk, J.
M., 2018. Occupational exposure to pesticides is associated with differential DNA
methylation. Occup. Environ. Med. 75 (6), 427–435. https://doi.org/10.1136/
oemed-2017-104787.
Wang, L., Liu, Z., Zhang, J., Wu, Y., Sun, H., 2016. Chlorpyrifos exposure in farmers and
urban adults: metabolic characteristic, exposure estimation, and potential effect of
oxidative damage. Environ. Res. 149, 164–170. https://doi.org/10.1016/j.
envres.2016.05.011.
Wei, M., Wang, P., Wan, Y., Jiang, Y., Song, W., He, Z., Wang, Q., 2022. Urinary parabens
and their derivatives associated with oxidative stress biomarkers in children from
South and Central China: repeated measures. Sci. Total Environ. 817 https://doi.
org/10.1016/j.scitotenv.2021.152639.
Zepeda-Arce, R., Rojas-García, A.E., Benitez-Trinidad, A., Herrera-Moreno, J.F., Medina-
Díaz, I.M., Barr´on-Vivanco, B.S., Villegas, G.P., Hern´andez-Ochoa, I., S´olis
Heredia, M. de J., Bernal-Hern´andez, Y.Y., 2017. Oxidative stress and genetic
damage among workers exposed primarily to organophosphate and pyrethroid
pesticides. Environ. Toxicol. 32 (6), 1754–1764. https://doi.org/10.1002/
tox.22398.
Zhu, Z.Z., Hou, L., Bollati, V., Tarantini, L., Marinelli, B., Cantone, L., Yang, A.S.,
Vokonas, P., Lissowska, J., Fustinoni, S., Pesatori, A.C., Bonzini, M., Apostoli, P.,
Costa, G., Bertazzi, P.A., Chow, W.H., Schwartz, J., Baccarelli, A., 2012. Predictors of
global methylation levels in blood DNA of healthy subjects: a combined analysis. Int.
J. Epidemiol. 41 (1), 126–139. https://doi.org/10.1093/ije/dyq154.
EnvironmentalResearch222(2023)1153689
| null |
10.1371_journal.pone.0263447.pdf
|
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
|
All relevant data are within the manuscript and its Supporting Information files.
|
RESEARCH ARTICLE
Stakeholder perceptions of bird-window
collisions
Georgia J. RiggsID*, Omkar Joshi, Scott R. Loss
Department of Natural Resource Ecology and Management, Oklahoma State University, Stillwater,
Oklahoma, United States of America
* [email protected]
Abstract
Bird-window collisions are a major source of human-caused avian mortality for which many
mitigation and prevention options are available. However, because very little research has
characterized human perspectives related to this issue, there is limited understanding about
the most effective ways to engage the public in collision reduction efforts. To address this
research need, we: (1) evaluated how two stakeholder groups, homeowners and conserva-
tion practitioners, prioritize potential benefits and obstacles related to bird-window collision
management, (2) compared priorities between these groups, and (3) evaluated potential
conflicts and collective strength of opinions within groups. We addressed these objectives
by merging the strengths, weaknesses, opportunities, and threats (SWOT) and analytic
hierarchy process (AHP) survey approaches. Specifically, survey respondents made pair-
wise comparisons between strengths and weaknesses (respectively, direct outcomes and
barriers related to management, such as fewer collisions and increased costs) and opportu-
nities and threats (indirect outcomes and barriers, such as increased bird populations and
fewer resources for other building-related expenses). Both homeowners and conservation
practitioners ranked strengths and opportunities higher than weaknesses and threats, indi-
cating they have an overall positive perception toward reducing bird-window collisions. How-
ever, key obstacles that were identified included costs of management and a lack of policy
and guidelines to require or guide management. These results suggest that substantial
advances can be made to reduce bird-window collisions because both homeowners and
conservation practitioners had positive views, suggesting their receptivity toward collision
management measures. However, because of more neutral views and conflicting
responses within the homeowner group, results also highlight the importance of targeting
homeowners with education materials that provide information about bird-window collisions
and solutions that reduce them. Because bird-window collisions are a human-caused phe-
nomenon, such information about human perspectives and priorities will be crucial to
addressing this threat and thus benefitting bird populations.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Riggs GJ, Joshi O, Loss SR (2022)
Stakeholder perceptions of bird-window collisions.
PLoS ONE 17(2): e0263447. https://doi.org/
10.1371/journal.pone.0263447
Editor: Christopher A. Lepczyk, Auburn University,
UNITED STATES
Received: May 31, 2021
Accepted: January 19, 2022
Published: February 10, 2022
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0263447
Copyright: © 2022 Riggs et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: This research was funded by Oklahoma
State University Department of Natural Resource
Ecology and Management (https://go.okstate.edu/
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
1 / 20
PLOS ONE undergraduate-academics/majors/natural-
resource-ecology-and-management.html) and
Hatch Grant funding (grant numbers: OKL02915,
OKL03150) from the USDA National Institute of
Food and Agriculture (https://nifa.usda.gov/).
Funding was obtained by SRL. The funders had no
role in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Stakeholder perceptions of bird-window collisions
Introduction
As earth’s human population continues to grow [1], human actions and ways of life increas-
ingly affect wildlife and their habitats, and the many sources of unintended, direct wildlife
mortality are a major component of these human impacts [2–4]. Among direct sources of
avian mortality, collisions of birds with buildings and their windows are a top global threat.
Window collisions cause between 365 and 988 million bird deaths annually in the United
States alone [5] and are also a top threat to birds in other countries (e.g., Canada, Mexico, Bra-
zil, Spain, Singapore, South Korea) [6–11]. Birds collide with glass because they are unable to
perceive it as a barrier due to its reflective and transparent qualities [12], and because artificial
light at night confuses and draws migrating birds near buildings, elevating collision risk [13,
14]. Bird collisions occur at a wide variety of building types; tall buildings such as skyscrapers
have higher per-building collision rates, but smaller and far more abundant residential build-
ings account for higher cumulative mortality despite lower per building collision rates [5, 7].
Many studies have identified factors that lead to spatiotemporal variation in bird-building
collisions. Temporal factors include weather, seasonality, migration phenology, and fluctua-
tions in bird abundance [15–17]. Spatial factors include building-related features like amount
of glass, building shape, and nearby vegetation [18–20], as well as broader landscape features
like surrounding greenspace and urbanization intensity [21]. Research into correlates of bird-
window collisions has led to development of recommendations and management approaches
that can be used to reduce collisions. Technologies and commercially available products that
reduce glass reflection and transparency have been developed, tested, and marketed, and
guidelines to make newly constructed buildings bird-friendly (e.g., by reducing amount of
glass or using opaque, fritted, or colored glass) have also been summarized [18, 22, 23]. Munic-
ipal, state, and federal policy guidelines and regulations to implement such bird-friendly
approaches have also been adopted or are under consideration. These include, for example,
Standards for Bird-Safe Buildings in San Francisco, California, U.S.A [24], Buildings, Bench-
marks, and Beyond in Minnesota, U.S.A. [25], Best Practices for Bird-friendly Glass and Best
Practices for Effective Lighting in Toronto, Canada [26], and the Bird Safe Buildings Act of 2021
currently under consideration by the U.S. federal government [27].
Bird-window collisions occur in areas with human infrastructure, and humans regularly
encounter the bird carcasses that result. However, although significant resources have gone
into designing and testing mitigation approaches to reduce bird-window collisions, and into
developing and implementing bird-friendly policies and guidelines, only two studies have eval-
uated human perceptions and priorities related to these practices. In fact, there is a general
lack of human dimensions research for nearly all sources of direct, human-caused bird mortal-
ity, including other kinds of bird collisions (e.g., with wind turbines, communication towers,
and vehicles; but see studies of wildlife predation by domestic cats) [28, 29]. One of the studies
that evaluated human perspectives related to bird-window collisions examined the Canadian
public’s willingness to pay (WTP) to reduce collisions at their homes [30] and found that WTP
was positively associated with homeowner age, income, and interest in birds, among other fac-
tors. The other study investigated public perceptions and knowledge about this issue in Costa
Rica and concluded that participants were aware of bird-window collisions but not of the large
magnitude of the problem [31]. Clarifying how people perceive bird-window collisions, and
how much they support mitigation and prevention techniques, is crucial for bird conservation
because implementing effective practices generally entails adoption of new products and tech-
nologies on buildings, and therefore, requires buy-in from multiple stakeholder groups (e.g.,
residential homeowners, owners/managers of commercial buildings, building architects,
policymakers).
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
2 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
We began to address this major research gap by exploring and quantifying perceptions and
priorities related to bird-window collisions among a diverse pool of respondents in North
America. Our objectives were to: (1) evaluate how two important stakeholder groups (owners
of individual residences, i.e., “homeowners,” and conservation practitioners in state, federal,
and non-government conservation organizations) perceive and prioritize potential benefits
and obstacles related to bird-window collision management, (2) compare priority rankings for
benefits and obstacles to management between homeowners and conservation practitioners,
and (3) evaluate potential conflicts in priorities within each stakeholder group, as well as the
collective strength of group opinions. To address objectives 1 and 2, we merged the strengths,
weaknesses, opportunities, and threats (SWOT) and analytical hierarchy process (AHP) analy-
ses; the approach of merging these two analyses is frequently used to quantitatively assess and
rank perceived benefits and obstacles related to management actions and decisions [32–35].
To address objective 3, we used Manfredo et al.’s [36] potential for conflict index (PCI) to visu-
alize within-group conflicts and strength of group opinions, information that can lend addi-
tional insight into factors potentially limiting progress in managing bird-window collisions.
Methods
Study design
This study, the survey distribution strategy, and the survey contents were approved by and
comply with the Oklahoma State University Institutional Review Board’s (IRB) standards and
regulations (approved IRB protocol # IRB-20-202). All survey participants gave consent for
participation upon completion of surveys, and data were also analyzed anonymously. To
address objectives 1 and 2, we used a combined SWOT-AHP perception analysis approach
(i.e., a strengths, weaknesses, opportunities, and threats analysis linked with an analytic hierar-
chy process analysis). This merged approach is often used to quantify and rank perceptions
about major benefits and obstacles related to issues, actions, and decisions of interest, and to
compare benefit and obstacle rankings among diverse stakeholder groups, including for issues
in conservation and natural resource management like renewable energy, ecotourism, and
land management and policy [32–35, 37, 38]. In the SWOT framework [39], there are 4 catego-
ries of factors related to the issue, action, or decision under consideration: strengths, weak-
nesses, opportunities, and threats. Strengths and weaknesses are considered internal to an
issue, action, or decision. In our case, strengths are direct, immediate outcomes of implement-
ing bird-window collision management (e.g., fewer bird collisions) and weaknesses are direct
barriers or obstacles to implementing management (e.g., the financial cost of management).
Opportunities and threats are considered external to an issue, action, or decision. In our case,
opportunities are non-immediate and/or secondary outcomes that indirectly result from
implementing management (e.g., increased bird populations due to fewer collisions), and
threats are barriers that are not directly related to management but that could arise as manage-
ment is carried out (e.g., with collision management expenses, reduced financial resources for
other building management-related costs). We used the SWOT approach to ask surveyed
stakeholders to prioritize strengths, weaknesses, opportunities, and threats related to bird-win-
dow collision mitigation and prevention (the specific factors used for each of these 4 SWOT
categories are under “Survey Questionnaire Details”). The ultimate goal of a SWOT analysis is
to determine perceptions of stakeholders to help develop a strategy that optimizes the tradeoff
between strengths and weaknesses of various options, while considering both internal and
external factors. When used alone, SWOT does not allow quantitative ranking of factors within
or across different categories, making it difficult to draw conclusions about perceptions. The
AHP, however, is a generalized method to rank decision problems that assumes independence
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
3 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
among options; when combined with SWOT, AHP allows quantitative comparisons of differ-
ent SWOT factors, which helps determine the relative importance of a decision [39]. As a
multi-criteria decision-making tool, AHP assigns relative weights to factors of interest based
on 2-way comparisons between factors [40]; this allows objective evaluation of the degree of
agreement (or disagreement) between factors.
Stakeholder groups and strategy to distribute survey questionnaire
Initially, we sought to investigate priorities of four stakeholder groups: architects, home-
owners, and conservation practitioners in both government agencies and non-governmental
organizations (NGOs). Each of these groups can play a key role in managing bird-window col-
lisions. Architects can help reduce collisions by working from the top down to incorporate
mitigation and prevention measures, within policy parameters, into design and construction
of new buildings [41, 42]. Homeowners act from the bottom-up as consumers by expressing
their values and desires, buying and living in houses, and deciding whether to manage their
properties in ways that benefit birds (e.g., feeding birds or applying films/decals to windows to
reduce collisions) [42, 43]. Government and NGO conservation practitioners are both knowl-
edgeable about and advocate for wildlife, but these two groups may enact change in different
ways. Government (federal, state/provincial, and tribal) practitioners help inform policy devel-
opment with research and management, and while NGOs can also help inform policy, they
typically engage members of the public through activities such as education campaigns, volun-
teering, and public funding [41, 42].
To recruit respondents from all stakeholder groups (architects, homeowners, government
conservation practitioners, and NGO conservation practitioners) and from as broad of a geo-
graphic area as possible, we used snowball sampling, a nonprobability sampling method that
uses gateway contacts who can take the survey themselves and are asked to forward the survey
invitation to relevant contacts in their stakeholder group [44]. For this study, gateway contacts
were the authors’ personal or professional contacts in each stakeholder group, including 17
architects, 66 homeowners, 36 government practitioners, and 20 NGO conservation practi-
tioners. Most of these contacts lived and worked in the United States (18 U.S. states repre-
sented), but Canada was also represented. Recruitment emails were tailored to each
stakeholder group and sent from the authors to gateway contacts; these emails contained a
brief overview of the project, a request for participation, a link to sign up to take the survey, a
link to a recruitment video on YouTube, and a request that respondents share recruitment
materials with colleagues [45]. The recruitment video contained a brief overview about the
issue of bird-window collisions and the objectives of this research project, as well as a request
for participation and to forward the recruitment materials. In addition to using gateway con-
tacts, we also actively recruited respondents using social media platforms, including Facebook
and Twitter [46, 47]. Recruitment via Facebook and Twitter included brief posts on the
authors’ profile pages, which are followed by numerous professional contacts with formal posi-
tions in conservation science and management (including government and NGO conservation
practitioners), and by nonprofessional contacts that include numerous homeowners. These
Facebook and Twitter posts contained information about the project, the recruitment video, a
call for participation, a link to sign up to take the survey, and a request to share recruitment
materials. Of note, mixed data collection methods involving focus group meetings, web sur-
veys, and email contacts have been commonly adopted in SWOT-AHP based studies [34, 37,
48]. Accordingly, to broaden participation and increase replication of responses from mem-
bers of the homeowner stakeholder group, we reached out to multiple neighborhood home-
owner’s associations (HOA) in Stillwater, Oklahoma, USA, the location of the authors’ home
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
4 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
institution (Oklahoma State University). We used this approach because we expected that
snowball sampling would result in recruitment of relatively few homeowners. Recruitment
materials were sent to publicly available email addresses of HOA board member contacts;
again, we requested participation in the survey and dissemination of recruitment materials to
other HOA board members and neighborhood residents.
Survey questionnaire details
Using the merged SWOT-AHP approach first entails development of a survey that contains a
list of top strengths, weaknesses, opportunities, and threats regarding the issue at hand. These
SWOT lists are often developed from a longer list of candidate factors with assistance of sub-
ject-matter experts [37]. We created a list of candidate SWOT factors related to bird-window
collision management based on our own subject matter expertise, which includes familiarity
with the scientific and gray literature on this topic, and years of interactions and collaborations
with key stakeholders in federal/state agencies and NGOs. After drafting the initial list of can-
didate SWOT factors, we asked three external bird-window collision experts to rank them by
importance. Expert responses for each candidate factor were counted and weighted based on
ranking to create a final SWOT list containing the four top-ranked factors in each category
(Table 1).
Following methodology used by similar SWOT studies, we next solicited stakeholder opin-
ions in two rounds of surveys, with each containing multiple pairwise comparisons between
SWOT factors using a scale of one to nine [32, 37, 49]. Specifically, a value of 1 indicated an
opinion that one factor was “extremely important,” a value of 9 indicated an opinion that the
other factor was extremely important, and a value of 5 indicated an opinion that the two fac-
tors were “equally important” (see Fig 1 for visual representation of scale). For Survey 1, all
possible pairwise comparisons were made between factors within (but not between) SWOT
categories. For example, all possible 2-way comparisons were made among strengths (e.g.,
Fewer collisions compared to Fewer bird carcasses to clean up), but in this survey, strengths
were not compared to weaknesses, opportunities, or threats (see example comparison in Fig 1
and S1 File for full Survey 1 contents). We created Survey 2 based on top-ranked factors
Table 1. List of all SWOT factors.
Strengths
Weaknesses
S1: Fewer collisions
W1: No economic incentives building for bird-friendly
buildings
S2: Fewer carcasses to clean up
W2: Lack of architect experience in bird-friendly design
S3: Fewer people witnessing collisions
W3: Lack of availability of expert consultation for bird-
friendly design
S4: Fewer stunned birds that die of other causes while
recovering from colliding
W4: Financial burden of treating glass or including bird-
friendly design in building process
Opportunities
O1: Recovering bird populations
O2: Public exposure to bird-friendly options
Threats
T1: Unknown social acceptance of bird-friendly treatments
and design
T2: Lack of understanding of federal/state policy on bird-
window collisions
O3: Consideration of birds in building design
becoming a norm/standard
T3: Reduced resources available to spend on other facilities
maintenance/improvements
O4: Greater energy efficiency of buildings
T4: No federal/state policy in many areas
Finalized list of strengths, weaknesses, opportunities, and threats (SWOT) containing the top four factors for each
category that were used to evaluate stakeholder perceptions regarding bird-window collisions.
https://doi.org/10.1371/journal.pone.0263447.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
5 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
Fig 1. SWOT survey example. Examples of pairwise comparisons within the strengths category of the strengths, weaknesses,
opportunities, and threats (SWOT) analysis; this example illustrates the format of Survey 1 distributed to stakeholder groups to evaluate
their perceptions and priorities regarding bird-window collision management.
https://doi.org/10.1371/journal.pone.0263447.g001
calculated from Survey 1 for each SWOT category (see details of these calculations under
“Data Analysis”). These calculations were made separately for each stakeholder group, which
allowed us to tailor Survey 2 to each group, a standard practice for SWOT studies. In Survey 2,
respondents were asked to make pairwise comparisons of all top-ranking factors between
SWOT categories. For example, within the homeowner group, the factor Fewer collisions was
identified as the top strength in Survey 1, and No federal/state policy in many areas was the top
threat. Thus, respondents were asked to compare these two factors (see S2 and S3 Files for full
Survey 2 contents for each stakeholder group).
All surveys were administered using the online platform Qualtrics [50], and both surveys
had the same general format. Both surveys contained an introductory page displaying informa-
tion about the study, including the study’s purpose, what to expect, risks associated with par-
ticipating, and a confidentiality statement. Next, the survey asked respondents to indicate
which stakeholder group they belonged to. The following section contained a brief introduc-
tion to the issue of bird-window collisions (to give respondents introductory background or to
reorient them to the issue), as well as a table containing all of the SWOT factors. To minimize
the collection of personally identifiable information and to retain survey anonymity, we only
collected contact information (names and emails) of potential respondents during the initial
recruitment period (i.e., the period during which we asked stakeholders to sign up to take the
survey, but before the survey was distributed). During survey periods, surveys were completed
anonymously; therefore, we could not monitor which people who signed up (including gate-
way contacts and other people reached through snowball and purposive sampling) actually
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
6 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
completed the surveys. For Survey 2, all individuals who signed up to take Survey 1 were again
contacted, but we requested that only those that completed Survey 1 complete Survey 2. Survey
1 was administered from 1 June 2020 to 30 June 2020, and Survey 2 was administered from 13
July 2020 to 12 August 2020. For all stakeholder groups and sampling approaches, we waited
two weeks before sending one reminder to complete the survey to allow adequate time for par-
ticipants to respond to the original request [51].
Data analysis
Analyses of survey response data followed methods of other SWOT-AHP studies (e.g., Starr
et al. 2019 and Joshi et al. 2020) [37, 52] that adapted their analyses from Saaty [40]. The same
general procedures were used to analyze results from Survey 1 (comparisons within SWOT cate-
gories) to determine factor priorities for Survey 2, and to analyze results from Survey 2 (compari-
sons of top-ranked factors between SWOT categories). First, to calculate the weighted geometric
mean for each factor in each SWOT category, and also separately for each stakeholder group, we
collated response data for each pairwise comparison into counts according to the selection scale
of one to nine (See S1 Dataset for calculated geometric means). Counts were then weighted
reciprocally, multiplied, and taken to the power of one over the total number of counts [53].
Each weighted geometric mean was entered into a standard reciprocal matrix, and values were
then normalized and placed into a weighted reciprocal pairwise matrix. The weighted reciprocal
pairwise matrix was used to calculate factor priority values for each factor in each SWOT cate-
gory and stakeholder group; these values were used to evaluate relative importance of factors
within each SWOT category (all factor priority values within a category sum to one). The stan-
dard reciprocal matrix and factor priority values for each category were also used to calculate a
consistency index, which when used with a predetermined random index (based on the number
of SWOT factors within a category) determines the consistency ratio, a metric indicating the
consistency of responses among respondents within a stakeholder group [39, 52]. Pairwise com-
parisons within each SWOT category were determined to be internally consistent if the consis-
tency ratio (calculated for each SWOT category within each stakeholder group and for both
surveys) was less than 10%; however, consistency ratios up to 20% are considered acceptable [34,
40, 49, 52]. When we conducted preliminary analyses of Survey 1 responses, we calculated unac-
ceptably high consistency ratios within the architect and NGO practitioner groups that were
most likely attributable to small sample sizes of recruited respondents (n = 12 for each group).
We therefore excluded data for architects, and due to similarities between the groups and to pre-
vent data loss, we combined government practitioners (n = 26) and NGO practitioners into a
single group (conservation practitioners, n = 38). Thus, our final analysis of Survey 1 (and subse-
quently, Survey 2) included two stakeholder groups, homeowners (Survey 1: n = 52; Survey 2:
n = 33) and conservation practitioners (Survey 1: n = 38; Survey 2: n = 41). Our receipt of more
conservation practitioner responses for Survey 2 than Survey 1 was unexpected because we only
asked recruits to complete the second survey if they had already completed the first survey. This
result likely arose because we had to exclude a small number of Survey 1 responses that were
incomplete or contained response errors (7 surveys excluded for homeowners; 4 for conserva-
tion practitioners). Regardless of the explanation, we have no reason to believe that receiving
slightly more Survey 2 results biased our results.
The last steps in the SWOT-AHP analysis were to calculate global and group priority values.
Global priority values rank individual SWOT factors among all categories for each stakeholder
group; these values allow for comparison among stakeholders’ perceptions and priorities, as
well as evaluation of SWOT factor priority rankings against each other [32, 37, 49]. Global pri-
ority values within each SWOT category were then added together to create group priority
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
7 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
values that represent the priority of each SWOT category as a whole. We also followed previ-
ous literature (e.g., Dwivedi & Alavalapati 2009 and Joshi et al. 2018) [32, 34] to generate per-
ception maps, which illustrate differences in global priority values and allow direct
comparisons among all SWOT factors and between stakeholder groups.
To address objective 3, we applied Manfredo et al.’s [36] potential for conflict index (PCI)
to the Survey 2 responses (see S1 Dataset for PCI calculations); the PCI allows visualization of
potential conflicts in perceptions within stakeholder groups, and of the collective strength (vs.
neutrality) of group opinions [54], information that can lend additional insight into factors
potentially limiting progress in addressing bird-window collisions. We used the PCI2, an
extension of PCI that is used for response data from a scalar survey to visually display degree
of conflict (i.e., opposite of agreement) in responses among respondents in a stakeholder
group, as well as neutrality of responses [36, 54]. In this case, the scalar survey questions were
pairwise comparisons that respondents completed in Survey 2. With regard to neutrality, pair-
wise comparisons that are near five for a stakeholder group indicate factors perceived as
Equally important (indicated as bubbles close to the x-axis on PCI graphs). Comparisons that
are lower (near one) or higher (near nine) toward either of the factors being compared repre-
sent an average group perception that one factor is Extremely important relative to the other
(bubbles farther from the x-axis). Regarding degree of conflict, this value ranges between 0 and
1, with values close to 0 indicating little conflict (strong agreement on a pairwise comparison
among respondents in a group, indicated as small bubbles), and values close to 1 indicating
complete conflict (i.e., responses on a pairwise comparison equally divided between the two
extreme values on the response scale, indicated as large bubbles) [36, 55].
Results
Stakeholder priorities for different SWOT categories
Our survey likely had a nationwide or even broader scope, as our gateway contacts represented
at least 18 U.S. states and Canada. However, the exact geographic distribution of survey
respondents is unknown because surveys were completed anonymously to minimize collection
of personally identifiable information, and because the snowball sampling method we used
entailed recruitment of additional respondents beyond our gateway contacts. For all SWOT
categories except two in the conservation practitioner group for Survey 1, consistency ratios
were <10%, indicating consistent responses within stakeholder groups. For conservation prac-
titioners, the weaknesses and opportunities categories had consistency ratios of 19% and 18%,
respectively, indicating some inconsistency. Nonetheless, consistency ratios <20% are consid-
ered acceptable for drawing inferences [34, 49].
A summary of SWOT factor, group, and global priorities for homeowners and conservation
practitioners is in Table 2. Group priorities for homeowners for strengths, weaknesses, oppor-
tunities, and threats were 24%, 15%, 40%, and 21%, respectively, and group priorities for con-
servation practitioners were 24%, 15%, 52%, and 9%, respectively. For homeowners and
conservation practitioners, perceptions about potential outcomes of bird-window collision
mitigation and prevention were generally positive, as evidenced by summed percentages of
group priorities for strengths and opportunities (64% and 76% for homeowners and conserva-
tion practitioners, respectively). As indicated by group priority values for threats, homeowners
gave greater priority (21%) to threats than did conservation practitioners (9%).
Stakeholder priorities for different factors within SWOT categories
As evident from the above-presented group priority values, homeowners prioritized opportu-
nities overwhelmingly over strengths, weaknesses, and threats. Among opportunities,
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
8 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
Table 2. Factor, global, and group priorities for all SWOT factors for each stakeholder group.
SWOT Factors
Factor Priority
Global Priority
Homeowner
Conservation
Practitioner
Homeowner
Conservation
Practitioner
S1: Fewer collisions
S2: Fewer carcasses to clean up
S3: Fewer people witnessing collisions
S4: Fewer stunned birds that die of other causes while recovering from colliding
Group Priorities for Strengths
W1: No economic incentives for building for bird-friendly buildings
W2: Lack of architect experience in bird-friendly design
W3: Lack of availability of expert consultation for bird-friendly design
W4: Financial burden of treating glass or including bird-friendly design in
building process
Group Priorities for Weaknesses
O1: Recovering bird populations
O2: Public exposure to bird-friendly options
O3: Consideration of birds in building design becoming a norm/standard
O4: Greater energy efficiency of buildings
Group Priorities for Opportunities
T1: Unknown social acceptance of bird-friendly treatments and design
T2: Lack of understanding of federal/state policy on bird-window collisions
T3: Reduced resources available to spend on other facilities maintenance/
improvements
T4: No federal/state policy in many areas
Group Priorities for Threats
0.46
0.11
0.09
0.34
0.23
0.18
0.31
0.28
0.34
0.18
0.25
0.23
0.19
0.25
0.25
0.31
0.60
0.06
0.07
0.27
0.36
0.13
0.26
0.25
0.45
0.15
0.20
0.21
0.14
0.16
0.36
0.35
0.11
0.03
0.02
0.08
0.24
0.03
0.03
0.05
0.04
0.15
0.14
0.07
0.10
0.09
0.40
0.04
0.05
0.05
0.07
0.21
0.15
0.02
0.02
0.07
0.24
0.05
0.02
0.04
0.04
0.15
0.23
0.08
0.10
0.11
0.52
0.01
0.01
0.03
0.03
0.09
Summary of factors used in strengths, weaknesses, opportunities, and threats (SWOT) analyses related to perceptions and potential outcomes of bird-window collision
mitigation and prevention. Factor priority values indicate the relative importance of a single factor within a SWOT category among other factors in the same category
(boldfaced factor priority values are the highest prioritized factor for each SWOT category). Global priority values rank individual SWOT factors among all factors and
can be compared across SWOT categories. Group priority values (the boldfaced values in “Global Priority” columns) are the sum of global priority values within each
SWOT category and are used to compare categories against each other.
https://doi.org/10.1371/journal.pone.0263447.t002
Recovering bird populations was the top factor priority (34%), followed by Consideration of
birds in building design becoming a norm/standard (25%) and Greater energy efficiency of build-
ings (23%). Homeowners prioritized strengths next; highest priority strengths were Fewer colli-
sions (46%) and Fewer stunned birds that die of other causes while recovering from colliding
(34%). The anthropocentric strengths received lower priority, including: Fewer carcasses to
clean up (11%) and Fewer people witnessing collisions (9%). For threats, which homeowners
prioritized only slightly behind strengths, the top factor was No federal/state policy in many
areas (31%), followed by two equally ranked (25%) priorities: Lack of understanding of federal/
state policy on bird-window collisions and Reduced resources available to spend on other facilities
maintenance/improvements. Homeowners prioritized weaknesses lowest, with Lack of avail-
ability of expert consultation for bird-friendly design being the top priority (31%) within this
category (Table 2).
Based on group priority values, conservation practitioners also prioritized opportunities as
most important; among opportunities, Recovering bird populations was the top-priority factor
(45%). Strengths was the second-highest prioritized category, and top factors in this category
were Fewer collisions (60%) and Fewer stunned birds that die of other causes while recovering
from colliding (27%). Conservation practitioners gave weaknesses and threats lowest priority.
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
9 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
The most highly prioritized weakness was No economic incentives for building bird-friendly
buildings (36%); the two top threats were Reduced resources available to spend on other facilities
maintenance/improvements (36%) and No federal/state policy in many areas (35%) (Table 2).
Stakeholder priorities for different factors across SWOT categories
Perception maps (Fig 2A and 2B) illustrate differences in global priorities and allow direct
comparisons among all SWOT factors and between stakeholder groups. For homeowners, the
opportunity Recovering bird populations (O1) received the highest global priority among all
SWOT factors, closely followed by the strength Fewer collisions (S1). Although homeowner
priorities for weaknesses and threats were lower than for strengths and opportunities, all
threats and some weaknesses still received higher global priorities than the strengths Fewer
people witnessing collisions (S2) and Fewer carcasses to clean up (S3). The opportunity Recover-
ing bird populations (O1) followed by the strength Fewer collisions (S1) also received the two
highest global priorities for conservation practitioners. Additionally, this group prioritized
weaknesses over threats while homeowners ranked these categories in the opposite order.
Although the two groups had similar broad priorities, such as valuing strengths and oppor-
tunities over weaknesses and threats, conservation practitioners gave higher priority to the top
factor in some categories, suggesting stronger perceptions toward these factors. Specifically,
although Recovering bird populations (O1) was the highest global priority among all SWOT
factors for both stakeholder groups, it received a greater global priority value for conservation
practitioners (0.23) than homeowners (0.14). Similarly, the top strength (and second highest
global priority among all SWOT factors) for both stakeholder groups (Fewer collisions; S1)
received a greater global priority value for conservation practitioners (0.15) than for home-
owners (0.11) (Table 2). Global priorities also illustrated that both homeowners and conserva-
tion practitioners gave low priority to Fewer people witnessing collisions (S2) and Fewer
carcasses to clean up (S3) relative to other strengths and many other weakness and threats.
Potential for conflict and strength of opinions within stakeholder groups
Regarding potential for conflict indices (PCI2) for Survey 2, comparison of the bubbles for
homeowners (Fig 3A) and conservation practitioners (Fig 3B) for each pairwise comparison
illustrates there was more conflict among responses for homeowners than conservation practi-
tioners for 4 of 6 comparisons. Additionally, relative locations of bubbles on the y-axis (which
indicates the difference in preference for each priority in a pairwise comparison) illustrate that
homeowners were more neutral than conservation practitioners for all 6 pairwise
comparisons.
Discussion
Our results suggest that both homeowners and conservation practitioners have an overall posi-
tive perception toward potential benefits related to bird-window collision mitigation and pre-
vention measures. This indicates stakeholders may believe that benefits of implementing
management to reduce bird-window collisions outweigh any obstacles that may impede such
measures. Although generally similar in their positive views, the two stakeholder groups dis-
played some differences in their specific priorities regarding strengths, weaknesses, opportuni-
ties, and threats surrounding this issue. Specifically, homeowners gave greater priority than
conservation practitioners to threats, indicating more concern among homeowners about
external obstacles (financial and policy related) that may impede bird-window collision man-
agement efforts.
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
10 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
Fig 2. Perception maps of SWOT global priorities for each stakeholder group. Perception maps illustrating
homeowner (a) and conservation practitioner (b) strength, weakness, opportunity, and threat (SWOT) global priorities
for a study evaluating perceptions about potential outcomes of bird-window collision mitigation and prevention.
Factors with the highest global priority are farthest from the origin. S1: Fewer collisions; S2: Fewer carcasses to clean
up; S3: Fewer people witnessing collisions; S4: Fewer stunned birds that die of other causes while recovering from
colliding. W1: No economic incentives for building for bird-friendly buildings; W2: Lack of architect experience in
bird-friendly design; W3: Lack of availability of expert consultation for bird-friendly design; W4: Financial burden of
treating glass or including bird-friendly design in building process. O1: Recovering bird populations; O2: Public
exposure to bird-friendly options; O3: Consideration of birds in building design becoming a norm/standard; O4:
Greater energy efficiency of buildings. T1: Unknown social acceptance of bird-friendly treatments and design; T2:
Lack of understanding of federal/state policy on bird-window collisions; T3: Reduced resources available to spend on
other facilities maintenance/improvements; T4: No federal/state policy in many areas.
https://doi.org/10.1371/journal.pone.0263447.g002
Stakeholder perceptions about bird-window collision management
Results indicate that the homeowner and conservation practitioner groups, while in general
agreement on their positive perceptions about managing bird-window collisions, each have
unique aspects of their perceptions that are important to consider in order to make headway
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
11 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
Fig 3. Potential for conflict indices from survey 2 for each stakeholder group. Illustration of the potential for
conflict index (PCI2) based on homeowner (a) and conservation practitioner (b) responses to Survey 2 in a study
evaluating perceptions about potential outcomes of bird-window collision mitigation and prevention. Bubble size and
values correspond and indicate the dispersion (conflict) among respondent answers (larger bubbles/numbers indicate
greater conflict). The location of the bubble indicates the scale mean or the direction respondents lean in their answers
to pairwise comparisons (e.g., 5 indicates completely neutral; values lower and higher than 5 indicate more non-
neutral perceptions). Each bubble is an individual pairwise comparison indicated by the labels. Pairwise comparisons
correspond visually to the y-axis scale (e.g., for S1-W3, 1 corresponds to S1 and 9 corresponds to W3). For a
description of all strengths (S), weaknesses (W), opportunities (O), and threats (T), see Table 1.
https://doi.org/10.1371/journal.pone.0263447.g003
in addressing this conservation issue. As evidenced by the PCI analysis, homeowners had
more conflict in their responses to pairwise comparisons than conservation practitioners, indi-
cating differing opinions within the group. PCI analysis also indicated that homeowners were
more neutral than conservation practitioners in their responses, demonstrating differing or a
potential lack of strong opinions within the group. Although we provided contextual informa-
tion about this project in the survey’s introductory materials, a lack of prior knowledge about
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
12 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
the issue—which was anecdotally revealed from comments made by gateway contacts in the
homeowner group—could have contributed to their relatively neutral perceptions and con-
flicting responses. The less-conflicting responses within the conservation practitioner group
could be due to greater knowledge about the issue or more cohesion within the group due to a
shared field of profession and its associated sources of information. Specifically, those in the
field of wildlife conservation likely have greater, and perhaps more consistent, exposure to
major bird conservation issues through training opportunities, professional conferences, social
media networks, newsletters, and scientific publications. It is important to note that the home-
owner group included gateway contacts from a wide variety of professional backgrounds,
which could explain the lesser degree of agreement within the group.
As evidenced by high group priority values for the strength and opportunity categories, as
well as high global priority values for individual strengths and opportunities, our results indi-
cate that both stakeholder groups have positive views about bird-window collision mitigation
and prevention measures. Members of these groups may therefore be willing to participate in
or support implementation of measures to reduce bird collisions. Because the top ranked
strengths and opportunities capture outcomes related to bird conservation and welfare (e.g.,
recovering bird populations), not anthropocentric benefits (e.g., no longer having to clean up
or observe collisions), our results suggest that stakeholders value mitigating and preventing
collisions for the sake of the birds themselves. This result demonstrates that stakeholders may
have a general sense of caring and responsibility for birds—and/or that they view birds as aes-
thetically, culturally, or economically valuable [56, 57]—which lends additional support to the
potential acceptability and implementation of management measures. Due to a greater degree
of neutrality and lack of strong opinions within the homeowner group (as illustrated by the
PCI), and because some homeowners in our study were not previously aware of bird-window
collisions and underlying challenges, our findings suggest a strong need for public education
on this issue.
Advantageously, the positive perceptions about reducing bird-window collisions, and the
apparently bird-centric reasons behind these positive perceptions, suggest that members of the
public may be receptive to further education about this issue. Menacho-Odio [31] also investi-
gated public perception and knowledge of bird-window collisions in Monteverde, Costa Rica,
and concluded that while participants had general knowledge of the issue, few were aware of
the magnitude of the problem. This previous study recommended targeted education that
informs people about the large number of bird-window collisions that occur, as well as effec-
tive methods for preventing collisions. There are multiple publicly available resources from
which individuals can learn about bird-window collisions and ways to reduce them. For exam-
ple, the American Bird Conservancy (ABC) has published a website geared toward the public
[58], a Bird-Friendly Building Design booklet targeting all types of building owners and man-
agers, as well as architects [22], interactive web resources and educational materials for home-
owners and architects, and a framework to help policy makers develop ordinances and
legislation to reduce collisions. Similar and complementary resources to improve stakeholder
knowledge about bird-window collisions have also been developed by other conservation orga-
nizations and agencies (e.g., USFWS 2021; National Audubon Society 2021; FLAP Canada
2019) [59–61]. While many resources are available, active education on this topic would also
be beneficial. Specifically, increased funding and staffing to expand the delivery and interpreta-
tion of such resources to stakeholders, along with research to improve understanding of how
best to develop and distribute these resources to ensure they are used, are needed to make fur-
ther headway in reducing bird-window collisions.
As evident from the factor and global priority values for threats, homeowners highly priori-
tized policy-related obstacles to bird-window collision mitigation and prevention. However,
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
13 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
importantly, multiple states, cities, and municipalities across North America have already
enacted policies designed to reduce bird-window collisions, including San Francisco, Califor-
nia, U.S.A. [24] and Minnesota, U.S.A. [25]. The U.S. House of Representatives also approved
legislation (Bird Safe Buildings Act of 2021) that would require bird-friendly measures at
many new and renovated U.S. federal buildings; however, this act has not yet passed the U.S.
Senate [27]. Thus, while there is concern among homeowners about potential policy-related
obstacles, many may not know that relevant policies already exist. This points again to the
importance of education, as increasing awareness of existing and proposed policies could
increase support for them among the public, and therefore, among policymakers.
Beyond educating homeowners about existing and planned policies related to bird-window
collisions, homeowners could also be informed that implementing bird-friendly measures at
homes might be their responsibility even with policies in existence. To date, no legislation and
policies have focused on residential structures, and the proposed U.S. federal bill only focuses
on public buildings. Thus, there are no formal mechanisms to ensure that collisions are
reduced at residences, even though residences collectively cause a large proportion of total
bird collisions [5, 7]. Although public education may encourage some homeowners to expend
their own resources on measures to reduce bird-window collisions, formal programs to
encourage these actions may be necessary to ensure that a large proportion of homes become
bird-friendly in the future, especially for lower income residents that lack expendable
resources to pay for such measures. Examples of such potential programs include conservation
grants/subsidies that help pay for materials that make existing windows more bird-friendly,
and revisions to existing sustainability or wildlife-friendly certification programs to specifically
incorporate considerations related to reducing bird-window collisions.
Our analysis identified other potential barriers to widespread bird-window collision man-
agement. For example, homeowners identified a lack of availability of expert consultation as
another top threat. Although the above-mentioned education campaigns could help empower
homeowners to reduce collisions themselves, this result suggests that widespread adoption of
collision management practices at homes may require increased training of consultants and
outreach professionals that convey information about collision management. Conservation
practitioners identified a lack of resources available to spend on other facilities/maintenance
improvements as a top threat arising from the costs of collision management. In addition to
emphasizing the need for low-cost management options, this result suggests that approaches
that reduce collisions while meeting other facilities-related needs may be especially likely to be
adopted. Notably, some approaches that are highly effective in reducing bird-window colli-
sions, including reducing nighttime lighting [14] and some of the films, coatings, and decals
adhered to windows to make them more visible [22], also may contribute to reducing build-
ing-related energy costs. Communicating the dual benefits of such approaches may lead to
greater adoption of bird-friendly building management techniques.
Limitations and future research
While this research provides valuable information to advance efforts to manage bird-window
collisions, there were some limitations and potential biases related to our analyses. We were,
for example, unable to analyze perspectives of architects as an independent stakeholder group
due to limited recruitment for participation in our surveys. Architects are a crucial stakeholder
in the issue of bird-window collisions, and further research should seek to thoroughly evaluate
their perceptions about this topic. The low number of respondents for architects leads to the
question of how best to reach and engage with this stakeholder group. Potential routes to
engage architects include having bird-window collision researchers present at architectural
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
14 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
society conferences, creating publication materials geared toward architects, or reaching out
directly to architectural societies or firms about bird-window collisions.
Another limitation concerns the representativeness of our sample of survey respondents,
which relates both to the limited sample size of respondents and mixed-data collection
approach that used gateway contacts and recruitment through social media platforms. Nota-
bly, the AHP approach does not require large sample sizes to result in statistically robust
results that are useful for understanding stakeholder perceptions and informing management
decisions [62]. Instead, reliability of results from this approach is interpreted using consistency
ratios, which indicate the degree of consistency of responses within stakeholder groups. Con-
sistency ratios for groups used in our analyses were considered acceptable [63], suggesting our
results are reliable. However, because many of the gateway contacts we recruited for the home-
owner group were our personal and professional contacts, our sample of homeowners could
have been biased toward bird enthusiasts rather than providing full representation of the diver-
sity within this group. Nonetheless, our homeowner sample contained many respondents
beyond the gateway contacts that we did not know personally, indicating that there may have
been variation in levels of interest or support for bird-window collision management and wild-
life conservation more broadly. Although our approach does not require large sample sizes, we
caution against making broad generalizations from our results, especially for the homeowner
group, due to these potential issues regarding sample representativeness.
Our results lay a foundation for future research into stakeholder perceptions, priorities, and
potential disputes and conflicts related to bird-window collision management. Conducting
research to better understand motivations and barriers to behavioral change will be crucial for
designing collision management programs that garner broad support and participation from
the public. In this study, we examined stakeholder perceptions and priorities, but other impor-
tant factors that influence behavioral changes (e.g., social and cultural norms, institutional and
economic factors) should also be evaluated [64]. Further, research that identifies social-psy-
chological barriers that may lead to conflicts among groups (e.g., conservation organizations
recommending collision management approaches vs. building management entities resistant
to recommendations) could facilitate more-rapid adoption of bird-friendly building design,
and similar research related to the green building movement may be instructive for this issue
[65]. We did not collect demographic information from respondents, nor did we know the
geographic representation of our sample other than for gateway contacts. Because the factors
that influence behaviors, perceptions, and conflicts can vary regionally and among demo-
graphic groups (e.g., among different age groups), future research could evaluate how percep-
tions about bird-window collisions vary regionally and in relation to various demographic
factors.
Another essential area of future research is to evaluate stakeholders’ willingness to pay
(WTP) for measures to reduce collisions. Our study shows that the stakeholder groups we eval-
uated are receptive to bird-window collision management, but that does not necessarily trans-
late into a willingness to pay for these measures, especially if doing so at private residences is
the responsibility of homeowners. Past research evaluating WTP for conservation practices
indicates that the public is often receptive to wildlife conservation and willing to pay for it [66–
69]. The public’s WTP for conservation practices can be heavily influenced by sense of place,
or the value and meaning that individuals attach to a physical location [70, 71]. This suggests
that informational materials that tie the issue of bird-window collisions to an individual’s loca-
tion or experience may be a particularly effective way to increase WTP. For example, educa-
tional materials could highlight the likely number of collisions that occur in areas where
residents live and how collisions may be affecting locally important bird species. Another
study found that while members of the public were willing to pay for bird conservation, they
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
15 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
believed the government should also play a role [68], a finding that lends additional support to
grant, subsidy, and/or certification programs specifically geared toward reducing bird-window
collisions. Although homeowners are a critical group to examine with regard to WTP to
reduce bird-window collisions, other stakeholders such as business owners and agencies oper-
ating in larger buildings are also important stakeholders to study.
Birds face multiple human-related threats, including climate change, habitat loss, and other
direct mortality sources (e.g., cat predation, other types of collisions) [3]. While it is important
to investigate bird-window collisions specifically, understanding human perceptions of other
threats is also necessary because this may lead to insights about which conservation actions are
most and least likely to be supported and implemented by the public. Understanding percep-
tions of different threats, as well as willingness to pay and/or willingness to change behaviors
in ways that mitigate these threats, could also lead to more effective conservation strategies
that optimize the tradeoff between addressing the most substantial threats and addressing the
threats for which substantial management inroads are possible.
Conclusions
This study provides novel insight about how important stakeholder groups view and prioritize
benefits and obstacles related to bird-window collision mitigation and prevention. Our
research suggests that substantial advances can be made to reduce bird-window collisions
because both homeowners and conservation practitioners had positive views, suggesting their
receptivity toward and acceptability of collision management measures. However, because of
the more neutral views and more conflicting responses within the homeowner group, our
results also highlight the importance of targeting these stakeholders with education materials
that provide information about bird-window collisions and policies and publicly available
solutions that reduce them. Homeowners are a critical stakeholder group because a large pro-
portion of collisions occur at residential buildings; having their support and participation in
bird-window collision mitigation and prevention could help significantly reduce collisions.
Future research needs related to human dimensions of bird-window collisions and other avian
mortality sources include evaluating perceptions of other stakeholder groups (e.g., architects
and policymakers), studying social-psychological barriers to reducing collisions, determining
willingness to pay for collision mitigation and prevention, and clarifying relative perceptions
about impacts and management of human-related threats other than bird-window collisions.
Because bird-window collisions are a human-caused phenomenon, understanding human per-
spectives and priorities about this issue will be crucial to addressing this threat and thus
benefitting bird populations.
Supporting information
S1 File. SWOT Survey 1. Strengths, weaknesses, opportunities, and threats (SWOT) survey
distributed to all respondents (i.e., Survey 1 described in main text) consisting of all pairwise
comparisons between factors in each SWOT category using a scale of one to nine. For this sur-
vey, all possible pairwise comparisons were made between factors within (but not between)
each SWOT category (e.g., all strengths compared to each other, but strengths not compared
to weaknesses, opportunities, and threats). Analysis of responses to this survey revealed top-
ranked SWOT factors in each category, which were unique to each stakeholder group and
used to generate comparisons in Survey 2.
(PDF)
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
16 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
S2 File. SWOT Survey 2 for homeowners. Strengths, weaknesses, opportunities, and threats
(SWOT) survey distributed to respondents in the homeowner stakeholder group (i.e., Survey 2
for homeowners described in main text) based on their responses to Survey 1. For this survey,
all possible pairwise comparisons were made between the top-ranking factors from each
SWOT category for homeowners (e.g., top homeowner strength compared to top weakness,
opportunity, and threat).
(PDF)
S3 File. SWOT Survey 2 for conservation practitioners. Strengths, weaknesses, opportuni-
ties, and threats (SWOT) survey distributed to respondents in the conservation practitioner
stakeholder group (i.e., Survey 2 for conservation practitioners described in main text) based
on their responses to Survey 1. For this survey, all possible pairwise comparisons made
between the top-ranking factors from each SWOT category for conservation practitioners
(e.g., top conservation practitioner strength compared to top weakness, opportunity, and
threat).
(PDF)
S1 Dataset. SWOT and PCI data analysis. This file contains all response data generated from
strengths, weaknesses, opportunities, and threats (SWOT) Surveys 1 and 2 (see main text and
S1–S3 Files for details about these surveys) along with data and analysis for the potential for
conflict index (PCI).
(XLSX)
Acknowledgments
We thank Christine Sheppard, Daniel Klem, and Stephen Hager for providing preliminary
rankings of candidate SWOT factors, Samantha Cady, Jared Elmore, and Timothy O’Connell
for insightful feedback on methods and an earlier version of the manuscript, and all survey
respondents for their participation. We also thank the handling editor and two anonymous
reviewers for their constructive feedback and suggestions that greatly improved the
manuscript.
Author Contributions
Conceptualization: Georgia J. Riggs, Omkar Joshi, Scott R. Loss.
Data curation: Georgia J. Riggs.
Formal analysis: Georgia J. Riggs.
Funding acquisition: Scott R. Loss.
Investigation: Georgia J. Riggs, Scott R. Loss.
Methodology: Georgia J. Riggs, Omkar Joshi.
Supervision: Omkar Joshi, Scott R. Loss.
Validation: Omkar Joshi.
Writing – original draft: Georgia J. Riggs.
Writing – review & editing: Omkar Joshi, Scott R. Loss.
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
17 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
References
1. United Nations, Department of Economic and Social Affairs, Population Division. World Population
Prospects 2019: Ten Key Findings; 2019.
2. Calvert A, Bishop C, Elliot R, Krebs E, Kydd T, Machtans C, et al. A synthesis of human-related avian
mortality in Canada. Avian Conservation and Ecology. 2013; 8(2).
3.
Loss SR, Will T, Marra PP. Direct mortality of birds from anthropogenic causes. Annual Review of Ecol-
ogy, Evolution, and Systematics. 2015; 46.
4. Nyhus PJ. Human–wildlife conflict and coexistence. Annual Review of Environment and Resources.
2016; 41:143–71.
5.
Loss SR, Will T, Loss SS, Marra PP. Bird–building collisions in the United States: estimates of annual
mortality and species vulnerability. The Condor. 2014; 116(1):8–23.
6. Kim E-M, Choi C-Y, Kang C-W. Causes of injury and mortality of Fairy Pitta Pitta nympha on Jeju Island,
Republic of Korea. Forktail. 2013; 29:145–8.
7. Machtans C, Wedeles C, Bayne E. A first estimate for Canada of the number of birds killed by colliding
with building windows. Avian Conservation and Ecology. 2013;8(2).
8. Aymı´ R, Gonza´lez Y, Lo´ pez T, Gordo O. Bird-window collisions in a city on the Iberian Mediterranean
coast during autumn migration. Catalan Journal of Ornithology. 2017; 33:17–28.
9. David J, Yong DL, Low BW, Owyong A, Alfred C. Anthropogenic Sources of Non-Migratory Avian Mor-
talities In Singapore. The International Journal of Tropical Veterinary and Biomedical Research. 2017; 2
(2):17–27.
10. Santos LPS, de Abreu VF, de Vasconcelos MF. Bird mortality due to collisions in glass panes on an
Important Bird Area of southeastern Brazil. Revista Brasileira de Ornitologia. 2017; 25(2):90–101.
11. Go´mez-Martı´nez MA, Klem D, Rojas-Soto O, Gonza´ lez-Garcı´a F, MacGregor-Fors I. Window strikes:
bird collisions in a Neotropical green city. Urban Ecosystems. 2019; 22(4):699–708.
12. Klem D. Bird: window collisions. The Wilson Bulletin. 1989:606–20.
13. Winger BM, Weeks BC, Farnsworth A, Jones AW, Hennen M, Willard DE. Nocturnal flight-calling
behaviour predicts vulnerability to artificial light in migratory birds. Proceedings of the Royal Society B.
2019; 286(1900):20190364. https://doi.org/10.1098/rspb.2019.0364 PMID: 30940055
14.
Lao S, Robertson BA, Anderson AW, Blair RB, Eckles JW, Turner RJ, et al. The influence of artificial
night at night and polarized light on bird-building collisions. Biological Conservation. 2020; 241:108358.
15. Kahle LQ, Flannery ME, Dumbacher JP. Bird-window collisions at a west-coast urban park museum:
analyses of bird biology and window attributes from Golden Gate Park, San Francisco. PLoS One.
2016; 11(1):e0144600. https://doi.org/10.1371/journal.pone.0144600 PMID: 26731417
16. Nichols KS, Homayoun T, Eckles J, Blair RB. Bird-building collision risk: an assessment of the collision
risk of birds with buildings by phylogeny and behavior using two citizen-science datasets. PloS One.
2018; 13(8):e0201558. https://doi.org/10.1371/journal.pone.0201558 PMID: 30092094
17.
Loss S, Lao S, Eckles J, Anderson A, Blair R, Turner R. Factors influencing bird-building collisions in
the downtown area of a major North American city. PloS One. 2019; 14(11):e0224164–e. https://doi.
org/10.1371/journal.pone.0224164 PMID: 31693699
18. Klem D. Preventing bird–window collisions. The Wilson Journal of Ornithology. 2009; 121(2):314–22.
19. Hager SB, Cosentino BJ, McKay KJ, Monson C, Zuurdeeg W, Blevins B. Window area and develop-
ment drive spatial variation in bird-window collisions in an urban landscape. PloS One. 2013; 8(1):
e53371. https://doi.org/10.1371/journal.pone.0053371 PMID: 23326420
20. Riding CS, O’Connell TJ, Loss SR. Building fac¸ade-level correlates of bird–window collisions in a small
urban area. The Condor. 2020; 122(1):duz065.
21. Hager SB, Cosentino BJ, Aguilar-Go´mez MA, Anderson ML, Bakermans M, Boves TJ, et al. Continent-
wide analysis of how urbanization affects bird-window collision mortality in North America. Biological
Conservation. 2017; 212:209–15.
22. Sheppard C, Fry M, Parr M, Law A. Bird-friendly building design: American Bird Conservancy (ABC);
2011.
23. Sheppard CD. Evaluating the relative effectiveness of patterns on glass as deterrents of bird collisions
with glass. Global Ecology and Conservation. 2019; 20:e00795.
24. San Francisco Planning. Standards for Bird-Safe Buildings. 2019; Available from: https://sfplanning.
org/standards-bird-safe-buildings.
25. Regents of the University of Minnesota. B3 –Building Benchmarks and Beyond. 2017; Available from:
https://www.b3mn.org/.
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
18 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
26. City of Toronto. Bird-friendly Guidelines. 2021; Available from: https://www.toronto.ca/city-government/
planning-development/official-plan-guidelines/design-guidelines/bird-friendly-guidelines/.
27. United States Congress. H.R.1986 –Bird Safe Buildings Act of 2021. 2021; Available from: https://www.
congress.gov/bill/117th-congress/house-bill/1986?s=1&r=3.
28. Wald DM, Jacobson SK, Levy JK. Outdoor cats: Identifying differences between stakeholder beliefs,
perceived impacts, risk and management. Biological Conservation. 2013; 167:414–24.
29. Gramza A, Teel T, VandeWoude S, Crooks K. Understanding public perceptions of risk regarding out-
door pet cats to inform conservation action. Conservation Biology. 2016; 30(2):276–86. https://doi.org/
10.1111/cobi.12631 PMID: 26379227
30. Warren EJ. A clear threat to conservation: using public policy to reduce bird collisions with windows in
homes. 2013.
31. Menacho-Odio RM. Local perceptions, attitudes, beliefs, and practices toward bird-window collisions in
Monteverde, Costa Rica. UNED Research Journal. 2018; 10.
32. Dwivedi P, Alavalapati JR. Stakeholders’ perceptions on forest biomass-based bioenergy development
in the southern US. Energy Policy. 2009; 37(5):1999–2007.
33. Darshini D, Dwivedi P, Glenk K. Capturing stakeholders views on oil palm-based biofuel and biomass
utilisation in Malaysia. Energy Policy. 2013; 62:1128–37.
34.
Joshi O, Parajuli R, Kharel G, Poudyal NC, Taylor E. Stakeholder opinions on scientific forest manage-
ment policy implementation in Nepal. PloS One. 2018; 13(9):e0203106. https://doi.org/10.1371/journal.
pone.0203106 PMID: 30183725
35. Kurttila M, Pesonen M, Kangas J, Kajanus M. Utilizing the analytic hierarchy process (AHP) in SWOT
analysis—a hybrid method and its application to a forest-certification case. Forest Policy and Econom-
ics. 2000; 1(1):41–52.
36. Manfredo M, Vaske J, Teel T. The potential for conflict index: A graphic approach to practical signifi-
cance of human dimensions research. Human Dimensions of Wildlife. 2003; 8(3):219–28.
37. Starr M, Joshi O, Will RE, Zou CB. Perceptions regarding active management of the Cross-timbers for-
est resources of Oklahoma, Texas, and Kansas: a SWOT-ANP analysis. Land Use Policy. 2019;
81:523–30.
38. Arsić S, Nikolić D, Zˇ ivković Zˇ . Hybrid SWOT-ANP-FANP model for prioritization strategies of sustain-
able development of ecotourism in National Park Djerdap, Serbia. Forest Policy and Economics. 2017;
80:11–26.
39. Saaty TL. Decision making—the analytic hierarchy and network processes (AHP/ANP). Journal of Sys-
tems Science and Systems Engineering. 2004; 13(1):1–35.
40. Saaty TL. A scaling method for priorities in hierarchical structures. Journal of Mathematical Psychology.
1977; 15(3):234–81.
41. Klem D. Bird–window collisions: a critical animal welfare and conservation issue. Journal of Applied Ani-
mal Welfare Science. 2015; 18(sup1):S11–S7. https://doi.org/10.1080/10888705.2015.1075832 PMID:
26440494
42. Snep RP, Kooijmans JL, Kwak RG, Foppen RP, Parsons H, Awasthy M, et al. Urban bird conservation:
presenting stakeholder-specific arguments for the development of bird-friendly cities. Urban Ecosys-
tems. 2016; 19(4):1535–50.
43. Kummer J, Bayne E. Bird feeders and their effects on bird-window collisions at residential houses.
Avian Conservation and Ecology. 2015; 10(2).
44. Bernard HR. Research methods in anthropology: Qualitative and quantitative approaches: Rowman &
Littlefield; 2017.
45. YouTube. San Bruno, CA. 2022; Available from: https://www.youtube.com/.
46.
47.
48.
Facebook. Menlo Park, CA. 2022; Available from: https://www.facebook.com/.
Twitter. San Francisco, CA. 2022; Avalable from: https://twitter.com/.
Tumpach C., Dwivedi P., Izlar R., & Cook C. Understanding perceptions of stakeholder groups about
Forestry Best Management Practices in Georgia. Journal of Environmental Management. 2018;
213:374–381. https://doi.org/10.1016/j.jenvman.2018.02.045 PMID: 29502022
49. Catron J, Stainback GA, Dwivedi P, Lhotka JM. Bioenergy development in Kentucky: a SWOT-ANP
analysis. Forest Policy and Economics. 2013; 28:38–43.
50. Qualtrics. Provo, UT. 2022; Available from: https://www.qualtrics.com/.
51. Dillman DA, Smyth JD, Christian LM. Internet, phone, mail, and mixed-mode surveys: the tailored
design method: John Wiley & Sons; 2014.
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
19 / 20
PLOS ONE Stakeholder perceptions of bird-window collisions
52.
Joshi O, Chapagain B, Kharel G, Poudyal NC, Murray BD, Mehmood SR. Benefits and challenges of
online instruction in agriculture and natural resource education. Interactive Learning Environments.
2020:1–12.
53. Krejčı´ J, Stoklasa J. Aggregation in the analytic hierarchy process: why weighted geometric mean
should be used instead of weighted arithmetic mean. Expert Systems with Applications. 2018; 114:97–
106.
54. Vaske JJ, Beaman J, Barreto H, Shelby LB. An extension and further validation of the potential for con-
flict index. Leisure Sciences. 2010; 32(3):240–54.
55.
Lute ML, Carter NH, Lo´pez-Bao JV, Linnell JD. Conservation professionals agree on challenges to
coexisting with large carnivores but not on solutions. Biological Conservation. 2018; 218:223–32.
56. Sekercioglu C¸ H, Wenny DG, Whelan CJ. Why birds matter: avian ecological function and ecosystem
services: University of Chicago Press; 2016.
57. Hedblom M, Knez I, Gunnarsson B. Bird diversity improves the well-being of city residents. Ecology and
Conservation of Birds in Urban Environments: Springer; 2017. p. 287–306.
58. American Bird Conservancy. Glass collisions: preventing bird window strikes. 2022; Available from:
https://abcbirds.org/glass-collisions/.
59. U.S. Fish and Wildlife Service. Buildings & glass. 2021; Available from: https://www.fws.gov/birds/bird-
enthusiasts/threats-to-birds/collisions/buildings-and-glass.php.
60. National Audubon Society. Reducing collisions with glass. 2021; Available from: https://www.audubon.
org/news/reducing-collisions-glass.
61.
Fatal Light Awareness Program Canada. FLAP Canada. 2019; Available from: https://flap.org/.
62. Darko A, Chan APC, Ameyaw EE, Owusu EK, Pa¨ rn E, Edwards DJ. Reivew of application of analytic
hierarchy process (AHP) in construction. International Journal of Construction Management. 2019; 19
(5):436–452.
63. Margles SW, Masozera M, Rugyerinyange L, Kaplin BA. Participatory planning: using SWOT-AHP
analysis in buffer zone management planning. Journal of Sustainable Forestry. 2010; 29(6–8):613–
637.
64. Kollmuss A, Agyeman J. Mind the gap: why do people act environmentally and what are the barriers to
pro-environmental behavior?. Environmental Education Research. 2002; 8(3):239–60.
65. Hoffman AJ, Henn R. Overcoming the social and psychological barriers to green building. Organization
& Environment. 2008; 21(4):390–419.
66.
67.
Tsi E, Ajaga N, Wiegleb G, Mu¨ hlenberg M. The willingness to pay (WTP) for the conservation of wild
animals: Case of the Derby Eland (Taurotragus derbianus gigas) and the African wild dog (Lycaon pic-
tus) in North Cameroon. African Journal of Environmental Science and Technology. 2008; 2(3):51–8.
Zander KK, Ainsworth GB, Meyerhoff J, Garnett ST. Threatened bird valuation in Australia. PloS One.
2014; 9(6):e100411. https://doi.org/10.1371/journal.pone.0100411 PMID: 24955957
68. Steven R, Morrison C, Castley JG. Exploring attitudes and understanding of global conservation prac-
tice among birders and avitourists for enhanced conservation of birds. Bird Conservation International.
2017; 27(2):224–36.
69. Getzner M, Meyerhoff J, Schla¨pfer F. Willingness to pay for nature conservation policies in state-owned
forests: An Austrian case study. Forests. 2018; 9(9):537.
70. Kudryavtsev A, Stedman RC, Krasny ME. Sense of place in environmental education. Environmental
Education Research. 2012; 18(2):229–50.
71. Nielsen-Pincus M, Sussman P, Bennett DE, Gosnell H, Parker R. The influence of place on the willing-
ness to pay for ecosystem services. Society & Natural Resources. 2017; 30(12):1423–41.
PLOS ONE | https://doi.org/10.1371/journal.pone.0263447 February 10, 2022
20 / 20
PLOS ONE
| null |
10.1038_s41598-021-03032-1.pdf
|
Data availability
All data are available in the manuscript.
|
Data availability All data are available in the manuscript. Received: 24 May 2021; Accepted: 17 November 2021
|
OPEN
In‑vitro propagation
and phytochemical profiling
of a highly medicinal and endemic
plant species of the Himalayan
region (Saussurea costus)
Ajmal Khan1, Azhar Hussain Shah2* & Niaz Ali1
Efficient protocols for callus induction and micro propagation of Saussurea costus (Falc.) Lipsch were
developed and phytochemical diversity of wild and in‑vitro propagated material was investigated.
Brown and red compact callus was formed with frequency of 80–95%, 78–90%, 70–95% and 65–80%
from seeds, leaf, petiole and root explants, respectively. MS media supplemented with BAP (2.0
mgL−1), NAA (1.0 mgL−1) and GA3 (0.25 mgL−1) best suited for multiple shoot buds initiation (82%),
while maximum shoot length was formed on media with BAP (1.5 mgL−1), NAA (0.25 mgL−1) and
Kinetin (0.5 mgL−1). Full strength media with IAA (0.5 mgL−1) along with IBA (0.5 mgL−1) resulted
in early roots initiation. Similarly, maximum rooting (87.57%) and lateral roots formation (up to
6.76) was recorded on full strength media supplemented with BAP (0.5 mgL−1), IAA (0.5 mgL−1) and
IBA (0.5 mgL−1). Survival rate of acclimatized plantlets in autoclaved garden soil, farmyard soil,
and sand (2:1:1) was 87%. Phytochemical analysis revealed variations in biochemical contents i.e.
maximum sugar (808.32 µM/ml), proline (48.14 mg/g), ascorbic acid (373.801 mM/g) and phenolic
compounds (642.72 mgL−1) were recorded from callus cultured on different stress media. Nonetheless,
highest flavenoids (59.892 mg/g) and anthocyanin contents (32.39 mg/kg) were observed in in‑vitro
propagated plants. GC–MS analysis of the callus ethyl acetate extracts revealed 24 different
phytochemicals. The variability in secondary metabolites of both wild and propagated plants/callus
is reported for the first time for this species. This study may provide a baseline for the conservation
and sustainable utilization of S. costus with implications for isolation of unique and pharmacologically
active compounds from callus or regenerated plantlets.
Plants have been essential sources of medicine for thousands of years and nearly 80% of the world’s popula-
tion still relies on traditional medicine for their primary healthcare1. Saussurea costus is an endemic species in
geographically limited places of the Himalayas, where it grows on moist slopes at altitudes of 2500–4000 m. The
species is critically endangered and is listed in Appendix I of the Convention on International Trade in Endan-
gered Species of Wild Fauna and Flora (CITES). In addition, it is one of the 37 endangered and highly medicinal
plants of the Himalayas, and has been prioritized for both in-situ and ex-situ conservation2. S. costus is a highly
prized medicinal plant in the Kaghan valley Pakistan. Roots of S. costus have sweet and strong aromatic odor
with bitter taste and are used as antiseptics as well as for treating bronchial asthma, especially of the vagotonic
type. The roots of S. costus have been widely used for curing diarrhea, jaundice, stomachache, respiratory tract
infections, antispasmodic agents against spasms caused by asthma, cholera, rheumatism, chronic skin diseases
and leprosy3. Further, oil extracted from the roots (referred to as Costus oil) is used for making high grade
perfumes and hair oils4. In addition, many studies have shown that extracts of S. costus have potent anti-cancer,
anti-inflammatory and anti-ulcer properties5. Because of the high demands for roots, most natural populations
of S. costus are on the verge of extinction6.
In order to avoid the future loss of endangered, endemic and rare species, conservation of plant genetic
resources has long been realized as an integral part of biodiversity conservation. Plant cell and tissue culture
1Department of Botany, Hazara University Mansehra, Mansehra, KP, Pakistan. 2Department of Biotechnology and
Genetic Engineering, Hazara University Mansehra, Mansehra, KP, Pakistan. *email: [email protected]
Scientific Reports | (2021) 11:23575
| https://doi.org/10.1038/s41598-021-03032-1
1
Vol.:(0123456789)www.nature.com/scientificreports Samples
CPM
CPM-1
CPM-2
CPM-3
CPM-4
Kinetin (mgL-1)
2,4-D (mgL-1)
Callus stress
0.5
0.5
0.5
0.5
0.5
1.0
1.0
1.0
1.0
1.0
–-
D-Sorbitol (60 gL-1)
D-Manitol (60 gL-1)
Poly ethylene glycol 600 (5 gL-1)
Sucrose (60 gL-1)
Table 1. Different stress for phytochemical comparison of callus grown on simple callus promoting media
(CPM) supplemented with Kinetin (0.5 mgL-1) and 2.4-D (1.0 mgL-1) and callus subjected to various stresses
i.e. 60 gL-1 D-Sorbitol stress (CPM-1), 60 gL-1 D-Manitol (CPM-2), 5 gL-1 Poly ethylene glycol 600 (CPM-3)
and 60 gL-1 Sucrose (CPM-4).
has been a powerful tool for rapid propagation and biomass production of valuable species. To overcome envi-
ronmental constraints in-vitro cultures (cell, callus, buds and shoot) provide the best alternative choice for the
smooth and constant supply of plant active ingredients7. However, there are no effort in literature for ex-situ
conservation and micro propagation of S. costus. Further, phytochemical composition of the wild (natural) S.
costus and tissue culture generated material is totally non-existent. The purpose of this research was to establish
an effective and efficient in vitro regeneration protocol for S. costus and to compare the photochemical variability
in the aqueous extracts of induced callus, in-vitro propagated plants with the wild/natural collections.
Materials and Methods
Plant material and sterilization procedure. Mother plant was collected from wild populations in
Makra, Kaghan valley, Pakistan (lat 34.57439º N, long 073.49580º E, alt 3,878 m). The specimen was identified
by Dr. Abdul Majid Department of Botany Hazara University Mansehra, and scientific name validated online
(http:// www. thepl antli st. org/). Voucher specimen was submitted to the Herbarium Department of Botany Haz-
ara University, Mansehra. Healthy plant parts (explant) were separated from the mother plant, washed and
sterilized following Yesmin et al. (2016)8.
Culture conditions. The basal MS media9 was used with various concentration and composition of growth
regulators (BAP, IAA, IBA, NAA, 2,4-D and kinetin). All culture media were agitated with 7% technical agar and
3% sucrose. The pH of media was set to 5.8 before addition of agar. These media were autoclaved at 121 °C for
20 min at 15 psi. Cultures were maintained in a culture room incubated with a 16-h light cycle and temperature
maintained at 25 ± 2 °C with 50% humidity.
Callus induction. Growth regulators such as 2,4-D (0.25, 0.5, and 1.0 mgL−1) in combination with varied
concentration of Kinetin (0.5, 1.0, 1.5 and 2.0 mgL−1), and four explants types (seeds, leaf, petiole and internode)
were compared for callus induction. Explants were subjected to two subcultures at an interval of fourteen days10.
Shoot bud initiation. Full strength MS media with different concentration of BAP (0.5, l.0, 2.0 mgL−1),
NAA (0, 0.25, l.0 mgL−1) and GA3 (0, 0.25, 1.0 mgL−1) were compared for shoot buds initiation. The percentage
of shoot induction, time taken for bud initiation and the growth state of the buds were measured after four weeks
of culturing.
Shoot proliferation. Nodal segments (1–2 cm long) were excised from cultured plant and transferred into
MS media agitated with BAP (0.5, 1.0, 1.5 mgL−1) in combination with NAA (0.25, 0.25 and 0.5 mgL−1) and
Kinetin (0, 0.25 and 0.5 mgL−1) in order to maximize shoot multiplication. In addition, basic MS media with dif-
ferent plant growth regulators were compared during the phase of subculture, and the optimal media for shoot
proliferation was selected.
Root initiation. To optimize root induction media, full-strength MS media was supplemented with differ-
ent combination and concentrations of IAA (0.5, 1.0 mgL−1) and IBA (0.5, 0.1 mgL−1) along BAP (0.5 mgL−1).
The time to root initiation was observed and recorded after every two days. Data on average root numbers and
length were recorded after 45 days of culturing.
Photochemical analysis. Treatments of in-vitro callus. After subculture for eight cycles, fourteen days
old callus was subjected to four different stresses each having five (05) replications. Callus was cultured on callus
promoting media (CPM) having 0.5 and 1.0 mgL−1 Kinetin and 2.4-D respectively. In addition, callus was cul-
tured on media agitated with 60 gL−1 D-Sorbitol stress (CPM-1), 60 gL−1 D-Manitol stress (CPM-2), 5 gL−1 Poly
ethylene glycol 600 stress (CPM-3) and 60 gL−1 Sucrose stress (CPM-4) for 120 days, while the callus cultured on
CPM alone was considered as control (Table 1). Similarly, different concentrations of growth hormones given to
calli for bud, shoot or root induction are given (see Tables 2–5).
Samples preparation. Samples for spectrometry (BMS, UV-1900) were prepared following Storey and Jones
(1975)11. Total sugars contents were analyzed following Dubois et al. (1956)12, proline content was assessed fol-
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
2
Vol:.(1234567890)www.nature.com/scientificreports/ Treatments
Conc. 2,4-D/Kinetin mgL-1
Type of explants used
T1
0.25/0.5
T2
0.5/0.5
T3
0.5/1.0
T4
0.5/1.5
T5
0.5/2.0
T6
1.0/2.0
Seeds
Leaf
Petiole
Root
Seeds
Leaf
Petiole
Root
Seeds
Leaf
Petiole
Root
Seeds
Leaf
Petiole
Root
Seeds
Leaf
Petiole
Root
Seeds
Leaf
Petiole
Root
Means days to callus induction
(x- ± SE)
14.00 ± 1.09A
15.20 ± 0.73C
15.80 ± 0.734BC
16.40 ± 0.97A
17.60 ± 0.67A
17.20 ± 0.37C
17.80 ± 0.37AB
19.20 ± 0.37AB
13.40 ± 0.60A
15.80 ± 1.07C
15.60 ± 0.50C
17.40 ± 0.86BC
16.60 ± 0.60A
18.60 ± 1.07AB
18.60 ± 0.50A
17.20 ± 0.86BC
17.40 ± 1.07A
18.20 ± 0.66AB
18.40 ± 0.50A
19.40 ± 0.74AB
18.20 ± 0.80A
19.60 ± 0.81A
19.01 ± 0.89A
20.80 ± 0.73A
Callus growth after 30 days
(x- ± SE)
0.22 ± 0.07D
1.50 ± 0.08AB
0.91 ± 0.02D
0.91 ± 0.05BC
1.39 ± 0.06CD
1.34 ± 0.04AB
1.27 ± 0.08BC
0.92 ± 0.02B
1.86 ± 0.06A
1.65 ± 0.06A
1.42 ± 0.05A
1.114 ± 0.07A
1.62 ± 0.05B
1.32 ± 0.08AB
1.33 ± 0.02AB
0.75 ± 0.08C
1.54 ± 0.08BC
1.23 ± 0.06BC
1.26 ± 0.03BC
0.90 ± 0.01BC
1.65 ± 0.06B
0.92 ± 0.20C
1.18 ± 0.06C
0.84 ± 0.03BC
Table 2. In-vitro callus induction and callus growth after 30 days of culturing of S. costus using seed, leaf,
petiole and root as an explants.
lowing Bates et al. (1973)13, flavenoids were assessed as per Csepregi et al. (2013)14. Antioxidant activity was as
described in Re et al. (1999)15, total phenol contents was measured following Singleton and Rossi (1965)16 and
total anthocyanin content was determined following Giusti and Wrolstad (2001)17.
Preparation of solvent extraction for GC–MS. Callus subjected to different stresses (Table 1) as well as grown
on CPM was shade dried and grounded to fine powder using mortar and pestle. For solvent preparation 1 g (dry
weight) of powder was soaked in 10 ml of ethyl acetate for 2 days. The sample was centrifuged at 8,000 rpm for
5 min and the supernatant collected was stored at 4 °C for further analyses18.
Gas chromatography-mass spectrometry (GC–MS) analysis. Chemical analysis of ethyl acetate extract was car-
ried out using gas chromatography coupled with mass spectrometry (GC–MS) with a Hewlett Packard GC–
MS system (PerkinElmer precisely, Carlus 600C). The relative percentage of each component was calculated by
comparing the average GC chromatogram peak to the total area. The mass detector used in this analysis was
Turbo-Mass Gold-Perkin-Elmer, and the software adopted to handle mass spectra and chromatograms was a
Turbo-Mass ver-5.419.
Identification of compounds.
Interpretation on mass spectrum GC–MS was conducted using the database of
National Institute Standard and Technology (NIST). The spectrum of a component was compared with the spec-
trum of the known components stored in the NIST library. Similarly, name, molecular weight and structure of
the components of the test materials were ascertained19.
Statistical analysis. Statistical analysis was performed with Statistic 8.1 (Trial version). Results were presented
as mean ± standard error (SE), and the data was analyzed by one way Analysis of variance (ANOVA) at 0.05%
confidence level (p < 0.01). All in-vitro propagation treatments had 5 replications whereas; the phytochemical
analyses had three replications for each treatment.
Results and discussion
Callus induction. Callus response was influenced by hormonal combinations as well as the type of explant
used. The callus response varied i.e. 80–95%, 78–90%, 70–95% and 65–80% for seeds, leaf, petiole and root
explants, respectively (Fig. 1A-H). Similarly, explants were grown on MS media alone (as control) for 14 days
and no callus induction or regeneration was observed and therefore, these results are not included. Maximum
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
3
Vol.:(0123456789)www.nature.com/scientificreports/ Figure 1. In-vitro callus induction of S. costus using seeds (A, B, C), leaf (D, E), roots (F) and nodes (G) as an
explant.
amount of callus tissue per seed explant was formed on MS media agitated with 2,4-D (0.5 mgL−1) and Kinetin
(1.0 mgL−1) as demonstrated in Fig. 1A-C. The colour of callus ranged from white to dark brown. Successful
callus initiation was observed after 13, 15, 15 and 16 day of culturing from seed, leaf, petiole and root explants
respectively (Table 2). It was also noted that subculture of callus into new media increased the callus biomass.
Maximum callus growth from seed (1.86 g), leaf (1.65 g), petiole (1.42 g) and root (1.14 g) were record at 2, 4-D
(0.5 mgL−1) and Kinetin (1.0 mgL−1) after twenty-eight days of culture (Table 2). Higher concentration of 2,4-D
reduced callus induction and it was observed that the colour changed to brown with hard texture, followed by
necrosis. Although 2,4-D is a synthetic plant growth regulator, its role in callus induction is highlighted for S.
costus. Previous studies have also reported the efficacy of exogenous 2,4-D in other medicinal plants. Hassan
et al. (2009)20 and Sen et al. (2014)21 have shown the positive role of 2,4-D plant growth hormones in culture
media of W. somnifera, I. obscura, A. precatorius and C. halicacabum and their results are in agreement to those
mentioned here. The effect of 2,4-D in combination with Kinetin demonstrated the potential of a synthetic plant
growth regulators in the production of callus from seeds, leaf, petiole as well as root explants of S. costus as a
potent plant growth regulator.
Shoot bud initiation. Auxiliary buds induction was observed after 15 to 20 days of culturing (Fig. 2A,B).
The earliest shoot bud initiations were observed on media agitated with BAP (2 mgL−1), NAA (1 mgL−1) and
GA3 (0.25 mgL−1). Higher concentration of BAP resulted in earlier buds induction. The analysis revealed BAP
had a marked influence on the rate of induction. Similarly, BAP in low concentration, the induction rate was
64% and the lateral buds sprouted late. In addition, new buds were relatively thinner and delicate. ANOVA
showed that shoot bud initiation was highly significant among the treatments (Table 3). Previous studies have
also indicated that high level of BAP and low GA3 induced greater response to shoot buds initiation22. Similarly,
BAP here was most effective for bud induction. GA3 contributes to the initiation and elongation of auxiliary buds
and expansion of leaves23. Further, GA3 regulates the growth and development of plants, mainly by stimulating
mitotic division and cell elongation24. It was found that high level of GA3 effectively increased stem length, while
lower GA3 concentration inhibited potato shoot growth25. Further, GA3 has long been used to break dormancy
and to stimulate shoot elongation in different species of magnolias26. In line with the previous reports, it was also
observed that BAP in combination with GA3 was important for bud initiation, reducing time for buds initiation
as well as resulted in stronger buds27.
Shoot bud proliferation. Full strength media augmented with BAP (0.1 mgL−1), NAA (0.25 mgL−1) and
Kinetin (0.25 mgL−1) proved best for shoot bud proliferation and elongation (Fig. 2C). Significant differences
were observed in multiplication rate and numbers of shoots between T7, T8 and T9, although T13 is significantly
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
4
Vol:.(1234567890)www.nature.com/scientificreports/ Figure 2. In-vitro Propagation of S. costus using nodal explant (A) Auxiliary buds, (B) Shoot buds initiation,
(C) Multiple shoot initiation, (D, E, F) Mature plantlets, (E, F) Roots initiation, (G, H) Acclimatization of plants.
Treatment
T7
T8
T9
T10
T11
T12
BAP mgL-1) NAA mgL-1 GA3 mgL-1
0.5
0
0
0.5
1.0
1.0
2.0
2.0
0.25
0
0.25
1.0
1.0
0
0.25
0.25
0.25
1.0
19 Days
Bud induction rate (x- ± SE) Time to bud initiation Growth state of bud
19.40 ± 1.50A
17.80 ± 1.02AB
17.60 ± 0.37AB
17.00 ± 0.44AB
15.60 ± 0.87B
16.40 ± 0.67B
15 Days
18 Days
17 Days
18 Days
16 Days
+ + +
+ + +
+ + +
+ + +
+ +
+
Table 3. Influence of different plant growth regulators on buds initiation and Range analysis. Vigorous and
green buds (+ + +); healthy buds (+ +); weak bud ( +). Each Value represents the mean ± SE of five replicates.
Significant deference at P ≤ 0.05, x- ± Sd- average ± Standard deviation, x ± SE – average ± Standard error.
Treatments
BAP (mgL-1) NAA (mgL-1) Kinetin (mgL-1)
shoots numbers per plant
(≧0.5 cm) (mean ± SE)
T13
T14
T15
0.5
1.0
1.5
0.25
0.25
0.25
0
0.25
0.5
2.20 ± 0.44
4.00 ± 0.70
2.80 ± 0.83
shoots numbers per plant
(≧0.5 cm) (mean ± SE)
2.20 ± 0.20B
4.00 ± 0.31A
2.80 ± 0.37B
Total shoot length (≧0.5 cm)
(mean ± SE)
2.19 ± 0.26B
2.50 ± 0.30AB
3.11 ± 0.33A
Table 4. The influence of different concentrations of plant growth regulators on bud proliferation of S. costus.
vigorous and green buds (+ + +); healthy buds (+ +); weak buds ( +). Each value represents the mean ± SE of five
replicates.
different from T14 and T15, while T13 and T14 are not significantly different (Table 4). Das et al. (2020)28 has
also recorded maximum number of shoot/explant of B. polystachyon with combination of BAP (13.32 μM) and
NAA (0.53 μM). Several medicinal plants such as, C. paniculatus 29 and C. blumei 30, have shown similar results
and BAP with NAA have been reported as being the most effective in direct organogenesis. Our results are in
alignment to those of Kaur et al. (1998)31, where 8–10 shoot/explants of A. catechu from nodal segment on media
containing BAP (4.0 mgL−1) with NAA (0.5 mgL−1) were reported.
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
5
Vol.:(0123456789)www.nature.com/scientificreports/ Treatment
BAP (mgL-1)
IAA (mgL-1)
IBA (mgL-1)
T16
T17
T18
0.5
0.5
0.5
1.0
0
0.5
0
1.0
0.5
Days to root initation
(mean ± SE)
18.80 ± 1.01A
16.20 ± 0.80AB
13.80 ± 0.91B
Number of roots
(mean ± SE)
3.84 ± 0.508B
5.32 ± 0.531AB
6.76 ± 0.733A
Total root length
(cm) (mean ± SE)
1.50 ± 0.19B
2.53 ± 0.21A
2.27 ± 0.24A
Table 5. The effect of different concentrations of IAA and IBA along with BAP on roots initiation, number of
roots and total root length of S. costus. Each value represents mean ± SE of five replicates.
Total shoot length. Average shoot length ranged from 2.19 to 3.11 cm among the treatments. Maximum
shoot length was recorded for media fortified with BAP (1.5 mgL−1), NAA (0.25 mgL−1) and Kinetin (0.5 mgL−1).
While, the minimum shoot length was recorded in media with BAP (0.5 mgL−1), NAA (0.25 mgL−1) without
Kinetin (Fig. 2D,E). ANOVA revealed significant variation in T15 compared to T13 and T14 (Table 4). The
addition of even smaller amounts of BAP or NAA help inducing adventitious shoot formation by increasing
propagation coefficient32. Other researchers have also reported that highest shoot length (3.73 ± 0.14 cm) of S.
rebaudiana was observed on MS media supplemented with BAP (2.0 mgL−1) and IAA (0.25 mgL−1) after 15 days
of culturing33. Additionally, higher concentrations of BAP reduces shoot length, which is in agreement to the
known literature34.
Root initiation, number of roots and total root length. Roots induction is a critical step in successful
in-vitro propagation experiments; here combination of IAA (0.5 mgL−1) with IBA (0.5 mgL−1) resulted in earliest
roots initiation (13 days), while IAA (0.1 mgL−1) delayed late root formation (19 days) (Fig. 2D-E). Further, IAA
(0.5 mgL−1) in combination with IBA (0.5 mgL−1) resulted in earliest as well as plenty of lateral roots formation
(Table. 5). ANOVA showed that TI6 was significantly different, while T17 and T18 had no significant variation
(Table 5). IBA is a highly stable and potential auxin for roots induction35. Maximum numbers of roots (6.76)
were recorded on full strength media supplemented with BAP (0.5 mgL−1), IAA (0.5 mgL−1) and IBA (0.5 mgL−1)
(Fig. 2E). On the contrary, least number of roots per plant (3.84) were formed on media supplemented with
IAA (1 mgL−1). Statistical analysis revealed that T16 and T18 varied significantly (Table 5). The in-vitro derived
shoots on MS medium were supplemented with a range of concentrations of two auxins (IAA and IBA) for
75 days, it was observed that the lower concentrations of BAP (0.5 mgL−1) in combination with IBA (1 mgL−1)
resulted in a higher root length (2.53 cm), while IAA (1mgL−1) and IBA (1 mgL−1) alone induced roots length of
(1.5 cm) and (2.27 cm) respectively. Results showed that IAA in comparison to IBA reduced roots length when
compared at the same concentration (Table 4). Statistical analysis showed that root length at T16 was signifi-
cantly different from T17 and T18. Cheepala et al. (2004)36 reported that IAA is a widely used auxin for rooting
in A. stenosperma and A. villosa. In several other plants species the promoting effect of IBA in rooting has also
been reported37. In contrast, induction of rooting of G. scabra was obtained on NAA (0.3 mgL−1) and IAA (0.1
mgL−1) containing media38. Similarly higher percentage of rooting were obtained in half strength MS media
with NAA (1.0 mgL−1), were as full strength medium with NAA (1.5 mgL−1) was the best media for rooting10.
Bekheet (2013)39 has indicated that addition of IAA, IBA or NAA (1 mgL−1) induced rooting of in-vitro grown
P. dactylifera. However, in the present study, IAA in combination with IBA was found to be the most efficient in
multiple shoots induction, followed by IBA alone.
Acclimatization. The ultimate success of all in-vitro micro propagation endeavors heavily relies on the
higher survival rates of such plantlets. Direct field transfers of the plantlets do not allow acclimatization of the
in-vitro generated plants as they fail to establish successful interactions with the soil microbes and/or to sus-
tain the environmental conditions40. Here, well rooted micro propagated plantlets were transferred into plastic
pots containing autoclaved garden soil, farmyard soil, and sand (2:1:1) as shown in Fig. 2F–H. The plants were
then acclimatized in the growth room at 27 °C temperature for 2 weeks followed by another 3 weeks at room
temperature under laboratory conditions. Finally, 35–40 days old plantlets were transferred to nursery where,
morphological anatomical and growth characteristics were observed (results not shown) and survival efficiency
recorded. Out of 92 plantlets, 80 (87%) could successfully acclimatize and the relatively low mortality rate here
is likely to be due to the biohardening of the micropropagated plants achieved prior to their nursery transfer.
Similarly, we have given water to the plantlets after 6 days interval and that too very close to the roots and have
avoided leaves. This approach has been previously reported beneficial for in-vitro raised plants41 and the survival
rate could be raised significantly higher if biotization of the explants is attempted42.
Phytochemical variation. Total sugar contents. Total sugar contents revealed significant variation with
treatments. Maximum sugar contents (808.326 µM/ml) was observed in callus cultured on CPM-4 supplement-
ed with 60 gL−1 sucrose, while the lowest sugar contents (16.64 µM/ml) was noted in wild plants (Fig. 3A).
Accumulation of sugars contents in different parts of plants increases in response to a variety of environmental
stresses43. The accumulation of total sugars is associated with adaptation of plants to various environmental
stresses44. The results shown here are in agreement to earlier findings where salinity increased total sugar con-
tents in leaves of in-vitro propagated P. euphratica. Similarly, addition of NaCl (250 mmoll−1) increased sugars
contents by 2.7 times45. In calli of M. arborea total sugars account for about 90% of the total dry weight and there
were no significant differences. The remarkable differences between the embryogenic and non-embryogenic
calli of M. arborea, was the amount of sugar found in embryogenic calli46. A similar trend with total sugars ac-
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
6
Vol:.(1234567890)www.nature.com/scientificreports/ Figure 3. Variation in the Total Sugar contents (A) and proline contents (B) of S. costus collected from wild,
in-vitro propagated plant and induced callus. Different stress for phytochemical comparison of callus grown on
simple callus promoting media (CPM) supplemented with Kinetin (0.5 mgL-1) and 2.4-D (1.0 mgL-1) and callus
subjected to various stresses i.e. 60 gL-1 D-Sorbitol stress (CPM-1), 60 gL-1 D-Manitol (CPM-2), 5 gL-1 Poly
ethylene glycol 600 (CPM-3) and 60 gL-1 Sucrose (CPM-4).
cumulation was also detected in P. kurroa47. In line results are also shown for the total sugars in selected calli of
D. caryophyllus subjected to different concentrations of culture filtrate that were significantly higher than those
of non selected calli48.
Proline content. Since, callus promoting media was used as a control; the stresses imposed increased proline
content in callus from 1.63 to 48.14 mg/g F.Wt. The variability in proline content among the different treatments
were highly significant as shown in Fig. 3B. Maximum proline contents was noticed in CPM-2 agitated with
60 gL−1 D-Manitol (48.14 mg/g) whereas, minimum was observes in CPM-3 supplemented with 5 gL−1 Poly
ethylene glycol 600 (1.63 mg/g). In brief, different stresses enhanced proline contents in S. costus callus as fol-
lows: callus treated with 60 gL−1 D-Manitol (48.14 µM/g) > 60 gL−1 D-Sorbitol (18.45 µM/g) > 60 gL−1 Sucrose
(17.79 µM/g) > 5 gL−1 Poly ethylene glycol 600 (1.63 µM/g). Results presented here are in general agreement to
earlier reports where, authors have reported proline accumulation in calli of sugarcane grown on different con-
centration of PEG49. Similarly, total proline level of 20% PEG selected calli was reported to be 17 times higher
than the non-selected calli of O. sativum50. Pradhan et al. (2021)51 reported on the increasing trends in proline
contents (0.798 µMg−1) in mango callus subjected to 15% PEG stress as compare to control (no PEG) with the
value of 0.080 µM g−1 FW. Similar increase in proline contents is also mentioned for H. annuus52 as well as rice
in response to PEG stress53. Previously, D-sorbitol stress has resulted in more than four-fold increase in proline
level in maize seedling54 and these results are in full agreement to those reported here.
Total flavonoids. Favonoids have protective functions for plants growing in soils that are rich in toxic met-
als. Here flavonoids contents showed significant variation among wild and in-vitro propagated plant as well as
induced callus of S. costus. This variation in flavonoids content ranged from 0.90 to 59.89 mg/g. Results showed
that in-vitro propagated plants had the highest flavonoids contents (59.892 mg/g) followed by plants collected
from wild (49.199 mg/g). Similarly, lowest flavonoids contents were reported in callus agitated on CPM-4 sup-
plemented with 60 gL−1 Sucrose (Fig. 4A). Comparing wild and in-vitro propagated plants, callus contained rela-
tively scarce amount of flavonoids, and this is most likely due to plants were grown in laboratory environment
with very much uniform environmental conditions. In natural habitats plants are well adapted and have evolved
mechanisms to minimize injuries under extreme environmental conditions. The accumulation of flavonoids in
the cells resulted by osmotic stress are often associated with a mechanisms that allow plants to tolerate harmful
effects of water shortage. Further, accumulation of these solutes lower the osmotic potential of plant tissues at
cellular level and hence allowing plants to sustain growth in stressful environments55. Ibrahim et al. (2018)56 have
also noted maximum level of total flavonoids in wild or natural P. barbatus as compared to in-vitro propagated
plants and callus57.
Ascorbic acid contents. Maximum accumulation of Ascorbic acid (373.801 mM/g) was recorded in callus cul-
tured on CPM-1 supplemented with 60 gL−1 D-Sorbitol, followed by CPM-2 that had subjected to (60 gL−1)
D-Manitol stress (373.801 mM/g), while minimum amount was observed in callus grown on CPM (104.95 mM/g)
(see Fig. 4b). Kamal et al. (2020)58 have studied the optimization of suitable media for callus induction and ascor-
bic acids accumulation in Chinese cabbage cultivars. The authors have found maximum ascorbic acid accumula-
tion in callus of root explant cultured on TDZ (1.0 mgL−1), NAA (0.25 mgL−1) and AgNO3 (5.0 mgL−1), while
minimum ascorbic acid was noted for callus grown from hypocotyl tissues cultured on TDZ (1.0 mgL−1), NAA
(1.0 mgL−1) and AgNO3 (9.0 mgL−1). Likewise, using the DPPH assay, free radical scavenging and antioxidant
potential of in-vitro propagated S. corymbosa plants were compared with the wild plants in where the wild plants
have shown highest free radical scavenging activity compared to the in-vitro propagated plants59.
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
7
Vol.:(0123456789)www.nature.com/scientificreports/ Figure 4. Variation in the Total Flavenoids (A) and Ascorbic acid (B) contents of S. costus collected from wild,
micro propagated plant and induced callus grown on simple callus promoting media (CPM) supplemented with
Kinetin (0.5 mgL-1) and 2.4-D (1.0 mgL-1) and callus subjected to various stresses i.e. 60 gL-1 D-Sorbitol stress
(CPM-1), 60 gL-1 D-Manitol (CPM-2), 5 gL-1 Poly ethylene glycol 600 (CPM-3) and 60 gL-1 Sucrose (CPM-4).
Figure 5. Variation in the Total Phenolics (A) and Anthocyanin (B) of S. costus collected from wild, micro
propagated plant and induced callus. Different stresses for phytochemical comparison of callus grown on
simple callus promoting media (CPM) supplemented with Kinetin (0.5 mgL-1) and 2.4-D (1.0 mgL-1) and callus
subjected to various stresses i.e. 60 gL-1 D-Sorbitol stress (CPM-1), 60 gL-1 D-Manitol (CPM-2), 5 gL-1 Poly
ethylene glycol 600 (CPM-3) and 60 gL-1 Sucrose (CPM-4).
Total phenolic compounds. Phenolics compounds represent a diverse array of plant secondary metabolites,
which are predominantly used as powerful scavengers of free radicals (Pietta, 2000)60. Here, highest phenolic
contents (642.72 mgL−1) accumulated in calli cultured on CPM when compared to wild or in-vitro propagated
plantlet. Similarly, lowest levels of phenolic compounds (420 mgL−1) were recorded in plants collected from
wild (Fig. 5A). Increase in phenolic compounds accumulation (37% and 34%) was observed in callus treated
with 100 mgL−1 yeast extract and 50 mgL−1 salicylic acid24. These finding are supported by those given in El-
Nabarawy et al. (2015)61, where the culture medium supplemented with low concentration of yeast extract
increased phenolic accumulation in micro propagated plants. Furthermore, Gorni and Pacheco (2016)62 have
reported that A. millefolium treated with 0.5 and 1.0 mM salicylic acid significantly increases phenolic contents.
A slight increase in total phenolic content was found in callus treated with glycine (200 mgL−1), yeast extracts
(500 mgL−1) and salicylic acid (100 mgL−1). This increase of phenolic contents in callus cultures was related
to mitochondrial activity; that is, while the cell dehydrogenase activity (FADH2/NADH) and the cytochrome
C-oxidase decrease, the production of phenolic compounds increases63. On the other hand, variation in total
phenolics within the mother source plant, micropropagated plants and callus subjected to different stresses may
be attributed to changes in the levels of various phytohormones or other endogenous physiological pathways
that occur in plant64. Also synthetic plant growth regulators used during the micro propagation pathways make
a significant contribution in the production of secondary metabolites within the in-vitro cultured cells and tis-
sues by controlling the expression of genes involved in the synthesis of secondary metabolites such as shikimate
and flavonoids65.
Total anthocyanin. Anthocyanin contents were detected in the form of Pelargonidin-3-glucoside per kilogram
of fresh sample. In the current analyses, in-vitro propagated plant possessed highest amounts of anthocyanin
(32.39 mg/kg) followed by wild (31.84 mg/kg) whereas, lowest amount of anthocyanin was recorded in callus
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
8
Vol:.(1234567890)www.nature.com/scientificreports/ RT
10.13
13.04
14.23
14.41
14.92
14.98
15.24
16.64
18.4
18.31
18.32
19.13
20.7
21.6
21.7
21.9
23.67
24.65
25.24
25.59
26.4
27.52
27.73
28.7
Compound name
Molecular mass Cas number
% area % area
% area
% area
% area
CPM
CPM-1 CPM-2 CPM-3 CPM-4
Dodecane,2,6,10-
trimethyl
Hexadecane
Bezene,1,4-bis(1,1-
dimethylethyl)
Benzene,
1,3bis(dimethylethyl)
Nonadecane,2,6,10,14
tetramethyl
Octacosane,1-Iodo
Octadecane-2,6,10,14-
tetramethyl
Hentriacontane
Heptadecane,2,6,10,15-
Tetramethyl
Tritetracontane
Dodecane,1-fluoro
Carbonic acid,decyl
undecyl ester
6-tetradecanesulfonicacid,
butyl ester
Nonadecane,2,6,10,14,18-
pentamethyl
Nonadecane,2,6,10,14
-tetramethyl
Eicosane,2,6,10,14,18-
pentamethyl
Heneisane
Propanic acid,2-
methyl-,3,7-dimethyl-
2,6-octadienyl ester
Alpha-maaliene
Dotriacontane
Tetrapentacotane
Dotricotane
Triacotane
Selina-3,7(11)-diene
212
226
190
190
324
520
310
436
296
604
188
356
334
338
324
352
296
224
204
450
758
450
422
204
3891–98-3
544–76-3
1012–72-2
1.35
2.48
3.045
5.92
1014–60-4
2.26
55,124–80-6
900,406–32-2
54,964–82-8
630–04-6
54,833–54-6
7098–21-7
334–68-9
4.08
3.66
6.85
7.68
7.90
1.97
4.57
5.31
3.73
5.17
7.55
4.31
9.40
3.87
5.38
3.20
1.89
5.46
8.67
12.34
12.34
7.47
9.89
900,383–16-0
3.11
20,028,280–27-4
10.30
10.62
55,191–61-2
13.39
12.76
10.47
18.60
55,124–80-6
22.43
16.90
14.28
51,794–16-2
13.39
16.01
629–94-7
4.28
7.04
12.89
2345–24-6
64.24
489–28-1
544,854
5856–66-6
544,854
638–68-6
6813–21-4
100
9.31
7.795
12.29
14.37
5.92
5.84
10.11
7.04
18.81
100
100
Total identification
100
99.99
Table 6. Phytocomponents identified in ethyl acetate extract of S. costus callus grown on simple callus
promoting media (Kinetin: 0.5mgL-1 and 2.4-D: 1.0 mgL-1) and callus subjected to various stresses i.e. 60gL-1
D-Sorbitol stress (CPM-1), 60gL-1 D-Manitol (CPM-2), 5gL-1 Poly ethylene glycol 600 (CPM-3) and 60gL-1
Sucrose (CPM-4).
grown on CPM (with no stress see Fig. 5B). A similar trend in callus cultures of A. cordata anthocyanin accu-
mulations was achieved with a combination of either NAA or 2,4-D and Kinetin in comparison with auxins or
cytokinins66. Similarly, maximum total anthocyanin contents (3.3 to 7.4 CV/g) was reported from cultures on
moderate level (40 and 50 mM) of total nitrogen67. Our results are in agreements with several authors where,
highest values of all estimated anthocyanin were recorded in shootlets of A. leptopus obtained from MS supple-
mented with 2iP (0.4 mg/l) and IBA (0.1 mgL−1)68. Similarly, calli derived from style of Crocus sativus showed
anthocyanin pigment of 3.75 × 10–7 mg g −1 on media supplemented with NAA (2 mgL−1) and TDZ (1 mgL−1)
compared to that of calli produced from corm 2.52 × 10–7 mg g−1 DW69.
GC–MS analysis of callus ethyl acetate extract. GC–MS chromatograms of ethyl acetate extract of callus and
callus subjected to different stresses revealed the presence of 24 compounds. The active principles with their
retention time, molecular weight and peak area (%) of the identified compounds that could contribute the
medicinal quality of the plant are summarized in Table 6. The major components in the CPM extract were Pro-
panic acid, 2-methyl-,3,7-dimethyl-2,6-octadienyl ester and Selina-3,7(11)-diene. The analysis of GC–MS chro-
matogram showed peaks of various phytochemical constituents present in ethyl acetate CPM extracts (Fig. 6).
In contrast, major components identified in CPM-1 were Nonadecane, 2,6,10,14,18-pentamethyl, Nonade-
cane,2,6,10,14–tetramethyl and 6-tetradecane sulfonic acid, butyl ester (see Table 6, Fig. 7). In CPM-2, major
phytocomponents were Octacosane,1-Iodo, Octadecane-2,6,10,14-tetramethyl, Nonadecane, 2,6,10,14,18-pen-
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
9
Vol.:(0123456789)www.nature.com/scientificreports/ 100
%
0
24.65
2229711
3.06
137379
4.86
26259
8.86
38781
10.13
46888
15.24
141548
16.64
127153
13.04
85977
22.19
470509
19.13
108066
28.70
652992
5.00
10.00
15.00
20.00
25.00
30.00
35.00
Time
Figure 6. GC–MS Chromatogram of ethyl acetate extracts of callus grown on simple callus promoting media
(CPM) supplemented with Kinetin (0.5 mgL-1) and 2.4-D (1.0 mgL-1).
100
3.06
50826
4.45
135909
8.87
66709
10.12
68479
%
0
12.94
133241
15.01
154059
16.27
172706
20.70
231682
21.90
301137
16.53
50718
20.56
26204
20.00
25.24
224616
23.67
111384
27.73
295008
33.36
91965
25.00
30.00
35.00
Time
5.00
10.00
15.00
Figure 7. GC–MS Chromatogram of ethyl acetate extracts of callus subjected to 60 gL-1 D-Sorbitol stress
(CPM-1).
tamethyl and Nonadecane ,2,6,10,14–tetramethyl. Similarly, in CPM-3 and CPM-4 major phytocomponents
recorded were Nonadecane,2,6,10,14–tetramethyl, Eicosane,2,6,10,14,18-pentamethyl, Tetrapentacotane and
Nonadecane,2,6,10,14,18-pentamethyl, Nonadecane, 2,6,10,14–tetramethyl and Heneisane respectively (see
supplementary Figs. 8–10 and Table 6). While, the phytocomponents such as Octadecane-2,6,10,14-tetramethyl
and Hentriacontane were present in all the tasted samples. High amount of Octadecane-2,6,10,14-tetramethyl
was observed in CPM-2, while the amount of Hentriacontane was higher in CPM-1. Previously, Gwari et al.
(2013)3 has reported 41 aromatic compounds from essential oil of S. costus roots extracts. Among the identified
compounds Aldehyde like (7Z, 10Z, 13Z)-7, 10, 13-hexadecaterinal, ketones like dehydrocostus lactone, alcohols
like elemol, g-costol, vulgarol B, valerenol, and terpinen-4-ol, etc. were found a major component. In addition,
Esters and acids were found to be completely absent in root extracts of S. costus. Srinivasan et al. (2016)70 stud-
ied the chemical compounds in costus oil and observed that n-hexadecanoic acid to be the major constituent
in all examined essential oil accompanied with other fatty acids, hydrocarbons and mono-,di- sesquiterpenes.
Recently, Deabes et al. (2021)71 identified 14 components from S. costus ethyl acetate extracts. Compounds like
Butanedioic acid and 2TMS derivative were recorded in highest percentage followed by D-(-)-Fructofuranose,
pentakis (trimethylsilyl) ether (isomer1), Androstan-17-one, 3-ethyl-3-hydroxy-, (5.alpha.)-, Caffeic acid, 3TMS
derivative and L-(-)- Sorbofuranose and pentakis (trimethylsilyl) ether. This great variation in phytocomponents
of S. costus may be attributed to factors related to ecotype, chemotype, phenophases and the variations in envi-
ronmental conditions such as temperature, relative humidity, irradiance and photoperiod. Moreover, the genetic
background may also affect the chemistry of secondary metabolites of plants72. Furthermore, exposure to vari-
ous type stresses may result in drastic epigenetic modifications thereby, changing the transcriptional activities
and the overall transcriptomic profile73. Recently it has been shown that stable phenotypes can be generated
through epigenetic modifications and thereby increasing the success and survival of plants in their natural habi-
tats. Although, we have not studied any such epigenetic modifications here, but these are very likely targets and
are important consideration to be included in future studies.
Conclusion
Efficient protocols for large scale callus induction of four explants (seeds, leaf, petiole and internodes) as well as
micro propagation from auxiliary buds of S. costus were developed. Callus formation was greatly influenced by
type of explant used and maximum callus tissue with minimum time taken was record for seed explants. The best
response to direct organogenesis was observed on media fortified with BAP (2.0 mgL−1), NAA (1.0 mgL−1) and
GA3 (0.25 mgL−1). Micropropagated plantlets suffer high mortality due to their slow acclimatization to ex-vitro
conditions. In spite of the prior limited success with Asteraceae members in inducing roots during tissue culture
and acclimatization; here, the regenerated plantlets had 87% of survival rate. We argue this survival rate could
be further improved through biotization of micro propagated plants with endophytic bacteria and fungi. Here,
phytochemical characterization and variability in metabolites such as total sugars, proline, flavonoids, ascorbic
acid, phenolics and anthocyanin is recorded from callus, wild as well as micro propagated plantlets. It is also
demonstrated that S. costus callus is rich source of various bioactive compounds as indicated in the GC–MS
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
10
Vol:.(1234567890)www.nature.com/scientificreports/ profiles. The remarkable variation in the secondary metabolites may be partly explained by the preexisting
genetic variation within the populations of S. costus, understanding the role of epigenetic regulation in response
to environmental stimuli, particularly in response to stresses is of paramount significance to the stability and
survival of the plants in their natural habitat. The current work on this critically endangered species provides
a baseline for future work including application of the newly evolved biotechnological tools that may speed up
and ensure sustainability of the plant species, thereby enhancing conservation and management of S. costus in
natural ecosystems.
Approval and compliance with regulation
The study was formally authorized by the Directorate of Academics and Research Hazara University Mansehra,
Pakistan. Experimental research and field studies on selected plant, including the collection of plant material
comply with relevant institutional guidelines and legislation.
Statement for submission of specimen to University herbarium
Specimen was collected from Makra, mountain peak (alt 3,878 m), situated in District Mansehra, KP Pakistan.
Date 10–07-2019, GPS (lat 34.57439º N, long 073.49580º E), collected by Ajmal Khan, Azhar Hussain Shah, and
Abdul Majid, 57 (HUP).
Data availability
All data are available in the manuscript.
Received: 24 May 2021; Accepted: 17 November 2021
References
1. Vines, G. Herbal harvests with a future: towards sustainable sources for medicinal plants. (Plantlife International, 2004).
2. Zahara, K. et al. A review of therapeutic potential of Saussurea lappa-An endangered plant from Himalaya. Asian Pac. J. Trop. Med.
7, S60–S69 (2014).
3. Gwari, G., Bhandari, U., Andola, H. C., Lohani, H. & Chauhan, N. Volatile constituents of Saussurea costus roots cultivated in
Uttarakhand Himalayas India. Pharmacognosy Res. 5, 179 (2013).
4. Bisht, A. S. & Chauhan, R. S. Phenological Studies of High Value Endangered Medicinal Herbs: Picrorhiza kurroa and Saussurea
costus in Sub-alpine Regions of Garhwal Hiamalya, Uttarakhand India.
5. Cho, J. Y. et al. Cytotoxic and pro-apoptotic activities of cynaropicrin, a sesquiterpene lactone, on the viability of leukocyte cancer
cell lines. Eur. J. Pharmacol. 492, 85–94 (2004).
6. Peng, Z. et al. Costunolide and dehydrocostuslactone combination treatment inhibit breast cancer by inducing cell cycle arrest
and apoptosis through c-Myc/p53 and AKT/14-3-3 pathway. Sci. Rep. 7, 1–16 (2017).
7. Davies, K. M. & Deroles, S. C. Prospects for the use of plant cell cultures in food biotechnology. Curr. Opin. Biotechnol. 26, 133–140
(2014).
8. Yasmin, S. et al. Plant growth promotion and suppression of bacterial leaf blight in rice by inoculated bacteria. PloS one 11,
e0160688 (2016).
9. Murashige, T. & Skoog, F. A revised medium for rapid growth and bio assays with tobacco tissue cultures. Physiol. Plant. 15,
473–497 (1962).
10. Hossain, A., Mishty, A. & Mian, A. Numerical analysis for design optimization of microcantilever beams for measuring rheological
properties of viscous fluid. Finite Elem. Anal. Des. 68, 1–9 (2013).
11. Storey, R. & Jones, R. W. Betaine and choline levels in plants and their relationship to NaCl stress. Plant Sci. Lett. 4, 161–168 (1975).
12. Dubois, M., Gilles, K. A., Hamilton, J. K., Rebers, P. T. & Smith, F. Colorimetric method for determination of sugars and related
substances. Anal. Chem. 28, 350–356 (1956).
13. Bates, L. S., Waldren, R. P. & Teare, I. Rapid determination of free proline for water-stress studies. Plant Soil 39, 205–207 (1973).
14. Csepregi, K., Kocsis, M. & Hideg, É. On the spectrophotometric determination of total phenolic and flavonoid contents. Acta Biol.
Hung. 64, 500–509 (2013).
15. Re, R. et al. Antioxidant activity applying an improved ABTS radical cation decolorization assay. Free Radical Biol. Med. 26,
1231–1237 (1999).
16. Singleton, V. L. & Rossi, J. A. Colorimetry of total phenolics with phosphomolybdic-phosphotungstic acid reagents. Am. J. Enol.
Vitic. 16, 144–158 (1965).
17. Giusti, M. M. & Wrolstad, R. E. Characterization and measurement of anthocyanins by UV‐visible spectroscopy. Curr. Protocols
Food Anal. Chem., F1. 2.1-F1. 2.13 (2001).
18. Deventhiran, M. et al. Comparative phytochemical analysis of wild and micropropagated Cleome viscosa L. J. Appl. Pharm. Sci.
7, 83–88 (2017).
19. Rosy, B. A. & Rosakutty, P. GC-MS analysis of methanol wild plant and callus extracts from three Cissus species, family Vitaceae.
J. Chem. Pharm. Res. 4, 3420–3426 (2012).
20. Hassan, M. M., Azam, F. S., Chowdhury, M. H. & Rahmatullah, M. Callus induction of Abrus precatorius: screening of phytohor-
mones. Am. Eur. J. Sustain. Agric., 512–519 (2009).
21. Sen, M. K., Nasrin, S., Rahman, S. & Jamal, A. H. M. In vitro callus induction and plantlet regeneration of Achyranthes aspera L,
a high value medicinal plant. Asian Pac. J. Trop. Biomed. 4, 40–46 (2014).
22. Ardelean, M. et al. Cytological aspects and anthocyanin accumulation observed in Sedum telephium ssp. maximum L. callus.
Roman. Biotechnol. Lett. 22, 13025 (2017).
23. Khoo, I.-C. Liquid crystals. Vol. 64 (John Wiley & Sons, 2007).
24. Ali, S. et al. In vitro effects of GA 3 on morphogenesis of CIP potato explants and acclimatization of plantlets in field. In Vitro Cell.
Dev. Biol. Plant 54, 104–111 (2018).
25. Sabeti, M., Zarghami, R. & Zadeh, M. Effects of explants and growth regulators on callogenesis and somatic embryogenesis of
Agria potato cultivar. Int. J. Agrisci. 3, 213–221 (2013).
26. Iralu, V. & Upadhaya, K. Dormancy, storability, and germination of seeds in Magnolia punduana (Magnoliaceae). Botany 94,
967–973 (2016).
27. Wojtania, A., Skrzypek, E. & Gabryszewska, E. Morphological and biochemical responses to gibberellic acid in
MagnoliaבSpectrum’in vitro. Acta Biologica Cracoviensia. Series Botanica 58 (2016).
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
11
Vol.:(0123456789)www.nature.com/scientificreports/ 28. Das, S. et al. Core–shell structured catalysts for thermocatalytic, photocatalytic, and electrocatalytic conversion of CO 2. Chem.
Soc. Rev. 49, 2937–3004 (2020).
29. Martin, G. et al. An efficient micropropagation system for Celastrus paniculatus Willd.: a vulnerable medicinal plant. J. For. Res.
11, 461–465 (2006).
30. Rani, G., Talwar, D., Nagpal, A. & Virk, G. Micropropagation of Coleus blumei from nodal segments and shoot tips. Biol. Plant.
50, 496–500 (2006).
31. Kaur, K., Verma, B. & Kant, U. Plants obtained from the Khair tree (Acacia catechu Willd.) using mature nodal segments. Plant
Cell Rep. 17, 427–429 (1998).
32. Sivaram, L. & Mukundan, U. In vitro culture studies on Stevia rebaudiana. In vitro Cell. Dev. Biol. Plant 39, 520–523 (2003).
33. Rafiq, M., Dahot, M. U., Mangrio, S. M., Naqvi, H. A. & Qarshi, I. A. In vitro clonal propagation and biochemical analysis of field
established Stevia rebaudiana Bertoni. Pak. J. Bot 39, 2467–2474 (2007).
34. Dhar, U., Upreti, J. & Bhatt, I. D. Micropropagation of Pittosporum napaulensis (DC.) Rehder & Wilson–a rare, endemic Himalayan
medicinal tree. Plant Cell, Tissue Organ Culture 63, 231–235 (2000).
35. Guo, Y.-X., Zhao, Y.-Y., Zhang, M. & Zhang, L.-Y. Development of a novel in vitro rooting culture system for the micropropagation
of highbush blueberry (Vaccinium corymbosum) seedlings. Plant Cell, Tissue and Organ Culture (PCTOC) 139, 615–620 (2019).
36. Cheepala, S., Sharma, N. & Sahi, S. Rapid in vitro regeneration of Sesbania drummondii. Biol. Plant. 48, 13–18 (2004).
37. Kumar, S., Kumaria, S. & Tandon, P. Efficient in vitro plant regeneration protocol from leaf explant of Jatropha curcas L—a promis-
ing biofuel plant. J. Plant Biochem. Biotechnol. 19, 273–275 (2010).
38. Wen, W. & Yang, J. Study on the tissue culture and propagation system of Gentiana scabra Bunge. Med. Plant 1, 13–23 (2010).
39. Bekheet, S. Direct organogenesis of date palm (Phoenix dactylifera L.) for propagation of true-to-type plants. Sci. Agric. 4, 85–92
(2013).
40. Chandra, S., Bandopadhyay, R., Kumar, V. & Chandra, R. Acclimatization of tissue cultured plantlets: from laboratory to land.
Biotech. Lett. 32, 1199–1205 (2010).
41. Deb, C. & Imchen, T. An efficient in vitro hardening technique of tissue culture raised plants. Biotechnology 9, 79–83 (2010).
42. del Rosario Espinoza-Mellado, M. et al. Biotization and in vitro plant cell cultures: plant endophyte strategy in response to heavy
metals knowledge in assisted phytoremediation. Microbe Med. Remed. Environ. Contamin., 27–36 (2021).
43. Prado, R., Huerta, G. & West, M. Bayesian time-varying autoregressions: Theory, methods and applications. Resenhas do Instituto
de Matemática e Estatística da Universidade de São Paulo 4, 405–422 (2000).
44. Wang, Z. & Stutte, G. W. The role of carbohydrates in active osmotic adjustment in apple under water stress. J. Am. Soc. Hortic.
Sci. 117, 816–823 (1992).
45. Zhang, F., Yang, Y., He, W., Zhao, X. & Zhang, L. Effects of salinity on growth and compatible solutes of callus induced from Populus
euphratica. In Vitro Cell. Dev. Biol. Plant 40, 491–494 (2004).
46. Martin, K. Rapid in vitro multiplication and ex vitro rooting of Rotula aquatica Lour., a rare rhoeophytic woody medicinal plant.
Plant Cell Rep. 21, 415–420 (2003).
47. Malik, A. A., Williams, C. A., Weston, K. L. & Barker, A. R. Perceptual and prefrontal cortex haemodynamic responses to high-
intensity interval exercise with decreasing and increasing work-intensity in adolescents. Int. J. Psychophysiol. 133, 140–148 (2018).
48. Mehta, R., Sharma, S. & Nath, A. K. In vitro selection and biochemical characterization of carnation (Diathus caryophyllus L.)
callus culture tolerant to Alternaria dianthi. Indian J. Plant Physiol. 12, 120 (2007).
49. Felix, K. et al. Hydrogen-induced tolerance against osmotic stress in alfalfa seedlings involves ABA signaling. Plant Soil 445,
409–423 (2019).
50. Shah, A. H., Shah, S. H., Ahmad, H. & Swati, Z. A. Adaptation to polyethylene stress maintains totipotency of cell lines of Oryza
sativa L. CV Swat-1 for a longer period. Pak. J. Bot 44, 313–316 (2012).
51. Pradhan, S. et al. Poly ethylene glycol mediated in vitro screening and physico-biochemical changes induced in mango callus due
to moisture stress. Plant Cell, Tissue Organ Culture PCTOC 145, 155–172 (2021).
52. Hassan, A. & Roy, S. Micropropagation of Smilax zeylanica L., a perennial climbing medicinal shrub, through axillary shoot
proliferation. Bangladesh J. Life Sci 16, 33–39 (2004).
53. Ahmad, M. S. A., Javed, F. & Ashraf, M. Iso-osmotic effect of NaCl and PEG on growth, cations and free proline accumulation in
callus tissue of two indica rice (Oryza sativa L.) genotypes. Plant Growth Reg. 53, 53–63 (2007).
54. Jain, M., Tiwary, S. & Gadre, R. Sorbitol-induced changes in various growth and biochemici parameters in maize. Plant Soil Environ.
56, 263–267 (2010).
55. Razavizadeh, R. & Adabavazeh, F. Effects of sorbitol on essential oil of Carum copticum L. under in vitro culture. Rom. Biotechnol.
Lett. 22, 12281–12289 (2017).
56. Ibrahim, S. R. et al. Lipoxygenase inhibitors flavonoids from Cyperus rotundus aerial parts. Rev. Bras 28, 320–324 (2018).
57. Khan, T. et al. Melatonin-enhanced biosynthesis of antimicrobial AgNPs by improving the phytochemical reducing potential of
a callus culture of Ocimum basilicum L var thyrsiflora. RSC Adv. 7, 38699–38713 (2017).
58. Kamal, O. M., Shah, S. H. A., Li, Y., Hou, X. & Li, Y. Production of ascorbic acid, total protein, callus and root in vitro of non-
heading Chinese cabbage by tissue culture. Mol. Biol. Rep. 47, 6887–6897 (2020).
59. Narayan, M., Thimmaraju, R. & Bhagyalakshmi, N. Interplay of growth regulators during solid-state and liquid-state batch cultiva-
tion of anthocyanin producing cell line of Daucus carota. Process Biochem. 40, 351–358 (2005).
60. Pietta, P.-G. Flavonoids as antioxidants. J. Nat. Prod. 63, 1035–1042 (2000).
61. El-Nabarawy, M., El-Kafafi, S., Hamza, M. & Omar, M. The effect of some factors on stimulating the growth and production of
active substances in Zingiber officinale callus cultures. Ann. Agric. Sci. 60, 1–9 (2015).
62. Gorni, P. H. & Pacheco, A. C. Growth promotion and elicitor activity of salicylic acid in Achillea millefolium L. Afr. J. Biotech. 15,
657–665 (2016).
63. Dias, M. I., Sousa, M. J., Alves, R. C. & Ferreira, I. C. Exploring plant tissue culture to improve the production of phenolic com-
pounds: a review. Ind. Crops Prod. 82, 9–22 (2016).
64. Surveswaran, S., Cai, Y.-Z., Xing, J., Corke, H. & Sun, M. Antioxidant properties and principal phenolic phytochemicals of Indian
medicinal plants from Asclepiadoideae and Periplocoideae. Nat. Prod. Res. 24, 206–221 (2010).
65. Sakakibara, H., Takei, K. & Hirose, N. Interactions between nitrogen and cytokinin in the regulation of metabolism and develop-
ment. Trends Plant Sci. 11, 440–448 (2006).
66. Murthy, H. N., Lee, E.-J. & Paek, K.-Y. Production of secondary metabolites from cell and organ cultures: strategies and approaches
for biomass improvement and metabolite accumulation. Plant Cell, Tissue Org. Culture PCTOC 118, 1–16 (2014).
67. Irshad, M. et al. Accumulation of anthocyanin in callus cultures of red-pod okra [Abelmoschus esculentus (L.) Hongjiao] in
response to light and nitrogen levels. Plant Cell, Tissue Organ Cul PCTOC 134, 29–39 (2018).
68. Youssef, N. M., Taha, L. S., El-Khalek, A. & Nagy, S. Secondary metabolites characterization of in vitro propagated Antigonon
leptopus cultures. Egypt. J. Chem. 64, 923–932 (2021).
69. Moradi, A., Zarinkamar, F., Caretto, S. & Azadi, P. Influence of thidiazuron on callus induction and crocin production in corm
and style explants of Crocus sativus L. Acta Physiol. Plant. 40, 1–8 (2018).
70. Srinivasan, J., Liu, C. & Kelly, R. Geometric evolution of flux from a corroding one-dimensional pit and its implications on the
evaluation of kinetic parameters for pit stability. J. Electrochem. Soc. 163, C694 (2016).
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
12
Vol:.(1234567890)www.nature.com/scientificreports/ 71. Deabes, M., Aboulthana, W., Marzouk, E.E.-D., Mohamed, M. I. & Ahmed, K. A. Evaluation of hepato-and neuroprotective effect
of chemical constituents in Saussurea costus extract against the toxicity induced by chloropyrofos ethyl in rats. Egypt. J. Chem. 64,
631–647 (2021).
72. Negi, J. S. et al. Antidiarrheal activity of methanol extract and major essential oil contents of Saussurea lappa Clarke. Afr. J. Pharm.
Pharmacol 7, 474–477 (2013).
73. Thiebaut, F., Hemerly, A. S. & Ferreira, P. C. G. A role for epigenetic regulation in the adaptation and stress responses of non-model
plants. Front. Plant Sci. 10, 246 (2019).
Author contributions
A.K. collected the data carried out lab work and drafted the manuscript. A.H.S. and N.A. conceived the overall
project, reviewed, edited and finalized the manuscript.
Funding
The authors deeply acknowledge Higher Education Commission (HEC), Islamabad, Pakistan for providing
financial assistance to this project (NRPU Project No. 20-4253).
Competing interests
The authors declare to have no conflict of interest.
Additional information
Supplementary Information The online version contains supplementary material available at https:// doi. org/
10. 1038/ s41598- 021- 03032-1.
Correspondence and requests for materials should be addressed to A.H.S.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
© The Author(s) 2021
Scientific Reports | (2021) 11:23575 |
https://doi.org/10.1038/s41598-021-03032-1
13
Vol.:(0123456789)www.nature.com/scientificreports/
| null |
10.3201_eid2507.181794.pdf
| null | null |
RESEARCH
Asymptomatic Dengue Virus
Infections, Cambodia, 2012–2013
Sowath Ly,1 Camille Fortas,1 Veasna Duong, Tarik Benmarhnia, Anavaj Sakuntabhai, Richard Paul,
Rekol Huy, Sopheak Sorn, Kunthy Nguon, Siam Chan, Souv Kimsan, Sivuth Ong, Kim Srorn Kim,
Sowathy Buoy, Lim Voeung, Philippe Dussart, Philippe Buchy,1,2 Arnaud Tarantola1
We investigated dengue virus (DENV) and asymptomatic
DENV infections in rural villages of Kampong Cham Prov-
ince, Cambodia, during 2012 and 2013. We conducted
perifocal investigations in and around households for 149
DENV index cases identified through hospital and village
surveillance. We tested participants 0.5–30 years of age by
using nonstructural 1 rapid tests and confirmed DENV infec-
tions using quantitative reverse transcription PCR or non-
structural 1–capture ELISA. We used multivariable Poisson
regressions to explore links between participants’ DENV in-
fection status and household characteristics. Of 7,960 study
participants, 346 (4.4%) were infected with DENV, among
whom 302 (87.3%) were <15 years of age and 225 (65.0%)
were <9 years of age. We identified 26 (7.5%) participants
with strictly asymptomatic DENV infection at diagnosis and
during follow-up. We linked symptomatic DENV infection
status to familial relationships with index cases. During the
2-year study, we saw fewer asymptomatic DENV infections
than expected based on the literature.
testing of dengue-like cases in referral pediatric hospitals
in Cambodia likely underestimate the true disease burden
(4). By definition, syndromic surveillance does not detect
asymptomatic DENV infections, which increase vector
transmission potential (5). Mammen et al. used both den-
gue-positive and dengue-negative index cases of febrile
children to initiate perifocal investigations and found no
cases in proximity to dengue-negative index cases (6). To
maximize the number of recruited cases, we investigated
homes around preidentified, dengue-positive index cases,
as per a previous study (7). Our objectives were to docu-
ment the proportion of strictly asymptomatic infections
in this region of Cambodia; characterize human, sociode-
mographic, household-level, and mosquito control–related
factors associated with DENV infection; and identify fac-
tors associated with asymptomatic DENV infection.
Methods
Annually, ≈390 million people in >100 countries are
infected with dengue virus (DENV); 70% of cases
occur in countries in Asia (1). DENV is a flavivirus trans-
mitted by Aedes aegypti and Ae. albopictus anthropo-
philic female mosquitoes. DENV has 4 distinct serotypes,
DENV-1–4 (2); DENV infections can range from asymp-
tomatic to life-threatening.
In Cambodia, the national dengue surveillance system
reported 60,000 cases and 135 deaths attributed to DENV
in 2012 and 2013 (3). Syndromic surveillance and random
Author affiliations: Institut Pasteur du Cambodge, Phnom Penh,
Cambodia (S. Ly, C. Fortas, V. Duong, S. Sorn, K. Nguon,
S. Chan, S. Kimsan, S. Ong, P. Dussart, P. Buchy, A. Tarantola);
University of California, San Diego, California, USA
(T. Benmarhnia); Institut Pasteur, Paris, France (A. Sakuntabhai,
R. Paul); Malaria National Center, Phnom Penh (R. Huy);
Kampong Cham Provincial Hospital, Kampong Cham, Cambodia
(K.S. Kim); Prey Chhor District Referral Hospital, Kampong Cham
(S. Buoy); Tboung Khmum District Referral Hospital, Thoung
Khmum, Cambodia (L. Voeung)
DOI: https://doi.org/10.3201/eid2507.181794
Ethics Considerations
The study protocol was approved by the Cambodian
National Ethics Committee on Health Research. We ob-
tained informed consent from participants or their guard-
ians documented during hospital or village surveillance or
perifocal investigations.
Study Site
We conducted a study in rural villages of Kampong Cham
Province, 120 km northeast of Cambodia’s capital, Phnom
Penh. The study area included 368 villages with ≈60,000
households and 3 hospitals within a 30-km radius. Dengue
is endemic in the region and mainly affects children <15
years of age during the annual rainy season (June–October).
Identification of Dengue Index Cases in
Hospitals and Villages
During June 1–October 31, 2012 and 2013, we identified
DENV index cases in 3 referral hospitals and 26 villages
1These authors contributed equally to this article.
2Current affiliation: GlaxoSmithKline Vaccines Research and
Design, Singapore.
1354
Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 25, No. 7, July 2019
under active surveillance for febrile illness (5). We targeted
persons 0.5–30 years of age. In the 3 hospitals, blood sam-
ples were drawn at admission and discharge for all patients
suspected of having DENV infection on the basis of clinical
assessment and platelet count. In the 26 villages, volunteers
monitored eligible residents weekly, measuring axillary body
temperature using a digital thermometer to identify persons
with temperatures >38°C. Blood samples were drawn 1–2
days after fever onset, as described elsewhere (4,8,9). All sam-
ples were screened for DENV infection by using a nonstruc-
tural (NS) 1 IgM/IgG combination rapid test. We confirmed
DENV by using quantitative reverse transcription PCR (qRT-
PCR) or NS1-capture ELISA and included case-patients with
confirmed DENV infection as index cases in the study.
Perifocal Investigations
Within 1–2 days of identifying an index case, whether from
village or hospital surveillance, we began a perifocal inves-
tigation of the index case-patient’s village of origin (7). For
each perifocal investigation, we used a rapid dengue test
kit to screen eligible residents in the index case-patient’s
household for DENV and completed a baseline question-
naire on individual symptoms, socioeconomic status, and
household characteristics. We did the same in 20 house-
holds in a 100-meter radius of the index case-patient’s
household. We included persons 0.5–30 years of age who
consented or whose guarantor consented. We tested adults
>20 years of age during the first year of the study but found
no DENV-positive cases and did not test this age group
during the second year. All consecutive cases were eligible
for inclusion. To avoid bias through overlapping investi-
gations of a potentially common source of infection, we
did not conduct a perifocal investigation within 1 week of
a previous investigation for >2 index cases consecutively
detected from the same village.
DENV Testing and Case Definitions
To screen for DENV infection during surveillance and
perifocal investigations, investigators tested all blood
samples on-site using SD BIOLINE Dengue Duo kit (Stan-
dard Diagnostics, https://www.alere.com), according to
the manufacturer's instructions. Investigators interpreted
results within 15–20 minutes and ruled out possible cas-
es if the control band was negative. Blood samples from
DENV-positive participants were sent to Institut Pasteur du
Cambodge (Phnom Penh, Cambodia) for qRT-PCR testing,
as described previously (10,11), or confirmation using an
NS1-capture ELISA (11,12) with positive controls diluted
to the limit of detection, negative, and nontemplate controls
used during extraction and PCR steps to reduce inaccura-
cies (10). We considered cases confirmed when a blood
sample tested positive by NS1 rapid test and was confirmed
by NS1-capture ELISA or qRT-PCR. During the first year,
Asymptomatic Dengue Virus Infections, Cambodia
we also tested participants for Japanese encephalitis virus
(JEV) and chikungunya virus (CHIKV) IgM antibodies
by ELISA and confirmed IgM-positive results using spe-
cific RT-PCR to ensure that symptoms were not related to
CHIKV, JEV, or co-infections (11–13).
Symptomatic DENV-confirmed case-patients had fe-
ver, muscle or joint pain, rash, bleeding, prolonged head-
aches, or digestive signs. We asked participants whether
they had taken antipyretics in the previous 24 hours. We
termed afebrile all symptomatic DENV-positive partici-
pants without a fever and no antipyretic use. We consid-
ered participants asymptomatic when they had confirmed
DENV infection, no antipyretic use, and no signs or symp-
toms, including fever. Participants who were symptomatic
at initial diagnosis on day 0 received follow-up monitoring
on days 2 and 7. We monitored asymptomatic participants
daily on days 0–7 and again on day 10 using a question-
naire to document signs and symptoms of DENV. In our
analyses, we recategorized participants who were asymp-
tomatic at baseline to symptomatic if they reported any
symptoms during the follow-up period.
Statistical Analysis
We described DENV infection attack rates for perifocal
investigations and the proportion of asymptomatic cases
among all DENV infections and circulating serotypes. To
explore participant- and household-level factors associated
with DENV infection, we conducted a multivariable Pois-
son regression estimating attack rate ratios (ARRs) (14),
excluding index cases. We built explanatory models around
each participant-level and household-level factor, with and
without adjusting for covariates. Participant-level factors
included age, sex, occupation or schooling, and relation-
ship to an index case-patient. Because we found collinear-
ity between age and occupation, we adjusted only for age.
We placed participants 0.5–1 year of age into a specific cat-
egory to account for differences in immunity and exposure
to vectors due to reduced mobility. Household-level fac-
tors included the main source of income, source of water,
measures against mosquitoes, and environmental factors
favorable to mosquito development. We further divided the
source of water into 2 categories: piped water (from indoor
or outdoor taps with a tube well and pump) or nonpiped
water (from a pond, river, lake, or a well without pump).
Considering the limited flight range of a female Aedes
mosquito, we assumed that the probability of DENV trans-
mission would be higher within a household than across
households. To account for this factor and measure poten-
tial clustering, we developed a random-effects multilevel
model. We computed the intraclass correlation coefficient
as the proportion of the variability in the probability of
infection attributable to differences between households
versus differences within households (15). We excluded
Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 25, No. 7, July 2019
1355
RESEARCH
19 participants with missing covariates or predictors from
the regression analyses. We explored associations between
asymptomatic DENV infection and DENV serotype, par-
ticipant-level factors, and the main source of income as so-
cioeconomic indicators. We used the Fisher exact test for
comparing proportions, the Student t test for means, and
an empty multilevel model to search for a cluster effect.
We conducted analyses using Stata version 13 (StataCorp,
https://www.stata.com).
Results
Dengue Surveillance for Index Case Identification
We identified 1,294 suspected DENV-infected persons,
834 (64.5%) among hospital inpatients and 460 (35.5%)
through febrile illness surveillance in villages. Our testing
confirmed 555 (66.5%) DENV-positive cases among hos-
pital patients and 36 (7.8%) DENV-positive cases through
febrile illness surveillance in villages.
Perifocal Investigations
From the 591 DENV-positive patients, we selected 149
(25.2%) consecutive cases for which we conducted perifocal
investigations: 131 from hospital patients, termed PI-H, and
18 from village febrile surveillance, termed PI-V. Perifocal
investigations took place in 104 villages over the 2 rainy sea-
sons and documented 7,960 participants, 6,811 (86%) male
and 1,149 (14%) female, in 2,988 households (Figure).
We found 346 (4.3%) persons who were positive for
DENV infection, 225 (65.0%) of whom were <9 years
of age. We determined attack rates of 14.7/1,000 partici-
pants (14/952) in PI-V and 47.4/1,000 (332/7,008) in PI-H
(p<0.05). The attack rate over the 2 outbreak seasons in-
creased marginally from 37/1,000 persons 0.5–30 years of
age during the 2012 season to 46/1,000 among those 0.5–
20 years of age during 2013 (p = 0.056). Only 26 (7.5%)
of 346 DENV-positive participants remained strictly
asymptomatic during the 10-day follow-up, an asymptom-
atic DENV-infection attack rate of 3.3/1,000 over the 2
years of our study. The proportion of asymptomatic infec-
tions was 21.4% (3/14) in PI-V and 6.9% (23/332) in PI-H.
Besides headache and fever, symptomatic case-
patients mainly experienced muscle, retro-orbital, and
joint pain. Although fever is considered a typical symptom
of DENV infection, careful interview, rigorous clinical
assessment, and follow-up interviews showed that partici-
pants remained afebrile in 110 (31.8%) of the 320 symp-
tomatic DENV infections, even without antipyretics. Only
6 (1.7%) of the DENV-positive case-patients required hos-
pitalization, 2 with bleeding.
The 2 annual outbreaks were dominated by DENV-1.
However, DENV-2 and DENV-4 emerged in 2013, and
we detected DENV-3 sporadically (Table 1). During the
first year of the study, samples from all symptomatic and
asymptomatic DENV cases were negative for CHIKV by
MAC-ELISA. Because we diagnosed no CHIKV in year
1, and our national surveillance system also did not detect
any CHIKV cases (data not shown), we did not perform
CHIKV testing during year 2. Of 26 asymptomatic cases
confirmed by qRT-PCR or NS1-capture ELISA, 6 had
positive JEV serology and also were positive for DENV
IgM. We could not conclude whether JEV-positive results
were indicative of a recent or acute JEV co-infection or the
result of cross-reaction among flaviviruses. Among hospi-
talized patients, 2 had positive JEV results without detect-
able DENV IgM, even though qRT-PCR or NS1-capture
ELISA was positive. These results could suggest a recent
or acute JEV co-infection. During perifocal investigations,
42 participants tested positive for JEV by MAC-ELISA
Figure. Participant screening and data flowchart for perifocal
investigations for asymptomatic DENV infection, Cambodia,
2012–2013. Initial DENV screening of febrile cases was
conducted using nonstructural (NS) 1 IgM/IgG combo rapid test.
Perifocal investigations took place in villages of index cases; we
screened all persons in 20 households within a 100-m radius of an
index case household. We excluded persons <0.5 and >30 years
of age. Laboratory confirmation of DENV was conducted through
quantitative reverse transcription PCR or NS1-capture ELISA.
DENV, dengue virus; PI-H, perifocal investigations conducted for
index cases identified through hospital surveillance; PI-V, perifocal
investigations conducted for index cases identified through village
febrile surveillance.
1356
Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 25, No. 7, July 2019
Asymptomatic Dengue Virus Infections, Cambodia
with negative DENV results, NS1, and qRT-PCR, support-
ing evidence of JEV co-circulation in the country (16).
Screened participants had a mean age (+ SD) of 11.7
(+ 7.9; median 10; interquartile range 6–16); 6,207 (77.9%)
were schoolchildren, university students, or nonschooled
children. The main sources of household income were
planting crops (61.0%), working in a factory (14.3%), and
keeping a shop (13.4%). Participants reported low use of
protective measures against mosquitoes, including mos-
quito coils in 787 (26.3%) households, insecticide sprays
in 557 (18.6%) households, and larvicidal temephos in
374 (12.5%) households. Our investigation found uncov-
ered water jars in 1,867 (62.7%) households and mosquito
larvae in water containers of 1,663 (55.7%) households
(Table 2).
Among DENV-positive cases, boys and girls were
equally affected at a mean (+ SD) age of 8.5 (± 5.7) years.
Compared with persons 15–30 years of age, we found that
children 1–10 years of age had a higher ARR of DENV
infection (ARR 4.04 [95% CI 2.72–5.98] for those 1–5
years of age and ARR 3.83 [95% CI 2.59–5.67] for those
6–10 years of age). Siblings and cousins of index case-
patients were more prone to DENV infection than neigh-
bors were; siblings were 2.24 (95% CI 1.42–3.53) times
and cousins 1.40 (95% CI 1.02–1.90) times more at risk for
infection than neighbors. Participants who used piped wa-
ter had a higher risk for DENV infection (ARR 1.35 [95%
CI 1.06–1.71]) than did those who used nonpiped water.
Households in which the main source of income was fish-
ing, farming, or animal husbandry also had higher risks for
infection (ARR 2.02 [95% CI 1.18–3.45]). Households re-
porting mosquito control–related parameters did not have a
lower risk for DENV infection (Table 2).
The main source of income was similarly distributed
between households with ≥1 case and households with no
cases (p = 0.272). Our multilevel model showed a notable
clustering effect at the household level after adjustment (in-
traclass correlation coefficient 40.8%).
We found 26 (7.5%) case-patients, 17 (65.4%) male
and 9 (34.6%) female, who were positive for DENV in-
fection but remained asymptomatic. We found serotypes
DENV-1, DENV-2, and DENV-4 in our study group (Ta-
ble 3). We used a multilevel approach to explore the role
of specific serotypes and participant-level factors, such as
age, gender, and relationship to the index case-patient, a
proxy for common genetic background, with being DENV-
positive and asymptomatic. We found that only family
relationship to the index case-patient was associated with
asymptomatic infection. We did not identify a cluster effect
or associated factors.
Discussion
We screened 7,960 participants in communities in Cambo-
dia during 2012 and 2013 and found 346 (4.3%) participants
Table 1. Surveillance data from perifocal investigations for asymptomatic dengue virus, Cambodia, 2012–2013*
Surveillance data
No. participants
No. villages investigated
No. perifocal investigations conducted
Mean no. participants per perifocal investigation
Confirmed infections, no. (%)
Strictly asymptomatic
Afebrile
Symptomatic
Attack rate/1,000 participants, %
Asymptomatic infections
Symptomatic infections
Afebrile infections
2012
2,391
35
47
51
88
5 (5.7)
33 (37.5)
83 (94.3)
36.8
2.1
34.7
13.8
83
55 (66.2)
52 (62.7)
16 (19.3)
17 (20.5)
17 (20.5)
15 (18.1)
13 (15.7)
3 (3.5)
88
2013
5,569
77
102
55
258
21 (8.1)
77 (29.8)
237 (91.9)
46.3
3.8
42.6
13.8
237
180 (75.9)
169 (71.3)
73 (30.8)
73 (30.8)
68 (28.7)
53 (22.4)
50 (21.1)
8 (3.3)
258
82 (98.8)
1 (1.2)
0
0
0
5
188 (72.9)
36 (13.9)
2 (0.8)
31 (12.0)
1 (0.4)
0
Total
7,960
104
149
53
346
26 (7.5)
110 (31.2)
320 (92.5)
43.5
3.3
40.2
13.8
320
236 (73.8)
221 (69.1)
89 (27.8)
90 (28.1)
85 (26.5)
68 (21.3)
63 (19.7)
11 (3.3)
346
270 (78.0)
37 (10.7)
2 (0.6)
31 (9.0)
1 (0.3)
5
Symptoms at diagnosis or follow-up, no. (%)
Fever
Headache
Muscle pain
Retro-orbital pain
Joint pain
Rash
Any bleeding
Hospitalizations, no. (%)
DENV infections
Serotype, no. (%)
DENV-1
DENV-2
DENV-3
DENV-4
DENV-1 and DENV-2
Missing
*Participants 0.5–30 years of age in 2012 and 0.5–20 years of age in 2013. DENV, dengue virus.
Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 25, No. 7, July 2019
1357
RESEARCH
infected by DENV; 26 (7.5%) remained asymptomatic be-
fore, during, and after our study. We found comparable at-
tack rates, 37/1,000 persons in 2012 and 46/1,000 persons in
2013, to other community investigations conducted in Cam-
bodia. For instance, another study reported DENV attack
rates of 13.4–57.8 cases/1,000 persons during 2006–2008
(4). Previous studies only included participants ≤20 years
of age, but we included persons 0.5–30 years of age with
confirmed DENV infection, even symptomatic but afebrile
case-patients, who were 31.8% of the DENV infections in
Table 2. Participant and household characteristics with unadjusted and adjusted attack rate ratios for factors potentially associated
with dengue virus infection, Cambodia, 2012–2013*
Characteristics
Participants
Sex
M
F
Age, y†
0.5–<1
1–<5
5–<10
10–<15
15–30
Mean (+ SD, median)
Occupation‡
Student, school or university
Preschool or unschooled
Planting crops
Other
Missing
Relationship to index case-patient§
Neighbor
Cousin
Sibling
Other
Missing
Households
Water source#
Nonpiped
Piped
Primary source of income**
Planting crops
Working in a factory
Shopkeeping
Fishing, farming, animal husbandry
Working in government
Other
Mosquito control measures††
Temephos
Larvivorous fish
Treated mosquito netting
Treated jar cover
Coils
Insecticide spray
Environmental factors**
Vegetable garden
Water collection around house
Uncovered water jars
Larvae in water containers
Distance from house to nearest
water jar, m (+ SD)
Missing for all items
Infected
346
Uninfected
7,614
171
175
4,103
3,511
Total
7,960
4,272
3,686
9 (2.6)
108 (31.2)
126 (36.4)
71 (20.5)
32 (9.3)
150 (2.0)
1,701 (22.3)
2,083 (27.4)
1,675 (22.0)
2,005 (26.3)
8.5 (+ 5.7, 7) 11.9 (+ 8.0, 10) 11.7 (+ 7.9, 10)
159 (2.0)
1,809 (22.7)
2,209 (27.8)
1,746 (21.9)
2,037 (25.6)
171 (49.8)
149 (43.2)
20 (5.8)
5 (1.5)
1 (0.2)
260 (75.4)
58 (16.8)
23 (6.7)
5 (1.2)
1 (0.2)
292¶
3,588 (47.2)
2,299 (30.2)
910 (12.0)
809 (10.6)
8 (0.1)
6,309 (83.0)
991 (13.0)
251 (3.3)
55 (0.7)
8 (0.1)
2,706
3,759 (47.3)
2,448 (30.8)
930 (11.7)
814 (10.2)
9 (1.1)
6,569 (82.6)
1,049 (13.2)
274 (3.5)
59 (0.7)
9 (1.1)
2,988
Unadjusted ARR
(95% CI)
Adjusted ARR
(95% CI)
Referent
Referent
1.14 (0.92–1.40) 1.01 (0.82–1.24)
3.47 (1.65–7.32) 3.53 (1.67–7.46)
3.98 (2.69–5.90) 4.04 (2.72–5.98)
3.79 (2.56–5.60) 3.83 (2.59–5.67)
2.59 (1.70–3.94) 2.55 (1.67–3.88)
Referent
–
Referent
–
2.14 (1.35–3.41) 2.14 (1.34–3.41)
2.84 (1.79–4.54) 2.84 (1.78–4.54)
Referent
Referent
0.28 (0.10–7.76) 0.28 (0.10–7.76)
Referent
Referent
1.38 (1.01–1.89) 1.40 (1.02–1.90)
2.11 (1.33–3.34) 2.24 (1.42–3.53)
1.66 (0.59–4.65) 1.76 (0.34–4.90)
108 (36.3)
184 (63.7)
1,186 (43.7)
1,520 (56.3)
1,284 (43.0)
1,704 (57.0)
Referent
Referent
1.32 (1.03–1.69) 1.35 (1.06–1.71)
176 (60.9)
42 (14.5)
37 (12.8)
14 (4.8)
5 (1.7)
15 (5.2)
26 (9.0)
26 (9.0)
27 (9.3)
3 (1.0)
77 (26.6)
44 (15.2)
57 (9.7)
126 (43.6)
178 (61.6)
168 (58.1)
1.5 (+ 2.2)
1,648 (61.0)
384 (14.2)
362 (13.4)
55 (2.0)
57 (2.1)
193 (7.2)
348 (12.9)
214 (7.9)
311 (11.5)
47 (1.7)
710 (26.3)
513 (19.0)
1,824 (61.0)
426 (14.3)
399 (13.4)
69 (2.3)
62 (2.1)
208 (7.0)
Referent
Referent
1.16 (0.84–1.62) 1.20 (0.87–1.66)
0.97 (0.67–1.40) 1.03 (0.72–1.48)
1.98 (1.15–3.43) 2.02 (1.18–3.45)
0.94 (0.38–2.30) 0.99 (0.41–2.37)
0.76 (0.43–1.32) 0.85 (0.50–1.46)
374 (12.5)
240 (8.3)
338 (11.3)
50 (1.7)
787 (26.3)
557 (18.6)
0.70 (0.47–1.06) 0.73 (0.48–1.10)
1.14 (0.75–1.74) 1.18 (0.78–1.79)
0.78 (0.52–1.17) 0.82 (0.55–1.21)
0.73 (0.24–2.24) 0.77 (0.26–2.27)
1.08 (0.82–1.41) 1.16 (0.89–1.51)
0.79 (0.57–1.10) 0.88 (0.63–1.22)
546 (20.2)
1,180 (43.7)
1,689 (62.6)
1,495 (55.4)
1.3 (+ 2.0)
603 (20.2)
1,306 (44.7)
1,867 (62.5)
1,663 (55.7)
1.3 (+ 2.0)
0.89 (0.66–1.21) 0.89 (0.66–1.20)
0.91 (0.71–1.15) 0.88 (0.70–1.12)
0.96 (0.75–1.22) 0.97 (0.76–1.23)
1.09 (0.86–1.39) 1.07 (0.85–1.37)
1.00 (0.98–1.02) 1.00 (0.98–1.02)
3
7
10
*Values are no. or no. (%) except as indicated. ARR, attack rate ratio; DENV, dengue virus.
†Participants 0.5–30 years of age in 2012 and 0.5–20 years of age in 2013.
‡Adjusted for sex.
§Adjusted for age.
¶No. housesholds with >1 DENV case.
#Nonpiped water comes from a river, pond, lake, or a well that does not have a pump; piped water comes from indoor or outdoor taps with a tube well and
pump.
**Adjusted for sex and occupation.
††Adjusted for age, relationship to index case-patient, occupation, and primary source of income.
1358
Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 25, No. 7, July 2019
Table 3. Univariate tests for associations between
sociodemographic factors and infecting serotypes with
asymptomatic dengue virus infections, Cambodia, 2012–2013*
Asymptomatic,
n = 26
Symptomatic,
n = 320
p value†
17
9
154
166
Factor
Sex
M
F
Age, y
0.5 to <1
1–5
6–10
11–14
15–30
Mean
(+ SD, median)
0
9 (34.6)
9 (34.6)
5 (19.2)
3 (11.5)
9.2
(+ 7.2, 8)
Relationship to index case-patient
17 (65.4)
Neighbor
5 (19.2)
Cousin
1 (3.9)
Sibling
Other
3 (11.5)
Source of household income
Planting crops
Working in a
14 (53.8)
3 (11.5)
0.09
0.976
0.004
0.812
9 (2.8)
99 (30.9)
117 (36.6)
66 (20.6)
29 (9.1)
11.0
(+ 7.1, 10)
243 (76.0)
53 (16.6)
22 (6.9)
1 (0.3)
192 (60.0)
53 (16.6)
factory
0
6 (1.9)
3 (11.5)
5 (19.2)
1 (3.9)
36 (11.3)
17 (5.3)
Shopkeeping
Fishing, farming,
animal husbandry
Working in
government
Other or missing
DENV serotype‡
21 (80.8)
DENV-1
2 (7.7)
DENV-2
0
DENV-3
DENV-4
3 (11.5)
*Values are no. (%) patients except as indicated. DENV, dengue virus.
†By Fisher test or 2 test.
‡Data for 5 symptomatic patients were missing, and another patient was
excluded from analysis because of co-infection with DENV-1 and DENV-2.
249 (79.1)
35 (11.1)
2 (0.6)
28 (8.9)
16 (5.0)
0.892
this study. We noted that attack rates were lower in PI-V,
14.7/1,000 participants (14/952), than in PI-H, 47.4/1,000
participants (332/7,008). Circulation of DENV around fe-
brile index case-patients identified through PI-V was less
intense, but with more asymptomatic cases, than around in-
dex case-patients identified through PI-H. Aside from pos-
sible detection biases (17), multiple factors could explain
this observation and deserve further research.
Our study documented cases of DENV infection in trans-
mission clusters located around index case-patients. We found
that 26.6% of DENV-confirmed case-patients reported clini-
cal symptoms, including headache and muscle pain, but no
fever even in the absence of antipyretics, comparable to data
from Thailand, where 40.4% of the DENV cases remained
afebrile (17). The appearance of afebrile DENV-infected pa-
tients raises potential concerns for case definitions for detec-
tion, especially of imported cases in at-risk countries.
An additional 7.5% of DENV-confirmed case-patients
had no symptoms during the 10-day course of clinical
monitoring, a considerably lower rate than estimates from
other prospective studies (5,18–21). Published sources refer
Asymptomatic Dengue Virus Infections, Cambodia
to inapparent infections, often defined as afebrile clinical
complaints with biologic evidence of DENV infection, rang-
ing from 20% to 80% of cases (19,22,23). Previous studies
used different definitions of asymptomatic infection than
ours, but the major difference lies in follow-up monitoring.
Other retrospective studies used school or work absentee-
ism as a basis for follow-up (19). Strictly asymptomatic
patients, such as those we describe, escape detection by sur-
veillance or control measures, infect mosquitoes, and might
disproportionately contribute to DENV transmission (5).
The DENV burden documented through hospital-
based surveillance of febrile case-patients in Thailand and
Vietnam showed a shift to older age groups (24,25). Our
active, systematic case-finding system to identify DENV in
villages in Cambodia found the attack rate was highest in
children <10 years of age, which is what we expected in a
dengue-endemic country with frequent outbreaks demon-
strated in other careful studies (26). This finding raises con-
cerns because recommendations for the only licensed den-
gue vaccine are for use in persons 9–45 years of age with
demonstrated evidence of past DENV infection (27). Our
study demonstrates that children in rural Cambodia might
have undergone >1 DENV infection before 9 years of age,
reducing the potential cost-effectiveness of vaccination.
Few studies have explored the role of socioeconom-
ic status, which might be a proxy for peridomestic envi-
ronmental management, on DENV infection in Southeast
Asia. Often, the direction of the association is unclear and
socioeconomic status has entirely different associations de-
pending on the setting (28). Our study shows that the ad-
justed risk for DENV infection was highest in households
in which the main source of income was from fishing, farm-
ing, or animal husbandry, activities associated with lower
average household income in Cambodia.
We found temephos provided no additional protection
against DENV infection after adjusting for other factors.
Although temephos is effective in reducing Aedes spp. lar-
val populations in water storage jars, its use did not cor-
relate with lower DENV transmission in Cambodia or else-
where (8), due to incorrect distribution coverage, dosage,
and placement (29) or multiple vector breeding sites. In ad-
dition, increases in temephos-resistant A. aegypti mosquito
larvae have been documented in Cambodia (30).
Unexpectedly, we found a higher risk for DENV with
piped water as a main water source after adjusting for other
factors, contrary to a study in Thailand (6). However, piped
water with suboptimal sanitation in Cambodia might contrib-
ute to collection of water in or around households that could
become breeding sites for DENV-transmitting mosquitoes.
We found that 40.8% of the variability in probabil-
ity of being DENV infected was explained by differences
between households. Those living in the same household
as an index case-patient were 2.11 times more likely to be
Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 25, No. 7, July 2019
1359
RESEARCH
infected, consistent with other published sources. A study
in Mexico found that the risk for infection for those liv-
ing with an index case-patient was twice that of someone
living in a 50-meter radius of an index case-patient (31).
This relationship was further described in a cluster study
in Thailand that showed decreased risk for infection with
increasing distance from the index case household (31).
This clustering effect around an index case, however,
seems to occur only on a short temporal scale, at least in
urban settings (32).
Rates and severity of illness after infection with the
different DENV serotypes differ widely (33,34). The only
notable epidemiologic factor associated with asymptomatic
DENV infections in our study was being family-related to
the index case-patient. Another study showed that adaptive
immune responses against DENV differ between persons
with symptomatic and asymptomatic DENV infection (35),
which might explain our observations. We found no other
associated epidemiologic factor, including age or cluster
effects. Although the ratio of male to female participants
was twice as high among asymptomatic than symptomatic
participants, this finding was not statistically significant,
likely due to sample size.
Although the size and duration of our study confer
strength to our data, it might suffer from biases and limita-
tions, especially due to the small number of strictly asymp-
tomatic DENV-positive participants after stratifying by
DENV serotype. We found dengue incidence rates high-
est in young children. These data might be biased because
we focused on investigating clusters around an index case,
perhaps overestimating the incidence in the general popula-
tion. DENV circulation, however, is intense in children in
Cambodia, and these figures remain comparable to those
found in dengue studies that use different methods, ranging
from 20 to 80 per 1,000 person-seasons (1). Furthermore,
we did not capture cases referred to the private sector, low-
ering our estimates somewhat. Healthy male workers often
were away at the time of the investigations, possibly lead-
ing to an overestimation of DENV incidence. These work-
ers, however, are >18 years of age, but DENV infections
occur mainly in persons <15 years of age in Cambodia (4).
Documentation bias might also have pulled our risk
factor estimates toward the null. We did not document solid
waste disposal in our study, but comparatively high Bre-
teau index values have been reported in Cambodia (29). In
addition, we could have missed details or misrepresented
implementation of mosquito-control measures. Despite the
potential misclassification, mosquito-control measures re-
main nondifferential and likely had no major effect on our
risk estimates.
Further, 7.5% of our DENV-infected participants re-
mained strictly asymptomatic. Aside from case definition
issues we discuss, our method of screening for DENV
around symptomatic cases might have underestimated the
number of asymptomatic DENV infections. In addition,
we did not enroll persons who tested negative for DENV
IgM, NS1-capture ELISA, and qRT-PCR. Some of these
persons might have been infected but not yet mounted an
IgM response, so that NS1 and viral RNA titers had al-
ready receded to undetectable levels when we tested them.
This strict case definition might have underestimated the
incidence of asymptomatic cases, but a precise retrospec-
tive documentation of such cases would be extremely
difficult. Similarly, we retrospectively conducted MAC-
ELISA on samples collected during perifocal investiga-
tions and identified 11 cases of IgM seroconversion in the
absence of PCR- or NS1-positive tests. Even in the context
of JEV cocirculation, some of these cases could have been
true DENV infections, but including them would not have
changed the overall estimated attack rate. Previous stud-
ies suggested virus serotype might affect severity and types
of symptoms and observed that DENV-1 infections more
frequently were associated with clinically apparent illness
(36,37). Virus molecular analysis studies are ongoing to
determine whether specific strains cause more asymptom-
atic infection than others. Furthermore, DENV infection in
Cambodia occurs mainly in children who might be more
likely to answer positively to daily-repeated questions on
dengue symptoms, somewhat underestimating asymptom-
atic cases. Having implemented careful and thorough 10-
day clinical assessment of objective symptoms in each as-
ymptomatic DENV-positive participant, we believe these
figures reflect the true proportion of strictly asymptomatic
DENV infection in our setting. However, we collected
our findings mainly in children with DENV-1 infection in
Cambodia. Whether these findings are directly applicable
to other epidemiologic settings, populations, or virus sero-
types or genotypes remains to be determined (33).
Finally, vaccination against JEV might have led to
cross-protection against symptomatic dengue. Data on JEV
vaccination were not collected during perifocal investiga-
tions. According to local health centers, however, JEV vac-
cine has been provided only recently and only for children
9–24 months of age. In our study, only 3 children were
DENV-positive in that age category.
Our study demonstrates that systematically relying on
fever for DENV case definition can underestimate cases
and hinder control efforts in areas with potential vectors
and at risk for DENV introduction. We found 7.5% of
DENV-infected participants remained strictly asymptom-
atic, which has wide-ranging epidemiologic consequences.
Undetected sources can increase transmission (5), a factor
that must be taken into account in future vaccine coverage
and vaccine effectiveness studies. The attack rate differ-
ences observed around febrile index case-patients detected
in village surveillance and index case-patients detected in
1360
Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 25, No. 7, July 2019
hospital surveillance deserve further study. In-depth virus
(36) and human genetic studies could contribute useful in-
sights (33,35). Our strict definition of asymptomatic DENV
infections should be considered when designing studies
that aim to elucidate the pathophysiological mechanisms
of dengue disease.
Acknowledgments
The authors gratefully acknowledge participating hospitals,
villages, and study participants, as well as Tineke Cantaert for
her editorial comments and suggestions.
Dengue Framework for Resisting Epidemics in Europe studies
were funded by a grant (no. 282378) from the European Union
7th FP.
P.B. is a former head of virology at Institut Pasteur du
Cambodge and is an employee of GSK Vaccines, Singapore.
About the Authors
Dr. Ly is a medical doctor and epidemiologist at the
Epidemiology & Public Health Department at Institut Pasteur
du Cambodge; his primary research interests are epidemiology
of endemic and epidemic arboviruses and zoonoses in the
Mekong Region. Ms. Fortas is an epidemiologist whose research
interests are tropical infectious diseases in low- and middle-
income countries.
References
1. Bhatt S, Gething PW, Brady OJ, Messina JP, Farlow AW,
Moyes CL, et al. The global distribution and burden of dengue.
Nature. 2013;496:504–7. http://dx.doi.org/10.1038/nature12060
2. Simmons CP, Farrar JJ, van Vinh Chau N, Wills B. Dengue.
N Engl J Med. 2012;366:1423–32. http://dx.doi.org/10.1056/
NEJMra1110265
3. Western Pacific Regional Office of the World Health Organization.
Dengue situation updates [cited 2017 Jul 12]. http://www.wpro.
who.int/entity/emergencies/wp_dengue_biweekly/en
4. Vong S, Goyet S, Ly S, Ngan C, Huy R, Duong V, et al.
Under-recognition and reporting of dengue in Cambodia: a capture-
recapture analysis of the National Dengue Surveillance System.
Epidemiol Infect. 2012;140:491–9. http://dx.doi.org/10.1017/
S0950268811001191
5. Duong V, Lambrechts L, Paul RE, Ly S, Lay RS, Long KC,
et al. Asymptomatic humans transmit dengue virus to mosquitoes.
Proc Natl Acad Sci U S A. 2015;112:14688–93. http://dx.doi.org/
10.1073/pnas.1508114112
6. Mammen MP, Pimgate C, Koenraadt CJM, Rothman AL,
Aldstadt J, Nisalak A, et al. Spatial and temporal clustering of
dengue virus transmission in Thai villages. PLoS Med.
2008;5:e205. http://dx.doi.org/10.1371/journal.pmed.0050205
7. Dussart P, Baril L, Petit L, Beniguel L, Quang LC, Ly S, et al.
Clinical and virological study of dengue cases and the members
of their households: the multinational DENFRAME Project.
PLoS Negl Trop Dis. 2012;6:e1482. http://dx.doi.org/10.1371/
journal.pntd.0001482
8. Huy R, Buchy P, Conan A, Ngan C, Ong S, Ali R, et al. National
dengue surveillance in Cambodia 1980–2008: epidemiologi-
cal and virological trends and the impact of vector control. Bull
Asymptomatic Dengue Virus Infections, Cambodia
World Health Organ. 2010;88:650–7. http://dx.doi.org/10.2471/
BLT.09.073908
9. Vong S, Khieu V, Glass O, Ly S, Duong V, Huy R, et al. Dengue
incidence in urban and rural Cambodia: results from population-
based active fever surveillance, 2006–2008. PLoS Negl Trop Dis.
2010;4:e903. http://dx.doi.org/10.1371/journal.pntd.0000903
10. Hue KDT, Tuan TV, Thi HTN, Bich CTN, Anh HHL, Wills BA,
et al. Validation of an internally controlled one-step real-time
multiplex RT-PCR assay for the detection and quantitation of
dengue virus RNA in plasma. J Virol Methods. 2011;177:168–73.
http://dx.doi.org/10.1016/j.jviromet.2011.08.002
11. Andries A-C, Duong V, Ngan C, Ong S, Huy R, Sroin KK, et al.
Field evaluation and impact on clinical management of a rapid
diagnostic kit that detects dengue NS1, IgM and IgG. PLoS Negl
Trop Dis. 2012;6:e1993. http://dx.doi.org/10.1371/
journal.pntd.0001993
12. Andries A-C, Duong V, Ong S, Ros S, Sakuntabhai A,
Horwood P, et al. Evaluation of the performances of six commercial
kits designed for dengue NS1 and anti-dengue IgM, IgG and IgA
detection in urine and saliva clinical specimens. BMC Infect Dis.
2016;16:201. http://dx.doi.org/10.1186/s12879-016-1551-x
13. Duong V, Andries A-C, Ngan C, Sok T, Richner B,
Asgari-Jirhandeh N, et al. Reemergence of chikungunya virus in
Cambodia. Emerg Infect Dis. 2012;18:2066–9. http://dx.doi.org/
10.3201/eid1812.120471
14. Zou G. A modified Poisson regression approach to prospective
studies with binary data. Am J Epidemiol. 2004;159:702–6.
http://dx.doi.org/10.1093/aje/kwh090
15. Wu S, Crespi CM, Wong WK. Comparison of methods for
estimating the intraclass correlation coefficient for binary responses
in cancer prevention cluster randomized trials. Contemp Clin Trials.
2012;33:869–80. http://dx.doi.org/10.1016/j.cct.2012.05.004
16. Tarantola A, Goutard F, Newton P, de Lamballerie X, Lortholary O,
Cappelle J, et al. Estimating the burden of Japanese encephalitis
virus and other encephalitides in countries of the Mekong region.
PLoS Negl Trop Dis. 2014;8:e0002533. https://doi.org/10.1371/
journal.pntd.0002533
17. Yoon I-K, Srikiatkhachorn A, Hermann L, Buddhari D, Scott TW,
Jarman RG, et al. Characteristics of mild dengue virus infection
in Thai children. Am J Trop Med Hyg. 2013;89:1081–7.
http://dx.doi.org/10.4269/ajtmh.13-0424
18. Porter KR, Beckett CG, Kosasih H, Tan RI, Alisjahbana B,
Rudiman PIF, et al. Epidemiology of dengue and dengue
hemorrhagic fever in a cohort of adults living in Bandung,
West Java, Indonesia. Am J Trop Med Hyg. 2005;72:60–6.
http://dx.doi.org/10.4269/ajtmh.2005.72.60
19. Endy TP, Chunsuttiwat S, Nisalak A, Libraty DH, Green S,
Rothman AL, et al. Epidemiology of inapparent and symptomatic
acute dengue virus infection: a prospective study of primary
school children in Kamphaeng Phet, Thailand. Am J Epidemiol.
2002;156:40–51. http://dx.doi.org/10.1093/aje/kwf005
20. Kosasih H, Alisjahbana B, Nurhayati, de Mast Q, Rudiman IF,
Widjaja S, et al. The epidemiology, virology and clinical findings
of dengue virus infections in a cohort of Indonesian adults in
western Java. PLoS Negl Trop Dis. 2016;10:e0004390.
http://dx.doi.org/10.1371/journal.pntd.0004390
21. Yoon I-K, Rothman AL, Tannitisupawong D, Srikiatkhachorn A,
Jarman RG, Aldstadt J, et al. Underrecognized mildly symptomatic
viremic dengue virus infections in rural Thai schools and villages.
J Infect Dis. 2012;206:389–98. http://dx.doi.org/10.1093/infdis/jis357
22. Balmaseda A, Standish K, Mercado JC, Matute JC, Tellez Y,
Saborío S, et al. Trends in patterns of dengue transmission over
4 years in a pediatric cohort study in Nicaragua. J Infect Dis.
2010;201:5–14. http://dx.doi.org/10.1086/648592
23. Endy TP, Anderson KB, Nisalak A, Yoon I-K, Green S,
Rothman AL, et al. Determinants of inapparent and symptomatic
Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 25, No. 7, July 2019
1361
RESEARCH
dengue infection in a prospective study of primary school children
in Kamphaeng Phet, Thailand. PLoS Negl Trop Dis. 2011;5:e975.
http://dx.doi.org/10.1371/journal.pntd.0000975
24. Anders KL, Nga H, Thuy NTV, Ngoc TV, Tam CT, Tai LTH,
et al. Households as foci for dengue transmission in highly
urban Vietnam. PLoS Negl Trop Dis. 2015;9:e0003528.
http://dx.doi.org/10.1371/journal.pntd.0003528
25. Limkittikul K, Brett J, L’Azou M. Epidemiological trends of
dengue disease in Thailand (2000–2011): a systematic literature
review. PLoS Negl Trop Dis. 2014;8:e3241. http://dx.doi.org/
10.1371/journal.pntd.0003241
26. Hadinegoro SR, Arredondo-García JL, Capeding MR, Deseda C,
Chotpitayasunondh T, Dietze R, et al.; CYD-TDV Dengue Vaccine
Working Group. Efficacy and long-term safety of a dengue vaccine
in regions of endemic disease. N Engl J Med. 2015;373:1195–206.
http://dx.doi.org/10.1056/NEJMoa1506223
27. Dengue vaccine: WHO position paper, September 2018—
recommendations. Vaccine. 2018;S0264-410X(18)31339-2.
28. Mulligan K, Dixon J, Joanna Sinn C-L, Elliott SJ. Is dengue a
disease of poverty? A systematic review. Pathog Glob Health. 2015;
109:10–8. http://dx.doi.org/10.1179/2047773214Y.0000000168
29. Khun S, Manderson LH. Abate distribution and dengue control in
rural Cambodia. Acta Trop. 2007;101:139–46. http://dx.doi.org/
10.1016/j.actatropica.2007.01.002
30. Polson KA, Curtis C, Seng CM, Olson JG, Chantha N, Rawlins SC.
Susceptibility of two Cambodian populations of Aedes aegypti
mosquito larvae to temephos during 2001. Dengue Bull. 2001;
25:79–83.
31. Martínez-Vega RA, Danis-Lozano R, Díaz-Quijano FA,
Velasco-Hernández J, Santos-Luna R, Román-Pérez S, et al.
Peridomestic infection as a determining factor of dengue
transmission. PLoS Negl Trop Dis. 2015;9:e0004296.
http://dx.doi.org/10.1371/journal.pntd.0004296
32. Yoon I-K, Getis A, Aldstadt J, Rothman AL, Tannitisupawong D,
Koenraadt CJM, et al. Fine scale spatiotemporal clustering of
dengue virus transmission in children and Aedes aegypti in rural
Thai villages. PLoS Negl Trop Dis. 2012;6:e1730. http://dx.doi.org/
10.1371/journal.pntd.0001730
33. Grange L, Simon-Loriere E, Sakuntabhai A, Gresh L, Paul R,
Harris E. Epidemiological risk factors associated with high global
frequency of inapparent dengue virus infections. Front Immunol.
2014;5:280. http://dx.doi.org/10.3389/fimmu.2014.00280
34. Clapham HE, Cummings DAT, Johansson MA. Immune status
alters the probability of apparent illness due to dengue virus
infection: evidence from a pooled analysis across multiple cohort
and cluster studies. PLoS Negl Trop Dis. 2017;11:e0005926.
http://dx.doi.org/10.1371/journal.pntd.0005926
35. Simon-Lorière E, Duong V, Tawfik A, Ung S, Ly S, Casadémont I,
et al. Increased adaptive immune responses and proper
feedback regulation protect against clinical dengue. Sci Transl
Med. 2017;9:eaal5088. http://dx.doi.org/10.1126/scitranslmed.
aal5088
36. Li D, Lott WB, Lowry K, Jones A, Thu HM, Aaskov J.
Defective interfering viral particles in acute dengue infections.
PLoS One. 2011;6:e0019447. https://doi.org/10.1371/journal.
pone.0019447
37. Yung C-F, Lee K-S, Thein T-L, Tan L-K, Gan VC, Wong JGX,
et al. Dengue serotype-specific differences in clinical manifestation,
laboratory parameters and risk of severe disease in adults,
Singapore. Am J Trop Med Hyg. 2015;92:999–1005.
http://dx.doi.org/10.4269/ajtmh.14-0628
Address for correspondence: Arnaud Tarantola, Institut Pasteur du
Cambodge, Epidemiology and Public Health Unit, PO Box 983,
Phnom Penh, Cambodia; email: [email protected]
EID Podcast:
Antimicrobial
Drug Resistance
and Gonorrhea
Neisseria gonorrhoeae, the causative pathogen of gon-
orrhea, has been designated an urgent antimicrobial drug
resistance threat by the Centers for Disease Control and
Prevention. Since the introduction of antimicrobial drugs
in the first half of the 20th century, N. gonorrhoeae has
successively developed resistance to each antimicrobial
agent recommended for gonorrhea treatment. In the Unit-
ed States, the prevalence of resistance in N. gonorrhoeae
often varies by sex of partner and by geographic region.
Prevalence is often greater in isolates from gay, bisexual,
and other men who have sex with men than those from
men who have sex only with women, and prevalence is
often highest in the West and lowest in the South. Resis-
tant strains, in particular penicillinase-producing N. gon-
orrhoeae, fluoroquinolone-resistant N. gonorrhoeae, and
gonococcal strains with re-
duced cephalosporin suscep-
tibility, seemed to emerge ini-
tially in the West (Hawaii and
the West Coast) before spread-
ing eastward across the coun-
try. These geographic patterns
seem to support the idea that
importation of resistant strains
from other regions of
the
world, such as eastern Asia, is
a primary factor of the emer-
gence of resistant gonococci
in the United States. Whereas
antimicrobial drug prescribing patterns have been clearly
associated with the emergence of resistance in other bac-
terial pathogens, the degree to which domestic antimicro-
bial use and subsequent selection pressure contributes to
the emergence of gonococcal antimicrobial resistance in
the United States is unclear. Using an ecologic approach,
we sought to investigate the potential geographic and tem-
poral association between antimicrobial drug susceptibil-
ity among US N. gonorrhoeae isolates and domestic out-
patient antimicrobial drug prescribing rates in the United
States during 2005–2013.
Visit our website to listen:
https://www2c.cdc.gov/podcasts/
player.asp?f=8647449
®
1362
Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 25, No. 7, July 2019
| null |
10.1371_journal.pcbi.1008853.pdf
|
cation.
Data Availability Statement: BCI FDP data are
available from https://repository.si.edu/handle/
10088/11. The authors do not own the data and are
unable to share it in a public repository, as the
majority of the data is held by various government
agencies. However interested researchers are able
to access all the data through the Smithsonian
Institution’s ForestGeo project (https://forestgeo.si.
edu/explore-data) The EAA analysis programs,
written in R, along with their documentation, are
1 Divis
|
BCI FDP data are available from https://repository.si.edu/handle/ 10088/11 . The authors do not own the data and are unable to share it in a public repository, as the majority of the data is held by various government agencies. However interested researchers are able to access all the data through the Smithsonian Institution's ForestGeo project ( https://forestgeo.si .
|
RESEARCH ARTICLE
Interactions between all pairs of neighboring
trees in 16 forests worldwide reveal details of
unique ecological processes in each forest,
and provide windows into their evolutionary
histories
4,5, Yi Jin6,
7, Jill ThompsonID
12, Sara GermainID
7, Heming Liu13, Joseph SmokeyID
1☯*, Bin Wang2☯, Shuai Fang3, Yunquan WangID
8, Kyle E. Harms9, Sandeep Pulla10,11,
Christopher WillsID
James LutzID
Bonifacio PasionID
Hsin Su15, Nathalie ButtID
20, H. S. DattarajaID
YangID
Shameema EsufaliID
Chang-Fu Hsieh27, Fangliang He18, Stephen Hubbell28, Zhanqing Hao3‡, Akira Itoh29,
30, Buhang Li18, Xiankun Li2, Keping Ma5, Michael MorecroftID
David KenfackID
Xiangcheng Mi5, Yadvinder Malhi32, Perry Ong33†‡, Lillian Jennifer RodriguezID
35, Raman SukumarID
10,34, I Fang SunID
S. SureshID
13, Xugao Wang3, T. L. YaoID
Maria Uriarte38, Xihua WangID
16,17, Chengjin Chu18, George ChuyongID
23,
21, Stuart Davies22, Sisira EdiriweeraID
31,
33, H.
14, Sheng-
19, Chia-Hao Chang-
10, Sylvester Tan36‡, Duncan Thomas37,
25, Jess Zimmermann39
24, Christine Dawn Fletcher25, Nimal Gunatilleke26, Savi Gunatilleke26,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Wills C, Wang B, Fang S, Wang Y, Jin Y,
Lutz J, et al. (2021) Interactions between all pairs
of neighboring trees in 16 forests worldwide reveal
details of unique ecological processes in each
forest, and provide windows into their evolutionary
histories. PLoS Comput Biol 17(4): e1008853.
https://doi.org/10.1371/journal.pcbi.1008853
Editor: Mercedes Pascual, University of Chicago,
UNITED STATES
Received: May 28, 2020
Accepted: March 3, 2021
Published: April 29, 2021
Copyright: This is an open access article, free of all
copyright, and may be freely reproduced,
distributed, transmitted, modified, built upon, or
otherwise used by anyone for any lawful purpose.
The work is made available under the Creative
Commons CC0 public domain dedication.
Data Availability Statement: BCI FDP data are
available from https://repository.si.edu/handle/
10088/11. The authors do not own the data and are
unable to share it in a public repository, as the
majority of the data is held by various government
agencies. However interested researchers are able
to access all the data through the Smithsonian
Institution’s ForestGeo project (https://forestgeo.si.
edu/explore-data) The EAA analysis programs,
written in R, along with their documentation, are
1 Division of Biological Sciences, University of California, San Diego, La Jolla, California, United States of
America, 2 Guangxi Key Laboratory of Plant Conservation and Restoration Ecology in Karst Terrain, Guangxi
Institute of Botany, Guangxi Zhuang Autonomous Region and Chinese Academy of Sciences, Guilin, 3 Key
Laboratory of Forest Ecology and Management, Institute of Applied Ecology, Chinese Academy of Sciences,
Shenyang, 4 College of Chemistry and Life Sciences, Zhejiang Normal University, Jinhua, 5 State Key
Laboratory of Vegetation and Environmental Change, Institute of Botany, Chinese Academy of Sciences, 20
Nanxincun, Xiangshan, Beijing, 6 College of Life Sciences, Zhejiang University, Hangzhou, 7 Department of
Wildland Resources, Utah State University, Logan, Utah, United States of America, 8 Center for Ecology &
Hydrology, Penicuik, Midlothian, Scotland, 9 Department of Biological Sciences, Louisiana State University,
Baton Rouge, Los Angeles, United States of America, 10 Divecha Centre for Climate Change, Indian Institute
of Science, Bengaluru, India, 11 National Centre for Biological Sciences, GKVK Campus, Bengaluru, India,
12 Center for Integrative Conservation, Xishuangbanna Tropical Botanical Garden, Chinese Academy of
Sciences, Menglun, Mengla, Yunnan, 13 Zhejiang Tiantong Forest Ecosystem National Observation and
Research Station, School of Ecological and Environmental Sciences, East China Normal University,
Shanghai, 14 Department of Biology, Memorial University of Newfoundland, Newfoundland, Canada,
15 Institute of Ecology and Evolutionary Biology, National Taiwan University, Taipei, 16 School of Geography
and the Environment, University of Oxford, Oxford, United Kingdom, 17 School of Biological Sciences, The
University of Queensland, St. Lucia, Queensland, Australia, 18 Department of Ecology, State Key Laboratory
of Biocontrol and School of Life Sciences, Sun Yat-sen University, Guangzhou, 19 Department of Botany and
Plant Physiology, University of Buea, Cameroon, 20 Department of Biological Sciences, National Sun Yat-
sen University, Kaohsiung, 21 National Centre for Biological Sciences, Bengaluru, India, 22 Center for
Tropical Forest Science, Smithsonian Institution, Washington, DC, United States of America, 23 Faculty of
Science and Technology, Uva Wellassa University, Badulla, Sri Lanka, 24 Department of Botany, University
of Peradeniya, Peradeniya Sri Lanka, 25 Forest Research Institute Malaysia, Kepong Selangor, Malaysia,
26 Dept. of Botany, Faculty of Science, University of Peradeniya, Peradeniya Sri Lanka, 27 Taiwan Forestry
Research Institute, Taipei, 28 Department of Ecology and Evolutionary Biology, University of California, Los
Angeles, Los Angeles, California, United States of America, 29 Graduate School of Science, Osaka City
University, Sumiyoshi Ku, Osaka, Japan, 30 Center for Tropical Forest Science–Forest Global Earth
Observatory (CTFS-ForestGEO), Smithsonian Tropical Research Institute, NMNH—MRC, Washington, DC,
United States of America, 31 Natural England Mail Hub, County Hall, Worcester, United Kingdom, 32 School
of Geography and the Environment, Oxford University Centre for the Environment, University of Oxford,
Oxford, United Kingdom, 33 Institute of Biology, College of Science, University of the Philippines Diliman,
Diliman, Quezon City, Philippines, 34 Centre for Ecological Sciences, Indian Institute of Science, Bengaluru,
India, 35 Department of Natural Resources and Environmental Studies, National Dong Hwa University,
Hualien, 36 Forest Department Sarawak, Bangunan Wisma Sumber Alam, Jalan Stadium, Petra Jaya,
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
1 / 33
freely available from https://github.com/
wangbinzjcc/EAAr.
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist. Authors Perry
Ong, Zhanqing Hao and Sylvester Tan were unable
to confirm their authorship contributions. On their
behalf, the corresponding author has reported their
contributions to the best of their knowledge.
Each of 16 forests shows a unique pattern of between-tree interactions
Kuching, Sarawak, Malaysia, 37 Department of Biology, Washington State University, Vancouver,
Washington State, United States of America, 38 Department of Ecology, Evolution, and Environmental
Biology, Columbia University, New York city, New York, United States of America, 39 Dept of Environmental
Sciences, University of Puerto Rico, Rio Piedras, San Juan, PR, United States of America
☯ These authors contributed equally to this work.
† Deceased.
‡ Unavailable.
* [email protected]
Abstract
When Darwin visited the Galapagos archipelago, he observed that, in spite of the islands’
physical similarity, members of species that had dispersed to them recently were beginning
to diverge from each other. He postulated that these divergences must have resulted pri-
marily from interactions with sets of other species that had also diverged across these other-
wise similar islands. By extrapolation, if Darwin is correct, such complex interactions must
be driving species divergences across all ecosystems. However, many current general eco-
logical theories that predict observed distributions of species in ecosystems do not take the
details of between-species interactions into account. Here we quantify, in sixteen forest
diversity plots (FDPs) worldwide, highly significant negative density-dependent (NDD) com-
ponents of both conspecific and heterospecific between-tree interactions that affect the
trees’ distributions, growth, recruitment, and mortality. These interactions decline smoothly
in significance with increasing physical distance between trees. They also tend to decline in
significance with increasing phylogenetic distance between the trees, but each FDP exhibits
its own unique pattern of exceptions to this overall decline. Unique patterns of between-spe-
cies interactions in ecosystems, of the general type that Darwin postulated, are likely to
have contributed to the exceptions. We test the power of our null-model method by using a
deliberately modified data set, and show that the method easily identifies the modifications.
We examine how some of the exceptions, at the Wind River (USA) FDP, reveal new details
of a known allelopathic effect of one of the Wind River gymnosperm species. Finally, we
explore how similar analyses can be used to investigate details of many types of interactions
in these complex ecosystems, and can provide clues to the evolution of these interactions.
Author summary
Worldwide, ecosystems are collapsing or in danger of collapse, but the precise causes of
these collapses are often unknown. Observational and experimental evidence shows that
all ecosystems are characterized by strong interactions between and among species, and
that these webs of interactions can be important contributors to the preservation of eco-
system diversity. But many of the interactions–such as those involving pathogenic micro-
organisms and the chemical defenses that are mounted by their prey–are not easily
identified and analyzed in ecosystems that may have hundreds or thousands of species.
Here we use our equal-area-annulus analytical method to examine census data from over
three million trees in forest plots from around the world. We show how the method can
be used to flag pairs and groups of species that exhibit unusual levels of interaction and
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
2 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
that are likely on further investigation to yield information about their causative mecha-
nisms. We give a detailed example showing how some of these interactions can be traced
to defense mechanisms that are possessed by one of the tree species. We explore how our
method can be used to identify the between-species interactions that play the largest roles
in the maintenance of ecosystems and their diversity.
Introduction
The] inhabitants of each separate [Galapagos] island, though mostly distinct, are related in
an incomparably closer degree to each other than to the inhabitants of any other part of the
world‥‥ [Dissimilarities] between the endemic inhabitants of the islands may be used as an
argument against my views; for it may be asked, how has it happened in the several islands
situated within sight of each other, having the same geological nature, the same height, cli-
mate, &c., that many of the immigrants should have been differently modified, though only
in a small degree. This long appeared to me a great difficulty: but it arises in chief part from
the deeply-seated error of considering the physical conditions of a country as the most
important for its inhabitants; whereas it cannot, I think, be disputed that the nature of the
other inhabitants, with which each has to compete, is at least as important, and generally a
far more important element of success.
Charles Darwin, The Origin of Species, 1st ed. 1859, p. 400.
In this passage from the Origin, Darwin effectively founded the field of evolutionary ecology.
He was faced with the difficulty of explaining recent adaptive radiations that sometimes
resulted in distinct species on the different islands of the Gala´pagos archipelago, even though
the islands have similar physical properties. The solution, he suggested, must lie in these evolv-
ing populations’ interactions with other species, the mix of which should differ among the
individual islands. (And those other species, by his reasoning, would simultaneously be evolv-
ing in their own unique directions as a result of their own sets of between-species interactions.)
But his claim that the importance of such between-species interactions "cannot be disputed"
was far from being demonstrated at the time.
In the century and a half since the Origin, ecologists and evolutionary biologists have
explored the many interactions among species that share the same ecological community, in
ever-greater detail and with ever-more-sophisticated tools [1]. Modeling has pointed the way
[2–6]. Even so, such interactions must involve many more species, occupying a variety of dif-
ferent trophic levels, than those that can be examined in a typical study. Host-pathogen inter-
actions were early postulated to be important in the maintenance of species diversity [7], and
were soon realized to have a high likelihood of contributing to negative density-dependent
(NDD) interactions between host species [8, 9]. Such interactions have been detected in the
relatively simple ecosystems of the Gala´pagos [10] and in complex ecosystems such as tropical
forests [11–13], coral reefs [14], and lacustrine fish communities [15].
The classic Lotka-Volterra model, based on competition coefficients, examines species that
compete directly for resources, and shows that the species can coexist if each has resources that
other species cannot access regardless of their numbers [2]. A second important group of mod-
els involves NDD interactions, in which a selective advantage to species that are locally rare
switches to a selective disadvantage when those species become locally common. NDD can
lead to multiple stable internal density equilibria that permit numerous species to occupy the
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
3 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
same ecosystem [3–6]. Possible mechanisms for NDD effects can include species interactions
with both physical and biological factors.
Many general ecological theories seeking to explain the distributions of species that occupy
the same or similar trophic levels have tended to gloss over such complexities. In 2010, McGill
[16] surveyed six “unified theories of biodiversity,” all of which had shown success in predict-
ing observed species abundance distributions and species-area curves at scales of 100 m and
above. He showed that all these theories employ three assumptions: intraspecific clumping,
intraspecific variation in global abundance, and—most importantly for the present study—
spatial independence of the distributions of different coexisting species.
Darwin had postulated that species populating a multitude of trophic levels in an ecosystem
are continuously interacting, and that these interactions contribute to evolutionary divergence.
Given the growing evidence for such interactions (see [17] for an extreme example), it is sur-
prising that apparently successful general ecological theories can be constructed using the
assumption that species-species interactions are irrelevant to the overall structure of communi-
ties. General unified theories of ecosystems may indeed be congruent with the distributions of
component species that happen to be easily countable, provided that the scale is 100 m and
above. But they incorporate no information about the existence of biotic and abiotic interac-
tions at smaller scales, which is where most between-species interactions are likely to take place.
How common, how complex, and how significant in their effects are the fine-structure
between-species interactions that Darwin postulated? Can an understanding of these interac-
tions lead to more complete theories that underlie ecosystem structures and their evolutionary
trajectories? Here we test the spatial independence assumption that McGill shows is basic to
the most general ecological distribution theories. We examine sixteen multiply-censused forest
diversity plots (FDPs) that are scattered over a wide variety of biogeographic regions (Table 1),
and find that the assumption does not hold at the scale of meters. We also show that the pat-
tern of departures from independence can reveal new information about between-species
interactions.
We use the Equal-Area Annulus (EAA) [18] point-pattern method to visualize and quantify
non-random patterns of tree clustering, distributions of tree recruitment and mortality, and
the influence of surrounding trees on tree growth. We show details of how these interactions
occur not only between conspecifics, where they are well-known [5, 13, 19, 20], but also
between species that are separated across a wide range of phylogenetic distances. We show
that, although the interactions decrease smoothly in significance with increasing physical dis-
tance between trees, they exhibit complex relationships with phylogenetic distance that are
unique to each of the study’s FDPs. Such complexities are not dealt with in the global theories
examined by McGill.
Because of the many differences between EAA and the commonly-used regression-based
methods that are used to detect NDD effects, and because of the many modifications that have
been made to EAA since its original publication (subheads 1–8 in the Modifications to the
Original Method section), we have chosen to place the extensive Materials and Methods sec-
tion immediately following this introduction. We emphasize how the EAA method avoids the
statistical biases [21] that are inherent in regression-based methods. The "regression dilution"
issue flagged by that paper is not a problem in our analyses because a bias towards zero makes
the EAA method less—not more—likely to detect NDD (i.e. errors in predictors would reduce
the statistical power of the method instead of increasing the Type I error rate).
The EAA method also incorporates a number of desirable features that ideally should be
exhibited by methods designed to detect density-dependent effects, as discussed in a recent
review [22]. A method should (1) measure the relative magnitudes of conspecific NDD and
heterospecific NDD and how they vary among different between-species interactions, (2)
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
4 / 33
PLOS COMPUTATIONAL BIOLOGY Table 1. Some characteristics of the FDPs examined in this paper, arranged by latitude.
FDP
Dim (m) No. of census
intervals
Total of species
recorded
Avg. tree density (trees/m2)
& no. annuli used
Annual Rainfall
(mm)
Latitude/
Longitude
Min/max or avg.
temp (oC)
Each of 16 forests shows a unique pattern of between-tree interactions
Pasoh (Peninsular
Malaysia)
Lambir (Sarawak,
Malaysia)
Korup (Cameroon)
1000 X
500
1040x500
1000 x
500
Sinharaja (Sri Lanka)
500 x 500
Barro Colorado Island
(Panama)
Mudumalai (India)
1000 x
500
1000 x
500
Palanan (Philippines)
400 x 400
Luquillo (Puerto Rico)
320x500
Nonggang
500 x 300
Heishiding
Fushan
1000 x
500
500 x 500
Gutianshan
600 x 400
Tiantong
500 x 400
Changbaishan
500 x 500
Wind River (US)
800 x 340
Wytham Woods (UK)
300x600
5
2
2
2
6
898
1180
494
239
320
2 (4-yr intervals)
76
3
4
1
1
1
2
1
2
1
2
319
163
217
245
111
159
154
52
26
24
https://doi.org/10.1371/journal.pcbi.1008853.t001
0.671
20
0.665
10
0.656
10
0.829
10
0.470
20
0.035
10
0.210
10
0.289
10
0.453
10
0.546
10
0.463
10
0.586
10
0.604
10
0.155
10
0.116
10
0.112
10
2000
2700
2.98N/102.3E
25.8/28.3
4.2N/114E
31.4/22.1
5500 (seasonal)
5.1N/8.9E
22.7/30.6
5000
6.4N/80.4E
20.4/24.7
2600 (seasonal)
9.15N/79.85W 23/32
1300 (seasonal)
11.6N/76.5E
16.4/27.4
3200 (typhoons)
17.0N/122.4E
26.1
3500 (hurricanes)
18.3N/65.8W 23.0
1300 (seasonal)
22.4N/107.0E
19.0/27.2
1700 (seasonal)
23.3N/111.5E
10.6/28.4
4300 (typhoons)
24.8N/121.6E
18.2
2000
29.1N/118.1E
4.3/27.9
5000 (some
typhoons)
700
29.8N/121.8E
16.2
42.4N/128.1E
3.6
2300 (seasonal)
45.8N/122W
-2/27
700
51.8N/1.34W 10
evaluate the relative roles of conspecific and heterospecific NDD in the maintenance of eco-
logical diversity, (3) remove the biases inherent in statistical methods that do not compare the
actual data with appropriate null models, (4) distinguish the relative sizes of the contributions
to NDD of species with different abundances and life histories, and the contributions of bio-
geographic factors such as latitude and rainfall, (5) follow the contributions of organisms at
different stages in their life histories, (6) provide a route for further examination of the details
of the NDD mechanisms themselves and the long-term evolutionary implications of these
mechanisms.
Unlike regression-based methods, EAA uses null models that isolate the variables being
examined while leaving all the other properties of these extensive data sets unchanged. This
enables us to present details of the species-species interactions with high statistical confidence.
For clarity, we provide a step-by-step illustration of an EAA analysis (Fig 1). We also provide an
example showing that the method is highly sensitive to small deliberately-introduced changes in
the FDP data (subhead 11 of the Modifications to the Original Method section and Fig 2).
In the Results section, we present EAA analyses of all sixteen FDPs (Figs 3–7), and examine
in detail the causes of the heterospecific interaction peaks and valleys that are observed at the
temperate Wind River (Washington State, USA) FDP (Fig 8). We show how EAA analysis
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
5 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Fig 1. An overview of a typical EAA analysis. At top is a diagram of a large focal tree in the Lambir (Malaysia) FDP, surrounded by 10
concentric annuli each of area 50 m2. For simplicity, the trees shown in the diagram are only a sample of some of the annular surviving trees
(S), recruits (R), and trees that die (D). In the diagram the trees shown are sampled from among the trees that lie at zero or at 185 Ma
(mega-annum) phylogenetic distances from their LCA with the focal tree, although of course all the trees in the annuli are used in the entire
EAA analysis. Generalized Additive Model (GAM) fits to patterns of clustering of surviving annular trees, using data from the closest and
the furthest annulus, are shown in the two-dimensional graphs on the right of the diagram. In this analysis, the observed annular clustering
is presented as deviations (z-values) from a null model expectation for four quantiles of focal tree diameter. The null model is generated by
repeated shuffling of the focal tree diameters within species, so that any positive z-values for some of the focal-annular size quantiles must be
balanced by negative values for others. Such positive-negative balances are expected in analyses of recruitment, clustering and mortality, but
not in analyses of growth (Materials and Methods). The 95% confidence intervals of the GAM curves are shown in gray. Brown horizontal
lines show the 95% confidence intervals around zero z-values. To help orient the viewer, gray arrows connect some of the closely-related
and distantly-related survivors in the diagram to the places at which their data contributes to the largest-quantile focal tree lines (red) on the
two-dimensional graphs. Each data point in the 2D graphs represents the difference between the actual and the null-model data for all focal
trees in a given census period that have annular trees within a specific phylogenetic range. The null-model data have been generated by
repeated shuffling of focal tree properties (size or growth rate) within species. A new point is generated for each of the ten replicates of the
actual-null comparisons and for each of the census periods at the FDP. The points in the graphs form clusters because, with each replicate,
species pairs separated by similar phylogenetic distances are shuffled at random in order to fill each of the phylogenetic distance quantiles.
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
6 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Thus, each of the gray arrows that shows the contribution of an individual tree simply shows where the tiny amount of information that is
contributed by that tree’s focal-annular interactions ends up in the data points in the graphs. The three-dimensional graphs show GAM fits
of the largest-quantile (graph A) and smallest-quantile (graph B) focal tree size data across all ten annuli. Regions of the surfaces that lie
within the range of non-significant z-values along the z-axis are gray; those that lie outside this range, and that therefore represent
significant z-values, are colored. The colors start with green and shade through blue as the significance of the positive or negative z-values
increases. The orientations of the three-dimensional graphs presented here sometimes differ, in order to reveal details of the surfaces. As
with the lines on the two-dimensional graphs, the 3D surfaces themselves have confidence intervals, but the confidence intervals are not
shown here. Typical confidence intervals on the 3D surfaces, which tend to be small, are visualized more easily if these three-dimensional
graphs can be rotated by the viewer. A sampling of such rotatable graphs is presented as html files in S3–S11 Figs.
https://doi.org/10.1371/journal.pcbi.1008853.g001
provides new details of the role played by strong allelopathic effects of a gymnosperm, the
western hemlock Tsuga heterophylla, on some but not all of the angiosperms in the FDP to
which it is very distantly related [23]. This example demonstrates the potential of EAA to
examine the roles of heterospecific interactions that may involve a wide variety of tree charac-
teristics and microenvironmental factors. EAA provides a tool to measure the sizes of the con-
tributions of these variables to species distributions and life history patterns. Our preliminary
findings show that EAA results and the experiments that they suggest will help to pinpoint
unusually significant interactions that can be investigated further through field observations
and experiments. These findings will in turn enable us to unravel the true complexity and the
evolutionary histories of the many between-species interactions that, as Darwin had believed,
“cannot be disputed.”
Materials and methods
FDP data
This paper surveys demographic and tree-distribution data from 16 repeatedly-censused forest
dynamics plots (FDPs) (https://forestgeo.si.edu/sites-all). The FDPs have been established in
locations ranging from tropical to high-temperate latitudes, and encompass a wide range of
seasonal and non-seasonal rainfall patterns (Table 1). Repeated censuses of the FDPs include
all trees present during each census that have a diameter of 1 cm or greater at a height of 1.3 m.
The EAA method
The EAA method [18] examines interactions between "focal" trees, made up of all the surviving
trees during a census period in the FDP, and the "annular" trees that occupy successive concen-
tric annuli of equal area around the focal trees. The use of these successive annuli, which each
consist of similar amounts of data that can be analyzed with the same statistical power, permits
unbiased statistical comparisons of the significance of interactions over a range of physical
focal-annular tree distances.
EAA draws on many previous studies that have employed quadrat or point pattern analysis,
coupled with a null modeling or Monte Carlo approach to generating control distributions in
which specific components of the data have been randomized [11, 24–26]. The EAA method is
an extension of neighborhood density functions such as the O-ring spatial statistic [26] and the
Dx index [27]. EAA is similar to bivariate mark correlation analyses [26, 28]. The r-mark func-
tion is a non-parametric estimator of the response of the growth of small trees to the presence
of a large tree at distance r. Point pattern methods are not based on regression analysis, like
many of the methods that have been used to search for positive or negative density-dependent
effects in forest data. These regression-based methods have recently been criticized as suscepti-
ble to over- or underestimation of the magnitudes of the effects that are being searched for [21].
EAA, in contrast, compares the distributions of the actual data with distributions in null models
in which only the variable or variables of interest are repeatedly randomized and all other
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
7 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
parameters of the FDPs are left untouched. Each iteration of the null model is analyzed in the
same way as the real data, and the entire distribution of these null model results is used to test
the difference between the real and randomized data. In each of these iterated replicates of the
null-model data, any heteroskedasticity of the distributions of within-species focal tree growth
rates and focal tree sizes is left unaltered and therefore cannot bias the results.
A diagrammatic example of EAA analysis
Fig 1 shows, in diagrammatic form, the steps of a typical EAA analysis, in this case the relation-
ship between focal tree size and clustering of annular trees in the Lambir FDP (Sarawak, Bor-
nean Malaysia).
Overview of the EAA tests
Table 2 lists the current EAA tests, the focal and annular tree properties that each test exam-
ines, and the expected results if focal-annular NDD interactions are present. Each of the tests
is carried out as illustrated in Fig 1. Additional details of each test, including details of the null
models used, are given below.
Modifications of the original EAA method
Overview of the modifications
The EAA method has been redesigned since its initial publication. In addition to the use of
equal-area annuli surrounding focal trees, the method now divides sets of FDP data into equal
Table 2. Summary of the properties of, and expectations for, the EAA analyses that are used in this paper. The expected results for each test are for NDD focal-annu-
lar interactions; the PDD expectation is the opposite.
Focal tree properties
Annular tree properties
Null Model Comparison
Expected Results
Focal-annular properties employed, and expectations shared by all tests:
Species, position, diameter, growth
rate, recruitment and mortality for
all trees in each census period
For each annulus: species, basal area,
distance from focal tree in Ma, whether
trees were recruited or died during census
period
A single focal-tree attribute is
repeatedly shuffled within
species to serve as a control
Focal-annular differences from null model
should decline in significance as either physical
or phylogenetic focal-annular distance
increases
Properties examined and expect results for each null model comparison test, assuming NDD focal-annular interactions:
Test 1) Relationship between focal survivor sizes and their annular survivor summed basal area
Focal survivor size
Summed basal area of annular tree survivors
that fall within a given quantile of
phylogenetic distance from focal tree
Shuffle focal tree survivor sizes
within species
Negative relationship between focal tree size
and annular survivor summed basal area
Test 2) Relationship between focal survivor sizes and their annular recruit fraction
Focal survivor size
Fraction of trees in the annulus and
phylogenetic distance quantile that recruit
Shuffle focal tree survivor sizes
within species
Negative relationship between focal tree size
and annular recruit fraction
Test 3) Relationship between focal survivor sizes and their annular mortality fraction
Focal survivor size
Fraction of trees in the annulus and
phylogenetic distance quantile that die
Shuffle focal tree survivor sizes
within species
Positive relationship between focal tree size
and fraction of annular trees that die
Test 4) Relationship between a focal tree’s growth rate (normalized within species) and its annular tree basal area
Focal survivor growth rate
Summed annular survivor basal area in the
phylogenetic distance quantile
Shuffle normalized focal tree
survivor growth rates within
species
Negative relationship between focal tree
growth rate and annular tree summed basal
area
Test 5) Differences between focal trees that do and do not recruit and their annular recruit fractions
Focal recruits vs. other focal trees
Fraction of trees in the annulus and
phylogenetic distance quantile that recruit
Shuffle properties of all focal
trees within species
Higher fraction of annular recruits around
focal recruits
Test 6) Differences between focal trees that do and do not die and their annular mortality fractions
Focal trees that die (separated into
small and large) vs. other focal trees
Fraction of trees in the annulus and
phylogenetic distance quantile that die
Shuffle properties of all focal
trees within species
Higher fraction of annular trees that die
around focal trees that die
https://doi.org/10.1371/journal.pcbi.1008853.t002
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
8 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
quantiles, so that the statistical power of the analyses of all the subdivisions of the data in a
given FDP are equivalent. Two- and three-dimensional GAM curves are now fit to the data.
These curves reveal details and significance of effects traceable to phylogenetic distances
between species. A uniform method of estimating phylogenetic distances between species is
applied to all the FDPs.
Equations that quantify the analyses of focal-annular NDD-influenced effects are derived in
[18]. Significances of the effects, compared to the mean of 1,000 iterations of the null models,
are estimated using z-scores. The GAM analyses we employ here use spline-based smooth
terms [29]. GAM fits of curves to the data use the formula y ~s(x, k = k-value), where s estab-
lishes the parameters of spline-based smooth terms and the k-value is the number of smooth
terms employed. The optimal k-value is determined using gam.check [30]. Edge-effect correc-
tions have been applied to all the FDPs [31]. The z-values obtained by all the analyses are
adjusted for the discovery of false positives, using the Benjamini-Hochberg correction for
independent statistics [32]. The data used to generate the GAM graphs are given in Supporting
Information compressed data files S1 and S2 Datas.
Division of the data into quantiles
Because we are comparing different FDPs that exhibit a wide range of species numbers, tree
size distributions and densities, plot sizes, and distributions of phylogenetic distances between
species, we have introduced standardized protocols for subdividing the data. Our goal is to
ensure that each of the subdivisions of a set of data are of approximately equal size, so that the
statistical power of the EAA tests remains the same across successive concentric annuli, phylo-
genetic distance intervals, and subdivisions of the focal and annular trees.
Annuli vary from 5 to 20 in number in the different FDPs, depending on overall tree
density, but the total area around each focal tree encompassed by the annuli is 500 m2 in all
FDPs. Focal tree diameters at the start of each census period for each species are divided
into four equal quantiles. Totals of annular tree biomasses in each of the annuli (approxi-
mated by the sum of the areas at standardized height) are divided into five equal quantiles.
Phylogenetic distances between species are also divided into quantiles, but this division
poses special problems. First, the amount of data varies among FDPs. Therefore, in order to
ensure that there is sufficient data for analysis, we have used different numbers of phyloge-
netic distance quantiles in different FDPs. We have been able to use as many as 20 quantiles
in large, species-rich FDPs such as BCI and Pasoh, but have been limited to as few as 5
quantiles in smaller, less speciose FDPs such as Wind River and Wytham Woods. Second,
because the distribution of pairwise phylogenetic distances between species is different in
each FDP, and these distributions are far from uniform, subdivision of quantile differences
often means that many species pairs that are separated by the same or similar phylogenetic
distances will fall into different quantiles. Each analysis for each census period, therefore, is
repeated ten times, each with 100 iterations of the null model. With each repetition, the
order within each set of focal-annular pairs that share the same phylogenetic distance value
is shuffled. This ensures that if large numbers of species pairs in an FDP share the same phy-
logenetic distance, subdivision into quantiles in the replicated analyses will have placed dif-
ferent random mixes of these pairs in adjacent quantiles.
Measurement of effect of annular trees on focal tree growth rates
We define focal survivors as trees that are present at the beginning and end of a five-year cen-
sus interval. For each data graph, data from all focal trees of all species are pooled. Normalized
focal tree growth rates are calculated for each FDP as standard deviations from the mean of a
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
9 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
given decile of diameters of the focal trees of a given species within a census period. This
approach avoids the confounding effects of tree size differences, species differences, and secu-
lar trends over time on the growth rates of focal trees.
Summed annular tree basal areas are estimated as summed area at "breast height," the sum
of the trees’ cross-sectional areas at a height of 1.30 m. The areas of multi-stemmed trees are
summed. Basal areas and focal trees are each divided into five size quantiles. In the data pre-
sented in this paper, only the annular tree growth effects on the smallest size quintile of focal
trees are examined, though as reported earlier there is a smaller but often significant negative
effect of annular trees’ summed basal area on the growth of larger focal trees [18].
In these analyses, as in the analyses of recruitment, mortality, and annular tree clustering,
each annulus is examined separately. Thus, an analysis of fifth-annulus annular tree effects on
focal tree growth begins by examining all focal trees of the smallest diameter quintile. These
focal trees either have, or do not have, annular trees in their fifth annulus that fall within a
given quantile of Ma values to their last common ancestor (LCA) with the focal tree. Focal
trees that have no such annular trees in their fifth annulus, regardless of whether or not they
have such trees in their other annuli, form the comparison group. The focal trees that have
such annular trees in their fifth annulus are divided into the five quintiles of summed annular
tree basal areas and their growth rates are compared to those of the controls.
The null model used for comparison is generated by repeated randomization, within spe-
cies, of the growth rates of the focal trees.
Annulus number and the number of phylogenetic distance quantiles used in each FDP
analysis are adjusted to ensure that within each of the concentric annuli there will be a substan-
tial number of such control annuli.
Focal-annular phylogenetic distances
We estimated DNA-based divergences times between focal and annular species (last com-
mon ancestor (LCA) in mega-anna (Ma)) using the S.PhyloMaker program written by YJ
(available at https://github.com/jinyizju/S.PhyloMaker). Table 3 presents the proportion of
species in each FDP that are present in S.PhyloMaker’s Phytophylo DNA dataset. These
proportions vary from 100% at Wind River to 14% at Sinharaja. The majority of phyloge-
netic distances between species must therefore be estimated at the genus rather than the
species level in most of the FDPs. In this study the estimation was made by using Scenario
2 of S.PhyloMaker. For the species for which only genus-level information is known, this
scenario picks uniformly-distributed distances from the interval from the present back to
its genus’ LCA.
There is unavoidably some noise in the phylogenetic distances, especially at FDPs such
as Lambir and Pasoh where few species have been characterized genetically (Table 3). Fur-
ther, and unavoidably, we are forced to add more noise because we divide the pairwise dis-
tances to the LCA into Ma interval quantiles that vary in number according to the amount
of information in the FDP. This division has the advantage that it equalizes the amount of
information in each quantile, but the disadvantage that it can add further noise to focal and
annular species that are separated by pairwise distances that fall in a sparsely population
region of the range of pairwise distances at the FDP. The noise problem can be overcome to
some extent because we repeatedly sample the pairwise distances during the analysis. We
have concluded that the advantage of having equal statistical power in each of the pairwise
distance quantiles outweighs the disadvantage of introduced noise. This is because, when
we have exhaustively analyzed FDPs more than once using this methodology, the results are
essentially indistinguishable.
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
10 / 33
PLOS COMPUTATIONAL BIOLOGY Table 3. Numbers and fraction of species in each of the FDPs in this study that are found in Phytophylo.
Each of 16 forests shows a unique pattern of between-tree interactions
Species Found in Phytophylo
Species Not Found in Phytophylo
Fraction of Total Species found in Phytophyo
FDP
BCI
Changbaishan
Fushan
Gutianshan
Heishiding
Korup
Lambir
Luquillo
Mudumalai
Nonggang
Palanan
Pasoh
Sinharaja
Tiantong
Windriver
Wytham
https://doi.org/10.1371/journal.pcbi.1008853.t003
266
26
45
117
132
117
209
130
33
82
49
190
33
105
17
19
54
26
66
42
166
390
1127
32
50
135
267
708
206
51
0
5
0.8313
0.5
0.4054
0.7358
0.443
0.2308
0.1564
0.8025
0.3976
0.3779
0.1551
0.2116
0.1381
0.6731
1
0.7917
Null models for clustering, recruitment and mortality
In order to isolate the influence of focal tree diameter on annular tree properties, the
observed distribution of clustering, recruitment or mortality of annular trees around differ-
ent diameter classes of focal trees is compared with 1,000 iterations in which focal tree
diameters are randomized within species within census intervals in an FDP which is other-
wise identical to the original FDP. Thus, these null models leave all other properties of the
FDPs intact, including the positions and species identifications of all of the annular trees
and the distributions of sizes of each species of focal tree. The only real-data components of
clustering, recruitment and mortality that are measured are in the form of z-values of differ-
ences between the real data and the means of the Monte Carlo randomizations of focal tree
diameters within species within census periods. This avoids the introduction of possible
unknown variables, which is a problem with the regression analyses that are commonly
used to search for density-dependent effects on recruitment and mortality [21]. Regression
analyses search for differences in the properties of trees that surround trees that recruit or
those that die, compared with those surrounding survivors, but the resulting regressions
may have many sources traceable to the distributions of tree positions and tree properties
that will vary across species.
Null models and the influence of storage effects
There is a built-in “delay” in NDD-influenced factors that affect recruitment, mortality and
clustering. This delay results from spatial and temporal storage effects [4, 6], allowing relatively
dense clusters of trees of the same or phylogenetically related species to become established in
regions where physical resources are initially plentiful and where species-specific pathogens,
browsers, and seed-predators are initially few. As the trees in the clusters grow older and
larger, their species-specific pathogens begin to accumulate, browsers and seed-predators
become increasingly attracted to the area, and species-specific physical resources become lim-
iting [33]. As a consequence, the trees in the clusters thin out over time, so that new clusters of
saplings of the same species as those in the clusters can only be established elsewhere. By
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
11 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
combining the original Janzen-Connell theory with spatial and temporal storage theory, it has
been possible to explain the apparently contradictions between the Janzen-Connell model and
the fact that many species of tree in forests are clustered rather than overdispersed [4, 6, 11].
Thus, null models that randomize only focal tree size and leave annular tree clustering or over-
dispersal intact are central to the EAA analyses, because these storage effects are the same in
both the real and the randomized data.
NDD- influenced focal-annular interactions
Many focal-annular interactions can be examined by the EAA method. We chose the interac-
tions, listed below, that were shown in our previous study [18] to have an NDD component
that remains significant across a wide range of focal-annular phylogenetic distances.
The effect of focal tree size on annular tree recruitment. NDD predicts that recruitment
of annular trees at any phylogenetic distance from the focal tree should tend to be highest
around small focal trees and diminish as the focal trees increase in size. Conditions favoring
annular recruitment result from the accumulation of NDD effects of pathogens and parasites
shared between focal and annular trees [8, 9], and from the depletion of shared physical and
biologically-generated resources (niche-complementarity) [34, 35]. As focal trees grow, as
shared pathogens accumulate, and as shared resources become scarcer, annular recruitment
should diminish [3, 8, 9, 19]. Seedling data are not available for these data sets, and we there-
fore use the fraction of annular trees that have achieved a diameter of 1 cm during a census
period as a proxy for recruitment rates.
The effect of focal tree size on annular tree mortality. The same NDD processes should
result in low mortality among the annular trees that surround small focal trees and high mor-
tality among annular trees that surround large focal trees [8, 9, 36].
The effect of focal tree size on annular tree clustering. The combination of NDD effects
influencing recruitment and mortality should over time result in high summed basal area of
annular trees around small focal trees and lower summed basal area of annular trees around
large focal trees, again across a wide range of focal-annular phylogenetic separations.
The effect of annular tree basal area on focal tree growth.
If a tree’s growth rate is slo-
wed by the effects of competition for physical resources with nearby conspecific or phylogenet-
ically related annular trees, or through the sharing of pathogens and predators with these
annular trees, there may be a negative impact on its fitness [37]. Trees growing in regions that
have few related trees nearby exhibit a fitness advantage over those growing in regions where
there are many related trees nearby [24].
As noted above, and as in the original EAA analyses [18], focal tree growth rates are nor-
malized within species, censuses, and focal tree size classes.
Additional analyses
Other analyses based on comparisons of the actual data with randomized null models may be
applied to these data. For example, the focal trees that recruit or die during a census period can
be examined to determine the fraction of their annular trees that are also recruits or trees that
die (tests 5 and 6 of Table 2). The null models for these tests determine the same ratios after
the positions of the focal trees that recruit, survive and die are repeatedly shuffled within spe-
cies and the properties of the annular trees are left untouched. These tests, however, only
examine how interactions between focal trees that recruit or die and their annular trees that
recruit or die, which are small fractions of all the focal-annular interactions, differ from those
of the remainder of the focal-annular interactions.
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
12 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Interpretation of the graphical results
In interpreting the graphs of tests for these classes of focal-annular interactions, recall that in
EAA analyses the z-values for the annular tree properties express differences from null-model
data sets that are randomized only with respect to focal tree size. The z-values from the four quan-
tiles of focal tree sizes are therefore symmetrically distributed around the mean of all the z-values.
Unlike the results for clustering, recruitment and mortality, the growth results are not sym-
metrically distributed around zero. Instead, each line shows the difference in z-scores between
(a) the normalized growth rates of small focal trees that have, in a specified annulus, a specified
summed basal area of annular trees within a specified quantile of phylogenetic distances from
the focal tree, and (b) the normalized growth rates of small focal trees with no such annular
trees in the specified annulus. In general, throughout the FDPs, the larger the summed basal
area of the specified subset of annular trees, and the closer to the focal tree they are in either
physical or phylogenetic distance, the greater the expected negative effect of the annular trees
on normalized focal tree growth.
A test for a possible relationship between tree size and patterns of tree
mortality in this study
Mortality varies across life cycle, with NDD effects being most pronounced in the smallest
trees [38]. NDD patterns of seedling mortality are primarily mediated by fungal pathogens [20,
39]. Mammals, foliar herbivores and foliar pathogens tend to contribute little to mortality at
these early stages [40]. Phylogenetic distance is known to play an important role in the inci-
dence of seedling mortality, which decreases as the phylogenetic distance between focal and
surrounding trees increases [41–43].
The smallest trees for which mortality is measured in the present study have a dbh of 1 cm.
We hypothesized that such small trees are more likely to die if they are near large focal trees,
and less likely to die if they are near small focal trees. Larger trees that die might show a weaker
association with focal tree size, because there might be a greater influence on the mortality of
large trees of factors such as wind, fire and large herbivores that may have a small NDD com-
ponent. To test this possibility, we divided annular trees of each species that died into two
equal-sized groups, designated small and large.
A control manipulation of the BCI data to check the sensitivity of the EAA
analyses
Fig 2 below shows that focal-annular phylogenetic curves undergo the expected changes when
FDP data are deliberately manipulated. EAA analyses are therefore highly sensitive to small
differences in the data sets.
A pronounced reduction in significance in the BCI FDP data (marked with a circled
numeral 1 at slightly over 100 Ma focal-annular phylogenetic distance in Fig 2 below) indicates
that focal-annular species pairs separated by this phylogenetic distance are only interacting at
low levels. This “valley” in significance values is seen clearly in focal-annular clustering,
recruitment and the mortality of large annular trees. It is not apparent in mortality of small
annular trees, or in focal tree growth.
To investigate the validity of this signal, we determined the set of phylogenetic distances
between pairs of species at this FDP that lie between 108 and 117 Ma, and selected the interac-
tions involving the three commonest species that are separated by these distances for our sensi-
tivity test. Our reasoning was that, because these species are common in the plot, they are
likely to contribute disproportionately to between-species interactions. The three, in
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
13 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Fig 2. Test of the sensitivity of EAA analysis. The circled 1 and 2 mark unusual valleys and peaks respectively in the
significance of focal-annular interactions. Asterisks mark the appearance of a new “valley” in significance levels after
deliberate manipulation of the data (see text). Other legends as in Fig 1.
https://doi.org/10.1371/journal.pcbi.1008853.g002
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
14 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
descending order of abundance, are Trichilia tuberculata (Meliaceae) (5.0% of the stems),
Mouriri myrtilloides (Melastomataceae) (3.4%), and Tetragastris panamensis (Burseraceae)
(2.5%). The focal-annular phylogenetic distances to their LCA that fall in this window are T.
tuberculata with M. myrtilloides (111 Ma) and T. panamensis with M. myrtilloides (111 Ma),
while T. tuberculata with T. panamensis (72 Ma) falls outside the window.
We modified each of the focal-annular phylogenetic distances separating these three species
to 50 Ma, and repeated the EAA analysis. The right-hand set of graphs in Fig 2 show the
appearance of a new region of lowered significance (indicated with an asterisk) close to the 50
Ma mark for the clustering, recruitment and large-tree mortality graphs. The new lowered-sig-
nificance signal does not appear precisely at 50 Ma, because these altered between-species dis-
tances now join a large group of other distances that contribute to the phylogenetic distance
quantile that includes the 50 Ma distance. The original “valley” in significance values remains,
though somewhat reduced in size, demonstrating that additional species pairs contribute to it.
The growth graph shows no new signal, however, suggesting that the causes of the reduction
in between-species interactions do not influence focal tree growth. Selective removals of spe-
cies or focal-annular data from the data sets, such as those which were carried out at Wind
River that are presented below, are therefore capable of providing consistent and detailed
information about the extent of individual between-species interactions.
Results
Each of the 16 forest diversity plots (FDPs) examined in this study exhibits smooth declines
across increasing physical focal-annular distance (in m) in the significance of each of the four
focal-annular interactions tested. There is one exception: recruitment at Mudumalai shows com-
plexity across physical distance, possibly because of recent influence of elephants and/or fires (S1
Fig). Clustering at Mudumalai, which measures the results of longer-term processes, shares with
the other FDPs the common pattern of a smooth decline with increasing physical distance.
These interactions also tend to show an overall decline in significance with increasing phy-
logenetic distances (in Ma) between the species, but there are many localized exceptions to this
decline that result in unique patterns of “peaks” and “valleys” in the magnitude of significance
along the phylogenetic distance axis for each FDP.
Table 4 summarizes the expected (from Table 2) and the observed results from the four
EAA tests employed in this paper. In general, the results agree with NDD expectation, but
there are many localized departures from a smooth decrease in significance with phylogenetic
distance. In addition, there is a puzzling weak PDD signal seen for small annular trees that die
and that warrants further investigation.
In Fig 3 we present the results, for each of the first four NDD-influenced focal-annular
interactions that are listed in Table 2, at the BCI (Panama) FDP. The results are presented as
two- and three-dimensional graphs generated from generalized additive model (GAM) analy-
ses. The three-dimensional graphs in the figure present data from GAM analyses along both
the physical (focal-annular distance in meters) and phylogenetic (Ma back to the last common
ancestor (LCA)) axes, permitting a comparison of the effects of physical and phylogenetic dis-
tance in a single graph. Note that in these 3D graphs the physical distance decline remains
smooth across all concentric annuli, while the irregularities in the phylogenetic distance curve
are preserved across all concentric annuli. The smooth and gradual decline in significance
with increasing physical focal-annular distance is clearly distinguishable from the complex
fluctuations in significance that are seen across the range of phylogenetic distances. Figs S1
and S2 show that these patterns are seen in 3D analyses across FDPs, with the exception of
recruitment at Mudumalai that was noted earlier.
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
15 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Table 4. Expected and observed results for the EAA tests, assuming NDD focal-annular interactions.
Expected (from Table 2)
Observed
Results expected and observed in all tests:
Focal-annular differences from null model should
decline in significance as either physical or phylogenetic
focal-annular distance increases
Smooth decline in significance with increasing physical
distance. A general decline as phylogenetic distance
increases, but many FDPs show significant peaks and
valleys
Results expected and observed in each test, given NDD expectation:
Test 1) Relationship between focal survivor sizes and their annular survivor summed basal area
Negative relationship between focal tree size and
annular survivor summed basal area
Generally negative, as predicted, but there are some
localized switches into PDD (positive) territory in some
FDPs
Test 2) Relationship between focal survivor sizes and their annular recruit fraction
Negative relationship between focal tree size and
annular recruit numbers
Generally negative, as predicted. Peaks and valleys along
the phylogenetic axis often match annular survivor
curves
Test 3) Relationship between focal survivor sizes and their annular mortality fraction
Positive relationship between focal tree size and
numbers of annular trees that die
Positive, as predicted for large annular trees that die, but
weakly negative for small annular trees that die
Test 4) Relationship between a focal tree’s growth rate (normalized within species) and its annular tree basal
area
Negative relationship between focal tree growth rate
and annular tree summed basal area
Negative relationship, as predicted. Significance falls
with decreasing annular summed basal area, with many
significant peaks and valleys along the phylogenetic axis
Test 5) Relationship between focal trees that do and that do not recruit. and their annular recruit fractions
Higher fraction of annular recruits around focal recruits At BCI, a higher fraction, as predicted. Peaks and valleys
along the phylogenetic axis do not match those for
surviving focal trees
Test 6) Relationship between focal trees that do and do not die and their annular mortality fractions
Higher fraction of annular trees that die around focal
trees that die
At BCI, a higher fraction, as predicted, for both small
and large focal trees that die
https://doi.org/10.1371/journal.pcbi.1008853.t004
Fig 4 shows phylogenetic distance results from a preliminary application of Tests 5 and 6 to
the data from BCI.
The results are highly significant and in agreement with NDD expectation (Table 4, Tests 5
and 6). As with the other analyses, there is a smooth decline with increasing focal-annular physical
distance (not shown), and a complex decline with increasing phylogenetic distance. The curves
for mortality are in agreement with those shown for large annular tree mortality in Fig 3. How-
ever, the phylogenetic significance decline for recruitment appears to show a different pattern of
hills and valleys from the equivalent analysis in Fig 3, suggesting that focal-annular interactions of
focal trees that recruit or die with annular trees that recruit or die may be different from those of
focal survivors. In addition, the phylogenetic peaks and valleys for recruitment are less complex
than those found in the Fig 3 analysis, possibly because in Test 5 data are being examined from a
smaller fraction of the focal trees. As these and other tests are explored further, they will provide
additional windows of opportunity for investigation of detailed focal-annular interactions.
First-annulus two-dimensional graphs for Tests 1–4 in the remaining fifteen FDPs, which
illustrate each FDP’s unique phylogenetic distance patterns, are shown in Figs 5, 6 and 7.
NDD-influenced focal-annular interactions for clustering and recruitment are present in all of
the FDPs, and NDD-influenced focal tree growth interactions with the summed basal areas of
annular trees are also found in 15 of the FDPs.
Three-dimensional GAM graphs for clustering, recruitment and focal growth in all of the
FDPs are shown in S1and S2 Figs. These three-dimensional graphs show that, with the
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
16 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Fig 3. A comparison of two- and three-dimensional graphs of the z-values for NDD-influenced focal-annular
patterns at the BCI (Panama) FDP. The 2D graphs show first-annulus z-values of these patterns for all subdivisions
of focal tree sizes and annular tree biomasses, compared to null models (Materials and Methods). The 3D graphs show
the surfaces formed by the z-values across both physical and phylogenetic distances between focal and annular trees.
The surfaces shown are for the largest focal trees or annular biomasses (red lines in the 2D graphs) and the smallest
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
17 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
focal trees or annular biomasses (blue lines in the 2D graphs). Circled numerals 1 and 2 mark unusually low or high z-
values that are found at certain focal-annular phylogenetic distances. The k-values are the optimized number of
smooth terms used in the GAM analyses (Materials and Methods). The 95% confidence limits for the expected 2D z-
values of zero are shown as brown horizontal lines, and the 95% confidence intervals of the 2D lines are shown in gray.
The regions of significant z-values on the 3D surfaces are heat-map colored as in Fig 1, and the non-significant regions
are gray.
https://doi.org/10.1371/journal.pcbi.1008853.g003
exception of recruitment at Mudumalai, the significance of NDD-related effects declines
smoothly rather than irregularly with increasing physical distance, while the patterns of peaks
and valleys along the phylogenetic axis that are unique to each FDP are preserved across the
range of physical focal-annular distances. These idiosyncratic phylogenetic distance patterns
show that the species in each of the FDPs have in the past been shaped by distinct evolutionary
trajectories that have led to this wide variety of patterns of species-species interactions.
Some of the exceptions to a smooth decline in significance with increasing focal-annular phylo-
genetic distance are observed at the same phylogenetic distances of an FDP’s recruitment, cluster-
ing, and/or focal tree growth analyses. In Figs 3 and 5–7, which present first-annulus 2D graphs
for all sixteen FDPs, the most pronounced of these exceptions are marked with circled numbers 1
or 2 to mark significant local reductions or increases in significance respectively. Fig 6 shows a par-
ticularly striking example in the subtropical Luquillo FDP. At this FDP a valley and a peak, cen-
tered at 100 and 150 Ma back to the LCA respectively, are seen in clustering, recruitment and
growth. If such exceptions are found at the same phylogenetic distance in more than one type of
focal-annular interaction, they may have underlying causes in common (see the detailed analysis
of the Wind River patterns below). Note that these resemblances are not the result of correlations
in the numbers of recruits and survivors in the annuli, because these correlations are preserved in
the null-model data to which the real data are compared (Materials and Methods).
Mortality results show a more complex and difficult-to-interpret pattern
Thirteen of the sixteen FDPs show significant deviations from the null model in the pattern of
small annular tree mortality. The pattern, however, is the opposite of that seen in previous
studies of seedling mortality [20, 39]. More small annular trees than expected die around small
focal trees, and fewer die than expected around large focal trees. This pattern, which is consis-
tent with positive rather than negative density dependence, is largely confined to conspecific
annular trees, except at the Nonggang, Tiantong and Wind River FDPs at which the PDD
effect extends to some heterospecifics. This test is different from the BCI analysis of Fig 4 (d),
which shows that more annular trees than expected die near the focal trees that die regardless
of their size. The Fig 4 (d) analyses are consistent with NDD effects, in which mortality—espe-
cially mortality among closely-related trees—is expected to be spatially clustered.
In Figs 3, 5, 6 and 7 the mortality pattern exhibited by large annular trees that die is the
reverse of that seen in the small annular trees, and like the Fig 4 (c) analysis is consistent with
NDD effects. In fifteen of the FDPs, with the exception of the dry tropical forest plot at Mudu-
malai, there is a deficiency of mortality in large annular trees around small focal trees and an
excess around large focal trees, as NDD would predict. We explore some possible reasons for
the different outcomes of these different mortality tests in the Discussion.
Known between-species interactions account for some of the variation in
significance levels along the phylogenetic distance axis at the Wind River
FDP
At the Wind River (Washington State, USA) FDP, which has only 26 tree species, EAA analysis
reveals an unusually large exception to declines of NDD effects as focal-annular phylogenetic
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
18 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Fig 4. Tests at the BCI FDP of the proportions of recruits or trees that die in Annulus 1 around focal recruits or
focal trees that die, compared with the proportions expected from a null model in which the properties (recruits,
survivors, died) of the focal trees are shuffled repeatedly within species.
https://doi.org/10.1371/journal.pcbi.1008853.g004
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
19 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Fig 5. This figure and the following two figures show two-dimensional GAM analyses of NDD-influenced focal-annular interactions for the
remaining fifteen FDPs in the study (excluding BCI, which is presented in Fig 3). This figure shows results from the lowest-latitude FDPs. Legends as
in Figs 1 and 3. The k-values are the optimized number of smooth terms. Circled numbers 1 and 2 represent local reductions or increases respectively in
the significance of the effects along the phylogenetic distance axis.
https://doi.org/10.1371/journal.pcbi.1008853.g005
distance increases. At the largest focal-annular phylogenetic distances in this FDP, there are
large increases in the significance of deviations of annular clustering and recruitment, and a
large negative effect on focal tree growth rate. First-annulus data are used in the analysis
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
20 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Fig 6. Continuation of the first-annulus two-dimensional phylogenetic distance analyses of all the FDPs, showing FDPs at intermediate
latitudes. Legends as in Figs 1 and 3.
https://doi.org/10.1371/journal.pcbi.1008853.g006
presented in Fig 7, but as with the other FDPs these unique patterns of phylogenetic peaks and
valleys at Wind River are preserved across more distant annuli.
A likely factor contributing to the unusual EAA pattern found in this FDP may be allelo-
pathic inhibition. Large-diameter trees of the western hemlock Tsuga heterophylla have dense
canopies and have been shown to have allelopathic needles [23]. These properties reduce the
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
21 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Fig 7. The highest-latitude set of the first-annulus two-dimensional phylogenetic distance analyses of all the FDPs. Legends as in Figs 1 and 3.
https://doi.org/10.1371/journal.pcbi.1008853.g007
growth rate of trees of other species when T. heterophylla are nearby, and also reduce cluster-
ing and recruitment of trees of a range of species around large T. heterophylla [23, 43].
The effects of removing different combinations of species from the Wind River EAA analy-
sis are shown in Fig 8. Trees of the commonest species, the vine maple Acer circinatum, and
those of the second most common species, Tsuga heterophylla, are close to each other in num-
bers. Together, these two species make up two-thirds of the stems, and those of all the other
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
22 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
species make up the remainder. We therefore divided the species into three categories: A. circi-
natum, T. heterophylla, and a third category made up of all the other species. We examined the
effects of removing different combinations of focal or annular trees, or both, that fall into these
three categories.
The row of graphs (1) at the top of Fig 8 show (A) annular clustering, (B) annular recruit-
ment, and (C) focal tree growth GAM-generated curves for all the Wind River first-annulus
data. All three of these focal-annular interactions show increases in significance among species
separated by 150 Ma or more from their LCA, in the direction expected from NDD models.
This is in striking contrast to the patterns of declining significance with increasing focal-annu-
lar phylogenetic distance that would be predicted if the strength of focal-annular interactions
decreases with increasing phylogenetic distance between them. For focal growth, however, a
group of species pairs that are separated by a little more than 100 Ma shows strong positive
effects of annular trees on the growth of focal trees, consistent with PDD.
The next row of graphs (2) show results from only focal T. heterophylla and all annular spe-
cies except for A. circinatum. The anomalous results at the most distant phylogenetic intervals
are retained for clustering and recruitment, but disappear for focal tree growth (as does the
equally anomalous PDD peak at more intermediate distances). This is the pattern to be expected
if focal T. heterophylla are suppressing recruitment of distantly-related annular species (pre-
dominantly Angiosperms), but if the presence of these distantly-related annular species is not
affecting the growth of the focal T. heterophylla in either a positive or a negative direction.
Graphs in row (3) show the results for only focal A. circinatum and all annular species
except for T. heterophylla. These graphs present only the interactions between focal trees of the
commonest species in the plot and annular species that are not known to have allelopathic
effects. These interactions show a general decline of NDD-influenced effects with increasing
phylogenetic distance across all phylogenetic distances. Thus, at Wind River, such a pattern—
seen, though with many localized exceptions, at most of the FDPs—is revealed when only the
interactions between focal trees of the commonest species A. circinatum and all annular species
except for T. heterophylla are analyzed. When the entire data set is examined, however, this
pattern is masked because of the strong allelopathic effects of T. heterophylla.
Row (4) conditions are the same as in Row (3), except that T. heterophylla has been added
back to the annular trees. Clustering and recruitment patterns around focal A circinatum are
little changed from Row (3), showing a lack of interactions between annular T. heterophylla
and focal A. circinatum. There may be a small negative effect of T. heterophylla on the growth
of focal A. circinatum, but it is at the margin of significance.
The graphs in Row (5) show results from the dataset consisting of all focal species except for
T. heterophylla and all annular species except for A. circinatum. These results are the mirror
image of Row (2). Clustering and recruitment of annular species are not influenced by allelop-
athy from focal trees other than T. heterophylla, and therefore do not show enhanced effects at
extreme phylogenetic distances. But growth of distantly related focal species is slowed, as
expected, by the allelopathic effects of annular T. heterophylla. And the striking peak in growth
of intermediate-distance focal trees is again apparent, strongly indicating that annular T. het-
erophylla may be responsible for this effect as well.
Taken together, these results show that T. heterophylla influences at least some distantly
related species negatively, and may influence trees at intermediate phylogenetic distances posi-
tively, but it does not influence the commonest species A. circinatum. Removal of other combi-
nations of focal and annular species yields results that are consistent with this interpretation
(not shown).
Previously-published measurements of the allelopathic effects of T. heterophylla had not
detected A. circinatum’s immunity to T. heterophylla’s effects [23, 43].
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
23 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Fig 8. EAA two-dimensional first-annulus analysis of allelopathic effects of Tsuga heterophylla on other species in
the Wind River FDP. Three-dimensional analyses (not shown) show smooth declines in significance with increasing
physical distance but preservation of localized phylogenetic distance features across annuli. See text for interpretation.
Legends and confidence intervals as in Fig 1.
https://doi.org/10.1371/journal.pcbi.1008853.g008
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
24 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
The positive effects on focal growth seen in rows (1) and (4) of Fig 8 suggest that T. hetero-
phylla has positive density-dependent effects on the growth of focal trees separated from it by
intermediate phylogenetic distances. This interesting result warrants further investigation.
Discussion
The FDPs examined here span a wide range of environments, from tropical to northern tem-
perate. Across this diversity of ecosystems, the EAA results reinforce the growing body of evi-
dence that NDD interactions between even distantly related species are common [44]. Here
we provide a brief summary of conclusions that can be drawn and questions that can be raised
from EAA analyses.
Peaks and valleys in the significance of focal-annular species interactions
along the phylogenetic axes for each FDP may result from unusual rates of
evolutionary divergence and convergence between pairs or groups of
species
For example, a valley in significance of annular tree effects on focal tree growth might result
from the convergent evolution of distantly-related species on the ability to obtain resources
from the environment [45]. A peak in significance in the NDD component of recruitment of
distantly-related annular trees around focal trees might result from convergent evolution that
has led to susceptibilities to similar pathogens [41] or browsers [46], while a valley in these sig-
nificance levels between distantly-related focal and annular trees might result from divergent
evolution in these susceptibilities. Evolution of allelopathy and of other offensive or defensive
mechanisms in particular species, as in the Wind River FDP (Fig 8), must also play an impor-
tant role in generating peaks and valleys.
We emphasize that the evolutionary changes leading to the peaks and valleys that we
observe are not likely to have originated in the FDPs being examined, but must have had a
much longer history of natural selection that took place in the various ecosystems that were
inhabited by these species’ ancestors. Regardless of these anomalies’ precise origins, EAA can
detect the between-species interactions that are most significant and that are therefore most
likely with further study to yield useful information about the species’ evolutionary histories.
Mortality interactions that have an NDD component may be confounded
with effects that do not have an NDD component
Mortality of seedlings is known to have a strong NDD component [41, 42]. On the assumption
that factors unrelated to NDD effects might play an important role in the mortality of large
trees (such as large herbivores, strangler figs, windstorms, etc.), we divided the focal trees of
each species that died in each FDP into two equally numerous groups (small and large) accord-
ing to their diameters. When Test 3 of Table 2 was applied, the mortality pattern differences in
these two groups were striking and unexpected.
For small annular trees that die, most of the FDPs show patterns consistent with positive
density-dependence. There is low mortality around the largest focal trees and high mortality
around the smallest focal trees. The FDPs Changbaishan and Palanan show no significant
effects. In contrast, when the larger annular trees that die are examined, a positive density-
dependent pattern is only seen at Changbaishan. Instead, a pattern that is consistent with
NDD effects is seen at BCI, Fushan, Gutianshan, Heishiding, Korup, Lambir, Nonggang, Sin-
haraja, Tiantong, Wytham Wood and possibly Wind River. Examination of the causes of
small-tree mortality may reveal the source or sources of the conspecific PDD effects.
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
25 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Storm-damaged FDPs show a variety of EAA patterns
The FDPs Palanan, Fushan and Luquillo are repeatedly damaged by cyclonic wind storms.
These three FDPs have recently been compared with each other and with BCI, and were
shown to have significant differences in mortality, growth and recruitment [47]. The present
study found that all four FDPs also show differences in NDD patterns of mortality, growth and
recruitment (Fig 6).
Palanan is a species-rich tropical forest in the Philippines that has been battered by three
supertyphoons and numerous other storms during the 18 years that it has been censused. (It
was hit with another supertyphoon in 2018, with effects yet to be measured.) This FDP shows
weak focal-annular interactions that do not decline with increasing phylogenetic distance.
Although Luquillo is at the same latitude as Palanan and was damaged by two severe hurri-
canes during the period covered by this study, it shows a different set of focal-annular interac-
tions. Luquillo, like Palanan, exhibits weak clustering and recruitment NDD effects that persist
across most phylogenetic distances, but unlike at Palanan these effects diminish markedly at
the largest distances.
Fushan, at a higher latitude, is a similarly storm-battered submontane FDP. It shows an
interaction pattern more typical of the majority of FDPs, with strong interactions among con-
specifics and an overall decline (with some dramatic exceptions) with increasing focal-annular
phylogenetic distance.
Palanan and Luquillo have lower tree densities than Fushan, even though Fushan lies fur-
ther north. In part this is the result of the more severe effects of the storms at Palanan and
Luquillo, which often rip away the tops of canopy trees. These events open up the areas around
large trees to higher levels of successful recruitment of all species [48, 49]. Enhanced recruit-
ment may help explain why NDD-associated recruitment and clustering patterns are weakly
significant and often change little over most phylogenetic distance in these FDPs. But at
Luquillo, the reduction in the significance of clustering and recruitment at extreme phyloge-
netic distances is more consistent with the majority of FDPs.
The three-dimensional GAM graphs show the persistence of peaks and
valleys in species-species interactions across a range of physical distances
The unique shapes of the three-dimensional surfaces along their phylogenetic distance axis are
retained across annuli at each FDP (Fig 3 and S1 and S2 Figs). They are retained even in distant
annuli in which the means of the z-values that the surfaces represent are not themselves signifi-
cantly different from zero (gray regions of the 3D surfaces). The surfaces can retain their phy-
logenetic-distance shapes, even in the gray regions, because the 95% confidence limits on the
surfaces are small. As noted above, the rotatable three-dimensional GAM graphs from Fushan,
Lambir and Pasoh that are presented in S3–S11 Figs allow the viewer to assess the relationships
between the surfaces and their confidence intervals. In each case these intervals are small com-
pared to the confidence intervals of the z-values themselves. Thus, the GAM analyses of the
EAA data are able to detect the details of focal-annular interactions, even if the z-values them-
selves may be below the level of significance.
Species with different properties do not on average contribute
disproportionately to the results
In the first EAA paper [18] it was shown for the BCI FDP that between-species NDD interac-
tions are of approximately equal strength across a variety of groupings of species into subsets
that have different phenotypic and ecological properties. In that analysis it was decided not to
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
26 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
divide species into phenotypic classes that would have led to unequal-sized subdivisions,
because this would have made the tests for NDD-influenced patterns less directly comparable to
each other. To preserve the statistical power of the tests, and to make them comparable to each
other, the BCI species were subdivided using two different criteria. The first criterion sorted the
species according to their abundance, and the second sorted them according to the CV of their
diameters across all the stems of the species. After ranking of the species according to each of
these criteria, the pooled trees of the ranked species were then divided into thirds, each third
consisting of equal numbers of individual trees. EAA analyses of each of these six subdivisions
showed that all six exhibited the same EAA NDD patterns as the BCI FDP as a whole, though as
expected the subdivision results were less significant than they were in the total data. Thus, in
this FDP, EAA analysis of species with a wide range of properties yields similar results.
Detailed examinations of NDD interactions between conspecifics of individual species of differ-
ent abundances at BCI, however, have suggested substantial differences in the strength of NDD-
influenced patterns, to the point that the least-influenced species may be in danger of local extinc-
tion [50]. EAA can be used to investigate the effects of the removal of information about individual
species that have similar abundances in an FDP, in order to test this observation further.
EAA analyses can be used to investigate ecological-evolutionary processes
in detail
It is possible to use EAA to examine interactions between focal-annular species pairs that are
based, not only on the phylogenetic distance between them, but also on other quantifiable
characteristics: differences in the species’ physical and biochemical phenotypes, in their defen-
sive and allelopathic mechanisms, in their shared interactions with different classes of patho-
gens, herbivores and parasites, and in their associations with the plots’ topographies and soil
types. The only requirement for such an analysis is the ability to arrange focal-annular species
pairs on a scale of numerical values for the character, from least divergent to most divergent.
The EAA approach can then be modified to "sieve out" focal-annular species combinations
that show the greatest discordance between phylogenetic distance and such scalable pheno-
typic and environmental characteristics. A large focal-annular phylogenetic distance, coupled
with a small focal-annular distance in a simultaneously measured phenotypic or niche-related
character, would suggest evolutionary convergence in the character being measured, while the
reverse situation would suggest an unusually high rate of evolutionary divergence. This
approach will permit the isolation of the characteristics that are most likely to be associated
with cases of unusual focal-annular divergence or convergence in FDPs and in similar complex
ecosystems. The nature of these interactions can be explored further by experimental manipu-
lations in the field or in greenhouse experiments.
Most of the thousands of species in these plots are rare. Their aggregate contributions to
NDD effects may be substantial, but most of these contributions are unlikely to be detectable
at the species level. EAA provides us with a tool for finding the species that are likely to repay
further study, while still applying the rigorous EAA standard of examining the effects of only
one variable at a time.
The EAA approach may also be sensitive enough to detect small changes over time in the
focal-annular dynamics of multiply-censused FDPs such as BCI and Pasoh that may be corre-
lated with climatic change.
EAA and Darwin’s hypothesis
EAA analyses can be used to test an important ecological-evolutionary prediction that follows
from Darwin’s observations and from his speculations that were quoted at the beginning of
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
27 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
this paper. When species have lived in close physical proximity to each other within an ecologi-
cal community for some time, some of these species’ evolutionary changes should have been
driven by the direct and indirect interactions among them. Even if particular pairs of tree spe-
cies are not physically close in the ecosystem, they will share symbionts, pollinators, pathogens,
parasites, predators, and herbivores that often have high levels of dispersal. These shared bio-
logical agents must also have undergone evolutionary change as a result of their interactions
with their hosts. It is likely that the variety of evolutionary trajectories in the NDD-influenced
focal-annular effects that are seen in the FDPs of this study stems in part from such biotic
interactions. EAA can detect the pairs of tree species in an FDP that are likely candidates for
detailed studies of these interactions, which will in turn help towards eventual clarification of
their true causes and their evolutionary histories. At the same time, such extended studies will
reveal the proportion of these interactions that are responses to physical factors in the species’
environment, and provide firm evidence for or against Darwin’s hypothesis that biological fac-
tors play a large role in ecosystem evolution.
EAA and ecosystem preservation
In order to preserve threatened ecosystems, we must understand the mechanisms that main-
tain their diversity. EAA can be used to flag between-species interactions that are unusually
strong or weak and are therefore likely to yield significant results when subjected to experi-
mental manipulation. Each such case that is understood in depth will increase our understand-
ing of the kinds of interactions that must be preserved in order to maintain the overall
structure of both intact and endangered ecosystems. General ecological models that ignore
these interactions are of little help in understanding which aspects of ecosystems are important
in their long-term preservation.
Supporting information
S1 Fig. Three-dimensional graphs showing the differences between the physical (m) and
phylogenetic (Ma) axes for z-values that measure the significance of the NDD-influenced
component or annular tree clustering and recruitment for the sixteen FDPs. The graphs
show the patterns seen around the largest quantile of focal tree sizes. Legends and surface col-
ors as in Figs 1 and 3. Note that levels of significance decrease smoothly with increasing focal-
annular physical distance and irregularly with increasing focal-annular phylogenetic distance
at each FDP. In addition, the shapes of the phylogenetic distance curves for clustering and
recruitment often resemble each other, for reasons discussed in the Results section of the main
paper.
(TIF)
S2 Fig. Three-dimensional graphs showing the effect of the largest quintile of annular
trees on the growth of focal trees for all 16 FDPs. Legend as in Figs 1 and 3.
(TIFF)
S3 Fig. bci_focal_growth_largest_annular_biomass_type_factor.html. An interactive rotat-
able graph of a three-dimensional GAM analysis of focal growth data from the BCI FDP. The
graph allows the viewer to examine the surfaces formed by the data from all vantages, in order
to visualize the differences between the physical and the phylogenetic distance axes. The areas
of the surface on the graph that lie in regions of significant z-values are shown in color. The
areas that lie in regions of non-significant z-values are in gray. Because of the size of the graph,
it is possible to show clearly the magnitude of the 95% confidence intervals of the surface itself,
which tend to be small. The surface with its confidence interval also shows that irregular
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
28 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
features along the surface’s phylogenetic distance axis retain their shapes and their significance
even in parts of the surface that lie within regions where the individual z-values themselves are
non-significant.
(HTML)
S4 Fig. bci_annular_clust_largest_focal_trees_type_factor.html. An interactive rotatable
graph of a three-dimensional GAM analysis of clustering data from the BCI FDP. Legend as in
S3 Fig.
(HTML)
S5 Fig. bci_annular_born_largest_focal_trees_type_factor.html. An interactive rotatable
graph of a three-dimensional GAM analysis of recruitment data from the BCI FDP. Legend as
in S3 Fig.
(HTML)
S6 Fig. fushan_focal_growth_largest_annular_biomass_type_factor.html. An interactive
rotatable graph of a three-dimensional GAM analysis of focal growth data from the Fushan
FDP. Legend as in S3 Fig.
(HTML)
S7 Fig. fushan_annular_clust_largest_focal_trees_type_factor.html. An interactive rotat-
able graph of a three-dimensional GAM analysis of clustering data from the Fushan FDP. Leg-
end as in S3 Fig.
(HTML)
S8 Fig. fushan_annular_born_largest_focal_trees_type_factor.html. An interactive rotat-
able graph of a three-dimensional GAM analysis of recruitment data from the Fushan FDP.
Legend as in S3 Fig.
(HTML)
S9 Fig. luquillo_focal_growth_largest_annular_biomass_type_factor.html. An interactive
rotatable graph of a three-dimensional GAM analysis of focal growth data from the Luquillo
FDP. Legend as in S3 Fig.
(HTML)
S10 Fig. luquillo_annular_clust_largest_focal_trees_type_factor.html. An interactive rotat-
able graph of a three-dimensional GAM analysis of clustering data from the Luquillo FDP.
Legend as in S3 Fig.
(HTML)
S11 Fig. Luquillo_annular_born_largest_focal_trees_type_factor.html. An interactive
rotatable graph of a three-dimensional GAM analysis of recruitment data from the Luquillo
FDP. Legend as in S3 Fig.
(HTML)
S1 Data. Data for FDP’s Mudumalai, Wytham Woods, Heishiding, Nonggang and Pala-
nan.
(ZIP)
S2 Data. Data for remainder of FDP’s in the study. These zipped files contain all the data
used to plot the figure graphs, including information needed to generate the errors on the
graphs, in .csv format.
(ZIP)
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
29 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Acknowledgments
We are most grateful for valuable advice on the paper provided by Scott Rifkin, Joshua Kohn,
Bret Elderd, Jim Dalling and Will Pearse. We also gratefully acknowledge help in data analysis
from Mahidhar Tatineni, Jerry Greenberg, Ron Hawkins and Paul Rodriguez of the San Diego
Supercomputer Center, and assistance from Sandra L. Yap and Edwino S. Fernando.
Dedication
Dedicated to the memories of Abdul Rahman bin Kassim (1963–2018) (Pasoh FDP) and of
Perry S. Ong (1960–2019) (Palanan FDP), who made many valued contributions to these and
to many other forest plot studies.
Author Contributions
Conceptualization: Christopher Wills, Shuai Fang, Kyle E. Harms.
Data curation: Shuai Fang, Yunquan Wang, James Lutz, Jill Thompson, Sandeep Pulla, Boni-
facio Pasion, Sara Germain, Heming Liu, Joseph Smokey, Sheng-Hsin Su, Nathalie Butt,
Chengjin Chu, George Chuyong, Chia-Hao Chang-Yang, H. S. Dattaraja, Stuart Davies,
Sisira Ediriweera, Shameema Esufali, Christine Dawn Fletcher, Nimal Gunatilleke, Savi
Gunatilleke, Chang-Fu Hsieh, Fangliang He, Stephen Hubbell, Zhanqing Hao, Akira Itoh,
David Kenfack, Buhang Li, Xiankun Li, Keping Ma, Michael Morecroft, Xiangcheng Mi,
Yadvinder Malhi, Perry Ong, Lillian Jennifer Rodriguez, H. S. Suresh, I Fang Sun, Raman
Sukumar, Sylvester Tan, Duncan Thomas, Maria Uriarte, Xihua Wang, Xugao Wang, T. L.
Yao, Jess Zimmermann.
Formal analysis: Christopher Wills, Bin Wang, Yunquan Wang, Yi Jin, James Lutz, Jill
Thompson, Kyle E. Harms, Sandeep Pulla, Bonifacio Pasion, Sara Germain, Heming Liu,
Joseph Smokey, Nathalie Butt, Chengjin Chu, Chia-Hao Chang-Yang, Sisira Ediriweera,
Shameema Esufali, Christine Dawn Fletcher, Nimal Gunatilleke, Savi Gunatilleke, Chang-
Fu Hsieh, Fangliang He, Zhanqing Hao, Akira Itoh, Buhang Li, Xiankun Li, Keping Ma,
Xiangcheng Mi, Raman Sukumar, Sylvester Tan.
Funding acquisition: Heming Liu, Sheng-Hsin Su, George Chuyong, Chia-Hao Chang-Yang,
Stuart Davies, Shameema Esufali, Christine Dawn Fletcher, Nimal Gunatilleke, Savi Guna-
tilleke, Chang-Fu Hsieh, Fangliang He, Stephen Hubbell, Zhanqing Hao, Akira Itoh, David
Kenfack, Buhang Li, Yadvinder Malhi, Perry Ong, Lillian Jennifer Rodriguez, H. S. Suresh,
I Fang Sun, Raman Sukumar, Duncan Thomas, Xihua Wang, Xugao Wang, T. L. Yao, Jess
Zimmermann.
Investigation: Bin Wang, Shuai Fang, James Lutz, Joseph Smokey, Chia-Hao Chang-Yang, H.
S. Dattaraja, Savi Gunatilleke, Stephen Hubbell, H. S. Suresh.
Methodology: Shuai Fang, Yunquan Wang, Yi Jin, Sandeep Pulla, Maria Uriarte.
Project administration: Yunquan Wang, Nathalie Butt, Chengjin Chu, George Chuyong, H.
S. Dattaraja, Stuart Davies, Sisira Ediriweera, Shameema Esufali, Christine Dawn Fletcher,
Nimal Gunatilleke, Savi Gunatilleke, Chang-Fu Hsieh, Fangliang He, Stephen Hubbell,
Zhanqing Hao, Akira Itoh, David Kenfack, Buhang Li, Xiankun Li, Keping Ma, Michael
Morecroft, Xiangcheng Mi, Yadvinder Malhi, Perry Ong, Lillian Jennifer Rodriguez, H. S.
Suresh, I Fang Sun, Raman Sukumar, Sylvester Tan, Duncan Thomas, Maria Uriarte, Xihua
Wang, Xugao Wang, T. L. Yao, Jess Zimmermann.
Resources: Shuai Fang, Sheng-Hsin Su, Nathalie Butt, Chengjin Chu, Chia-Hao Chang-Yang,
H. S. Dattaraja, Stuart Davies, Sisira Ediriweera, Keping Ma, Michael Morecroft.
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
30 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
Software: Christopher Wills, Bin Wang, Yi Jin.
Supervision: Christopher Wills, Bonifacio Pasion.
Validation: Jill Thompson, Heming Liu.
Visualization: Bin Wang.
Writing – original draft: Christopher Wills.
Writing – review & editing: Christopher Wills, Bin Wang, Shuai Fang, James Lutz, Jill
Thompson, Kyle E. Harms, Sandeep Pulla, Bonifacio Pasion, Sara Germain.
References
1. Cavender-Bares J, Kozak KH, Fine PVA, Kembel SW. The merging of community ecology and phyloge-
netic biology. 2009; 12(7):693–715. https://doi.org/10.1111/j.1461-0248.2009.01314.x PMID:
19473217
2. Broekman MJE, Muller-Landau HC, Visser MD, Jongejans E, Wright SJ, de Kroon H. Signs of stabilisa-
tion and stable coexistence. Ecol Lett. 2019; 22(11):1957–75. https://doi.org/10.1111/ele.13349
PubMed PMID: WOS:000477122900001. PMID: 31328414
3. Bagchi R, Henrys PA, Brown PE, Burslem D, Diggle PJ, Gunatilleke CVS, et al. Spatial patterns reveal
negative density dependence and habitat associations in tropical trees. Ecology. 2011; 92(9):1723–9.
https://doi.org/10.1890/11-0335.1 PubMed PMID: WOS:000294152800003. PMID: 21939068
4. Chesson P. General theory of competitive coexistence in spatially-varying environments. Theoretical
Population Biology. 2000; 58(3):211–37. https://doi.org/10.1006/tpbi.2000.1486 PubMed PMID:
WOS:000166247900004. PMID: 11120650
5. Marden JH, Mangan SA, Peterson MP, Wafula E, Fescemyer HW, Der JP, et al. Ecological genomics
of tropical trees: how local population size and allelic diversity of resistance genes relate to immune
responses, cosusceptibility to pathogens, and negative density dependence. Mol Ecol. 2017; 26
(9):2498–513. https://doi.org/10.1111/mec.13999 PubMed PMID: WOS:000400335200008. PMID:
28042895
6. Chesson P. Updates on mechanisms of maintenance of species diversity. Journal of Ecology. 2018;
106(5):1773–94. https://doi.org/10.1111/1365-2745.13035 PubMed PMID: WOS:000452178600002.
7. Gillett JB. Pest pressure, an underestimated factor in evolution. In: Nichols D, editor. Taxonomy and
Geography, a Symposium. London: London Systematics Association; 1962. p. 37–46. PMID:
13884603
8.
Janzen DH. Herbivores and the number of tree species in tropical forests. Am Nat. 1970; 104:501–29.
9. Connell JH. On the role of natural enemies in preventing competitive exclusion in some marine animals
and in rain forest trees. In: Den Boer PJ, Gradwell G, editors. Dynamics of Populations. New York:
PUDOC; 1971. p. 298–312.
10.
Lindstrom KM, Foufopoulos J, Parn H, Wikelski M. Immunological investments reflect parasite abun-
dance in island populations of Darwin’s finches. Proc R Soc B-Biol Sci. 2004; 271(1547):1513–9.
https://doi.org/10.1098/rspb.2004.2752 PubMed PMID: WOS:000222600600011. PMID: 15306324
11. Wills C, Condit R, Foster R, Hubbell S. Strong density- and diversity-related effects help to maintain
tree species diversity in a neotropical forest. Proceedings of the National Academy of Sciences. 1997;
94(4):1252–7.
12. Wills C, Condit R. Similar non-random processes maintain diversity in two tropical rainforests. Proc R
Soc B-Biol Sci. 1999; 266(1427):1445–52. https://doi.org/10.1098/rspb.1999.0799 PubMed PMID:
WOS:000081833200006. PMID: 10457617
13. Comita LS, Queenborough SA, Murphy SJ, Eck JL, Xu KY, Krishnadas M, et al. Testing predictions of
the Janzen-Connell hypothesis: a meta-analysis of experimental evidence for distance- and density-
dependent seed and seedling survival. Journal of Ecology. 2014; 102(4):845–56. https://doi.org/10.
1111/1365-2745.12232 PubMed PMID: WOS:000338027500003. PMID: 25253908
14. Marhaver KL, Vermeij MJA, Rohwer F, Sandin SA. Janzen-Connell effects in a broadcast-spawning
Caribbean coral: distance-dependent survival of larvae and settlers. Ecology. 2013; 94(1):146–60.
PubMed PMID: WOS:000316186900017. https://doi.org/10.1890/12-0985.1 PMID: 23600249
15. Karvonen A, Wagner CE, Selz OM, Seehausen O. Divergent parasite infections in sympatric cichlid
species in Lake Victoria. J Evol Biol. 2018; 31(9):1313–29. Epub 2018/06/27. https://doi.org/10.1111/
jeb.13304 PMID: 29944770.
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
31 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
16. McGill BJ. Towards a unification of unified theories of biodiversity. Ecol Lett. 2010; 13(5):627–42.
https://doi.org/10.1111/j.1461-0248.2010.01449.x PubMed PMID: WOS:000276790200011. PMID:
20337695
17. Kaliszewicz A, Uchmanski J. A cross-phyla response to Daphnia chemical alarm substances by an
aquatic oligochaete. Ecological Research. 2009; 24(2):461–6. https://doi.org/10.1007/s11284-008-
0522-0 PubMed PMID: WOS:000263504100022.
18. Wills C, Harms KE, Wiegand T, Punchi-Manage R, Gilbert GS, Erickson D, et al. Persistence of Neigh-
borhood Demographic Influences over Long Phylogenetic Distances May Help Drive Post-Speciation
Adaptation in Tropical Forests. PLOS ONE. 2016; 11(6):e0156913. https://doi.org/10.1371/journal.
pone.0156913 PMID: 27305092
19. Harms KE, Wright SJ, Calderon O, Hernandez A, Herre EA. Pervasive density-dependent recruitment
enhances seedling diversity in a tropical forest. Nature. 2000; 404(6777):493–5. https://doi.org/10.
1038/35006630 PubMed PMID: WOS:000086257700047. PMID: 10761916
20. Bagchi R, Gallery RE, Gripenberg S, Gurr SJ, Narayan L, Addis CE, et al. Pathogens and insect herbi-
vores drive rainforest plant diversity and composition. Nature. 2014; 506(7486):85–8. https://doi.org/10.
1038/nature12911 http://www.nature.com/nature/journal/v506/n7486/abs/nature12911.
html#supplementary-information. PMID: 24463522
21. Detto M, Visser MD, Wright SJ, Pacala SW. Bias in the detection of negative density dependence in
plant communities. Ecol Lett. 2019; 22(11):1923–39. https://doi.org/10.1111/ele.13372 PubMed PMID:
WOS:000486708600001. PMID: 31523913
22. Hu¨lsmann L, Chisholm RA, Hartig F. Is Variation in Conspecific Negative Density Dependence Driving
Tree Diversity Patterns at Large Scales? Trends Ecol Evol. 2020. https://doi.org/10.1016/j.tree.2020.
10.003 PMID: 33589047
23.
Lutz JA, Larson AJ, Furniss TJ, Donato DC, Freund JA, Swanson ME, et al. Spatially nonrandom tree
mortality and ingrowth maintain equilibrium pattern in an old-growth Pseudotsuga-Tsuga forest. Ecol-
ogy. 2014; 95(8):2047–54. https://doi.org/10.1890/14-0157.1 PubMed PMID: WOS:000340215900003.
PMID: 25230456
24. Wills C, Harms KE, Condit R, King D, Thompson J, He F, et al. Nonrandom Processes Maintain Diver-
sity in Tropical Forests. Science. 2006; 311(5760):527–31. https://doi.org/10.1126/science.1117715
PMID: 16439661
25. Wiegand T, Gunatilleke S, Gunatilleke N. Species associations in a heterogeneous Sri lankan diptero-
carp forest. Am Nat. 2007; 170(4):E77–E95. https://doi.org/10.1086/521240 PubMed PMID:
WOS:000249307700001. PMID: 17891727
26. Wiegand T, Moloney KA. Handbook of Spatial Point-Pattern Analysis in Ecology. Ch Hall Crc App Env.
2014:1–495. PubMed PMID: WOS:000364179900008.
27. Stoyan D, Bertram U, Wendrock H. Estimation Variances for Estimators of Product Densities and Pair
Correlation-Functions of Planar Point-Processes. Ann I Stat Math. 1993; 45(2):211–21. https://doi.org/
10.1007/Bf00775808 PubMed PMID: WOS:A1993LL84100002.
28.
Illian J, Penttinen A, Stoyan H, Stoyan D. Statistical Analysis and Modelling of Spatial Point Patterns.
Chichester, UK: Wiley; 2008.
29. Wood SN. Generalized additive models: an introduction with R. Second edition. ed. Boca Raton: CRC
Press/Taylor & Francis Group; 2017. xx, 476 pages p.
30.
Team RC. R: A language and environment for statistical computing https://www.R-project.org/: Founda-
tion for Statistical Computing, Vienna, Austria; 2019.
31. Goreaud F, Pelissier R. On explicit formulas of edge effect correction for Ripley’s K-function. Journal of
Vegetation Science. 1999; 10(3):433–8. https://doi.org/10.2307/3237072 PubMed PMID:
WOS:000082205100014.
32. Benjamini Y, Hochberg Y. On the Adaptive Control of the False Discovery Rate in Multiple Testing With
Independent Statistics. 2000; 25(1):60–83. https://doi.org/10.1016/s0165-0270(99)00194-6 PMID:
10720676
33.
Fang S, Munoz F, Ye J, Lin F, Yuan ZQ, Kuang X, et al. Deterministic processes drive functional and
phylogenetic temporal changes of woody species in temperate forests in Northeast China. Annals of
Forest Science. 2019; 76(2). doi: ARTN 42 https://doi.org/10.1007/s13595-019-0830-2 PubMed PMID:
WOS:000465547100002.
34. Hector A, Schmid B, Beierkuhnlein C, Caldeira MC, Diemer M, Dimitrakopoulos PG, et al. Plant diversity
and productivity experiments in European grasslands. Science. 1999; 286(5442):1123–7. https://doi.
org/10.1126/science.286.5442.1123 PubMed PMID: WOS:000083534200033. PMID: 10550043
35.
Loreau M. Biodiversity and ecosystem functioning: recent theoretical advances. Oikos. 2000; 91(1):3–
17. https://doi.org/10.1034/j.1600-0706.2000.910101.x PubMed PMID: WOS:000165237400001.
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
32 / 33
PLOS COMPUTATIONAL BIOLOGY Each of 16 forests shows a unique pattern of between-tree interactions
36. Paine CET, Norden N, Chave J, Forget P-M, Fortunel C, Dexter KG, et al. Phylogenetic density depen-
dence and environmental filtering predict seedling mortality in a tropical forest. Ecol Lett. 2012; 15
(1):34–41. https://doi.org/10.1111/j.1461-0248.2011.01705.x PMID: 22004454
37. Davi H, Cailleret M, Restoux G, Amm A, Pichot C, Fady B. Disentangling the factors driving tree repro-
duction. Ecosphere. 2016; 7(9). doi: UNSP e01389 https://doi.org/10.1002/ecs2.1389 PubMed PMID:
WOS:000387210100001.
38. Green PT, Harms KE, Connell JH. Nonrandom, diversifying processes are disproportionately strong in
the smallest size classes of a tropical forest. Proc Natl Acad Sci U S A. 2014; 111(52):18649–54.
https://doi.org/10.1073/pnas.1321892112 PubMed PMID: WOS:000347444400061. PMID: 25512498
39.
Johnson DJ, Bourg NA, Howe R, McShea WJ, Wolf A, Clay K. Conspecific negative density-dependent
mortality and the structure of temperate forests. Ecology. 2014; 95(9):2493–503. PubMed PMID:
WOS:000342340200012.
40. Mangan SA, Schnitzer SA, Herre EA, Mack KML, Valencia MC, Sanchez EI, et al. Negative plant–soil
feedback predicts tree-species relative abundance in a tropical forest. Nature. 2010; 466(7307):752–5.
https://doi.org/10.1038/nature09273 PMID: 20581819
41. Gilbert GS, Webb CO. Phylogenetic signal in plant pathogen-host range. Proc Natl Acad Sci U S A.
2007; 104(12):4979–83. https://doi.org/10.1073/pnas.0607968104 PubMed PMID:
WOS:000245256700040. PMID: 17360396
42. Parker IM, Saunders M, Bontrager M, Weitz AP, Hendricks R, Magarey R, et al. Phylogenetic structure
and host abundance drive disease pressure in communities. Nature. 2015; 520(7548):542–4. https://
doi.org/10.1038/nature14372 http://www.nature.com/nature/journal/v520/n7548/abs/nature14372.
html#supplementary-information. PMID: 25903634
43. Das AJ, Larson AJ, Lutz JA. Individual species-area relationships in temperate coniferous forests. Jour-
nal of Vegetation Science. 2018; 29(2):317–24. https://doi.org/10.1111/jvs.12611 PubMed PMID:
WOS:000431503000019.
44.
LaManna JA, Mangan SA, Alonso A, Bourg NA, Brockelman WY, Bunyavejchewin S, et al. Plant diver-
sity increases with the strength of negative density dependence at the global scale. Science. 2017; 356
(6345):1389–92. https://doi.org/10.1126/science.aam5678 PubMed PMID: WOS:000404351500039.
PMID: 28663501
45. Hubbell SP. Neutral theory and the evolution of ecological equivalence. Ecology. 2006; 87(6):1387–98.
https://doi.org/10.1890/0012-9658(2006)87[1387:ntateo]2.0.co;2 PubMed PMID:
WOS:000238658400006. PMID: 16869413
46.
Forrister DL, Endara MJ, Younkin GC, Coley PD, Kursar TA. Herbivores as drivers of negative density
dependence in tropical forest saplings. Science. 2019; 363(6432):1213-+. https://doi.org/10.1126/
science.aau9460 PubMed PMID: WOS:000461329000048. PMID: 30872524
47. Hogan J, Zimmerman J, Thompson J, Uriarte M, Swenson N, Condit R, et al. The Frequency of
Cyclonic Wind Storms Shapes Tropical Forest Dynamism and Functional Trait Dispersion. 2018; 9
(7):404. https://doi.org/10.3390/f9070404
48. Hubbell SP, Foster RB, O’Brien ST, Harms KE, Condit R, Wechsler B, et al. Light-gap disturbances,
recruitment limitation, and tree diversity in a neotropical forest. Science. 1999; 283(5401):554–7.
https://doi.org/10.1126/science.283.5401.554 PubMed PMID: WOS:000078203300042. PMID:
9915706
49. Comita LS, Uriarte M, Thompson J, Jonckheere I, Canham CD, Zimmerman JK. Abiotic and biotic driv-
ers of seedling survival in a hurricane-impacted tropical forest. Journal of Ecology. 2009; 97(6):1346–
59. https://doi.org/10.1111/j.1365-2745.2009.01551.x PubMed PMID: WOS:000270787100025.
50. Stump SM, Comita LS. Interspecific variation in conspecific negative density dependence can make
species less likely to coexist. Ecol Lett. 2018; 21(10):1541–51. https://doi.org/10.1111/ele.13135
PubMed PMID: WOS:000444413300009. PMID: 30129216
PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021
33 / 33
PLOS COMPUTATIONAL BIOLOGY
| null |
10.1088_1361-6463_ad146b.pdf
|
Data availability statement
All data that support the findings of this study are included
within the article (and any supplementary files).
|
Data availability statement All data that support the findings of this study are included within the article (and any supplementary files).
|
J. Phys. D: Appl. Phys. 57 (2024) 115204 (20pp)
Journal of Physics D: Applied Physics
https://doi.org/10.1088/1361-6463/ad146b
Liquid treatment with a plasma jet
surrounded by a gas shield: effect of the
treated substrate and gas shield
geometry on the plasma effluent
conditions
Pepijn Heirman1,∗, Ruben Verloy1,2, Jana Baroen1,2, Angela Privat-Maldonado1,2,
Evelien Smits2 and Annemie Bogaerts1,∗
1 Research group PLASMANT, Department of Chemistry, University of Antwerp, Antwerp, Belgium
2 Center for Oncological Research (CORE), IPPON, University of Antwerp, Antwerp, Belgium
E-mail: [email protected] and [email protected]
Received 18 October 2023, revised 30 November 2023
Accepted for publication 11 December 2023
Published 20 December 2023
Abstract
The treatment of a well plate by an atmospheric pressure plasma jet is common for in vitro
plasma medicine research. Here, reactive species are largely produced through the mixing of the
jet effluent with the surrounding atmosphere. This mixing can be influenced not only by the
ambient conditions, but also by the geometry of the treated well. To limit this influence and
control the atmosphere, a shielding gas is sometimes applied. However, the interplay between
the gas shield and the well geometry has not been investigated. In this work, we developed a
2D-axisymmetric computational fluid dynamics model of the kINPen plasma jet, to study the
mixing of the jet effluent with the surrounding atmosphere, with and without gas shield. Our
computational and experimental results show that the choice of well type can have a significant
influence on the effluent conditions, as well as on the effectiveness of the gas shield.
Furthermore, the geometry of the shielding gas device can substantially influence the mixing as
well. Our results provide a deeper understanding of how the choice of setup geometry can
influence the plasma treatment, even when all other operating parameters are unchanged.
Keywords: atmospheric pressure plasma jet, 2D fluid modeling, gas shield, in vitro treatment,
plasma-liquid
1. Introduction
An atmospheric pressure plasma jet (APPJ) is a typical source
of cold atmospheric plasma (CAP) used in plasma medicine
research [1]. The field of plasma medicine investigates the
interaction between CAP and cells or tissue [2], and has shown
promise for several applications, such as sterilization [2],
∗
Authors to whom any correspondence should be addressed.
wound healing [3, 4], and cancer treatment [5]. Additionally,
the use of APPJs is gaining attention in other fields, such as
nitrogen fixation [6] and polymer treatment [7]. An import-
ant example of an APPJ specifically developed for plasma
medicine applications is the kINPen®. This medically certified
plasma jet operates with argon as the feed gas and has been the
subject of numerous studies [8–10].
Reactive oxygen and nitrogen species (RONS) produced by
the plasma act as the major players in the biological effects
induced by plasma treatment [11, 12]. In a plasma jet that uses
1
© 2023 IOP Publishing Ltd
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
a noble gas as the feed gas, like the kINPen, these reactive
species can originate from two possible sources: either from
impurities (or deliberate admixtures) in the feed gas [13, 14],
or from mixing of the plasma effluent with the surround-
ing atmosphere [15]. The latter makes plasma treatment with
an APPJ susceptible to the ambient conditions during treat-
ment, such as the relative humidity [16], which determines
how much water can diffuse into the active plasma zone. This,
in turn, plays a role in the RONS production and treatment
reproducibility. To prevent this, a shielding gas device can be
employed. First introduced in the context of plasma medicine
by Reuter et al [17], a gas shield induces a second, concentric
gas flow that surrounds the jet and separates the plasma efflu-
ent from the surrounding air. The composition of the shielding
gas can be controlled, thus allowing control over the gasses
that are in contact with the plasma effluent. Since its introduc-
tion, the use of a gas shield with different gas compositions has
been the subject of several studies. Although its effect on the
RONS deposited in a treated liquid and its effect on treated
cells has been studied [18, 19], its effect on RONS produc-
tion in the gas phase is usually analyzed with operation of the
free jet, i.e. without a substrate being treated. In fact, most dia-
gnostic investigations of the kINPen plasma and its effluent
have been performed on the free jet, without accounting for
the effects caused by the presence of a substrate [8].
The treatment of biological tissue with an APPJ can be
investigated in vivo [20–22], and some clinical trials have been
reported [23, 24]. However, most biomedical research is still
performed in vitro [25, 26], where cell treatment is typic-
ally performed in well plates of varying sizes. Treatment of
a well plate with a plasma jet, as opposed to e.g. a flat surface,
inherently causes a backflow towards the jet outlet, of which
the flow pattern logically depends on the well geometry. As
such, this backflow may in turn influence the dynamics of both
the jet effluent and, when used, the shielding gas. Moreover,
cells in vitro are usually covered by a liquid layer such as
cell medium. The gas flow over this liquid causes evaporation,
which forms an additional source of water vapor that can affect
the chemical pathways in the effluent, and thus the plasma
treatment [27, 28].
The interaction between a plasma jet and a liquid forms a
very complicated system, with many phenomena influencing
each other [29]. For this reason, simulations can be valuable
for elucidating the different processes and their effect on the
treatment [5]. For the kINPen, and plasma jets in general, 0D
models have been employed to investigate the chemical reac-
tion pathways leading to RONS formation [30–33]. To account
for spatial phenomena, like convection and diffusion, multi-
dimensional models have been developed to study e.g. mix-
ing of the jet effluent with ambient species [34, 35] and the
effects of e.g. flow rate [36] or molecular admixtures [13] on
the RONS production. The influence of using a gas shield
has also been investigated computationally [17, 37–39]. Like
gas phase diagnostics, however, these computational invest-
igations are often performed for the free jet, without a sub-
strate. Some computational efforts to investigate the interac-
tion of a plasma jet with a liquid substrate have been repor-
ted. Lindsay et al [40] presented a 2D axisymmetric model
of a jet-like system above a petri dish, showing the import-
ance of the induced convection in the liquid, while Verlackt
et al [41] developed a model with a more extensive chemistry
set to investigate transport phenomena in a liquid-filled beaker
treated by a plasma jet. Semenov et al [42] presented a mod-
eling study on the description of convection and diffusion in
the system and at the gas-liquid interface. In our previous work
[43], we reported on a combined 0D/2D modeling approach to
describe both the plasma chemistry and the transport of neutral
species into the liquid. This model was applied to the kINPen
above a well plate with buffered liquid water and investigated
the stability of dissolved species after the treatment.
In the present work, we investigate how in vitro treatment
with a plasma jet is affected by the choice of the used substrate
(i.e. well size), as well as the interplay of this effect with a gas
shield during treatment. For this purpose, we employ a com-
putational 2D-axisymmetric model of the kINPen plasma jet
above a liquid water surface, with and without a shielding gas
device. Simulations are performed for various ambient condi-
tions, such as temperature and relative humidity. This allows
us to assess whether the gas shield is effective in eliminating
variation caused by the ambient conditions for different setup
geometries. Finally, attention is given to the geometry of the
shielding gas device itself. Since the first report of a gas shield,
different designs have been employed in different works [19,
35, 44]. However, the possible effects of this change in geo-
metry on, e.g., mixing of the shielding gas with the plasma
jet effluent have rarely been addressed. Our simulations reveal
that the design of the shielding gas device can have a sub-
stantial effect on the conditions in the plasma effluent and the
effectiveness of the shielding.
2. Methods
2.1. Computational methods
The description of the gas and liquid phase in the simulated
system, i.e. a plasma jet above a well plate containing a liquid,
is based on the modeling approach presented in [43]. Here, we
constructed the 2D-axisymmetric fluid dynamics model with
COMSOL Multiphysics (version 6.0) [45]. Figure 1 shows the
simulated geometry. The plasma jet geometry is based on that
of the kINPen, while the shielding gas device is based on a
commercially available gas shield made by Neoplas GmbH
specifically for the kINPen. The dimensions of the treated
well in figure 1 are those of a 24-well plate containing 1 mL
of liquid, though different well sizes are used throughout the
paper. Because of the impinging gas from the plasma jet, the
liquid surface is deformed, causing a dimple of which the
shape and depth depend on the treatment setup [46]. The shape
of the gas-liquid interface in our model is therefore based on
observations made in our lab for the same setup geometry.
Note that we treat the liquid surface in the model as stationary,
and oscillations of the surface over time, as present in reality,
are not yet accounted for. However, as we focus mainly on
the effect of the substrate geometry on the gas phase dynam-
ics, this assumption will not significantly influence our results.
The distance between the jet nozzle and the liquid surface (in
2
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
(−→
u · ∇
ρ
) −→
u = ∇ · [−pI + K]
−→
µ the velocity vector (m s
With ρ the density (kg m
−1)
−3),
and p the pressure (Pa). I is the identity matrix, while K is
the viscous stress tensor (Pa). When applying a gas flow rate
of 2 SLM through the jet, the jet is in a turbulent regime [8].
Therefore, we employed Menter’s shear stress transport turbu-
lence model [47] as implemented in the COMSOL computa-
tional fluid dynamics (CFD) module.
Walls are treated with the no-slip boundary condition. At
the open boundaries, a pressure of 1 atm is prescribed as a
normal stress, while at the inlet of both the jet and the gas
shield the inflow is prescribed as a fully developed flow pro-
file. Finally, the gas-liquid interface is treated as a slip wall,
by implementing a no-penetration boundary condition, as in
reality the flowing gas sets the liquid in motion.
2.1.2. Time-dependent calculation of heat
transfer and
species transport. We calculated the temperature in the
gas and liquid phase by solving the conservation of energy:
ρCp
∂T
∂t
+ ∇ · −→
q + ρCp
−→
u · ∇T = Q
where Cp is the heat capacity at constant pressure, T is the
−→
q = −K∇T is the conductive heat
absolute temperature and
flux, with k the thermal conductivity. Q represents additional
heat sources like, in this model, viscous dissipation and heat
loss due to water evaporation.
The transport of chemical species is calculated by solving
the continuity equation:
+ ∇ ·
−→
Ji + ρ
)
(−→
u · ∇
ni = Ri
ρ
∂ni
∂t
Here, ni is the number density of species i, while
−→
Ji is its
diffusive flux. Ri represents the net production or consump-
tion of chemical species (equal to zero in case no chemical
reactions are included). The continuity equation is solved for
each species, except argon, of which the density is determ-
ined through the fact that the sum of all mass fractions must
be equal to 1.
The diffusive flux is defined as:
−→
Ji = −Dm,i
∇ni
− DT,i
∇ni
with DT,i =
µT
ρ · ScT
where Dm,i is the mixture-averaged diffusion coefficient [48].
DT,i
is the eddy diffusivity; this additional diffusive flux
accounts for the turbulent mixing by eddies that are not
resolved by the turbulence model. DT,i depends on the turbu-
lent viscosity µT, and on the turbulent Schmidt number ScT,
calculated by COMSOL based on [49]. The binary diffusion
coefficients used to determine Dm,i are calculated within the
model as described in [50], using tabulated data from [51].
3
Figure 1. General model geometry. The geometry components are
(1) the plasma jet, (2) the shield gas device, (3) gas phase and (4)
liquid phase. The pin-electrode (5) and plasma plume (6) are also
indicated. The boundary conditions applied at the edges of the
model are indicated. As the geometry is 2D axisymmetric, only the
right part is simulated by the model, with the symmetry axis
indicated in red.
its original state) is 20 mm. We use a gas flow rate of 2 stand-
ard liters per minute (SLM) for the jet, and 4 SLM for the gas
shield (when applied). We selected these conditions to make
sure that, in the experiments performed with this geometry,
there would be no liquid splashing out of the well, and the
jet would operate in a non-touching regime, i.e. no direct dis-
charge onto the liquid occurs. Both phenomena would signi-
ficantly complicate the modeling description.
The plasma discharge itself is not simulated by the model.
Instead, we study the species that typically lead to RONS
formation in this system, i.e. the surrounding O2, N2 and
H2O, and their mixing with the feed gas from the jet. In this
way, we hope to provide a deeper understanding of how the
choice of setup geometry, like the treated well and the use of
a gas shield, can influence the treatment even when all fur-
ther operating parameters (which define the plasma discharge)
are unchanged. The modeling approach is as follows: first, we
calculate the stationary state of the flow field in the system.
Using this stationary flow field as the input, we simulate the
temperature and transport of species in the system in a time-
dependent manner, for a treatment of 10 s. These simulations
are performed for different geometries, of both gas shield and
well plate, to elucidate their effect on the plasma jet effluent
conditions.
The flow field
2.1.1. Stationary calculation of the fluid flow.
in the system is calculated by solving the time-independent,
incompressible Navier–Stokes equations:
ρ∇ · −→
u = 0
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
Other species-specific parameters Cp and k are also calculated
within the model, as described in [52, 53], respectively.
At the jet inlet, argon enters the domain, containing impur-
ities (1 ppm O2, 4 ppm N2, 3 ppm H2O [32, 43]). The tem-
perature of the gas flowing into the system via this inlet is
set to 327 K as in [43]. The gas shield inlet supplies dry air
(79% N2, 21% O2) into the system, while at the open bound-
aries we specify a constant concentration of N2, O2, and H2O
equal to the initial conditions in the gas phase, i.e. humid air
with a H2O concentration that depends on the specified rel-
ative humidity and temperature. The temperature at both the
gas shield inlet and the open boundaries is set to the initial
ambient temperature, taken to be 293 K in the general case,
but (when indicated) varied between 283 K and 303 K for dif-
ferent simulations. Walls are treated as thermally insulating,
with a no penetration boundary condition for the chemical spe-
cies. At the gas-liquid interface, water evaporation is accoun-
ted for with a flux that keeps the water density at the interface
in line with the vapor pressure of water, which is in turn cal-
culated via Antoine’s law. Evaporative cooling, i.e. the heat
loss at the liquid interface due to the evaporation of water, is
implemented by prescribing a loss of heat at the interface as
follows:
Qwater evap = JH2O
· Hvap
With J the molar flux of water due to evaporation and H
−1).
the latent heat of evaporation (for water = 2260 kJ kg
Note that in our previous work [43], we confirmed the occur-
rence of this evaporative cooling experimentally, and imple-
mented a correction factor to prevent large overestimation of
the evaporative cooling at the gas-liquid interface in the model.
In the present work, this correction factor is no longer needed.
Briefly, the earlier overestimation stemmed from assuming
JH2O = Jz,H2O, i.e. the total axial flux of H2O at the interface,
as described in [40], which is an overestimation for a non-flat
surface due to the large contribution of fluid flow in the axial
direction.
Transport of chemical species other than H2O over the
gas-liquid interface is implemented with a flux governed by
Henry’s law:
ci, aq = Hi
· RT · ci, g
where ci, aq and are the concentrations of species i in the liquid
and gas phase, respectively, and Hi is the temperature depend-
ent Henry’s constant [54].
2.2. Experimental methods
We performed experiments with the kINPen® MED to validate
the computational results, using the same setup geometry as in
the simulations, i.e. 2 SLM argon flow rate with a 20 mm gap
between the kINPen nozzle and the liquid surface. The treat-
ment time was 60 s in all experiments. Afterwards, deionized
water was added to counter evaporation. Specifically, we com-
pared the liquid volume in the different well types before and
after treatment, showing that the treatment caused the evapor-
ation of up to 2% of the liquid during the applied treatment
time.
2.2.1. Determination of RONS in the treated liquid.
Concentrations of long-lived RONS (H2O2, HNO2 and HNO3)
were determined after treatment of a 12-, 24-, 48-, and 96-well
plate (respectively 665180, Greiner; 10062–896, Avantor;
677180, Greiner and 655180, Greiner), containing 2 ml, 1 ml,
0.5 ml and 0.1 ml phosphate-buffered saline (PBS) per well,
respectively. These volumes are typically used in experiments,
as e.g. outlined in the standardized protocol by Tornin et al
[55] and correspond to solution depths of ca. 0.5 cm for the
12-, 24- and 48-well plates, and ca. 0.3 cm for the 96-well
plate. Experiments were performed on three separate days,
with three technical replicates each. Quantification of hydro-
gen peroxide in plasma-treated PBS (pPBS) was performed
with the Fluorometric Hydrogen Peroxide Assay Kit from
Sigma-Aldrich (MAK165–1KT), according to the supplier’s
instructions. The samples were diluted according to a 1:100
ratio in untreated PBS. The fluorescence intensity was meas-
ured at an excitation wavelength of 540 nm and an emission
wavelength of 590 nm with the Tecan Spark Cyto 600. A
standard curve was used to determine the concentration.
For nitrate and nitrite, quantification in pPBS was done
with the Nitrate/Nitrite Colorimetric Assay Kit from Cayman
Chemical (780001), according to the supplier’s instructions.
The samples were not diluted, except for the samples treated
in a 96 well, which were diluted according to a 1:10 ratio in
untreated PBS. The absorbance was measured at 540–550 nm
with the Tecan Spark Cyto 600. Calibration curves were used
to determine the concentrations.
−1 penicillin, 100 µg ml
The human cancer cell line A375
2.2.2. Cell experiments.
(melanoma, ATCC®) was used in this study to determine the
effect of the treatment on the cell viability. Cancer cells were
cultured in Dulbecco’s Modified Eagle Medium (10938025,
Gibco) supplemented with 10% fetal bovine serum (Gibco),
−1 streptomycin (15140122,
100 U ml
Gibco) and 4 mM L-glutamine. The kINPen® MED was used
to treat 2D monolayers of 96000 A375 cells per well in a
24-flat well plate (10062–896, Avantor) with or without gas
shield. The plasma setup was controlled by an automated
stage using the program WinPC-NC. Hoechst (200 nM, 62249,
Life Technologies) and Cytotox Green (60 nM, 4633, Essen
Bioscience) were added to measure both confluence and cell
death with fluorescence imaging at 4 h, 24 h, and 72 h.
Experiments were performed on three separate days, with
three technical replicates each.
3. Results and discussion
3.1. General modeling case
As the general case in our investigation, we chose the treat-
ment of a 24-well plate containing 1 ml of liquid. Figure 2(A)
shows the calculated stationary flow field in the gas phase,
4
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
Figure 2. Calculated flow field (A) and air density (B) for the treatment of a 24-well plate, without (left panels) and with (right panels) gas
shield. In (A), white arrows represent the flow field vectors, while in (B) they represent the main paths through which air mixes with the jet
effluent.
both without and with the gas shield (left vs right panel). As
expected, the gas flow out of the shielding gas device com-
pletely envelopes the plasma effluent. As shown previously
[43], without a gas shield the argon flow from the jet quickly
displaces the gas that was initially in the well. This does not
mean that the well becomes completely devoid of air; the vor-
tex created in the well traps some of the surrounding air and
mixes it with the effluent, keeping e.g. the N2 density in the
−3. Adding a gas shield provides a
well around 1.8 × 1017 cm
constant inflow of air into the well, keeping the concentrations
of O2 and N2 much higher, as shown in figure 2(B).
The kINPen produces
3.1.1. N2/O2 in the plasma effluent.
a plasma plume of around 10 mm starting from the outlet of
the jet nozzle. In this region, ambient species N2, O2 and H2O
are converted by electron impact reactions or reactions with
excited Ar atoms into the primary reactive species that drive
the RONS formation in the effluent [56]. For this reason, we
are most interested in the conditions in the immediate effluent
of the jet, resulting from mixing of the plasma effluent with the
ambient. Because of the high velocity of the gas flow exiting
the jet, axial convection is the dominant transport mechanism
over radial diffusion, even when turbulent mixing is taken into
account. We can thus expect strong radial concentration gradi-
ents in this region. To obtain a picture of the entire effluent as
opposed to, e.g. only the very center on the symmetry axis,
we focus on the conditions at five different radial positions
spanning the entire width of the jet nozzle. Figure 3(A) depicts
the N2 number density at these five radial positions, as a func-
tion of distance from the pin electrode. O2 is not depicted here,
but behaves in the same way as N2, with absolute values four
times lower, reflecting the N2/O2 ratio in air. Near the edge
of the jet effluent, at r = 0.75 mm, the N2 number density
rises very steeply immediately when the gas exits the jet nozzle
(see first vertical dotted line). The ambient N2 however only
reaches the center of the effluent near the end of the plasma
plume (second vertical dotted line). The number density does
reach the same level over the entire width of the effluent before
the gas flow reaches the liquid surface.
With a shielding gas (figure 3(B)), we can see the same
qualitative behavior, but the gas shield significantly increases
the absolute number densities of N2 (and by extension, O2) in
the effluent. It is not surprising that the addition of a gas shield
causes an increase of the number density of these species,
especially near the edge of the jet effluent. However, even in
the core plasma region the difference reaches up to two orders
of magnitude, depending on the axial position. In diagnostic
investigations of the kINPen, a gas shield is sometimes used
not to exclude influence from the surrounding atmosphere,
but to simulate the presence of a surrounding atmosphere of
known composition [16], or to provide a surrounding atmo-
sphere when directing the jet into a closed-off chamber for
e.g. FTIR measurements [31, 57, 58]. Our simulations reveal
that mixing of the jet effluent with the gas surrounding it is
significantly enhanced in the entire plasma effluent, when that
5
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
we performed these simulations for three different ambient
temperatures (283 K, 293 K, and 303 K) and for three different
ambient relative humidities (0%, 50% and 100%), yielding a
total of nine conditions. The results are shown in figures 4(C)
and (D). For clarity, only the density at a radial position of
0.4 mm is shown here, halfway between the center and the
edge of the plasma jet effluent. The changes at this radial pos-
ition can be used as a measure for the other radial positions in
the effluent, as observed in figures 4(A) and (B).
Two main conclusions can be drawn from figure 4. First,
we can see that the gas shield reduces the H2O concentration
throughout the effluent. Indeed, since the gas shield is com-
posed of dry air, it causes the H2O concentration in the efflu-
ent to drop up to a factor 20 compared to the case without gas
shield. Still, as observed from figure 4(D), it does not elim-
inate the variation in the H2O concentration caused by differ-
ent ambient conditions. In fact, the relative difference stays
unchanged: both with and without the gas shield, there is a
factor three difference between H2O density in the effluent at
an ambient temperature of 283 K and 303 K. A second obser-
vation is that the ambient temperature is by far the main cause
of the different H2O densities in the effluent, both with and
without the gas shield. Indeed, while the temperature determ-
ines which H2O concentration in the air corresponds to a cer-
tain relative humidity, our model predicts almost the same rise
in H2O density for 0% and 100% relative humidity at each
temperature. Therefore, the atmospheric water vapor cannot
be the cause of the variation.
An additional source of H2O in this treatment setup is the
water that evaporates from the treated liquid surface. As gas
from the jet flows over this surface, it takes with it the vapor
that is present just above the surface and mixes it into the
vortex present in the well. The vapor pressure of water is
determined by the temperature, implemented into our model
via Antoine’s law. To confirm that it is the evaporated water
that gives rise to the H2O present in the jet effluent as opposed
to the ambient water, we adapted our model so that it treats
the H2O originating from different sources (i.e. from impur-
ities in the feed gas, from the ambient atmosphere and from
evaporation at the liquid surface) as different species. The res-
ults are presented in figure 5. Indeed, most of the water vapor
in the jet effluent is evaporated H2O from the treated liquid.
Even without the gas shield, the ambient H2O is kept relat-
ively far away from the effluent, due to the backflow created
by the well. In a way, this backflow induced by treatment of a
well already acts as its own gas shield. Implementation of the
actual shielding gas device then enhances the shielding effect,
keeping ambient species even further away from the effluent
(cf figure 5(A)). Additionally, the gas shield changes the flow
field so that most evaporated water is guided outside of the
well, decreasing the concentration of water vapor in the jet
effluent (cf figure 5(B)).
Apart from using a gas shield, another approach that can be
used to reduce the effect of (different amounts of) water vapor
in the ambient is by admixing H2O into the feed gas of the
plasma jet, in much larger amounts than the impurity levels
typically already present. It was shown in literature that this
admixed water vapor has a much larger influence than different
Figure 3. Calculated N2 number density in the jet effluent without
(A) and with (B) a shielding gas. The first (x = 0.35 cm) and second
(x = 1.35 cm) vertical dotted lines indicate the jet nozzle and the
end of the plasma plume, respectively, while the horizontal dotted
line indicates the number density in the surrounding air (1 atm and
293 K).
surrounding gas originates from a gas shield as opposed to the
ambient atmosphere. This means that caution must be taken
when diagnostics in the presence of a gas shield are compared
to experiments without a gas shield, as they do not entail the
same (quantitative) conditions in the active plasma zone.
The main incentive to
3.1.2. H2O in the plasma effluent.
employ a gas shield is to limit the variation caused by differ-
ent atmospheric conditions. The concentration of water in the
atmosphere depends on the temperature and relative humid-
ity. Hence, unlike O2 or N2, the amount of H2O that mixes
with the kINPen effluent can change from day to day, irre-
spective of the setup. The presence of H2O substantially influ-
ences the CAP chemistry and can change the biological effects
of CAP treatment [16, 59]. Figures 4(A) and (B) illustrate the
mixing of ambient H2O with the plasma jet effluent in the
same way as plotted for N2 in figure 3, i.e. for different radial
positions. In addition, to assess the efficacy of the gas shield
in eliminating variation caused by the surrounding humidity,
6
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
Figure 4. Calculated H2O number density in the jet effluent, for different radial positions (A) and (B) and different ambient conditions (C)
and (D). Results are shown for the case without (A) and (C) and with (B) and (D) gas shield. The first (x = 0.35 cm) and second
(x = 1.35 cm) vertical dotted lines indicate the jet nozzle and the end of the plasma plume, respectively. The horizontal dotted line,
indicating the number density in the surrounding air (at 1 atm and 293 K), is only shown for (A) and (B), as it changes for different
conditions. The different relative humidities in (C) and (D) overlap.
Figure 5. Calculated H2O number density in the jet effluent originating from different sources. 2D profiles show the H2O originating from
the ambient humidity (A) or evaporation (B), both without and with gas shield. White arrows represent the main paths through which water
mixes with the jet effluent. 1D profiles show the number densities originating from the different sources, i.e., ambient humidity, evaporation,
and feed impurity, at a radial distance of 0.4 mm, without (C) and with (D) gas shield.
7
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
humidities surrounding the effluent [57], effectively making a
variation in ambient humidity less relevant. When humidify-
ing the feed gas (or shielding gas), the humidity should how-
ever still be actively controlled with e.g. a hygrometer [57],
in order to obtain precise, reproducible treatments. Usually,
H2O admixtures are introduced into the feed gas of a plasma
jet by leading (part of) the dry feed though a water bubbler.
Since the vapor pressure of the water, and thus the amount
of gaseous H2O, depends on the temperature, different ambi-
ent temperatures can still influence the amounts of H2O in the
plasma, unless the bubbler is temperature controlled. In addi-
tion, depending on the liquid volume, continuous evaporation
from the bubbler will cause the liquid water to cool down,
changing the amount of H2O in the gas over time. In practice,
the kINPen is mostly used without admixed H2O, and feed gas
impurity forms only a small contribution. In these cases, most
of the H2O entering the active plasma zone originates from the
ambient, and even more from evaporated water when treating
a liquid-filled well (cf figure 5; note the logarithmic scale).
Thus, it should be kept in mind that different ambient condi-
tions, and especially the temperature, can cause variations in
the treatment [16].
3.2. Effect of the treated well size
In the previous section we demonstrated that the backflow
created by the treated well plays a role in determining the
conditions in the plasma jet effluent, as it can induce a ‘self-
shielding’ effect. In literature, different well sizes are used for
treatment of liquid and/or cells with a plasma jet. For the gen-
eration of plasma-treated liquids, even larger containers like
beakers or petri dishes are often treated. To investigate the
effect of the substrate geometry, we simulated the plasma jet
over a 12-, 24-, 48-, and 96-well plate.
In this section, we
3.2.1. Well size effect without gas shield.
investigate how the geometry of the treated well influences the
conditions experienced by the plasma jet, and thus the treat-
ment itself. Figure 6(A) shows the N2 density as calculated for
the treatment of the four different well sizes, without a shield-
ing gas. To indicate how the well geometry influences the
flow field, we also show the main streamlines originating from
the jet.
Figures 6(B) and (C) depict the density of N2 and H2O
in the jet effluent between the pin electrode and the liquid
surface, for the different well types, as well as for a free jet.
The difference in behavior with the free jet (i.e. without well
plate) will be discussed below. Like in figure 5, densities are
shown for a radial position of 0.4 mm, but the behavior can
be extended to the entire width of the effluent (cf figure 3).
Clearly, the chosen well type significantly affects the N2 dens-
ity in the gas phase. As the well diameter decreases, the vortex
in the well becomes more confined. This causes an increase
in the velocity by which the backflow exits the well, and dir-
ects the backflow less towards the jet, effectively improving
the ‘self-shielding’ effect of the well-induced backflow. As a
result, less ambient species, like N2, enter the well and mix
with the effluent. This trend is clearly visible up to the 48-well
plate. For the 96-well plate, however, the N2 density in the
well and the effluent is the highest of all simulated wells (see
figure 6). The reason for this is twofold. First, because the well
is so small, the gas flow is more turbulent than for the other
wells, causing more turbulent mixing in the region between
the well interior and the surrounding gas. For example, the
gas flow exiting the 48-well plate has a maximum turbulent
−4 Pa·s, while for the 96-well
dynamic viscosity of 1.8 × 10
−4 Pa·s, causing a factor 9 and a factor
plate this is 7.1 × 10
32 increase, respectively, in the diffusivity of N2 over the nor-
mal molecular diffusion constant (cf section 2.1.2). Secondly,
whereas the other well types have the same depth (16.5 mm),
the 96-well plate is more shallow (10.9 mm). Mohades et al
[60] have previously shown that the so-called rim-height of a
well influences the amount of ambient gas that enters it, when
treated with a plasma jet. As this rim height is lower for the
96-well plate than for the other investigated well plates, more
N2 is able to reach the interior of the well. It should be noted
that, since the rim height depends on the amount of liquid in
the well, the results from figure 6(B) will be quantitatively dif-
ferent for smaller or larger liquid volumes. Indeed, more liquid
will result in a lower rim height, causing more ambient gas to
enter the well. We investigated the volumes typically used in
experiments, as e.g. outlined in the standardized protocol by
Tornin et al [55].
While the N2 in the effluent originates from the surrounding
atmosphere, the H2O in the effluent is mainly evaporated water
from the treated liquid, as explained in the previous section. As
the simulations were performed for the same temperature and
thus the same vapor pressure, the H2O densities in the efflu-
ent are similar regardless of the well type. Hence, the curves
overlap, with only a small difference for the 96-well plate.
To assess whether
the
3.2.2. Experimental validation.
change in effluent conditions caused by the well geometry in
fact leads to a different treatment result for the treated liquid,
we measured the concentration of long-lived RONS, i.e. H2O2,
HNO2 and HNO3, in PBS after 60 s of treatment with the kIN-
Pen. Figure 7(A) shows the measured number of moles of each
measured species in the different wells. As can be seen, the
smaller wells take up less RONS. Yan et al [61] also showed
that the choice of well type, when treating a liquid with the
kINPen, can change its reactive species uptake, and we see the
same trend as the one they reported [61]. However, while Yan
et al [61] attributed the lower species uptake by the smaller
wells to the decreasing surface area through which the liquid
can exchange species with the gas, figure 7(B) clearly demon-
strates that this cannot be the only explanation. If the surface
area was the only cause, the RONS uptake per unit of sur-
face area should be the same for each well type. Instead, more
RONS seem to be taken up by the liquid in the smallest well
per unit of surface area than by the largest well, and the surface
area of the liquid and its reactive species uptake do not vary in
the same way for the different well plates. Indeed, whereas
the 12-well plate contains twice as much H2O2 as the 48-well
plate, its liquid surface area is four times as large. Since the
8
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
Figure 6. Influence of the well type on the flow field and N2 density in 2D (A), and on the N2 (B) and H2O (C) density profiles in the jet
effluent in 1D, without a shielding gas at 293 K. The flow lines in (A) originating from the jet are plotted in white. In B and C, comparison is
also made with the result for a free jet. The H2O density curves for the 12-, 24- and 48-well plates overlap, since the plasma effluent mainly
mixes with evaporated water from the well.
Figure 7. Experimentally measured RONS in the treated liquid for different well types, in absolute number of moles (A), and per unit of
liquid surface area (B). Data is shown as the mean value and standard deviation of the replicates. For easy comparison between the different
well types, the RONS uptake per surface area is plotted as fold change with respect to the 12-well plate, and the statistical significance of
∗∗∗ = p < 0.001). Note that due to their different size, the wells contained
this difference is indicated (
different volumes of liquid (cf section 2.2), meaning that the number of moles and the concentration do not follow the same trend. For
instance, while the number of moles H2O2 as measured in a 12- and 48-well plate is 0.06 µmol and 0.032 µmol, respectively, their H2O2
concentrations after treatment are 30 µM and 65 µM.
∗∗ = p < 0.01,
∗ = p < 0.05,
9
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
uptake of species by a liquid is determined by the concentra-
tion of that species above the liquid surface, this indicates that
the concentrations of the species in the gas phase above the
liquid are different, even though the sole difference is the type
of well that is treated.
The main chemical reaction leading to the formation of
H2O2 in the gas phase by the kINPen is [32, 58]:
OH + OH + M → H2O2 + M (R1)
)−0.8
(
k = 8.010
−31cm6s
−1
T
300 K
where M stands for a neutral collision partner, such as Ar.
OH radicals are formed in the active plasma by the dissoci-
ation of water. It has been reported in literature that OH rad-
ical densities produced by a plasma jet correlate linearly with
the H2O density within a range relevant for the current study
[58, 62]. From the results shown in figure 7(B), we can deduce
that a smaller size of the treated well leads to a higher gas
phase concentration of H2O2 above the liquid. However, since
the H2O density in the jet effluent is the same regardless of
the well type, according to our model (cf figure 6), the rise
in H2O2 production cannot be due to reaction (R1). Contrary
to H2O, the calculated N2 and O2 densities in the effluent
change significantly depending on the chosen well type (cf
figure 6). Increasing levels of air entering the plasma will lead
to increased production of primary and secondary RONS such
as N and NO [56]. Both species in turn react with OH via the
following reactions [30, 32]:
N + OH → NO + H (R2)
k = 4.710
−11cm3s
−1
NO + OH + M → HNO2 + M (R3)
)−2.4
(
k = 7.410
−31cm6s
−1
T
300 K
These reactions will compete with reaction (R1) and thus
reduce the amount of H2O2 formed via OH radical recom-
bination. This can explain the increasing H2O2 trend seen for
the 12-, 24- and 48-well plate seen in figure 7(B). Indeed,
decreasing N2(/O2) densities in the smaller well sizes can lead
to a higher rate of (R1), due to less competition with (R2)
and (R3). On the other hand, the 96-well does not follow this
trend. This is unexpected, since our model predicts the N2(/O2)
density in the effluent to be the highest when treating a 96-
well plate (cf figure 6), while experimentally the highest aver-
age H2O2 concentration was found in the liquid. It is possible
that our model does not yet accurately describe all mechan-
isms at play. Especially for the 96-well plate, which is a very
small system, certain assumptions in the model may play a lar-
ger role, such as the stationary liquid surface that may affect
the flow field in the gas. Still, the trends seen for the other
well types do agree well between our model and experiments.
Additionally, chemical reactions in the liquid may influence
10
−
the results. Indeed, plasma-treatment of PBS can cause forma-
with atomic oxygen [63].
tion of ClO
through reaction of Cl
−
H2O2 (and HNO2) can, in turn, react with ClO
, via e.g. the
following reactions:
−
H2O2 + ClO
− → Cl
− + H2O + O2 (R4)
NO2
− + ClO
− → Cl
− + NO3
− (R5)
−
However, the kINPen, using Ar as the feed gas, leads to
−
in PBS compared to other
much smaller amounts of ClO
CAP devices, e.g., the COST-jet operating with He/O2 as feed
gas [64]. In addition, Van Boxem et al [32] showed the sta-
bility of H2O2 in PBS for 2 h after treatment with the kIN-
cannot
Pen. Still, the possibility of a small production of ClO
be excluded. The surface-to-volume ratio of the liquid in the
investigated wells (i.e. the surface through which RONS can
enter the liquid, compared to the volume in which they can
react) is nearly the same for all wells, except for the 96-well
plate. Thus, this may be an additional reason why its results
in figure 7 do not follow the qualitative trend explained above.
Finally, it must also be noted that the experimental results for
the 96-well plate in figure 7 simply have a very large uncer-
tainty. For example, the difference in HNO2 uptake per sur-
face area of the 96-well plate, compared to that of the 12-well
plate, is not statistically significant. This large uncertainty can
be explained in two ways. First, we showed in figure 6 that, due
to the higher turbulence and lower rim-height, more ambient
gas mixes with the plasma effluent for the 96-well plate com-
pared to the other well plates. This means that treatment of a
96-well plate will be inherently more susceptible to different
relative humidities compared to the other well types. Second,
the diameter of a 96-well is only 4.3 times larger than that of
the kINPen nozzle. Small deviations from a perfectly centered
position above the well may thus significantly influence the
flow field in the well. As our model describes an ideal geo-
metry, this may additionally form an explanation for why our
model and our experiments do not follow the same trend for
the 96-well plate only.
The trends plotted in figure 7 for HNO2 and HNO3 are less
straightforward to explain. Like for H2O2, the liquid takes up a
higher amount per unit of surface area for decreasing well size,
indicating higher gas phase concentrations. However, because
the formation of HNO2 and HNO3 requires the presence of
both N2 and O2, and the concentrations of these species in
the effluent decrease with smaller well sizes, as shown in
figure 6(B), one would expect HNO2 and HNO3 to follow this
trend. Instead, however, we see the opposite. HNO2 is mainly
formed via reaction (R3), while HNO3 is formed through [30]:
NO2 + OH + M → HNO3 + M (R6)
k = 4.610
−29cm6s
−1
(
T
300 K
)−0.8
(
exp
− 1180
T
)
where NO2 in turn is formed from NO. It was noted by Van
Gaens et al [33] that NO2 production by the kINPen rises with
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
Figure 8. Influence of the well type on the flow field and N2 density in 2D (A), and on the N2 (B) and H2O (C) density profiles in 1D, in the
jet effluent, with a shielding gas at 293 K. The flow lines in (A) originating from the jet and the gas shield are plotted in white and black,
respectively. In (B) and (C), comparison is also made with the result for a free jet with gas shield. For the 48-well plate, a neighboring well
is shown in (A), for clarity. For the other well types, the inclusion of a neighboring well has no effect on the flow field. The N2 density
curves for the 12-, 24- and 96-well plates overlap, since the plasma effluent mainly mixes with air supplied by the shielding gas.
increasing amounts of N2 or O2 only up to a certain point:
above 0.1% O2 content and 0.15% N2 content, the NO2 pro-
duction decreases again. Mohades et al [60] calculated lower
concentrations of NO and NO2 near the liquid in geometries
where more N2/O2 from the ambient diffused into the well.
Though both these observations are in line with the behavior
we see here, they cannot be directly compared. In [33], N2
and O2 were supplied to the jet as admixtures in the feed
gas, while in [60] a different type of plasma jet was sim-
ulated. A more in-depth analysis is thus necessary to fully
explain the observed trends, which will be part of future work.
Nevertheless, both our experimental and computational results
show that the choice of treated well type can have a signific-
ant influence on the effluent conditions and, by extension, the
treatment itself. This is important to keep in mind for plasma
medicine applications.
3.2.3. Well size effect with gas shield. We also investigated
how the chosen well geometry impacts the effluent conditions
when a gas shield is employed. The results are depicted in
figure 8 in a similar way as figure 6, for the different well types,
11
as well as for a free jet. The difference in behavior with the
free jet will also be discussed below. When implementing a
gas shield, it is clear that the flow field changes drastically for
different treated wells. An extreme case is the treatment of a
48-well plate. Because the edge of the well in this case is posi-
tioned at a similar radial position as the nozzle of the shielding
gas device, the shielding gas does not blow into the well being
treated, but instead into the wells surrounding it. The neigh-
boring well for the 48-well plate is shown in figure 8(A) to
illustrate this. It should be noted that this second well is in
fact treated in the model as a ring-shaped cavity surrounding
the treated well, because the model is axisymmetric. However,
this still illustrates the effect on the neighboring wells. For the
other well types, figure 8(A) shows the results of simulations
without the neighboring well present. In these cases, we also
performed simulations that included the neighboring well, but
its inclusion had no effect on the results, and the flow did not
enter the neighboring well like for the 48-well plate.
We can thus expect that for the 48-well plate the gas shield
will not actually provide a shielding effect, as is clear from
the flow field in figure 8(A). Indeed, when comparing figure 8
with figure 6 we can see that there is little difference between,
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
respectively, the N2 and the H2O density for the 48-well with
or without a gas shield. For the other wells, the shielding gas
does envelop the plasma jet effluent. Because in these cases,
almost all N2 that mixes with the effluent is supplied by the
shielding gas, the N2 density in the effluent is the same for the
12-, 24- and 96-well plate. The H2O density is also very similar
for these three well types. As discussed in section 3.1, the H2O
that mixes with the effluent is mainly evaporated water from
the treated surface. The H2O density is lower with shielding,
because the flow field caused by the shielding gas for the three
wells is such that it guides the evaporated water away from the
jet nozzle, instead of towards it like in the cases without shield-
ing. Overall, our results indicate that the gas shield is able to
induce a controlled environment surrounding and mixing with
the jet effluent, regardless of the treated well type, although in
some cases, like for the 48-well plate in this setup, the sub-
strate geometry can influence the flow field in such a way that
the shielding gas does not at all behave as intended.
The fact
3.2.4. Unintended effects on neighboring wells.
that the shielding gas blows into the wells surrounding the
treated well, for the 48-well plate shown in figure 8, does not
only cause the gas shield to not operate as intended. Along
with the shielding gas, much of the effluent from the plasma
jet passes through the neighboring well. This means that in this
geometry, wells could be unintendedly affected by the treat-
ment of other wells close by.
The RONS produced by a plasma jet can be divided into
three groups [43]. These are (i) short-lived species, which
quickly react away after the end of the active plasma zone
(some even before reaching the liquid), (ii) long-lived species
with a high Henry’s constant, and (iii) long-lived species with
a low Henry’s constant. For a flow field such as that of the
48-well plate in figure 8, short-lived species are unlikely to
reach the neighboring wells, because they react away quickly.
Long-lived species with a high Henry’s constant can survive
long enough, but are also unlikely to reach the neighboring
wells because they will mostly be taken up by the liquid in the
treated well, more so for higher Henry’s constants. However,
long-lived species with a low Henry’s constant, that are thus
only taken up by the treated liquid in small amounts, are able to
reach the surrounding wells and dissolve in the liquid therein.
−3
To illustrate this, we performed simulations where a 1015 cm
O3 was supplied at the inlet of the plasma jet, similar to pro-
duced amounts of O3 reported in literature [56], while the
neighboring well (also shown in figure 8(A)) was filled with
liquid. Figure 9(A) shows the O3 density in the gas and liquid
phase, for a 48-well plate. Indeed, the flow field in this case
guides a substantial amount of O3 to the liquid surface in the
neighboring well, where it subsequently dissolves.
The effect occurs when the radius of the treated well is com-
parable to (or smaller than) the radius of the gas shield nozzle,
in combination with a strong enough backflow from the well
to push away the shielding gas. For instance, for a 96-well
plate with 20 mm gap between the liquid and the plasma jet
Figure 9. Ozone density in the gas and liquid phase of both the
treated and neighboring well when the setup geometry causes the
shielding gas to blow into the wells surrounding the treated well. (A)
48-well plate with a treatment gap of 2 cm. (B) 96-well plate with a
−3 is supplied
treatment gap of 1.5 cm. An ozone density of 1015 cm
by the jet, based on reported produced densities in literature [56].
nozzle, as discussed in section 3.2.3, the flow does not affect
the neighboring well, while for a smaller gap of 15 mm, and
thus a stronger backflow, the same effect as for the 48-well
plate with 20 mm gap can be seen (cf figure 9(B)).
3.2.5. Comparison with the free jet. As is clear from the pre-
vious sections, the choice of treated substrate will influence
the treatment of the substrate itself. To emphasize this influ-
ence, figures 6(B) and 8(B) also show the density profiles of
N2 and H2O as calculated for the free jet, i.e., without treated
substrate. Due to the lack of a backflow from the treated well,
the N2 density without a shielding gas is significantly differ-
ent for the free jet (cf black curve in figure 6(B)), which con-
firms again that caution should be taken when comparing res-
ults (like produced RONS densities) for the free jet with those
gathered when treating a well. With a shielding gas, the N2
density is independent of the well geometry (except for the
48-well), and the same as for the free jet, since in these cases
most N2 mixing with the effluent is supplied by the shielding
gas, confirming again that the shield is able to create a con-
trolled environment. However, it should be kept in mind that
the effect is not the same as a controlled atmosphere around
the jet.
The H2O density without shielding is lower for the free jet
than for the treatment of any well type, which is to be expec-
ted, since without a treated liquid surface no evaporated water
is present, and all H2O in the effluent stems from the ambient
12
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
humidity. The density is however higher than the H2O density
coming from the ambient shown in figure 5(B), because the
‘self-shielding’ effect caused by the induced backflow from
the well is absent here. This also means that the free jet is
far more susceptible to varying ambient humidity, which is
shown in figure 10(A). With the gas shield, one would expect
that the lack of an evaporating liquid surface beneath the jet,
and the presence of a shielding flow separating the effluent
from the ambient, would cause the H2O density to barely rise
above the impurity level in the feed gas. Surprisingly, as shown
in figure 10, the H2O density is similar with and without a
gas shield. The position from which concentrations start to
rise is increased slightly, but by the end of the plasma plume
(x = 1.35 cm) the densities are at the same level as without
shielding gas. This indicates that the gas shield does not effi-
ciently shield the jet effluent from the surrounding atmosphere
at all. In the following section, the reason for this will be
elucidated.
3.3. Effect of gas shield geometry
In literature, the effectiveness of using a gas shield was shown
by Reuter et al [17], both experimentally and with a simple
CFD model. The most notable difference between their invest-
igation and our study is that a different shielding gas device
was used, with a different geometry. While in [17] the device
was made in-house, we opted (with the interest of reproducib-
ility and a clinical setting in mind) for a device specifically
made for the kINPen by Neoplas [65]. Importantly, when we
adapt our model to the geometry shown in [17], our computa-
tional results show efficient shielding for the free jet: any influ-
ence from the ambient atmosphere is pushed back by the gas
shield until after the plasma plume has ended as illustrated in
figure 11(B). Moreover, the different shield geometry also sig-
nificantly changes the conditions in the plasma effluent above a
well plate, as shown in figure 11(C): the rise in H2O only hap-
pens after the end of the plasma plume (i.e. providing better
shielding), and the amount of N2 that mixes with the effluent
is reduced.
To con-
3.3.1. Experimentally testing shielding efficiency.
firm that the shielding gas device used in our model so far
does not provide efficient shielding, we experimentally tested
different shielding gas compositions for treatment of a liquid
sample, followed by measurement of long-lived RONS in
the treated liquid. It has been shown in literature that when
the plasma jet effluent is efficiently shielded from the envir-
onment, different shielding gas compositions can signific-
antly change the RONS-composition of a treated liquid. For
example, by using a shielding gas devoid of N2, the production
of nitrogen-containing species can effectively be prevented
[19, 66]. However, the results in figure 12(A) show that HNO2
and HNO3 are still produced in large amounts when an Ar/O2
shielding gas is used. Moreover, the production of the meas-
ured RONS is similar regardless of the shielding gas compos-
ition. This shows that despite the shielding, species from the
Figure 10. H2O density in the jet effluent as calculated for the free
jet, without (A) and with (B) gas shield, for three different ambient
temperatures and three different relative humidities. As all water
mixing with the jet originates from the ambient humidity, there is no
rise in H2O density for 0% relative humidity (see dotted line), and
the curves for the three temperatures overlap.
ambient, like N2, still make it into the jet effluent in significant
amounts with this gas shield. This is also reflected by the meas-
ured response of melanoma cell line A375 to the plasma treat-
ment, shown in figure 12(B): there is little difference between
the results for the different shielding gas compositions. The
difference with the treatment without shielding gas mirrors
the measured H2O2 uptake, shown in figure 12(A). Note that
the RONS measurements and the cell experiments were per-
formed in different liquids. For the species measurements,
PBS was used to prevent consumption of the long-lived RONS
before measurement. For the cell experiments, the cells were
treated in cell medium, which contains organic biomolecules
that can act as scavengers for plasma-produced RONS. The
response seen in figure 12(B) is thus not only the result of the
primary species produced by the plasma, but also of secondary
13
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
Figure 11. Calculated flow field (A) and H2O number density in the jet effluent, for a free jet (B), for the shielding gas geometry used in
[17]. Number density of N2 and H2O in the effluent above a 24-well plate is also shown for both the shield from [17] and the shield from
Neoplas (used throughout the paper), for 293 K and 50% relative humidity (C). It is important to note that the shape of the liquid water
surface was kept unchanged for the different shield geometry. This will not be entirely correct: the different geometry changes the flow field,
and e.g. slows stagnation of the effluent velocity, which will in turn induce a different liquid surface shape. However, we believe this effect
is of minor importance for the message of this figure.
Figure 12. (A) Experimentally measured RONS uptake in treated liquid after 60 s of treatment time, for different shielding gas
compositions. (B) Response of A375 cancer cells to the treatment, using the same treatment conditions as in (A). Data is shown as the mean
value and standard deviation of the replicates. Cell death was measured 4, 24 and 72 h after the treatment.
species such as oxidized biomolecules, which also affect the
cancer cells [67].
3.3.2. Underlying reason for the different shielding efficien-
The question remains what the reason is for the large
cies.
difference in shielding efficacy seen when comparing figure 11
with figure 4(B). One of the most notable differences between
the two shield geometries is that the radial distance between
the shielding nozzle and the jet nozzle is larger for the shield
geometry used in the present study. At first glance, one would
expect that increasing the radial position of the shielding
nozzle could be beneficial, as it keeps the ambient further away
from the plasma jet. However, as was visible in figure 2, a
vortex forms in the region between both nozzles. This vor-
tex effectively ‘pulls in’ species towards the jet effluent. In
fact, this vortex acts in a similar way to the recirculation zone
generated by bluff-body stabilization, used in combustion. In
this field, the formation of a recirculation zone between the
two nozzles is used to enhance fuel-air mixing and stabilize
the flame [68, 69].
For the free jet, the species that are pulled in are those from
the ambient atmosphere, explaining why in figure 10 there is
only a small difference between the case with and without
gas shield. When the jet is positioned above a well plate, the
species that are pulled in by the vortex are mainly those in the
backflow coming from the well (explaining why the evapor-
ated water so significantly affects the H2O density in the jet
effluent, while for the shield geometry in figure 11 this is far
less the case). However, species from the ambient also still
make it into the effluent, as seen in figures 5 and 12. Finally,
for a 96-well plate, we can see in figure 8 that backflow coming
14
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
causes formation of a vortex, like discussed previously in
section 3.3.2, that accumulates species from the ambient atmo-
sphere. As the radial distance decreases, this vortex becomes
smaller, and the density of ambient species in it decreases.
At very small radial distance between the nozzles, the vor-
tex disappears completely. This trend is clearly visible for the
H2O density, on which the radial distance (and by extension,
the size of the formed vortex) has a very large influence. For
larger radial distances, the H2O density in the plasma efflu-
ent rises much faster. Additionally, larger radial distances also
decrease the axial position at which the rise in density starts.
For the largest simulated radial distance (8 mm), the H2O dens-
ity in the plasma effluent (at a radial position of 0.4 mm) starts
rising at 7 mm from the pin electrode. When the radial distance
is small enough to prevent formation of the vortex, this rise
in H2O density only occurs near the very end of the plasma
plume. This means that smaller radial distances between the
jet and shielding nozzle, or in other words, a more confined
shielding curtain around the plasma effluent, provides a better
shielding effect. The same general trend is present for the N2
density. A larger radial distance between the jet and shield-
ing nozzle decreases the distance from the pin electrode at
which the N2 density in the plasma effluent starts to rise, and
it increases the amount of N2 in the effluent. Only at the end
of the visible afterglow, around 10 mm from the pin electrode,
the behavior is different, and no clear trend is visible.
Figure 14(C) shows the influence of the width of the shield-
ing gas nozzle. It can be seen that increasing the gas cur-
tain width causes a strong drop in the amount of H2O that is
able to mix with the plasma effluent, as the ambient is kept
further away. This trend is opposite to that for an increasing
radial distance between the plasma jet nozzle and the shield-
ing gas nozzle. Indeed, increasing the radial position of the
shield gas nozzle also keeps the ambient further away from
the plasma jet, but this actually increases the amount of ambi-
ent H2O that mixes with the jet effluent, due to the vortex cre-
ation. This further emphasizes the importance of the vortex
that can form between the jet effluent and the shield effluent
(note that in figures 14(B) and (C) the outer diameter of the
gas shield is the same for each simulated condition, respect-
ively). For the N2 density, two behaviors can be seen. For small
shield nozzle widths (0.5–1 mm) the N2 mixes with the plasma
effluent in higher amounts compared to large nozzle widths
(2–8 mm). This can be attributed to the fact that, for a large
nozzle width, the flow is far less turbulent, causing much less
turbulent mixing between the N2 from the shielding gas and
the plasma effluent. These two behaviors are in fact also vis-
ible for the H2O density, indicating that turbulent mixing also
plays a role in the mixing of the plasma effluent with the ambi-
ent atmosphere.
Finally, figure 14(D) depicts the influence of the axial pos-
ition of the shielding gas nozzle. The N2 density in the plasma
effluent is mostly unaffected by this change in geometry. For
the H2O density, increasing the axial position (i.e., if the shield
nozzle position is higher than the jet nozzle position in the
geometry of figure 14(A)) has only a small effect. Lowering
the axial position of the shield nozzle however significantly
reduces how quickly the ambient gas, as seen by the H2O
Figure 13. H2O number density in the jet effluent as calculated for
the 96-well plate with shielding gas.
from the well is directed outwards, unlike for the other well
types where the induced backflow is directed upwards. This
means that the gas next to the vortex between the jet nozzle and
the shield nozzle is mainly the ambient atmosphere. Indeed,
as shown in figure 13, the ambient humidity plays a far larger
role in this case than for the other well types, more akin to the
situation without a treated well.
To
3.3.3. Effect of the shield nozzle position and width.
investigate the effect of different gas shield geometries in more
detail, we present here a systematic study of the effect of the
gas shield geometry. Indeed, since its introduction by Reuter
et al [17], different gas shield geometries have been used in
both modeling and experimental works [19, 35, 44]. To the
best of our knowledge, however, the influence of a change in
shielding gas geometry on its efficacy has rarely been repor-
ted. In the following, we will use our model to investigate
the effect of the gas shield geometry on the conditions in
the plasma effluent, and its ability to shield the plasma efflu-
ent from the ambient. To investigate this in a general way,
we simplified the model geometry, as shown in figure 14(A).
Like in the previous sections, we will discuss the H2O density
and the N2 density in the effluent (while O2 acts in the same
way as N2).
Four main parameters can be adjusted in the shielding gas
device geometry, i.e. (i) the axial and (ii) radial position of the
shielding nozzle compared to the plasma jet nozzle, (iii) the
width of the shielding nozzle, and (iv) the flow direction of
the shielding gas relative to the plasma effluent. This section
will present the effect of the first three parameters, while the
flow direction will be discussed in section 3.3.4 below.
Figure 14(B) illustrates the effect of the radial position of
the shielding gas nozzle compared to the outer diameter of the
plasma jet nozzle. A large radial distance between both nozzles
15
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
Figure 14. Effect of the gas shield geometry on the N2 density and H2O density in the plasma effluent. (A) shows the basic geometry of the
jet nozzle and gas shield nozzle. The investigated parameters are (B) the radial position, (C) width and (D) axial position of the gas shield
nozzle relative to the plasma jet nozzle. The 2D plots show the ambient H2O density, with white arrows representing the flow field. In (A)
the black box indicates which part of the geometry is shown in these 2D plots. The 1D plots depict the N2 and H2O number density in the
plasma effluent, at a radial position of 0.40 mm, as a function of distance from the pin electrode for the different simulated gas shield
geometries. In (C), the curves for 2.0, 4.0 and 8.0 mm overlap. Note that the color scale is logarithmic to clearly visualize the differences.
density, mixes with the plasma effluent. This is to be expec-
ted, because it simply takes longer before the ambient can
start diffusing towards the jet effluent. Though the results
in figure 14(D) are only shown for the case where a vortex
between the jet nozzle and shield nozzle forms, the trends are
the same for the case where no vortex can form.
It should be kept in mind that these simulations were per-
formed for the free jet. When treating a well plate, the back-
flow from the well can in some cases push back the shield-
ing gas, which may cause unintended effects, as discussed
in section 3.2.4. To emphasize this, figure 15(A) shows the
effect of the radial position of the shielding gas nozzle, like
16
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
Figure 15. Effect of the well-induced backflow on a gas shield, for different radial distances of the shielding gas nozzle. (A) Ambient H2O
density in the system. For the case with a radial distance of 8.0 mm, the neighboring well is shown, for clarity. (B) Evaporated H2O density
in the system, for the gas shield with a radial distance of 4.0 mm. White arrows represent the flow field.
also depicted in figure 14(B), but for the treatment of a 24-
well plate (with a 2 cm treatment gap). For a radial dis-
tance of 0.5 and 4.0 mm, the shielding gas flow is unaf-
fected, although in these cases most of the H2O that mixes
with the plasma effluent (and accumulates in the vortex, when
formed) is evaporated water as opposed to H2O from the
ambient air, as shown in figure 15(B). For a radial distance
of 8.0 mm, however, the radial position of the shielding
gas nozzle becomes similar to that of the well edge, and
no vortex forms. Instead, the shielding gas blows into the
wells surrounding the treated well. Additionally, although
figure 14(C) indicates that increasing the shielding nozzle
width gives a better shieling effect, a wider shielding curtain
makes it easier for the backflow from the well to push back the
shielding gas.
3.3.4. Effect of the flow direction. Apart from the position
of the shielding gas nozzle relative to the plasma jet nozzle,
and its width, the flow direction relative to the plasma effluent
can be adapted. In literature, different gas shields are depicted
that direct the shielding curtain parallel to the plasma effluent
[35], or direct it towards the plasma effluent diagonally [44,
70] or perpendicularly [17]. Figure 16 shows the influence of
the flow direction of the shielding gas, for a small radial dis-
tance between the jet nozzle and the shielding gas nozzle. It
can be seen that both the diagonal and perpendicular flow dir-
ection induce less mixing of the jet effluent with H2O from
the ambient, compared to the parallel flow. The reason is that
for these two cases there is much less turbulent mixing, sim-
ilar to the effect of the shielding nozzle width discussed in
the previous section, which causes the ambient air to penet-
rate into the jet effluent far slower. This behavior was also
noted by Gazzah and Belmabrouk [71], who reported reduced
mixing in a jet with co-flow by directing the co-flow towards
the jet effluent. This also explains the N2 profiles plotted in
figure 16: less N2 mixes with the jet effluent for the diagonal
Figure 16. Effect of the flow direction of the gas shield, relative to
the flow from the plasma jet, on the N2 density and H2O density in
the plasma effluent. The 2D plots show the ambient H2O density,
with white arrows representing the flow field. The 1D plots show the
N2 and H2O number density in the plasma effluent as a function of
distance from the pin electrode for the different simulated flow
direction of the gas shield.
and perpendicular compared to the parallel flow direction.
Note that here, the same two behaviors can be seen as in
figure 14(C).
17
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
4. Conclusion
We investigated how in vitro treatment with a plasma jet is
affected by the geometry of the treatment setup, i.e., the chosen
well type, and the use of a shielding gas device. For this pur-
pose, we developed a computational 2D-axisymmetric model
of the kINPen plasma jet above a liquid water surface, with
and without gas shielding, to investigate the mixing of the
plasma jet effluent with the ambient N2, O2 and H2O. These
molecules drive the formation of RONS by the plasma, and
thus their mixing with the plasma influences the treatment
effect. Simulations were performed for different ambient tem-
peratures and relative humidities.
Both our computational and experimental results show that
the choice of treated well type can significantly influence the
effluent conditions and, by extension, the treatment itself. The
backflow created by the treatment of a well plate plays a
role in determining the conditions in the plasma jet effluent,
as it can induce a ‘self-shielding’ effect. Because this flow
field depends on the size of the treated well, mixing of the
plasma with the ambient will be different for different treated
wells, causing changes in the RONS formation. Additionally,
because the self-shielding keeps the surrounding atmosphere
away from the plasma effluent, evaporation of water in the
treated well forms the main contributor to the H2O that enters
the plasma plume. The use of a shielding gas provides a con-
sistent supply of gas, and is able to induce a controlled environ-
ment surrounding and mixing with the jet effluent, regardless
of the treated well type. However, in some cases, like for the
48-well plate in this setup, the substrate geometry can influ-
ence the flow field in such a way that the shielding gas does
not at all behave as intended. When the radius of the treated
well is comparable to (or smaller than) the radius of the gas
shield nozzle, in combination with a strong enough backflow
from the well, the shielding gas can be pushed away. In this
case, the long-lived RONS with a low Henry’s constant, such
as O3, may enter the wells surrounding the treated well, caus-
ing unintended effects of the treatment to these neighboring
wells. Furthermore, it should be taken into account that the
flow of a gas shield can enhance the amount of N2 and O2 that
mixes with the plasma effluent. This means that caution must
be taken when diagnostics in the presence of a gas shield are
compared to experiments without a gas shield, as they do not
entail the same (quantitative) conditions in the active plasma
zone.
Finally, we systematically investigated the effect of differ-
ent gas shield geometries, i.e., the radial and axial position of
the shielding gas nozzle, its width, and the flow direction rel-
ative to the plasma jet flow. Overall, the largest effect was seen
for the radial position of the shielding gas nozzle. When this
increases, a recirculation zone can arise between the shielding
curtain and the plasma effluent, which pulls in species from
the ambient, severely changing the conditions in the plasma
effluent. In this way, we showed that the gas shield design can
have a substantial effect on the shielding efficiency.
Altogether, our results provide a deeper understanding of
how the choice of setup geometry, such as the treated well
and the use of a gas shield, can influence the conditions in the
plasma effluent and, by extension, the plasma treatment itself.
Data availability statement
All data that support the findings of this study are included
within the article (and any supplementary files).
Acknowledgments
We acknowledge financial support from the Fund for Scientific
Research (FWO) Flanders (Grants ID 1100421N, G033020N
and 1SD6522N). This article is based upon work from COST
Action CA20114 PlasTHER ‘Therapeutical Applications of
Cold Plasmas’, supported by COST (European Cooperation
in Science and Technology). We also thank I Tsonev, S Van
Hove, R De Meyer and R Vertongen for their help with the
model development, and valuable input.
ORCID iDs
Pepijn Heirman https://orcid.org/0000-0003-0210-9053
Ruben Verloy https://orcid.org/0000-0002-4248-1258
Jana Baroen https://orcid.org/0000-0002-8983-5893
Angela Privat-Maldonado https://orcid.org/0000-0002-
5616-8182
Evelien Smits https://orcid.org/0000-0001-9255-3435
Annemie Bogaerts https://orcid.org/0000-0001-9875-
6460
References
[1] Bran´y D, Dvorská D, Halaˇsová E and ˇSkovierová H 2020 Cold
atmospheric plasma: a powerful tool for modern medicine
Int. J. Mol. Sci. 21 2932
[2] Laroussi M 2018 Plasma medicine: a brief introduction
Plasma 1 47–60
[3] Bekeschus S, von Woedtke T, Emmert S and Schmidt A 2021
Medical gas plasma-stimulated wound healing: evidence
and mechanisms Redox Biol. 46 102116
[4] Boeckmann L et al 2020 Cold atmospheric pressure plasma in
wound healing and cancer treatment Appl. Sci. 10 6898
[5] Keidar M 2020 Plasma Cancer Therapy (Springer)
[6] Gorbanev Y, Vervloessem E, Nikiforov A and Bogaerts A
2020 Nitrogen fixation with water vapor by nonequilibrium
plasma: toward sustainable ammonia production ACS
Sustain. Chem. Eng. 8 2996–3004
[7] Onyshchenko I, De Geyter N and Morent R 2017
Improvement of the plasma treatment effect on PET with a
newly designed atmospheric pressure plasma jet Plasma
Process. Polym. 14 1600200
[8] Reuter S, von Woedtke T and Weltmann K-D 2018 The
kINPen—a review on physics and chemistry of the
atmospheric pressure plasma jet and its applications J.
Phys. D: Appl. Phys. 51 233001
[9] Bekeschus S et al 2019 Risk assessment of kINPen plasma
treatment of four human pancreatic cancer cell lines with
respect to metastasis Cancers 11 1237
18
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
[10] Bekeschus S, Schmidt A, Weltmann K-D and von Woedtke T
2016 The plasma jet kINPen–a powerful tool for wound
healing Clin. Plasma Med. 4 19–28
[11] Graves D B 2012 The emerging role of reactive oxygen and
nitrogen species in redox biology and some implications for
plasma applications to medicine and biology J. Phys. D:
Appl. Phys. 45 263001
[12] Privat-Maldonado A, Schmidt A, Lin A, Weltmann K-D,
Wende K, Bogaerts A and Bekeschus S 2019 ROS from
physical plasmas: redox chemistry for biomedical therapy
Oxid. Med. Cell. Longevity 2019 1–29
[13] Lietz A M and Kushner M J 2018 Molecular admixtures and
impurities in atmospheric pressure plasma jets J. Appl.
Phys. 124 153303
[14] Iséni S, Reuter S and Weltmann K-D 2014 NO2 dynamics of
an Ar/Air plasma jet investigated by in situ quantum
cascade laser spectroscopy at atmospheric pressure J. Phys.
D: Appl. Phys. 47 075203
[15] Reuter S, Tresp H, Wende K, Hammer M U, Winter J,
Masur K, Schmidt-Bleker A and Weltmann K-D 2012 From
RONS to ROS: tailoring plasma jet treatment of skin cells
IEEE Trans. Plasma Sci. 40 2986–93
[16] Winter J, Wende K, Masur K, Iseni S, Dünnbier M,
Hammer M, Tresp H, Weltmann K-D and Reuter S 2013
Feed gas humidity: a vital parameter affecting a cold
atmospheric-pressure plasma jet and plasma-treated human
skin cells J. Phys. D: Appl. Phys. 46 295401
[17] Reuter S, Winter J, Schmidt-Bleker A, Tresp H, Hammer M U
and Weltmann K-D 2012 Controlling the ambient air
affected reactive species composition in the effluent of an
argon plasma jet IEEE Trans. Plasma Sci. 40 2788–94
[18] Bekeschus S, Iseni S, Reuter S, Masur K and Weltmann K-D
2015 Nitrogen shielding of an argon plasma jet and its
effects on human immune cells IEEE Trans. Plasma Sci.
43 776–81
[19] Schmidt A, Bekeschus S, Jablonowski H, Barton A,
Weltmann K-D and Wende K 2017 Role of ambient gas
composition on cold physical plasma-elicited cell signaling
in keratinocytes Biophys. J. 112 2397–407
[20] Pasqual-Melo G et al 2020 Plasma treatment limits cutaneous
squamous cell carcinoma development in vitro and in vivo
Cancers 12 1993
[21] Schmidt A, Nießner F, von Woedtke T and Bekeschus S 2020
Hyperspectral imaging of wounds reveals augmented tissue
oxygenation following cold physical plasma treatment
in vivo IEEE Trans. Radiat. Plasma Med. Sci. 5 412–9
[22] Arndt S, Schmidt A, Karrer S and von Woedtke T 2018
Comparing two different plasma devices kINPen and Adtec
SteriPlas regarding their molecular and cellular effects on
wound healing Clin. Plasma Med. 9 24–33
[23] Metelmann H-R et al 2018 Clinical experience with cold
plasma in the treatment of locally advanced head and neck
cancer Clin. Plasma Med. 9 6–13
[24] Stratmann B et al 2020 Effect of cold atmospheric plasma
therapy vs standard therapy placebo on wound healing in
patients with diabetic foot ulcers: a randomized clinical trial
JAMA Netw. Open 3 e2010411–e
[25] Von Woedtke T, Metelmann H R and Weltmann K D 2014
Clinical plasma medicine: state and perspectives of in vivo
application of cold atmospheric plasma Contrib. Plasma
Phys. 54 104–17
[26] Dubuc A, Monsarrat P, Virard F, Merbahi N, Sarrette J-P,
Laurencin-Dalicieux S and Cousty S 2018 Use of
cold-atmospheric plasma in oncology: a concise systematic
review Ther. Adv. Med. Oncol. 10 1758835918786475
[27] Hansen L, Schmidt-Bleker A, Bansemer R, Kersten H,
Weltmann K-D and Reuter S 2018 Influence of a liquid
surface on the NOx production of a cold atmospheric
pressure plasma jet J. Phys. D: Appl. Phys. 51 474002
[28] Yagi I, Ono R, Oda T and Takaki K 2014 Two-dimensional
LIF measurements of humidity and OH density resulting
from evaporated water from a wet surface in plasma for
medical use Plasma Sources Sci. Technol. 24 015002
[29] Bruggeman P J et al 2016 Plasma–liquid interactions: a review
and roadmap Plasma Sources Sci. Technol. 25 053002
[30] Schmidt-Bleker A, Winter J, Bösel A, Reuter S and
Weltmann K-D 2015 On the plasma chemistry of a cold
atmospheric argon plasma jet with shielding gas device
Plasma Sources Sci. Technol. 25 015005
[31] Schmidt-Bleker A, Winter J, Iseni S, Dünnbier M,
Weltmann K and Reuter S 2014 Reactive species output of a
plasma jet with a shielding gas device—combination of
FTIR absorption spectroscopy and gas phase modelling J.
Phys. D: Appl. Phys. 47 145201
[32] Van Boxem W, Van der Paal J, Gorbanev Y, Vanuytsel S,
Smits E, Dewilde S and Bogaerts A 2017 Anti-cancer
capacity of plasma-treated PBS: effect of chemical
composition on cancer cell cytotoxicity Sci. Rep. 7 16478
[33] Van Gaens W, Iseni S, Schmidt-Bleker A, Weltmann K-D,
Reuter S and Bogaerts A 2015 Numerical analysis of the
effect of nitrogen and oxygen admixtures on the chemistry
of an argon plasma jet operating at atmospheric pressure
New J. Phys. 17 033003
[34] Reuter S, Winter J, Schmidt-Bleker A, Schroeder D, Lange H,
Knake N, Schulz-von der Gathen V and Weltmann K-D
2012 Atomic oxygen in a cold argon plasma jet: TALIF
spectroscopy in ambient air with modelling and
measurements of ambient species diffusion Plasma Sources
Sci. Technol. 21 024005
[35] Schmidt-Bleker A, Norberg S A, Winter J, Johnsen E,
Reuter S, Weltmann K and Kushner M J 2015 Propagation
mechanisms of guided streamers in plasma jets: the
influence of electronegativity of the surrounding gas
Plasma Sources Sci. Technol. 24 035022
[36] Norberg S A, Johnsen E and Kushner M J 2015 Formation of
reactive oxygen and nitrogen species by repetitive
negatively pulsed helium atmospheric pressure plasma jets
propagating into humid air Plasma Sources Sci. Technol.
24 035026
[37] Li J, Guo H, Zhang X-F and Li H-P 2018 Numerical and
experimental studies on the interactions between the
radio-frequency glow discharge plasma jet and the shielding
gas at atmosphere IEEE Trans. Plasma Sci. 46 2766–75
[38] Lin P, Zhang J, Nguyen T, Donnelly V M and Economou D J
2020 Numerical simulation of an atmospheric pressure
plasma jet with coaxial shielding gas J. Appl. Phys.
54 075205
[39] Semenov I and Weltmann K 2020 Modelling of turbulent
reacting flow for a cold atmospheric pressure argon plasma
jet Plasma Sources Sci. Technol. 29 055001
[40] Lindsay A, Anderson C, Slikboer E, Shannon S and Graves D
2015 Momentum, heat, and neutral mass transport in
convective atmospheric pressure plasma-liquid systems and
implications for aqueous targets J. Phys. D: Appl. Phys.
48 424007
[41] Verlackt C, Van Boxem W and Bogaerts A 2018 Transport and
accumulation of plasma generated species in aqueous
solution Phys. Chem. Chem. Phys. 20 6845–59
[42] Semenov I, Weltmann K and Loffhagen D 2019 Modelling of
the transport phenomena for an atmospheric-pressure
plasma jet in contact with liquid J. Phys. D: Appl. Phys.
52 315203
[43] Heirman P, Van Boxem W and Bogaerts A 2019 Reactivity
and stability of plasma-generated oxygen and nitrogen
species in buffered water solution: a computational study
Phys. Chem. Chem. Phys. 21 12881–94
[44] Winter J, Sousa J S, Sadeghi N, Schmidt-Bleker A, Reuter S
and Puech V 2015 The spatio-temporal distribution of He
19
J. Phys. D: Appl. Phys. 57 (2024) 115204
P Heirman et al
(23S1) metastable atoms in a MHz-driven helium plasma jet
is influenced by the oxygen/nitrogen ratio of the
surrounding atmosphere Plasma Sources Sci. Technol.
24 025015
[45] Comsol multiphysics (available at: www.comsol.com)
[46] Van Rens J F, Schoof J T, Ummelen F C, Van Vugt D C,
Bruggeman P J and Van Veldhuizen E 2014 Induced liquid
phase flow by RF Ar cold atmospheric pressure plasma jet
IEEE Trans. Plasma Sci. 42 2622–3
[47] Menter F R, Kuntz M and Langtry R 2003 Ten years of
industrial experience with the SST turbulence model
Turbulence Heat Mass Transfer 4 625–32
[48] Kee R J, Coltrin M E and Glarborg P 2005 Chemically
Reacting Flow: Theory and Practice (Wiley)
[49] Kays W M 1994 Turbulent Prandtl number. Where are we?
ASME J. Heat Transfer 116 284–95
[50] Brokaw R S 1969 Predicting transport properties of dilute
gases Ind. Eng. Chem. Process Des. Dev. 8 240–53
[51] Smith G P et al GRI-Mech 3.0 (available at: www.me.
berkeley.edu/gri_mech/)
[52] McBride B J 2002 NASA Glenn Coefficients for Calculating
Thermodynamic Properties of Individual Species (National
Aeronautics and Space Administration, John H. Glenn
Research Center at Lewis Field)
[53] Stiel L I and Thodos G 1964 The thermal conductivity of
nonpolar substances in the dense gaseous and liquid regions
AlChE J. 10 26–30
[54] Sander R 2015 Compilation of Henry’s law constants (version
4.0) for water as solvent Atmos. Chem. Phys. 15 4399–981
[55] Tornin J, Labay C, Tampieri F, Ginebra M-P and Canal C 2021
Evaluation of the effects of cold atmospheric plasma and
plasma-treated liquids in cancer cell cultures Nat. Protocols
16 2826–50
[56] Schmidt-Bleker A, Bansemer R, Reuter S and Weltmann K D
2016 How to produce an NOx-instead of Ox-based
chemistry with a cold atmospheric plasma jet Plasma
Process. Polym. 13 1120–7
[57] Reuter S, Winter J, Iséni S, Schmidt-Bleker A, Dünnbier M,
Masur K, Wende K and Weltmann K-D 2014 The influence
of feed gas humidity versus ambient humidity on
atmospheric pressure plasma jet-effluent chemistry and skin
cell viability IEEE Trans. Plasma Sci. 43 3185–92
[58] Winter J et al 2014 Tracking plasma generated H2O2 from gas
into liquid phase and revealing its dominant impact on
human skin cells J. Phys. D: Appl. Phys. 47 285401
[59] Hähnel M, von Woedtke T and Weltmann K D 2010 Influence
of the air humidity on the reduction of Bacillus spores in a
defined environment at atmospheric pressure using a
dielectric barrier surface discharge Plasma Process. Polym.
7 244–9
[60] Mohades S, Lietz A M, Kruszelnicki J and Kushner M J 2020
Helium plasma jet interactions with water in well plates
Plasma Process. Polym. 17 1900179
[61] Yan D, Talbot A, Nourmohammadi N, Cheng X, Canady J,
Sherman J and Keidar M 2015 Principles of using cold
atmospheric plasma stimulated media for cancer treatment
Sci. Rep. 5 18339
[62] Verreycken T, Mensink R, Van Der Horst R, Sadeghi N and
Bruggeman P J 2013 Absolute OH density measurements in
the effluent of a cold atmospheric-pressure Ar–H2O RF
plasma jet in air Plasma Sources Sci. Technol.
22 055014
[63] Gorbanev Y, Van der Paal J, Van Boxem W, Dewilde S and
Bogaerts A 2019 Reaction of chloride anion with atomic
oxygen in aqueous solutions: can cold plasma help in
chemistry research? Phys. Chem. Chem. Phys.
21 4117–21
[64] Wenske S, Lackmann J-W, Busch L M, Bekeschus S, von
Woedtke T and Wende K 2021 Reactive species driven
oxidative modifications of peptides—Tracing physical
plasma liquid chemistry J. Appl. Phys. 129 193305
[65] Neoplas Optional accessories for kINPen®IND (available at:
www.neoplas.eu/en/accessorieskinpenind.html)
[66] Tresp H, Hammer M U, Weltmann K-D and Reuter S 2013
Effects of atmosphere composition and liquid type on
plasma-generated reactive species in biologically relevant
solutions Plasma Med. 3 45–55
[67] Tampieri F, Gorbanev Y and Sardella E 2023 Plasma-treated
liquids in medicine: let’s get chemical Plasma Process.
Polym. 20 e2300077
[68] Roquemore W, Tankin R, Chiu H and Lottes S 1986 A study
of a bluff-body combustor using laser sheet lighting Exp.
Fluids 4 205–13
[69] Tong Y, Liu X, Chen S, Li Z and Klingmann J 2018 Effects of
the position of a bluff-body on the diffusion flames: a
combined experimental and numerical study Appl. Therm.
Eng. 131 507–21
[70] Kapaldo J, Han X and Ptasinska S 2019
Shielding-gas-controlled atmospheric pressure plasma jets:
optical emission, reactive oxygen species, and the effect on
cancer cells Plasma Process. Polym. 16 1800169
[71] Gazzah M H and Belmabrouk H 2014 Directed co-flow effects
on local entropy generation in turbulent heated round jets
Comput. Fluids 105 285–93
20
| null |
10.1371_journal.pgen.1010857.pdf
|
Data Availability Statement: Hi-C data were
deposited in the Gene Expression Omnibus
(accession no. GSE225771). The scripts used for
analyzing Hi-C data have been deposited to Github
(https://github.com/hbbrandao/Borrelia_HiC_
Analysis). All other relevant data are within the
paper and its Supporting Information files.
|
Hi-C data were deposited in the Gene Expression Omnibus (accession no. GSE225771). The scripts used for analyzing Hi-C data have been deposited to Github ( https://github.com/hbbrandao/Borrelia_HiC_ Analysis ). All other relevant data are within the paper and its Supporting Information files.
|
RESEARCH ARTICLE
Organization and replicon interactions within
the highly segmented genome of Borrelia
burgdorferi
Zhongqing RenID
WagnerID
2,3,4*, Xindan WangID
1*
1☯, Constantin N. TakacsID
2,3,4☯¤, Hugo B. BrandãoID
5, Christine Jacobs-
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 Department of Biology, Indiana University, Bloomington, Indiana, United States of America, 2 Department
of Biology, Stanford University, Stanford, California, United States of America, 3 Sarafan ChEM-H Institute,
Stanford University, Stanford, California, United States of America, 4 Howard Hughes Medical Institute,
Stanford, California, United States of America, 5 Illumina Inc., 5200 Illumina Way, San Diego, California,
United States of America
☯ These authors contributed equally to this work.
¤ Current address: Department of Biology, College of Science, Northeastern University, Boston,
Massachusetts, USA
* [email protected] (CJW); [email protected] (XW)
OPEN ACCESS
Abstract
Borrelia burgdorferi, a causative agent of Lyme disease, contains the most segmented bac-
terial genome known to date, with one linear chromosome and over twenty plasmids. How
this unusually complex genome is organized, and whether and how the different replicons
interact are unclear. We recently demonstrated that B. burgdorferi is polyploid and that the
copies of the chromosome and plasmids are regularly spaced in each cell, which is critical
for faithful segregation of the genome to daughter cells. Regular spacing of the chromosome
is controlled by two separate partitioning systems that involve the protein pairs ParA/ParZ
and ParB/Smc. Here, using chromosome conformation capture (Hi-C), we characterized
the organization of the B. burgdorferi genome and the interactions between the replicons.
We uncovered that although the linear chromosome lacks contacts between the two replica-
tion arms, the two telomeres are in frequent contact. Moreover, several plasmids specifically
interact with the chromosome oriC region, and a subset of plasmids interact with each other
more than with others. We found that Smc and the Smc-like MksB protein mediate long-
range interactions on the chromosome, but they minimally affect plasmid-chromosome or
plasmid-plasmid interactions. Finally, we found that disruption of the two partition systems
leads to chromosome restructuring, correlating with the mis-positioning of chromosome
oriC. Altogether, this study revealed the conformation of a complex genome and analyzed
the contribution of the partition systems and SMC family proteins to this organization. This
work expands the understanding of the organization and maintenance of multipartite bacte-
rial genomes.
Citation: Ren Z, Takacs CN, Brandão HB, Jacobs-
Wagner C, Wang X (2023) Organization and
replicon interactions within the highly segmented
genome of Borrelia burgdorferi. PLoS Genet 19(7):
e1010857. https://doi.org/10.1371/journal.
pgen.1010857
Editor: Frederic Boccard, Centre National de la
Recherche Scientifique, FRANCE
Received: April 5, 2023
Accepted: July 5, 2023
Published: July 26, 2023
Copyright: © 2023 Ren et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Hi-C data were
deposited in the Gene Expression Omnibus
(accession no. GSE225771). The scripts used for
analyzing Hi-C data have been deposited to Github
(https://github.com/hbbrandao/Borrelia_HiC_
Analysis). All other relevant data are within the
paper and its Supporting Information files.
Funding: The support for this work comes in part
from the National Institutes of Health
R01GM141242 and R01GM143182 (X.W.), and
the Pew Innovation Fund (C.J.-W.). This research is
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
1 / 27
PLOS GENETICS a contribution of the GEMS Biology Integration
Institute, funded by the National Science
Foundation DBI Biology Integration Institutes
Program, Award #2022049 (X.W.). The funders
had no role in study design, data collection and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: I have read the journal’s
policy and the authors of this manuscript have the
following competing interests: C.J.-W. is an
investigator of the Howard Hughes Medical
Institute. H.B.B. is an employee of Illumina, Inc.
Organization of a highly segmented bacterial genome
Author summary
Genomes are highly organized in cells to facilitate biological processes. Borrelia burgdor-
feri, an agent of Lyme disease, carries one linear chromosome and more than twenty plas-
mids in what is known as one of the most segmented bacterial genomes. How the
different replicons interact with each other is unclear. Here we investigate the organiza-
tion of this highly segmented genome and the protein factors that contribute to this orga-
nization. Using chromosome conformation capture assays, we determined the
interactions within the chromosome, between the chromosome and plasmids, and
between the plasmids. We found that the two telomeres of the linear chromosome interact
with each other; a subset of plasmids interact with the chromosomal replication origin
region; and a subset of plasmids preferentially interact with one another. Finally, we
revealed that two structural maintenance of chromosomes (SMC) family proteins, Smc
and MksB, promote long-range DNA interactions on the chromosome, and the two parti-
tion systems, ParA/ParZ and ParB/Smc, contribute to chromosome structure. Altogether,
we characterized the conformation of a complex genome and investigated the functions of
different genome organizers. Our study advances the understanding of the organization
of highly segmented bacterial genomes.
Introduction
Borrelia burgdorferi causes Lyme disease, the most prevalent vector-borne infectious disease in
Europe and North America [1,2]. Although the B. burgdorferi genome is only ~1.5 Megabase
pairs in size, it includes one linear chromosome and more than 20 plasmids (circular and lin-
ear) and is, to our knowledge, the most segmented bacterial genome [3–6]. Recently, using
fluorescence microscopy to visualize loci on the chromosome and 16 plasmids, we found that
B. burgdorferi contains multiple copies of its genome segments per cell, with each copy regu-
larly spaced along the cell length [7].
In bacteria, the broadly conserved parABS partitioning system plays an important role in
the segregation of chromosome and plasmids [8–15]. ParA dimerizes upon ATP binding and
non-specifically binds to the DNA [16–19]. Centromeric ParB proteins bind to the parS
sequences scattered around the origin of replication and spread several kilobases to nearby
regions, forming a nucleoprotein complex [20–25]. The ParB-DNA nucleoprotein complex
interacts with DNA-bound ParA-ATP dimers and stimulates the ATPase activity of ParA,
leading to the release of ParA from the DNA and the formation of a ParA concentration gradi-
ent along the nucleoid [12, 15, 17, 26]. It is thought that repeated cycles of ParA and ParB
interaction and release, together with the translocating forces from elastic chromosome
dynamics [27–30] or the chemical ParA gradient [31, 32], promote the segregation of the two
newly replicated ParB-origin complexes from one another [27, 29]. In addition, ParB plays a
separate role in recruiting the broadly conserved SMC complex onto the chromosomal origin
region [13, 14]. Once loaded, SMC complexes move away from the loading sites and typically
tether the two replication arms together, facilitating the resolution and segregation of the two
sister chromosomes [33–35].
We discovered that in B. burgdorferi, the segregation and positioning of the replication ori-
gin (oriC) of the multicopy chromosome require the concerted actions of the ParB/Smc system
and a newly discovered ParA/ParZ system [7]. ParZ, a centromere-binding protein, substitutes
ParB to work with ParA and plays a major role in chromosome segregation [7]. Although B.
burgdorferi ParB does not appear to partner with ParA, it is still required to recruit Smc to
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
2 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
oriC. Smc in turn contributes to oriC positioning [7]. Overall, these previous findings
advanced our understanding of oriC segregation in B. burgdorferi. However, the information
on the organization of the bulk of the chromosome and the interactions among the various
genome segments in this bacterium is still lacking.
Chromosome conformation capture assays (Hi-C) have significantly advanced our
understanding of bacterial genome folding and interactions [34, 36–41]. Along bacterial
genomes, short-range self-interacting domains called chromosome interaction domains
(CIDs) have been observed and are shown to be dictated mostly by transcription, with
domain boundaries correlating with highly transcribed genes. In bacteria that contain the
canonical SMC complex, the two replication arms of the chromosome are juxtaposed
together, whereas bacteria that only encode SMC-like MukBEF and MksBEF proteins do
not show inter-arm interactions [37, 39].
More recent efforts have begun to reveal the genome conformation of bacteria containing
multiple replicons. In Agrobacterium tumefaciens, the origins of the four replicons are clus-
tered together, which regulates DNA replication and drives the maintenance of this multipar-
tite genome [41, 42]. Similarly, the two origins of Brucella melitensis chromosomes also
showed frequent interactions [43]. In Vibrio cholerae, the origin of Chromosome 2 (Ch2)
interacts with the crtS region on Chromosome 1 (Ch1) for replication control, and the termi-
nus regions of Ch1 and Ch2 interact for coordinated replication termination and terminus
segregation [40, 44]. These findings suggest that multipartite genomes harness inter-replicon
interactions as a mechanism for replication regulation and genome maintenance. In this
study, we aimed at understanding how B. burgdorferi organizes its ~20 replicons and how the
partitioning proteins and Smc homologues contribute to genome organization.
Results
The organization of the linear B. burgdorferi chromosome
To determine the organization of the highly segmented genome of B. burgdorferi, we per-
formed chromosome conformation capture (Hi-C) on exponentially growing cultures of the
infectious, transformable strain S9, hereafter used as our wild-type (WT) strain (S1 Table and
Figs 1A, 1B, and S1). Hi-C experiments measure the frequency of DNA contacts captured by
formaldehyde, which is a one-carbon crosslinker that covalently links protein-protein, pro-
tein-DNA, and DNA-DNA when these molecules are in spatial proximity [45]. A high fre-
quency of contact in a Hi-C map indicates that the DNA pieces are either in physical contact
or in spatial proximity, which may happen on their own or be mediated by protein factors. In
this study, we refer to “high frequency of contact between the DNA pieces in the Hi-C maps”
as “interactions” for simplicity.
After mapping the reads and plotting the data, we observed many white lines on the Hi-C
map, especially in regions corresponding to the plasmids (Fig 1B). These white lines indicated
the presence of repetitive sequences on the affected replicons, which were omitted during
sequence mapping. The genome-wide Hi-C interaction map (Fig 1B) has four distinct regions:
an intra-chromosomal interaction map in the lower left quadrant, a plasmid-chromosome
interaction map with identical, mirrored copies in the upper left and lower right quadrants,
and a plasmid-plasmid interaction map in the upper right quadrant. The chromosome dis-
played strong short-range interactions as shown on the primary diagonal (Fig 1B, lower left
quadrant). To better present the short-range interactions on the chromosome, we plotted the
Hi-C data in a different color scale (S1 Fig). Similar to what has been reported in other bacteria
[34, 36–38], chromosome interaction domains (CIDs) were present along the chromosome
(S1A Fig), with the strongest CIDs boundaries largely correlated with highly transcribed genes
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
3 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Fig 1. Genome-wide organization of B. burgdorferi replicons. (A) The B. burgdorferi S9 wild-type strain has one linear
chromosome (Chr), eight circular plasmids, and ten linear plasmids. The replication origin of the chromosome is labeled as oriC.
The sizes (in kb) and relative copy numbers of the plasmids are listed. The copy numbers of each plasmid were previously
measured using whole genome sequencing analysis [7], and were shown relative to the copy number of oriC. (B) Normalized Hi-
C interaction map showing interaction frequencies for pairs of 5-kb bins across the genome of B. burgdorferi strain S9. The x-
and y-axes show genome positions. The chromosome and the plasmids are indicated by red and blue bars, respectively. oriC is
labeled on the x-axis. The boundaries between the chromosome and the plasmids are indicated by black dotted lines. The white
lines indicate the presence of repetitive sequences omitted during sequence mapping. The black arrows point to the interactions
between the telomere regions. The plasmids are ordered alphabetically from cp26 to lp54, from left to right on the x-axis and
bottom to top on the y-axis. The whole map was divided into four regions: the lower left region shows intra-chromosomal
interactions, the upper left and lower right regions show plasmid-chromosome interactions, and the upper right region
represents plasmid-plasmid interactions. We used the same convention for all whole-genome Hi-C and downstream analyses in
this study. The color scale depicting Hi-C interaction scores in arbitrary units is shown at the right. The same Hi-C map with a
different color scale is shown in S1 Fig.
https://doi.org/10.1371/journal.pgen.1010857.g001
revealed by RNA-seq performed in a different study [46] (S1B Fig). Interestingly, a secondary
diagonal representing inter-arm interactions was absent from the Hi-C map (Figs 1B and S1,
lower left quadrant). This was unexpected as B. burgdorferi encodes an Smc protein homolog
and all Smc-carrying bacteria tested so far display inter-arm interactions on the chromosome
[34, 36, 38, 39, 41, 47, 48]. Notably, although B. burgdorferi contains a homolog of the ScpA
subunit of the SMC complex, it does not encode the other subunit, ScpB [3]. Thus, the absence
of the Smc-ScpAB holo-complex might explain the absence of chromosome arm alignment in
B. burgdorferi (see Discussion). Additionally, the two ends of the linear chromosome, the left
and right telomeres (terCL and terCR), displayed a high frequency of contact (Fig 1B, black
arrows in lower left quadrant). It is unclear whether terCL and terCR regions were physically
interacting through specific factors, or some unknown properties of these chromosome ends
increased the probability of contact between these two DNA regions. In addition, since B.
burgdorferi is polyploid [7], we do not know whether the interacting terCL and terCR were
located on the same chromosome or on adjacent chromosome copies.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
4 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Interactions between the chromosome and 18 plasmids
Qualitatively, plasmid-chromosome interactions (Figs 1B and S1, upper left and lower right
quadrants) were weaker than short-range interactions within the chromosome (Figs 1B and
S1, the primary diagonal of the lower left quadrant), but were stronger than long-range inter-
actions within the chromosome (Figs 1B and S1, outside of the primary diagonal on the lower
left quadrant). We plotted the distribution of these types of interaction frequencies and found
that the differences were statistically significant (Fig 2). To better show the plasmid-chromo-
some interactions (Fig 3A), we analyzed the interaction of each plasmid with each 5-kb bin on
the chromosome by adding up the interaction scores that belonged to the same plasmid
(Fig 3B). Interestingly, a subset of the linear plasmids, namely lp17, lp21, lp25, and lp28-3,
showed higher contact frequency with the chromosome, especially in the oriC region compared
with the rest of the chromosome (Fig 3B). We also observed that cp32-3, cp32-7, cp32-9 had
overall lower interactions with the chromosome seen as “blue stripes” in Fig 3B, which was cor-
related with their higher plasmid-plasmid interactions (see below). To examine the plasmid-
chromosome interactions without the influence of intra-chromosomal and plasmid-plasmid
Fig 2. Hi-C contact frequencies for different types of interactions. Distributions of Hi-C contact frequencies
measured for different types of interactions are shown as violin plots. Blue lines indicate standard deviations of the
values. Orange lines indicate the median, 5th and 95th percentile of the data. The p-values were computed using a
Mann-Whitney U test. All comparisons were done for data binned at 5-kb resolution.
https://doi.org/10.1371/journal.pgen.1010857.g002
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
5 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Fig 3. Plasmid-chromosome interactions. (A) Enlarged Hi-C map of plasmid-chromosome interactions in WT B.
burgdorferi strain S9 from Fig 1B. The x-axis shows positions on the chromosome, and the y-axis shows the plasmids
with their relative lengths. The white lines indicate repetitive sequences omitted during sequence mapping. oriC is
labeled on the x-axis. The color scale depicting Hi-C interaction scores in arbitrary units is shown at the right. We note
that on plasmid lp25 of WT B. burgdorferi strain S9, the bbe02 gene was disrupted by a PflaB-aadA streptomycin
resistance cassette. Therefore, there were two copies of PflaB, one on lp25 and one at the endogenous chromosomal
locus at ~150 kb. The B31 genome sequence used for Hi-C mapping contained only the endogenous copy of PflaB.
Thus, short-range interactions on lp25 involving the ectopic copy of PflaB artifactually appeared as interactions between
lp25 and the chromosome at ~150 kb. (B) The calculated interaction scores between each plasmid and chromosome
locus. The Hi-C interaction scores in consecutive bins were summed according to each plasmid before plotting. The
plot shows averaged data of two replicates. The x-axis indicates the genome position on the chromosome. The y-axis
specifies the different plasmids. The color scale depicting interaction scores in arbitrary units is shown at the right. We
note that these values were calculated from (A), which was part of Fig 1B. The data were normalized including all the
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
6 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
interactions in the genome (i.e. intra-chromosomal, plasmid-chromosome and plasmid-plasmid interactions). (C)
Renormalized plasmid-chromosome interactions following iterative correction to remove the contributions of intra-
chromosomal and plasmid-plasmid interactions (see Materials and methods). The data were normalized such that
each row had the same total score, and each column had the same total score.
https://doi.org/10.1371/journal.pgen.1010857.g003
interactions, we renormalized the data by iterative correction (see Materials and methods) on
Fig 3B and generated Fig 3C. While this renormalization removed the blue stripes seen in
Fig 3B, the positive interactions between the four plasmids (lp17, lp21, lp25 and lp28-3) and
oriC were still evident albeit less intense (Fig 3C). The plasmid-oriC interactions observed by
Hi-C are reminiscent of the origin clustering interactions mediated by centromeric proteins in
A. tumefaciens, which are critical for the replication and maintenance of the secondary replicons
in that bacterium [41, 42]. Notably, the plasmid-chromosome interactions observed here are
weaker than those observed in A. tumefaciens, and only four out of 18 plasmids showed these
specific interactions with the chromosome, thus the biological function of these interactions is
unclear (see Discussion).
Plasmid-plasmid interactions
Plasmid-plasmid interactions are depicted in the upper right quadrant of the Hi-C map (Figs
1B and S1) and appeared stronger than plasmid-chromosome interactions (Fig 1B, upper left
quadrant, and Fig 2) and long-range interactions within the chromosome (Fig 1B, outside of
the primary diagonal on the lower left quadrant, and Fig 2). The primary diagonal of the plas-
mid-plasmid interaction quadrant showed that each plasmid formed an interaction domain
on its own (Fig 4A). We note that the sizes of the 18 plasmids range from 17 kb to 54 kb [3, 4]
(Fig 1A) and that many plasmids have repetitive sequences omitted during Hi-C mapping
(Fig 4A). Therefore, our Hi-C map with a bin size of 5 kb does not have high enough resolu-
tion to describe detailed intra-plasmid interactions.
To better examine the interactions between every two plasmids, we recalculated the interac-
tion frequencies by adding up interaction scores that belonged to the same plasmid (Fig 4B).
To remove the influence of plasmid-chromosome interactions, we renormalized the data by
iterative correction (see Materials and methods) on Fig 4B to obtain Fig 4C. These analyses
revealed higher interactions among the seven cp32 plasmids (cp32-1, cp32-3, cp32-4, cp32-6,
cp32-7, cp32-8, cp32-9) (Fig 4B and 4C). To a lesser degree, the circular cp26 plasmid and the
ten linear plasmids interacted more among themselves than with the cp32 plasmids (Fig 4C).
The sizes of the plasmids range from 17 to 54 kb (Fig 1A). Their copy number had been previ-
ously determined by microscopy and whole genome sequencing, which ranged from 0.5 to 1.4
relative to the copy number of the oriC locus [7] (Fig 1A). To test whether the sizes and copy
numbers of the plasmids might contribute to plasmid-plasmid interactions, we used these
numbers to simulate the plasmid-plasmid interaction frequencies, assuming that all the plas-
mids were randomly interacting with each other and were freely diffusing in the cytoplasm
(see Materials and methods for simulation details). Before any corrections, our simulations
showed that plasmids that have a bigger size or a higher copy number interacted more with
other plasmids (S2A and S2B Fig, top panels). However, these preferential interactions did
not show up after our standard procedure of iterative corrections which were also applied to
the experimental Hi-C maps [49] (S2A and S2B Fig, middle panels, Fig 4D), unless we used a
very fine color scale (S2A and S2B Fig bottom panels). Thus, the preferential interactions
between plasmids we observed in our experiment (Fig 4B and 4C) could not be explained
solely by random plasmid-plasmid interactions after plasmid size and copy number differences
were accounted for. Since repetitive sequences within the plasmids were removed during
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
7 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Fig 4. Plasmid-plasmid interactions. (A) Enlarged Hi-C map of plasmid-plasmid interactions in WT B. burgdorferi strain S9 from Fig 1B. The x- and y-axes
show the plasmids with their relative lengths. The white lines indicate repetitive sequences omitted during sequence mapping. The color scale depicting Hi-C
interaction scores in arbitrary units is shown at the right. (B) The calculated interaction scores between each pair of plasmids. The Hi-C interaction scores in
consecutive bins were summed according to each plasmid prior to plotting. The plot shows averaged data of two replicates. The color scale depicting
interaction scores in arbitrary units is shown at the right. We note that these values were calculated from (A), which was part of Fig 1B. The data were
normalized including all the interactions in the genome (i.e. intra-chromosomal, plasmid-chromosome and plasmid-plasmid interactions). (C) Renormalized
plasmid-plasmid interactions following iterative correction to remove the contributions of plasmid-chromosome interactions (see Materials and methods).
The data were normalized such that each row had the same total score, and each column had the same total score. (D) The simulated interaction frequencies
between plasmids based on random collisions accounting for plasmid copy numbers and plasmid sizes (see Materials and methods). The data went through
iterative correction in the same way as the experimental data shown in (C). The simulated maps before iterative correction or after iterative correction but in a
finer color scale can be found in S2A Fig.
https://doi.org/10.1371/journal.pgen.1010857.g004
mapping, we believe that these higher-than-expected interactions observed in our experiment
are genuine and not due to erroneous mapping or normalization. The molecular mechanism
for plasmid-plasmid interactions remains to be determined.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
8 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Clustering analysis of smc and par mutants
The highly conserved SMC family proteins and the DNA partitioning proteins are central
players in bacterial chromosome organization and segregation [50–53]. B. burgdorferi has a
canonical Smc protein, encoded by gene bb0045, as well as an MksB protein, encoded by gene
bb0830, but lacks the genes encoding the accessory proteins ScpB, MksE, and MksF [3]. Addi-
tionally, B. burgdorferi employs two partition systems for the positioning of its multicopy oriC
loci: ParB/Smc and ParA/ParZ [7]. In our previous study, we built a collection of mutants car-
rying the following gene deletions: ΔparB, ΔparS, ΔparBS, ΔparA, ΔparZ, ΔparAZ, ΔparAZBS,
or Δsmc [7]. In these strains, the genes of interest were disrupted and replaced with a gentamy-
cin or kanamycin resistance gene. A control strain CJW_Bb284 was also built, which had the
gentamycin marker inserted in a non-coding region located between the convergently-ori-
ented parZ and parB genes, in an otherwise WT parAZBS locus. We have previously shown
that the mutant strains have similar growth rates compared with the S9 WT and control
strains, except for the ΔparAZBS mutant, which grows slower [7]. Quantitative imaging has
also indicated that all of these mutants have a similar cell length distribution [7]. Using either
ParZ-msfGFP or mCherry-ParB as a marker for oriC localization, we have previously shown
that the control strains have ~10 copies of oriC per cell, but this number decreases to ~9 for
ΔparA, 7–8 for ΔparBS, ΔparZ, ΔparAZ, and Δsmc, and ~6 for ΔparAΔparBS [7]. Additionally,
ΔparBS and ΔparAZ both disrupt the even spacing of oriC in the polyploid cells, but ΔparAZ
has the more pronounced effect that is similar to that of ΔparZ [7]. Importantly, the ΔparAΔ-
parBS mutant has a much stronger defect in origin spacing than ΔparA, ΔparBS, ΔparZ, or
ΔparAZ, lending support to the conclusion that ParA works with ParZ in a pathway separate
from ParB/parS [7]. Although ParB/parS does not seem to interact with ParA in B. burgdorferi,
our previous work has shown that ParB binds to parS and recruits Smc to the origin region
[7], highlighting that the parS-ParB-Smc interactions in B. burgdorferi are similar to those
demonstrated in other bacterial species [13, 14, 38, 39, 41]. Thus, in B. burgdorferi, the regular
spacing of chromosome copies is controlled by two separate partitioning systems that involve
the protein pairs ParA/ParZ and ParB/Smc [7].
To understand the contribution of ParB/parS/Smc, ParA/ParZ, and additionally MksB to B.
burgdorferi genome organization, we performed Hi-C on these mutants (S1 Table) and the con-
trol strain and compared the results with those of the WT. Hi-C experiments on every strain were
done in two biological replicates which showed nearly identical results (S3 Fig). To compare the
different mutants, we performed a clustering analysis using the contact probability curves of our
22 Hi-C samples (S4 Fig) so that mutants that had similar profiles of contact probabilities would
be grouped together (Fig 5). Using the Silhouette method [54], we found that the mutants could
be divided into six groups (Fig 5A and 5B) (see Materials and methods), which was largely consis-
tent with Principal Component Analysis [54] (S5 Fig) and t-distributed stochastic neighbor
embedding [54] (S6 Fig): group 1 included the WT and the control strain CJW_Bb284 (Figs 5B,
5C and S7); group 2 included Δsmc (Fig 5B and 5D); group 3 included ΔmksB (Fig 5B and 5E);
group 4 included ΔparB, ΔparS and ΔparBS (Fig 5B and 5F); group 5 included ΔparA, ΔparZ and
ΔparAZ (Fig 5B and 5G); and group 6 included ΔparAZBS (Fig 5B and 5H).
This grouping analysis based on Hi-C results indicates that the control strain CJW_Bb284
behaves the same as its parental WT strain (S7 Fig); Smc and MksB have different effects on
chromosome folding; ParB and parS work as a unit; ParA and ParZ work together; and ParB/
parS and ParA/ParZ have additive effects because ΔparAZBS formed its own group. Therefore,
the grouping of mutants based on Hi-C analysis here (Fig 5B) is largely consistent with our
previous cytological characterization of these mutants [7]. This agreement shows the robust-
ness of our assays.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
9 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Fig 5. Clustering analysis of different mutants. (A) Determination of the optimal number of clusters of contact probability curves, Pc(s), for k-means
clustering (see Materials and methods). Only intra-chromosomal interactions were used to calculate the Pc(s) curves. The number of clusters was determined
by identifying the peak in Silhouette score. This analysis found six optimal groupings, which is indicated by the red circle and black dotted line. (B) Pc(s) curves
of all the samples plotted in the same graph. Pc(s) curves show the average contact frequency between all pairs of loci on the chromosome separated by set
distance (s). The x-axis indicates the genomic distance of separation in kb. The y-axis represents the averaged contact frequency. The curves were computed for
intra-chromosomal interactions binned at 5 kb. Grouping result of the 11 strains was listed on the right. Two biological replicates of each strain were plotted.
Individual Pc(s) curves can be found in S4 Fig. Principal Component Analysis (PCA) and T-distributed stochastic neighbor embedding (t-SNE) results can be
found in S5 and S6 Figs, respectively. (C-I) Curves belonging to the same groups in (B) were plotted in different panels. Two biological replicates of each strain
were plotted.
https://doi.org/10.1371/journal.pgen.1010857.g005
Smc and MksB mediate long-range interactions within the chromosome
In our clustering analysis, the two biological replicates of Δsmc fell in one group (group 2) and
replicates of ΔmksB fell into a separate group (group 3) (Fig 5B, 5D and 5E). To understand
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
10 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Fig 6. Smc and MksB mediate long-range DNA interactions. (A-C) Normalized Hi-C interaction maps of the control (CJW_Bb284), Δsmc (CJW_Bb609)
and ΔmksB (CJW_Bb605) strains. Black dotted lines mark the boundary between the chromosome and the plasmids. The color scale depicting Hi-C interaction
scores in arbitrary units is shown at the right. (D-F) Log2 ratio plots comparing different Hi-C matrices. Log2(matrix 1/matrix 2) was calculated and plotted in
the heatmaps. Identities of matrix 1/ matrix 2 are shown at the top of each plot. The color scale is shown at the right of panel (F). Black arrows point to terCL-
terCR interactions. Black trapezoids indicate reduced interactions in the mutants. (G-I) Contact probability decay Pc(s) curves of indicated Hi-C matrices
taken from Fig 5B. The intersection points of mutant and control curves are indicated by black dotted lines.
https://doi.org/10.1371/journal.pgen.1010857.g006
how Δsmc and ΔmksB affect genome contacts, we analyzed the log2 ratios of the Hi-C maps
between each mutant strain and the relevant control (Fig 6A–6F). We observed that both
Δsmc and ΔmksB strains had decreased long-range DNA contact compared with the control
(Fig 6D–6F, blue pixels in black trapezoid). Specifically, as seen on the Hi-C contact probabil-
ity decay curves, in Δsmc, loci separated by ~50 kb or greater had decreased frequency of con-
tacts compared with the control (Fig 6G and 6H), and in ΔmksB, loci separated by ~100 kb or
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
11 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
greater had decreased frequency of contact compared with the control (Fig 6G and 6I, black
dotted lines). These data indicate that both Smc and MksB promote long-range DNA contacts
and that their effects are different enough to fall into different groups in our clustering analysis.
We noted that B. burgdorferi is missing the ScpB subunit of the SMC complex, as well as the
MksE and MksF subunits of the MksBEF complex. However, previous work showed that puri-
fied B. subtilis Smc protein (in the absence of ScpA and ScpB) is able to form DNA loops in
vitro [55]. Our results suggest that in B. burgdorferi, the incomplete SMC/Mks complexes may
form DNA loops. Alternatively, it is possible that B. burgdorferi uses unknown factors instead
of ScpB and MksEF. Curiously, the absence of MksB, and to a lesser degree, the absence of
Smc, enhanced the terCL-terCR interactions (Fig 6E and 6F, black arrows). Since this trend is
the opposite of the overall reduction of long-range DNA interactions seen in the Δsmc and
ΔmksB strains (Fig 6E and 6F, black trapezoids), these results suggest that MksB and Smc spe-
cifically reduce the contacts between the telomeres. In addition, when the data were normal-
ized to remove intra-chromosomal interactions, we did not find evidence of MksB or Smc
affecting plasmid-chromosome (S8, S9 and S10 Figs) or plasmid-plasmid interactions (S11
and S12 Figs), suggesting that these proteins act primarily within the chromosome and not
between replicons. Finally, we do not know whether MksB and Smc affected the intra-replicon
contacts within each plasmid because our 5-kb resolution was too low for the small sizes of the
plasmids.
Contribution of ParB/parS and ParA/ParZ to genome organization
In the grouping analysis, ΔparS, ΔparB and ΔparBS fell in the same group (group 4) (Fig 5B
and 5F), consistent with the previous finding that ParB and parS act as a unit [7]. Compared
with the control, the absence of parB and/or parS caused similar changes to genome interac-
tions (Fig 7A–7F): terCL-terCR interactions decreased (Fig 7D–7F, blue pixels indicated by
black arrows); longer range (>150 kb) interactions within the chromosome increased
(Fig 7D–7F, red pixels within black trapezoid); and short-range interactions (50–150 kb)
decreased (Fig 7D–7F, blue pixels between black trapezoid and the red line). These trends are
opposite to those observed in Δsmc or ΔmksB (Fig 6E and 6F). We postulate that the effect of
ParB/parS on global chromosome conformation might be due to their effect on Smc distribu-
tion. Our previous work showed that ParB recruits Smc to the oriC region in B. burgdorferi,
and the loss of parBS caused Smc localization to be more dispersed on nucleoid [7]. Thus, the
increase of long-range interactions in the absence of ParB/parS suggests that non-specific load-
ing of Smc to the chromosome outside of the oriC region (i.e. independent of ParB/parS) con-
tributes greatly to long-range chromosome interactions.
Group 5 contains ΔparA, ΔparZ, ΔparAZ (Figs 5B, 5G and 7G–7I), consistent with the
idea that ParA and ParZ work in the same pathway [7]. The absence of parA and/or parZ
caused two major changes in chromosome folding: loci separated by 100 to 300 kb had
increased interactions (Fig 7K–7M, red pixels below the black line) and loci separated by 300
kb or more had decreased interactions (Fig 7K–7M, blue pixels above the black line). Thus,
ParA/ParZ acts to reduce mid-range (100–300 kb) and enhance long-range (>300 kb) DNA
interactions on the chromosome. Since ParA/ParZ promotes chromosome segregation and
spacing, we speculate that loss of ParA acting on DNA caused these changes in DNA
interactions.
Finally, ΔparAZBS, which lacked both parBS and parAZ, formed its own group (group 6)
(Figs 5B, 5H, 7J and 7N), consistent with its physiological and cytological behavior being the
most severe in all of the mutants tested [7]. In Hi-C experiments, this mutant essentially exhib-
ited an additive effect of ΔparBS (Fig 7C and 7F) and ΔparAZ (Fig 7I and 7M): decreased
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
12 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Fig 7. Disruption of the partition systems re-structures the genome. (A-C) Normalized Hi-C interaction maps of the ΔparB (CJW_Bb353), ΔparS
(CJW_Bb354), and ΔparBS (CJW_Bb285) strains. Black dotted lines indicate the boundary between the chromosome and the plasmids. The color scale
depicting Hi-C interaction scores in arbitrary units is shown at the right. (D-F) Log2 ratio plots comparing ΔparB (CJW_Bb353), ΔparS (CJW_Bb354), or
ΔparBS (CJW_Bb285) with the control (CJW_Bb284) strain as indicated. Black arrows point to blue pixels of terCL-terCR interactions. Black trapezoids
indicate area of red pixels. Red lines indicate the boundary between red and blue pixels. The color scale is shown at the right. (G-J) Normalized Hi-C
interaction maps of the ΔparA (CJW_Bb366), ΔparZ (CJW_Bb286), ΔparAZ (CJW_Bb287) and ΔparAZBS (CJW_Bb288) strains. Black arrows indicate terCL-
terCR interactions. (K-N) Log2 ratio plots comparing indicated strains. Solid black lines mark the boundaries between red and blue pixels. Black arrows
indicate terCL-terCR interactions.
https://doi.org/10.1371/journal.pgen.1010857.g007
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
13 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
interactions below 150 kb (like in ΔparBS), increased mid-range (100–300 kb) interactions (as
seen in ΔparAZ), and a complete loss of terCL-terCR interactions (Fig 7J and 7N, black
arrows). These effects can be explained by the independent actions of ParB/parS and ParA/
ParZ that we discussed above.
Overall, our Hi-C analyses of these mutants indicate that the perturbation of genome inter-
actions is correlated with the previously observed cytological defects in chromosome position-
ing and segregation [7]. Interestingly, although DNA interactions within the chromosome
were changed in cells missing parBS or parAZ, the interactions between replicons (plasmid-
chromosome and plasmid-plasmid interactions) remained similar to the control (S8–S12
Figs). Only in ΔparAZBS, plasmid-chromosome interactions were reduced, and plasmid-plas-
mid interactions were more evened out. It is possible that in ΔparAZBS, the entanglement of
different copies of chromosomes in the polyploid cells [7] affected the interactions between
replicons.
Discussion
In this study, we characterized the organization of the highly segmented genome of B. burgdor-
feri and the contribution of the chromosome partitioning proteins and Smc homologs to this
organization. Even though B. burgdorferi expresses an Smc protein, we found that the chromo-
some does not have inter-arm interactions, which are observed in other Smc-carrying bacteria
[34, 36, 38, 39, 41, 47, 48]. Nonetheless, Smc and the Smc-like MksB protein increase long-
range DNA contacts possibly through DNA looping. Since B. burgdorferi lacks ScpB and
MksEF thus cannot form complete SMC and Mks complexes, it is possible that the loop forma-
tion mechanism by the incomplete complexes is different from the loop-extrusion activity of
the holocomplexes [55–59]. For instance, Smc or MksB alone might facilitate long-range loop
formation by bridging only DNA segments that are already in proximity. Alternatively, just as
ParA works with ParZ instead of ParB in B. burgdorferi, it is also possible that Smc and MksB
recruit other factors instead of ScpB and MksEF in this organism.
The B. burgdorferi strain used in this study contains 18 plasmids, which showed differential
interactions with the chromosome. Namely, plasmids lp17, lp21, lp25, and lp28-3 displayed
higher frequency of contact with the chromosome especially at the oriC region (Figs 3A and
S8). This pattern was highly reproducible in different mutants (S8–S10 Figs), suggesting that
these plasmid-chromosome contacts are real, specific interactions that might be mediated by
unknown protein factors. We did not detect specific plasmid-oriC colocalization in our previ-
ous imaging-based analysis [7]. This is likely because these interactions are transient, and such
weak but reproducible interactions are more easily captured in Hi-C experiments where mil-
lions of cells are averaged than in microscopy experiments where fewer cells are analyzed.
What are the molecular mechanism and biological function of these plasmid-chromosome
interactions? In A. tumefaciens, the secondary replicons cluster with the primary replicon at
their origin regions through the interactions between ParB homologs [41, 42], which prevents
the loss of the secondary replicons [42]. In B. burgdorferi, we note that these interactions did
not require ParB/parS or ParA/ParZ (S8–S10 Figs), suggesting that the molecular mechanism
for these interactions is different from the centromeric clustering observed in A. tumefaciens.
Although it is still possible that the four plasmids that interact with the chromosome may “pig-
gyback” the chromosome to facilitate their own segregation and maintenance, it is also possi-
ble that these plasmid-chromosome interactions have functions unrelated to plasmid
segregation. Indeed, 14 out of 18 plasmids did not interact with the chromosome origin, indi-
cating that B. burgdorferi plasmids segregate largely independently from the chromosome.
Notably, B. burgdorferi is polyploid with unequal copy number for each replicon [7] while A.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
14 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
tumefaciens newborn cells are haploid [41]. We postulate that the difference in ploidy might be
one underlying factor accounting for the difference in organizing strategies between these two
species. Our findings suggest that different species might take diverse strategies to organize
and maintain segmented genomes.
We found that the interactions between the plasmids on average are more frequent than
plasmid-chromosome interactions and long-range intra-chromosomal interactions (Figs 1B
and 2). Interestingly, all seven circular cp32 plasmids interacted more frequently with one
another (Fig 4B and 4C); the remaining 11 plasmids, including the circular cp26 plasmid and
the ten linear plasmids, preferentially interacted with one another, though to a lesser degree
(Fig 4C). These groupings cannot be simply explained by plasmid size, topology, or copy num-
ber (Figs 1A and 4D). In addition, all the B. burgdorferi plasmids are thought to use members
of the PF32, PF49, PF50 and PF57/62 gene clusters for replication and partitioning [4, 60–62]:
PF32 belongs to the ParA protein family, PF50 and PF57/62 are homologs of replication initia-
tor proteins, while PF49 likely serves as a ParB-like centromeric protein [63]. Therefore, their
replication and partitioning systems cannot explain the grouping of the plasmids, either. Curi-
ously, cp32 plasmids resemble the genomes of certain tailed bacteriophages [5, 64–66] and
cp32 DNA was found to be packaged in bacteriophage particles isolated from B. burgdorferi
cultures [67]. Thus, it is conceivable that the grouping of cp32 plasmids might be related to the
process of bacteriophage assembly, although the phage proteins are expressed at minimal level
without induction [5, 68]. The exact mechanism for the preferential interactions between plas-
mids remains to be explored.
Unlike in other bacteria studied to date, in B. burgdorferi, there are two partitioning systems,
ParA/ParZ and ParB/parS, which co-regulate the spacing of the oriC copies in the cell. ParA/
ParZ plays a more important role than ParB/parS. While removing ParB/parS only causes very
mild defects in oriC spacing in the presence of ParA/ParZ, deleting both parA and parBS further
disrupts the spacing pattern [7]. By Hi-C, we observed a similar trend in genome reorganization
in these mutants: removing parAZ caused a significant increase of the medium-range (100–300
kb) interactions but double deletion of parAZ and parBS led to an additive increase in these
interactions. Thus, the segregation defect is correlated with increased mid-range genome inter-
actions. The causal relationship between chromosome segregation and genome folding is
unclear and remains to be examined. We speculate that the tension exerted through the parti-
tioning system leads to the change in DNA folding over the length of the chromosome, which is
a decrease of DNA interactions in the 100–300 kb range observed here.
Despite the absence of inter-arm interactions on the chromosome, the two ends of the lin-
ear chromosome, terCL and terCR, displayed a high contact frequency, which required ParA/
ParZ and ParB/parS. The contribution of ParA/ParZ and ParB/parS to terCL-terCR interac-
tions might be through different mechanisms. ParA/ParZ is required for the spacing of oriC
copies [7]. Thus, it is possible that mis-positioning of chromosome copies reduces the fre-
quency of terCL-terCR contacts. For ParB/parS, although it does not contribute much to the
spacing of chromosome copies [7], it recruits Smc to the origin. Since Smc reduced terCL-
terCR contacts (Fig 5F), it is possible that ParB-mediated recruitment of Smc to the oriC-prox-
imal parS site and away from chromosome arms lifts Smc’s inhibitory role in terCL-terCR
interactions.
Altogether, our study identified intra-chromosomal, plasmid-chromosome, and plasmid-
plasmid interactions of the most segmented bacterial genome known to date. We explored the
contribution of SMC-family proteins and two partitioning systems to the folding and interac-
tions of the genome. Although the exact mechanism for replicon interactions remains to be
investigated, our study represents one step forward in the understanding of multipartite
genome architecture and maintenance.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
15 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Materials and methods
General methods
The B. burgdorferi strains used in this study are listed in S1 Table. Cells were maintained in
exponential growth in complete Barbour-Stoenner-Kelly (BSK)-II liquid medium at 34˚C in a
humidified incubator and under 5% CO2 atmosphere [69, 70]. Complete BSK-II medium con-
tained 50 g/L bovine serum albumin (Millipore, Cat. 810036), 9.7 g/L CMRL-1066 (US Biolog-
ical, Cat. C5900-01), 5 g/L Neopeptone (Difco, Cat. 211681), 2 g/L Yeastolate (Difco, Cat.
255772), 6 g/L HEPES (Millipore, Cat. 391338), 5 g/L glucose (Sigma-Aldrich, Cat. G7021), 2.2
g/L sodium bicarbonate (Sigma-Aldrich, Cat. S5761), 0.8 g/L sodium pyruvate (Sigma-Aldrich,
Cat. P5280), 0.7 g/L sodium citrate (Fisher Scientific, Cat. BP327), 0.4 g/L N-acetylglucosamine
(Sigma-Aldrich, Cat. A3286), 60 mL/L heat-inactivated rabbit serum (Gibco, Cat.16120), and
had a final pH of 7.60. When noted, the following antibiotics were used: gentamicin at 40 μg/
mL, streptomycin at 100 μg/mL, and kanamycin at 200 μg/mL [71–73]. Lists of plasmids, oli-
gonucleotides and next-generation-sequencing samples can be found in S2–S4 Tables.
Growing cells for Hi-C
For Hi-C biological replicates, pairs of 100 mL cultures of each strain were inoculated and
grown for two or three days. The cultures were fixed by adding 37 mL 37% formaldehyde
(Sigma-Aldrich, Cat. F8775) which resulted in 10% final concentration. This formaldehyde
concentration was chosen because the BSK-II medium used in this study was rich in primary
amines (see General methods above) which reacted with formaldehyde. 10% formaldehyde
gave us highly reproducible Hi-C results without signs of over-crosslinking such as inefficient
lysis or digestion. For crosslinking, the cultures were rocked at room temperature for 30 min.
Formaldehyde was quenched using 7 mL 2.5 M glycine at room temperature for 5 min with
rocking. The samples were chilled on ice for 10 min, then pelleted at 4˚C and 4,300 x g for 30
min in an Allegra X-14R centrifuge (Beckman Coulter) equipped with a swinging bucket
SX4750 rotor. The cell pellet was resuspended in 1 mL ice-cold HN buffer (50 mM NaCl, 10
mM HEPES, pH 8.0) [74], then pelleted at 4˚C and 10,000 x g for 10 min. The pellet was resus-
pended in 400 μL cold HN buffer, and 100 μL aliquots were frozen in a dry ice ethanol bath
then stored at below -80˚C.
Hi-C
The detailed Hi-C procedure for B. burgdorferi was adapted from previously described proto-
cols for B. subtilis [34] and A. tumefaciens [41]. Briefly, 5x108 B. burgdorferi cells were used for
each Hi-C reaction. Cells were lysed using Ready-Lyse Lysozyme (Epicentre, R1802M) in TE
for 60 min, followed by 0.5% SDS treatment for 30 min. Solubilized chromatin was digested
with DpnII for 2 hours at 37˚C. The digested chromatin ends were repaired with Klenow and
Biotin-14-dATP, dGTP, dCTP, dTTP. The repaired products were ligated in dilute reactions
by T4 DNA ligase at 16˚C overnight (about 20 hrs). Ligation products were incubated at 65˚C
overnight to reverse crosslinking in the presence of EDTA, 0.5% SDS and proteinase K. The
DNA was then extracted twice with phenol/chloroform/isoamylalcohol (25:24:1) (PCI), pre-
cipitated with ethanol, and resuspended in 40 μL 0.1XTE buffer. Biotin at non-ligated ends
was removed using T4 polymerase (4 hrs at 20˚C) followed by extraction with PCI. The DNA
was then resuspended in 105 μL ddH2O and sheared by sonication for 12 min with 20% ampli-
tude using a Qsonica Q800R2 water bath sonicator. The sheared DNA was used for library
preparation with the NEBNext UltraII kit (E7645) following the manufacturer’s instructions
for end repair, adapter ligation, and size selection. Biotinylated DNA fragments were purified
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
16 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
using 5 μL streptavidin beads (Invitrogen 65-001) following the manufacturer’s instructions.
All DNA-bound beads were used for PCR in a 50 μL reaction for 14 cycles. PCR products were
purified using Ampure beads (Beckman, A63881) and sequenced at the Indiana University
Center for Genomics and Bioinformatics using a NextSeq 500 sequencer.
Hi-C analysis
Paired-end sequencing reads were mapped to the genome file of B. burgdorferi B31 (NCBI Ref-
erence Sequence GCA_000008685.2 ASM868v2) using the default setting with MAPQ30 filter
of Distiller (https://github.com/open2c/distiller-nf). Plasmids are arranged in this order: cp26,
cp32-1, cp32-3, cp32-4, cp32-6, cp32-7, cp32-8, cp32-9, lp17, lp21, lp25, lp28-1, lp28-2, lp28-3,
lp28-4, lp36, lp38 and lp54. Plasmids cp9, lp5 and lp56 are absent from our strain.
The mapped Hi-C contact frequencies were stored in multi-resolution cooler files [75] and
the Hi-C matrices were balanced using the iterative correction and eigenvector decomposition
method [49]. The iterative correction method is a standard way to balance the Hi-C map such
that the rows and columns sum to a constant value (typically 1), which helps to correct for
biases in genomic coverage, for example some genomic regions might be easier to amplify
than other regions. The iterative correction process can be roughly summarized as follows.
Each individual value within a row is divided by the sum of values for that row to achieve a
sum of 1 for every row. However, this normalization of the rows breaks the required symmetry
of the Hi-C matrix. Therefore, row normalization is followed by column normalization in
which each individual value in a column is divided by the resulting sum of values for that col-
umn, which subsequently "unbalances" the rows and the row sum is no longer 1. As such, the
process is iteratively repeated until the row and column sums converge to 1 within a pre-
defined error tolerance for which we used the default value of 10−5. This results in a balanced
Hi-C matrix in which genomic coverage biases are minimized. We described the process start-
ing with normalization of rows followed by columns. However, the procedure could equally
have been applied by starting with columns instead of rows since the Hi-C matrix is symmetric
about the primary diagonal. Unless otherwise specified, all Hi-C plots and downstream analy-
ses were performed with this iterative correction. For the renormalization of plasmid-chromo-
some and plasmid-plasmid interactions (Figs 3C, 4C, S9 and S12), the same procedure of
iterative correction was used.
Plots were generated with R or Python 3.8.15 using Matplotlib 3.6.2 [76]. Data were
retrieved for plotting at 5-kb resolution. Pc(s) curves show the averaged contact frequency
between all pairs of loci on the chromosome separated by set distance (s). The x-axis indicates
the genomic distance of separation in kb. The y-axis represents the averaged contact frequency
in a logarithmic scale. The curves were computed for data binned at 5 kb. For the log2 ratio
plots, the Hi-C matrix of each mutant was divided by the matrix of the control. Then, log2(mu-
tant/control) was calculated and plotted in a heatmap using R.
Indicating highly transcribed genes on a Hi-C map
The RNA-seq data of the B. burgdorferi B31-S9 strain growing in culture from a recent pub-
lished study (SRR22149536) [46] were mapped to WT B. burgdorferi B31 genome (NCBI
GCA_000008685.2_ASM868v2) using CLC Genomics Workbench (QIAGEN) as previously
described [7]. RNA-seq analysis was performed using the default setting of the built-in package
of CLC Genomics Workbench. Genes were ranked by transcripts per kilobase per million
reads (TPM). For the top 50 highly transcribed genes, the first nucleotide of each gene was
indicated with fine dotted lines and plotted on to the Hi-C map using R (S1B Fig).
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
17 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Clustering of strains based on Hi-C data
Clustering of strains based on the contact probability curves was done using the scikit-learn
1.1.3 k-means algorithm [54]. The optimal number of clusters was determined using the maxi-
mum of the Silhouette score. The Silhouette score, s(i) is a metric that determines, for some
collection of objects {i}, how well each individual object, i, matches the clustering at hand [77].
In our case, the collection of objects were the log-transformed contact frequency Pc(s) curves,
which were computed as the average value of the contact frequency of pairs of loci separated
by a fixed genomic distance. Average Silhouette scores were computed for data clustered using
k-means with varying the number of clusters ranging from 2 to 21. We found that the number
of clusters that maximized the average Silhouette score was six, suggesting that six is the opti-
mal number of clusters in the data.
To better visualize the results of the k-means clustering and Silhouette method of identify-
ing the optimal number of clusters, we visualized the data clusters using two different methods:
Principal Component Analysis (PCA) and t-distributed stochastic neighbor embedding (t-
SNE). PCA was performed using scikit learn 1.2.2 (sklearn.decomposition.PCA) [54] on the
log-transformed Pc(s) curves (computed for the chromosome only, ignoring plasmids) for
each of the 22 different Hi-C maps (11 strains, with 2 biological replicates each). To visualize
how the data clusters together, we projected the Pc(s) curve values from each experiment onto
the first two principal components, which explained approximately 85% of the total data vari-
ance (48% for component 1 and 37% for component 2). t-SNE was performed using scikit
learn 1.2.2 (sklearn.maniforld.TSNE) [54] on the same input data used for the PCA (see
above). We ran the t-SNE using the following parameters: n_components = 2, perplexity = 5,
init = "random", n_iter = 2000, random_state = 0. The results were subsequently plotted in a
two-dimensional graph, and the points of the scatter plot were labelled using the group classifi-
cations from application of the k-means clustering in Fig 5B.
Simulating plasmid-plasmid interaction frequencies based on plasmid sizes
and copy numbers
Plasmid-plasmid interaction frequencies were simulated assuming random collisions. We
accounted for either plasmid copy numbers alone, or in combination with information on the
plasmid lengths (Fig 1A). Plasmid copy numbers were previously determined using marker fre-
quency analysis [7], which yielded values ranging between 0.5 and 1.4 relative to the oriC (see
Fig 1A). Plasmid sizes ranged from 17–54 kb [3] (see Fig 1A), which covered 3–11 of 5-kb bins.
For the simulated plasmid-plasmid contact map using both the copy numbers and plasmid
lengths (S2A Fig), we first multiplied the average plasmid copy number by the plasmid lengths
in numbers of 5-kb bins and rounded the resulting number to the nearest integer, np for each
plasmid p. The values of np ranged between 2 and 14, and the total sum over all the plasmids,
p, was N = ∑pnp = 80. The simulated plasmid-plasmid “contact frequency” matrix was com-
puted using the probability of randomly drawing a given pair of plasmids. The probability for
drawing a plasmid, p, is np/N. The resulting probability matrix from this calculation can be
seen in S2A Fig (top panel). To best compare the simulated plasmid-plasmid contact probabil-
ity map with the experimental Hi-C data, we applied the iterative correction procedure [49] to
this map. The resulting matrix is shown both with the same scale bar as the experimental Hi-C
map (S2A Fig, middle panel) and with a very fine color scale (S2A Fig, bottom panel). We
note that the iterative correction scheme tends to minimize the effects of copy number varia-
tion from one genome segment to another and this is why the simulated plasmid-plasmid con-
tact map looks largely uniform when plotted with the same dynamic range as experimental
data (Figs 4D and S2 middle panel).
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
18 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
The simulated plasmid-plasmid contact map computed using only copy numbers was made
in a similar fashion (S2B Fig). For this method, instead of multiplying copy number by the
length of the plasmid, a fixed integer number was used (in our case, 10) to convert the relative
ratios into integer numbers. The method of computation was the same as that described above.
We made two assumptions for this simulation: 1) plasmids constitute independent units of
interaction, and 2) plasmids are “well mixed”. The “independence of contact” assumption
implies that there are no restrictions on how many DNA segments may be simultaneously in
contact with one another and the identity of the DNA segments in contact does not matter.
The “well mixed” assumption stipulates that independent DNA segments interact with equal
probability with other DNA segments. Together, these assumptions allow us to compute the
plasmid-plasmid interaction frequencies while safely ignoring other types of contacts such as
plasmid-chromosome and chromosome-chromosome contacts. Our simulation does not con-
sider the cytoplasmic volume.
Plasmid construction
Plasmid pΔmksB(gent) was generated in the following manner: (i) nucleotides 874996 through
876527 of the B31 chromosome were PCR-amplified with primers NT968 and NT969; (ii) the
gentamicin cassette of pKIGent_parSP1_phoU [7] was PCR-amplified with primers NT970
and NT971; (iii) nucleotides 879168 through 880691 of the B31 chromosome were PCR-ampli-
fied with primers NT972 and NT973; (iv) the suicide vector backbone of pΔparA(kan) [7] was
PCR-amplified with primers NT974 and NT975; and (v) the four PCR fragments listed above
were digested with DpnI (New England Biolabs), gel-purified, and subjected to Gibson assem-
bly [78] using New England Biolabs’ platform. The assembled plasmid was introduced into
Escherichia coli strain NEB 5-alpha (New England Biolabs) by heat shocking. The resulting
strain (CJW7512) was grown at 30˚C on LB plates or in Super Broth liquid medium with shak-
ing, while 15 μg/mL gentamicin was used for selection.
Strain construction
To generate strain CJW_Bb605, 75 μg of plasmid pΔmksB(gent) were digested with ApaLI (New
England Biolabs) in a 500 μL reaction volume for 4 hours. The DNA was then ethanol precipi-
tated [79], dried, and resuspended in 25 μL sterile water. The resulting DNA suspension was
then electroporated at 2.5 kV, 25 μF, 200 O, in 2 mm-gap cuvette [80, 81] into 100 μL of electro-
competent cells made [82] using B. burgdorferi strain S9. The electroporated bacteria were trans-
ferred immediately to 6 mL BSK-II medium and allowed to recover overnight at 34˚C. The next
day, a fraction of the culture was embedded in 25 mL of semisolid BSK-agarose medium con-
taining gentamicin per 10-cm round Petri dish, as previously described [83]. The semisolid BSK-
agarose mix was made by mixing 2 volumes of 1.7% agarose in water, sterilized by autoclaving,
then melted and pre-equilibrated at 55˚C, with 3 volumes of BSK-1.5 medium, which was also
equilibrated at 55˚C for at most 5 minutes. BSK-1.5 contained 69.4 g/L bovine serum albumin,
12.7 g/L CMRL-1066, 6.9 g/L Neopeptone, 3.5 g/L Yeastolate, 8.3 g/L HEPES, 6.9 g/L glucose,
6.4 g/L sodium bicarbonate, 1.1 g/L sodium pyruvate, 1.0 g/L sodium citrate, 0.6 g/L N-acetylglu-
cosamine, and 40 mL/L heat-inactivated rabbit serum, and had a final pH of 7.50. After 10 days
of growth in the BSK-agarose semisolid matrix, an individual colony was expanded in liquid cul-
ture and confirmed by PCR to have undergone correct double crossover homologous recombi-
nation of the suicide vector, thus yielding strain CJW_Bb605. This strain was also confirmed by
multiplex PCR [84] to contain all endogenous plasmids contained by its parent.
Requests for strains, plasmids, resources, reagents should be directed to and will be fulfilled
by the corresponding authors with appropriate Material Transfer Agreements.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
19 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Supporting information
S1 Fig. Hi-C interaction map of B. burgdorferi strain S9 shown in a different color scale.
(A) To better show the intra-chromosomal interactions in Fig 1B, the normalized Hi-C inter-
action map is shown in a different color scale. Black arrows point to a few examples of strong
CID boundaries that overlap with highly transcribed genes shown in (B). The color scale
depicting Hi-C interaction scores in arbitrary units is shown at the right. (B) The positions of
the top 50 highly transcribed chromosomal genes found by RNA-seq [46] are indicated using
fine black dotted lines. A recent study [46] published RNA-seq data of the B. burgdorferi
B31-S9 strain grown in culture. We mapped the data to the B. burgdorferi B31 genome, calcu-
lated the number of transcripts per kilobase per million reads for each gene, and indicated the
top 50 highly transcribed genes on the Hi-C map. Although the growth condition in our study
was different from the RNA-seq study [46], strong CIDs boundaries (black arrows in A)
largely overlap with highly transcribed genes.
(TIF)
S2 Fig. Simulated plasmid-plasmid interaction frequencies. The contact probability between
plasmids was simulated under the assumptions that plasmids are randomly interacting, inde-
pendent of one another, and are “well mixed” within the cytoplasm (see Materials and meth-
ods). The calculation was performed accounting for plasmid copy numbers and plasmid
lengths together (A) or only plasmid copy numbers (B). Top panels, the raw contact frequency
expected between plasmids without normalization. Middle panels, the simulated contact fre-
quency after normalization using iterative correction. Bottom panels, the same as middle pan-
els, but shown with a much finer color scale. The color scales depicting contact frequencies in
arbitrary units are shown at the right. We note that there is residual resemblance between bot-
tom and top panels, and in the bottom panel, the row or column sums do not appear to be the
same. This is because the iterative correction procedure stops when the row and column sums
approach 1 within a pre-defined error tolerance (see Materials and methods), but not exactly
at 1.
(TIF)
S3 Fig. Hi-C samples used in this study. The normalized Hi-C interaction maps of all 22
experiments done for this study. The color scale depicting Hi-C interaction scores is shown at
the bottom right.
(TIF)
S4 Fig. Individual Pc(s) curves of all the samples analyzed in this study. Pc(s) curves of all
22 Hi-C experiments done in this study. The x-axis indicates genomic distance while the y-axis
shows averaged contact frequency. Only intra-chromosomal interactions were used to calcu-
late the Pc(s) curves.
(TIF)
S5 Fig. Principal Component Analysis (PCA) with groups from k-means clustering results.
To better visualize the results of the k-means clustering generated by the Silhouette method,
we performed Principal Component Analysis (PCA) and labeled the clustering results (see
Materials and methods). The plots with up to six clusters gave nicely visually segregated
groups. Beyond six, the two-dimensional projections from PCA showed poor segregation of
the data points, and biological replicates were separated to different groups.
(TIF)
S6 Fig. T-distributed stochastic neighbor embedding (t-SNE) with groups from k-means
clustering results. To better visualize the results of the k-means clustering generated by the
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
20 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
Silhouette method, we performed t-distributed stochastic neighbor embedding (t-SNE) and
labeled the clustering results (see Materials and methods). Similar to PCA, the plots with up to
six clusters gave nicely visually segregated groups. Beyond six, the two-dimensional projections
from t-SNE showed poor segregation of the data points, and biological replicates were sepa-
rated to different groups.
(TIF)
S7 Fig. Comparison of WT and control strains. (A-B) Normalized Hi-C interaction maps of
B. burgdorferi strains S9 (WT) and the control strain CJW_Bb284. Two biological replicates of
each strain (rep1 and rep2) are shown. The color scale depicting Hi-C interaction scores in
arbitrary units is shown at the right. (C) Pc(s) curves of the four samples. Pc(s) curves show
the averaged contact frequency between all pairs of loci on the chromosome separated by set
distance (s). The x-axis indicates the genomic distance of separation in kb. The y-axis repre-
sents the averaged contact frequency. The curves were computed for data binned at 5 kb. Only
intra-chromosomal interactions were used to calculate the Pc(s) curves. (D-F) Log2 ratio plots
comparing different Hi-C matrices. Log2(matrix 1/matrix 2) was calculated and plotted in the
heatmaps. The identities of matrix 1/matrix 2 are shown at the top of each plot. The color scale
is shown at the right of panel (F).
(TIF)
S8 Fig. Plasmid-chromosome interactions in different mutants. Calculated plasmid-chro-
mosome interaction frequencies are shown. The x-axis shows chromosome location in kb. The
y-axis specifies the different plasmids analyzed. The color indicates the contact frequency
between each plasmid and chromosome locus. Each graph plots the mean value of the two bio-
logical replicates shown in S3 Fig. Data are binned at 5-kb resolution. The data were normal-
ized including all the interactions in the genome (i.e. intra-chromosomal, plasmid-
chromosome and plasmid-plasmid interactions).
(TIF)
S9 Fig. Renormalized plasmid-chromosome interactions in different mutants. Plasmid-
chromosome interactions from S8 Fig were renormalized using iterative correction to remove
the influence of intra-chromosomal and plasmid-plasmid interactions (see Materials and
methods). The data were normalized such that each row had the same total score, and each col-
umn had the same total score.
(TIF)
S10 Fig. Plasmid-chromosome interactions in different mutants organized by plasmids.
Calculated plasmid-chromosome interaction frequencies are shown. The x-axis shows the
chromosome location in kb. The y-axis specifies the different mutants. The color indicates the
contact frequency between each plasmid and chromosome locus. Each graph plots the mean
value of the two biological replicates shown in S3 Fig. Data are binned at 5-kb resolution.
(TIF)
S11 Fig. Plasmid-plasmid interactions in different mutants. Calculated plasmid-plasmid
contact frequencies in different strains. The x- and y-axes indicate the plasmids analyzed. The
color shows the computed contact frequency. Each graph plots the mean of the two biological
replicates shown in S3 Fig. The data were normalized including all the interactions in the
genome (i.e. intra-chromosomal, plasmid-chromosome and plasmid-plasmid interactions).
(TIF)
S12 Fig. Renormalized plasmid-plasmid interactions in different mutants. Plasmid-plas-
mid contact frequencies from S11 Fig were renormalized without plasmid-chromosome
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
21 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
interactions. The data were normalized such that each row had the same total score, and each
column had the same total score.
(TIF)
S1 Table. Bacterial strains used in this study.
(DOCX)
S2 Table. Plasmids used in this study.
(DOCX)
S3 Table. Oligonucleotides used in this study.
(DOCX)
S4 Table. Next-generation-sequencing samples used in this study.
(DOCX)
Acknowledgments
We thank the Wang and Jacobs-Wagner labs for discussions and support, and the Indiana
University Center for Genomics and Bioinformatics for assistance with high-throughput
sequencing.
Author Contributions
Conceptualization: Zhongqing Ren, Constantin N. Takacs, Christine Jacobs-Wagner, Xindan
Wang.
Data curation: Zhongqing Ren, Constantin N. Takacs.
Formal analysis: Zhongqing Ren, Constantin N. Takacs, Hugo B. Brandão.
Funding acquisition: Christine Jacobs-Wagner, Xindan Wang.
Investigation: Zhongqing Ren, Constantin N. Takacs.
Methodology: Zhongqing Ren, Constantin N. Takacs, Hugo B. Brandão.
Project administration: Christine Jacobs-Wagner, Xindan Wang.
Software: Zhongqing Ren, Hugo B. Brandão.
Supervision: Christine Jacobs-Wagner, Xindan Wang.
Validation: Zhongqing Ren, Constantin N. Takacs.
Visualization: Zhongqing Ren, Hugo B. Brandão.
Writing – original draft: Zhongqing Ren, Xindan Wang.
Writing – review & editing: Zhongqing Ren, Constantin N. Takacs, Hugo B. Brandão,
Christine Jacobs-Wagner, Xindan Wang.
References
1. Mead P. Epidemiology of Lyme Disease. Infect Dis Clin North Am. 2022; 36(3):495–521. Epub 2022/
09/19. https://doi.org/10.1016/j.idc.2022.03.004 PMID: 36116831.
2. Kugeler KJ, Schwartz AM, Delorey MJ, Mead PS, Hinckley AF. Estimating the Frequency of Lyme Dis-
ease Diagnoses, United States, 2010–2018. Emerg Infect Dis. 2021; 27(2):616–9. Epub 2021/01/27.
https://doi.org/10.3201/eid2702.202731 PMID: 33496229; PubMed Central PMCID: PMC7853543.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
22 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
3.
Fraser CM, Casjens S, Huang WM, Sutton GG, Clayton R, Lathigra R, et al. Genomic sequence of a
Lyme disease spirochaete, Borrelia burgdorferi. Nature. 1997; 390(6660):580–6. Epub 1997/12/24.
https://doi.org/10.1038/37551 PMID: 9403685.
4. Casjens S, Palmer N, van Vugt R, Huang WM, Stevenson B, Rosa P, et al. A bacterial genome in flux:
the twelve linear and nine circular extrachromosomal DNAs in an infectious isolate of the Lyme disease
spirochete Borrelia burgdorferi. Mol Microbiol. 2000; 35(3):490–516. Epub 2000/02/15. https://doi.org/
10.1046/j.1365-2958.2000.01698.x PMID: 10672174.
5. Schwartz I, Margos G, Casjens SR, Qiu WG, Eggers CH. Multipartite Genome of Lyme Disease Borre-
lia: Structure, Variation and Prophages. Curr Issues Mol Biol. 2021; 42:409–54. Epub 2020/12/18.
https://doi.org/10.21775/cimb.042.409 PMID: 33328355.
6.
7.
diCenzo GC, Finan TM. The Divided Bacterial Genome: Structure, Function, and Evolution. Microbiol
Mol Biol Rev. 2017; 81(3). Epub 2017/08/11. https://doi.org/10.1128/MMBR.00019-17 PMID:
28794225; PubMed Central PMCID: PMC5584315.
Takacs CN, Wachter J, Xiang Y, Ren Z, Karaboja X, Scott M, et al. Polyploidy, regular patterning of
genome copies, and unusual control of DNA partitioning in the Lyme disease spirochete. Nat Commun.
2022; 13(1):7173. Epub 2022/12/01. https://doi.org/10.1038/s41467-022-34876-4 PMID: 36450725;
PubMed Central PMCID: PMC9712426.
8. Baxter JC, Funnell BE. Plasmid Partition Mechanisms. Microbiol Spectr. 2014; 2(6). Epub 2015/06/25.
https://doi.org/10.1128/microbiolspec.PLAS-0023-2014 PMID: 26104442.
9. Guilhas B, Le Gall A, Nollmann M. Physical Views on ParABS-Mediated DNA Segregation. Adv Exp
Med Biol. 2020; 1267:45–58. Epub 2020/09/08. https://doi.org/10.1007/978-3-030-46886-6_3 PMID:
32894476.
10.
Jalal ASB, Le TBK. Bacterial chromosome segregation by the ParABS system. Open Biol. 2020; 10
(6):200097. Epub 2020/06/17. https://doi.org/10.1098/rsob.200097 PMID: 32543349; PubMed Central
PMCID: PMC7333895.
11. Surovtsev IV, Jacobs-Wagner C. Subcellular Organization: A Critical Feature of Bacterial Cell Replica-
tion. Cell. 2018; 172(6):1271–93. Epub 2018/03/10. https://doi.org/10.1016/j.cell.2018.01.014 PMID:
29522747; PubMed Central PMCID: PMC5870143.
12. Ptacin JL, Lee SF, Garner EC, Toro E, Eckart M, Comolli LR, et al. A spindle-like apparatus guides bac-
terial chromosome segregation. Nat Cell Biol. 2010; 12(8):791–8. Epub 2010/07/27. https://doi.org/10.
1038/ncb2083 PMID: 20657594; PubMed Central PMCID: PMC3205914.
13. Sullivan NL, Marquis KA, Rudner DZ. Recruitment of SMC by ParB-parS organizes the origin region
and promotes efficient chromosome segregation. Cell. 2009; 137(4):697–707. Epub 2009/05/20.
https://doi.org/10.1016/j.cell.2009.04.044 PMID: 19450517; PubMed Central PMCID: PMC2892783.
14. Gruber S, Errington J. Recruitment of condensin to replication origin regions by ParB/SpoOJ promotes
chromosome segregation in B. subtilis. Cell. 2009; 137(4):685–96. Epub 2009/05/20. https://doi.org/10.
1016/j.cell.2009.02.035 PMID: 19450516.
15.
Fogel MA, Waldor MK. A dynamic, mitotic-like mechanism for bacterial chromosome segregation.
Genes Dev. 2006; 20(23):3269–82. Epub 2006/12/13. https://doi.org/10.1101/gad.1496506 PMID:
17158745; PubMed Central PMCID: PMC1686604.
16. Gerdes K, Howard M, Szardenings F. Pushing and pulling in prokaryotic DNA segregation. Cell. 2010;
141(6):927–42. Epub 2010/06/17. https://doi.org/10.1016/j.cell.2010.05.033 PMID: 20550930.
17.
Leonard TA, Butler PJ, Lowe J. Bacterial chromosome segregation: structure and DNA binding of the
Soj dimer—a conserved biological switch. EMBO J. 2005; 24(2):270–82. Epub 2005/01/07. https://doi.
org/10.1038/sj.emboj.7600530 PMID: 15635448; PubMed Central PMCID: PMC545817.
18. Motallebi-Veshareh M, Rouch DA, Thomas CM. A family of ATPases involved in active partitioning of
diverse bacterial plasmids. Mol Microbiol. 1990; 4(9):1455–63. Epub 1990/09/01. https://doi.org/10.
1111/j.1365-2958.1990.tb02056.x PMID: 2149583.
19. Vecchiarelli AG, Han YW, Tan X, Mizuuchi M, Ghirlando R, Biertumpfel C, et al. ATP control of dynamic
P1 ParA-DNA interactions: a key role for the nucleoid in plasmid partition. Mol Microbiol. 2010; 78
(1):78–91. Epub 2010/07/28. https://doi.org/10.1111/j.1365-2958.2010.07314.x PMID: 20659294;
PubMed Central PMCID: PMC2950902.
20. Rodionov O, Lobocka M, Yarmolinsky M. Silencing of genes flanking the P1 plasmid centromere. Sci-
ence. 1999; 283(5401):546–9. Epub 1999/01/23. https://doi.org/10.1126/science.283.5401.546 PMID:
9915704.
21. Osorio-Valeriano M, Altegoer F, Steinchen W, Urban S, Liu Y, Bange G, et al. ParB-type DNA Segrega-
tion Proteins Are CTP-Dependent Molecular Switches. Cell. 2019; 179(7):1512–24 e15. Epub 2019/12/
14. https://doi.org/10.1016/j.cell.2019.11.015 PMID: 31835030.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
23 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
22. Murray H, Ferreira H, Errington J. The bacterial chromosome segregation protein Spo0J spreads along
DNA from parS nucleation sites. Mol Microbiol. 2006; 61(5):1352–61. Epub 2006/08/24. https://doi.org/
10.1111/j.1365-2958.2006.05316.x PMID: 16925562.
23.
24.
Lee MJ, Liu CH, Wang SY, Huang CT, Huang H. Characterization of the Soj/Spo0J chromosome segre-
gation proteins and identification of putative parS sequences in Helicobacter pylori. Biochem Biophys
Res Commun. 2006; 342(3):744–50. Epub 2006/02/24. https://doi.org/10.1016/j.bbrc.2006.01.173
PMID: 16494844.
Jakimowicz D, Chater K, Zakrzewska-Czerwinska J. The ParB protein of Streptomyces coelicolor A3(2)
recognizes a cluster of parS sequences within the origin-proximal region of the linear chromosome. Mol
Microbiol. 2002; 45(5):1365–77. Epub 2002/09/05. https://doi.org/10.1046/j.1365-2958.2002.03102.x
PMID: 12207703.
25. Soh YM, Davidson IF, Zamuner S, Basquin J, Bock FP, Taschner M, et al. Self-organization of parS
centromeres by the ParB CTP hydrolase. Science. 2019; 366(6469):1129–33. Epub 2019/10/28.
https://doi.org/10.1126/science.aay3965 PMID: 31649139; PubMed Central PMCID: PMC6927813.
26. Radnedge L, Youngren B, Davis M, Austin S. Probing the structure of complex macromolecular interac-
tions by homolog specificity scanning: the P1 and P7 plasmid partition systems. EMBO J. 1998; 17
(20):6076–85. Epub 1998/10/17. https://doi.org/10.1093/emboj/17.20.6076 PMID: 9774351; PubMed
Central PMCID: PMC1170934.
27. Surovtsev IV, Campos M, Jacobs-Wagner C. DNA-relay mechanism is sufficient to explain ParA-
dependent intracellular transport and patterning of single and multiple cargos. Proc Natl Acad Sci U S
A. 2016; 113(46):E7268–E76. Epub 2016/11/02. https://doi.org/10.1073/pnas.1616118113 PMID:
27799522; PubMed Central PMCID: PMC5135302.
28. Surovtsev IV, Lim HC, Jacobs-Wagner C. The Slow Mobility of the ParA Partitioning Protein Underlies
Its Steady-State Patterning in Caulobacter. Biophys J. 2016; 110(12):2790–9. Epub 2016/06/23.
https://doi.org/10.1016/j.bpj.2016.05.014 PMID: 27332137; PubMed Central PMCID: PMC4919595.
29. Hu L, Vecchiarelli AG, Mizuuchi K, Neuman KC, Liu J. Brownian Ratchet Mechanism for Faithful Segre-
gation of Low-Copy-Number Plasmids. Biophys J. 2017; 112(7):1489–502. Epub 2017/04/14. https://
doi.org/10.1016/j.bpj.2017.02.039 PMID: 28402891; PubMed Central PMCID: PMC5390091.
30.
Lim HC, Surovtsev IV, Beltran BG, Huang F, Bewersdorf J, Jacobs-Wagner C. Evidence for a DNA-
relay mechanism in ParABS-mediated chromosome segregation. Elife. 2014; 3:e02758. Epub 2014/05/
27. https://doi.org/10.7554/eLife.02758 PMID: 24859756; PubMed Central PMCID: PMC4067530.
31. Walter JC, Dorignac J, Lorman V, Rech J, Bouet JY, Nollmann M, et al. Surfing on Protein Waves: Pro-
teophoresis as a Mechanism for Bacterial Genome Partitioning. Phys Rev Lett. 2017; 119(2):028101.
Epub 2017/07/29. https://doi.org/10.1103/PhysRevLett.119.028101 PMID: 28753349.
32. Sugawara T, Kaneko K. Chemophoresis as a driving force for intracellular organization: Theory and
application to plasmid partitioning. Biophysics (Nagoya-shi). 2011; 7:77–88. Epub 2011/09/11. https://
doi.org/10.2142/biophysics.7.77 PMID: 27857595; PubMed Central PMCID: PMC5036777.
33. Wang X, Brandao HB, Le TB, Laub MT, Rudner DZ. Bacillus subtilis SMC complexes juxtapose chro-
mosome arms as they travel from origin to terminus. Science. 2017; 355(6324):524–7. Epub 2017/02/
06. https://doi.org/10.1126/science.aai8982 PMID: 28154080; PubMed Central PMCID: PMC5484144.
34. Wang X, Le TB, Lajoie BR, Dekker J, Laub MT, Rudner DZ. Condensin promotes the juxtaposition of
DNA flanking its loading site in Bacillus subtilis. Genes Dev. 2015; 29(15):1661–75. Epub 2015/08/09.
https://doi.org/10.1101/gad.265876.115 PMID: 26253537; PubMed Central PMCID: PMC4536313.
35.
36.
37.
Tran NT, Laub MT, Le TBK. SMC Progressively Aligns Chromosomal Arms in Caulobacter crescentus
but Is Antagonized by Convergent Transcription. Cell Rep. 2017; 20(9):2057–71. Epub 2017/08/31.
https://doi.org/10.1016/j.celrep.2017.08.026 PMID: 28854358; PubMed Central PMCID: PMC5583512.
Le TB, Imakaev MV, Mirny LA, Laub MT. High-resolution mapping of the spatial organization of a bacte-
rial chromosome. Science. 2013; 342(6159):731–4. Epub 2013/10/26. https://doi.org/10.1126/science.
1242059 PMID: 24158908; PubMed Central PMCID: PMC3927313.
Lioy VS, Cournac A, Marbouty M, Duigou S, Mozziconacci J, Espeli O, et al. Multiscale Structuring of
the E. coli Chromosome by Nucleoid-Associated and Condensin Proteins. Cell. 2018; 172(4):771–83
e18. Epub 2018/01/24. https://doi.org/10.1016/j.cell.2017.12.027 PMID: 29358050.
38. Bohm K, Giacomelli G, Schmidt A, Imhof A, Koszul R, Marbouty M, et al. Chromosome organization by
a conserved condensin-ParB system in the actinobacterium Corynebacterium glutamicum. Nat Com-
mun. 2020; 11(1):1485. Epub 2020/03/22. https://doi.org/10.1038/s41467-020-15238-4 PMID:
32198399; PubMed Central PMCID: PMC7083940.
39.
Lioy VS, Junier I, Lagage V, Vallet I, Boccard F. Distinct Activities of Bacterial Condensins for Chromo-
some Management in Pseudomonas aeruginosa. Cell Rep. 2020; 33(5):108344. Epub 2020/11/05.
https://doi.org/10.1016/j.celrep.2020.108344 PMID: 33147461.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
24 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
40. Cockram C, Thierry A, Gorlas A, Lestini R, Koszul R. Euryarchaeal genomes are folded into SMC-
dependent loops and domains, but lack transcription-mediated compartmentalization. Mol Cell. 2021;
81(3):459–72 e10. Epub 2021/01/01. https://doi.org/10.1016/j.molcel.2020.12.013 PMID: 33382984.
41. Ren Z, Liao Q, Karaboja X, Barton IS, Schantz EG, Mejia-Santana A, et al. Conformation and dynamic
interactions of the multipartite genome in Agrobacterium tumefaciens. Proc Natl Acad Sci U S A. 2022;
119(6). Epub 2022/02/02. https://doi.org/10.1073/pnas.2115854119 PMID: 35101983.
42. Ren Z, Liao Q, Barton IS, Wiesler EE, Fuqua C, Wang X. Centromere Interactions Promote the Mainte-
nance of the Multipartite Genome in Agrobacterium tumefaciens. mBio. 2022; 13(3):e0050822. Epub
2022/05/11. https://doi.org/10.1128/mbio.00508-22 PMID: 35536004; PubMed Central PMCID:
PMC9239152.
43. Huang YF, Liu L, Wang F, Yuan XW, Chen HC, Liu ZF. High-Resolution 3D Genome Map of Brucella
Chromosomes in Exponential and Stationary Phases. Microbiol Spectr. 2023:e0429022. Epub
20230227. https://doi.org/10.1128/spectrum.04290-22 PMID: 36847551.
44. Val ME, Marbouty M, de Lemos Martins F, Kennedy SP, Kemble H, Bland MJ, et al. A checkpoint con-
trol orchestrates the replication of the two chromosomes of Vibrio cholerae. Sci Adv. 2016; 2(4):
e1501914. Epub 2016/05/07. https://doi.org/10.1126/sciadv.1501914 PMID: 27152358; PubMed Cen-
tral PMCID: PMC4846446.
45. Hoffman EA, Frey BL, Smith LM, Auble DT. Formaldehyde crosslinking: a tool for the study of chromatin
complexes. J Biol Chem. 2015; 290(44):26404–11. Epub 20150909. https://doi.org/10.1074/jbc.R115.
651679 PMID: 26354429; PubMed Central PMCID: PMC4646298.
46. Sapiro AL, Hayes BM, Volk RF, Zhang JY, Brooks DM, Martyn C, et al. Longitudinal map of transcrip-
tome changes in the Lyme pathogen Borrelia burgdorferi during tick-borne transmission. Elife. 2023;
Reviewed Preprint. https://doi.org/10.7554/eLife.86636 PMID: 37449477
47.
Lioy VS, Lorenzi JN, Najah S, Poinsignon T, Leh H, Saulnier C, et al. Dynamics of the compartmental-
ized Streptomyces chromosome during metabolic differentiation. Nat Commun. 2021; 12(1):5221.
Epub 2021/09/03. https://doi.org/10.1038/s41467-021-25462-1 PMID: 34471117; PubMed Central
PMCID: PMC8410849.
48. Szafran MJ, Malecki T, Strzalka A, Pawlikiewicz K, Dulawa J, Zarek A, et al. Spatial rearrangement of
the Streptomyces venezuelae linear chromosome during sporogenic development. Nat Commun.
2021; 12(1):5222. Epub 2021/09/03. https://doi.org/10.1038/s41467-021-25461-2 PMID: 34471115;
PubMed Central PMCID: PMC8410768.
49.
Imakaev M, Fudenberg G, McCord RP, Naumova N, Goloborodko A, Lajoie BR, et al. Iterative correc-
tion of Hi-C data reveals hallmarks of chromosome organization. Nature methods. 2012; 9(10):999–
1003. https://doi.org/10.1038/nmeth.2148 PMID: 22941365; PubMed Central PMCID: PMC3816492.
50. Uhlmann F. SMC complexes: from DNA to chromosomes. Nat Rev Mol Cell Biol. 2016; 17(7):399–412.
Epub 2016/04/15. https://doi.org/10.1038/nrm.2016.30 PMID: 27075410.
51. Yatskevich S, Rhodes J, Nasmyth K. Organization of Chromosomal DNA by SMC Complexes. Annu
Rev Genet. 2019; 53:445–82. Epub 2019/10/03. https://doi.org/10.1146/annurev-genet-112618-
043633 PMID: 31577909.
52. Wang X, Montero Llopis P, Rudner DZ. Organization and segregation of bacterial chromosomes. Nat
Rev Genet. 2013; 14(3):191–203. Epub 2013/02/13. https://doi.org/10.1038/nrg3375 PMID: 23400100;
PubMed Central PMCID: PMC3869393.
53. Badrinarayanan A, Le TB, Laub MT. Bacterial chromosome organization and segregation. Annu Rev
Cell Dev Biol. 2015; 31:171–99. https://doi.org/10.1146/annurev-cellbio-100814-125211 PMID:
26566111; PubMed Central PMCID: PMC4706359.
54. Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: Machine learn-
ing in Python. the Journal of machine Learning research. 2011; 12:2825–30.
55. Kim H, Loparo JJ. Multistep assembly of DNA condensation clusters by SMC. Nat Commun. 2016;
7:10200. https://doi.org/10.1038/ncomms10200 PMID: 26725510; PubMed Central PMCID:
PMC4725763.
56. Davidson IF, Bauer B, Goetz D, Tang W, Wutz G, Peters JM. DNA loop extrusion by human cohesin.
Science. 2019; 366(6471):1338–45. Epub 2019/11/23. https://doi.org/10.1126/science.aaz3418 PMID:
31753851.
57. Ganji M, Shaltiel IA, Bisht S, Kim E, Kalichava A, Haering CH, et al. Real-time imaging of DNA loop
extrusion by condensin. Science. 2018; 360(6384):102–5. Epub 2018/02/24. https://doi.org/10.1126/
science.aar7831 PMID: 29472443.
58. Kim Y, Shi Z, Zhang H, Finkelstein IJ, Yu H. Human cohesin compacts DNA by loop extrusion. Science.
2019; 366(6471):1345–9. Epub 2019/11/30. https://doi.org/10.1126/science.aaz4475 PMID:
31780627.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
25 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
59.
Terakawa T, Bisht S, Eeftens JM, Dekker C, Haering CH, Greene EC. The condensin complex is a
mechanochemical motor that translocates along DNA. Science. 2017; 358(6363):672–6. Epub 2017/
09/09. https://doi.org/10.1126/science.aan6516 PMID: 28882993; PubMed Central PMCID:
PMC5862036.
60. Chaconas G, Norris SJ. Peaceful coexistence amongst Borrelia plasmids: getting by with a little help
from their friends? Plasmid. 2013; 70(2):161–7. Epub 2013/06/04. https://doi.org/10.1016/j.plasmid.
2013.05.002 PMID: 23727020; PubMed Central PMCID: PMC3737319.
61.
62.
Lin T, Gao L, Zhang C, Odeh E, Jacobs MB, Coutte L, et al. Analysis of an ordered, comprehensive
STM mutant library in infectious Borrelia burgdorferi: insights into the genes required for mouse infectiv-
ity. PLoS One. 2012; 7(10):e47532. Epub 2012/11/08. https://doi.org/10.1371/journal.pone.0047532
PMID: 23133514; PubMed Central PMCID: PMC3485029.
Tilly K, Checroun C, Rosa PA. Requirements for Borrelia burgdorferi plasmid maintenance. Plasmid.
2012; 68(1):1–12. Epub 2012/02/01. https://doi.org/10.1016/j.plasmid.2012.01.009 PMID: 22289894;
PubMed Central PMCID: PMC3367046.
63. Burns LH, Adams CA, Riley SP, Jutras BL, Bowman A, Chenail AM, et al. BpaB, a novel protein
encoded by the Lyme disease spirochete’s cp32 prophages, binds to erp Operator 2 DNA. Nucleic
Acids Res. 2010; 38(16):5443–55. Epub 2010/04/28. https://doi.org/10.1093/nar/gkq284 PMID:
20421207; PubMed Central PMCID: PMC2938228.
64. Hayes SF, Burgdorfer W, Barbour AG. Bacteriophage in the Ixodes dammini spirochete, etiological
agent of Lyme disease. J Bacteriol. 1983; 154(3):1436–9. https://doi.org/10.1128/jb.154.3.1436-1439.
1983 PMID: 6853449; PubMed Central PMCID: PMC217620.
65. Neubert U, Schaller M, Januschke E, Stolz W, Schmieger H. Bacteriophages induced by ciprofloxacin
in a Borrelia burgdorferi skin isolate. Zentralbl Bakteriol. 1993; 279(3):307–15. https://doi.org/10.1016/
s0934-8840(11)80363-4 PMID: 8219501.
66. Wachter J, Cheff B, Hillman C, Carracoi V, Dorward DW, Martens C, et al. Coupled induction of pro-
phage and virulence factors during tick transmission of the Lyme disease spirochete. Nat Commun.
2023; 14(1):198. Epub 20230113. https://doi.org/10.1038/s41467-023-35897-3 PMID: 36639656;
PubMed Central PMCID: PMC9839762.
67. Eggers CH, Samuels DS. Molecular evidence for a new bacteriophage of Borrelia burgdorferi. J Bacter-
iol. 1999; 181(23):7308–13. Epub 1999/11/26. https://doi.org/10.1128/JB.181.23.7308-7313.1999
PMID: 10572135; PubMed Central PMCID: PMC103694.
68.
Zhang H, Marconi RT. Demonstration of cotranscription and 1-methyl-3-nitroso-nitroguanidine induction
of a 30-gene operon of Borrelia burgdorferi: evidence that the 32-kilobase circular plasmids are pro-
phages. J Bacteriol. 2005; 187(23):7985–95. https://doi.org/10.1128/JB.187.23.7985-7995.2005 PMID:
16291672; PubMed Central PMCID: PMC1291276.
69. Barbour AG. Isolation and cultivation of Lyme disease spirochetes. Yale J Biol Med. 1984; 57(4):521–5.
Epub 1984/07/01. PMID: 6393604; PubMed Central PMCID: PMC2589996.
70.
Zuckert WR. Laboratory maintenance of Borrelia burgdorferi. Curr Protoc Microbiol. 2007;Chapter 12:
Unit 12C 1. Epub 2008/09/05. https://doi.org/10.1002/9780471729259.mc12c01s4 PMID: 18770608.
71. Bono JL, Elias AF, Kupko JJ 3rd, Stevenson B, Tilly K, Rosa P. Efficient targeted mutagenesis in Borre-
lia burgdorferi. J Bacteriol. 2000; 182(9):2445–52. Epub 2000/04/13. https://doi.org/10.1128/JB.182.9.
2445–2452.2000 PMID: 10762244; PubMed Central PMCID: PMC111306.
72.
Frank KL, Bundle SF, Kresge ME, Eggers CH, Samuels DS. aadA confers streptomycin resistance in
Borrelia burgdorferi. J Bacteriol. 2003; 185(22):6723–7. Epub 2003/11/05. https://doi.org/10.1128/JB.
185.22.6723-6727.2003 PMID: 14594849; PubMed Central PMCID: PMC262111.
73. Elias AF, Bono JL, Kupko JJ 3rd, Stewart PE, Krum JG, Rosa PA. New antibiotic resistance cassettes
suitable for genetic studies in Borrelia burgdorferi. J Mol Microbiol Biotechnol. 2003; 6(1):29–40. Epub
2003/11/01. https://doi.org/10.1159/000073406 PMID: 14593251.
74. Nowalk AJ, Gilmore RD Jr., Carroll JA. Serologic proteome analysis of Borrelia burgdorferi membrane-
associated proteins. Infect Immun. 2006; 74(7):3864–73. Epub 2006/06/23. https://doi.org/10.1128/IAI.
00189-06 PMID: 16790758; PubMed Central PMCID: PMC1489744.
75. Abdennur N, Mirny LA. Cooler: scalable storage for Hi-C data and other genomically labeled arrays. Bio-
informatics. 2020; 36(1):311–6. Epub 2019/07/11. https://doi.org/10.1093/bioinformatics/btz540 PMID:
31290943; PubMed Central PMCID: PMC8205516.
76. Hunter JD. Matplotlib: A 2D graphics environment. Comput Sci Eng. 2007; 9(3):90–5. https://doi.org/10.
1109/Mcse.2007.55 WOS:000245668100019.
77. Rousseeuw PJ. Silhouettes—a Graphical Aid to the Interpretation and Validation of Cluster-Analysis. J
Comput Appl Math. 1987; 20:53–65. https://doi.org/10.1016/0377-0427(87)90125-7 WOS:
A1987L111800005.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
26 / 27
PLOS GENETICS Organization of a highly segmented bacterial genome
78. Gibson DG, Young L, Chuang RY, Venter JC, Hutchison CA 3rd, Smith HO. Enzymatic assembly of
DNA molecules up to several hundred kilobases. Nature methods. 2009; 6(5):343–5. Epub 2009/04/14.
https://doi.org/10.1038/nmeth.1318 PMID: 19363495.
79. Green MR, Sambrook J. Precipitation of DNA with Ethanol. Cold Spring Harb Protoc. 2016;2016(12).
Epub 2016/12/10. https://doi.org/10.1101/pdb.prot093377 PMID: 27934690.
80. Samuels DS. Electrotransformation of the spirochete Borrelia burgdorferi. Methods Mol Biol. 1995;
47:253–9. Epub 1995/01/01. https://doi.org/10.1385/0-89603-310-4:253 PMID: 7550741; PubMed
Central PMCID: PMC5815860.
81. Samuels DS, Drecktrah D, Hall LS. Genetic Transformation and Complementation. Methods Mol Biol.
2018; 1690:183–200. Epub 2017/10/17. https://doi.org/10.1007/978-1-4939-7383-5_15 PMID:
29032546; PubMed Central PMCID: PMC5806694.
82.
83.
Tilly K, Elias AF, Bono JL, Stewart P, Rosa P. DNA exchange and insertional inactivation in spirochetes.
J Mol Microbiol Biotechnol. 2000; 2(4):433–42. Epub 2000/11/15. PMID: 11075915.
Takacs CN, Scott M, Chang Y, Kloos ZA, Irnov I, Rosa PA, et al. A CRISPR interference platform for
selective downregulation of gene expression in Borrelia burgdorferi. Appl Environ Microbiol. 2020; 87
(4). Epub 2020/12/02. https://doi.org/10.1128/AEM.02519-20 PMID: 33257311; PubMed Central
PMCID: PMC7851697.
84. Bunikis I, Kutschan-Bunikis S, Bonde M, Bergstrom S. Multiplex PCR as a tool for validating plasmid
content of Borrelia burgdorferi. J Microbiol Methods. 2011; 86(2):243–7. Epub 2011/05/25. https://doi.
org/10.1016/j.mimet.2011.05.004 PMID: 21605603.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1010857 July 26, 2023
27 / 27
PLOS GENETICS
| null |
10.1038_s41598-020-65766-8.pdf
|
Data availability
Data are available at http://bicresources.jcbose.ac.in/ssaha4/drag/browse.php. Supporting figures and tables are
included in Supplementary Files 1–3.
|
Data availability Data are available at http://bicresources.jcbose.ac.in/ssaha4/drag/browse.php . Supporting figures and tables are included in Supplementary Files 1-3.
|
open
Survey of drug resistance
associated gene mutations in
Mycobacterium tuberculosis,
eSKApe and other bacterial species
Abhirupa Ghosh1,2, Saran n.1,2 & Sudipto Saha1 ✉
tuberculosis treatment includes broad-spectrum antibiotics such as rifampicin, streptomycin
and fluoroquinolones, which are also used against other pathogenic bacteria. We developed Drug
Resistance Associated Genes database (DRAGdb), a manually curated repository of mutational data of
drug resistance associated genes (DRAGs) across ESKAPE (i.e. Enterococcus faecium, Staphylococcus
aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter
spp.) pathogens, and other bacteria with a special focus on Mycobacterium tuberculosis (MTB). Analysis
of mutations in drug-resistant genes listed in DRAGdb suggested both homoplasy and pleiotropy to be
associated with resistance. Homoplasy was observed in six genes namely gidB, gyrA, gyrB, rpoB, rpsL
and rrs. For these genes, drug resistance-associated mutations at codon level were conserved in MTB,
ESKAPE and many other bacteria. Pleiotropy was exemplified by a single nucleotide mutation that
was associated with resistance to amikacin, gentamycin, rifampicin and vancomycin in Staphylococcus
aureus. DRAGdb data also revealed that mutations in some genes such as pncA, inhA, katG and
embA,B,C were specific to Mycobacterium species. for inhA and pncA, the mutations in the promoter
region along with those in coding regions were associated with resistance to isoniazid and pyrazinamide
respectively. In summary, the DRAGdb database is a compilation of all the major MTB drug resistance
genes across bacterial species, which allows identification of homoplasy and pleiotropy phenomena of
DRAGs.
There is a rise in the use of broad spectrum antibiotics such as rifamycins, aminoglycosides and fluoroquinolo-
nes against tuberculosis (TB), as well as common bacterial infections such as gastro-intestinal infections1–3. The
multi- and extensively drug-resistant (MDR and XDR) Mycobacterium tuberculosis (MTB) pose a global threat
to public health as new resistance mechanisms are developing and making the treatment for patients prolonged
and expensive. Drug resistance is not restricted to TB, but also observed in common bacterial infections such as
pneumonia and foodborne infections4,5. Genome-wide analysis of MDR and XDR MTB reveals that drug resist-
ance arises due to mutations in the gene and/or the promoter region. Drug resistance associated mutations are
linked to increasing drug efflux, modifications of the drugs or their targets6–8. The accessibility to next-generation
sequencing technologies and characterization of bacteria specific drug resistance allows the extensive study of
other pathogenic bacteria as well9–11. Antibiotic resistance mutations specific to pathogenic bacteria are available.
The Infectious Diseases Society of America has grouped Enterococcus faecium, Staphylococcus aureus, Klebsiella
pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa and Enterobacter spp. as ESKAPE pathogens that
are capable of ‘escaping’ the actions of antibiotics thereby developing antibiotic resistance12. The ESKAPE patho-
gens are the leading cause of Hospital-Acquired Infection (HAI) or nosocomial infection13. Thus, it is important
to understand the drug resistance mutations across ESKAPE species against tuberculosis drugs. Three major data-
bases, Tuberculosis Drug Resistance Mutation database (TBDReaMDB), MUtation BioInformatics Identification
(MUBII-TB-DB), Tuberculosis Drug resistance Database (TBDR) are currently available for mutations associated
with drug resistance in MTB14–16. TBDReaMDB lists information on mutations in 51 genes across both first and
second line TB drugs14. The major drawback of this database is that it has not been updated after 2009. Other
databases such as MUBII-TB-DB and TBDR cover only a small set of genes15,16.
1Division of Bioinformatics, Bose Institute, Kolkata, India. 2These authors contributed equally: Abhirupa Ghosh and
Saran N. ✉e-mail: [email protected]
Scientific RepoRtS | (2020) 10:8957 | https://doi.org/10.1038/s41598-020-65766-8
1
www.nature.com/scientificreports Drugs
Genes
DRAGdb
TBDReaMDB
MUBII-TB-DB
No. of mutations
No. of Novel mutations*
in DRAGdb
Ethambutol
embA,B,C
Fluoroquinolone
Isoniazid
Pyrazinamide
Rifampicin
Streptomycin
gyrA
gyrB
inhA
katG
pncA
rpoB
gidB
rpsL
rrs
273
105
69
30
542
1200
710
178
113
201
11
17
18
13
273
278
133
21
16
25
—
17
16
11
—
277
130
—
—
7
56
1
—
17
55
125
32
37
6
26
Table 1. Mycobacterium tuberculosis (MTB) gene mutations reported in DRAGdb compared with
TBDReaMDB and MUBII-TB-DB. *Reported only in DRAGdb.
Prolonged usage of broad spectrum antibiotics against TB may affect the lung microbiome as well as the intes-
tinal microbiome, which are connected by the “gut-lung axis”17–20. In addition, there may be potential horizontal
transfer of antibiotic resistant genes in the human microbiome21–23. Therefore, there is a need to combine the
information from all organism-specific studies into a single platform to have a complete idea of antibiotic resist-
ance associated mutations across bacterial species. In order to facilitate the characterization of mutations in Drug
Resistance Associated Genes (DRAGs) across bacterial species, we present DRAGdb, a manually curated database
that has enlisted DRAG mutations across bacterial communities focusing on drugs used to treat tuberculosis.
DRAGdb provides mutation information related to 6 drugs, a few of which are broad spectrum antibiotics and
12 associated genes across bacterial species including MTB, ESKAPE and other pathogens such as Escherichia
coli and Salmonella enterica. It also provides drug resistance patterns of non-pathogenic bacteria including
Staphylococcus epidermidis and Bifidobacterium species24,25. The mutational gene data analysis of DRAGdb high-
lights the concepts of homoplasy and pleiotropy26. Homoplasy is described as a phylogenetic event when a resist-
ance determining mutation arises in phylogeny under selection pressure across species or strains27. Another
major phylogenetic event occurs when a resistance determining mutation causes pleiotropic effects on resistance
to other drugs in a bacteria due to resistance selection28. In summary, DRAGdb is a manually curated database of
drug resistant genes of bacteria with a focus on TB drugs, which reveals that at least 6 genes carry drug resistance
mutations across bacterial species, whereas some drug resistance genes are specific to Mycobacterium species.
Results
Overview of DRAGdb. Database content. DRAGdb is a database of mutational information of DRAGs
across MTB clinical strains, ESKAPE bacteria and other pathogenic and non-pathogenic bacterial species with
special reference to MTB H37Rv. A systematic curation of mutations found in drug resistant bacteria from exist-
ing literature was compiled to create the database. Each mutation entry comprises of organism name, gene name
and corresponding identifier from Ensembl Bacteria database, the nucleotide position, the nucleotide change,
the amino acid codon position, the codon change, the type of mutation at the amino acid level, the sequenc-
ing method used to detect the mutation, the strain of the bacterial species, the geographical location of the
sample and PubMed identifier of the literature referred to. The PROVEAN scores predicting the functionality
of gene-mutations in different bacteria were added in DRAGdb and the full list of each entry is available in
Supplementary File. 2. DRAGdb contains 4653 mutation entries associated with 12 genes and 6 drugs across 126
bacterial species.
DRAGdb statistics. The basic statistics of DRAGdb is shown in Tables 1 and 2. In Table 1, the MTB gene
mutations were compared with existing MTB mutation databases such as TBDReaMDB and MUBII-TB-DB. It
was observed that DRAGdb has comparatively higher numbers of mutations for each gene than the other two
databases. Table 2 includes MTB genes that were also observed in ESKAPE pathogens and other pathogenic and
non-pathogenic bacterial species.
Mutation trends from DRAGdb. The literature survey led to compilation of mutational data across dif-
ferent bacterial species for the genes such as gidB, gyrA, gyrB, rpoB, rpsL and rrs (Table 3). Mutations in these
genes associated with drug resistance were observed in different bacterial species. However, the genes inhA, katG,
embA, embB, embC and pncA are specific to Mycobacterium species. The genes associated with drug resistance
across different bacterial species, were re-numbered using multiple sequence alignment at codon level with refer-
ence to MTB H37Rv, in order to obtain the most frequently mutated codon positions. The frequencies of impor-
tant drug resistance associated mutations with positions at codon level are represented in bar plots for gyrA, gyrB,
rpoB, and rpsL in Fig. 1(A–D).
common set of drug resistance genes across bacterial species. DRAGdb focuses on drugs associ-
ated with TB treatment regimens, and lists mutations in associated genes across bacterial genera. It also lists drug
resistance genes that are specific to MTB. MTB is relatively “young” from an evolutionary standpoint. It does not
Scientific RepoRtS | (2020) 10:8957 | https://doi.org/10.1038/s41598-020-65766-8
2
www.nature.com/scientificreportswww.nature.com/scientificreports/ Drugs
Genes
Bacterial Pathogens
No. of mutations
Fluoroquinolone
gyrA
gyrB
Rifampicin
rpoB
gidB
Streptomycin
rpsL
rrs
ESKAPE
Others
ESKAPE
Others
ESKAPE
Others
ESKAPE
Others
ESKAPE
Others
ESKAPE
Others
41
282
39
168
73
346
2
116
4
129
—
13
Table 2. ESKAPE and other bacterial species gene mutations reported in DRAGdb.
Gene
gidB
gyrA
gyrB
rpoB
rpsL
rrs
No. of organisms
No. of drugs
13
51
41
62
37
07
01
14
13
15
04
08
Table 3. Genes associated with resistance across a variety of organisms and resistance to a number of drugs.
Figure 1. (A–C) The frequency plots for gyrA, gyrB and rpoB respectively show mutational frequencies in the
top codon positions among all antibiotic resistance determining region codons in each gene. The frequency bars
are plotted for ESKAPE pathogens, Mycobacterium species and all other bacterial species in each gene. (D) The
frequency plot of rpsL shows mutational frequency of top codon positions among all reported codon positions
of mutation. RRDR stands for Rifampicin Resistance Determining Region; QRDR stands for Quinolone
Resistance Determining Region.
Scientific RepoRtS | (2020) 10:8957 | https://doi.org/10.1038/s41598-020-65766-8
3
www.nature.com/scientificreportswww.nature.com/scientificreports/ carry plasmids and is thus thought not to be engaged in horizontal gene transfer. However, it was observed that
mutations in DRAGs of MTB and other bacterial species including ESKAPE pathogens, and other pathogenic and
non-pathogenic bacteria, occur usually at the same codon position.
gidB. gidB also known as rsmG, was found to be associated with streptomycin resistance across 13 bacterial
species including Mycobacterium species, an ESKAPE pathogen i.e. S. aureus and other bacteria.
gyrA. DRAGdb lists gyrA mutations associated with resistance to second and third generation fluoroquinolo-
nes, nalidixic acid and triclosan. gyrA mutations were found in 42 bacteria including different Mycobacterium
species, all 6 ESKAPE pathogens and other bacterial species. The frequency plot of 3 important mutations in the
Quinolone Resistance Determining Region (QRDR) of gyrA at codon positions 90, 91 and 94 is shown in Fig. 1A
and the data is shown in Supplementary File 1: Table S1. The data shows that mutation at the 90th codon position
was more dominant in ESKAPE pathogens whereas mutations at the 91st and 94th codon positions occurred more
frequently in drug resistant MTB.
gyrB. Similar to gyrA, gyrB was also related to fluoroquinolone resistance. DRAGdb indicates that similar to
gyrA, most of the gyrB mutations were associated with resistance to nalidixic acid and various fluoroquinolones.
However, some gyrB mutations were associated with resistance to aminocoumarins, a group of gyrase inhibi-
tors which include novobiocin and coumermycin. gyrB mutations were found in 36 bacteria including differ-
ent Mycobacterium species, 5 ESKAPE pathogens and other bacterial species. The frequency plot of 4 important
mutations in the QRDR of gyrB is shown in Fig. 1B and the data is shown in Supplementary File 1: Table S2. The
codon at the 499th position was most frequently mutated in ESKAPE, MTB as well as other drug resistant bacteria.
rpoB. DRAGdb indicates that mutations in rpoB were not only responsible for resistance to the rifamycin
class of drugs including rifampicin, rifabutin, rifalazil, rifapentine and rifaximin, but also resistance to 10 other
drugs of various drug families in 62 bacterial species including Mycobacterium species, the ESKAPE pathogens
Acinetobacter baumannii, Enterococcus faecium, Pseudomonas aeruginosa, Staphylococcus aureus and many other
bacteria. The frequency plot of 3 crucial mutations in the Rifampicin Resistance Determining Region (RRDR) of
rpoB (as shown in Fig. 1C and in Supplementary File 1: Table S3) shows that mutations at codon positions 435,
445 and 450 exerted an additive effect on drug resistance. Thus no single dominant mutation is alone responsible
for resistance to rifamycins in MTB and ESKAPE.
rpsL is primarily associated with streptomycin resistance. However, mutations in rpsL also cause resistance
rpsL.
to other aminoglycosides such as kanamycin, amikacin and paromomycin. rpsL mutations were present across
37 bacteria including Mycobacterium species, an ESKAPE pathogen Kleibsella pneumoniae and other bacterial
species. The frequency plot of two dominant drug resistance associated mutations in rpsL at codon positions 43
and 88 is shown in Fig. 1D and the data is shown in Supplementary File 1: Table S4.
rrs.
rrs encodes 16S rRNA in bacteria and is associated with streptomycin resistance. DRAGdb shows its
involvement in resistance to 5 other aminoglycosides as well. Mutations in rrs were found in 7 bacterial species.
No mutation has been reported for the ESKAPE pathogens.
Homoplasy and pleiotropy. Multiple sequence alignments of the protein sequences corresponding to each
gene across the reported bacteria were performed as shown in Supplementary File 1: Fig. S1(A–D). Interestingly,
in some genes such as rpoB, gyrA, gyrB, gidB and rpsL similar points of mutations associated with drug resistance,
were observed across bacterial species. This could be due to common mechanisms associated with the bacterial
response to an antibiotic/drug29. Such occurrence of identical genotypes across drug resistant bacterial species is
termed here as homoplasy. The MTB H37Rv numbering system was used as reference in our analysis. An example
of homoplasy is a point mutation, Asp to Asn in rpoB at codon position 435 (MTB numbering). This mutation
was found in 12 bacterial species including MTB, ESKAPE pathogens such as Actinobacter baumannii and S.
aureus, other pathogenic bacteria including Helicobacter pylori and Haemophilus influenza and non-pathogenic
bacteria, for example, Deinococcus radiodurans and Streptomyces lividans. In Fig. 2A, the circular plot illustrates
some examples of homoplasy events in mutated codon positions across some of the reported bacterial species.
The data curated for DRAGdb also indicates in some bacteria, the presence of cross resistance towards mul-
tiple drugs due to a single point mutation. This phenomenon in which a single locus influences resistance to two
or more distinct drugs is defined here as pleiotropy. Some of the mutations in rpoB across bacterial species were
known to be associated with resistance to rifampicin and/or other rifamycins. However, an instance was found
in S. aureus where mutation in rpoB at codon position 477 and nucleotide position 1430 (S. aureus numbering)
was responsible for resistance to rifampicin, daptomycin, vancomycin and oxacillin30. In Fig. 2B, the circular plot
provides examples of nucleotide positions in genes in specific organisms where a single point mutation is associ-
ated with multi-drug resistance.
Drug resistance genes specific to MTB. MTB is assumed to engage very little in horizontal gene trans-
fer and thus considered inert or relatively young in evolutionary terms31. It also has an additional layer in its
outer membrane composed of novel lipids and polysaccharides such as mycolic acid that makes it an acid fast
bacterium32. The frequency analysis of 4 Mycobacterium specific genes with mutation entries, namely; inhA,
embB, katG and pncA are shown in Fig. 3(A–D) and numbers are shown in Supplementary File 1: Tables S5–S8.
Other than katG, all three lacked specific drug resistance determining regions. Drug resistance associated muta-
tions were present all through their coding and non-coding (promoter) regions. Mutation type distribution in
Scientific RepoRtS | (2020) 10:8957 | https://doi.org/10.1038/s41598-020-65766-8
4
www.nature.com/scientificreportswww.nature.com/scientificreports/ Figure 2. (A) The circular plot depicts examples of Homoplasy observed in DRAGdb. Rainbow colored
chords connect amino acid positions of the genes (rainbow colored grids) to bacteria (grey colored grids) that
have the same point mutation in a specific position associated with resistance to the same drug.The grid name
pattern is “gene + codon position”. (B) The circular plot depicts examples of Pleiotropy observed in DRAGdb
where rainbow colored chords connect nucleotide positions of the genes of specific bacteria (rainbow colored
grids) to drug names (grey colored grids) showing that a single nucleotide mutation causes multiple drug
resistance. The grid name pattern is “abbreviation of bacterial name + gene + nucleotide position”. SA stands
for Staphylococcus aureus, MS stands for Mycobacterium smegmatis and EF stands for Enterococcus faecium. The
circular plots were drawn using circlizeR package in R.
Figure 3. (A–D) The frequency plot of inhA, embB, katG and pncA respectively show mutational frequency
of the top codon [cyan] or promoter nucleotide [green] positions among all reported mutation points in each
gene. The frequency bars are plotted for each gene in all Mycobacterium species. The 11points in (D) includes -10,
10, 51, 54, 57, 68, 71, 96, 103, 132 and 142.
MTB specific genes namely inhA, embB, katG and pncA is shown in Supplementary File 1: Tables S9-S12 and
Supplementary File 1: Figures S2(A–D). It was observed that overall non-synonymous mutations in genic regions
dominated over other types of mutations. However, in inhA and pncA, mutations in promoter regions were also
associated with drug resistance.
Scientific RepoRtS | (2020) 10:8957 | https://doi.org/10.1038/s41598-020-65766-8
5
www.nature.com/scientificreportswww.nature.com/scientificreports/ inhA.
inhA codes for enoyl-ACP reductase and is the primary target of the first-line tuberculosis drug isoni-
azid33. Mutations in the -8 and -15 positions in the promoter region of inhA were found in 44% of the isoniazid
resistant clinical isolates of Mycobacterium species. The frequency plot is shown in Fig. 3A and the data is shown
in Supplementary File 1: Table S5. The distribution of mutation types in inhA associated with isoniazid resistance
is shown in Supplementary File 1: Figure S2A and Table S9.
embB. embB codes for arabinosyl transferase, an enzyme that plays a role in the polymerization of arabinose
into the arabinan of arabinogalactan34. It is one of the primary targets of the first line tuberculosis drug etham-
butol. Ethambutol inhibits the transfer of arabinogalacton to the cell wall. Mutations in codons 306 and 406 were
found in 25% of the ethambutol resistant MTB as shown in Fig. 3B. The data is shown in Supplementary File 1:
Table S6. The mutations were mainly observed in the coding region of embB as shown in Supplementary File 1:
Figure S2B and Table S10.
katG. katG encodes for a bifunctional enzyme with both catalase and peroxidase activity. It plays a role in
protecting Mycobacterium against toxic reactive oxygen species as well as in activating the first line drug isoni-
azid35,36. Mutations in 6 codon positions taken together account for 40% of the drug resistant clinical isolates as
shown in Fig. 3C. The data is shown in Supplementary File 1: Table S7. Mutation at codon 315 was found in 30%
of drug resistant MTB. The distribution of mutation types in katG associated with isoniazid resistance is shown in
Supplementary File 1: Figure S2C and Table S11.
pncA. pncA gene codes for pyrazinamidase, which converts the first line tuberculosis drug, pyrazinamide into
its active form, pyrazinoic acid37. Mutations in pncA and its promoter region results in resistance to pyrazinamide.
On comparing mutations from pyrazinamide resistant clinical strains of Mycobacterium species, it was observed
that the mutations were scattered throughout the promoter and the coding region. Overall, 11 sites of mutation
accounted for 21% of the mutations in clinical isolates of Mycobacterium species. This is shown in Fig. 3D and
data is given in Supplementary File 1: Table S8. The mutations in pncA were observed to be diverse in nature
(shown in Supplementary File 1: Figure S2D and Table S12).
Comparison with other databases and tools. There are several antibiotic resistance related databases as
listed in Supplementary File 1: Table S13 obtained from PubMed literature search. The contents of the databases such
as the bacterial species focused, data types, availability of mutation data were thoroughly studied and compared. Out
of these 17 databases, three were beta-lactamases related resources38–40, five were specific to single species such as
uCARE41 is for E.coli and TBDB, ReSeqTB42, TBDReaMDB and MUBII-TB-DB were dedicated to MTB. A compar-
ison of TBDreamDB and MUBII-TB-DB, two well-known databases for mutations associated with drug resistance
in MTB with DRAGdb is presented in Table 1. DRAGdb lists a higher number of gene mutations. There were nine
multispecies antimicrobial databases. Among them, MEGARes43, BacMet44, Resfams45 and Pathosystems Resource
Integration Center (PATRIC)46 contain bacterial genome and drug resistance genes but no mutation data is available
in them. There were two deprecated databases such as ARG-ANNOT47 and ARDB - Antibiotic Resistance Genes
Database48, however, ARG-ANNOT gene list is incorporated in MEGARes and ARDB is upgraded as CARD49.
Finally it was observed that only three databases contain updated drug resistance causing mutation data across spe-
cies namely; CARD, BARRGD50 and PointFinder51. The mutation data available in these three databases were down-
loaded for further analysis. A case study to compare the rpoB gene mutations associated with drug resistance in MTB
and ESKAPE pathogens along with Mycobacterium leprae, Escherichia coli, Enterococcus faecalis present in these
three databases and DRAGdb was done. A venn diagram in Fig. 4 shows that DRAGdb had 174 unique SNPs in rpoB
gene compared to other three databases. The unique list of rpoB mutations in DRAGdb as shown in supplementary
File. 3 comprises mainly of mutations in bacteria like Enterococcus faecium, Acinetobacter baumannii, Pseudomonas
aeruginosa and Enterococcus faecalis that were only available in DRAGdb and few mutation points in other bacteria
also. There are also some tools for the prediction of antibiotic resistance genes such as meta-MARC that predicts drug
resistance from metagenomics data52, and AMRFinder that uses hidden Markov model of BARRGD sequence data-
base to identify the genes related to drug resistance53. DRAGdb use its own drug resistance associated gene sequence
database at bacterial species level for Basic Local Alignment Search Tool (BLAST) search. Thus it allows users to
identify the best hit mutant sequence at species level. This cannot be achieved with AMRFinder.
Utility and limitations of DRAGdb. The benefit of DRAGdb is that it provides information on antibiotic
resistance related mutations across various bacterial species in a single platform. As shown in the schematic
diagram of DRAGdb in Fig. 5, in addition to browsing the mutation database, the BLAST tool is integrated for
prediction of drug resistance from a query sequence. Compared to existing databases, DRAGdb contains higher
numbers of, as well as unique drug resistance associated gene mutations as shown in Table 1. The caveat of this
version of DRAGdb, is that all the double or multiple mutations in a drug resistant gene are considered as separate
entries for each species and thus the overall effect of all drug resistant mutations is not presented in a comprehen-
sive manner in a specific search. Further, the effect of mutations in multiple genes in MDR, for example gyrA and
rpoB, cannot be obtained in a single search. However, the BROWSE page of DRAGdb allows users to get all the
information in tabular format.
Discussion
In the recent past, due to the increasing availability of next generation technologies, a large number of stud-
ies have been carried out to unravel the specificity of drug resistance in many pathogenic bacteria. Here, we
describe DRAGdb a database that contains mutational data across MTB, ESKAPE pathogens, other path-
ogenic bacteria such as those causing sexually transmitted infections (Neisseria gonorrhoeae), foodborne
Scientific RepoRtS | (2020) 10:8957 | https://doi.org/10.1038/s41598-020-65766-8
6
www.nature.com/scientificreportswww.nature.com/scientificreports/ Figure 4. The venn diagram for the comparison of rpoB mutations in MTB and ESKAPE pathogens along
with Mycobacterium leprae, Escherichia coli, Enterococcus faecalis among CARD, BARRGD, PointFinder and
DRAGdb. BARRGD has rpoB mutations of Proteobacteria at phylum level thus it has no common entry at
species level.
Figure 5. Schematic architecture of DRAGdb online database.
infections (Campylobacter jejuni), skin infections (Streptococcus pyogenes), and non-pathogenic organisms such as
Bifidobacterium species. Compared to the existing TB mutation databases such as TBDReaMDB, MUBII-TB-DB
and TBDR, DRAGdb data carries more extensive mutational data14–16. DRAGdb data indicates the presence of
similar mutation patterns in 6 drug resistant genes, namely rpoB, gyrA, gyrB, gidB, rrs and rpsL across bacterial
ecosystems, that in turn highlights the drawbacks of using broad spectrum antibiotics for prolonged treatment of
diseases such as tuberculosis1–5. We suggest that prolonged exposure to drugs required for the treatment of TB,
leads to occurrence of resistance across bacterial populations in the gut microbiome that may hinder treatment
of other bacterial infections20. However, on a positive note, identifying a common cause of resistance across a
wide range of bacterial species opens up the possibility of designing diagnostic tools and identifying specific drug
targets for a wide range of bacterial infections. The data presented here points to the occurrence of resistance in
many pathogenic bacterial species along with the MTB clinical strains, the ESKAPE pathogens and commensal or
non-pathogenic bacteria. There is an urgent need to focus on the purportedly under-rated pathogens which may
cause severe health problems in the near future due to homoplasy and pleiotropy. DRAGdb also indicates that
for the MTB specific drug resistance genes pncA, inhA, katG and embA, B, C in addition to the non-synonymous
mutations in coding region, the non-coding regions also play important roles associated with drug resistance.
This brings an additional layer of complexity to the mechanisms of drug resistance. Further, a systematic analysis
of mutations responsible for drug resistance in a bacterial community against specific drugs, is required to under-
stand the evolution in drug resistance genes in response to drug exposure.
Scientific RepoRtS | (2020) 10:8957 | https://doi.org/10.1038/s41598-020-65766-8
7
www.nature.com/scientificreportswww.nature.com/scientificreports/ conclusions
Antibiotic/drug resistance is a natural phenomenon in microbial populations and is a global health threat mak-
ing the usage of antibiotics to treat life threatening infections such as tuberculosis and pneumonia less and less
effective. Tuberculosis treatment requires broad spectrum antibiotic classes such as rifamycins, aminoglycosides
and fluoroquinolones that are also extensively used against other bacterial infections. To contribute towards the
analysis of the development of antibiotic/drug resistance we have developed the DRAGdb database. It is a free
online repository of mutations in genes associated with broad spectrum antibiotics across Mycobacterium species,
ESKAPE pathogens and other pathogenic and non-pathogenic bacteria, along with MTB specific drug resistance
genes associated with drugs such as pyrazinamide, isoniazid and ethambutol. The database can be easily searched
and browsed at http://bicresources.jcbose.ac.in/ssaha4/drag. DRAGdb also includes a BLAST search option to
predict drug resistance related mutations. Comparison and analysis of mutations in DRAGs across bacterial spe-
cies give a clear indication of two phylogenetic phenomena namely homoplasy and pleiotropy. Six genes (gidB,
gyrA, gyrB, rpoB, rpsL and rrs) were associated with drug resistance not only in MTB but also in ESKAPE and
other bacterial pathogens. For these genes, we analyzed coding regions using MSA where MTB H37Rv was used
as reference genome. Some genes (inhA, embB, katG and pncA) were specific to MTB. The promoter regions of
inhA and pncA were involved in drug resistance along with their genic regions. The study clearly indicates that
under the stress of drug exposure, the response is not random. Instead it follows a defined pattern across bacterial
communities.
Methods
Database implementation. DRAGdb comprises of a single table where each mutation entry is uniquely
identified with DRAGDB_ID as the primary key. The NUCLEOTIDE_POSITION, NUCLEOTIDE_CHANGE,
AMINOACID_POSITION, AMINOACID_CHANGE define the mutation point at both levels. The PUBMED_
ID provides PubMed identifier, hyperlink to PubMed database and ENSEMBL_BACTERIA_ID provides the gene
identifier.
DRAGdb was developed using the Apache HTTP 2.2.15 web server and MySQL 5.1.69. The PHP 5.3.3,
HTML, JavaScript and CSS were used to build the web interfaces of the database. The PHP-based web interfaces
execute the SQL queries dynamically. It is freely accessible at http://bicresources.jcbose.ac.in/ssaha4/drag.
Data curation. The PubMed database (till March 2018) was searched for studies that reported at least one
mutation in rpoB, pncA, inhA, katG, embA, embB, embC, gidB, rpsL, rrs, gyrA and gyrB associated with resistance
to rifampicin, pyrazinamide, isoniazid, ethambutol, streptomycin and fluoroquinolones respectively in MTB,
ESKAPE and other bacterial species. The literature was searched using advance search option of PubMed with the
terms: “Gene name (Abstract/title) AND Resistance (Abstract/title) AND mutation (Abstract/title) AND/ NOT
tuberculosis (Abstract/title)”. The combination of search terms helped to obtain instances with cross resistance
and multiple resistances. In total, 2548 unique publications were obtained from this search. The publications
that were missing full English text in public domain, or lacked relevant data or had ambiguous data were filtered
out. Around 604 publications were systematically reviewed to obtain mutational information. All the mutations
described in drug resistant bacterial strains in the literature were manually read, further curated and compiled in
the database. The devised methodology is given as workflow in Supplementary File 1: Figure S3.
Mutation data analysis with reference to MTB H37Rv. All the gene mutations reported in the litera-
ture across bacterial species have different numbering systems (NS) thus leading to genetic location inconsistency
and conflict. One of the examples of NS discrepancy is of gyrA in MTB, for which 4 different NS were found in
the literature54. For better understanding and comparison across species of a single gene, Mycobacterium tuber-
culosis H37Rv was selected as reference organism, further multiple sequence alignment (MSA) was performed at
amino acid codon level for each drug resistance gene to have single numbering system across all organisms. MSA
was performed on on-line Clustal Omega platform using default iterated mBed-like Clustering Guide-tree55,56.
The rational for choosing MTB as reference genome was due to the fact that exposure of 3–6 antibiotics includ-
ing broad spectrum antibiotics during TB treatment for 6 months results in known multiple drug resistance
phenotypes. The MSA of the regions of interest for genes such as gyrA, gyrB, rpoB, and rpsL were shown in
Supplementary File 1: Figures S1(A–D). The common reference number at the amino acid codons level of drug
resistance genes across bacterial species helped in calculating frequency of mutated codon positions in DRAGdb.
The frequency percentage was calculated using the following formulae –
F
xi
=
∑
N
xi
t
N
xi
i
=
j
×
100
where Fxi is the frequency percentage of ith codon or nucleotide position in a gene of x th group, x can be all organ-
isms, Mycobacterium, ESKAPE pathogens or other bacteria. Nxi is the number of mutation entries of ith codon or
nucleotide position in the gene of x th group in DRAGdb. ∑ = Ni
xi is the total number of mutation entries in the
drug resistance determining region (DRDR) of the gene, j is the starting codon or nucleotide position and t is the
end codon or nucleotide position of DRDR. The number of mutation entries was calculated based on report of a
single mutation across various PubMed literature. We assume that the higher the number of publications report-
ing a particular drug resistance determining gene mutation, the higher is the confidence of that mutation entry.
t
j
Functional effects of the mutations. The functional effects of the unique SNPs in drug-resistance genes
in different bacteria were predicted using PROVEAN webserver with Score thresholds for prediction as of −2.5.
Scientific RepoRtS | (2020) 10:8957 | https://doi.org/10.1038/s41598-020-65766-8
8
www.nature.com/scientificreportswww.nature.com/scientificreports/ The variants with score equal to or below of −2.5 were considered “deleterious”, and the variants with score of
above −2.5 were considered “neutral”57.
Blast search. A customized BLAST database was created with wild type and mutated small nucleotide
stretches of drug resistance determining regions of associated genes. The mutated sequences were modified wild
type sequences with incorporation of single mutations enlisted in DRAGdb. blastall, a package for BLAST search
was used58. formatdb utility from that package was used for converting nucleotide FASTA sequences to BLAST
database. blastn program was used to find similar sequences to query sequences in the BLAST database.
DRAGdb user interface. The ‘HOME’ page of DRAGdb web interface provides two different search options: 1)
keyword search: a single keyword can be searched specific to bacteria, resistant drugs, genes, geographical loca-
tion or ‘ALL’ option to search in any category. 2) Advance search: three fields are present where bacteria and gene
name are mandatory and drug name is optional. Both the search options will generate a table giving details of the
mutations related to the search and also provide the number of specific entries. The DRAGdb result pages also
contain hyperlinked Ensembl Bacteria IDs, PROVEAN score and PubMed IDs. To keep with the open access
policy, the result table can be downloaded by the users. The ‘BROWSE’ page allows users to browse DRAGdb
data in three categories: 6 drugs, 12 genes, and 126 bacterial species. It shows the comparison of DRAGdb data
with other tuberculosis databases namely, TBDReaMDB and MUBII-TB-DB. The ‘Organisms’ section is further
divided into 3 parts: ‘Mycobacterium tuberculosis’, ‘ESKAPE’ and ‘others’ which includes other bacterial species.
The entries within the three categories are linked to DRAGdb table and provide specific results with details of the
gene mutations. The nucleotide BLAST search with customized BLAST database is incorporated in the ‘TOOL’
page to determine whether the users input bacterial gene sequence is drug resistant. Users can define the ‘E-value’
for BLAST operation. The output page shows the user input sequence, the DRAGDB_ID of the best hit, the
BLAST score and E-value of the hit. ‘OTHER LINKS’ page is also included to help users find popular TB and
antibiotic resistance related databases and webservers. To guide users through DRAGdb, a ‘HELP’ page is also
presented in the online web server.
Data visualization. The bar plots for representation of frequency % of various codon level mutations of drug
resistance genes across bacterial species were drawn using Microsoft office excel. The circular plots for representa-
tions of homoplasy and pleiotropy were drawn using ‘circlize’ R package59.
Data availability
Data are available at http://bicresources.jcbose.ac.in/ssaha4/drag/browse.php. Supporting figures and tables are
included in Supplementary Files 1–3.
Received: 14 February 2019; Accepted: 9 April 2020;
Published: xx xx xxxx
References
1. Krause, K. M., Serio, A. W., Kane, T. R. & Connolly, L. E. Aminoglycosides: An Overview. Cold Spring Harb Perspect Med 6, https://
doi.org/10.1101/cshperspect.a027029 (2016).
2. Rothstein, D. M. R, Alone and in Combination. Cold Spring Harb Perspect Med 6, https://doi.org/10.1101/cshperspect.a027011
(2016).
3. Redgrave, L. S., Sutton, S. B. & Webber, M. A. & Piddock, L. J. Fluoroquinolone resistance: mechanisms, impact on bacteria, and role
in evolutionary success. Trends Microbiol. 22, 438–445, https://doi.org/10.1016/j.tim.2014.04.007 (2014).
4. World Health Organization. Antibiotic resistance, http://www.who.int/news-room/fact-sheets/detail/antibiotic-resistance (2018).
5. Sommer, M. O. A., Munck, C., Toft-Kehler, R. V. & Andersson, D. I. Prediction of antibiotic resistance: time for a new preclinical
paradigm? Nat. Rev. Microbiol. 15, 689–696, https://doi.org/10.1038/nrmicro.2017.75 (2017).
6. Kim, J. S. & Wood, T. K. Tolerant, Growing Cells from Nutrient Shifts Are Not Persister Cells. MBio 8, https://doi.org/10.1128/
mBio.00354-17 (2017).
7. McKeegan, K. S., Borges-Walmsley, M. I. & Walmsley, A. R. Microbial and viral drug resistance mechanisms. Trends Microbiol. 10,
S8–14 (2002).
8. Brauner, A., Fridman, O., Gefen, O. & Balaban, N. Q. Distinguishing between resistance, tolerance and persistence to antibiotic
treatment. Nat. Rev. Microbiol. 14, 320–330, https://doi.org/10.1038/nrmicro.2016.34 (2016).
9. Crofts, T. S., Gasparrini, A. J. & Dantas, G. Next-generation approaches to understand and combat the antibiotic resistome. Nat. Rev.
Microbiol. 15, 422–434, https://doi.org/10.1038/nrmicro.2017.28 (2017).
10. Kumar, P. et al. Molecular Insights into Antimicrobial Resistance Traits of Multidrug Resistant Enteric Pathogens isolated from
India. Sci. Rep. 7, 14468, https://doi.org/10.1038/s41598-017-14791-1 (2017).
11. Lakin, S. M. et al. MEGARes: an antimicrobial resistance database for high throughput sequencing. Nucleic Acids Res. 45,
D574–D580, https://doi.org/10.1093/nar/gkw1009 (2017).
12. Pendleton, J. N., Gorman, S. P. & Gilmore, B. F. Clinical relevance of the ESKAPE pathogens. Expert. Rev. Anti Infect. Ther. 11,
297–308, https://doi.org/10.1586/eri.13.12 (2013).
13. Quainoo, S. et al. Whole-Genome Sequencing of Bacterial Pathogens: the Future of Nosocomial Outbreak Analysis. Clin. Microbiol.
Rev. 30, 1015–1063, https://doi.org/10.1128/CMR.00016-17 (2017).
14. Sandgren, A. et al. Tuberculosis drug resistance mutation database. PLoS Med. 6, e2, https://doi.org/10.1371/journal.pmed.1000002
(2009).
15. Flandrois, J. P., Lina, G. & Dumitrescu, O. MUBII-TB-DB: a database of mutations associated with antibiotic resistance in
Mycobacterium tuberculosis. BMC Bioinforma. 15, 107, https://doi.org/10.1186/1471-2105-15-107 (2014).
16. Coll, F. et al. Rapid determination of anti-tuberculosis drug resistance from whole-genome sequences. Genome Med. 7, 51, https://
doi.org/10.1186/s13073-015-0164-0 (2015).
17. Samuelson, D. R., Welsh, D. A. & Shellito, J. E. Regulation of lung immunity and host defense by the intestinal microbiota. Front.
Microbiol. 6, 1085, https://doi.org/10.3389/fmicb.2015.01085 (2015).
18. Schuijt, T. J. et al. The gut microbiota plays a protective role in the host defence against pneumococcal pneumonia. Gut 65, 575–583,
https://doi.org/10.1136/gutjnl-2015-309728 (2016).
Scientific RepoRtS | (2020) 10:8957 | https://doi.org/10.1038/s41598-020-65766-8
9
www.nature.com/scientificreportswww.nature.com/scientificreports/ 19. Sze, M. A. et al. Changes in the bacterial microbiota in gut, blood, and lungs following acute LPS instillation into mice lungs. PLoS
One 9, e111228, https://doi.org/10.1371/journal.pone.0111228 (2014).
20. Gupta, N., Kumar, R. & Agrawal, B. New Players in Immunity to Tuberculosis: The Host Microbiome, Lung Epithelium, and Innate
Immune Cells. Front. Immunol. 9, 709, https://doi.org/10.3389/fimmu.2018.00709 (2018).
21. Pal, C., Bengtsson-Palme, J., Kristiansson, E. & Larsson, D. G. The structure and diversity of human, animal and environmental
resistomes. Microbiome 4, 54, https://doi.org/10.1186/s40168-016-0199-5 (2016).
22. Feng, J. et al. Antibiotic resistome in a large-scale healthy human gut microbiota deciphered by metagenomic and network analyses.
Env. Microbiol. 20, 355–368, https://doi.org/10.1111/1462-2920.14009 (2018).
23. Smillie, C. S. et al. Ecology drives a global network of gene exchange connecting the human microbiome. Nature 480, 241–244,
https://doi.org/10.1038/nature10571 (2011).
24. Fajardo-Cavazos, P. & Nicholson, W. L. Cultivation of Staphylococcus epidermidis in the Human Spaceflight Environment Leads to
Alterations in the Frequency and Spectrum of Spontaneous Rifampicin-Resistance Mutations in the rpoB Gene. Front. Microbiol. 7,
999, https://doi.org/10.3389/fmicb.2016.00999 (2016).
25. Kiwaki, M. & Sato, T. Antimicrobial susceptibility of Bifidobacterium breve strains and genetic analysis of streptomycin resistance
of probiotic B. breve strain Yakult. Int. J. Food Microbiol. 134, 211–215, https://doi.org/10.1016/j.ijfoodmicro.2009.06.011 (2009).
26. Wake, D. B., Wake, M. H. & Specht, C. D. Homoplasy: from detecting pattern to determining process and mechanism of evolution.
Science 331, 1032–1035, https://doi.org/10.1126/science.1188545 (2011).
27. Walker, T. M. et al. Whole-genome sequencing for prediction of Mycobacterium tuberculosis drug susceptibility and resistance: a
retrospective cohort study. Lancet Infect. Dis. 15, 1193–1202, https://doi.org/10.1016/S1473-3099(15)00062-6 (2015).
28. Guerillot, R. et al. Convergent Evolution Driven by Rifampin Exacerbates the Global Burden of Drug-Resistant Staphylococcus
aureus. mSphere 3, https://doi.org/10.1128/mSphere.00550-17 (2018).
29. Poole, K. Bacterial stress responses as determinants of antimicrobial resistance. J. Antimicrob. Chemother. 67, 2069–2089, https://doi.
org/10.1093/jac/dks196 (2012).
30. Baek, K. T. et al. Stepwise decrease in daptomycin susceptibility in clinical Staphylococcus aureus isolates associated with an initial
mutation in rpoB and a compensatory inactivation of the clpX gene. Antimicrob. Agents Chemother. 59, 6983–6991, https://doi.
org/10.1128/AAC.01303-15 (2015).
31. Sreevatsan, S. et al. Restricted structural gene polymorphism in the Mycobacterium tuberculosis complex indicates evolutionarily
recent global dissemination. Proc. Natl Acad. Sci. USA 94, 9869–9874 (1997).
32. Cole, S. T. et al. Deciphering the biology of Mycobacterium tuberculosis from the complete genome sequence. Nature 393, 537–544,
https://doi.org/10.1038/31159 (1998).
33. Banerjee, A. et al. inhA, a gene encoding a target for isoniazid and ethionamide in Mycobacterium tuberculosis. Science 263,
227–230 (1994).
34. Belanger, A. E. et al. The embAB genes of Mycobacterium avium encode an arabinosyl transferase involved in cell wall arabinan
biosynthesis that is the target for the antimycobacterial drug ethambutol. Proc. Natl Acad. Sci. USA 93, 11919–11924, https://doi.
org/10.1073/pnas.93.21.11919 (1996).
35. Johnsson, K., Froland, W. A. & Schultz, P. G. Overexpression, purification, and characterization of the catalase-peroxidase KatG
from Mycobacterium tuberculosis. J. Biol. Chem. 272, 2834–2840 (1997).
36. Zhao, X. et al. Hydrogen peroxide-mediated isoniazid activation catalyzed by Mycobacterium tuberculosis catalase-peroxidase
(KatG) and its S315T mutant. Biochemistry 45, 4131–4140, https://doi.org/10.1021/bi051967o (2006).
37. Zhang, H. et al. Characterization of Mycobacterium tuberculosis nicotinamidase/pyrazinamidase. FEBS J. 275, 753–762, https://doi.
org/10.1111/j.1742-4658.2007.06241.x (2008).
38. Bush, K. & Jacoby, G. A. Updated functional classification of beta-lactamases. Antimicrob. Agents Chemother. 54, 969–976, https://
doi.org/10.1128/AAC.01009-09 (2010).
39. Naas, T. et al. Beta-lactamase database (BLDB) - structure and function. J. Enzyme Inhib. Med. Chem. 32, 917–919, https://doi.org/
10.1080/14756366.2017.1344235 (2017).
40. Srivastava, A., Singhal, N., Goel, M., Virdi, J. S. & Kumar, M. CBMAR: a comprehensive beta-lactamase molecular annotation
resource. Database 2014, bau111, https://doi.org/10.1093/database/bau111 (2014).
41. Saha, S. B., Uttam, V. & Verma, V. u-CARE: user-friendly Comprehensive Antibiotic resistance Repository of Escherichia coli. J.
Clin. Pathol. 68, 648–651, https://doi.org/10.1136/jclinpath-2015-202927 (2015).
42. Ezewudo, M. et al. Integrating standardized whole genome sequence analysis with a global Mycobacterium tuberculosis antibiotic
resistance knowledgebase. Sci. Rep. 8, 15382, https://doi.org/10.1038/s41598-018-33731-1 (2018).
43. Doster, E. et al. MEGARes 2.0: a database for classification of antimicrobial drug, biocide and metal resistance determinants in
metagenomic sequence data. Nucleic Acids Res. 48, D561–D569, https://doi.org/10.1093/nar/gkz1010 (2020).
44. Pal, C., Bengtsson-Palme, J., Rensing, C., Kristiansson, E. & Larsson, D. G. BacMet: antibacterial biocide and metal resistance genes
database. Nucleic Acids Res. 42, D737–743, https://doi.org/10.1093/nar/gkt1252 (2014).
45. Gibson, M. K., Forsberg, K. J. & Dantas, G. Improved annotation of antibiotic resistance determinants reveals microbial resistomes
cluster by ecology. ISME J. 9, 207–216, https://doi.org/10.1038/ismej.2014.106 (2015).
46. Gillespie, J. J. et al. PATRIC: the comprehensive bacterial bioinformatics resource with a focus on human pathogenic species. Infect.
Immun. 79, 4286–4298, https://doi.org/10.1128/IAI.00207-11 (2011).
47. Gupta, S. K. et al. ARG-ANNOT, a new bioinformatic tool to discover antibiotic resistance genes in bacterial genomes. Antimicrob.
Agents Chemother. 58, 212–220, https://doi.org/10.1128/AAC.01310-13 (2014).
48. Liu, B. & Pop, M. ARDB–Antibiotic Resistance Genes Database. Nucleic Acids Res. 37, D443–447, https://doi.org/10.1093/nar/
gkn656 (2009).
49. Alcock, B. P. et al. CARD 2020: antibiotic resistome surveillance with the comprehensive antibiotic resistance database. Nucleic Acids
Res. 48, D517–D525, https://doi.org/10.1093/nar/gkz935 (2020).
50. https://www.ncbi.nlm.nih.gov/pathogens/isolates#/refgene.
51. Zankari, E. et al. PointFinder: a novel web tool for WGS-based detection of antimicrobial resistance associated with chromosomal
point mutations in bacterial pathogens. J. Antimicrob. Chemother. 72, 2764–2768, https://doi.org/10.1093/jac/dkx217 (2017).
52. Lakin, S. M. et al. Hierarchical Hidden Markov models enable accurate and diverse detection of antimicrobial resistance sequences.
Commun. Biol. 2, 294, https://doi.org/10.1038/s42003-019-0545-9 (2019).
53. Feldgarden, M. et al. Validating the AMRFinder Tool and Resistance Gene Database by Using Antimicrobial Resistance Genotype-
Phenotype Correlations in a Collection of Isolates. Antimicrob Agents Chemother 63, https://doi.org/10.1128/AAC.00483-19 (2019).
54. Maruri, F. et al. A systematic review of gyrase mutations associated with fluoroquinolone-resistant Mycobacterium tuberculosis and
a proposed gyrase numbering system. J. Antimicrob. Chemother. 67, 819–831, https://doi.org/10.1093/jac/dkr566 (2012).
55. Clustal Omega < Multiple Sequence Alignment < EMBL-EBI. Ebi.ac.uk (2020). at https://www.ebi.ac.uk/Tools/msa/clustalo/ EBI.
Vol. 2018 (2017).
56. Sievers, F. & Higgins, D. G. Clustal Omega for making accurate alignments of many protein sequences. Protein Sci. 27, 135–145,
https://doi.org/10.1002/pro.3290 (2018).
57. Choi, Y., Sims, G. E., Murphy, S., Miller, J. R. & Chan, A. P. Predicting the functional effect of amino acid substitutions and indels.
PLoS One 7, e46688, https://doi.org/10.1371/journal.pone.0046688 (2012).
Scientific RepoRtS | (2020) 10:8957 | https://doi.org/10.1038/s41598-020-65766-8
1 0
www.nature.com/scientificreportswww.nature.com/scientificreports/ 58. Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J. Basic local alignment search tool. J. Mol. Biol. 215, 403–410, https://
doi.org/10.1016/S0022-2836(05)80360-2 (1990).
59. Gu, Z., Gu, L., Eils, R., Schlesner, M. & Brors, B. circlize Implements and enhances circular visualization in R. Bioinformatics 30,
2811–2812, https://doi.org/10.1093/bioinformatics/btu393 (2014).
Acknowledgements
We thank Prof. Joyoti Basu and Dr. Anupama Ghosh for critically reading the manuscript. AG would like to thank
DBT, Govt. of India for DBT-BINC-JRF Fellowship. SS thank Bose Institute for intramural fund.
Author contributions
A.G. and S.N. collected and compiled the data. A.G., S.N. and S.S. performed the data analysis and wrote the
manuscript. S.S. conceived the idea and supervised the overall study.
competing interests
The authors declare no competing interests.
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/s41598-020-65766-8.
Correspondence and requests for materials should be addressed to S.S.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
ative Commons license, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons license and your intended use is not per-
mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
© The Author(s) 2020
Scientific RepoRtS | (2020) 10:8957 | https://doi.org/10.1038/s41598-020-65766-8
1 1
www.nature.com/scientificreportswww.nature.com/scientificreports/
| null |
10.1007_s00127-023-02428-w.pdf
|
Data Availability The datasets generated during and/or analysed dur-
ing the current study are not publicly available under current ethical
approvals but are available from the corresponding author on reason-
able request.
|
Data Availability The datasets generated during and/or analysed during the current study are not publicly available under current ethical approvals but are available from the corresponding author on reasonable request.
|
Social Psychiatry and Psychiatric Epidemiology (2023) 58:569–579
https://doi.org/10.1007/s00127-023-02428-w
ORIGINAL PAPER
Five‑year illness trajectories across racial groups in the UK
following a first episode psychosis
Siân Lowri Griffiths1
Linda Everard2 · Peter B. Jones3 · David Fowler4 · Joanne Hodgekins5 · Tim Amos6 · Nick Freemantle7 ·
Paul McCrone8 · Swaran P. Singh9 · Max Birchwood9 · Rachel Upthegrove1
· Tumelo Bogatsu1 · Mia Longhi1 · Emily Butler1 · Beel Alexander1 · Mrunal Bandawar1 ·
Received: 2 May 2022 / Accepted: 12 January 2023 / Published online: 30 January 2023
© The Author(s) 2023
Abstract
Purpose Psychosis disproportionally affects ethnic minority groups in high-income countries, yet evidence of disparities
in outcomes following intensive early intervention service (EIS) for First Episode Psychosis (FEP) is less conclusive. We
investigated 5-year clinical and social outcomes of young people with FEP from different racial groups following EIS care.
Method Data were analysed from the UK-wide NIHR SUPEREDEN study. The sample at baseline (n = 978) included
White (n = 750), Black (n = 71), and Asian (n = 157) individuals, assessed during the 3 years of EIS, and up to 2 years post-
discharge (n = 296; Black [n = 23]; Asian [n = 52] and White [n = 221]). Outcome trajectories were modelled for psychosis
symptoms (positive, negative, and general), functioning, and depression, using linear mixed effect models (with random
intercept and slopes), whilst controlling for social deprivation. Discharge service was also explored across racial groups,
2 years following EIS.
Results Variation in linear growth over time was accounted for by racial group status for psychosis symptoms—positive
(95% CI [0.679, 1.235]), negative (95% CI [0.315, 0.783]), and general (95% CI [1.961, 3.428])—as well as for functioning
(95% CI [11.212, 17.677]) and depressive symptoms (95% CI [0.261, 0.648]). Social deprivation contributed to this vari-
ance. Black individuals experienced greater levels of deprivation (p < 0.001, 95% CI [0.187, 0.624]). Finally, there was a
greater likelihood for Asian (OR = 3.04; 95% CI [2.050, 4.498]) and Black individuals (OR = 2.47; 95% CI [1.354, 4.520])
to remain in secondary care by follow-up.
Conclusion Findings suggest variations in long-term clinical and social outcomes following EIS across racial groups; social
deprivation contributed to this variance. Black and Asian individuals appear to make less improvement in long-term recov-
ery and are less likely to be discharged from mental health services. Replication is needed in large, complete data, to fully
understand disparities and blind spots to care.
Keywords Outcomes · Early psychosis · Ethnicity · Deprivation · Inequities
Max Birchwood and Rachel Upthegrove shared joint senior
authorship.
* Siân Lowri Griffiths
[email protected]
1
Institute for Mental Health, University of Birmingham,
Edgbaston, Birmingham B15 2TT, UK
2 Birmingham and Solihull Mental Health Foundation Trust,
Birmingham, UK
3 Department of Psychiatry, University of Cambridge
and CAMEO, Cambridge and Peterborough NHS Foundation
Trust, Fulbourn, UK
4 Department of Psychology, University of Sussex, Brighton,
UK
5 Norwich Medical School, University of East Anglia,
Norwich, UK
6 Academic Unit of Psychiatry, University of Bristol, Bristol,
UK
7
8
Institute of Clinical Trials and Methodology, University
College London, London, UK
Institute for Life Course Development, University
of Greenwich, London, UK
9 Mental Health and Wellbeing Warwick Medical School,
University of Warwick, Coventry, UK
Vol.:(0123456789)1 3
570
Introduction
The incidence of psychosis disproportionally affects eth-
nic minority groups in high-income countries [1]. Black
Caribbean individuals are five times more likely to develop
psychosis in the UK, compared to the White British popu-
lation, but such incidence rates are not mirrored in Carib-
bean countries [2]. Further, for individuals of Pakistani,
Bangladeshi, or of Mixed ethnic backgrounds in England,
the incidence rates are twice as high compared to the
White population [1].
It is well documented that inequalities exist in access
to mental health care, for example, Black individuals are
more likely to experience adverse pathways to care [3].
Differences may also exist in the type of care offered and
received by ethnic minorities within mental health ser-
vices. Black Caribbean and Black African individuals
with psychosis are 15–30% less likely to receive Cogni-
tive–Behavioural Therapy (CBT) compared to White indi-
viduals with psychosis in the UK [4]. National clinical
audit data has also highlighted inequalities in the offer of
clozapine; Black individuals are up to 44% less likely to
be offered this evidenced-based medication for treatment-
resistant psychosis [5].
Despite this, less robust and consistent research has
been carried out on the impact of this potential disparity
on course and outcome of psychosis [6, 7]. A systematic
review has provided evidence that migrant groups are
more likely to achieve remission but have higher rates of
involuntary admission and disengagement compared to
host populations [8, 9]. In studies comparing outcomes
across ethnic groups, poorer social and clinical outcomes
for Black individuals are reported compared to White indi-
viduals [6, 7, 10–15]. For other racial groups, outcomes
have been reported to be more benign. For example, in
an exploratory study by Birchwood et al., relapse and
readmission rates were the highest for Black Caribbean
individuals, and lowest in those of Asian heritage, when
compared to White British individuals [10]. Family struc-
ture, quicker access to care, and employment status have
been proposed to mitigate these effects [10]. However,
inconsistencies and methodological constraints, such as
small sample size, high attrition, short follow-up, and ret-
rospective designs, make it difficult to draw on conclusions
regarding any differences in clinical and social outcomes
for ethnic minority individuals with psychosis, and ques-
tions remain over why such differences exist [7, 12].
In a more recent longitudinal study, the AESOP-10
cohort study investigated ethnic disparities in illness out-
comes between Black minority ethnic and White British
individuals 10 years following a first episode psychosis
(FEP) [6]. Compared to the White British group, the Black
Social Psychiatry and Psychiatric Epidemiology (2023) 58:569–579
Caribbean group had poorer clinical, social, and service
use outcomes. There was also some initial evidence sug-
gesting social disadvantage and isolation contributed to
the differences in symptoms and social outcomes [6].
It is important to understand malleable factors and social
inequalities related to illness incidence, but also whether
underlying factors continue to drive enduring impair-
ment and poorer outcomes. We aim to extend the evidence
on social inequalities and ethnic variation in outcomes after
FEP, using large, national longitudinal dataset of patients
receiving gold standard EIP care.
We wish to establish whether: (1) Black and Asian racial
minority individuals with FEP differ in their long-term
symptoms (psychosis symptoms and depression) and func-
tional outcomes, compared to White individuals; (2) social
deprivation contributes to later clinical and social outcomes
across racial groups; (3) discharge services 2 years after EIS,
differ by racial group.
Method
Study design
This was a secondary analysis of the National Evaluation of
the Development and Impact of Early Intervention Services
(NEDEN) study, a prospective longitudinal study of young
people with a first episode of psychosis (FEP), across 14
early intervention services (EIS) in the UK [16]. The details
of the original study methodology are reported elsewhere
[16], but in brief, participants were initially recruited and
assessed over the first 12 months of service as part of the
NEDEN study. SUPEREDEN is the follow-on study, pro-
spectively assessing the same cohort of individuals up until
discharge from EIS (approximately at 3 years from baseline),
and then up to 2 years post-discharge from EIS. Individuals
with lived experience were involved in the study implemen-
tation and delivery and were regularly consulted throughout
the SUPEREDEN project.
Sample
The initial sample had a total of 978 participants. Participants
were aggregated into 3 racial groups: a Black minority racial
group (n = 71; 6.9%), Asian minority racial group (n = 157;
15%), and a White racial group (n = 750; 73%). The Black
racial minority group included individuals who identified as
Black Caribbean, Black African, and Black ‘other’. The White
group included participants who identified as White British,
White Irish or White ‘other’. The Asian group included par-
ticipants who identified as Pakistani, Bangladeshi, Indian, or
other Asian background. Participants met diagnostic criteria
outlined in International Classification of Diseases under the
1 3 Social Psychiatry and Psychiatric Epidemiology (2023) 58:569–579
571
following codes: F20, F25, F29, F31, F32–F32.1, and F32.3
[17]. Written and verbal consent was obtained for all partici-
pants. Ethical approval was given by Suffolk Local Research
Ethics Committee, UK. REC reference number: 05/Q0102/44.
characteristics between racial groups at baseline and final
follow-up (approximately 5 years from baseline).
Model building
Measures
Outcome variables
Assessments were undertaken by research assistants who
were trained and had no clinical involvement with the partic-
ipants. A robust reliability protocol is detailed in the original
research [16]. The following measures were used to assesses
outcomes: Positive and Negative Syndrome Scale (PANSS)
[18], Calgary Depression Scale for Schizophrenia (CDSS)
[19], Global Assessment of Functioning Disability Scale
(GAF Disability) [20], and Duration of Untreated Psycho-
sis (DUP) [21].
Covariate: social deprivation
A social deprivation proxy was derived at each time point by
summing the presence of the following demographic factors:
(1) unemployed, (2) single marital status, (3) living alone,
and (4) living in temporary/supported accommodation or
social housing, with each of these factors being assigned a
score of 1 if present (maximum score = 4). A score of ‘1’ for
living alone may also be indicative of financial stability or
independence; however, a high score on our proxy measure
(i.e., score of 4) is within the context of being unemployed,
single and in supported or temporary accommodation, and
hence more likely to signify social deprivation.
To validate the summation of these items, reliability sta-
tistics were inspected. Given that reliability coefficients such
as Cronbach’s alpha are sensitive to the number of items in
a scale and often lower with a smaller number of items, we
interpreted this coefficient alongside the optimal mean inter-
item correlations (r = 0.2–0.4), and explored the dimension-
ality of the data using a factor analysis [22, 23]. Correlations
between items were significant (p < 0.01), and the mean
inter-item correlations fell within the recommended range
(r = 0.366), with a Cronbach’s alpha of 0.69 (Supplementary
Material 1) [24]. The exploratory factor analysis confirmed
the uni-dimensionality of the data, with all items loading
strongly on a single component (Supplementary Material 1).
Statistical analysis
Descriptive statistics
Chi-square tests for categorical, and between Analy-
sis of Variance (ANOVA) tests for continuous variables
were performed on the demographic, clinical, and social
To determine the longitudinal relationship between ethnic
status and clinical and social outcomes, hierarchical lin-
ear mixed effect models were constructed within Statisti-
cal Package for the Social Sciences (SPSS v.25). Multi-
level models were constructed in the following manner for
PANSS Positive, PANSS Negative, PANSS General, GAF
Disability, and Calgary Depression. At level 1, fixed and
randomly varying time components were added to the model
to examine the rate of change on the outcome for partici-
pants across the 5-year study period. Graphs were initially
inspected to provide an indication of the shape of the growth
trajectory alongside model fit indices to determine which
rate of growth provided the best model fit. Lower scores
on the Schwartz’s Bayesian Criterion indicated that a lin-
ear time component (coded as 0 for baseline and 1–4 for
subsequent follow-ups) provided better model fit for each
outcome and was therefore used to model the growth tra-
jectories (Supplementary Material 2). At level 2, race was
added to the covariance model to see if any variation in the
(random) time slopes and intercepts for each of the outcomes
were accounted for by racial group (Supplementary Mate-
rial 2). At level 3, a social deprivation proxy was added as
a covariate to determine its influence on the outcome when
all variables were added (and controlled for) in the model.
Models were estimated using a restricted maximum-likeli-
hood (REML) method. REML was selected as it provides
unbiased parameter estimates and is robust against large
missing data and unbalanced designs [25–27]. Simulation
studies have demonstrated that using REML to estimate the
linear mixed models is preferable to multiple imputation for
handling missing data when the mechanisms of missingness
is assumed to be random; data imputation introduces greater
noise into the models, rendering them more unstable [28,
29]. Results of the missing data analyses are reported on
page 9. Finally, a diagonal covariance structure was used
for the repeated and random effects which assumes heter-
ogenous variances and no correlation between any of the
elements [27].
Discharge services
A binary logistic regression was employed to explore the
discharge destinations of the racial minority groups com-
pared to the White racial group, 2 years following discharge
from EIS. The binary outcome was coded as ‘1’ for sec-
ondary care (i.e., specialist mental health service support),
or a ‘0’ for primary care (i.e., non-specialist community
care from a general physician, on a needs basis). Electronic
1 3 572
Social Psychiatry and Psychiatric Epidemiology (2023) 58:569–579
medical record data were accessed for this part of the analy-
sis meaning that more complete (85.2%) data were obtained
(n = 833; Black = 57; Asian = 147; White = 628).
to be missing at random. A restricted maximum-likelihood
method (REML) was considered appropriate to fit the linear
mixed models [25, 28].
Results
Sample description
Missing data
At baseline, outcome data were available for n = 912 partici-
pants (male = 632, 69.3%; mean age = 21.9 years), with an
average retention rate of 33% (n = 296) by the final follow-up
(5 years from baseline). This included data on 34% (n = 23)
of the Black racial group, 37% (n = 52) of the Asian group,
and 32% (n = 221) of the White group, by year 5. The great-
est attrition was observed when participants reconsented
into the SUPEREDEN study (Supplementary Material 3).
To determine any bias in the patterns of missingness on the
outcome variables, we conducted an exploratory analysis
comparing individuals who remained in the study compared
to those who did not. We did not find significant differences
on any of the outcome measures at baseline (Supplementary
Material 4), and there were no differences by racial group
(X2 = 1.165, p = 0.559). We therefore assumed that missing-
ness was not related to the outcomes of interest, and likely
Demographic and clinical characteristics
At baseline and at 2 years post-discharge, there was a higher
frequency of individuals within the Black racial group who
were living alone, single, and living in temporary or sup-
ported accommodation (Table 1). They were also more
likely to be unemployed at baseline, but there were no sig-
nificant differences by follow-up. There were no significant
differences in qualifications levels across racial groups.
The clinical characteristics of the sample are provided
in Table 2. There were no significant differences across the
groups with age of onset; however, the White group had a
significantly longer median DUP, and a significantly higher
percentage of the White racial group had reported self-harm
and used cannabis persistently. There were no significant dif-
ferences between the racial groups on medication adherence
and prescriptions of clozapine or psychological therapies
(Table 2). Over the follow-up period, the Black racial group
had a higher average score on our proxy measure of depri-
vation (b = 0.406, p < 0.001, 95% CI [0.187, 0.624]), whilst
the Asian group had a lower score (b = − 0.322, p < 0.001,
95% CI [− 0.477, − 0.168]) compared to the White group.
Table 1 Demographic
breakdown of racial groups
at baseline and 2 years
post-discharge from early
intervention service
Black N = 71
Asian N = 157
White N = 750
66 (93%)
23 (79.3%)
15 (21%)
12 (42.9%)
52 (73%)
21 (75%)
18 (27%)
32 (48%)
17 (25%)
4 (6%)
40 (62%)
25 (36%)
34 (49%)
11 (15%)
115 (73%)
25 (43.9%)
660 (88%)
196 (71%)
6 (3.8%)
5 (9.3%)
102 (65%)
34 (63%)
42 (28%)
59 (39%)
39 (26%)
12 (8%)
121 (79%)
117 (76%)
24 (16%)
13 (8%)
106 (14.1%)
80 (31.9%)
419 (56%)
149 (59.8%)
168 (23%)
293 (40%)
192 (26%)
79 (11%)
726 (97%)
396 (56%)
250 (35%)
66 (9%)
Statistical
signifi-
cance
p < 0.001
p < 0.001
p < 0.001
p < 0.001
p < 0.001
NS
NS
p < 0.001
p < 0.001
Single marital status
Baseline
2 years post-discharge
Living alone
Baseline
2 years post-discharge
Unemployed
Baseline
2 years post-discharge
Qualifications
None
GSCE/NVQ
A-level/BTEC
Degree
Place of birth: UK
Housing type
Owned/parents own
Rented
Temporary or supported
NS non-significant
1 3 Social Psychiatry and Psychiatric Epidemiology (2023) 58:569–579
Table 2 Clinical characteristics
across racial groups
Black N = 71
Asian N = 157
White N = 750
573
Statistical
signifi-
cance
Presentation factors
Delay of untreated psychosis
(weeks; median)a
Age of onset (years; mean/SD)
Ongoing factors
Cannabis useb (persistent)
Self-harm (n; %)c
Yes/no
Treatment factors
Medication
Non-adherenced
Clozapinee
Psychological therapyf
6.43
8.64
12.71
p < 0.05
21.72 (4.7)
21.05 (4.17)
21.4 (5.17)
NS
3 (4.2%)
1; 53 (1.9%)
8 (5.5%)
5; 117 (4.1%)
107 (14.7%)
85; 510 (14.3%)
p < 0.001
p < 0.002
10 (14.1%)
22 (14%)
89 (11.9%)
2 (2.8%)
13 (18.3%)
8 (5.1%)
28 (17.8%)
14 (1.86%)
125 (16.7%)
NS
NS
NS
NS non-significant
a Independent median test
b Persistent cannabis use = continued cannabis use over 12 months derived from the Drug Check [30]
c Client reported self-harm; any incidence of self-harm over the initial 12 months of treatment
d Medication adherence derived as an average score from the clinician-rated ‘Service Engagement Scale’
[31]
e Prescribed clozapine within the first year of EIS treatment
f Received an individualised form of therapy, e.g., cognitive behavioural therapy across the full study period
Table 3 Linear mixed model fixed effects analysis of recovery out-
comes over the 5-year study period
Beta
SE
p value Lower-95 Upper-95
PANSS positive − 0.502
PANSS negative − 0.335
PANSS general − 1.037
CDSS
GAF disability
0.077 < 0.001 − 0.652
0.071 < 0.001 − 0.474
0.126 < 0.001 − 1.284
− 0.473 − 0.065 < 0.001 − 0.600
1.582
0.243 < 0.001 1.104
− 0.352
− 0.197
− 0.789
− 0.345
2.060
PANSS Positive and Negative Syndrome Scale, CDSS Calgary
Depression Syndrome for Schizophrenia
Racial group differences on recovery outcomes
Linear time effect (level 1)
Over the follow-up period, there were significant main
effects of time for PANSS positive, PANSS negative,
PANSS general, and Calgary Depression, with symp-
toms decreasing on average over the follow-up period.
GAF disability scores on average increased over the study
period, with higher scores indicating improved function-
ing (Table 3).
Illness trajectories and race (level 2)
The random covariance analysis indicated significant vari-
ation in the intercepts and linear slopes across the racial
groups for PANSS positive (b = 0.140; 95% CI [0.679,
1.235]), negative (b = 0.497; 95% CI [0.315, 0.783]), and
general symptoms (b = 2.593; 95% CI [1.961, 3.428]),
as well as GAF disability (b = 14.078, 95% CI [11.212,
17.677]) and depression (b = 0.684; 95% CI [0.261, 0.648]).
The growth trajectories are summarised in Table 4, and visu-
alisation of the trajectories are provided in Figs. 1, 2, 3, 4
and 5 (see also Supplementary Material 5 for means and
standard deviations). Steeper slopes were observed for the
White racial group. The Black group showed no significant
variation in growth for PANSS positive and Calgary Depres-
sion. Lower symptom scores were observed for the Black
group at baseline, whilst the White group had higher scores,
except for negative symptoms, where the Asian group were
observed to have higher scores at baseline.
Social deprivation, race, and outcome (level 3)
A ‘social deprivation proxy’ was added as a covariate in
the linear mixed models for each of the outcome variables
described above. Social deprivation proxy score significantly
contributed to variance in outcomes across the racial groups.
Higher scores on the social deprivation proxy was associated
1 3 574
Social Psychiatry and Psychiatric Epidemiology (2023) 58:569–579
Fig. 1 Graphs depicting illness trajectories across the racial groups
on PANSS positive over the 5-year follow-up
Fig. 4 Graphs depicting illness trajectories across the racial groups
on CDSS depression symptoms over the 5-year follow-up
Fig. 2 Graphs depicting illness trajectories across the racial groups
on PANSS negative over the 5-year follow-up
Fig. 5 Graphs depicting illness trajectories across the racial groups
on GAF disability over the 5-year follow-up
with higher PANNS positive scores (b = 0.710, SE = 0.103,
p < 0.001, 95% CI [0.509, 0.912]), PANSS negative scores,
b = 0.875, SE = 0.106, p < 0.001, 95% CI [0.667, 1.083]),
PANNS general score (b = 1.390; SE = 0.174, p < 0.001, 95%
CI [1.050, 1.731]), and Calgary depression scores (b = 0.455,
SE = 0.099, p < 0.001, 95% CI [0.261, 0.648]). Finally, a
higher social deprivation score on our proxy measure was
associated with lower GAF scores (b = − 5.116, SE = 0.328,
p < 0.001, 95% CI [− 5.758, − 4.473]).
Discharge trajectories
A binary logistic regression comparing discharge services
across racial groups 2 years following EIS, showed that,
compared to their White counterparts, there was a greater
likelihood for the Asian (OR = 3.04; 95% CI [2.050, 4.498];
p = < 0.001) and Black racial group (OR = 2.47; 95% CI
Fig. 3 Graphs depicting illness trajectories across the racial groups
on PANSS general symptoms over the 5-year follow-up
1 3 Social Psychiatry and Psychiatric Epidemiology (2023) 58:569–579
575
Table 4 Linear growth parameters across racial groups over time for
each of the outcome variable
Beta
Wald Z p value Lower-95 Upper-95
PANSS positive
Intercepta
Group × timeb
Black
Asian
White
PANSS negative
Intercepta
Group × timeb
Black
Asian
White
PANSS general
Group × timeb
Black
Asian
White
CDSS
Intercepta
Group × timeb
Black
Asian
White
GAF disability
Intercepta
Group × timeb
Black
Asian
White
4.824
4.894
< 0.001
3.232
7.200
0.706
1.372
1.121
NS
1.935
3.4943 < 0.001
< 0.001
6.656
0.257
0.783
0.835
1.944
2.406
1.505
7.275
6.568
< 0.001
5.398
9.805
1.541
0.929
0.609
11.554a
2.988
2.695
2.944
2.500
2.838
4.487
4.381
2.418
3.175
6.784
0.0124
0.0045
< 0.001
< 0.001
0.704
0.466
0.394
7.387
0.016
0.002
< 0.001
1.329
1.454
2.205
3.374
1.85
0.943
18.073
6.721
4.997
3.930
6.1502a
8.770
< 0.001
4.9185
7.690
0.074
0.434
0.812
0.424
2.214
6.406
NS
0.033
< 0.001
0.001
0.1791
0.598
7.561
1.053
1.103
81.516
9.351
< 0.001
66.103
100.524
8.537
12.125
14.933
0.0471
1.986
3.530
< 0.001
7.8943 < 0.001
3.182
6.959
11.650
22.909
21.127
19.142
a Random covariance parameter for the intercepts across racial groups
b Random covariance slope parameter for time × racial group
NS non-significant at 0.05 alpha level
[1.354, 4.520]; p = < 0.001) to remain in secondary care
(i.e., treatment within mental health services) by follow-up.
Discussion
In this large, prospective FEP cohort, recovery outcomes
significantly improved across the follow-up period,
which included the duration of EIS care and up to 2 years
post-discharge.
The rate of improvement varied by racial group, with the
White group showing more growth in their recovery trajec-
tories. Social deprivation further contributed to this variance
in growth across racial groups. Two years following EIS
care, the Asian and Black individuals were less likely to be
discharged from mental health services.
To our knowledge, this is the first study to report long-
term outcomes across different racial minority groups fol-
lowing EIS care [6, 11–15]. Our findings hint at the potential
compounded impact of the intersectional challenges of racial
minority status and deprivation [6, 12, 32–34]. However, our
findings are nuanced; deprivation was not uniform across
minority racial groups. The Black group had significantly
greater levels of deprivation, whilst the Asian group experi-
enced less social deprivation over the study period.
Despite improving more, the White group typically had
similar levels of symptoms to the minority racial groups by
follow-up, possibly suggesting a ceiling effect in recovery
trajectories for the minority racial groups. This was likely
the case for the Black group who showed no change in
growth over time on the Calgary Depression Scale, but this
was in the context of low, stable symptoms across the time
frame. Similarly, self-harm was less frequent in the Black
group; a finding supported by previous research [6]6.
Nevertheless, we showed that minority individuals were
more likely to be receiving mental health treatment follow-
ing discharge from EIS, suggesting that they may not have
achieved the same level of recovery as their White counter-
parts. This possible enduring nature of psychosis for minor-
ity groups would support the work of Morgan et al., where
Black individuals were more likely to have a continuous,
non-remitting illness course, as opposed to an episodic tra-
jectory [6].
Confounding treatment factors
It is well documented that a prolonged delay of untreated
psychosis (DUP) is associated with poorer recovery out-
comes [35, 36]. We did not find a longer DUP for racial
minority groups. Instead, the White group had a significantly
longer DUP; a finding supported by other studies [37–40].
This may account for the higher symptom scores for the
White group at baseline, yet the White group typically
showed more growth in their trajectories over time. This
raises the question as to why the trajectories of the racial
minority groups may appear less responsive to the support
offered within current service models.
Linked to this notion, our initial inspection showed no
differences in treatment factors that are likely to influence
recovery outcomes, such as medication adherence, treat-
ment with clozapine, and receiving a psychological therapy
[41–43]. Further, we found no differences in age of onset of
illness, but there were significant differences in persistent
cannabis use, which was more frequent in the White group.
However, as previously reported by the EDEN consortium,
1 3 576
Social Psychiatry and Psychiatric Epidemiology (2023) 58:569–579
the influence of cannabis on poor outcomes was shown to
be independent of ethnic status [44].
Proposed mechanisms driving inequalities
in outcomes
Socio-economic status, experiences of racism, linguistic dis-
tance, and social exclusion and discrimination may lead to a
psychological ‘disempowerment’ [45] or ‘social defeat’ [46].
Such processes are likely to play an important role in the
aetiology and pathogenesis of psychosis [34, 46, 47]. Indeed,
in our study, for the Black racial group, deprivation was
already apparent at baseline, likely reflecting a longstand-
ing trajectory of deprivation. This not only exposes these
individuals to psychotic illness, but is likely to be mutually
reinforcing, where psychosis symptoms drive further depri-
vation and exclusion, and vice versa, resulting in enduring
impairment, marginalisation, and further feelings of disem-
powerment [6, 33, 48].
On the other hand, the Asian minority group expe-
rienced less deprivation compared to the other groups,
which suggests that other factors are also likely to play a
part. Indeed, previous studies have reported racial-ethnic
differences in receiving evidence-based interventions and
family psychoeducation once in treatment following a first
episode of schizophrenia [49]. Compulsory treatment is
also frequently reported [9]. Themes of mistrust in services,
stigma, and coerciveness have also featured in the narra-
tives of Black and minority individuals receiving mental
health treatment [50]. Thus, treatment trajectories, including
pathways out of EIS, warrant in-depth exploration, particu-
larly as our racial minority groups were less likely to be
discharged out of mental health service 2 years following
EIS discharge. The lived experiences of these individuals
will be essential to fully understand the processes behind
these disparities.
Strengths and limitations
There are several strengths to this study. The EDEN studies
comprised a large, perspective cohort of participants who
had experienced FEP across distinct and varied geographi-
cal areas in England, making it representative of the UK’s
diverse population, but also representing socioeconomic
variability. We add to past literature by further including a
comparison with individuals of Asian heritage, which has
not been robustly reported within the literature. Finally, we
explore a range of outcome variables and model the hetero-
geneity in illness trajectories across the duration of EIS care
and the subsequent 2 years following discharge. However,
there are important study limitations to consider.
First, whilst over a thousand participants originally con-
sented to the study, our target minority racial groups were
substantially smaller, reducing our statistical power. Given
the high prevalence of psychosis within ethnic minority
groups in high-income countries, our small group size in this
study may reflect lack of engagement of minority individu-
als in research, thus placing limit on the representativeness
of our findings and potentially biasing the sample. Second,
there were high levels of attrition across each time point,
potentially introducing bias in our findings. We were, how-
ever, able to demonstrate that missingness did not differ by
racial group, and there were no differences by racial group
on the main outcomes at baseline for those who continued
in the study compared to those who dropped out. In such
situations where mechanisms of missingness are assumed
to be random, the REML algorithm (used within the analy-
sis) is shown to be robust to large missing data and unbal-
anced designs [25, 28, 29]. Third, for reasons of statistical
power, we were not able to explore intergroup differences.
For example, there is evidence pointing to differential out-
comes in Black Caribbean, as opposed to Black African
individuals [6, 51]. We also did not include a mixed racial
group in our analysis because of the limited sample size; this
should be investigated further. Finally, as this was a second-
ary analysis of existing data, this restricted our examina-
tions into other potential factors influencing the observed
differences. This also meant that a proxy estimate was used
to quantify social deprivation. Future research may wish to
build on these findings using a more robust measure of social
deprivation, which also considers the premorbid levels of
deprivation, compared with the deprivation synergistically
linked to psychosis.
Implications and future directions
Methodological issues place limit on how much we can
extrapolate our findings, but they nevertheless add to a
growing body of research indicating differential outcomes
for racial minorities recovering from a first episode psycho-
sis. In addition to replication, further research is also needed
to understand the key drivers of these disparities that may
serve as pivotal points for intervention. Our findings may
suggest wider contextual and societal factors feeding into
illness trajectories. Systemic barriers and social structures
inherent within our society are likely to permeate into health
care and place limit on one’s outcome. Breaking this cycle
should not only be a priority for EIS, but a shared priority
for public health and social policy [6].
There is growing interest looking into area-level inter-
ventions to mitigate the psychological consequences of
belonging to a disempowered minority group. For example,
increasing access to social capital is proposed to dampen the
social stress associated with deprivation and discrimination,
and thus foster an environment that is more conducive to
1 3 Social Psychiatry and Psychiatric Epidemiology (2023) 58:569–579
577
recovery [11, 52, 53]. Though promising, implementing such
interventions is complex given their nuanced and context-
dependent nature [54]. At a service level, there may be a
need to develop clinicians’ cultural competencies, in addi-
tion to offering culturally sensitive interventions to improve
service provision for underserved groups. Co-produced work
will be an important step towards achieving this goal [55].
Finally, exploring the disempowerment experienced by such
individuals may also be an important target for clinical inter-
vention [47].
adaptation, distribution and reproduction in any medium or format,
as long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons licence, and indicate
if changes were made. The images or other third party material in this
article are included in the article's Creative Commons licence, unless
indicated otherwise in a credit line to the material. If material is not
included in the article's Creative Commons licence and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this licence, visit http:// creat iveco mmons.
org/ licen ses/ by/4. 0/.
Conclusion
In a large FEP cohort, our findings suggest variations in
long-term clinical and social outcomes following EIS for
racial minority groups. Social deprivation contributed to this
variance, with Black individuals experiencing the most dep-
rivation. Black and Asian individuals were also less likely
to be discharged from mental health service by follow-up.
Though replication is needed, our findings hint at the need
for targeted, and culturally sensitive service provision, that
mitigates the impact of discrimination and deprivation and
promotes long-term recovery following FEP.
Supplementary Information The online version contains supplemen-
tary material available at https:// doi. org/ 10. 1007/ s00127- 023- 02428-w.
Acknowledgements M.B. and S.P.S are part funded by the National
Institute for Health Research through the Applied Research Collabo-
ration West Midlands (ARC-WM). P.B.J. is part funded by the NIHR
ARC East of England. The views expressed in this publication are
those of the authors and not necessarily those of the NHS, NIHR,
or Department of Health. Birmingham and Solihull NHS Foundation
Trust acted as study sponsor. We would like to thank the participants
of the National EDEN study and the UK Clinical Research Network
for study support.
Data Availability The datasets generated during and/or analysed dur-
ing the current study are not publicly available under current ethical
approvals but are available from the corresponding author on reason-
able request.
Declarations
Conflict of interest RU reports grants from Medical Research Council,
grants from National Institute for Health Research: Health Technol-
ogy Assessment, grants from European Commission—Research: The
Seventh Framework Programme, and personal fees from Sunovion,
outside the submitted work.
Ethical standards Ethical approval was given by Suffolk
Local Research Ethics Committee, UK, in accordance with the ethi-
cal standards laid down in the 1964 Declaration of Helsinki and its
later amendments. Research Ethics Committee reference number: 05/
Q0102/44.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
References
1. Kirkbride JB et al (2012) Incidence of schizophrenia and other
psychoses in England, 1950–2009: a systematic review and meta-
analyses. PLoS One 7:e31660. https:// doi. org/ 10. 1371/ journ al.
pone. 00316 60
2. Bhugra D et al (1996) First-contact incidence rates of schizo-
phrenia in Trinidad and one-year follow-up. Br J Psychiatry
169:587–592. https:// doi. org/ 10. 1192/ bjp. 169.5. 587
3. Halvorsrud K, Nazroo J, Otis M, Brown Hajdukova E, Bhui K
(2018) Ethnic inequalities and pathways to care in psychosis
in England: a systematic review and meta-analysis. BMC Med
16:223. https:// doi. org/ 10. 1186/ s12916- 018- 1201-9
4. Morris RM et al (2020) Ethnicity and impact on the receipt of
cognitive-behavioural therapy in people with psychosis or bipo-
lar disorder: an English cohort study. BMJ Open 10:e034913.
https:// doi. org/ 10. 1136/ bmjop en- 2019- 034913
5. RCPsych, R. C. o. P. (2021) National clinical audit of psy-
chosis—national report for the early intervention in psychosis
audit 2020/2021. Healthcare Quality Improvement Partnership,
London
6. Morgan C et al (2017) Ethnicity and long-term course and out-
come of psychotic disorders in a UK sample: the ÆSOP-10
study. Br J Psychiatry 211:88–94. https:// doi. org/ 10. 1192/ bjp.
bp. 116. 193342
7. Chorlton E, McKenzie K, Morgan C, Doody G (2012) Course
and outcome of psychosis in black Caribbean populations and
other ethnic groups living in the UK: a systematic review. Int
J Soc Psychiatry 58:400–408. https:// doi. org/ 10. 1177/ 00207
64011 403070
8. Maguire J, Sizer H, Mifsud N, O’Donoghue B (2020) Outcomes
for migrants with a first episode of psychosis: a systematic
review. Schizophr Res 222:42–48. https:// doi. org/ 10. 1016/j.
schres. 2020. 05. 048
9. Maguire J et al (2021) Symptomatic, functional and service uti-
lization outcomes of migrants with a first episode of psychosis.
Soc Psychiatry Psychiatr Epidemiol 56:1389–1397. https:// doi.
org/ 10. 1007/ s00127- 020- 02011-7
10. Birchwood M et al (1992) The influence of ethnicity and family
structure on relapse in first-episode schizophrenia: a comparison
of Asian, Afro-Caribbean, and White patients. Br J Psychiatry
161:783–790. https:// doi. org/ 10. 1192/ bjp. 161.6. 783
11. Hodgekins J et al (2015) Investigating trajectories of social
recovery in individuals with first-episode psychosis: a latent
class growth analysis. Br J Psychiatry 207:536–543. https:// doi.
org/ 10. 1192/ bjp. bp. 114. 153486
12. Hall M-H, Holton KM, Öngür D, Montrose D, Keshavan MS
(2019) Longitudinal trajectory of early functional recovery in
patients with first episode psychosis. Schizophr Res 209:234–
244. https:// doi. org/ 10. 1016/j. schres. 2019. 02. 003
1 3 578
Social Psychiatry and Psychiatric Epidemiology (2023) 58:569–579
13. Morgan C et al (2008) Cumulative social disadvantage, ethnicity
and first-episode psychosis: a case-control study. Psychol Med
38:1701–1715
14. Li H, Eack SM, Montrose DM, Miewald JM, Keshavan M
(2011) Longitudinal treatment outcome of African American
and Caucasian patients with first episode psychosis. Asian J
Psychiatr 4:266–271. https:// doi. org/ 10. 1016/j. ajp. 2011. 08. 004
15. Kam SM, Singh SP, Upthegrove R (2015) What needs to follow
early intervention? Predictors of relapse and functional recov-
ery following first-episode psychosis. Early Interv Psychiatry
9:279–283
16. Birchwood M et al (2014) The UK national evaluation of the
development and impact of Early Intervention Services (the
National EDEN studies): study rationale, design and baseline
characteristics. Early Interv Psychiatry 8:59–67. https:// doi. org/
10. 1111/ eip. 12007
17. WHO (2004) ICD-10: international statistical classification of
diseases and related health problems: tenth revision. ICD-10
18. Kay SR, Fiszbein A, Opler LA (1987) The positive and nega-
tive syndrome scale (PANSS) for schizophrenia. Schizophr Bull
13:261–276. https:// doi. org/ 10. 1093/ schbul/ 13.2. 261
19. Addington D, Addington J, Schissel B (1990) A depression rating
scale for schizophrenics. Schizophr Res 3:247–251. https:// doi. org/
10. 1016/ 0920- 9964(90) 90005-R
20. Roche E et al (2016) The prognostic value of formal thought disorder
following first episode psychosis. Schizophr Res 178:29–34. https://
doi. org/ 10. 1016/j. schres. 2016. 09. 017
21. Larsen TK, McGlashan TH, Moe LC (1996) First-episode Schizo-
phrenia: I. Early course parameters. Schizophr Bull 22:241–256.
https:// doi. org/ 10. 1093/ schbul/ 22.2. 241
22. Briggs SR, Cheek JM (1986) The role of factor analysis in the devel-
opment and evaluation of personality scales. J Pers 54:106–148.
https:// doi. org/ 10. 1111/j. 1467- 6494. 1986. tb003 91.x
23. Piedmont RL (2014) In: Michalos AC (ed) Encyclopedia of quality
of life and well-being research. Springer Netherlands, Heidelberg,
pp 3303–3304
24. Pallant J (2001) SPSS survival manual
25. Gilmour AR, Thompson R, Cullis BR (1995) Average information
REML: an efficient algorithm for variance parameter estimation in
linear mixed models. Biometrics 51:1440–1450. https:// doi. org/ 10.
2307/ 25332 74
26. Snijders TA, Bosker RJ (2011) Multilevel analysis: An introduction
to basic and advanced multilevel modeling, Sage
27. Heck RH, Thomas SL, Tabata LN (2013) Multilevel and longitudi-
nal modeling with IBM SPSS. Routledge, New York
28. Twisk J, de Boer M, de Vente W, Heymans M (2013) Multiple impu-
tation of missing values was not necessary before performing a lon-
gitudinal mixed-model analysis. J Clin Epidemiol 66:1022–1028.
https:// doi. org/ 10. 1016/j. jclin epi. 2013. 03. 017
29. Peters SAE et al (2012) Multiple imputation of missing repeated
outcome measurements did not add to linear mixed-effects models. J
Clin Epidemiol 65:686–695. https:// doi. org/ 10. 1016/j. jclin epi. 2011.
11. 012
30. Kavanagh DJ et al (2011) The drugcheck problem list: a new screen
for substance use disorders in people with psychosis. Addict Behav
36:927–932
31. Tait L, Birchwood M, Trower P (2002) A new scale (SES) to meas-
ure engagement with community mental health services. J Ment
Health 11:191–198. https:// doi. org/ 10. 1080/ 09638 23002 00235 70-2
32. Suvisaari J et al (2018) Is it possible to predict the future in first-
episode psychosis? Front Psychiatry. https:// doi. org/ 10. 3389/ fpsyt.
2018. 00580
33. Puntis S, Oke J, Lennox B (2018) Discharge pathways and relapse
following treatment from early intervention in psychosis services.
BJPsych Open 4:368–374. https:// doi. org/ 10. 1192/ bjo. 2018. 50
34. Jongsma HE et al (2020) Social disadvantage, linguistic distance,
ethnic minority status and first-episode psychosis: results from the
EU-GEI case–control study. Psychol Med. https:// doi. org/ 10. 1017/
S0033 29172 00002 9X
35. Marshall M et al (2014) Impact of early intervention services on
duration of untreated psychosis: data from the National EDEN pro-
spective cohort study. Schizophr Res 159:1–6. https:// doi. org/ 10.
1016/j. schres. 2014. 07. 005
36. Drake RJ et al (2020) Effect of delaying treatment of first-episode
psychosis on symptoms and social outcomes: a longitudinal analysis
and modelling study. Lancet Psychiatry 7:602–610. https:// doi. org/
10. 1016/ S2215- 0366(20) 30147-4
37. Singh SP et al (2015) Ethnicity and pathways to care during first
episode psychosis: the role of cultural illness attributions. BMC
Psychiatry 15:287
38. Morgan C et al (2006) Duration of untreated psychosis and ethnicity
in the AESOP first-onset psychosis study. Psychol Med 36:239–247.
https:// doi. org/ 10. 1017/ s0033 29170 50066 04
39. Anderson KK, Flora N, Archie S, Morgan C, McKenzie K (2014)
Race, ethnicity, and the duration of untreated psychosis: a systematic
review. Soc Psychiatry Psychiatr Epidemiol 49:1161–1174. https://
doi. org/ 10. 1007/ s00127- 013- 0786-8
40. Bhui K, Ullrich S, Coid JW (2014) Which pathways to psychi-
atric care lead to earlier treatment and a shorter duration of first-
episode psychosis? BMC Psychiatry 14:72. https:// doi. org/ 10. 1186/
1471- 244x- 14- 72
41. Stokes I et al (2020) Prevalence of treatment resistance and clozap-
ine use in early intervention services. BJPsych Open 6:e107. https://
doi. org/ 10. 1192/ bjo. 2020. 89
42. Wykes T, Steel C, Everitt B, Tarrier N (2007) Cognitive behavior
therapy for schizophrenia: effect sizes, clinical models, and meth-
odological rigor. Schizophr Bull 34:523–537. https:// doi. org/ 10.
1093/ schbul/ sbm114
43. Demjaha A et al (2017) Antipsychotic treatment resistance in first-
episode psychosis: prevalence, subtypes and predictors. Psychol
Med 47:1981–1989. https:// doi. org/ 10. 1017/ s0033 29171 70004 35
44. Seddon JL et al (2016) Cannabis use is associated with increased
psychotic symptoms and poorer psychosocial functioning in first-
episode psychosis: a report from the UK national EDEN study.
Schizophr Bull 42:619–625. https:// doi. org/ 10. 1093/ schbul/ sbv154
45. Marmot MG (2006) Status syndrome: a challenge to medicine.
JAMA 295:1304–1307. https:// doi. org/ 10. 1001/ jama. 295. 11. 1304
46. Selten JP, Cantor-Graae E (2005) Social defeat: risk factor for schiz-
ophrenia? Br J Psychiatry 187:101–102. https:// doi. org/ 10. 1192/ bjp.
187.2. 101
47. Baker SJ, Jackson M, Jongsma H, Saville CWN (2021) The ethnic
density effect in psychosis: a systematic review and multilevel meta-
analysis. Br J Psychiatry J Ment Sci 219:632–643. https:// doi. org/
10. 1192/ bjp. 2021. 96
48. Griffiths SL, Wood SJ, Birchwood M (2018) Vulnerability to psy-
chosocial disability in psychosis. Epidemiol Psychiatr Sci. https://
doi. org/ 10. 1017/ s2045 79601 80004 95
49. Oluwoye O et al (2018) Racial-ethnic disparities in first-episode
psychosis treatment outcomes from the RAISE-ETP study. Psychiatr
Serv 69:1138–1145. https:// doi. org/ 10. 1176/ appi. ps. 20180 0067
50. Devonport TJ et al (2022) A systematic review of inequalities in the
mental health experiences of Black African, Black Caribbean and
black-mixed UK populations: implications for action. J Racial Ethn
Health Disparities. https:// doi. org/ 10. 1007/ s40615- 022- 01352-0
51. Ajnakina O et al (2017) Patterns of illness and care over the 5 years
following onset of psychosis in different ethnic groups; the GAP-5
study. Soc Psychiatry Psychiatr Epidemiol 52:1101–1111
52. Handley CE, Oakley D, Saville CWN (2020) Residential churn
moderates the relationship between economic deprivation and psy-
chiatric admission: evidence from Wales. J Epidemiol Community
Health 74:560–564. https:// doi. org/ 10. 1136/ jech- 2019- 213351
1 3 Social Psychiatry and Psychiatric Epidemiology (2023) 58:569–579
579
53. Das-Munshi J et al (2012) Ethnic density as a buffer for psychotic
experiences: findings from a national survey (EMPIRIC). Br J Psy-
chiatry 201:282–290
54. Kirkbride JBBJ, Ploubidis GB, Morgan C, Dazzan P, McKenzie
K, Murray RM, Jones PB (2008) Testing the association between
the incidence of schizophrenia and social capital in an urban area.
Psychol Med 38:1083–1094. https:// doi. org/ 10. 1017/ S0033 29170
70020 85
55. Naz S, Gregory R, Bahu M (2019) Addressing issues of race, ethnic-
ity and culture in CBT to support therapists and service managers
to deliver culturally competent therapy and reduce inequalities in
mental health provision for BAME service users. Cogn Behav Ther
12:e22. https:// doi. org/ 10. 1017/ S1754 470X1 90000 60
1 3
| null |
10.1088_1402-4896_ad0c11.pdf
|
Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).
|
Data availability statement All data that support the findings of this study are included within the article (and any supplementary files). ORCID iDs Yaseen Muhammad https://orcid.org/0000-0002-7042-1527
|
Phys. Scr. 98 (2023) 125967
https://doi.org/10.1088/1402-4896/ad0c11
PAPER
RECEIVED
3 May 2023
REVISED
18 September 2023
ACCEPTED FOR PUBLICATION
13 November 2023
PUBLISHED
27 November 2023
Comparative investigation of low and high pelletize pressure for
(Ag)x/CuTl-1223 nanoparticles-superconductor composites
Yaseen Muhammad1,2,∗
1 Materials Research Laboratory, Department of Physics, Faculty of Sciences (FOS), International Islamic University (IIU), H-10 Islamabad
, M Rahim1, M Mumtaz1, Nazir Hussain2 and Bahar Hussain1
44000, Pakistan
2 Department of Physics, Faculty of Engineering and Applied Sciences (FEAS), Riphah International University, I-14 Islamabad, Pakistan
∗ Author to whom any correspondence should be addressed.
E-mail: [email protected]
Keywords: Ag nano-particles, CuTl-1223 superconducting phase, nano-(Ag)x/CuTl-1223 composites, low and high pelletize pressure,
Superconducting properties, Ag nano-particles (NPs), direct current resistivity (dc-resistivity)
Abstract
This article experimentally investigates the impact of silver (Ag) nano-particles inclusion and high
pelletized pressure on the structural, morphological, and electrical properties of
Cu0.5Tl0.5Ba2Ca2Cu3O10−δ (noted CuTl-1223) bulk system. The nano-(Ag)x/CuTl-1223 composites
were synthesized using a two-step solid-state reaction process with added amount of Ag nano-particles
ranging from 0 to 2.0 wt% of the total mass. These nano-composites were produced at both low and
high pelletize pressure of 0.202 GPa. All prepared samples were characterized using valuable
techniques such as X-ray Diffraction (XRD), Scanning Electron Microscope (SEM), Energy Dispersive
Spectroscopy (EDS), Fourier Transform Infrared Spectroscopy (FTIR), and dc-resistivity measure-
ment at low as well as high pelletized pressure, respectively. The structural investigation via the XRD
technique indicated that Ag NPs did not affect the CuTl-1223 tetragonal structure, confirmed that Ag
nano-particles were settled across the grain boundaries. SEM examination revealed a fine distribution
of nano-sized silver (Ag) NPs among the CuTl-1223 grains, as well as improved weak-links and density
of void/pores. The position of distinct vibrational oxygen modes in FTIR spectra showed no
substantial alteration, indicating that the structural nature of the host CuTl-1223 phase was preserved.
The electrical properties were studied using the four-point probe technique, and the activation energy
r - measurements showed that pelletization
were determined using Arrhenius law. The results of
) of (Ag)x/CuTl-1223 composites.
pressure of 0.202 GPa have an impact on the critical temperature Tc(
The superconducting critical temperature T 0c ( ) was enhanced from 99 K to 107 K at high-pressure of
pelletization (0.202 GPa) as compared to low pressure, with x = 0 ∼ 2.0 wt% nano-particles addition
to the host CuTl-1223 phase.
T
)
(
1. Introduction
(
g = x
ab
x
The most desirable phase of the high temperature superconductors (HTSCs) family is
Cu0.5Tl0.5Ba2Ca2Cu3O10−δ (CuTl-1223) due to its higher critical temperature T ,c(
)
and longer coherence length (
,
superconducting anisotropy
superconducting phase can be successfully fabricated under both ambient and high pressure conditions [3, 4].
The enhancement of critical temperature Tc(
researchers are busy to explore the materials as well as the best synthesis process to increase the critical
temperature of desire superconducting compounds. Several research groups investigated the pure CuTl-1223
phase using various preparative techniques, and they reported significant critical temperature Tc(
) [1, 3]. It is
vital for high temperature superconductors (HTSCs) that the grains of the compounds are perfectly compacted
and oriented in the appropriate direction, leading to enhanced superconductivity [5]. The bulk CuTl-1223
) low value of
cx ) along c-axis [1, 2]. The CuTl-1223
) is the main challenge for these superconducting phase. The
c
© 2023 IOP Publishing Ltd
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
compounds are composed of grains that are weakly connected at their respective boundaries by weak links [6].
The strength and performance of CuTl-based superconductors are effected by dislocations, imperfection,
porosity and micro fractures [6]. The superconducting critical parameters are reduced by the existence of inter-
grain voids/pores, weak connections as well as motion of the vortices in these superconducting phase produced
resistances resulting in energy dissipation [7]. To improve the inter-grain voids and weak-link effect the
inclusion of nano-particles to the host CuTl-1223 matrix is the best techniques to produce nanoparticles
superconductor composites at ambient pressure [1–3, 5, 6]. The transport properties are improved after
introducing of different nano-particles to the CuTl-1223 phase are ascribed to an enhance inter-grains
connectivity by healing up the voids/pores [6]. These nano-particles are not member of host matrix and resides
at inter-crystallite sites. The impact of introducing various nano-particles inclusion on the structure and
superconducting properties of CuTl-1223 superconductor has been studied by numerous research teams
[5–16]. The literature evaluations have been indicated that the addition of nano-particles refine inter-grain
connectivity and somewhat enhance the superconducting parameters without changing the CuTl-1223 crystal
structure. The occurrence of nano-particles at the inter-grain sites of granular bulk CuTl-1223 superconductor
can enhance their superconducting properties at ambient/low pressure, but the size and uniform distributions
of these nano-particles is the still biggest challenge. For significantly improvement in weak-link effects and inter-
grain voids of nano-particles superconductor-composites, high-pressure synthesis is the efficient method for
enhancement of critical parameters in HTSCs [17–28]. The mechanism underling the increase in
superconducting parameters at high pelletize pressure corresponds with incorporation of nano-structures to fill
and squeeze the inter-grain voids/pores, leading to inter-grain connectivity as well as conduct the grain-
boundaries [21]. These improved inter-grain weak-links makes the change carriers more accessible for transfer
processes and reduced energy loss across the grain-sites [21]. In comparison to ambient/low pressure synthesis,
high pelletize pressure provided that the nano-doped-CuTl-1223 superconducting composites are mostly lager
CuTl-1223 grain size containing discrete nano-particles at the grain-sites. Thus, the uniform distribution of
nano-particles to the CuTl-1223 superconductor as well as use of high-pelletize pressure may lead to the creation
of superconducting regions at the inter-grain sites, this could be advancing the superconducting properties of
HTSCs [29].
In the present research work, the pressure-induced superconducting characteristics of (Ag)x/CuTl-1223
nano-particles superconducting-composites were investigated, using XRD, SEM, FTIR, and electrical resistivity
measurements. The major issue with the CuTl-1223 superconducting phase synthesized at ambient/low
pressure is the appearances of gapes and holes in higher density, which has an impact on the overall
superconducting critical parameters. When vital factors such as the shape, orientation, and dimension of the
inserted Ag nanoparticles are properly adopted, the transport properties of CuTl-1223 superconductor can be
enhance enormously. Consequently, this research focuses on the ability of Ag nano-particles to operate as pining
center/inter-grain connectivity enhancers in the CuTl-1223 superconductor at low and high pressure synthesis.
We studied their effects on superconducting parameters including: transition temperature T ,c(
onset
) holes concentration p( ) and activation energy U ,o(
temperature T
,
c
application scenarios of these superconducting materials. The (Ag)x/CuTl-1223; x = 0.0, 0.5, 1.0, 1.5, and
2.0 wt% nano-composites were manufactured via solid-state reaction route and characterized at low (100 MPa)
and high-pressure of 0.202 GPa. The structural, morphological, compositional, and transport characteristics of
(Ag)x/CuTl-1223 nano-composites under low & high pelletize pressure were analyzed, described, and
compared.
) onset critical
) which are functional parameters in
(
2. Experimental detail
2.1. Samples synthesis
2.1.1. Preparation of Cu0.5Ba2Ca2Cu3O10-δ precursor
The solid-state reaction technique was used to synthesize Cu0.5Ba2Ca2Cu3O10-δ precursor. Firstly, powder of
99.99 percent pure Cu2(CN)2.H2O, Ba(NO3)2, and Ca(NO3)2.4H2O were weighed by electronic balance in
accordance with the desired stoichiometric ratio. These three chemical compounds were combined and ground
in an agate mortar and pestle for 3 h. The pulverized powders were put onto quartz boat, calcined in a chamber
furnace for 24 h at 860 degree Celsius, and then cooled to room temperature. Following one hour of grinding,
the material was subjected to a second heat-treatment under similar condition. The final precursor
Cu0.5Ba2Ca2Cu3O10-δ was obtained. The step-by-step approach is illustrated in flow chart of figure 1.
2
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Figure 1. Schematic diagram of experimental work for the synthesis of (Ag)x/CuTl-1223 nano-particles superconductor composites
at low and high-pressure of 0.202 GPa.
2.1.2. Preparation of (Ag)x/CuTl-1223 (x = 0.0, 0.5, 1.0, 1.5, and 2.0 wt%) nano-composites at low & high pelletize
pressure of 0.202 GPa
The composites were prepared by carefully combining appropriate quantity of Tl2O3 and silver nano-(Ag)-
particles of 40–50 nm dimension to previously manufactured Cu0.5Ba2Ca2Cu3O10-δ precursor at the desired
weight percentage (wt%). The mixture was ground in agate mortar and pestle for 1 h. The disc-shaped pellets
were made with low hydraulic press at 100 MPa, and encased in (Au) capsule, sintered about 10 min in a furnace
at 860 °C. Finally, using a hydraulic-press machine i.e. Cold Isostatic Press (CIP), these processed samples were
treated to pelletize high-pressure of 0.202 GPa. In this manner, the required superconducting-composites
(Ag)x/CuTl-1223; x = 0.0, 0.5, 1.0,1.5, and 2.0 wt% were synthesized at both low (100 MPa) and pelletize high-
pressure (0.202 GPa) as presented in figure 1.
2.2. Experimental characterization techniques
X-ray diffraction (XRD) data of superconducting composites were obtained, using a ‘D/Max IIIC Rigaku
diffactrometer’ with a CuKα radiation source (λ = 1.546 Å) in the 2°– 60° range. The lattice parameters of
CuTl-1223 phase was identified by computer software Xʹ Pert HighScore Plus, and matched with ICCD record
data. The morphologies of the prepared samples were examined by a scanning electron microscopy (SEM)
‘JEOL, Model No.5910’. The elemental composition of the prepared samples were identified using energy
dispersive X-ray spectroscopy ‘EDS; Det: Octane Pro, Reso’. The various vibrational oxygen modes in the wave
number (400–700) cm
‘Nicolet: 5700 FTIR spectrometer’. The relationship between temperature T(
) and electrical resistivity ( )r of
bar-shaped (V = 1.2 × 1.0 × 4.0 mm3) samples in a liquid nitrogen source were measured, using the standard
four-terminal method with dc current. Four low-impedance connections were formed on the surface of the semi
disc-like pellets via silver paste. Throughout the (
T )
μA. Using an Electronic Balance, the mass of each sample was measured with a precision of 0.001 g. All
characterizations of prepared samples were carried out at low-pelletize pressure of 100 MPa, and pelletization
high-pressure i.e. 0.202 GPa.
−1 range were investigated, using Fourier transform infrared spectroscopy (FTIR),
r - measurement, the current I( ) was held constant at 10
3. Results and discussion
3.1. Structural analysis
The crystal structure and phase confirmation of (Ag)x/CuTl-1223 nanocomposites with (x = 0.0 ∼ 2.0 wt%)
were analyzed via X-ray diffraction technique in the range of 2°– 60° at low and high pelletized pressure with
CuKα radiation source (l = 1.546 Å). Figures 2 and 3 illustrated XRD data of (Ag)x/CuTl-1223 nano-
composites with concentrations, x = 0.0, 1.0, and 2.0 wt% at low as well as high pelletized pressure, respectively.
All diffraction peaks of CuTl-1223 phase with a tetragonal structure have a relativity high intensity and well
3
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Figure 2. X-ray diffraction patterns for (Ag)x/CuTl-1223; x = 0, 1.0, and 2.0 wt% composites at low pressure synthesis. All peaks of
CuTl-1223 phase with tetragonal (P4/mmm) symmetry. Diffraction indexes assigned to each peaks, which were used to determine the
lattice parameter. The peaks marked by asterisk ‘*’, hash ‘#’, and not ‘o’ are attributed to the secondary phases CuTl-1212, CuTl-1234
and impurities, respectively.
Figure 3. X-ray diffraction patterns for (Ag)x/CuTl-1223 composites with x = 0, 1.0, and 2.0 weight percent at high pelletized
pressure of 0.202 GPa. All CuTl-1223 peaks have tetragonal (P4/mmm) symmetry. Diffraction indices were assigned to each peak and
used to calculate the lattice parameter. The peaks denoted by the letters asterisk ‘*’, hash ‘#’, and not ‘o’ correspond to the secondary
phases CuTl-1212, CuTl-1234, and impurities, respectively.
indexed following (P4/mmm) space group. Several un-index minimum intensity XRD peaks associated with
impurities and other undesirable superconducting phases were also identified in the XRD pattern shown in
figures 2 and 3. The crystal symmetry of parent matrix (CuTl-1223) were preserved with the inclusion of nano-
sized particles, showing that these NPs constituted inter-crystallite location (grain-boundaries). Furthermore,
these (Ag) NPs resides in the spaces between the grains and can help to improve the inter-grain connections [5].
The XRD spectra of (Ag)x/CuTl-1223; x = 0.0, 1.0, and 2.0 wt% nano-composites revealed that the crystal
structure of host matrix was unaffected by application of pelletization high-pressure of 0.202 GPa, with the
exception of minor change in the lattice parameters, due to oxygen concentrations at the grain boundaries
[11, 30]. Figure 4 shows the variation of c-axis length of the CuTl-1223 phase pelletized at low and high pressure
of 0.202 GPa. The small decrement in the dimension of c-axis of the CuTl-1223 phase unit cell can be attributed
to stresses and strains caused by inserted Ag NPs as well as high pelletization pressure, thereby promoting charge
carrier mobility along the c-axis length and lowering anisotropy in the unit cell [29]. Table 1 shown the lattice
parameters (a-axis and c-axis) of the CuTl-1223 phase computed using software ‘Xʹ Pert High-score Plus’. To
determine the percent volume fraction of distinct superconducting phases and an unknown impurity in
4
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Figure 4. Comparison of c-axis length of nano-(Ag)x/CuTl-1223 composites with x = 0, 1.0, and 2.0 wt% at low pressure as well as
pelletization high-pressure of 0.202 GPa.
Table 1. Crystal parameters of CuTl-1223 phase in (Ag)x/CuTl-1223; x = 0,
1.0 and 2.0 wt% composites processed at low pressure and at high pelletize
pressure.
Lattice parameters of
CuTl-1223 phase at low
pressure
Lattice parameters of
CuTl-1223 phase at high
pressure
a-axis (Å)
c-axis (Å)
a-axis (Å)
c-axis (Å)
4.19
4.21
4.21
15.29
15.27
15.26
4.20
4.21
4.22
15.22
15.19
15.17
(Ag)x/CuTl-1223
(wt%)
x = 0
x = 1.0
x = 2.0
prepared samples at low and high pelletized pressure, equation (1) was used [31].
%
of CuTl
(
-
)
1223
=
%
of CuTl
(
-
)
1234
=
%
of CuTl
(
-
)
1212
=
(
I1223
å
+
I1234
(
I1223
å
+
I1234
(
I1223
å
+
I1234
å
+
)
(
I 1223
+
I1212
å
+
)
(
I 1234
+
I1212
å
+
)
(
I 1212
+
I1212
(
I Unkown impurity
)
(
I Unkown impurity
)
(
I Unkown impurity
)
´
100
´
100
´
100
%
of Unknown impurity
(
)
=
(
I1223
å
+
(
I Unknown impurity
å
I1234
I1212
+
+
)
(
I Unkown impurity
´
100
)
⎫
⎪
⎪
⎬
⎪
⎪
⎪
⎭
( )
1
-Where I(1223), I(1234), and I(1212) represents the peak intensities of associated ‘superconducting phases’, and
I(unknown impurity) is the peak intensity of ‘unknown impurities’ in the X-ray diffraction spectra of
(Ag)x/CuTl-1223 nano-composites with x = 0.0, 1.0, and 2.0 wt. percent, respectively. Figure 5 presents the
variation in the percent phase fraction of CuTl-1223 superconductor verses Ag NPs concentration (x = 0, 1.0,
and 2.0 wt%) in (Ag)x/CuTl-1223 nano-composites at both low and pelletization high-pressure. The inclusion
of silver (Ag) nano-particles along with pelletization high-pressure, restricted the formation of additional phases,
and maintained the crystal structure of main CuTl-1223 matrix [14]. It was noticeable from figure 5 that volume
fraction determined for (Ag)x/CuTl-1223; x = 2.0 wt% is the optimum level at both low and high pressure,
implying the distinct improvement in the superconductivity. Tables 2(a), (b) shows the calculated phase
fractions of different phases, including the parent superconducting phase, at low and pelletization high-pressure
of 0.202 GPa. These results indicates that the volume fractions increased by 7% for x = 0, 8% for x = 1.0 wt%,
and 7% for x = 2.0 wt% of (Ag)x/CuTl-1223 nano-composites at high pelletized pressure compared to low
pelletized pressure.
5
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Figure 5. The percentage volume fraction of CuTl-1223 phase versus Ag NPs contents (x) in (Ag)x/CuTl-1223; x = 0, 1.0, and
2.0 wt% composites synthesized at low and high pelletize pressure. The volume fraction of the CuTl-1223 phase increases with high-
pressure, indicating the greatest improvement in superconducting properties.
Table 2. Determined percent volume fraction of CuTl-1223, CuTl-1234, CuTl-1212 and unknown impurity for the (Ag)x/CuTl-1223; x= 0,
1.0 and 2.0 wt% composites sintered at 860 °C, using (a) low pelletization pressure (b) high pelletization pressure (0.202 GPa).
Ag Nano-particles con-
tents x( )(wt%)
% Volume fraction of
CuTl-1223 phase
% Volume fraction of
CuTl-1234 phase
% Volume fraction of
CuTl-1212 phase
% Volume fraction of
unknown impurity
(a) (Ag)x/CuTl-1223 composites prepared at Low Pressure
x = 0
x = 1.0
x = 2.0
85.02
87.20
90.05
7.23
3.05
2.71
(b) (Ag)x/CuTl-1223 composites prepared at High Pelletize Pressure
5.18
6.82
5.05
2.36
2.93
2.19
Ag Nano-particles con-
tent x( )(wt%)
% Volume fraction of
CuTl-1223 phase
% Volume fraction of
CuTl-1234 phase
% Volume fraction of
CuTl-1212 phase
% volume fraction of
unknown impurity
x = 0
x = 1.0
x = 2.0
88.51
90.41
91.37
3.90
2.01
3.09
4.74
5.02
3.66
3.53
2.65
1.88
3.2. Morphological analysis
Scanning electron microscopy (SEM) is a technique for examining the surface morphology of various materials.
Figures 6(a)–(b) reveals SEM micrographs of (Ag)x/CuTl-1223 nano-composites with x = 1.5 wt percent
synthesized at low and pelletization high-pressure of 0.202 GPa. The morphology of the sintered
(Ag)x/CuTl-1223 sample with x = 1.5 wt% at low pressure, confirmed that the host CuTl-1223 matrix
accommodates Ag nano-particles in its granular regions. It is assumed that these finer particles ‘Ag nano-
particles’ may act as nano-pining centers, because their dimension match with the coherence length of CuTl-
1223 superconductor system [11]. According to the SEM image in figure 6(a), the insertion of silver
nanoparticles decreased the number of voids/pores and dislocations in the CuTl-1223 phase. Therefore, the
presence of metallic Ag NPs (0.0–1.5 wt%), improved inter-grain connectivity across the grain-boundaries as
well as volume fraction of bulk CuTl-1223 phase [6]. These results are applicable with obtained data from XRD,
where the CuTl-1223 phase increases with increasing nano-(Ag)-content up to 1.5 wt%. When pelletized
pressure was increased to 0.202 GPa, the number of voids/pores and dislocations in (Ag)x/CuTl-1223;
x = 1.5 wt% nano-composites significantly reduced, as shown in SEM image in figure 6(b). Application of high-
pressure of 0.202 GPa increased CuTl-1223 precursor–Ag–nanoparticles coupling, contributing to improved
inter-grain weak-links in (Ag)x/CuTl-1223 nano-composites. Furthermore, the addition of silver-nanoparticles
up to 1.5 wt% at high-pressure synthesis, minimizes the density of voids and enhances the connection between
CuTl-1223 grains, resulting a decrease in sample porosity as well as dislocations [25]. A close examination of
6
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Figure 6. (a), (b) SEM micro-graphs of (Ag)x/CuTl-1223 composites with x = 1.5 wt%, prepared at (a) low pressure of 100 MPa, and
(b) high-pelletize pressure of 0.202 GPa, respectively. (a) SEM image showing spherical Ag NPs and the successful formation of a
rectangular-like bulk CuTl-1223 phase. (b) SEM image indicating, the spherical Ag NPs present between plates-like CuTl-1223 phases
with improved inter-grain weak link by squeezing the grains together.
(Ag)x/CuTl-1223 nanocomposites with x = 1.5 wt. percent in figure 6(b) at high pelletize pressure revealed that
the (Ag) NPs-doped sample is mostly made up of relatively large CuTl-1223 grains size associated with discrete
nano-sized particles, as compared to low pressure. Therefore, the addition of Ag nano-particles reduced the
density of defects, and use of pelletization high-pressure improved inter-grain weak connections, leading in an
increasing superconducting volume fraction, contributed well to enhancing the superconducting properties of
the CuTl-1223 system [29, 32–34].
3.3. Elemental composition analysis
Energy dispersive X-ray spectroscopy (EDS) is the preferred technique for detecting and measuring the
elemental compositions in a materials. Figures 7(a)–(b) presents the EDX spectra of synthesized
(Ag)x/CuTl-1223; x = 1.5 wt. percent nano-particles superconducting-composites, at low and high pressure of
0.202 GPa, along with SEM images of 1μm magnification. These spectra indicated that peaks in the
(Ag)x/CuTl-1223 superconducting composites of various elements existed, including: Thallium (Tl), Barium
(Ba), Copper (Cu), Calcium (Ca), Oxygen (O), Aluminum (Al), and Molybdenum (Mo) etc There is no
significant difference observed in the count of various elements as well as peak intensities in the EDX spectra of
(Ag)x/CuTl-1223 nano-composites, pelletized at both low and high-pressure of 0.202 GPa. Tables 3(a) and (b)
gives the compositional percentage of each element as estimated via EDX analysis in (Ag)x/CuTl-1223;
x = 1.5 wt% nano-composites at (a) low pressure, and (b) high pelletize pressure, respectively.
3.4. FTIR analysis
The study of several oxygen vibrational modes in various materials, especially play a vital part in the phenomena
of superconductivity is evaluated, using Fourier Transform Infrared Spectroscopy (FTIR). Figures 8(a)–(b)
illustrated the FTIR spectra of (Ag)x/CuTl-1223 NPs superconductor-composites with x = 0.0, 0.5, 1.0, and
−1). The
1.5 wt% processed at low and high-pressure of 0.202 GPa in the wave number range (400 to 700 cm
7
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Figure 7. (a), (b) Characteristic EDX spectra for (Ag)x/CuTl-1223 with x = 1.5 wt%, CuTl-1223, CuTl-1234, and CuTl-1212 phases
recorded in the SEM using an electron energy of 20 k eV at (a) low pressure (100 MPa) and (b) high pelletize pressure (0.202 GPa) . The
inset SEM images at 1 μm magnification clearly shows enhanced inter-grain connectivity at 0.202 GPa pressure.
Table 3. Distribution of elemental quantitative analysis by EDX of
synthesized (Ag)x/CuTl-1223 composites with x = 1.5 wt% concentration.
Elements
Weight %
Atomic %
Error %
(a) (Ag)x/CuTl-1223; x = 1.5 wt% composites synthesized at Low Pressure
O K
Ca K
Ba L
Cu K
Tl L
Ag L
Total
14.51
10.02
34.01
31.08
8.90
1.48
100
11.31
21.80
18.75
42.65
3.80
1.69
100
5.34
4.51
2.96
3.82
34.74
1.34
—
(b) (Ag)x/CuTl-1223; x = 1.5 wt% composites synthesized at High Pelletize
Pressure
O K
Ca K
Ba L
Cu K
Tl L
Mo L
Ag L
Total
09.02
07.86
48.87
17.39
13.37
01.98
1.51
100
10.41
19.84
32.02
27.69
06.62
01.69
1.73
100
6.56
4.41
2.33
4.11
15.72
10.77
1.05
—
−1), (541–600 cm
associated bands consist of three types: apical oxygen (OA), CuO2 planar oxygen (Op), and the charge reservoir
layers (Oδ) atoms, respectively. The absorption bonds in the (400–540 cm
−1) ranges are associated to ‘apical oxygen atoms’ (OA), ‘CuO2 planar oxygen atoms’ (Op), and ‘Oδ
(670–700 cm
atoms in the charge reservoir layer’. The CuTl-1223 phase unit cell with apical, planar, and Oδ oxygen modes, are
shown in figure 9. For pure CuTl-1223 phase processed at low pressure of 100 MPa, the apical oxygen (OA)
modes of categories ‘Tl–OA–Cu(2)’, ‘Cu(1)–OA–Cu(2)’, planar oxygen modes ‘Cu(2)–Op–Cu(2)’ and ‘Oδ’
−1, respectively.
−1, 467–515 cm
oxygen modes are reported to be 415–430 cm
Figure 8(a) indicates, the apical oxygen modes (OA), Oδ oxygen modes, and planar oxygen modes (Op) have
shifted slightly after inclusion of Ag nano-particles to the host superconducting matrix. The little shift in the
location of all oxygen modes: OA, Op, and Oδ are possibly due to compression and relaxation of planar bond-
length caused by stresses and strains induced, following the (Ag) nano-structures to the host CuTl-1223 matrix.
Furthermore, FTIR spectra also confirmed that these nano-(Ag)-particles resides in the inter-granular space
−1, and 631–694 cm
−1, 584 cm
−1), and
8
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Figure 8. (a), (b) FTIR spectra of (Ag)x/CuTl-1223; x = 0, 1.0, 1.5, and 2.0 wt% composites, with associated bands consist of apical
oxygen (OA) atoms, CuO2 planner oxygen (OP) atoms, and Oδ atoms in the charge reservoir layer in the wave number range from 400
−1 synthesized at (a) low-pelletize pressure of 100 MPa and (b) high-pelletize pressure of 0.202 GPa, respectively.
to 700 cm
Figure 9. Structure of unit cell of CuTl-1223 superconducting phase, displaying cell characteristics as well as multiple oxygen phonon
modes.
(grain-boundaries) [6]. Therefore the host CuTl-1223 phase’s crystal structure and stoichiometry were retained
after addition of Ag NPs due to unchanged position of these oxygen vibrational modes [34]. When
(Ag)x/CuTl-1223 nano-composites processed at high-pressure of 0.202 GPa, for x = 0, the apical oxygen modes
of categories ‘Tl–OA–Cu (2)’, ‘Cu(1)–OA–Cu(2)’, planar oxygen modes ‘Cu–Op–Cu(2)’, and ‘Oδ’ oxygen modes
−1, respectively. Figure 9(b)
are identified at 407–426 cm
shows, the position of apical oxygen modes (OA), Oδ oxygen modes, and planer oxygen modes (Op) changed
slightly, when (Ag)x/CuTl-1223 nano-composites were subjected to a pressure of 0.202 GPa [29]. The minimal
shifting of all oxygen modes: OA, Op, and Oδ within the range may explains the action of pelletize high-pressure
−1, and 662–688 cm
–1, 443–471 cm
–1 555–582 cm
9
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Figure 10. Temperature T(
synthesized at low pressure with x = 0, 1.0, 1.5, and 2.0 wt%. The inset represents: how T 0c ( ) varies with Ag NPs concentration, and
magnified resistivity versus temperature approaching T 0c ( ) for superconducting composites.
) dependence of the electrical resistivity ( )r from 30 K to 300 K for nano-(Ag)-CuTl-1223 composites
Figure 11. A variable temperature-resistivity
concentration (x= 0, 1.0, 1.5, and 2.0 wt%) prepared at high pelletized pressure of 0.202 GPa. The inset shows variation of T 0c ( ) with
nano-(Ag)-particles contents x( ) from
superconducting composites.
r - measurements, and a large scale transition region at lower temperature for
(
r - curves in (30–300) K for nano-(Ag)x/CuTl-1223 composites with various Ag
(
T
T
)
)
on the bond length of unit cell, which happens as a result of stress and strains produced in the nano-composites.
However, high pelletization pressure had no effect on the entire structure of parent CuTl-1223 superconducting
phase.
T
3.5. Electrical measurements analysis
3.5.1. Resistivity versus temperature measurement
r - data of nano-composites were measured via four-probe system at
(
The resistivity verses temperature
temperature ranging from 30 K to 300 K. Figure 10 and figure 11 shown the temperature dependent electrical-
o £ £ 300 K, for (Ag)x/CuTl-1223 nano-Ag-composites, with nanoparticles
resistivity
contents (x = 0.0, 1.0, 1.5, and 2.0 wt%) prepared at low as well as high pelletize pressure, respectively. All
synthesized nano-Ag-composites indicates metallic-like nature at high temperature above Tc in the normal state,
and a superconducting transition state as the temperature decreased below Tc [14]. The transition regions are
shown in large scale in the inset of figures 10 and 11 to clarify the T 0 ,
r - in the T
(
)
c ( ) and to show how the Tc
onset was
T
T
)
10
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Table 4. Comparative superconducting parameters of (Ag)x/CuTl-1223 with x = 0, 1.0 and 2.0 wt% composites synthesized under different
pelletized pressure.
Low pressure versus High pelletize pressure
Ag NPs con-
tents(wt%)
T 0c ( ) (K)
onset (K)
Tc
Tc∆ (K)
Hole con-
centration p( )
T 0c ( ) (K)
onset (K)
Tc
Tc∆ (K)
Hole concentra-
tion p( )
0
1.0
1.5
2.0
91
93
96
98
97.90
98.95
99.75
100.55
6.90
5.95
3.75
2.55
0.0990
0.1002
0.1025
0.1042
99
101
105
107
106.50
107.40
109.55
110.05
7.50
6.40
4.55
3.05
0.1050
0.1067
0.1103
0.1121
)
T
r - curve of Ag)x/CuTl-1223 nano-composites. The zero resistivity critical temperature
(
was reported to be approximately 91–98 K, at low-pressure synthesis. It is obvious that Tc increase as
determined using
))=
(
c(
T R
0
Ag NPs increases from 0.0 to 2.0 wt% for (Ag)x/CuTl-1223 nano-composites as presented in the inset of
figure 10. This rise in Tc with the addition of Ag nano-particles might be responsible for the increase in the phase
fraction of CuTl-1223, conducting character of nano-Ag particles as well as improvement in grain connectivity
at low pressure synthesis [35]. These results are strongly supported our acquired XRD and SEM data. The
enhanced Tc behavior of (Ag)x/CuTl-1223 nano-composites when synthesized (at P = 0.202 GPa) was similar to
that of reported at low pressure. T R
x = 0.0, 1.0, 1.5, and 2.0 wt. percent under high pelletize pressure (as seen in the inset of figure 11). The
)= at high pelletization pressure as a result of compressing the inter-grain voids, which improve inter-
c (
T R
grain connection and therefore enhancing the number of carriers, and so the cooper pairing [22]. Moreover, the
enhancement of Tc at high pelletize pressure attributed to the decreased porosity, decreased c-axis length and
increased superconducting volume fraction [16]. Table 4 provides the values of T R
)= as a function of Ag
c (
( )
NPs contents x .
)= is observed to be 99 K, 101 K, 105 K, and 107 K for samples with
c (
0
0
0
3.5.2. Calculation of superconducting transition width
To study the efficiency and purity of the manufactured samples, the superconducting transition width was
determined using following relation, and listed in table 4.
∆
T
=
onset
T
c
-
(
T R
c
=
)
0
( )
2
)
0
c (
)= and Tc
onset were calculated based on experimentally obtained resistivity against temperature
Where T R
r - curves at low and high pressure, as illustrated in figures 12(a)–(b). It is noticed that the samples with
(
T
x = 2.0 wt% has least values of T∆ and large value of Tc
0.202 GPa, respectively. These findings are in accordance with the XRD patterns results, which revealed that the
samples with x = 2.0 wt% has minimum impurities (table 2). On the other hand, a widening of the
onset frequently indicates the existence of microscopic
T(∆ ) and least T
superconducting transition width
c
inhomogeneity [36, 37]. Table 4 displays for (Ag)x/CuTl-1223 nano-composites with x < 2.0 wt%, T∆
increases due to impurities inside the grain boundaries, resulting in reduced inter-granular coupling at low as
well as pelletize high-pressure respectively.
onset pelletized at low as well as high-pressure of
,
3.5.3. Determination of holes concentration
Equation (3) was used to determine the holes concentration p( ) following experimental values [38, 39].
p
=
0.16
-
-
T
c
max
T
c
82.6
1
⎡
⎢
⎢
⎣
⎤
⎥
⎥
⎦
( )
3
max for CuTl-1223
max refers to the maximum critical temperature of CuTl-1223 superconductors. Tc
Tc
superconductor is 132 K. Figure 13 exhibits variation of holes concentration p( ) as a function of nano-(Ag)
concentration x( ) for (Ag)x/CuTl-1223 nano-composites processed both low and high pelletize pressure,
respectively. At low pressure preparation the holes concentration was observed to increase with increasing Ag
nano-particles concentration, ranging from 0.0980 to 0.1042. This increase in holes concentration at low
T(∆ ) as well as increase in T 0c ( ) and superconducting
pressure is due to relative decrease in transition width
volume fraction [6]. The values of holes concentration p( ) at pelletize high-pressure (0.202 GPa) were found
around 0.1050, 0.1067, 0.1103, and 0.1121 for (Ag)x/CuTl-1223 nanocomposites with (x = 0.0, 1.0, 1.5, and
2.0 wt%). The higher values of holes concentration p( ) of (Ag)x/CuTl-1223 nanoparticles SC-composites
synthesized at 0.202 GPa pressure is attributed to improved void/pores, decrease in c-axis length, and better
connectivity of grains across the grain-boundaries [29]. Table 4 clearly indicates the holes concentration
11
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Figure 12. (a), (b): Double–Y plot of Tc∆ (K) and Tc
composites. The samples with x = 2.0 wt% has smallest value of T∆ and large value of Tc
pressure of 0.202 GPa.
onset as a function of Ag NPs contents (x) in (Ag)x/CuTl-1223 NPs/superconductor
onset at both (a) low pressure (b) pelletize high
Figure 13. Comparative plot of hole concentration p( ) versus Ag NPs in (Ag)x/CuTl-1223 NPs superconductor composites with
x = 0, 0.5, 1.0, and 2.0 wt. percent under low (100 MPa), and high pressure (0.202 GPa) conditions.
12
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Figure 14. Arrhenius plots of (Ag)x/CuTl-1223 with x = 0, 1.0, 1.5, and 2.0 wt% composites obtained at low pressure. {The inset:
shows a zoomed Arrhenius plot of transition region, and variation of activation energy Uo(
concentration x( )}.
) in (eV) versus nano-(Ag)-particles
increased for all samples of (Ag)x/CuTl-1223 NPs/SC composites produced under high-pressure as compared
to low-pressure, confirming the optimum superconducting properties.
3.5.4. Activation energy
The activation energy (Uo) of superconductors in the transition region near Tc is determined using the Arrhenius
equation [40].
r
(
T
)
r=
o
e
- /
Uo kBT
r
where T(
indicates normal-state resistivity at
onset and ‘kB’ is Boltzmann constant [41–44].Rearranging (4) yields:
somewhat higher temperature then T
c
represents temperature-dependent resistivity, or = T(
,
r
)
)
ln
r
(
T
)
r=
ln
o
(
T
)
-
Uo
kB
.
1
T
( )
4
( )
5
)
r
)/
T1(
A plot of ln T(
against
Arrhenius plots i.e. the relationship between the ‘ln
superconductor-composites with various values of x ,
) that are represented as a straight part at the end of the curves are employed to
transition temperature Tc(
) [42–44]. In the insets of figures 14 and 15 indicates magnified resistive
determined activation energy Uo(
transition Arrhenius plots along with linear part of zero resistivity area for calculating activation energy.
)-
yields a straight line with slope of U .o
’ and ‘ T1/ ’ for (Ag)x/CuTl-1223 nanoparticles
r r/
(
o
( ) at both low and high pelletize pressure. The regions near
Figures 14 and 15 illustrates the
)
(
At low-pressure synthesis, the calculated activation Uo(
) against NPs content x( ) for (Ag)x/CuTl-1223 nano-
) increases as ‘ x’ increases
composites are varies from 0.016 eV to 0.35 eV. It was noted that activation energy Uo(
up to optimum level i.e. x = 2.0 wt%, represented in the inset of figure 14. The mechanism for the increase of Uo
at ambient pressure synthesis is defined by the presence of Ag NPs between the grains, which can serve as inter-
grain connectivity [23, 24]. Additional argument, the rise in activation energy could explain the interaction of
mobile carriers with nano-(Ag)-particles at inter-grain sites [14].
The activation energy Uo(
) of (Ag)x/CuTl-1223 nano-composites with contents (x = 0.0, 1.0, 1.5, and
2.0 wt%), processed under high-pressure observed from 0.023 eV to 0.038 eV, as displayed in the (inset) of
figure 15, respectively. The reason for the increase in ‘Uo’ under high pressure synthesis is due to the metallic Ag
NPs present in the inter-grain sites by filling voids, reduced pores, and strengthened inter-grains weak-links by
compressing the grains together, resulting further increasing the activation energy of nano-(Ag)x/CuTl-1223
composites [14, 29]. Table 5 indicates the calculated values of activation energy for the (Ag)x/CuTl-1223 samples
prepared at low and high-pressure of 0.202 GPa. The table 5 clearly demonstrates that the activation energy
increases at high pressure over low pressure, which is verified by SEM, XRD, and resistivity verses temperature
r - data. Therefore, the addition of metallic nano-(Ag)-particles and a high pelletize pressure can reduced
(
T
)
13
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
Figure 15. Arrhenius plots of (Ag)x/CuTl-1223; x= 0, 1.0, 1.5, and 2.0 wt%, samples prepared at high pelletize pressure of 0.202 GPa.
{The inset displays magnified Arrhenius plot of transition region, and the variation of activation energy Uo(
( )}
contents x .
) in (eV) versus Ag NPs
Table 5. Calculated values of activation energy { Uo(
0.202 GPa.
) (eV)} as a function of Ag NPs contents in (x, wt%) for both low & high pressure of
Low pressure
High pelletize pressure
Ag NPs concentration (x, wt%)
0
1.0
1.5
2.0
Activation Energy{ U ,o(
) (eV)}
/
r r
(
)
k T
ln
B
=-
)(
Activation Energy{ U ,o(
) (eV)}
/
r r
(
)
k T
ln
B
=-
)(
0.016
0.024
0.029
0.035
0.023
0.028
0.032
0.038
defects density and improve weak-links which promote better superconducting properties of various HTSCs
[5, 6, 14, 18, 22, 29].
4. Conclusion
Sol–gel technique was used to synthesize Ag nano-particles. Secondly, superconducting nano-composites
{(Ag)x/(Cu0.5Tl0.5)Ba2Ca2Cu3O10-δ, 0 x 2.0 wt%}, were successfully prepared by two-step solid-state
reaction approach at 860 °C under ambient/low pressure. The synthesized composites were thereafter subjected
to high pelletize pressure of 0.202 GPa, using Cold Isostatic Press (CIP). X-ray diffraction, SEM, EDS, FTIR, and
electrical resistivity measurements were used to characterize the nano-(Ag)x/CuTl-1223 composites at both low
and high pelletize pressure. XRD spectra revealed that nano-composites (obtained at low-pressure of 100 MPa
and high pressure of 0.202 GPa) have a tetragonal structure, and an identical phase composition consisting of
mainly CuTl-1223 phase as well as secondary phases i.e. CuTl-1223 and CuTl-1234. The SEM micrographs
showed that some micro-defects, such as pores and cracks were improved with the inclusion of NPs at relatively
low-pressure synthesis. The synthesized nano-(Ag)x/CuTl-1223 composites at high pressure of 0.202 GPa,
reflect further improvement in porosity, enhanced weak links, and improved grain coupling with maximum
superconducting volume fraction. The FTIR spectra indicated no pronounced alteration in the position of
different oxygen vibration modes with the inclusion of nano-particles as well as with the application of high
pelletization pressure, confirming that the stoichiometry of the original CuTl-1223 matrix was maintained. The
zero and T
onset According
r - data was used to determine the critical temperatures of T
(
electrical resistivity
c
c
to the results of XRD, SEM, and electrical resistivity analysis, the sample with inclusion of 2.0 wt% NPs (Ag) at
high-pressure (P = 0.202 GPa), shows a pronounced enhancement in phase formation, grains connectivity,
T
)
,
.
14
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
holes concentration p ,
synthesis of CuTl-1223 superconductor is an advantageous step toward enhancing superconducting properties.
) and critical temperature T R
( ) activation energy U ,o(
)= Hence high-pressure
0 .
c (
Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).
ORCID iDs
Yaseen Muhammad
https://orcid.org/0000-0002-7042-1527
References
[1] Ihara H, Tanaka K, Tanaka Y, Iyo A, Terada N, Tokumoto M, Ariyama M, Hase I, Sundaresan A, Hamada N and Miyashita S 2000
Mechanism of Tc enhancement in Cu1−xTlx-1234 and-1223 system with Tc > 130 K Physica C 341 487–8
[2] Watanabe T, Miyashita S, Ichioka N, Tokiwa K, Tanaka K, Iyo A, Tanaka Y and Ihara H 2000 Carrier doping and superconducting
properties in Cu-1234 and CuTl-1223 superconductors Physica B 284 1075–6
[3] Yamamoto H, Tanaka K, Tokiwa K, Hirabayashi H, Tokumoto M, Khan N A and Ihara H 1998 Synthesis of Cu1−xTlxBa2Ca2Cu3O11−y
(x ∼ 0.7) superconductor by hot press Physica C: Superconductivity 302 137–42
[4] Tanaka K et al 2000 Tc beyond 130 K on a high-pressure synthesized (Cu, Tl)-1223 superconductor Physica B 284 1079–80
[5] Khan N A and Mumtaz M 2008 How grain-boundaries influence the intergranular critical current density of
Cu1−xTlxBa2Ca3Cu4O12−δ superconductor thin films J. Low Temp. Phys. 151 1221
[6] Ali L, Mumtaz M, Ali I, Rehman M W and Jabbar A 2018 Metallic Cu nanoparticles added Cu0.5Tl0.5Ba2Ca2Cu3O10−δ superconductor
J. Supercond. Novel Magn. 31 561
[7] Barakat M M, Al-Sayyed N, Awad R and Abou-Aly A I 2016 Superconducting parameter determination for (Co0.5Zn0.5Fe
2O4)x/Cu0.5Tl0.5-1223 composite Journal of Advanced Ceramics 5 210–8
[8] Abou-Aly A I, Abdel Gawad M M H, Awad R and G-Eldeen I 2011 Improving the physical properties of (Bi, Pb)-2223 phase by SnO2
nano-particles addition J. Supercond. Novel Magn. 24 2077–84
[9] Şakiroğlu S and Kocabaş K 2011 The effect of silver substitution in Bi1.7Pb0.3Sr2Ca2−x AgxCu3Oy J. Supercond. Novel Magn. 24 1321–5
[10] Özkurt B 2013 Nickel substitution effect on superconducting properties of Bi-2212 ceramics J. Alloys Compd. 579 132–7
[11] Jabbar A, Qasim I, Waqee-ur-rehman M, Zaman M, Nadeem K and Mumtaz M 2015 Structural and superconducting properties of
(Al2O3)y/CuTl-1223 composites J. Electron. Mater. 44 110–6
[12] El-Makdah M H, El-Ghouch N, El-Dakdouki M H, Awad R and Matar M 2023 Structural, electrical and mechanical properties of the
(NdFeO3)x/(CuTl)-1223 superconductor phase Appl. Phys. A 129 265
[13] Dahiya M, Kumar R, Kumar D and Khare N 2022 Comparative study of dimensionality and superconducting parameters in
YBCO–NaNbO3 nanoparticles-and nanorods-added composite sample from excess conductivity analysis Appl. Phys. A 128 308
[14] Anas M, El-Shorbagy G A, Abou-Aly A I and Khalaf A 2022 Synthesis, Characterization and Activation energy of
Nano-(GO)x/(CuTl)-1234 Superconducting Composites J. Low Temp. Phys. 206 210–31
[15] Basma H, Abbas S, Labban W, Awad R and Matar M 2022 Investigation of the structural and electrical properties of CdO/(Bi, Pb)−
2212 superconducting phase Phys. Scr. 97 065801
[16] Cevizci E B and Kocabas K 2023 Comparative investigation of magnetic and mechanical properties of nano-Sb2O3 and nano-Y2O3
addition on bismuth-based superconducting materials J. Mater. Sci., Mater. Electron. 34 657
[17] Drozdov A P, Eremets M I, Troyan I A, Ksenofontov V and Shylin S I 2015 Conventional superconductivity at 203 kelvin at high
pressures in the sulfur hydride system Nature 525 73–6
[18] Yamamoto A, Takeshita N, Terakura C and Tokura Y 2015 High pressure effects revisited for the cuprate superconductor family with
highest critical temperature Nat. Commun. 6 8990
[19] Sadewasser S, Schilling J S, Paulikas A P and Veal B W 2000 Pressure dependence of Tc to 17 GPa with and without relaxation effects in
superconducting YBa2Cu3Ox Phys. Rev. B 61 741
[20] Peifeng G, Zhang R and Wang X 2017 Pressure induced self-doping and dependence of critical temperature in stoichiometry
YBa2Cu3O6.95 predicted by first-principle and BVS calculations AIP Adv. 7 035215
[21] Murayama C, Iye Y, Enomoto T, Fukushima A, Mori N, Yamada Y and Matsumoto T 1991 Is pressure dependence of Tc through
change in carrier density? Physica C 185 1293–4
[22] Marconi D, Lung C and Pop A V 2013 The influence of pelletization pressure on normal and superconducting properties of (Bi, Pb):
2223 bulk system J. Alloys Compd. 579 355–9
[23] Deng L, Zheng Y, Wu Z, Huyan S, Wu H C, Nie Y, Cho K and Chu C W 2019 Higher superconducting transition temperature by
breaking the universal pressure relation PNAS 116 2004
[24] Khorramshahi M and Mohammadizadeh M R 2021 Y3Ba5Cu8Ox superconductor under hydrostatic pressure J. Low Temp. Phys.
203 309
[25] Abdulridha A R, Bermany E A, Hashima F S and Alkhayatt A H O 2020 Synthesis and characterization and pelletization pressure effect
on the properties of Bi1.7Pb0.3Sr2W0.2Ca2Cu3 O10+δ superconductor system Intermetallics 127 106967
[26] Li G and Wang M 2021 Influence of preforming pressure on the morphology and properties of single-grain Y–Ba–Cu–O
superconductors by the 011-type infiltration growth technique Mater. Today Commun. 29 10277
[27] Liu S, Matsumoto R, Jha R, Yamashita A, Kawaguchi S I, Goto Y, Takano Y and Mizuguchi Y 2021 High-pressure effects on
superconducting properties and crystal structure of Bi-based layered superconductor La2O2Bi3Ag0.6Sn0.4S6 J. Phys. Condens. Matter 33
225702
[28] Öner B, Özkurt P, Madre M A, Özkurt B and Sotelo A 2022 Enhanced superconducting properties in Bi2Sr2Ca1Cu1.75Na0.25Oy
ceramics prepared by hot-pressing under different pressures and temperatures J. Supercond. Novel Magn. 35 1831–8
[29] Muhammad Y, Mumtaz M, Ali L, Ali M, Rahim M, Hussain N and Iqbal Z 2022 Impact of high pelletize pressure on superconducting
properties of (Cu)x/CuTl-1223 composites J. Supercond. Novel Magn. 35 1–10
15
Phys. Scr. 98 (2023) 125967
Y Muhammad et al
[30] Ngenefeme F T J, Eko N J, Mbom Y D, Tantoh N D and Rui K W M 2013 A one pot green synthesis and characterization of iron oxide-
pectine hybrid nanocomposite J. Compos. Mater. 3 30–7
[31] Halim S A, Khawaldeh S A, Mohamed S B and Azhan H 1999 Superconducting properties of Bi2−xPbxSr2Ca2Cu3Oy system derived via
sol-gel and solid state routes Mater. Chem. Phys. 61 251–9
[32] Fultz B and Howe J M 2008 Transmission Electron Microscopy and Diffractometry of Materials 3rd Ed. (Berlin: Springer)
[33] Goldstein J I, Newbury D E, Echlin P, Joy D C, Lyman C E, Lifshin E, Sawyer L and Michael J R 2003 Generation of x-rays in the SEM
specimen Scanning Electron Microscopy and X-Ray Microanalysis (Springer) pp 271–96
[34] Jabbar A, Qasim I, Khan K M, Ali Z, Nadeem K and Mumtaz M 2015 Synthesis and superconducting properties of (Au)x/CuTl-1223
composites J. Alloys Compd. 618 110
[35] Awad R 2008 Study of the influence of MgO Nano-Oxide addition on the electrical and mechanicalproperties of (Cu0.25Tl0.75)-1234
superconducting phase J. Supercond. Nov. Magn. 21 461–6
[36] Jayaram B, Lanchester P C and Weller M T 1991 Localization and interaction effects during superconductor-insulator transition of
Bi2Sr2Ca1-xGdxCu2O8+d Phys. Rev. B: Condens. Matter 43 5444–50
[37] Bouchoucha I, Azzouz F B and Salem M B 2011 Excess conductivity studies in Zn0.95Mn0.05O and ZnO added YBa2Cu3Oy
superconductors J. Supercond. Nov. Magn. 24 345–50
[38] Presland M R, Tallon J L, Buckley R G, Liu R S and Flower N E 1991 General trend in oxygen stoichiometery effects on Tc in Bi and Tl
superconductors Physica C 176 95–105
[39] Bilgili O, Selamet Y and Kocabaş K 2008 Effects of Li substitution in Bi-2223 superconductors Journal of Superconductivity Novel
Magnetism 21 439
[40] Batista-Leyva A J, Orlando M T D, Rivero L, Cobas R and Altshuler E 2003 The resistive transition of
(Hg0.85Re0.15)(Ba1-ySry)2Ca2Cu3O8+δ superconducting polycrystals Physica C 383 365–73
[41] Jaroszynski J et al 2008 Upper critical fields and thermally-activated transport of NdFeAsO0.7F0.3 single crystal Physical Reviews B 78
174523
[42] Jabbar A, Qasim I, Mumtaz M, Zubair M, Nadeem K and Khurram A A 2014 Suppression of activation energy and superconductivity
by addition of Al2O3 nanoparticles in CuTl-1223 matrix J. Appl. Phys. 115 203904
[43] Feng S J, Ma J, Zhou H D, Li G, Shi L, Liu Y, Fang J and Li X G 2003 Energy dissipation in Bi2Sr2CaCu2O8+δ single crystals Physica C 386
22–5
[44] Hussain G, Jabbar A, Qasim I, Mumtaz M, Nadeem K, Zubair M, Abbas S Q and Khurram A A 2014 Activation energy and excess
conductivity analysis of (Ag)x/CuTl-1223 nano-superconductors composites J. Appl. Phys. 116 103911
16
| null |
10.1371_journal.pone.0248369.pdf
|
Data Availability Statement: All relevant data are
within the paper and its Supporting information
files.
|
All relevant data are within the paper and its Supporting information files.
|
RESEARCH ARTICLE
Climbing since the early Miocene: The fossil
record of Paullinieae (Sapindaceae)
Nathan A. JudID
G. Chery5*
1,3*, Sarah E. AllenID
2, Chris W. Nelson3¤, Carolina L. Bastos4, Joyce
1 Department of Biology, William Jewell College, Liberty, MO, United States of America, 2 Department of
Biology, Penn State Altoona, Altoona, PA, United States of America, 3 Florida Museum of Natural History,
University of Florida, Gainesville, FL, United States of America, 4 Laboratory of Plant Anatomy, Department of
Botany, Instituto de Biociências, Universidade de São Paulo, São Paulo, SP, Brazil, 5 School of Integrative
Plant Sciences, Section of Plant Biology and the L.H. Bailey Hortorium, Cornell University, Ithaca, NY, United
States of America
¤ Current address: Gainesville, FL, United States of America
* [email protected] (NAJ); [email protected] (JGC)
Abstract
Paullinieae are a diverse group of tropical and subtropical climbing plants that belong to the
soapberry family (Sapindaceae). The six genera in this tribe make up approximately one-
quarter of the species in the family, but a sparse fossil record limits our understanding of
their diversification. Here, we provide the first description of anatomically preserved fossils
of Paullinieae and we re-evaluate other macrofossils that have been attributed to the tribe.
We identified permineralized fossil roots in collections from the lower Miocene Cucaracha
Formation where it was exposed along the Culebra Cut of the Panama Canal. We prepared
the fossils using the cellulose acetate peel technique and compared the anatomy with that
of extant Paullinieae. The fossil roots preserve a combination of characters found only in
Paullinieae, including peripheral secondary vascular strands, vessel dimorphism, alternate
intervessel pitting with coalescent apertures, heterocellular rays, and axial parenchyma
strands of 2–4 cells, often with prismatic crystals. We also searched the paleontological liter-
ature for other occurrences of the tribe. We re-evaluated leaf fossils from western North
America that have been assigned to extant genera in the tribe by comparing their morphol-
ogy to herbarium specimens and cleared leaves. The fossil leaves that were assigned to
Cardiospermum and Serjania from the Paleogene of western North America are likely
Sapindaceae; however, they lack diagnostic characters necessary for inclusion in Paulli-
nieae and should be excluded from those genera. Therefore, the fossils described here as
Ampelorhiza heteroxylon gen. et sp. nov. are the oldest macrofossil evidence of Paullinieae.
They provide direct evidence of the development of a vascular cambial variant associated
with the climbing habit in Sapindaceae and provide strong evidence of the diversification of
crown-group Paullinieae in the tropics by 18.5–19 million years ago.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Jud NA, Allen SE, Nelson CW, Bastos CL,
Chery JG (2021) Climbing since the early Miocene:
The fossil record of Paullinieae (Sapindaceae).
PLoS ONE 16(4): e0248369. https://doi.org/
10.1371/journal.pone.0248369
Editor: William Oki Wong, Indiana University
Bloomington, UNITED STATES
Received: October 16, 2020
Accepted: February 23, 2021
Published: April 7, 2021
Copyright: © 2021 Jud et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting information
files.
Funding: This study received support from the
National Science Foundation (NSF) Award Number
0966884. The funders had no role in study design,
data collection and analysis, decision to publish, or
preparation of the manuscript. No additional
external funding received for this study.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
1 / 22
PLOS ONE Fossil Paullinieae
Introduction
Paullinieae (Sapindaceae) are tropical and subtropical woody vines (i.e., lianas), herbaceous
climbers (i.e., vines), and seldom shrubs [1]. The six genera of Paullinieae–Paullinia L., Serja-
nia L., Cardiospermum Kunth., Urvillea Kunth., Lophostigma Radlk., and Thinouia Triana &
Planch–form a clade [2–4, 21] defined by their tendrilate climbing habit and presence of stip-
ules [21]. With approximately 475 species [21], they comprise nearly one quarter of all species
in Sapindaceae. The Paullinieae are one of the four successively nested tribes of the Supertribe
Paulliniodae sensu by Acevedo-Rodrı´guez et al. [21], however the other members–Athyaneae,
Bridgesieae, Thouinieae–are all trees and shrubs. Numerous members of Paullinieae undergo
developmental re-patterning during the production of secondary xylem (i.e., wood) and sec-
ondary phloem (i.e., inner bark), resulting in the formation of “vascular cambial variants,”
such as continuous or discontinuous successive cambia, neoformations forming peripheral
secondary vascular strands (i.e., corded [5]), compound stems, fissured xylem, divided xylem,
lobed xylem, and phloem wedges [5–19].
The monophyly of Paullinieae within the subfamily Sapindoideae is supported by morphol-
ogy [20] and molecular sequence data [2–4, 21, 22]. Molecular phylogenetic analyses have
repeatedly yielded a long branch subtending the Paullinieae [2–4], suggesting shifts in nucleo-
tide substitution rates potentially associated with the evolution of the climbing habit. Previous
efforts to calibrate the phylogeny of Sapindaceae have yielded Oligocene or Miocene estimates
for the age of crown-group Paullinieae [23–25]; however, critical evaluation of the fossil record
is necessary to constrain the timing of diversification and the evolution of morphology and
anatomy of Paullinieae.
Although the fossil record of Sapindaceae is rich e.g., [1, 26], macrofossils of Paullinieae
are rare and at least some previous identifications are unreliable. Here, we describe the first
anatomically preserved macrofossils of Paullinieae. The fossils are roots, but nonetheless pro-
vide strong evidence of the climbing habit based on wood anatomy associated with climbing
in Sapindaceae. Next, we evaluate fossil leaves that have been attributed to the tribe. Then, we
summarize the fossil record of the tribe with a focus on macrofossils and identify occurrences
best suited for calibrating time-trees [27]. Finally, we discuss the implications of our findings
for future studies of the evolution of Paullinieae.
Materials and methods
Geologic setting
Two fossil roots were identified in a collection from the Lirio East site in lower part of the
Cucaracha Formation along the Culebra Cut (Gaillard Cut) of the Panama Canal (Fig 1).
These collections were made in in 2007 by F. Herrera and S.R. Manchester. The lower Cucara-
cha Formation consists of deltaic and coastal swamp deposits laid down during the early Mio-
cene when the nearby Pedro-Miguel Volcanic Complex was active [28–31]. At the Lirio East
site, fossil fruits as well as woods with bark are preserved as calcareous permineralizations in a
poorly sorted, carbonate-cemented sandstone [32].
So far, remains of Sacoglottis (Humiriaceae) [33], Oreomunnia (Juglandaceae) [34], Parinari
(Chrysobalanaceae) [35], Mammea (Calophyllaceae) [36], Rourea (Connaraceae) [37], and
Spondias (Anacardiaceae) [38], have been described. Plant macrofossils from elsewhere in the
Cucaracha Formation include palm stem fragments [39], Guazuma-like Malvaceae [40],
legume woods [39, 41], and a Malpighialean wood [42]. Fossil pollen from the Cucaracha For-
mation includes at least 52 pollen types [43]. Together, these records suggest the vegetation
was primarily tropical rainforest near the paleoshoreline of central Panama [43].
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
2 / 22
PLOS ONE Fossil Paullinieae
Fig 1. Native distribution of Paullinieae and fossil occurrences. Modern occurrence data from the BIEN database [45, 46]. Red star indicates the location
of the Lirio East fossil site where the fossil roots were collected. Fossil pollen occurrence codes: 1 = Serjania sp., upper Miocene Paraje Solo Formation [47–
49]; 2 = Serjania sp. and Paullinia sp., lower-middle Miocene Me´ndez Formation [50]; 3 = Serjania sp. and Paullinia sp., upper Miocene Gatun Formation
[49, 51]; 4 = Serjania sp., Paullinia sp., and Cardiospermum sp., upper Eocene Gatuncillo Formation [48, 52] Occurrence data were extracted from BIEN
ver. 4.1 database using the RBIEN package [46], supplemented with C. pechuelii data from GBIF [53]. Cardiospermum spp. distribution data follows native
ranges determined by [54, 55] (excluding controversial range in India).
https://doi.org/10.1371/journal.pone.0248369.g001
Fossil preparation
We cut the fossils in transverse and tangential and radial longitudinal sections using a Micro-
slice 2 annular saw and prepared serial sections using the cellulose acetate peel technique [44].
Peels were mounted on 25 x 75 mm glass slides with Canada Balsam or Eukitt mounting
medium and examined using light microscopy. Images of microscopic features were captured
with a Canon EOS digital camera mounted on a Nikon compound microscope with transmit-
ted light and processed with Adobe Photoshop (San Jose, California, USA). All specimens,
peels, and microscope slides are curated at the Florida Museum of Natural History Paleobotan-
ical Collections, Gainesville, Florida, United States.
Terminology and measurement protocols for the wood anatomy generally follow the
IAWA Hardwood List [56] but we adapted our approach for characters particular to Paulli-
nieae [64]. Summary statistics for anatomical characters were calculated from 25 measure-
ments. The fossil exhibits vessel dimorphism; this term has been used for both highly skewed
distributions and bimodal distributions [57–59], so we measured all vessels in the central
xylem cylinder [14] of a single transverse peel (n = 162) from the holotype (UF 19391-63016)
to generate a histogram of the distribution of vessel diameters. Then, we used the densityM-
clust function in the package mclust [60] in R [61] to identify the modes in the distribution
that correspond to the narrow and wide vessel classes. We report “narrow vessel diameter”
and “wide vessel diameter” as two separate characters. All measurements were made in ImageJ
1.50a [62].
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
3 / 22
PLOS ONE Fossil Paullinieae
Table 1. Summary of pre-Quaternary macrofossils that have been assigned to Paullinieae.
Species
Ampelorhiza heteroxylon
Bohlenia spp.
“Cardiospermum” coloradensis
“Cardiospermum” terminale
“Serjania” rara
Serjania mezzalire
Serjania itaquaquecetubensis
Serjania laceolata
Organ
root
Formation
Cucaracha
leaf
leaf
leaf
leaf
leaf
leaf
leaf
Klondike Mountain
Green River
Florissant; Renova
Aycross; Bridger
Rio Claro
Itaquaquecetuba
Itaquaquecetuba
Age
Mi.
Eo.
Eo.
Eo.
Eo.
Ol.
Mi.
Mi.
Country
Panama
USA
USA
USA
USA
Brazil
Brazil
Brazil
References
This study
[85, 86]
[81]
[77, 78, 95]
[74, 75]
[89]
[87]
[87]
Status
accepted
rejected
rejected
rejected
rejected
uncertain
uncertain
uncertain
Each identification is classified as accepted, rejected, or uncertain (material is consistent with Paullinieae, but alternative interpretations have not been ruled out). Mi.:
Miocene, Ol.: Oligocene, Eo.: Eocene. See text for further justification of status.
https://doi.org/10.1371/journal.pone.0248369.t001
Fossil leaves
We searched the literature for fossils identified as Paullinieae (Table 1). Of the species we
found, we examined specimens and images for those from North America and we re-described
their morphology following the format of the Manual of Leaf Architecture [63]. For putative
occurrences from South America and Europe, we evaluated images and descriptions from the
published literature. We used herbarium collections and online images to survey angiosperm
families for leaves with organization, margin type, and venation patterns similar to the fossil
leaf taxa re-described here (originally assigned to modern genera within Paullinieae). Then, we
compared the morphology of the fossils with leaves from extant genera in Paullinieae and with
leaves of selected genera outside Sapindaceae that exhibit similarities in organization, shape,
margin, and venation patterns. Cuticle was not preserved on any of the fossil leaves we exam-
ined and we did not evaluate cuticle for diagnostic characters. Comparisons are based on dried
specimens in the University (UC) and Jepson (JEPS) Herbaria at the University of California
—Berkeley, the R. L. McGregor Herbarium (KANU) at the University of Kansas, images avail-
able online via JSTOR Global Plants, and cleared and stained leaves in the National Cleared
Leaf Collection (NCLC-H; https://collections.peabody.yale.edu/pb/nclc/).
Phylogenetic analysis
We obtained the concatenated multiple sequence alignments from [21] and [22]. From these
datasets, we exclusively selected species within the supertribe Paullionieae as described by Ace-
vedo-Rodrı´guez et al. [21], which includes Athyaneae, Bridgesieae, Thouinieae, and Paulli-
nieae, totalling 100 ITS and 88 trnL intron sequences from [21], and 115 ITS sequences from
[22]. We then combined the two ITS datasets and realigned them in Geneious Prime 2021.0.3
(https://www.geneious.com) using the MUSCLE v3.8.425 aligner under default settings; the
trnL intron sequences were realigned under the same settings.
We then obtained wood anatomy data for 11 terminals from [13] and 33 terminals
from [64], and one terminal from [20], now available on morphobank (morphobank.org/
permalink/?P3910), and scored the fossil for 22 out of the 27 anatomy characters. Finally, we
added the character “habit” (0 = self-supporter, 1 = climber) and scored it for all extant species.
Although the wood anatomy characters scored for extant species were observed in stems and
the fossils are roots, available evidence indicates that differences in wood anatomy between
stems and roots within individual plants tend to be quantitative rather than qualitative [16, 65,
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
4 / 22
PLOS ONE Fossil Paullinieae
66]. The resulting dataset (S1 Appendix) comprises 216 tips and 1517 characters with three
partitions: anatomy (1-28), ITS (29-882), and trnL intron (883-1517).
We estimated the phylogenetic position of the fossil taxon using a Bayesian analysis with
two runs each of four chains (three hot, one cold, temp = 0.02) in MrBayes 3.2.7 [67]. We
applied the GTR+G model of nucleotide evolution to the ITS and trnL intron partitions. The
Mk model with rates drawn from a lognormal distribution was applied to the anatomy parti-
tion. The analysis ran for 12 million generations, sampling trees every 1000th generation. The
analysis converged with a standard deviation of split frequencies of 0.007428 and the estimated
sample size (ESS) of all parameters exceeded 2108. All trees were generated using the post
burnin (25% of initial trees discarded) from the combined MrBayes runs. The allcompat con-
sensus tree (50% majority rule consensus with compatible groups added) was generated with
the MrBayes command: contype = allcompat and annotated using iToL v4 [68]. The maxi-
mum clade credibility (MCC) tree was generated with Tree Annotator v1.10.4 [69], and the
maximum a posteriori tree (MAP) was generated with RevBays v1.10 [70]. The MrBayes input
nexus file (data matrix), allcompat consensus, MCC, and MAP trees, and full accession list
with associated molecular and anatomical data references are provided in (S1 Appendix).
Nomenclature
ptThe electronic version of this article in Portable Document Format (PDF) in a work with an
ISSN or ISBN will represent a published work according to the International Code of Nomencla-
ture for algae, fungi, and plants, and hence the new names contained in the electronic publica-
tion of a PLOS ONE article are effectively published under that Code from the electronic edition
alone, so there is no longer any need to provide printed copies. The online version of this work is
archived and available from the following digital repositories: PubMed Central and LOCKSS.
Results
Fossil roots
Family. Sapindaceae Jussieu.
Subfamily. Sapindoideae Burnett.
Tribe. Paullinieae (Kunth) DC.
Genus. Ampelorhiza Jud, S.E. Allen, Nelson, Bastos & Chery gen. nov.
Generic diagnosis. Roots woody with neoformations forming peripheral secondary vas-
cular strands; vessels of two distinct size classes, wide vessels solitary and in tangential multi-
ples, narrow vessels in long radial multiples; intervessel pits alternate with slit-like coalescent
apertures on the walls of large vessels; heterocellular rays composed of mixed upright, square,
and procumbent cells; axial parenchyma strands 2–4 or more cells tall, often chambered with
prismatic crystals.
Type species. Ampelorhiza heteroxylon Jud, S.E. Allen, Nelson, Bastos & Chery gen. et sp. nov.
Specific diagnosis. As for genus.
Holotype. UF 19391-63016 (Figs 2 and 3).
Paratype. UF 19391-63026 (S1 Fig).
Repository. Florida Museum of Natural History (FLMNH), Gainesville, Florida, U.S.A.
Type locality. Panama; Culebra Cut, northeast side of the Panama Canal (N 9.051375˚, W
79.649027˚, WGS84).
Stratigraphic position and age. Cucaracha Formation; early Miocene, c. 18.5–19 Ma
[30, 31].
Etymology. The genus comes from the Greek ámpelos, meaning vine, and ríza meaning
root. The specific epithet comes from the Greek héteros meaning different and -xylon meaning
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
5 / 22
PLOS ONE Fossil Paullinieae
Fig 2. Cambial variant and vessel characters in Ampelorhiza heteroxylon. (A) Transverse section of the stem
showing diffuse-porous wood of the central cylinder (cc) and peripheral vascular strands (ps) in the external tissues.
Arrow indicates the position of the smaller of two peripheral vascular cylinders. Specimen UF 19391-63016 XS peel 10.
(B) Close up transverse section of the larger of two peripheral vascular strands. Specimen UF 19391-63016 XS peel 10.
(C) Transverse section of the smaller of two peripheral vascular cylinders. There is no pith. Specimen UF 19391-63016
XS peel 10. (D) Close up of A showing the primary vascular parenchyma. Specimen UF 19391-63016 XS peel 10. (E)
Tangential longitudinal section through the tall cells of the primary vascular parenchyma (center right), ray cells
(center left) and juvenile wood (far left). UF 19391-63016 LS peel 16. (F) Transverse section showing wide solitary
vessels (WV) and narrow vessels in long radial multiples (at arrow). Specimen UF 19391-63026 XS peel 6. (G)
Tangential longitudinal section (LS) showing coalescent pit apertures on the vessel wall. Specimen UF 19391-63016 LS
peel 6. (H) Tangential longitudinal section showing alternate polygonal pits on the vessel wall (at arrow). Specimen UF
19391-63016 LS peel 7. (I) Tangential longitudinal section showing narrow vessels (NV) with oblique end walls, and
wide vessels (WV) with simple perforation plates and end walls perpendicular to lateral walls (right arrow). Specimen
UF 19391-63026 TLS peel 1. Scale bars: A = 3 mm; B = 1 mm; C, F, I = 200 μm; D, E = 500 μm; G, H = 100 μm.
https://doi.org/10.1371/journal.pone.0248369.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
6 / 22
PLOS ONE Fossil Paullinieae
Fig 3. Wood anatomy in Ampelorhiza heteroxylon. (A) Tangential longitudinal section showing uniseriate pitting on the fiber walls. Specimen UF 19391-
63016 LS peel 5. (B) Tangential longitudinal section showing axial elements including narrow vessels and uniseriate rays (at arrow). Specimen UF 19391-
63016 LS peel 1. (C) Radial longitudinal section showing ray cells against a vessel. Note the partially preserved vessel-ray parenchyma pitting similar in size
to the intervessel pitting (at arrow). Specimen UF 19391-63016 LS peel 7. (D) Tangential longitudinal section showing uniseriate and biseriate rays (left
arrow) and axial elements with crystals (right arrow). Specimen UF 19391-63016 LS peel 5. (E) Radial longitudinal section showing upright (at arrow),
square, and procumbent ray cells. Specimen UF 19391-63026 LS peel 2. Scale bars: A = 70 μm; B = 150 μm; C = 40 μm; D, E = 100 μm.
https://doi.org/10.1371/journal.pone.0248369.g003
wood, referring to the different sizes of the peripheral secondary vascular strands found in
Paullinieae.
Description (descriptio generico-specifica). The holotype is an axis 1 cm wide and 3 cm
long; the paratype is an axis 0.5 by 1 cm wide and 2.5 cm long. Each consists of bark with one
or two peripheral secondary vascular strands (Fig 2A–2C), surrounding a central woody cylin-
der with a small core of primary vascular parenchyma (Fig 2A–2C). The peripheral vascular
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
7 / 22
PLOS ONE Fossil Paullinieae
strands consist of secondary xylem and phloem derived from C-shaped cambia that lack pri-
mary vascular parenchyma. In the holotype, the two preserved peripheral strands are of differ-
ent sizes. One is c. 3.3 mm by c. 2.0 mm in transverse section and the other is 0.7 mm by c. 0.4
mm (Fig 2A lower arrow, Fig 2C). Primary vascular parenchyma in the central cylinder of the
holotype is an eccentric collection of parenchyma cells 200 μm tall by 500 μm wide (Fig 2D).
Radial files of cells with dark contents also extend away from the center of the central cylinder
on one side (Fig 2D). The primary vascular parenchyma cells are tall (c. 150–300 μm), and
many have dark contents in the lumen (Fig 2E). Secondary xylem is diffuse porous (Fig 2A &
2F). Growth rings are indistinct (Fig 2A & 2F). Vessels are in two distinct size classes: wide ves-
sels 50-270 μm (mean: 104 μm) in tangential diameter, mostly solitary but also in tangential
multiples of 2–3; narrow vessels are 11–50 μm in tangential diameter and arranged in radial
multiples of 2–9 (Fig 2A & 2F). Vessel elements are 153–280 μm long (mean: 223 μm, n = 14).
Mean vessel frequency is 27 per mm2. Vessel element end walls are without scalariform bars;
perforation plates are simple (Fig 2I). Tyloses and helical thickenings were not observed. Inter-
vessel pits alternate with distinct borders and coalescent apertures on the walls of large vessels
(Fig 2G & 2F). Vessel-ray parenchyma pits were difficult to observe; we did not find large sim-
ple pits different from those on the vessel walls (Fig 3C). Fibers are poorly preserved but appear
non-septate with minutely bordered uniseriate pits on the radial walls (Fig 3A). Axial paren-
chyma is diffuse and scanty paratracheal, with strands at least 2–4 cells tall and often cham-
bered with prismatic crystals (Fig 3D). Rays are 1–2 (rarely three) cells wide, less than 1 mm
tall, and heterocellular with rows of procumbent square and upright cells mixed throughout
(Fig 3E). Secretory structures were not observed.
Remarks. Although cambial variants are often associated with the climbing habit, the
presence of peripheral vascular strands is not sufficient to identify the fossils as stems or roots.
Bastos et al. [16, 66] demonstrated that cambial variants can be found in both organs. In stems
of Paullinieae, the pith is conspicuously angular (e.g., triangular, pentangular) in transverse
section with primary vascular bundles at the corners [19]. By contrast, in roots the primary
vascular parenchyma is diarch and this region (i.e., the “medulla”) is oval and smaller than the
stem pith in transverse section (Fig 4). In Ampelorhiza heteroxylon, there is an eccentric oval-
shaped parenchymatous core c. 200 by 500 μm in diameter (Fig 2D); therefore, our interpreta-
tion is that the specimens are roots.
We initially recognized that these fossils might be lianas based on the diameter of the largest
vessels relative to the width of the axis. To illustrate this approach, we used logistic regression
to classify unknown fossil axes from Lirio East as climbers or self-supporters based on maxi-
mum vessel diameter and diameter of the central woody cylinder (S2 Fig). The model was
trained using a dataset of 71 samples obtained from Ewers et al. [71], and predicted the habit
of 22 fossil axes with woody cylinders greater than 5 mm in diameter from the Lirio East fossil
collections. Although the model did predict that the Ampelorhiza fossils (and the Rourea fossil
described by Jud and Nelson [37]) are climbers, the training dataset is only stem material and
therefore may not be suitable for classifying roots, given the patterns found by Ewers et al. [72]
when comparing stems and roots in lianas and trees. Further work on the relationship between
hydraulic path length, vessel diameter, and root diameter in lianas and self-supporters (as has
been done for stems [73]) would be useful for identifying lianas in the fossil record.
Fossil leaves
We found one fossil species assigned to Serjania and two assigned to Cardiospermum from
North America in the literature (Table 1). All three were described from fossils of isolated leaf-
lets or partially complete compound leaves (Fig 5). MacGinitie [74] described “Serjania” rara
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
8 / 22
PLOS ONE Fossil Paullinieae
Fig 4. Wood anatomy of the roots of extant Paullinieae species. A–B: Neoformations forming peripheral vascular strands in Serjania
caracasana (Jacq.) Willd. in transverse section. (A) Root macromorphology presenting a cambial variant. Arrows point to individual
peripheral vascular strands. (B) Close up of the juncture of the central cylinder (cc) and a peripheral vascular strand (ps) with a c-shaped
“pith” (i.e., primary vascular parenchyma of the root). (C) Secondary xylem of Thinouia scandens Triana & Planch. with vessel
dimorphism in transverse section. Note the wide vessels (WV) are solitary or in tangential (upper arrow) or radial multiples (lower
arrow), while the narrow vessels (NV) are in longer radial chains. (D) Primary vascular parenchyma in the center of the the diarch roots
(arrows towards protoxylem poles) of S. caracasana in transverse section. (E) Alternate intervessel pits (lower arrow) and those with
coalescent apertures (upper arrow) in S. caracasana in tangential longitudinal section. (F) Prismatic crystals in the axial parenchyma (�) of
S. caracasana in macerated material. Scale bars: A = 0.5 cm, B = 1 mm, C = 250 μm, D = 100 μm, E = 70 μm, F = 50 μm. �prismatic
crystals in axial parenchyma.
https://doi.org/10.1371/journal.pone.0248369.g004
based on leaves from the Eocene Aycross Formation in northwestern Wyoming. The same
species also occurs in the Eocene Bridger Formation in southwestern Wyoming [75]. “Cardios-
permum” terminale (Lesquereux) MacGinitie was first described from the Eocene Florissant
Formation in central Colorado by Lesquereux [76] as Lomatia. MacGinitie [77] transferred
these specimens and others to Cardiospermum based on the twice-ternate leaf organization
and architecture of lobes, teeth, and major vein framework of the leaflets. This species was
later reported from the late Eocene to early Oligocene Climbing Arrow Member of the Renova
Formation in southwestern Montana [78, 79] as well. Finally, “Cardiospermum” coloradensis
(Knowlton) MacGinitie was first described from the Eocene Green River Formation as Phyl-
lites by Knowlton [80]; and later transferred to Cardiospermum by MacGinitie [81]. This
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
9 / 22
PLOS ONE Fossil Paullinieae
Fig 5. Leaf fossils previously assigned to Paullinieae. (A) “Serjania” rara MacGinitie from the Bridger Formation,
Blue Rim site, Sweetwater County, Wyoming, UF 15761S-57786. (B) “Serjania” rara MacGinitie from the Bridger
Formation, Blue Rim site, Sweetwater County, Wyoming, UF 15761N-61430. (C) Paratype of “Serjania” rara
MacGinitie from the Aycross Formation, Kisinger Lakes site, northwestern Wyoming (Pl 25, Fig 3 in [74]), UCMP PA
108, 5698. (D) Hypotype of “Cardiospermum” coloradensis (Knowlton) MacGinitie from the Green River Formation,
west of Wardell Ranch site, Colorado (Pl 22, Fig 3 in [81]), UCMP PA 321, 20593. Arrow indicates marginal vein. (E)
“Cardiospermum” coloradensis (Knowlton) MacGinitie from the Green River Formation in Rainbow, UT, UCMP
PB02016, 201265. Arrow indicates marginal vein. (F) “Cardiospermum” terminale (Lesquereux) MacGinitie from the
Florissant Formation in central Colorado, FLFO 10147. Scale bars = 1 cm.
https://doi.org/10.1371/journal.pone.0248369.g005
species has been reported from throughout the Green River Formation [81–84]. Updated
descriptions of these three species are provided in the (S2 Appendix).
The extinct genus Bohlenia Wolfe & Wehr [85] was established for sapindaceous leaves and
fruits from the Eocene Republic flora (Klondike Mountain Formation) in Washington, USA
(Table 1). Wolfe and Wehr [85] suggested that B. americana (Brown) Wolfe & Wehr may
belong to Paullinieae based on the course of the secondary veins and on the assumption that
co-occurring samaras belonged to the same species; however, McClain and Manchester [86]
transferred the samaras to Dipteronia brownii McClain & Manchester and noted that Bohlenia
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
10 / 22
PLOS ONE Fossil Paullinieae
foliage is similar to Koelreuteria elegans (Seem.) A.C. Sm. Both of these fossil species are mem-
bers of Sapindaceae, but neither belong to Paullinieae.
We also found three species assigned to Serjania from the Cenozoic of Brazil in the litera-
ture (Table 1). Fittipaldi et al. [87] described Serjania lanceolata Fittipaldi, Simões Giulietti et
Pirani and Serjania itaquaquecetubensis Fittipaldi, Simões Giulietti et Pirani based on incom-
plete unlobed, toothed leaf blades from the Oligocene upper Itaquaquecetuba Formation. To
our knowledge, the characteristic pollen of Paullinieae has not been recognized in palynologi-
cal studies of this formation [88]. Finally, Serjania mezzalire Duarte et Rezende-Martins was
described from fossil leaves in the Miocene Rio Claro Formation [89, 90].
Edwards and Wannacot [91] compiled list of all fossil species that had been assigned to
Paullinia based on leaf morphology from Europe. They concluded that a close relationship to
extant Paullinieae can be rejected or is doubtful for all of them based on morphology or quality
of preservation. We concur, so we did not consider these further.
There is considerable variation in the blade shape, margin type, tertiary venation, and base
shape among extant Paullinieae (Fig 6). Leaf margins may be unlobed or lobed, toothed or
untoothed. Toothed margins may be serrate, dentate, or crenate. Secondary vein framework
may be craspedodromous, semicraspedodromous, brochidodromous, or eucamptodromous.
Leaf organization is also variable. Leaves may be simple, once or twice imparipinnate, or up to
thrice ternate (most commonly twice ternate). In compound leaves, the rachis may be winged
or unwinged. Axillary tendrils may be absent or present. Many of these characters also vary
across Sapindaceae. Based on our observations, isolated fossil leaves or leaflets of Paullinieae
may be recognizable if they preserve a combination of the following characters: Axial tendrils,
stipules, ternate compound organization, winged rachides, and absence of a marginal vein.
Morphological similarities between “Cardiospermum” coloradensis, “C.” terminale, “Serja-
nia” rara, and the leaves of some extant Paullinieae include 1. ternate-compound organization,
2. decurrent (Figs 5C, 5E and 5F and 6A and 6B) or complex leaflet bases (Fig 5A and 5B), 3.
irregular spacing of secondary veins, 4. secondary veins that terminate beyond the apex of
lobes/teeth, 5. secondary veins that terminate in angular (“V-shaped”) sinuses (Fig 5), and 6.
secondary veins that bifurcate around angular sinuses (Fig 5E). However, some or all of these
characters can be found in the leaves of other Sapindaceae (e.g., Thouinia Poit., Koelreuteria
Laxm., Dipterodendron Radlk., Dilodendron Radlk., and Athyana (Griseb.) Radlk) and in other
families (see Discussion section for further commentary); they are not diagnostic of Cardios-
permum, Serjania, nor Paullinieae. Furthermore, a prominent marginal vein like that present
in at least some specimens of the fossil species (Fig 5D and 5E) is not present in extant Serjania
and Cardiospermum (Fig 6A and 6B). The descriptions and images of “Serjania” lanceolata,
“S.” itaquaquecetubensis, and “S.” messalire show the shape of the blade, the presence of a ser-
rate margin, and craspedodromous secondary vein framework [87, 90]. Although these char-
acters are consistent with Serjania, their combination is not diagnostic of the genus.
Phylogenetic position of Ampelorhiza
We evaluated the placement of Ampelorhiza by observation of the allcompat consensus, MCC,
and MAP trees sampled from the posterior distribution. Ampelorhiza is always nested within
extant Paullinieae, however its relationship with extant genera differs based on the method
used to generate the tree, reflecting the uncertainty typical of taxa with a high proportion of
missing data. In the allcompat consensus tree (Fig 7) Ampelorhiza is nested within a clade with
Cardiospermum, Paullinia, and Serjania. The various positions of Ampelorhiza within this
clade is represented as a polytomy that includes several lineages of Serjania and Cardiosper-
mum. In the maximum a posteriori tree (S1 Appendix), Ampelorhiza is nested within Urvillea,
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
11 / 22
PLOS ONE Fossil Paullinieae
Fig 6. Extant leaves. Modern leaves for comparison with the putative Paullinieae fossils. Cleared leaves from the National Cleared Leaf Collection (NCLC).
(A) Serjania rhombea Radlk. (Coll.: W.H. Lewis, J.D. Dwyer, T.S. Elias, and R. Solı´s #72 (UC 1355158), 1966, Panama]. (B) Cardiospermum halicacabum L.
[Coll.: R.D.A. Baylis #5080 (UC 1409568), 1972, South Africa]. (C) Paullinia pinnata L., NCLC 0012. (D) Quercus nigra L., NCLC 0215. (E) Lycopersicum
esculentum L., NCLC 1640. (F) Beauprea balansae Brongn. & Gris, NCLC 6658. Scale bars = 1 cm.
https://doi.org/10.1371/journal.pone.0248369.g006
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
12 / 22
PLOS ONE Fossil Paullinieae
Fig 7. Phylogeny of supertribe Paulliniodae. (A) Majority rule consensus tree with all compatible groups (“allcompat”) of supertribe Paulliniodae sensu
Acevedo et al. [21], generated in MrBayes 3.2.7 from an anatomical and molecular combined dataset of 216 tips. Branch colors indicate posterior
probabilities. The outermost black line indicates the tribe Paullinieae. Note the position of the fossil taxon Ampelorhiza within Paullinieae indicated by the
arrow and the dagger. (B) Summary tree showing the same topology, but pruned to show genera only, assuming all genera are monophyletic. Numbers
above branches are posterior probabilities, dashes indicate genera represented by a single species in the “allcompat” consensus tree.
https://doi.org/10.1371/journal.pone.0248369.g007
whereas in the maximum clade credibility tree Ampelorhiza is nested within Serjania. These
results further supports our circumscription of Ampelorhiza as a distinct genus from extant
Paullinieae. The placement of Ampelorhiza within Paullinieae is supported by vessel dimor-
phism, heterocellular rays, and axial parenchyma strands typically 2-4 cells long. One synapo-
morphy of Paullinieae that we did not observe in the fossil is wide rays (ray dimorphism);
however, we only examined two root fragments and this character is observed in many, but
not all, samples from modern roots [16].
Discussion
Roots
The combination of peripheral vascular strands (Figs 2A–2C and 4A & 4B), vessel dimorphism
(Figs 2F & 2I and 4B–4D), wide vessels solitary or in tangential multiples of 2–3 (Fig 2F and
4C), narrow vessels in long radial multiples of 2–21 (Figs 2F and 4C & 4D), alternate interves-
sel pits with slit-like coalescent apertures (Figs 2G and 2H and 4E), heterocellular rays, pris-
matic crystals in axial parenchyma (Figs 3D and 4F), and dark content (possibly phenolic
compounds) in primary vascular parenchyma and ray parenchyma (Fig 2D and 2E) support
the inclusion of Ampelorhiza in Paullinieae [13, 16, 18, 64, 66, 92, 93, 94]. Two wood anatomi-
cal characters typical of extant Paullinieae were not observed in the fossils: 1) alternating bands
of thin and thick-walled regions in the wood which may either be axial parenchyma alternating
with ordinary fibers (e.g., Serjania spp.) or parenchyma-like fiber bands alternating with
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
13 / 22
PLOS ONE Fossil Paullinieae
ordinary fibers (e.g., Paullinia spp.) and 2) ray dimorphism. Because the bands are clearest in
sufficiently thin, stained sections or polished blocks, it may be that the thickness of the peels
and the absence of stain obscures this feature.
The cambial configuration in stems and roots is highly variable in Paullinieae. Chery et al.
[19] and Cunha Neto et al. [18] together distinguished six ontogenetic pathways in the stems
of Paullinia alone, and we expect that Serjania has the most variation in the tribe based on pre-
liminary observations of images in the Smithsonian Liana databases (Acevedo & Chery, per-
sonal observation). Furthermore, Bastos [16, 66] showed that roots may or may not also have
cambial variants, and when present they do not necessary mirror the configuration of the
stems. An asymmetrical distribution of peripheral secondary vascular strands of different
sizes, as in Ampelorhiza heteroxylon, occurs in the roots of Serjania caracasana (Fig 4A & 4B)
and the stems of some Paullinia [18]. Given the variation among stems and the paucity of data
on cambial variants in roots, the configuration of secondary growth in the fossils does not jus-
tify assignment to one of the extant genera.
Despite some anatomical differences among the genera of Paullinieae, the fossils of Ampe-
lorhiza do not preserve a combination of wood anatomy characters diagnostic of any extant
genus either, they are most similar to some Serjania. The wood of Serjania stems has banded
axial parenchyma, no septate fibers, and crystals confined to axial elements, whereas Paullinia,
Thinouia, and Cardiospermum have scanty axial parenchyma, abundant septate fibers, and
crystals in ray parenchyma. Thinouia differs from Paullinia and Cardiospermum by the
absence of crystals in axial elements [13], and some Paullinia can be recognized by a herring-
bone pattern in the wide rays when viewed in transverse section [13]. The fossils do not have
banded parenchyma, nor do they have wide rays with a herringbone pattern. They do have
crystals in the axial elements but we did not observe them in the rays, nor did we detect septate
fibers.
Leaves
We reject the generic assignments of Cardiospermum and Serjania species described from fos-
sil leaf material. Our search for leaves with organization, margin features, and venation archi-
tecture similar to “C.” coloradensis, “C.” terminale, and “S.” rara outside of Sapindaceae led to
comparisons with Anacardiaceae (e.g., Rhus L.), Fagaceae (e.g., Quercus L.), Proteaceae (e.g.,
Roupala Aubl., Lomatia R. Br., Beauprea Brongn. & Gris), Ranunculaceae (e.g., Clematis L.),
and Solanaceae (e.g., Hyoscyamus L., Chamaesaracha (A. Gray) Benth. & Hook. f., Physalis L.,
Lycopersicum Hill.). Some Rhus (Anacardiaceae) have similar shapes to the fossil material, but
secondary venation in Rhus varies from craspedodromous to cladodromous. Some Fagaceae
have similar blade shape, secondary veins, and major veins that project beyond the margin of
the blade; however, all Fagaceae have simple leaves and the sinuses are generally rounded
rather than angular as in the fossils. Previous authors (e.g., [76, 83]) have attributed fossils like
these to Proteaceae; however, although secondary veins in the Proteaceae are variable (e.g.,
brochidodromous to semicraspedodromous to festooned brochidodromous to festooned
semicraspedodromous), they are unlike the craspedodromous framework in the fossils and
again the sinuses between teeth are generally rounded in Proteaceae rather than angular. The
compound leaves of some lobed and toothed Clematis (Ranunculaceae) can be distinguished
from the fossils because they usually have festooned secondary venation. Finally, several Sola-
naceae have asymmetric blades and similarly shaped teeth and lobes; however, again the
sinuses tended to be rounded rather than angular as in the fossils.
Leaf architectural characters preserved in “C.” coloradensis, “C.” terminale, and “S.” rara
support inclusion in Sapindaceae, yet we consider a close relationship with Paullinieae unlikely
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
14 / 22
PLOS ONE Fossil Paullinieae
based on the presence of a prominent perimarginal vein in the fossils and the absence of co-
occurring fossil fruits or pollen despite decades of intensive sampling in the Green River For-
mation and the Florissant fossil beds. Similarly, in his update of the fossil flora of Florissant,
Manchester [95] doubted the generic assignment of “C.” terminale based on the rather coria-
ceous texture of the fossils compared to extant Cardiospermum and the absence of associated
fruits. Other extant Sapindaceae with similar leaf organization, margin type, teeth, and vena-
tion include: Thouinia Poit., Koelreuteria Laxm., Dipterodendron Radlk., Dilodendron Radlk.,
and Athyana (Griseb.) Radlk.
Evolution of Paullinieae
To our knowledge, the oldest reliable fossil evidence of Paullinieae is heteropolar hemi-tri-syn-
colpate pollen from the Gatuncillo Formation in Panama [52]. Some fossil species of the gen-
era Syncolporites and Proteacidites (used for dispersed pollen) may belong to Paullinieae (or
Proteaceae or Myrtaceae) [96]; however, a review of those species is beyond the scope of this
work. Heteropolar hemi-tri-syncolpate pollen is a synapomorphy of the clade that includes all
Paullinieae except Thinouia and Lophostigma [21, 97, 98]. Therefore, these fossils can be con-
sidered evidence of crown-group Paullinieae in the fossil record. Unfortunately, constraining
the age of these samples is challenging. Montes et al. [99] reported Late Eocene and Oligocene
foraminifera from the Gatuncillo Formation, consistent with the original age estimate from
Graham [52]. More recently, Ramı´rez et al. [100] obtained detrital zircons from two sites that
constrain the maximum age of deposition of the Gantuncillo Formation to Late Eocene, c. 41
Ma and c. 36 Ma respectively, but we do not know their position relative to Graham’s [52] pol-
len sample. Older putative occurrences of Middle Eocene pollen from the Wagon Bed Forma-
tion in Wyoming [101] and the Claiborne Group in northern Alabama [102] were not
described nor figured, and are not reliable [103]. Pollen from the Kisinger Lakes paleoflora in
Wyoming that MacGinitie compared with Serjania [74] was not described; however, one fig-
ure shows a single grain 24 μm across in polar view with a 3-(parasyncol)porate structure. It is
not possible to determine whether it was heteropolar and pollen grains in Paullinieae are larger
than 30 μm across [98, 104]. Therefore, we do not consider this a reliable fossil occurrence of
Paullinieae based on the available information. Younger occurrences include heteropolar
demisyncolpate pollen from the late Miocene Gatun Formation in Panama [43, 49] and the
Pliocene Paraje Solo Formation, also in Panama [47].
The transition to the liana habit occurred only once in Sapindaceae along the branch lead-
ing to crown-group Paullinieae [21]. Accordingly, all members of the tribe share anatomy
associated with the climbing habit such as abrupt changes in vessel diameter, vessel dimor-
phism, and numerous members have cambial variants [19, 105]. The combination of wood
anatomical characters and the presence of the peripheral vascular strands preserved in the fos-
sils provides strong evidence of the climbing habit in Paullinieae by the early Miocene.
Paleoecology
Lianas are a conspicuous element of tropical forests and their fossils contribute to reconstruc-
tions of paleoenvironments and paleocommunities. The Lirio East fossil assemblage includes
at least 32 plant morphotypes have been distinguished and assigned to family based on fossil
fruits, seeds, and woods [32–34, 36–38]. The discovery of Ampelorhiza brings the number of
liana species to a minimum of 8, or 25% of the local assemblage. This value is typical of lowland
tropical forests [106]. Three other potential liana axes were identified using logistic regression
(S2 Fig), but remain to be described (F. Herrera, pers. comm.). At least 31 additional fruit and
seed morphotypes have been distinguished but not yet identified to family [32]. In modern
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
15 / 22
PLOS ONE Fossil Paullinieae
tropical forests liana species richness is highest in seasonally dry tropical forests and locally
near forest edges or in treefall gaps [107–109]. Given the rarity of distinct growth rings in the
co-occurring fossil woods and the preference of Sacoglottis and Oreomunnea for humid tropi-
cal forests [33, 34], we hypothesize that the high proportion of lianas in the Lirio East assem-
blage is a signal of riparian zone disturbance and/or edge effects in a humid tropical forest on a
landscape shaped by nearby volcanic activity [31].
Conclusion
The discovery of Ampelorhiza reported here is the oldest reliable macrofossil evidence of Paul-
linieae. Fossil leaves from the Eocene of North America previously attributed to Cardiosper-
mum and Serjania likely belong to Sapindaceae, but are not reliable occurrences of Paullinieae.
Our findings support the conclusion that diversification of the tribe was underway by at least
18.5–19 Ma (early Miocene) and that the climbing habit had evolved by that time.
Supporting information
S1 Appendix. Folder containing the accession list, mrbayes infile.nex, mcc, map, allcom-
pat, and accession list.
(ZIP)
S2 Appendix. Revised descriptions of the leaf architecture. Descriptions of Bohlenia ameri-
cana, Bohlenia insignis, “Cardiospermum” coloradensis, “Cardiospermum” terminale, and “Ser-
jania” rara.
(PDF)
S1 Fig. Transverse section of the paratype, UF 19391-63026.
(TIF)
S2 Fig. Plot of lianas and self-supporting woody dicots. Filled points are fossil axes from the
Lirio East site classified as either climbers or self-supporters using logistic regression. We
applied a conservative decision threshold of 0.95 for classifying lianas.
(TIF)
Acknowledgments
We thank Bruce MacFadden, Jonathan Bloch, Steven Manchester, Carlos Jaramillo, and
Fabiany Herrera for support in the early phases of this project, Veronica Angylossy for super-
vision over Carolina Basto’s thesis work concerning the root anatomy of Paullinieae, and Lil-
lian Pearson for making initial peels of the fossil during her PCP-PIRE internship. Fabiany
Herrera discovered the fossil site at Lirio East. We thank Ricardo Martinez for donating the
vehicles used for fieldwork in Panama, and the Autoridad del Canal de Panama (ACP) for
access to the site where the fossils were collected. We also thank the staff of the herbaria at the
University of California Berkeley, the Florida Museum of Natural History, and the University
of Kansas, and the staff of the paleobotany collections at UC Berkeley, the Florida Museum of
Natural History, and the Smithsonian for assistance. We thank Sarah DeWitt for comments
on the figures. Finally, we thank the reviewers for helpful feedback and suggestions during the
review process. Any opinions, findings, conclusions, or recommendations expressed in this
article are those of the authors and do not necessarily reflect the views of the NSF.
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
16 / 22
PLOS ONE Fossil Paullinieae
Author Contributions
Conceptualization: Nathan A. Jud, Joyce G. Chery.
Data curation: Nathan A. Jud, Joyce G. Chery.
Formal analysis: Nathan A. Jud, Joyce G. Chery.
Investigation: Nathan A. Jud, Sarah E. Allen, Chris W. Nelson, Carolina L. Bastos, Joyce G.
Chery.
Methodology: Nathan A. Jud, Joyce G. Chery.
Resources: Carolina L. Bastos.
Supervision: Nathan A. Jud.
Writing – original draft: Nathan A. Jud, Sarah E. Allen, Chris W. Nelson, Joyce G. Chery.
Writing – review & editing: Nathan A. Jud, Sarah E. Allen, Joyce G. Chery.
References
1. Acevedo-Rodrı´guez P, Van Welzen PC, Adema F, Van Der Ham RWJM. Flowering Plants. Eudicots.
Sapindaceae. Springer p. 357–407. 2011.
2. Harrington MG, Edwards KJ, Johnson SA, Chase MW, Gadek PA. Phylogenetic Inference in Sapinda-
ceae sensu lato Using Plastid matK and rbcL DNA Sequences. Systematic Botany. 2005; 30(2):366–
382. https://doi.org/10.1600/0363644054223549
3. Buerki S, Forest F, Acevedo-Rodrı´guez P, Callmander MW, Nylander JAA, Harrington M, et al. Plastid
and nuclear DNA markers reveal intricate relationships at subfamilial and tribal levels in the soapberry
family (Sapindaceae). Molecular Phylogenetics and Evolution. 2009; 51(2):238–258. https://doi.org/
10.1016/j.ympev.2009.01.012 PMID: 19405193
4. Buerki S, Lowry PP II, Alvarez N, Razafimandimbison SG, Ku¨ pfer P, Callmander MW. Phylogeny and
circumscription of Sapindaceae revisited: molecular sequence data, morphology and biogeography
support recognition of a new family, Xanthoceraceae. Plant Ecology and Evolution. 2010; 143(2):148–
159. https://doi.org/10.5091/plecevo.2010.437
5.
Tamaio N, Somner GV. Development of corded vascular cylinder in Thinouia restingae Ferruci & Som-
ner (Sapindaceae: Paullinieae). The Journal of the Torrey Botanical Society. 2010; 137(4):319–326.
https://doi.org/10.3159/10-RA-047.1
6. Schenck H. Beitra¨ ge zur Biologie und Anatomie der Lianen im besonderen der in Brasilien einhei-
mischen Arten. 4-5. G. Fischer; 1892. f
7. Radlkofer L. Sapindaceae In: Engler A. & Prantl K. Die Natu¨ rlichen Pflazenfamilien. 1895; 3:277–366.
8. Pfeiffer HH. Das abnorme Dickenwachstum. vol. 9. Gebr. Borntraeger; 1926.
9. Radlkofer L, Engler A, Diels L. Sapindaceae. Engelmann 1933.
10.
Johnson MA, Truscott FH. On the Anatomy of Serjania. I. Path of the Bundles. American Journal of
Botany. 1956; 43(7):509–519. https://doi.org/10.1002/j.1537-2197.1956.tb10525.x
11. Obaton M. Les Lianes ligneuses a structure anormale des forêts denses d’Afrique occidentale.
Annales de scinences naturelles (Botanique) Ser. 12. 1960; 1:1–220
12. Van der Walt JJA, Van der Schijff HP, Schweickerdt HG. Anomalous secondary growth in the stem of
lianas Mikania cordata (Burm. F.) Robins (Compositae) and Paullinia pinnata Linn.(Sapindaceae). Kir-
kia. 1973; 9:109–138.
13. Klaassen R. Wood Anatomy of the Sapindaceae. IAWA J. suppl 2. International Association of Wood
Anatomists, Leiden; 1999.
14.
Tamaio N, Neves MF, Brandes AFN, Vieira RC. Quantitative analyses establish the central vascular
cylinder as the standard for wood-anatomy studies in lianas having compound stems (Paullinieae:
Sapindaceae). Flora—Morphology, Distribution, Functional Ecology of Plants. 2011; 206(11):987–
996. https://doi.org/10.1016/j.flora.2011.07.006
15.
Tamaio N. Wood anatomy of lianas of Sapindaceae commercially used in São Paulo-SP. Cerne.
2011; 17(4):533–540.
16. Bastos CL, Tamaio N, Angyalossy V. Unravelling roots of lianas: a case study in Sapindaceae. Annals
of Botany. 2016; 118(4):733–746. https://doi.org/10.1093/aob/mcw091
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
17 / 22
PLOS ONE Fossil Paullinieae
17.
Lopes WAL, De Souza LA, De Almeida OJG. Procambial and cambial variants in Serjania and Urvillea
species (Sapindaceae: Paullinieae). Journal of the Botanical Research Institute of Texas. 2017; 11
(2):421–432.
18. Cunha Neto ILD, Martins FM, Somner GV, Tamaio N. Successive cambia in liana stems of Paullinieae
and their evolutionary significance in Sapindaceae. Botanical Journal of the Linnean Society. 2018;
186(1):66–88. https://doi.org/10.1093/botlinnean/box080
19. Chery JG, Pace MR, Acevedo-Rodrı´guez P, Specht CD, Rothfels CJ. Modifications during early plant
development promote the evolution of nature’s most complex woods. Current Biology. 2020; 30
(2):237–244. https://doi.org/10.1016/j.cub.2019.11.003
20. Acevedo-Rodriguez P. A revision of Lophostigma (Sapindaceae). Systematic Botany. 1993; 18
(3):379–388. https://doi.org/10.2307/2419414
21. Acevedo-Rodrı´guez P, Wurdack KJ, Ferrucci MS, Johnson G, Dias P, Coelho RG, et al. Generic rela-
tionships and classification of tribe Paullinieae (Sapindaceae) with a new concept of supertribe Paulli-
niodae. Systematic Botany. 2017; 42(1):96–114. https://doi.org/10.1600/036364417X694926
22. Chery JG, Acevedo-Rodrı´guez P, Rothfels CJ, Specht CD. Phylogeny of Paullinia L. (Paullinieae:
Sapindaceae), a diverse genus of lianas with dynamic fruit evolution. Molecular Phylogenetics and
Evolution. 2019; 140:106577. https://doi.org/10.1016/j.ympev.2019.106577
23. Buerki S, Forest F, Alvarez N, Nylander JAA, Arrigo N, Sanmartı´n I. An evaluation of new parsimony-
based versus parametric inference methods in biogeography: a case study using the globally distrib-
uted plant family Sapindaceae. Journal of Biogeography. 2011; 38(3):531–550. https://doi.org/10.
1111/j.1365-2699.2010.02432.x
24. Buerki S, Forest F, Stadler T, Alvarez N. The abrupt climate change at the Eocene–Oligocene bound-
ary and the emergence of South-East Asia triggered the spread of sapindaceous lineages. Annals of
Botany. 2013; 112(1):151–160. https://doi.org/10.1093/aob/mct106
25. Muellner-Riehl AN, Weeks A, Clayton JW, Buerki S, Nauheimer L, Chiang YC, et al. Molecular phylo-
genetics and molecular clock dating of Sapindales based on plastid rbcL, atpB and trnL-trnF DNA
sequences. Taxon. 2016; 65(5):1019–1036. https://doi.org/10.12705/655.5
26. Erwin DM, Stockey RA. Sapindaceous flowers from the Middle Eocene Princeton chert (Allenby For-
mation) of British Columbia, Canada. Canadian Journal of Botany. 1990; 68(9):2025–2034. https://doi.
org/10.1139/b90-265
27. Gandolfo MA, Nixon KC, Crepet WL. Selection of fossils for calibration of molecular dating models.
Annals of the Missouri Botanical Garden. 2008; 95(1):34–42. https://doi.org/10.3417/2007064
28. Retallack GJ, Kirby MX. Middle Miocene global change and paleogeography of Panama. Palaios.
2007; 22(6):667–679. https://doi.org/10.2110/palo.2006.p06-130r
29. Kirby MX, Jones DS, MacFadden BJ. Lower Miocene stratigraphy along the Panama Canal and its
bearing on the Central American Peninsula PLoS One. 2008; 3(7):e2791 https://doi.org/10.1371/
journal.pone.0002791 PMID: 18665219
30. MacFadden BJ, Bloch JI, Evans H, Foster DA, Morgan GS, Rincon A, et al. Temporal calibration and
biochronology of the Centenario Fauna, early Miocene of Panama. The Journal of Geology. 2014; 122
(2):113–135. https://doi.org/10.1086/675244
31. Buchs DM, Irving D, Coombs H, Miranda R, Wang J, Coronado M, et al. Volcanic contribution to emer-
gence of Central Panama in the Early Miocene. Scientific reports. 2019; 9(1):1–16. https://doi.org/10.
1038/s41598-018-37790-2 PMID: 30723222
32. Herrera F. Revealing the floristic and biogeographic composition of Paleocene to Miocene Neotropical
forests. [PhD Thesis] University of Florida, Gainesville, FL; 2014.
33. Herrera F, Manchester SR, Jaramillo CA, MacFadden BJ, da Silva-Caminha SA. Phytogeographic his-
tory and phylogeny of the Humiriaceae. International Journal of Plant Sciences. 2010; 171(4):392–
408. https://doi.org/10.1086/651229
34. Herrera F, Manchester SR, Koll R, Jaramillo CA Fruits of Oreomunnea (Juglandaceae) in the early
Miocene of Panama Paleobotany and biogeography: A Festschrift for Alan Graham in his 80th year:
Monographs in systematic botany from the Missouri Botanical Garden. 2014; 128:124–133.
35.
Jud NA, Nelson CW, Herrera F. Fruits and wood of Parinari from the early Miocene of Panama and the
fossil record of Chrysobalanaceae. American Journal of Botany. 2016; 103(2):277–289. https://doi.
org/10.3732/ajb.1500425
36. Nelson CW, Jud NA Biogeographic implications of Mammea paramericana sp. nov. from the lower
Miocene of Panama and the evolution of Calophyllaceae International Journal of Plant Sciences.
2017; 178(3):241–257. https://doi.org/10.1086/689618
37.
Jud NA, Nelson CW. A liana from the lower Miocene of Panama and the fossil record of Connaraceae.
American Journal of Botany. 2017; 104(5):685–693. https://doi.org/10.3732/ajb.1700080
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
18 / 22
PLOS ONE Fossil Paullinieae
38. Herrera F, Carvalho MR, Jaramillo CA, Manchester SR. 19-Million-Year-Old Spondioid Fruits from
Panama Reveal a Dynamic Dispersal History for Anacardiaceae. International Journal of Plant Sci-
ences. 2019; 180(6):479–492. https://doi.org/10.1086/703551
39. Berry EW. Contributions to the geology and paleontology of the Canal Zone, Panama and geologically
related areas in Central America and the West Indies United States National Museum Bulletin. 1918;
103:15–44.
40. Rodrı´guez-Reyes O, Falcon-Lang HJ, Gasson P, Collinson ME, Jaramillo CA. Fossil woods (Malva-
ceae) from the lower Miocene (early to mid-Burdigalian) part of the Cucaracha Formation of Panama
(Central America) and their biogeographic implications. Review of Palaeobotany and Palynology
2014; 209:11–34. https://doi.org/10.1016/j.revpalbo.2014.05.006
41. Rodrı´guez-Reyes O, Gasson P, Falcon-Lang HJ, Collinson ME. Fossil legume woods of the Prioria-
clade (subfamily Detarioideae) from the lower Miocene (early to mid-Burdigalian) part of the Cucara-
cha Formation of Panama (Central America) and their systematic and palaeoecological implications
Review of Palaeobotany and Palynology. 2017; 246:44–61. https://doi.org/10.1016/j.revpalbo.2017.
06.005
42. Rodrı´guez-Reyes O, Gasson P, Thornton C, Falcon-Lang HJ, Jud NA. Panascleroticoxylon crystallosa
gen. et sp. nov.: a new Miocene malpighialean tree from Panama. IAWA Journal. 2017; 38(4):437–
455. https://doi.org/10.1163/22941932-20170178
43.
44.
Jaramillo CA, Moreno E, Ramı´rez V, da Silva-Caminha SAF, de la Barrera A, de la Barrera A, et al.
Palynological record of the last 20 million years in Panama. Paleobotany and biogeography: A fest-
schrift for Alan Graham in his 80th year. 2014; 128:134–251.
Joy KW, Willis AJ, Lacey WS. A rapid cellulose peel technique in palaeobotany. Annals of Botany.
1956; 20(80):635–637.
45. Enquist BJ, Condit R, Peet RK, Schildhauer M, Thiers BM. Cyberinfrastructure for an integrated botan-
ical information network to investigate the ecological impacts of global climate change on plant biodi-
versity. PeerJ Inc.; 2016. e2615v2. Available from: https://peerj.com/preprints/2615.
46. Maitner BS, Boyle B, Casler N, Condit R, Donoghue J, Dura´ n SM, et al. The bien r package: A tool to
access the Botanical Information and Ecology Network (BIEN) database. Methods in Ecology and
Evolution. 2018; 9(2):373–379. https://doi.org/10.1111/2041-210X.12861
47. Graham A. Studies in neotropical paleobotany. II. The Miocene communities of Veracruz, Mexico.
Annals of the Missouri Botanical Garden. 1976; 63(4):787–842. https://doi.org/10.2307/2395250
48. Graham A. Tropical American Tertiary floras and paleoenvironments: Mexico, Costa Rica, and Pan-
ama. American Journal of Botany. 1987; 74(10):1519–1531. https://doi.org/10.1002/j.1537-2197.
1987.tb12143.x
49. Graham A. Studies in Neotropical Paleobotany. X. The Pliocene Communities of Panama–Composi-
tion, Numerical Representations, and Paleocommunity Paleoenvironmental Reconstructions. Annals
of the Missouri Botanical Garden. 1991; 78(2):465–475. https://doi.org/10.2307/2399574
50. Chavez RP, Rzedowski J. Estudio palinolo´gico de las floras fo´ siles del Mioceno inferior y principios del
Mioceno medio de la regio´n de Pichucalco, Chiapas, Me´xico. Acta Bota´ nica Mexicana. 1993; 24:1–96.
51. Hendy AJW. Spatial and stratigraphic variation of marine paleoenvironments in the middle-upper Mio-
cene Gatun Formation, Isthmus of Panama. PALAIOS. 2013; 28(4):210–227. https://doi.org/10.2110/
palo.2012.p12-024r
52. Graham A. Studies in Neotropical Paleobotany. IV. The Eocene communities of Panama. Annals of
the Missouri Botanical Garden. 1985; 72(3):504–534. https://doi.org/10.2307/2399101
53. GBIF.org [11 January 2021] GBIF Occurrence Download https://doi.org/10.15468/dl.7yscgt
54. Gildenhuys E, Ellis AG, Carroll SP, Le Roux JJ. The ecology, biogeography, history and future of two
globally important weeds: Cardiospermum halicacabum Linn. and C. grandiflorum NeoBiota. 2013;
19:45–65. https://doi.org/10.3897/neobiota.19.5279
55. Gildenhuys E, Ellis AG, Carroll SP, Le Roux JJ. Combining natal range distributions and phylogeny to
resolve biogeographic uncertainties in balloon vines (Cardiospermum, Sapindaceae). Diversity and
Distributions. 2015; 21(2):163–174. https://doi.org/10.1111/ddi.12261
56. Wheeler EA, Baas P, Gasson PE, et al. IAWA list of microscopic features for hardwood identification.
IAWA Journal. 1989.
57. Carlquist S. Wood anatomy of Nepenthaceae. Bulletin of the Torrey Botanical Club. 1981; 108
(3):324–330. https://doi.org/10.2307/2484711
58. Carlquist S. Observations on Functional Wood Histology of Vines and Lianas. Aliso: A Journal of Sys-
tematic and Evolutionary Botany. 1985; 11(2):139–157. https://doi.org/10.5642/aliso.19851102.03
59. Carlquist S. Anatomy of vine and liana stems. In: The Biology of Vines. Cambridge University Press;
1991. p. 53–72.
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
19 / 22
PLOS ONE Fossil Paullinieae
60. Scrucca L, Fop M, Murphy TB, Raftery AE. mclust 5: clustering, classification and density estimation
using Gaussian finite mixture models. The R journal. 2016; 8(1):289. https://doi.org/10.32614/RJ-
2016-021
61. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria; 2020.
Available from: https://www.R-project.org/.
62. Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of image analysis. Nature
methods. 2012; 9(7):671–675. https://doi.org/10.1038/nmeth.2089
63. Ellis B, Daly DC, Hickey LJ, Mitchell JD, Johnson KR, Wilf P, et al. Manual of Leaf Architecture. First
edition ed. Ithaca, NY: Cornell University Press; 2009.
64. Chery JG, da Cunha Neto IL, Pace MR, Acevedo-Rodrı´guez P, Specht CD, Rothfels CJ. Wood anat-
omy of the neotropical liana lineage Paullinia L. (Sapindaceae). IAWA Journal. 2020; 41(3):278–300.
https://doi.org/10.1163/22941932-bja10027
65. Patel RN. A comparison of the anatomy of the secondary xylem in roots and stems. Holzforschung-
International Journal of the Biology, Chemistry, Physics and Technology of Wood. 1965; 19(3):72–79.
66. Bastos CL. Decifrando raı´zes e caules de lianas de Sapindaceae: diversidade macrosco´ pica e con-
exões vasculares [PhD Thesis]. Universidade de São Paulo; 2015.
67. Ronquist F, Teslenko M, Van Der Mark P, Ayres DL, Darling A, Ho¨ hna S, et al. MrBayes 3.2: efficient
Bayesian phylogenetic inference and model choice across a large model space. Systematic biology.
2012; 61(3)539–542. https://doi.org/10.1093/sysbio/sys029 PMID: 22357727
68.
Letunic I, Bork P. Interactive Tree Of Life (iTOL) v4: recent updates and new developments. Nucleic
Acids Research. 2019; 47(W1):W256–W259. https://doi.org/10.1093/nar/gkz239
69. Bouckaert R, Heled J, Ku¨ hnert D, Vaughan T, Wu C-H, Xie D, et al. BEAST 2: a software platform for
Bayesian evolutionary analysis. PLoS Comput Biol. 2014; 10(4):e1003537. https://doi.org/10.1371/
journal.pcbi.1003537 PMID: 24722319
70. Ho¨ hna S, Landis MJ, Heath TA, Boussau B, Lartillot N, Moore BR, et al. RevBayes: Bayesian phyloge-
netic inference using graphical models and an interactive model-specification language. Systematic
biology. 2016; 65(4):726–736. https://doi.org/10.1093/sysbio/syw021 PMID: 27235697
71. Ewers FW, Fisher JB, Chiu S-T. A survey of vessel dimensions in stems of tropical lianas and other
growth forms. Oecologia 1990; 84(4):544–552. https://doi.org/10.1007/BF00328172
72. Ewers FW, Carlton MR, Fisher JB, Kolb KJ, Tyree MT. Vessel diameters in roots versus stems of tropi-
cal lianas and other growth forms. IAWA Journal. 1997; 18(3):261–279. https://doi.org/10.1163/
22941932-90001490
73. Rosell JA, Olson ME. Do lianas really have wide vessels? Vessel diameter–stem length scaling in non-
self-supporting plants. Perspectives in Plant Ecology, Evolution and Systematics. 2014; 16(6):288–
295. https://doi.org/10.1016/j.ppees.2014.08.001
74. MacGinitie HD. An early middle Eocene flora from the Yellowstone-Absaroka volcanic province, north-
western Wind River Basin, Wyoming. First edition ed. Berkeley: University of California Press; 1974.
75. Allen SE. The uppermost Lower Eocene Blue Rim flora from the Bridger Formation of Southwestern
Wyoming: Floristic Composition, Paleoclimate, and Paleoecology [PhD Thesis]. University of Florida.
Gainesville, FL; 2017.
76.
Lesquereux L. Contributions to the flora of the Western Territories III. The Cretaceous and Tertiary flo-
ras. Rept US Geol Surv Terr. 1883; 8:1–283.
77. MacGinitie HD. Fossil plants of the Florissant beds, Colorado. vol. 599 of Carnegie Institution of Wash-
ington Publication. Carnegie Institution of Washington; 1953.
78. Becker HF. Oligocene plants from the upper Ruby River Basin, southwestern Montana. vol. 82. Geo-
logical Society of America; 1961.
79.
Lielke K, Manchester SR, Meyer H, and others. Reconstructing the environment of the northern Rocky
Mountains during the Eocene/Oligocene transition: constraints from the palaeobotany and geology of
south-western Montana, USA. Acta palaeobotanica. 2012; 52(2):317–358.
80. Knowlton FH. Revision of the flora of the Green River formation, with descriptions of new species;
1923. 131–F. Available from: http://pubs.er.usgs.gov/publication/pp131F.
81. MacGinitie HD. The Eocene Green River flora of northwestern Colorado and northeastern Utah. Uni-
versity of California Press; 1969.
82. Brown RW. Additions to the flora of the Green River Formation. US Geological Survey Professional
Paper. 1929;154:279–292.
83. Brown RW. The recognizable species of the Green River flora. United States. Government Printing
Office. 1934.
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
20 / 22
PLOS ONE Fossil Paullinieae
84.
Johnson KR, Plumb C. Common plant fossils from the Green River Formation at Douglas Pass, Colo-
rado, and Bonanza, Utah. In: The Green River Piceance Creek and eastern Unita Basin: Grand Junc-
tion Colorado. Grand Junction Geological Society. 1995; p. 121–130.
85. Wolfe JA, Wehr W. Middle Eocene dicotyledonous plants from Republic, northeastern Washington.
US Geological Survey Bulletin 1597. 1987;p. 1–25.
86. McClain AM, Manchester SR. Dipteronia (Sapindaceae) from the Tertiary of North America and impli-
cations for the phytogeographic history of the Aceroideae. American Journal of Botany. 2001; 88
(7):1316–1325. https://doi.org/10.2307/3558343
87.
Fittipaldi FC, Simões MG, Giulietti AM, Pirani JR. Fossil plants from the Itaquaquecetuba Formation
(Cenozoic of the São Paulo Basin) and their possible paleoclimatic significance. Boletim IG-USP Publi-
cac¸ão Especial. 1989; 7:183–203.
88. Dos Santos D, Garcia M, Saad A, Bistrichi C. Itaquaquecetuba Formation Palynostratigraphy, São
Paulo Basin, Brazil. Revista Brasileira de Paleontologia. 2010; 13:205–220.
89. Duarte L, Pimentel de Rezende-Martins A. Contribuı´c¸ão ao conhecimento da flora Cenozo´ica do Bra-
sil: Jazigo Vargem Grande do Sul, SP: se´ rie taubate´ . I. Anais da Academia Brasileira de Ciências.
1983; 55(1):109–121.
90. Dos Santos MA, Bernardes-de Oliveira MEC. Taxonomia da tafoflora Neogena da Formac¸ão Rio
Claro, Jaguariuna, Estado de Sao Paulo, Brasil. Revista Brasileira de Paleontolgia. 2013; 16(3):465–
486. https://doi.org/10.4072/rbp.2013.3.07
91. Edwards WN, Wonnacott F. Sapindaceae. In: Fossilum Catalogus 2, Plantae. pars. 14. W. Junk; 1928
p. 1–84.
92. Cunha Neto IL, Martins FM, Somner GV, Tamaio N. Secretory structures in stems of five lianas of
Paullinieae (Sapindaceae): morphology and histochemistry Flora. 2017; 235:29–40
93. Pellissari LCO, Barros CF, Medeiros H, Tamaio N. Cambial patterns of Paullinia (Sapindaceae) in
southwestern Amazonia, Brazil. Flora. 2018; 246-247:71–82. https://doi.org/10.1016/j.flora.2018.07.
002
94. Metcalfe CR, Chalk L. Anatomy of the Dicotyledons, Vols. 1 & 2. Clarendon Press; 1950.
95. Manchester SR. Update on the megafossil flora of Florissant, Colorado. Denver Museum of Nature
and Science. 2001; 4:137–161.
96.
Jaramillo C, Rueda MJ. A Morphological Electronic Database of Cretaceous-Tertiary and Extant Pol-
len and Spores from Northern South America, v. 2020. Smithsonian Tropical Research Institute. http://
biogeodb.stri.si.edu/jaramillosdb/web/morphological/ (accessed 18 Jan 2021)
97.
Ferrucci MS, Anzotegui LM. El polen de Paullinieae (Sapindaceae). Bonplandia. 1993;p. 211–243.
98. Van der Ham R, Tomlik A. Serjania pollen and the origin of the tribe Paullinieae (Sapindaceae).
Review of Palaeobotany and Palynology. 1994; 83(1-3):43–53. https://doi.org/10.1016/0034-6667(94)
90056-6
99. Montes C, Cardona A, McFadden R, Moro´ n SE, Silva CA, Restrepo-Moreno S, et al. Evidence for mid-
dle Eocene and younger land emergence in central Panama: Implications for Isthmus closure. Geolog-
ical Society of America Bulletin. 2012; 124(5-6):780–799. https://doi.org/10.1130/B30528.1
100. Ramı´rez DA, Foster DA, Min K, Montes C, Cardona A, Sadove G. Exhumation of the Panama base-
ment complex and basins: Implications for the closure of the Central American seaway. Geochemistry,
Geophysics, Geosystems. 2016; 17(5):1758–1777. https://doi.org/10.1002/2016GC006289
101.
Leopold EB, MacGinitie HD. Development and affinities of Tertiary floras in the Rocky Mountains. In:
Floristics and Paleofloristics of Asia and Eastern North America. Elsevier Publishing Company; 1972.
p. 147–200.
102. Gray J. Temperate pollen genera in the Eocene (Claiborne) flora, Alabama. Science. 1960; 132
(3430):808–810. https://doi.org/10.1126/science.132.3430.808
103. Muller J. Fossil pollen records of extant angiosperms. The Botanical Review. 1981; 47(1):1–142.
https://doi.org/10.1007/BF02860537
104. Bellonzi TK, Dutra FV, de Souza CN, Rezende AA, Gasparino EC. Pollen types of Sapindaceae from
Brazilian forest fragments: apertural variation. Acta Botanica Brasilica. 2020; 34(2):327–341. https://
doi.org/10.1590/0102-33062020abb0022
105. Carlquist S. Comparative Wood Anatomy: Systematic, Ecological, and Evolutionary Aspects of Dicoty-
ledon Wood. Springer Science & Business Media; 2001.
106. Gentry AH. Distribution and evolution of climbing plants. In: The Biology of Vines. Cambridge Univer-
sity Press; 1991. p. 3–49.
107. Schnitzer SA, Bongers F. The ecology of lianas and their role in forests. Trends in Ecology & Evolution.
2002; 17(5):223–230. https://doi.org/10.1016/S0169-5347(02)02491-6
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
21 / 22
PLOS ONE Fossil Paullinieae
108. Schnitzer SA. A mechanistic explanation for global patterns of liana abundance and distribution. The
American Naturalist. 2005; 166(2):262–276. https://doi.org/10.1086/431250
109. Van Der Heijden GMF, Phillips OL. Environmental effects on Neotropical liana species richness. Jour-
nal of Biogeography. 2009; 36(8):1561–1572. https://doi.org/10.1111/j.1365-2699.2009.02099.x
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
22 / 22
PLOS ONE
| null |
10.1371_journal.pone.0298960.pdf
|
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
|
All relevant data are within the manuscript and its Supporting Information files.
|
RESEARCH ARTICLE
Prevalence and associated factors of
refractive error among adults in South
Ethiopia, a community-based cross-sectional
study
Marshet Gete Abebe1☯, Abiy Maru Alemayehu2☯, Minychil Bantihun Munaw2☯, Mikias
Mered Tilahun2☯, Henok Biruk AlemayehuID
1☯*
1 Department of Ophthalmology and Optometry, Hawassa University, Comprehensive Specialized Hospital,
Hawassa, Ethiopia, 2 Department of Optometry, School of Medicine, University of Gondar, Comprehensive
Specialized Hospital, Gondar, Ethiopia
☯ These authors contributed equally to this work.
* [email protected]
Abstract
Introduction
The increasing prevalence of refractive error has become a serious health issue that needs
serious attention. However, there are few studies regarding the prevalence and associated
factors of refractive error at the community level in Ethiopia as well as in the study area.
Therefore, providing updated data is crucial to reduce the burdens of refractive error in the
community.
Objective
To assess the prevalence and associated factors of refractive error among adults in
Hawassa City, South Ethiopia, 2023.
Method
A community-based cross-sectional study was conducted on 951 adults using a multistage
sampling technique from May 8 to June 8, 2023, in Hawassa City, South Ethiopia. A pre-
tested, structured questionnaire combined with an ocular examination and a refraction pro-
cedure was used to collect data. The collected data from the Kobo Toolbox was exported to
a statistical package for social sciences for analysis. Binary and multivariable logistic regres-
sion analyses were performed. A P-value of less than 0.05 was considered statistically sig-
nificant in the multivariable analysis.
Result
A total of 894 study participants were involved in this study with a 94.1% response rate. The
prevalence of refractive error was 12.3% (95% CI: 10.2, 14.5%). Regular use of electronic
devices (adjusted odds ratio = 3.64, 95% CI: 2.25, 5.91), being diabetic (adjusted odds ratio
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Abebe MG, Alemayehu AM, Munaw MB,
Tilahun MM, Alemayehu HB (2024) Prevalence and
associated factors of refractive error among adults
in South Ethiopia, a community-based cross-
sectional study. PLoS ONE 19(3): e0298960.
https://doi.org/10.1371/journal.pone.0298960
Editor: Fidan Aghayeva, Chiemsee Augen
Tagesklinik, Technical University of Munich,
GERMANY
Received: August 31, 2023
Accepted: February 1, 2024
Published: March 25, 2024
Copyright: © 2024 Abebe et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: he author(s) received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
1 / 14
PLOS ONE Prevalence and associated factors of refractive error among adults in South Ethiopia
= 4.02, 95% CI: 2.16, 7.48), positive family history of refractive error (adjusted odds ratio =
2.71, 95% CI 1.59, 4.61) and positive history of cataract surgery (adjusted odds ratio = 5.17,
95% CI 2.19, 12.4) were significantly associated with refractive error.
Conclusion and recommendation
The overall magnitude of refractive error in our study area was high. Regular use of elec-
tronic devices, being diabetic, positive family history of refractive error, and a positive history
of cataract surgery were associated with refractive error.
Introduction
Refractive error (RE) is a condition where the optical system of the eye fails to focus parallel
rays of light on the retina. The RE occurs when there is an imbalance between the axial length
and the refractive power of the eye [1]. Symptoms of RE include blurring of vision, headaches,
eyestrain, and problems with focusing and seeing details at any distance. Globally, the preva-
lence of RE was 12% [2]. The prevalence of RE ranges from 6% to 72% in developed countries
[3, 4]. In Sub-Saharan Africa, the prevalence of RE was approximately 46% [5, 6]. Hospital-
based studies conducted in Gondar, Borumeda, and Arba Minch, Ethiopia showed that the
prevalence of RE was 76.3%, 18.3%, and 27.5% respectively [7–9].
Globally, 2.2 billion people suffer from visual impairment (VI), and RE accounts for 88.4
million cases [10]. RE is the most common cause of visual impairment worldwide. Around
50% of the world’s vision impairment and blindness caused by RE are found in Asia [11].
According to Ethiopian national surveys, RE accounts for 33.4% of low vision and is the sec-
ond leading cause of VI after cataracts [12]. RE can undermine individual performance,
reduce social participation, and reduce employability. RE can also increase the economic
burden on the country. Approximately US$202 billion is attributed to VI due to uncor-
rected RE [13]. Those above conditions result in a reduced quality of life for individuals
with RE [11]. Among the top 20 causes of disability-adjusted life years, RE is one of the four
non-fatal disorders [14].
Some of the factors, such as age, educational level, history of cataract surgery, family history
of RE, and history of diabetes mellitus were associated with the development of RE, as reported
by studies [15, 16]. Although RE cannot be completely prevented, it can be treated easily. RE
can be treated with spectacle, contact lens, or refractive surgery [17].
To address the issue, multi-tiered points of delivery for refractive care services and optical
dispensing units were established, together with highly qualified optometry personnel [18].
Ethiopia launched the Vision 2020 global initiative to develop a comprehensive and sustain-
able eye care system that will eliminate the major causes of avoidable blindness [19].
The increasing prevalence of RE in both developed and developing nations remains an
urgent public health problem that needs serious attention [10, 11]. Although RE is prevalent
across the world, there is limited evidence on the burden and predictors of RE among adults at
the community level in Ethiopia. Hence, conducting the prevalence and associated factors
gives updated information that contributes to reducing the burden of RE. In addition, this
study can be used as baseline information for policymakers, the Ministry of Health, and other
researchers to allocate resources for eye care service delivery.
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
2 / 14
PLOS ONE Prevalence and associated factors of refractive error among adults in South Ethiopia
Method and materials
Study design
A community-based cross-sectional study was conducted.
Study area and period
The study was conducted in Hawassa City, South Ethiopia from May 8, 2023, to June 8, 2023.
Hawassa is the capital city of the Southern Nations, Nationalities, and Peoples Region as well
as the Sidama Regional State. It is located 273 kilometers (170 miles) south of Addis Ababa.
According to the Ethiopian National Housing and Census Statistical Agency, the population
of Hawassa city administration is expected to be 403,025 people, and out of this, 266,331 peo-
ple live in the urban with an estimated household of 63,412 [20]. There are 20 kebeles (The
smallest administrative unit of Ethiopia, contained within a woreda) in the city. Five govern-
ment health centers and four hospitals are found in Hawassa City. In general, there are four
private eye clinics and one comprehensive, specialized hospital that provides a comprehensive
eye care service that serves more than 16 million people in the catchment area. In addition,
there is one general hospital that provides eye care services.
Source and study population
All adults who lived in Hawassa City were the source population and all adults aged �18 years
who lived for at least 6 months in households of selected kebeles in Hawassa City were the
study population.
Inclusion and exclusion criteria
All adults aged �18 years who lived for at least 6 months in households of selected kebeles in
Hawassa city were included in the study and adults aged �18 years with ocular comorbidities
(like corneal opacity, and active eye infection) that obscure retinoscopy reflex during the
refraction, adults aged �18 years with an absolute blind eye, adults aged �18 years who were
unable to respond due to serious illness, and mental illness were excluded from the study.
Sample size and sampling procedure
Sample size determination. A single population proportion formula was used by consid-
ering the following assumptions:
n ¼
ðZa=2Þ2Pð1 (cid:0) PÞ
d2
Where;
n = sample size
Z = Value of z statistic at 95% confidence interval = 1.96
α (level of significance) = 5%
P = proportion of RE from a study in Eritrea 6.4% [21]
d = allowable maximum margin of error 2%
Sample size ¼
3:84 � 0:064 � 0:936
0:022
¼ 576
Design effect = 1.5 and Non response rate = 10%
The final sample size was 951
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
3 / 14
PLOS ONE Prevalence and associated factors of refractive error among adults in South Ethiopia
Fig 1. Schematic presentation of sampling technique and procedures for prevalence and associated factors of refractive error
among adults in Hawassa City, South Ethiopia, 2023.
https://doi.org/10.1371/journal.pone.0298960.g001
Sampling technique and procedure.
In Hawassa city, there are 20 kebeles. A multistage
sampling technique was employed to select a representative sample from the city. The list of
the total of kebeles was obtained from the Hawassa city administration. The four kebeles were
chosen by lottery using simple random sampling. The selected four kebeles contained 12,363
of the city’s total households (63,412). The appropriate household was then picked by system-
atic random sampling with a K interval after the sample size was proportionally assigned based
on the household size of each selected kebele Fig 1. The K interval was calculated by dividing
the number of total households in the selected kebele by the total sample size (i.e., 12,363 / 951;
K = 13).
Then, at random, we chose a number between 1 and 13 to choose the first family to be
included in the sample, and every 13th household was included after that. For families with
more than one person eligible for the study, a lottery approach was used to choose study par-
ticipants. When the eligible individual was not present at the time of data collection, the resi-
dence was revisited twice. When there were no eligible persons who met the inclusion criteria
in the selected household, a household listed immediately was evaluated.
Operational definitions
RE was defined as a spherical equivalent of > +0.50 or < -0.50 diopter in either eye on subjec-
tive refraction. Myopia was defined as a spherical equivalent of < -0.50 D. High myopia was
defined as a spherical equivalent of > -6.00 D [22]. Hyperopia was defined as a spherical equiv-
alent of > +0.50 D. Astigmatism was defined as cylinder power > 0.50 D, without taking the
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
4 / 14
PLOS ONE Prevalence and associated factors of refractive error among adults in South Ethiopia
direction of the axis into account [23]. Smoking was defined as those who smoked one stick of
cigarette within the last month [24]. Sleeping Duration was defined as a longer duration when
an individual sleeps for 6 hours or more and a short duration when an individual sleeps for
less than 6 hours [25]. History of cataract surgery was defined as the examiner, facing the
patient, shining the light source on the patient’s eye to see Purkinje’s reflexes like small shining
bubbles. Regular use of electronic devices was defined as using mobile phones or computers,
and other electronic devices at least once a day for at least two hours [26]. Family history of RE
was defined as a family member (mother, father, brother, and sister) of RE diagnosed by pro-
fessionals or any spectacle use [27]. History of diabetes mellitus and hypertension was defined
if the individual has/had a diagnosed diabetic mellitus/ hypertension or undergoing anti-dia-
betes mellitus/antihypertensive treatment [28].
Data collection tools, procedures, and quality control
Data collection tools, procedures.
In this study, data were collected in three sections
which were face-to-face interviews, ocular examinations, and refraction procedures. The data
were collected by five qualified and well-trained Optometrists. A brief explanation of the pur-
pose of the study was provided then verbal informed consent was obtained before collecting
the information. An electronic data collection tool called Kobo Toolbox version 2022.4.4 was
used to collect the data. A pre-tested and semi-structured interviewer-administered question-
naire adapted from previous studies [9, 29, 30] was used to conduct the data collection. The
questionnaires consist of several questions to assess socio-demographic characteristics, behav-
ioral factors, systemic co-morbidity, and clinical factors (S1 File). One supervisor (MSc in
Clinical Optometry) from Hawassa University supervises the data collector every day during
the data collection time.
Ophthalmic examination. Following the interview, all study participants received an
ophthalmic examination and refraction. Optometrists performed ophthalmic examinations,
which began with a VA test. Monocular and binocular unaided VA, and VA after refractive
correction were measured using reduced Snellen acuity charts measured at 3 meters under
normal illumination. When participants could not see a letter at 3 meters their VA was tested
by reducing the testing distance and when the participant could not see letters at 1 meter, VA
was determined by counting fingers, hand motion, light perception, and no light perception.
Following the recording of the VA, a torch was used to inspect for the presence of any corneal
opacity, cataracts, or pseudophakia/aphakia.
Finally, the optometrist set up a semi-dark room within the participant’s home for the static
retinoscopy technique and retinoscopy was performed for each study participant. Objective
refraction was performed using streak retinoscopy. The objective retinoscopy result was then
refined using monocular subjective refraction. Subjective refraction was then recorded for
each eye. Finally, the spherical equivalent was calculated for the result of subjective refraction.
Study participants with a spherical equivalent of > +0.50 or < -0.50 diopter in either eye were
categorized as having RE. Finally, for individuals with refractive problems, a spectacle pre-
scription was supplied to the participant.
Data quality control
To ensure the consistency of the data, the questionnaire was translated from English to
Amharic and back again. A pre-tested Amharic version of semi-structured questions was used
to ensure the reliability of the questionnaires. Before collecting data, a pretest of 48 participants
(5% of the sample size) was conducted in Yirgalem, Sidama, to ensure that the questionnaire
was clear, acceptable, and understandable.
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
5 / 14
PLOS ONE Prevalence and associated factors of refractive error among adults in South Ethiopia
To increase the quality of the data, the data collectors and one supervisor received one day
of training before the actual data collection day. Training on how to utilize the Kobo Toolbox,
examination procedures, and interviewing techniques was given. The supervisor closely moni-
tored the data collection activities in the field and ensured that the collected data was complete
and consistent.
Data processing and analysis
The data collected in the Kobo Toolbox was checked for completeness and consistency. The
data were exported to Microsoft Excel, cleaned, and coded with SPSS 26, and then further anal-
ysis was conducted by using SPSS. Descriptive statistics like percentage and frequency were
used to summarize demographic data and categorical variables. A binary logistic regression
was used to identify factors related to RE. In the bivariable analysis, variables having a P-value
of less than 0.2 were entered on the multivariable logistic regression (S2 File).
The variance inflation factor (VIF) and tolerance test have been used to determine whether
the independent variables were multi-collinear, and a value less than 1.05 with a tolerance less
than 0.955 was found. The model’s fitness was evaluated using the Hosmer and Lemeshow
goodness of fits, and the P-value was 0.76. To demonstrate the relationship between the inde-
pendent and dependent variables, an adjusted odds ratio with a 95% confidence interval was
computed. A P-value of less than 0.05 was considered statistically significant.
Result
Socio-demographic characteristics of study participants
A total of 894 participants were involved in the study, the remaining 57 individuals were non-
respondents making a response rate 94.1%. 3 cases with corneal opacity and 2 cases with infec-
tion were excluded during the study. The median age of the participant was 37 years, with an
interquartile range (IQR) (28–50). Out of 894 study participants, 466 (52.1%) were male,
(23.0%) were private employees and 478(53.5%) had college/university educational status
(Table 1).
Table 1. Socio-demographic characteristics of study participants among adults in Hawassa City, South Ethiopia, 2023 (n = 894).
Variables
Age (year)
Sex
Educational status
Occupational status
Categories
18–28
29–37
38–50
51–80
Male
Female
Unable to read and write
Read and write
Primary school
Secondary school
College/ University
Unemployed
Farmer
Housewife
Student
Merchant
Government employee
Private employee
n = sample size
https://doi.org/10.1371/journal.pone.0298960.t001
Frequency (N)
Percent (%)
238
197
242
217
466
428
15
63
71
267
478
106
22
128
93
140
199
206
26.6
22.0
27.1
24.3
52.1
47.9
1.7
7.0
7.9
29.9
53.5
11.9
2.4
14.3
10.4
15.7
22.3
23.0
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
6 / 14
PLOS ONE Table 2. Systemic comorbidities, clinical and behavioral characteristics of study participants among adults in Hawassa City, South Ethiopia, 2023 (n = 894).
Prevalence and associated factors of refractive error among adults in South Ethiopia
Categories
Frequency(N)
Percent (%)
Variables
Diabetes mellitus
Hypertension
Eye examination
Duration of eye examination(year) (n = 380)
Mode of an eye examination (n = 380)
Family history of RE
History of wearing spectacle
Having cataract
History of cataract surgery
Smoking
Yes
No
Yes
No
Yes
No
>3
� 3
Home
Traditional medicine
Hospital/clinic
Yes
No
Yes
No
Yes
No
Yes
No
Smoker
Non-Smoker
Sleeping duration (hour)
Longer duration shorter duration
Regular use of electronic devices
Yes
No
https://doi.org/10.1371/journal.pone.0298960.t002
69
825
58
836
380
514
27
353
2
3
375
124
770
24
870
85
809
30
864
32
862
608
286
201
693
7.7
92.3
6.5
93.5
42.5
57.5
7.1
92.9
0.5
0.8
98.7
13.9
86.1
2.7
97.3
9.5
90.5
3.4
96.6
3.6
96.4
68.0
32.0
22.5
77.5
Systemic comorbidities, clinical and behavioral characteristics of study
participants
This study reported that 69(7.7%), 58(6.5%), and 124(13.9%) of the study participants had a
history of diabetic mellitus, hypertension, and a family history of RE respectively. Besides, reg-
ular use of electronic devices was found among 201(22.5%) of the study participants (Table 2).
Prevalence of RE
Among the total of 894 participants, 110 (12.3%) [95% CI: 10.2, 14.5%] had a RE. The preva-
lence of uncorrected RE was 11.1%. This study revealed that from the total RE 43.8% of them
had myopia and 2.7% had high myopia (Fig 2).
Factors associated with RE
Bivariable and multivariable binary logistic regression was performed to identify the associated
factors with RE. In bivariable binary logistic regression analysis, older age, being male, regular
use of electronic devices, longer sleeping duration, positive history of diabetes mellitus, family
history of RE, having cataract, and history of cataract surgery were associated with RE.
Those variables in the bivariable analysis that had a P-value less than 0.2 were entered into a
multivariable binary logistic regression. A family history of RE, regular use of electronic
devices, a positive history of diabetes mellitus, and a history of cataract surgery were associated
with RE in multivariable logistic regression with a P-value of less than 0.05.
The odds of having RE among participants aged 51–80 years were two times more likely
compared with participants aged 18–28 years (AOR = 2.08, 95% CI: 1.01–4.31).
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
7 / 14
PLOS ONE Prevalence and associated factors of refractive error among adults in South Ethiopia
Fig 2. Types of refractive error among adults in Hawassa City, South Ethiopia, 2023 (n = 110).
https://doi.org/10.1371/journal.pone.0298960.g002
Regular use of electronic devices was also significantly associated with RE. The odds of having
RE among participants with regular use of electronic devices were 3.64 times higher compared to
participants who had no regular use of electronic devices (AOR = 3.64, 95% CI: 2.25–5.91).
The odds of having RE among participants who had a positive history of diabetes mellitus
were 4.02 times higher than those who had no history of diabetes mellitus (AOR = 4.02, 95%
CI: 2.16–7.48).
The odds of having RE among Participants who had a family history of RE were 2.71 times
more likely than participants who had no family history of RE (AOR = 2.71, 95% CI: 1.59–
4.61). The odds of having RE among participants who had a history of cataract surgery were
5.17 times higher compared to participants who had no history of cataract surgery
(AOR = 5.17, 95% CI: 2.19–12.4) (Table 3).
Discussion
The prevalence and associated factors of RE were assessed in this community-based cross-sec-
tional study among adults in Hawssa City, South Ethiopia.
The finding of this study revealed that the prevalence of RE was 12.3% (95% CI: 10.2–
14.5%). This result was in line with the study conducted in Bogota, Colombia 12.5% [29]. Both
studies used similar study designs, which may account for this similarity.
On the other hand, the finding of this study was lower than studies conducted in Gondar
Northwest Ethiopia 35.6% [31], Borumed Ethiopia 18.3% [7], and London United Kingdom
54% [32]. In this case, the discrepancy may be due to the socio-demographic characteristics of
the study population and the study setting. As an example, the study done in Gondar was con-
ducted among pregnant women. During pregnancy, corneal curvature and central corneal
thickness increase substantially, while intraocular pressure decreases. Those physiological
changes contribute to RE, which may lead to an increase in the prevalence of RE [33]. Further-
more, the study in Borumed, Ethiopia, was hospital-based. Given that most patients go to the
hospital for vision difficulties, this could overestimate the magnitude of RE. Furthermore, a
study in London, United Kingdom, was conducted among older persons, as age causes struc-
tural changes in the ocular system, which increase the magnitude of RE [34].
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
8 / 14
PLOS ONE Table 3. Bivariable and multivariable binary logistic regression analysis for factors associated with RE among adults in Hawassa City, South Ethiopia, 2023
(n = 894).
Prevalence and associated factors of refractive error among adults in South Ethiopia
Variable
Age (year)
50–80
38–50
29–37
18–28
Sex
Male
Female
Regular use of electronic devices (hours)
Yes
No
Sleeping duration (hour)
Longer
Shorter
Diabetes mellitus
Yes
No
Family history RE
Yes
No
Having cataract
Yes
No
History of cataract surgery
Yes
No
RE
Yes
41
26
23
20
64
46
51
59
85
25
30
80
31
79
17
93
16
94
NO
176
216
174
218
402
382
150
634
523
261
39
745
93
691
68
716
14
770
COR: crude odds ratio
AOR: adjusted odds ratio
https://doi.org/10.1371/journal.pone.0298960.t003
COR
(95%CI)
AOR
(95%CI)
2.53(1.43–4.49) 1.31(0.71–2.42)
1.44(0.76–2.70)
1.00
2.08(1.01–4.31) 1.51(0.76–2.99)
1.78(0.89–3.57)
1.00
1.32(0.88–1.98) 1.00
1.18(0.75–1.85) 1.00
P-value
0.047
0.237
0.100
0.457
3.65(2.41–5.53) 1.00
3.64(2.25–5.91) 1.00
< 0.001
1.69 (1.06–2.71) 1.00
1.38(0.83–2.30) 1.00
0.208
7.16 (4.2–12.1)
1.00
4.02(2.16–7.48) 1.00
< 0.001
2.91 (1.82–4.65) 1.00
2.71(1.59–4.61) 1.00
< 0.001
1.92 (1.08–3.41) 1.00
1.60(0.79–3.25) 1.00
0.187
9.36 (4.42–19.79) 1.00
5.17(2.19–12.4) 1.00
< 0.001
The current study’s results were greater than those obtained in Eritrea 6.4% [35], Kenya
7.4% [36], and Durban South Africa [37]. This difference may be due to variations in the
method they employed and cut-off points for RE. The study in Eritrea employed a definition
of RE with a VA of 6/12 or worse, which excluded participants who had RE with a VA better
than 6/12, which may reduce the prevalence of RE. A study done in Durban, South Africa only
included 15- to 24-year-olds, but this study included all persons 18 years and above. Several
ocular diseases (diabetic retinopathy, glaucoma, and cataracts) and structural changes (retinal
degeneration) in the ocular system are common among older adults and thus lead to RE. Since
ocular growth stabilizes at older ages, RE risk factors will likely differ from those of younger
ages due to ocular growth stability and slight changes in biometrics [34]. Because of age-related
ocular disorders that increase the prevalence of RE, the above condition causes an increase in
RE. Furthermore, the result of a study conducted in Bangladesh 4.7% [38] was lower than in
this study; this discrepancy might be caused by the difference in the study population.
The odds of having RE among participants who had a history of diabetes mellitus were 4.02
times higher compared to participants who had no diabetes mellitus. This result is comparable
with the studies conducted in Borumed, Ethiopia, and Yunnan, China [7, 39]. Clinical research
has demonstrated that transient RE shifts are related to blood glucose levels. Increasing glucose
may decrease the osmotic pressure of aqueous humor, leading to a flow of water from the
aqueous humor into the lens, resulting in functional and morphologic changes in the lens. As
a result of changes in lens refractive index, diabetics are more likely to develop RE [40, 41].
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
9 / 14
PLOS ONE Prevalence and associated factors of refractive error among adults in South Ethiopia
The odds of having RE among participants who underwent cataract surgery were 5.17
times higher than those participants who had no history of cataract surgery. This result was
supported by the study conducted in South India [42]. Cataract surgery induces RE in different
ways, which can be in preoperative (errors in biometry parameters, Pre-existing systemic &
ocular comorbidities, Pre-existing uncorrected corneal astigmatism >1.00 DC), intraoperative
(surgical variations of incision size, incision location, Use of sutures), or postoperative (shift in
IOL position) conditions [43–45].
The odds of having RE in participants who had a family history of RE was 2.71 times higher
than in participants who had no family history of RE. This result is comparable with the studies
conducted in Arba Minch, Ethiopia, and East China [9, 30]. Studies have found considerable
relationships between first-degree relatives’ RE. Research has shown that RE aggregates signifi-
cantly within families. It has been reported that the heritability of RE ranges from 50% to 90%
within various populations [46, 47].
The odds of having RE among participants who have regular use of electronic devices were
3.64 times higher than participants who have no regular use of electronic devices. This result
was consistent with a study conducted in Gondar, Northwest Ethiopia, and Rohtak India [48,
49]. Staring at the computer for an extended period causes prolonged accommodation and
muscle fatigue, which might result in a transient shift in the refractive status of the eye [50]. In
addition, staring at the computer for an extended time will cause dry eye, which will affect the
refractive power of the cornea [51].
Strengths and limitations of the study
Both objective and subjective full refraction procedure was performed to determine the refrac-
tive status of the eye. As the study is community-based it is more representative than institu-
tion-based studies.
A cross-sectional study design does not reveal a cause-and-effect relationship between
dependent and independent variables. Recall bias was another issue due to the nature of the
questionnaire to assess family history of RE and smoking.
Conclusion
As a conclusion, the prevalence of RE in this study area was 12.3%. A family history of RE, reg-
ular use of electronic devices, a positive history of diabetes mellitus, and a history of cataract
surgery were significantly associated with RE. Since most of these associated factors are modi-
fiable (regular use of electronic devices, a positive history of diabetes mellitus, and a history of
cataract surgery), eye care professionals should primarily focus on the prevention of these
modifiable causes. To mitigate the burden of RE, it is recommended that eye care professionals
prioritize early screening of individuals with diabetes. From a perspective of minimizing post-
operative RE following cataract surgery, there is a need to enhance preoperative evaluation
and intraoperative care.
Supporting information
S1 File. English version of questionnaire.
(DOCX)
S2 File. Data used for analysis including data on refractive error and associated factors.
(SAV)
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
10 / 14
PLOS ONE Prevalence and associated factors of refractive error among adults in South Ethiopia
Acknowledgments
The authors would like to acknowledge study participants for their participation in the study
and also, we would like to acknowledge data collectors (optometrists).
Ethical consideration
The University of Gondar, College of Medicine and Health Sciences, School of Medicine, and
the Ethical Review Committee provided us with ethical approval, approval ID 06/01/622/
2015EC, and the regional administrative office gave us a letter of support. All study partici-
pants provided verbal informed consent after they were provided with an information sheet
receiving a full explanation of the study’s objective and being informed that they have the right
to question and withdraw from the study at any moment during data collection. this was
approved by the IRB.
There was no reward or risk for the study participants who were chosen. By avoiding any
personal identifiers in the data-gathering tools and using password-protected data on a com-
puter, confidentiality was maintained. In addition, the collected data on the data collector’s
phone was discarded after sending the daily collected information to the principal investigator
to maintain confidentiality.
Author Contributions
Conceptualization: Marshet Gete Abebe, Mikias Mered Tilahun.
Data curation: Marshet Gete Abebe, Abiy Maru Alemayehu, Minychil Bantihun Munaw,
Henok Biruk Alemayehu.
Formal analysis: Marshet Gete Abebe, Henok Biruk Alemayehu.
Methodology: Marshet Gete Abebe, Henok Biruk Alemayehu.
Software: Marshet Gete Abebe, Abiy Maru Alemayehu, Minychil Bantihun Munaw, Henok
Biruk Alemayehu.
Supervision: Abiy Maru Alemayehu.
Validation: Minychil Bantihun Munaw, Mikias Mered Tilahun.
Visualization: Abiy Maru Alemayehu, Mikias Mered Tilahun, Henok Biruk Alemayehu.
Writing – original draft: Marshet Gete Abebe.
Writing – review & editing: Minychil Bantihun Munaw, Mikias Mered Tilahun, Henok Biruk
Alemayehu.
References
1. Alem KD, Gebru EA. A cross-sectional analysis of refractive error prevalence and associated factors
among elementary school children in Hawassa, Ethiopia. J Int Med Res. 2021; 49
(3):300060521998894. https://doi.org/10.1177/0300060521998894 PMID: 33752506
2. Hashemi H, Fotouhi A, Yekta A, Pakzad R, Ostadimoghaddam H, Khabazkhoob M. Global and regional
estimates of prevalence of refractive errors: Systematic review and meta-analysis. J Curr Ophthalmol.
2018; 30(1):3–22. https://doi.org/10.1016/j.joco.2017.08.009 PMID: 29564404
3.
Jonas JB, Bourne RR, White RA, Flaxman SR, Keeffe J, Leasher J, et al. Visual impairment and blind-
ness due to macular diseases globally: a systematic review and meta-analysis. Am J Ophthalmol.
2014; 158(4):808–15. https://doi.org/10.1016/j.ajo.2014.06.012 PMID: 24973605
4. Collaborators G RL. Causes of blindness and vision impairment in 2020 and trends over 30 years, and
prevalence of avoidable blindness in relation to VISION 2020: the Right to Sight: an analysis for the
Global Burden of Disease Study. Lancet Glob Health. 2021; 9(2):144–60.
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
11 / 14
PLOS ONE Prevalence and associated factors of refractive error among adults in South Ethiopia
5. Naidoo K, Gichuhi S, Basa´ ñez M-G, Flaxman SR, Jonas JB, Keeffe J, et al. Prevalence and causes of
vision loss in sub-Saharan Africa: 1990–2010. British Journal of Ophthalmology. 2014; 98(5):612–8.
https://doi.org/10.1136/bjophthalmol-2013-304081 PMID: 24568870
6.
Flaxman SR, Bourne RR, Resnikoff S, Ackland P, Braithwaite T, Cicinelli MV, et al. Global causes of
blindness and distance vision impairment 1990–2020: a systematic review and meta-analysis. Lancet
Glob Health. 2017; 5(12):1221–34. https://doi.org/10.1016/S2214-109X(17)30393-5 PMID: 29032195
7. Besufikad B, Hailemichael W, Tilahun L, Yimam W, Anteneh S. Refractive errors and associated factors
among patients visiting BoruMeda Hospital’s secondary eye Unit in Dessie Town, South Wollo Zone,
Ethiopia. BMC ophthalmology. 2022; 22(1):1–5.
8. Shiferaw Alemu D, Desalegn Gudeta A, Tsega Ferede A, Woretaw Alemu H. Prevalence and degrees
of myopia and hyperopia at Gondar university hospital tertiary eye care and training center, Northwest
Ethiopia. Clinical optometry. 2016:85–91. https://doi.org/10.2147/OPTO.S116535 PMID: 30214353
9. Worku S, Getachew T, Nagarchi K, Shewangizaw M. The Magnitude of Refractive Error and Its Associ-
ated Factors Among Patients Visiting Ophthalmology Clinics in Southern Ethiopia, 2022. Clin Ophthal-
mol. 2023; 17:1801–11. https://doi.org/10.2147/OPTH.S408610 PMID: 37383841
10. World Health Organization. Blindness and vision impairment 2022 [cited 2023 March]. Available from:
https://www.who.int/news-room/fact-sheets/detail/blindness-and-visual-impairment.
11.
Lou L, Yao C, Jin Y, Perez V, Ye J. Global patterns in health burden of uncorrected refractive error.
Invest Ophthalmol Vis Sci. 2016; 57(14):6271–7. https://doi.org/10.1167/iovs.16-20242 PMID:
27893092
12. Berhane Y, Worku A, Bejiga A, Adamu L, Alemayehu W, Bedri A, et al. National survey on blindness,
low vision and trachoma in Ethiopia: Methods and study clusters profile. Ethiop J Health Dev. 2007; 21
(3):185–203.
13.
Fricke T, Holden B, Wilson D, Schlenther G, Naidoo K, Resnikoff S, et al. Global cost of correcting vision
impairment from uncorrected refractive error. Bulletin of the World Health Organization. 2012; 90:728–
38. https://doi.org/10.2471/BLT.12.104034 PMID: 23109740
14. Mohammadi S, Farzadfar F, Pour PM, Ashrafi E, Lashay A, Mohajer B, et al. Prevalence and burden of
refractive errors at national and sub-national levels in Iran. J Ophthalmic Vis Res. 2022; 17(1):78.
https://doi.org/10.18502/jovr.v17i1.10173 PMID: 35194499
15. Ye H, Qian Y, Zhang Q, Liu X, Cai X, Yu W, et al. Prevalence and risk factors of uncorrected refractive
error among an elderly Chinese population in urban China: A cross-sectional study. BMJ Open. 2018; 8
(3):bmjopen–2017-021325.
16. Nae¨l V, Moreau G, Monferme´ S, Cougnard-Gre´ goire A, Scherlen A-C, Arleo A, et al. Prevalence and
associated factors of uncorrected refractive error in older adults in a population-based study in France.
JAMA ophthalmology. 2019; 137(1):3–11. https://doi.org/10.1001/jamaophthalmol.2018.4229 PMID:
30326038
17. Cochrane GM, du Toit R, Le Mesurier RT. Management of refractive errors. BMJ. 2010;340. https://doi.
org/10.1136/bmj.c1711 PMID: 20385718
18. Honavar SG. The burden of uncorrected refractive error. Indian J Ophthalmol. 2019; 67(5):577. https://
doi.org/10.4103/ijo.IJO_762_19 PMID: 31007210
19. Soboka JG, Teshome TT, Salamanca O, Calise A. Evaluating eye health care services progress
towards VISION 2020 goals in Gurage Zone, Ethiopia. BMC Health Serv Res. 2022; 22(1):1–9.
20. Agency CS. National Population and Housing Census of Ethiopia: Population Projection of Ethiopia for
All Regions, at Wereda Level from 2014–2017. Ethiopian Central Statistics Agency. 2018.
21. Chan VF, Mebrahtu G, Ramson P, Wepo M, Naidoo KS. Prevalence of refractive error and spectacle
coverage in Zoba Ma’ekel Eritrea: a rapid assessment of refractive error. Ophthalmic Epidemiol. 2013;
20(3):131–7. https://doi.org/10.3109/09286586.2013.783082 PMID: 23713915
22. Cumberland PM, Bountziouka V, Hammond CJ, Hysi PG, Rahi JS, Eye UB, et al. Temporal trends in
frequency, type and severity of myopia and associations with key environmental risk factors in the UK:
Findings from the UK Biobank Study. Plos one. 2022; 17(1):e0260993. https://doi.org/10.1371/journal.
pone.0260993 PMID: 35045072
23. Cheng F, Shan L, Song W, Fan P, Zhang L, Wang X, et al. Prevalence and risk factor for refractive error
in rural Chinese adults in Kailu, Inner Mongolia. Opht and Physiol Optics. 2021; 41(1):13–20. https://doi.
org/10.1111/opo.12745 PMID: 33104269
24. Nikaj S, Chaloupka FJ. The effect of prices on cigarette use among youths in the global youth tobacco
survey. Nicotine Tob Res. 2014; 16(Suppl_1):S16–S23. https://doi.org/10.1093/ntr/ntt019 PMID:
23709614
25. Na K-S, Park Y-G, Han K, Mok JW, Joo C-K. Prevalence of and risk factors for age-related and anterior
polar cataracts in a Korean population. PLoS One. 2014; 9(6):96461.
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
12 / 14
PLOS ONE Prevalence and associated factors of refractive error among adults in South Ethiopia
26. Sewunet SA, Aredo KK, Gedefew M. Uncorrected refractive error and associated factors among pri-
mary school children in Debre Markos District, Northwest Ethiopia. BMC Ophthalmol. 2014; 14:1–6.
27. Berhane MA, Demilew KZ, Assem AS. Myopia: an increasing problem for medical students at the Uni-
versity of Gondar. Clinical Ophthalmol. 2022:1529–39. https://doi.org/10.2147/OPTH.S365618 PMID:
35615078
28. Raman R, Pal SS, Adams JSK, Rani PK, Vaitheeswaran K, Sharma T. Prevalence and risk factors for
cataract in diabetes: Sankara Nethralaya Diabetic Retinopathy Epidemiology and Molecular Genetics
Study, report no. 17. Invest Ophthalmol Vis Sci. 2010; 51(12):6253–61. https://doi.org/10.1167/iovs.10-
5414 PMID: 20610838
29.
Luque LC, Naidoo K, Chan VF, Silva JC, Naduvilath TJ, Peña F, et al. Prevalence of refractive error,
presbyopia, and spectacle coverage in Bogota´ , Colombia: a rapid assessment of refractive error.
Optometry and Vision Science. 2019; 96(8):579–86.
30. Xu C, Pan C, Zhao C, Bi M, Ma Q, Cheng J, et al. Prevalence and risk factors for myopia in older adult
east Chinese population. BMC Ophthalmol. 2017; 17:1–11.
31. Diress M, Yeshaw Y, Bantihun M, Dagnew B, Ambelu A, Seid MA, et al. Refractive error and its associ-
ated factors among pregnant women attending antenatal care unit at the University of Gondar Compre-
hensive Specialized Hospital, Northwest Ethiopia. PLoS One. 2021; 16(2):0246174. https://doi.org/10.
1371/journal.pone.0246174 PMID: 33577552
32. Cumberland PM, Bao Y, Hysi PG, Foster PJ, Hammond CJ, Rahi JS, et al. Frequency and distribution
of refractive error in adult life: methodology and findings of the UK Biobank Study. PLoS One. 2015; 10
(10):0139780. https://doi.org/10.1371/journal.pone.0139780 PMID: 26430771
33. Agrawal N, Agarwal LT, Lavaju P, Chaudhary SK. Physiological ocular changes in various trimesters of
pregnancy. Nepal J Ophthalmol. 2018; 10(1):16–22. https://doi.org/10.3126/nepjoph.v10i1.21685
PMID: 31056572
34. Hashemi A, Khabazkhoob M, Hashemi H. High prevalence of refractive errors in an elderly population;
a public health issue. BMC Ophthalmology. 2023; 23(1):38. https://doi.org/10.1186/s12886-023-02791-
x PMID: 36707798
35. Chan VF, Mebrahtu G, Ramson P, Wepo M, Naidoo KS. Prevalence of refractive error and spectacle
coverage in Zoba Ma’ekel Eritrea: a rapid assessment of refractive error. Ophthalmic epidemiology.
2013; 20(3):131–7. https://doi.org/10.3109/09286586.2013.783082 PMID: 23713915
36. Bastawrous A, Mathenge W, Foster A, Kuper H. Prevalence and predictors of refractive error and spec-
tacle coverage in Nakuru, Kenya: a cross-sectional, population-based study. Int Ophthalmol. 2013;
33:541–8. https://doi.org/10.1007/s10792-013-9742-6 PMID: 23440405
37. Naidoo KS, Chinanayi FS, Ramson P, Mashige KP. Rapid assessment of refractive error in the eThe-
kwini Municipality of KwaZulu-Natal, Durban, South Africa. Clin Exp Optom. 2016; 99(4):360–5. https://
doi.org/10.1111/cxo.12377 PMID: 27161520
38. Muhit M, Minto H, Parvin A, Jadoon MZ, Islam J, Yasmin S, et al. Prevalence of refractive error, presby-
opia, and unmet need of spectacle coverage in a northern district of Bangladesh: Rapid Assessment of
Refractive Error study. Ophthalmic Epidemiol. 2018; 25(2):126–32. https://doi.org/10.1080/09286586.
2017.1370119 PMID: 28976783
39. Wang M, Cui J, Shan G, Peng X, Pan L, Yan Z, et al. Prevalence and risk factors of refractive error: a
cross-sectional Study in Han and Yi adults in Yunnan, China. BMC Ophthalmol. 2019; 19(1):33. https://
doi.org/10.1186/s12886-019-1042-0 PMID: 30683073
40. Song E, Qian Dj, Wang S, Xu C, Pan Cw. Refractive error in Chinese with type 2 diabetes and its associ-
ation with glycaemic control. Clin Exp Optom. 2018; 101(2):213–9. https://doi.org/10.1111/cxo.12606
PMID: 28975669
41. Kasˇtelan S, Gverović-Antunica A, Pelčić G, Gotovac M, Marković I, Kasun B, editors. Refractive
changes associated with diabetes mellitus. Seminars in Ophthalmology; 2018: Taylor & Francis. https://
doi.org/10.1080/08820538.2018.1519582 PMID: 30199309
42. Marmamula S, Barrenkala NR, Challa R, Kumbam TR, Modepalli SB, Yellapragada R, et al. Uncor-
rected refractive errors for distance among the residents in’homes for the aged’in South India–The
Hyderabad Ocular Morbidity in Elderly Study (HOMES). Ophthalmic Physiol Opt. 2020; 40(3):343–9.
https://doi.org/10.1111/opo.12684 PMID: 32207179
43.
Lundstro¨m M, Dickman M, Henry Y, Manning S, Rosen P, Tassignon MJ, et al. Risk factors for refrac-
tive error after cataract surgery: Analysis of 282 811 cataract extractions reported to the European Reg-
istry of Quality Outcomes for cataract and refractive surgery. J Cataract Refract Surg. 2018; 44(4):447–
52.
44. Khoramnia R, Auffarth G, Łabuz G, Pettit G, Suryakumar R. Refractive outcomes after cataract surgery.
Diagnostics. 2022; 12(2):243. https://doi.org/10.3390/diagnostics12020243 PMID: 35204334
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
13 / 14
PLOS ONE Prevalence and associated factors of refractive error among adults in South Ethiopia
45. Aristodemou P, Sparrow JM, Kaye S. Evaluating refractive outcomes after cataract surgery. Ophthal-
mology. 2019; 126(1):13–8. https://doi.org/10.1016/j.ophtha.2018.07.009 PMID: 30153943
46. Peet JA, Cotch M-F, Wojciechowski R, Bailey-Wilson JE, Stambolian D. Heritability and familial aggre-
gation of refractive error in the Old Order Amish. Invest Ophthalmol Vis Sci. 2007; 48(9):4002–6. https://
doi.org/10.1167/iovs.06-1388 PMID: 17724179
47. Wojciechowski R, Congdon N, Bowie H, Munoz B, Gilbert D, West SK. Heritability of refractive error
and familial aggregation of myopia in an elderly American population. Invest Ophthalmol Vis Sci. 2005;
46(5):1588–92. https://doi.org/10.1167/iovs.04-0740 PMID: 15851555
48. Diress M, Yeshaw Y, Bantihun M, Dagnew B, Ambelu A, Seid MA, et al. Refractive error and its associ-
ated factors among pregnant women attending antenatal care unit at the University of Gondar Compre-
hensive Specialized Hospital, Northwest Ethiopia. Plos one. 2021; 16(2):e0246174. https://doi.org/10.
1371/journal.pone.0246174 PMID: 33577552
49. Kumar N, Jangra B, Jangra MS, Pawar N. Risk factors associated with refractive error among medical
students. Int J Community Med Public Health. 2018; 5(2):634–8.
50. Kim S-H, Suh Y-W, Choi Y-M, Han J-Y, Nam G-T, You E-J, et al. Effect of watching 3-dimensional tele-
vision on refractive error in children. Korean Journal of Ophthalmology. 2015; 29(1):53–7. https://doi.
org/10.3341/kjo.2015.29.1.53 PMID: 25646061
51. Alemayehu A, Alemayehu MM. Pathophysiologic mechanisms of computer vision syndrome and its pre-
vention. World J Ophthalmol Vis Res. 2019; 2(5):1–7.
PLOS ONE | https://doi.org/10.1371/journal.pone.0298960 March 25, 2024
14 / 14
PLOS ONE
| null |
10.1016_j.cell.2023.05.028.pdf
|
lability—Plasmids generated in this study are available upon written request.
Data and Code availability
•
•
All raw and processed sequencing data in this study, including single-cell DNA
methylation sequencing, single-cell RNA-seq, ChIP-seq, Cut and Run assay
and Nanopore sequencing, have been deposited to the NCBI Gene Expression
Omnibus (GEO) database under accession GSE208449. All data are publicly
available as of the date of publication.
This paper analyses existing, publicly available data or available upon request
to the authors. These accession numbers for the datasets are listed in the key
resources table.
Cell. Author manuscript; available in PMC
|
• All raw and processed sequencing data in this study, including single-cell DNA methylation sequencing, single-cell RNA-seq, ChIP-seq, Cut and Run assay and Nanopore sequencing, have been deposited to the NCBI Gene Expression Omnibus (GEO) database under accession GSE208449. All data are publicly available as of the date of publication. • This paper analyses existing, publicly available data or available upon request to the authors. These accession numbers for the datasets are listed in the key resources table. • This paper does not report original code. All the scripts and mathematical algorithms used in this study will be available from the corresponding authors upon request. • All the versions of software packages used in this study are listed in the key resource table and noted in the data analysis method accordingly. • Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Experimental model Clinical Specimens-All patient samples were collected in this study after written informed consent, in accordance with Institutional Review Board (IRB) protocols (DF/HCC 05-300, 11-497, 13-217 or 14-375). For the CTC cohort, 10-20 ml of blood was drawn from patients with a diagnosis of metastatic prostate cancer, localized prostate cancer, or age-matched males without a diagnosis of cancer at Massachusetts General hospital (MGH). For the localized tumor tissue cohort, all samples were acquired from either core biopsies or surgical resection of untreated localized prostatic adenocarcinoma (Gleason scores 6 and 8) from patients at MGH. In cases with the lowest grade tumors (Gleason score 6), normal prostate tissue was also identified in the tissue specimen by a Genito-Urinary (GU) specialized pathologist and used as a source of matched normal prostate cells. Both normal and tumor tissue samples were de-identified, snap frozen and sectioned. Only tumor sections with >80% tumor content, as assessed by a specialized GU pathologist were used in this study. The clinical data of the patients with metastatic prostate cancer enrolled in the single-cell CTC analysis and patients with resected localized prostate cancer used for single nucleus analysis are described in Table S1 . The clinical data of the patients with localized prostate cancer and metastatic prostate cancer enrolled in Nanopore sequencing anlysis of CTC-enriched blood are described respectively in Table S4 and Table S5 . Cell culture-Human prostate cancer cell lines (LNCaP, VCaP, PC3 and 22Rv1), murine prostate cancer line (Myc-CaP), normal cultured prostate epithelial cells (HPrEC), benign prostatic hypertrophy cells (BPH-1) and murine Lewis lung carcinoma cells (LLC-1) were all obtained from ATCC, after authentication by short tandem repeat (STR) profiling. All cell lines used in the paper were derived from male mice or male human patients. They were cultured in the following media at 37°: RPMI-1640 (ATCC) medium supplemented with 10% FBS (Gibco) and 1X Pen/Strep (Gibco) (for LNCaP, VCaP, PC3, 22Rv1 and BPH-1 cells); Prostate Epithelial Cell medium (ATCC) with 6 nM L-glutamine (ATCC), 0.4% Extract P (ATCC), 1.0 mM Epinephrine (ATCC), 0.5 ng/ml rh-TGFα (ATCC), 100ng/ml hydrocortisone hemisuccinate (ATCC), 5 mg/ml rh-Insulin (ATCC), 5 mg/ml Apo-transferrin (ATCC), 33 μM Phenol red (ATCC) and 1X Pen/Strep/Ampho Solution (ATCC) (for HPrEC cells); DMEM high glucose medium (Gibco) with 10% FBS (Gibco) and 1X Pen/Strep (Gibco) (for Myc-CaP cells and LLC-1 cells). All the cell lines used in this study were checked for mycoplasma every 4 months using Mycoalert kit (Lonza). Mouse xenograft assays-All animal experiments were carried out in accordance with approved protocols by the MGH Subcommittee on Research Animal Care (IACUC). All the mice used in this study were maintained under a 12/12 h light/dark cycle in MGH animal facility. 6-8 weeks old FVB male mice (Jackson Laboratory, Strain#001800) or 6-8 weeks old male immunodeficient NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) mice (Jackson Laboratory, Strain#005557) were used for intraprostatic injection or subcutaneous injection of Myc-CaP cells stably expressing luciferase and mCherry. 6-8 weeks old C57BL/6 female mice (Jackson Laboratory, Strain#000664) were used for subcutaneous injection of LLC-1 cells stably expressing luciferase. Littermates of the same sex were randomly assigned to experimental groups. For intraprostatic inoculation, mice were first anesthetized using isoflurane, and a 1 cm skin incision was performed along the midline of the abdomen to expose the inner muscle layer, which was then separated. The tip of seminal vesicle was raised gently with forceps to expose the anterior lobe of the prostate gland. 50,000 Myc-CaP cells 1:1 mixed with Matrigel (v/v) (total volume: 30 μl) were slowly injected into the prostate lobe. All the tissues were then returned into the abdomen, and continuous sutures were used to close the inner muscle layer, followed by separate skin closure. For subcutaneous injections, mice were anesthetized, and 50,000 Myc-CaP cells or 1,000,000 LLC-1 cells 1:1 mixed with Matrigel (v/v) (total volume: 100 μl) were injected into the flank. Tumor cell-derived bioluminescent signal was quantified every other day for the Myc-CaP cells and 3 times a week for the LLC-1 for mice after either orthotopic injection or subcutaneous injection. At 2-3 weeks after inoculation, mice were sacrificed and tumors were harvested for flow cytometry and RNA extraction for the Myc-CaP experiments. Method Details CTC isolation-CTCs were isolated from fresh blood specimens drawn from patients with prostate cancer, following negative depletion of leukocytes using the microfluidic CTC-iChip as reported previously 26, 27 . Briefly, 10-20 ml of whole blood specimens were incubated with biotinylated antibody cocktails against CD45 (R&D Systems, clone 2D1), CD66b (AbD Serotec, clone 80H3), and CD16 (BD Biosciences), followed by incubation with Dynabeads MyOne Streptavidin T1 (Invitrogen) for magnetic labeling and depletion of leukocytes. After CTC-iChip processing, the CTC-enriched product was further stained with FITC-conjugated antibody against EpCAM (Cell Signaling Technology, clone VU1D9) and PE-conjugated antibody against CD45 (BD Biosciences, clone HI30). Single CTCs (FITC positive and PE negative) or white blood cells (WBCs, FITC negative and PE positive) were individually picked into PCR tubes containing 5 μl RNA/DNA lysis buffer using micromanipulator (Eppendorf TransferMan NK 2) and snap-frozen in liquid nitrogen. In total, 38 CTCs from 5 different patients (GU114, GU169, GU181, GU216 and GURa15) with metastatic prostate cancer were individually picked, sequenced and lineage-confirmed based on transcriptome and DNA copy number. One patient sample (GU169) had only one CTC, and it was therefore excluded from some downstream analyses focused on the four patients with multiple CTCs. Nuclei isolation from frozen tumor sections-Tumor tissue sections with high tumor content (>80%) and adjacent normal tissue section were micro-dissected and transferred into a pre-chilled Dounce homogenizer containing ice-cold 1 ml 1X HB buffer (0.26 M sucrose, 30 mM KCl, 10 mM MgCl2, 20 mM Tricine-KOH, 1 mM DTT, 0.5 mM Spermidine, 0.15 mM Spermine, 0.3% NP-40 and 1X complete protease inhibitor). Tissue was homogenized with ~10 strokes of 'A' loose pestle, followed by another ~10 strokes of 'B' tight pestle. The tissue homogenate was then filtered using a 70 μm strainer and pelleted by centrifugation. Nuclear pellets were resuspended and purified by density gradient centrifugation (top layer: 25% Iodixanol solution; middle layer: 30% Iodixanol solution; bottom layer: 40% Iodixanol solution). The nuclear band at the interface of 30% and 40% Iodixanol solutions was collected into a new Eppendorf tube and washed twice with ice-cold 1X PBS. 20% of the purified nuclei were used to isolate single nuclei using fluorescenceactivated cell sorting (FACS) for single-cell DNA methylation analysis, while the remaining 80% of the nuclei were subjected to ChIP-seq analysis. Western Blot-Cells or tumor tissues were lysed in Laemmli buffer (Sigma) and cleared. Protein concentration was determined using DC protein assay (Bio-rad). Proteins (25 μg) were separated on precast NuPAGE 4-12% Bis-Tris protein gels (ThermoFisher), and transferred onto nitrocellulose membranes (Bio-Rad). After blocking with 5% BSA buffer for 1 hour at room temperature, membranes were incubated with primary antibodies overnight at the recommended concentrations. HRP conjugated secondary antibodies (1:10,000; Bio-rad; Cat#5196-2504) were applied, and ultra-sensitive autoradiography film (Amersham) was used to detect the chemiluminescence signal. Primary antibodies used are H3K27me3 (1:1,000, Invitrogen Cat#MA5-11198) and H3 total (1:1,000, Abcam Cat#1791). 5-Azacytidine treatment, bisulfite sequencing and staining of chromatin marks -The human prostate epithelial cell line BPH-1 was cultured in the presence of 5 μM 5azacitidine (Selleck, #S1782). At serial time points (days 0, 1, 4 and 5), cells were collected for DNA extraction, confocal microscopy, or flow cytometric analysis. DMSO-treated cells were used as control at each time point. To quantify 5-azacitidine-induced demethylation at the genomewide level, we used the whole genome bisulfite sequencing (WGBS). Briefly, DNA ws extracted from BPH-1 cells upon 5-azacitidine treatment, 1 μg genomic DNA was used to sonicate into 300-500 bp fragments, DNA was end-polished, A-tailed and ligated with pre-methylated adaters before bisulfite conversion using EZ DNA methylation kit (Zymo, #D5001), bisulfite-converted DNA was amplified and sample index was introduced during amplification. To quantify 5-azacytidine-induced demethylation at the CD1A-IFI16 locus, DNA extracted from BPH-1 cells treated with 5-azacitidine was subjected to bisulfite conversion using EZ DNA methylation kit (Zymo, #D5001), and bisulfite-converted DNA was used for PCR amplification, applying bisulfite-specific PCR primers covering the human CD1A-IFI16 locus (see Table S3 ). PCR products were purified by 1% agarose gel and cloned using the Zero blunt PCR cloning kit (ThermoFisher, #K270020). 10 individual bacterial clones were randomly picked for Sanger sequencing. Sequencing data were analyzed and shown using online tool QUMA ( http://quma.cdb.riken.jp/ ) 61 . Nuclear accumulation of H3K27me3 was stained with H3K27me3 antibody (1:1000 dilution; CST#9733), in 5-azacytidine-treated cells. Images were acquired using a Zeiss LSM710 Lase Scanning Confocal and were quantified by quantitative image analysis of cells (ImageJ). Flow cytometry was also performed at serial time points on BD LSRFortessa machine to assess CD1d expression using human CD1d-APC antibody (1:100 dilution; BioLegend#350308, clone: 51.1). EZH2 inhibitor treatment-Human prostate cancer cell lines (22Rv1, LNCaP and VCaP) were cultured in the presence of the small molecule EZH2 inhibitor GSK126 (Selleckchem, #S7061) at the indicated concentration (0, 5 or 10μM). After 6 days of treatment, protein and RNA were harvested, for quantitation of H3K27me3 and total H3, using Western blotting and expression of individual genes within the CD1A-IFI16 locus by real time qPCR. Paired single-cell DNA methylation and RNA-seq-For these experiments, we used either single CTCs or WBCs individually picked from fresh blood specimens after CTC enrichment, and single cells from cultured prostate cell lines (either picked or FACS-sorted). These were subjected to paired single-cell DNA methylation and RNA-seq analysis to obtain the transcriptomes and DNA methylomes from the same single cells 33, 62 . Briefly, single cells were first lysed in 5 μl DNA/RNA lysis buffer; 0.5 μl Magnetic MyOne Carboxylic Acid Beads (Invitrogen, Cat#65011) were then added to each single cell lysate to facilitate segregation of nucleus versus cytoplasm. After centrifugation and magnetic separation, the supernatant (containing cytoplasmic RNA) was transferred into a new tube for single-cell RNA-seq amplification using the SMART-seq2 protocol 63 , while the pellet (aggregated beads with the intact nucleus) was resuspended in DNA methylation lysis buffer and subjected to single-cell whole genome methylation sequencing using the scBS-seq protocol 64 . Single nuclei sorted from the frozen primary prostate tumor sections were also subjected to the scBS-seq procedure. MNase native ChIP-seq-Purified nuclei from frozen tissue sections were subjected to MNase native ChIP-seq following the ULI NChIP procedure, as published elsewhere 65 . Briefly, nuclei were suspended in Nuclear Isolation Buffer (Sigma) supplemented with 1% TritonX 100, 1% Deoxycholate and 1X complete protease inhibitor. Chromatin was digested by MNase enzyme (NEB, 1:10 diluted) at 21°C for 7.5 min, and further diluted in Complete Immunoprecipitation Buffer, with 1X complete protease inhibitor. 2 μl ChIPgrade H3K27me3 (Active motif, Cat#39155) or H3K9me3 (Abcam, Cat#ab8898) antibody was incubated with the digested chromatin overnight at 4°C. DNA was then purified using protease K digestion followed by phenol-chloroform extraction. ChIP-seq sequencing libraries were prepared using NEBNext Ultra II DNA Library Prep Kit (NEB, Cat#E7645L). Cut and Run Assay-H3K27me3 and H3K9me3 Cut and Run assays were performed with cultured prostate cell lines (LNCaP, 22Rv1, BPH-1, HPrEC and Myc-CaP), using the CUT&RUN Assay kit (CST, Cat#86652S). Briefly, 100,000 freshly cultured prostate cells were collected and incubated with Concanavalin A Magnetic Beads. 2 μl ChIP-grade H3K27me3 (Active motif, Cat#39155) or H3K9me3 (Abcam, Cat#ab8898) or IgG (CST, Cat#66362S) antibody was added to the cell: bead suspension and incubated overnight at 4°C. 1.5 μl pAG-MNase enzyme was then added to the tube, which was rotated for 1 h at 4°C, followed by activation of pAG-MNase using 3 μl cold Calcium Chloride. The activation reaction was stopped and DNA was further diluted and collected for phenol-chloroform extraction. Cut and Run sequencing libraries were constructed using NEBNext Ultra II DNA Library Prep Kit (NEB, Cat#E7645L). Next generation sequencing-All the single-cell RNA-seq, single-cell DNA methylation, MNase ChIP-seq, Cut and Run samples and WGBS samples were molecularly barcoded, pooled together and sequenced on a HiSeq X sequencer to obtain 150 bp pairended reads (Novogene). RNA extraction, reverse transcription and quantitative PCR (qPCR)-RNA extracted from cultured prostate cells was prepared using the RNeasy Mini kit (QIAGEN) with DNase I digestion on the column. To extract RNA from mouse tumor tissues, these were first dissected to remove connective tissue and fat, and washed extensively with 1X PBS to remove excessive blood or necrotic tissues. Tumors were then homogenized in RLT RNA lysis buffer using a Dounce homogenizer, and passed through a QIAshredder column (QIAGEN). RNA from normal prostate of FVB mice were prepared following a similar method. RNA from tissue homogenate was extracted using the RNeasy Mini kit (QIAGEN) with DNase I digestion on the column. cDNA was synthesized from 50-200 ng RNA using SuperScript III One-Step qRT-PCR kit (Invitrogen). qPCR was performed using the primers listed in Table S3 . (Addgene#12260) packaging plasmids using Lipofectamine 2000 reagent (Invitrogen). 48-72 h after transfection, culture medium (containing lentiviral particles) was collected, filtered and concentrated using LentiX concentrator (Clontech). Concentrated virus was added to the Myc-CaP cells in presence of polybrene (Santa Cruz, 8 μg/ml as final concentration) overnight. FACS was used to select GFP positive cells as marker of Cd1d1 construct transduction in the Myc-CaP cells. The LLC-1 cells transduced with the Cd1d1 cloned in the the N174-MCS vector were selected using G418 (Sigma Aldrich #G8168) at 400 μg/mL for 4-6 days. To obtain stable Ifi204 overexpression, 10 μg/ml blasticidin (InvivoGen) was added to the medium for 5-7 days selection. Tumor immune infiltration assayed by flow cytometry-Mouse tumors generated by intraprostatic injection of control or Cd1d1-expressing Myc-CaP cells were dissected and washed to remove blood, fat and connective tissues. Tumor tissues were further mashed and digested in 5 ml digestion buffer (RPMI1640, 2.5 mg/ml collagenase D, 0.1 mg/ml DNase I) at 37°C for 30 min. Tissue digestion was stopped by adding another 5 ml RPMI1640 with 2% FBS, and then filtered through 70 μm strainers. The tissue cell suspension was obtained in the same way for tumors generated by subcutaneous injection of control or Ifi204 expressing Myc-CaP cells. To stain for NKT cell infiltration in prostate tumors with control or Cd1d1 expression, the singlecell suspension was first blocked with rat anti-mouse CD16/CD32 blocking reagent (BD#553142, Clone: 2.4G2) at 4°C for 30 min, followed by mouse NKT surface antibody cocktail staining at 4°C for another 30 min. The mouse NKT surface antibodies used in this study were: BV510-viability dye (BD#564406), APC-α-GalCer-mCD1d Tetramer Multiplex Oxford Nanopore native sequencing-Blood samples from either healthy donors or patients with localized or metastatic prostate cancer were subjected to CTC-ichip enrichment (10 4 -fold leukocyte depletion) 26, 27 . The enriched CTCs (ranging from 0.1% to 1% purity, admixed with residual leukocytes) were subjected to high molecule weight (HMW) DNA extraction using the HMW DNA extraction kit (QIAGEN), and then prepared for Oxford Nanopore sequencing using the rapid barcoding kit (Nanopore#SQK-RBK004). For each sequencing run, 11 blood samples (either from healthy donors or cancer patients), together with 1 lambda DNA (unmethylated control), were uniquely barcoded and pooled together. Sequencing was performed using a Nanopore MinION device with R9.4 flowcell for 48 h, per manufacturer instructions. Single-cell and bulk RNA-seq data analysis-Raw fastq reads generated from HiSeq X sequencer were first cleaned using TrimGalore (v0.4.3) ( https://github.com/FelixKrueger/ TrimGalore ) to remove the adapter-polluted reads and reads with low sequencing quality. Cleaned reads were aligned to the human (hg19) or mouse (mm9) genome using Tophat (v2.1.1) 66 . PCR duplicates were further removed using samtools (v1.3.1) 67 , gene counts were computed using HTseq (v0.6.1) 68 , gene expression level (FPKM) was further calculated using cufflinks (v2.1.1) 66 . Gene expression matrix was subjected to R (v3.1.2) or Prism9 for graphics. Single-cell and bulk DNA methylation sequencing data analysis-Raw fastq reads from both the single-cell and bulk DNA methylation sequencing were first trimmed using TrimGalore (v0.4.3) ( https://github.com/FelixKrueger/TrimGalore ), and cleaned reads were aligned to the human hg19 or mouse mm9 genome (in silico bisulfite converted) using Bismark tool (v0.17.0) 69 . Samtools (v1.3.1) 67 was used to remove PCR duplicates, and CpG methylation calls were extracted using the Bismark methylation extractor 69 . 0.1% lambda DNA was spiked in, prior to bisulfite treatment, for each sample to assess the bisulfite conversion efficiency. Only samples with more than 4 million unique CpG sites covered at least once and with a bisulfite conversion rate > 98% were used in this study. TCGA methylation array data reanalysis-Prostate DNA methylation datasets from TCGA analyzed by Illumina Infinium Human Methylation 450 K BeadChip were downloaded from the National Cancer Institute's GDC Data Portal (https:// portal.gdc.cancer.gov) for 502 tumor samples and 50 normal samples. CpG site-level methylation files (beta value, txt format) were first converted to hg19 coordinates using UCSC lift-over tool (https://genome.ucsc.edu/cgi-bin/hgLiftOver) for the downstream analysis. The data were binned to a fixed set of 10 kb nonoverlapping genomic windows by computing the average fraction methylation within each bin in each sample. Bins were excluded if they lacked coverage (i.e., had no probes on the Illumina Infinium Human Methylation 450 K BeadChip array) or had a mean normal-tissue methylation level, averaged across all the normal samples, of <70%. For each sample, the global methylation level was calculated as the fraction of bins having methylation >50%. The methylation level at the CD1A-IFI16 locus for each sample was calculated as the fraction of bins in the range chr1:158,130,000-158,340,000 (hg19) having methylation >50%. The gene expression data and clinical information of TCGA PRAD samples, including Gleason score, tumor stage and others, were all downloaded from cbioportal (https://www.cbioportal.org/). Tumor purity was calculated using ABSOLUTE algorithm 70 . DNA Methylation 450 K BeadChip datasets for other cancer types were also downloaded from the National Cancer Institute's GDC Data Portal ( https://portal.gdc.cancer.gov ) and CpG site-level methylation files (beta value, txt format) were also converted to hg19 coordinates using UCSC lift-over tool ( https:// genome.ucsc.edu/cgi-bin/hgLiftOver ) for the downstream analysis. Genomic element enrichment analysis-For analytical purposes, a promoter region was defined based on the relative position to a transcription start site (TSS): 1,500 bp upstream and 500 bp downstream. The annotations of TSS, exon, intron, intragenic regions, CpG islands (CGIs), repetitive elements and UCSC gap regions were all downloaded from UCSC genome table browser ( https://genome.ucsc.edu/cgi-bin/hgTables ) 71 . Enrichment analysis on different genomic elements was calculated using the Bioconductor package regioneR (v1.18.1) with overlapPermTest function 72 . DNA copy number analysis inferred by single-cell DNA methylation sequencing data-Single-cell DNA methylation sequencing reads were first aligned to the genome using Bismark. Uniquely aligned reads were extracted into a bed file and subsequently submitted to Ginkgo online tool 73 , http://qb.cshl.edu/ginkgo ) to infer the DNA copy number, using 5 Mb as the bin size. The processed integer copy number data from the Ginkgo website (SegCopy.tsv) was used to calculate the DNA Copy Number Variation (CNV) score. Given an assignment of a copy number to all the locations in a diploid genome, we define a CNV score for any given single cells as follows. Let ci be the copy number at the ith location of the genome. CNV score is then defined to be the average over all i in the genome of the absolute value of (ci-2). DNA copy number analysis inferred by single-cell RNA-seq data-Single-cell RNA-seq reads were aligned to human genome using TopHat, and large-scale chromosomal copy number alterations were determined by InferCNV ( https://github.com/broadinstitute/ infercnv ). MNase ChIP-seq and Cut and Run data analysis-ChIP-seq and Cut and Run reads were first trimmed by Trim Galore (v0.4.3) ( https://github.com/FelixKrueger/TrimGalore ) and then mapped to the human or mouse genome using BWA men 74 . Duplicated reads were marked by sambamba 75 and further removed using samtools 67 . MACS2 (v2.0.10) 76 was used to call the peaks and deepTools 77 were used to compute the ChIP-seq or Cut and Run signal around prostate PMDs. Determination of Partially Methylated Domains (PMDs)-The human genome was first binned into 100 kb windows placed at 200 bp offsets. Windows that intersected CGIs or UCSC gap regions were discarded. For each source (i.e., single CTCs from patients with prostate cancer, single WBCs from healthy donors, single cells from normal prostate or prostate cancer cell lines or normal prostate tissues 42 , the per-source methylation level of each window was calculated by taking the average over all cells from that source of the methylation level of the CpG sites within the given window. For each source the distribution of the per-source methylation level of the 100 kb windows was plotted. Normal cells showed a unimodal distribution, while prostate cancer cells showed a bimodal distribution. A threshold for hypomethylation determination was set at the lowest point of the valley in the histogram of the bimodal distribution for each prostate cancer patient or prostate cell line; if the distribution was unimodal, the threshold was set to 60%. The windows with methylation level lower than threshold were defined as hypomethylation windows and overlapping hypomethylation windows were merged into per-source PMDs. The 250 kb minimal length threshold was then applied to the per-source PMDs. The union of the per-source PMDs for all single CTCs from four prostate cancer patients (GU114, GU216, GURa15 and GU181) and for all single cells from four prostate cancer cell lines (LNCaP, VCaP, 22Rv1 and PC3) was defined as the total prostate PMDs (1,496 in total). Chromatin mark and genome element enrichment analyses were performed on these PMDs. To identify the genes that reside in the most consistently hypomethylated PMDs across all prostate cancer specimens analyzed (i.e., intersection), we quantile-normalized the DNA methylation levels for all PMDs among all CTCs from four prostate cancer patients (GU114, GU216, GURa15 and GU181) and all single cells from four prostate cancer cell lines (LNCaP, VCaP, 22Rv1 and PC3) and only used the PMDs (annotated as core prostate PMDs) with their averaged quantile-normalized DNA methylation level less than 25% across these 8 sources to extract the genes. Determination of Preserved Methylation Islands (PMIs)-After identification of PMDs for each of the eight sample sources [CTCs from four prostate cancer patients (GU114, GU216, GURa15 and GU181) and single cells from four prostate cancer cell lines (LNCaP, VCaP, 22Rv1 and PC3)], we defined small interspersed islands ('gaps') with preserved methylation (sample source PMIs) using the following criteria: (1) every PMI is flanked by defined PMDs in each given source; (2) length of each PMI should be >30 kb and <3 Mb. Total prostate PMIs were defined by taking the union of sample source PMIs across 8 sources (1,412 in total), while core prostate PMIs (44 in total) were defined by requiring the uniformity across sample sources: the genomic location of given PMI is overlapped in all 8 sample sources. Differential gene expression and hypergeometric gene set enrichment analysis (hGSEA)-Differential gene expression between TCGA prostate normal tissue and primary tumors was determined as follows: We started by considering the genes that reside in the most hypomethylated PMDs [as described in the section titled 'Determination of partially methylated domains (PMDs)']. Of those, genes with 95th percentile of normalized FPKM values less than 1 were discarded. A two-tailed variance-equal t-test was performed on each of the remaining genes. The p-values from those t-tests were used to generate a false-discovery rate (FDR) estimate for each gene by the Benjamini-Hochberg method. We considered genes for which the FDR estimate was less than 0.1 to be differentially expressed between normal prostate and prostate tumor samples. hGSEA was performed to determine the gene set and pathway enrichment using the phyper R function as reported elsewhere 26 . All gene sets and pathways evaluated in this study were obtained from MSigDB (v7.2) from the Broad Institute. Differential gene expression and hGSEA for genes in PMIs was performed in the same way. Heterogeneity assessment-Consistent with a previous publication 26 , means of correlation coefficients and jackknife estimates were used to assess the heterogeneity within and between subsets of samples. Nanopore data analysis-Nanopore sequencing reads (format: fast5) generated by Nanopore MinION device were first converted into fastq files using ONT Albacore software (v2.3.1) ( https://nanoporetech.com/community ). Demultiplexing was also performed during fast5 to fastq conversion. DNA methylation information was extracted from both fast5 and fastq files using Nanopolish software (v0.10.2) ( https://github.com/nanoporetech/ nanopolish ). Nanopolish output files (albacore_output.sorted.bam and methylation_calls.tsv) were used for downstream analysis. Every nanopore run was spiked in with lambda DNA, which was used as the negative control to assess the fidelity of Nanopore sequencing. To estimate CTC-derived hypomethylation signal in each Nanopore sequencing sample, stringent criteria were applied: (1) each Nanopore read should be long enough to harbor at least 30 CpG sites with confident methylation calls after Nanopolish; (2) the number of Nanopore reads aligned to prostate PMDs (pre-determined among CTCs isolated from 4 prostate cancer patients and 4 prostate cancer cell lines using single-cell whole genome bisulfite sequencing) should be no fewer than 300 for metastatic patients or no fewer than 400 for localized patients; (3) methylation level of spike-in lambda DNA in each run should be <1%. Following application of these criteria, microfluidic processed (leukocytedepleted) blood samples from seven patients with metastatic prostate cancer, six patients with localized prostate cancer. Since we required different number of Nanopore reads in the prostate PMDs for metastatic patients and localized patients, 23 age-matched healthy donors were validated for analysis in the metastatic cohort, and 21 were validated for localized cohort. In-silico mathematical modeling of Nanopore sequencing in detecting rare signal-To assess the ability to detect large hypomethylated domains in rare circulating tumor cells, we performed an analysis using Nanopore reads from a normally methylated non-cancer cell line (HUES64) with 1% in-silico spiked-in reads from a cancer cell line (HCT116). We assessed the ability to determine the correct cell line of origin for reads that aligned to predefined HCT116 PMDs based on their average methylation level by quantifying the precision and sensitivity of read classification using the PRROC 78 . Methylation was averaged across each read, considering only CpG sites that fall within PMDs and excluding those within CpG islands. Illustration-Illustrations were created with BioRender.com . Quantification and Statistical Analysis Statistical analyses for all experiments are described in the figure legends and the method details. Statistical analyses were performed using R (version 3.1.2) and GraphPad Prism 9.0. Supplementary Material Refer to Web version on PubMed Central for supplementary material. (D) IGV representation (hg19) of DNA methylation spanning chromosome 8, showing extensive PMDs (yellow) across 37 individual CTCs from four patients (GU114, GU216, GU181 and GURa15), and 17 cells from prostate cancer cell lines (LNCaP, PC3, VCaP, 22Rv1). As controls, 4 normal bulk prostate tissues (N.P.), 36 cells from two prostate epithelial cell lines (HPrEC, BPH-1) and normal leukocytes (WBCs) are shown. Normal methylation level (blue). (E-F) Higher resolution of chromosome 8 in IGV, showing precise PMD boundaries shared across individual CTCs and prostate cancer cell lines (panel E), with magnified view of the nested PMI, bracketing a few genes, with precise boundaries of preserved methylation flanked by profound hypomethylation (panel F). (G-H) Components of coding genes and classes of repeats differentially enriched in PMDs versus PMIs (panel G), with differences among subtypes of repeats (panel H). ns, not significant; *P<0.05; **P<0.01, assessed by permutation test. (C) Boxplot showing enrichment of Cut and Run signal for H3K27me3, but not H3K9me3, across prostate cancer PMDs between LNCaP cells and non-transformed cell lines (HPrEC and BPH-1). Pvalue, one-tailed Student's t-test. (D) IGV track showing representative cancer-associated PMD (DNA hypomethylation: yellow), with pronounced enrichment of H3K27me3 at PMD borders in cancer cells (LNCaP: red) versus non-transformed cells (HPrEC: blue, BPH-1: green), whereas PMDcentered H3K9me3 occupancy is unaltered. (E) Inter-and intra-patient heterogeneity of PMDs among single CTCs from four prostate cancer patients (red) and single cells from prostate cancer cell lines. Mean Jaccard index indicates heterogeneity, with higher mean score indicating less heterogeneity among samples. Error bar, mean with 95% confidence interval (CI). (F-G) IGV representation of total PMDs and core PMDs at chromosome 3 locus, across 8 sample sources (4 patients and 4 prostate cancer cell lines). Total PMDs (blue) are the union of PMDs defined in each sample source, while core PMDs (black) are shared across all 8 sample sources (panel F); representation of PMDs from the single-cell components of an individual sample source (22 CTCs from patient GU181) showing a core PMD shared across all sample sources (black) and neighboring non-core PMDs that are shared by >90% CTCs in this patient, but not across different sample sources (panel G). See Figure S2D and Methods for criteria in core PMD and PMI designation. (D) Quantitation of demethylation as a function of GS in TCGA prostate cancer methylation array data, showing early and progressive loss of methylation of core PMDs (red curve), with an attenuated trend for other PMDs (magenta). The core PMIs (blue) display stable DNA methylation pattern during prostate tumorigenesis. Statistical analysis as for panel C. (E-F) Gene set enrichment analysis (GSEA) of genes residing within core PMDs and downregulated in primary prostate cancer (E), and of genes residing within core PMIs with gene expression preserved (up-regulated and not significantly changed) in primary prostate cancer (F), compared with normal prostate. (FDR <0.1; two-tailed Student's t-test with FDR correction). (C) IGV screenshot of single-cell DNA methylation data showing progressive demethylation of CD1A-IFI16 locus (box with red dashed line) from normal prostate cells to localized (GS6 and GS8) and metastatic prostate cancer (CTCs). Heterogeneity of hypomethylation (shaded yellow) across single cells is evident at GS6, becoming more prevalent at GS8, and uniform in CTCs . (D) Plots showing suppressed expression of lipid antigen presentation and interferon inducible genes within the CD1A-IFI16 locus, during transition from normal prostate to low-grade GS6, with persistent silencing in higher grade GS7, 8 and 9 cancers (TCGA dataset). Error bar, mean with SEM. (E) Analysis of 33 different tumor types (TCGA) for DNA methylation differences at core prostate cancer PMDs, compared with corresponding normal tissues. 30 of 35 (86%) evaluable PMDs are hypomethylated across all tumor types (red circles), with the CD1A-IFI16 locus having the strongest hypomethylation. (F) Histograms of DNA methylation level within 100kb windows (200bp offsets) across the genome in normal prostate cells (BPH-1), following 5-azacytidine treatment (days 1 and 5), compared with DMSO control. (G) Quantitation of H3K27me3-related fluorescence intensity within single-cell nuclei (confocal microscopy). Error bar, mean with SEM. P-value, two-tailed Student's t-test. (H) Sequential reduction in CD1d protein expression in normal prostate cells (BPH-1) treated with 5-azacytidine, compared with DMSO control. Representative flow cytometry (left panel); median fluorescence intensity (right panel). Error bar, mean with SEM. P-value, two tailed Student's t-test. (I-J) Western blot showing reduced H3K27 trimethylation in 22Rv1 cells treated with EZH2 inhibitor GSK126 for 6 days (panel H); qPCR of genes within the CD1A-IFI16 cluster show induced expression (panel I), while non-PMD resident control genes (PP1A, HPRT and β-actin) remain unchanged. P-value, Tukey's multiple comparison tests, where GSK126 treatment conditions (red bars) were compared to controls (blue bar). n.s. not significant; ****P<0.0001. (D) Myc-CaP cells engineered as in (C) show no difference in tumor growth in immunedeficient NSG mice. Error bar, mean with SEM. (E) Flow cytometry of Cd1d-restored Myc-CaP tumors in FVB mice, showing recruitment of CD1d-restricted NKT cells (marked by α-GalCer CD1d Tetramer) and activated NKT cells (marked by CD69), compared with controls. Error bar, mean with SD. (F) Flow cytometry of Ifi204-restored Myc-CaP tumors in FVB mice, showing unaltered infiltration of total CD4 + and CD8 + T cells, but reduced immune infiltration by PD-1 + CD8 + T cells and increased presence of TNFα + CD8 + T cells, compared with controls. Error bar, mean with SD. P-values, two-tailed Student's t-test; ns, not significant. (E) Schematic of microfluidic CTC enrichment (followed by direct Nanopore sequencing of bulk cells (approximatly 0.1% CTC purity). HMW, high molecular weight. (F-G) Scatter plot quantitation of hypomethylation signal by Nanopore sequencing, comparing leukocyte-depleted blood samples from patients with either metastatic (panel F) or localized prostate cancer before surgical resection or radiation therapy (panel G), versus healthy age-matched male donors (HDs). Error bar denotes mean with SEM. P-value assessed by two-tailed Student's t-test. Dotted lines indicate thresholds of hypomethylation signal that encompass all healthy donors tested, with the fraction of cancer patients with hypomethylation signal above that threshold considered positive.
|
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
HHS Public Access
Author manuscript
Cell. Author manuscript; available in PMC 2023 August 18.
Published in final edited form as:
Cell. 2023 June 22; 186(13): 2765–2782.e28. doi:10.1016/j.cell.2023.05.028.
DNA hypomethylation silences antitumor immune genes in early
prostate cancer and CTCs
Hongshan Guo1,2,11,13, Joanna A. Vuille1,13, Ben S. Wittner1, Emily M. Lachtara1, Yu
Hou3,4,11, Maoxuan Lin1,4, Ting Zhao1,5, Ayush T. Raman1,4, Hunter C. Russell1, Brittany
A. Reeves1, Haley M. Pleskow1,6, Chin-Lee Wu1,5, Andreas Gnirke4, Alexander Meissner4,7,
Jason A. Efstathiou1,6, Richard J. Lee1,8, Mehmet Toner9,10, Martin J. Aryee1,5,12, Michael S.
Lawrence1,4,5, David T. Miyamoto1,6,*, Shyamala Maheswaran1,9,*, Daniel A. Haber1,2,8,14,*
1.Massachusetts General Hospital Cancer Center, Harvard Medical School, Charlestown, MA
02129, USA.
2.Howard Hughes Medical Institute, Chevy Chase, MD 20815, USA.
3.Evergrande Center for Immunologic Diseases, Brigham and Women’s Hospital, Harvard Medical
School, Boston, MA 02115, USA.
4.Broad Institute of MIT and Harvard, Cambridge, MA 02142, USA.
5.Department of Pathology, Massachusetts General Hospital, Harvard Medical School, Boston,
MA 02114, USA.
6.Department of Radiation Oncology, Massachusetts General Hospital, Harvard Medical School,
Charlestown, MA 02129, USA.
7.Department of Genome Regulation, Max Planck Institute for Molecular Genetics, Berlin 14195,
Germany.
This work is licensed under a Creative Commons Attribution 4.0 International License, which allows reusers to distribute, remix,
adapt, and build upon the material in any medium or format, so long as attribution is given to the creator. The license allows for
commercial use.
*Correspondence: [email protected] (D.T.M.), [email protected] (S.M.), [email protected]
(D.A.H.).
Author contributions
H.G., J.A.V., D.T.M., S.M. and D.A.H. conceived the project, provided leadership for the project and drafted the manuscript. H.G.,
J.A.V., Y.H., T.Z., H.C.R., B.A.R., H.M.P., C.W., J.A.E., R.J.L., M.T. and D.T.M. conducted all the experiments. H.G., B.S.W.,
E.M.L., M.L., A.T.R., A.G., A.M., M.S.L., and M.J.A analyzed all the data. All authors reviewed and edited the manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our
customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review
of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered
which could affect the content, and all legal disclaimers that apply to the journal pertain.
Declaration of interests
Massachusetts General Hospital (MGH) has applied for patents regarding the CTC-iChip technology and CTC detection signatures.
M.T., S.M. and D.A.H. are cofounders and have equity in Tell-Bio, which is not related to this work. The interests of these authors
were reviewed and managed by MGH and Mass General Brigham (MGB) in accordance with their conflict of interest policies. All
other authors declare no competing interests.
Inclusion and Diversity
We support inclusive, diverse, and equitable conduct of research.
Key Ressource Table (see document)
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 2
8.Department of Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, MA
02114, USA.
9.Department of Surgery, Massachusetts General Hospital, Harvard Medical School, Boston, MA
02114, USA.
10.Center for Engineering in Medicine and Shriners Hospital for Children, Harvard Medical School,
Boston, MA 02114, USA.
11.Present address: Bone Marrow Transplantation Center, First Affiliated Hospital, Zhejiang
University School of Medicine and Liangzhu Laboratory, Zhejiang University Medical Center,
Hangzhou, 310012, China.
12.Present address: Department of Data Science, Dana-Farber Cancer Institute, Harvard Medical
School, Boston, MA 02114, USA.
13.These authors contributed equally.
14.Lead contact.
Summary
Cancer is characterized by hypomethylation-associated silencing of large chromatin domains,
whose contribution to tumorigenesis is uncertain. Through high-resolution genome-wide
single-cell DNA methylation sequencing, we identify 40 core domains that are uniformly
hypomethylated from earliest detectable stages of prostate malignancy through metastatic
Circulating Tumor Cells (CTCs). Nested among these repressive domains are smaller loci
with preserved methylation that escape silencing and are enriched for cell proliferation genes.
Transcriptionally silenced genes within the core hypomethylated domains are enriched for
immune-related genes; prominent among these is a single gene cluster harboring all five CD1
genes that present lipid antigens to NKT cells, and four IFI16-related interferon-inducible
genes implicated in innate immunity. Re-expression of CD1 or IFI16 murine orthologs in
immunocompetent mice abrogates tumorigenesis, accompanied by activation of anti-tumor
immunity. Thus, early epigenetic changes may shape tumorigenesis, targeting co-located genes
within defined chromosomal loci. Hypomethylation domains are detectable in blood specimens
enriched for CTCs.
Graphical Abstract
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 3
In Brief
Analysis of circulating tumor cluster cells reveals how DNA hypomethylation during early
prostate tumorigenesis silences immune surveillance genes, while sparing proliferation-associated
genes.
Keywords
DNA hypomethylation; prostate cancer; circulating tumor cells; immune surveillance; single-cell
sequencing
Introduction
Cancer is characterized by two primary changes at the level of DNA methylation1–4.
Focal hypermethylation of CpG islands, often located within gene regulatory regions,
results in gene silencing, a well-established mechanism for inactivation of tumor suppressor
genes5–7. In addition, long-range hypomethylated regions, Partially Methylated Domains
(PMDs), coincide with nuclear Lamina-Associated Domains (LADs) and Large Organized
Chromatin lysine (K) (LOCK) domains8–10. These chromosomal loci are large (>100 kb),
gene-poor, correlated with late-replicating DNA, and topologically associated with nuclear
lamina. Repetitive sequences and retro-elements residing within PMDs may be de-repressed
in cancer, but the rare protein encoding genes are silenced11. Two repression-associated
chromatin modifications are evident: H3K9me3 is abundant within hypomethylated blocks,
while H3K27me3 denotes their boundaries12,13. Conflicting models have suggested that
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 4
hypomethylated blocks are either a direct consequence of cell transformation14, or an
incidental result of excessive cell proliferation13,15. The functional consequences of
hypomethylation-associated gene silencing, and potential selection pressures that shape such
domains, are not well understood. A recent study of advanced colon cancers proposed
an intrinsic tumor suppressive mechanism that may counter cell proliferation13, although
genome-wide hypomethylation is extensive in advanced cancers and may not reveal specific
targets contributing to early tumorigenesis.
Prostate cancer is noteworthy for its characteristically slow evolution from precancerous
lesions with low levels of cell proliferation to more invasive, and ultimately metastatic
malignancy. Localized prostate cancer may be classified as indolent (Gleason score (GS)
6) or clinically significant (GS≥7) based on histological grade, reflecting differences in
differentiation, proliferative index, and metastatic potential16,17. GS6 tumors are often
safely monitored without therapy, while the more aggressive GS7 and higher tumors are
resected surgically or treated with radiation in combination with androgen deprivation
therapy. GS8–10 denotes poorly differentiated tumors with an adverse prognosis and
high propensity for metastasis. Multiple heterogeneous foci of early tumors are often
dispersed throughout the prostate gland, complicating bulk molecular characterization
and necessitating careful dissection with single-cell analytic strategies. Conversely,
advanced metastatic prostate cancer predominantly affects bone, making it difficult to
perform biopsies to study disseminated tumor deposits. Circulating tumor cells (CTCs),
comprising potential metastatic precursors isolated from the bloodstream, thus enable
single-cell analysis of advanced prostate cancer. Immune checkpoint blockade (ICB) is
generally ineffective in treating prostate cancer18–21, possibly reflecting the stroma-rich,
immunosuppressive environment of primary prostate cancer, but tumor cell autonomous
mechanisms may also contribute, in both primary and metastatic disease. Epigenetic changes
affecting expression of immune regulatory genes and modulating the responsiveness of
prostate cancer to immunological therapies have not been characterized.
In addition to their biological significance, cancer-associated methylation changes are of
considerable molecular diagnostic interest for blood-based cancer detection. These rely
primarily on CpG island-enriched methylation within short DNA fragments (170 bp)
circulating in plasma, a fraction of which are tumor-derived (ctDNA)22–24. However, among
patients with localized prostate cancers, only 11.2% are detectable using plasma CpG island
hypermethylation assays25, leading us to ask whether the large genomic coverage provided
by hypomethylated domains within CTCs may provide complementary information. To
address these questions, we first established genome-wide, high-resolution single-cell
bisulfite sequencing of hypomethylated domains within individual prostate CTCs from
multiple patients and cancer cell lines, identifying 40 core PMDs, shared across metastatic
prostate cancers. The timing of DNA hypomethylation during prostate tumorigenesis reveals
that core PMDs are hypomethylated as early as indolent GS6 tumors, identifying a single
predominant genomic locus, the CD1A-IFI16 gene cluster, encompassing the entire family
of CD1 lipid antigen presentation genes and multiple interferon-inducible genes implicated
in innate immunity. Early hypomethylation-mediated gene silencing points to specific
tumorigenic pathways with both biological and diagnostic implications.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Results
Page 5
Identification of shared core PMDs and PMIs across single metastatic prostate cancer cells
To characterize genome-wide DNA methylation features of single metastatic prostate cancer
cells, we enriched CTCs from five patients with castration-resistant prostate cancer, all
with multiple bone metastases and disease refractory to hormonal therapy and performed
individual cell micromanipulation and single-cell sequencing26,27(Table S1, see Methods).
We compared 44 single CTCs with 40 single cells from four prostate cancer cell lines
(LNCaP, VCaP, PC3 and 22Rv1) and two non-transformed prostate epithelial cell lines
(Human Prostate Epithelial Cells (HPrEC) and Benign Prostate Hypertrophy cells (BPH-1)).
HPrECs represent normal prostate epithelium, while BPH-1 cells share luminal cell features
with cancer precursors28–31. As control for contaminating blood cells within CTC-enriched
clinical specimens, we compared single prostate cells with 13 microfluidic-processed single
leukocytes (WBCs) from four age-matched healthy men. To confirm the identity of single
CTCs, we adapted single-cell multiomics sequencing to enable separation of nucleus from
cytoplasm in individual cells, subjecting the former to single-cell whole genome bisulfite
sequencing (scBS-seq)32 and the latter to single-cell RNA-seq (SMART-seq2)33 (Figure
1A, see Methods). On average, we detected 9 million CpG sites for each single-cell DNA
methylation sequencing sample, and 5,790 genes (RPM>0) for each single-cell RNA-seq
library (Figures S1A and S1B). Transcriptomes of prostate CTCs confirm the expression of
expected lineage-specific and epithelial transcripts, and absence of hematopoietic markers
(Figures 1B and S1C). Unsupervised hierarchical clustering analysis of all single-cell RNA-
seq data reveals three distinct clusters: leukocytes, normal prostate, and prostate cancer
(including CTCs and prostate cancer cell lines) (Figure S1D). In addition to transcriptional
confirmation, all prostate CTCs demonstrate extensive DNA copy number variations (CNV)
inferred from single-cell DNA methylation sequencing (see Methods). These CNV patterns
are matched with those inferred from cytoplasmic RNA-seq from the same single cells
(Figures 1C, S1E and S1F). As controls, HPrEC cells and WBCs show normal diploid copy
numbers (Figure 1C, see Methods). As a final test, principal component analysis (PCA) of
promoter methylation patterns readily distinguishes all tumor cells from normal controls
(Figure S2A). Taken all together, we applied highly stringent criteria, including both
transcriptional and DNA copy number confirmation, to nominate 38/44 (86.4%) initially
selected CTCs as bona fide prostate CTCs for detailed single-cell genomic analyses.
We quantified methylation levels of individual cells by binning the genome into 100 kb
windows: the methylation distribution of normal cells is unimodal, with a single peak near
80% methylation, whereas virtually all tumor samples exhibit a bimodal distribution, with a
varying number of hypomethylated regions (Figures 1D–1F and S2B, see Methods). Overall,
DNA hypomethylation constitutes 20–40% of the genome in patient-derived prostate
CTCs and prostate cancer cell lines, but <2.5% in normal prostate cells or blood cells
(Figure S2C). In contrast to individual CpG islands (CGIs), which often demonstrate focal
hypermethylation around gene regulatory regions, the hypomethylated regions in prostate
tumor cells span very large gene-poor regions, consistent with previously described PMDs.
In total, we identified 1,496 PMDs with a mean size of 1.2 Mb (range 250 kb to 9.2
Mb) across the prostate cancer genome, a number consistent with previous measurements
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 6
based on bulk tumor sequencing in multiple advanced cancers8,12,34 (Figure S2D, Table
S2). Notably, on a chromosome-wide view and with the high resolution afforded by
single-cell methylation analysis, some PMDs are punctuated by smaller regions, where
DNA methylation is retained (Figures 1D–1F). We call these Preserved Methylation Islands
(PMIs, see defining criteria in Methods) (Figure S2D, Table S2). In contrast to large gene-
poor PMDs, the 1,412 PMIs interspersed within hypomethylated domains are gene-rich,
with sharp methylation boundaries that bracket a single gene or a small group of genes
(mean PMI size 1.3Mb; range 30.8 kb to 11.1 Mb) (Figures 1F and 1G). The identification
of PMIs raises the possibility that selection pressures may preserve methylation, and
potentially gene expression, at a small number of genes nested within PMDs.
As demonstrated in other cancers12,13,35, PMDs are gene-poor and have strong enrichment
of some endogenous retroviral elements (ERVs), notably Long Terminal Repeats (LTRs). In
contrast, PMIs in prostate cancer are gene-rich with relative absence of long interspersed
nuclear elements (LINEs) and LTRs (Figures 1G and 1H). Previous studies show that PMDs
in breast and colon cancers exhibit depletion of active chromatin marks (H3K4me1/3,
H3K27ac, H3K36me3) and enrichment of repressive histone modifications, including
H3K9me3 at the center of the domains and H3K27me3 at their borders12,13. To confirm
these chromatin changes in prostate cancer, we used cultured cell lines, to analyze chromatin
landscapes using ChIP assays. Analysis of prostate cancer cells (LNCaP and 22Rv1)
confirms the differential positioning of repressive H3K9me3 marks at the center and
H3K27me3 at the border of hypomethylated domains (Figures 2A, 2B and S3A). However,
direct comparison of cancer cells with non-transformed prostate epithelial and basal cells
(HPrEC and BPH-1) at the same PMDs indicates that changes associated with malignancy
primarily relate to H3K27me3 deposition. Indeed, Cut and Run assays show profound
enrichment of H3K27me3 at PMD borders in cancer cells compared with normal cells,
whereas central H3K9me3 marks are abundant at these loci, but invariant between normal
and cancer cells (Figures 2B–2D and S3A–S3C). Thus, hypomethylation-associated gene
silencing in cancer cells is primarily correlated with the acquisition of H3K27me3 histone
modification flanking these chromosomal domains. In contrast, genes within PMIs show
strong enrichment for activation (H3K4me1/3, H3K27ac and H3K36me3) and absence of
repression (H3K27me3 and H3K9me2/3) (Figure 2A).
At the single-cell level, both PMDs and PMIs show substantial intra-patient and inter-patient
heterogeneity (Figures 2E–2G and S3D–S3F), leading us to define common domains shared
across all single prostate cancer cells that may identify common and hence functionally
significant pathways. Of 1,496 PMDs, only 40 (2.7%) are universally hypomethylated, with
the mean quantile normalized methylation level <25%, across cells from all four patients
with metastatic prostate cancer and four prostate cancer cell lines (Figure S2D, Table S2, see
Methods). The 40 core PMDs have a mean size of 2.5 Mb (range 353.4 kb to 7.7 Mb) and
encompass 143 protein-encoding genes, a gene density of 1.44 gene/Mb. Hypomethylation
associated with cell proliferation is thought to be more rapid in loci that have reduced CpG
content15,36. Indeed, we note that the core prostate PMDs exhibit reduced CpG residue
content, compared with other PMDs across the genome (P<0.0066, Figure S3G), providing a
possible explanation for their universal hypomethylation. In the same single prostate cancer
cells, analysis of the 1,412 PMIs for intersection across all prostate cancer patients and
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 7
prostate cancer cell lines identifies 44 core PMIs (Figure S2D, Table S2, see Methods).
Core PMIs have a mean size of 371.7 kb (range 27 kb to 1.9 Mb) and harbor 255 protein-
encoding genes, with a gene density of 15.6 genes/Mb (Figure S3H).
Our single-cell analysis of prostate cancer cells identifies a small fraction of PMDs that are
universally shared, which we describe as core PMDs, and it also reveals that interspersed
within these large PMDs are small gene-rich islands with preserved DNA methylation, that
we call PMIs.
Hypomethylation of core PMDs is an early event in prostate tumorigenesis
DNA hypomethylation progresses during cancer evolution to ultimately encompass large
regions of the non-coding and gene-poor genome within advanced cancers37. By analogy
with early genetic driver mutations, however, non-random epigenetic silencing may play
an important role in initiating tumorigenesis, with selection pressures guiding recurrent
early events. Having defined core PMDs shared across single metastatic prostate cancer
cells, we sought to identify genomic loci that are consistently subject to early silencing
during tumorigenesis. Given the characteristic admixture of tumor and stromal cells in
localized prostate cancer, we obtained frozen tissue sections from prostatectomy specimens
and purified single nuclei for molecular analysis. Tumor origin of individual nuclei was
confirmed by CNV inferred from whole genome bisulfite sequencing, and we computed
a CNV score (absolute DNA copy number changes per Mb) to complement Gleason
histological scoring, as an independent measure of tumor progression (Figure 3A, see
Methods). In addition to Gleason histological scoring of localized prostate cancer, we
computed a CNV score (absolute DNA copy number changes per Mb) to quantify genomic
instability in single nuclei from different prostatectomy samples, as an independent measure
of tumor progression. In total, we profiled 38 primary tumor nuclei from five patients with
low grade (GS6) prostate cancer, 62 nuclei from another five patients with high grade
(GS≥8) disease, and 78 normal prostate cells from adjacent tissue sections, comparing
these with the 38 CTCs from patients with metastatic disease (Table S1). Inferred CNV
from our high resolution single nucleus analysis identifies Chr8p loss (containing NKX3–
1, BMP1, FGFR1 genes and multiple microRNAs) as one of the earliest genetic events
in prostate tumorigenesis, shared by >43% of cancer cells in GS6 tumors (Figure S4A).
Early allelic loss of this locus has been reported in prostate cancer38–41. Interestingly, GS6
prostate cancer cells with Chr8p loss show more hypomethylation across PMDs, pointing to
coordinated early timing of CNV and hypomethylation (Figure S4B). At the single-cell level
across different tumors, hypomethylation at prostate PMDs exhibits less heterogeneity than
do hypermethylated CpG promoter regions (Figures S4C and S4D).
Remarkably, core PMDs initially defined by their universal hypomethylation in metastatic
prostate cancer cells show profound enrichment at the earliest stages of tumorigenesis. In
early GS6 tumors, 77.5% (31/40) core PMDs are hypomethylated, compared with only 8%
(115/1,456) of non-core PMDs (Figure 3B). Indeed, mean quantitative methylation levels
within core PMDs decline from 78.4% (normal prostate), to 70.4% (GS6), 57.2% (GS8),
and 20.2% in metastatic CTCs. Comparable methylation levels across all prostate PMDs
decline more slowly: 82.2% (normal prostate), 80.9% (GS6), 74.7% (GS8) and 57.6%
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 8
(CTCs) (Figure 3C). By contrast, methylation at interspersed core PMIs shows little change
from normal prostate nuclei to GS6, GS8, and metastatic prostate CTCs. Compared with
hypomethylation of large chromosomal domains, focal hypermethylation of CpG islands
within gene regulatory regions increases gradually from 27.5% (normal prostate), to 30.7%
(GS6), 31.9% (GS8), and 34.3% (CTCs) (Figure S4E), as does aneuploidy measured by
CNV score (Figure S4F). We observed no confounding correlation (FDR>0.1) between
CNV and DNA methylation for core PMDs (Figure S4G). Our observations of accelerated
progressive demethylation of core PMDs in early prostate cancer are confirmed by analysis
of TCGA prostate cancer methylation array data stratified by Gleason Score (Figure 3D),
as well as whole genome bisufite sequencing in primary and metastatic prostate tumors34,42
(Figure S4H). Core PMIs show preserved methylation patterns independent of Gleason
Score (Figures 3D and S4H).
Taken together, core PMDs begin to lose DNA methylation within indolent GS6 prostate
cancers, one of the earliest identifiable lesions in prostate tumorigenesis. This early
timing explains their universal hypomethylation in advanced cancers, compared with more
heterogeneous hypomethylation domains that emerge during subsequent tumor progression.
Silencing of immune-related genes within core PMDs and persistent expression of
proliferative genes within PMIs
To address the functional consequences of early DNA hypomethylation, we identified
protein-encoding genes localized to core PMDs that display loss of expression across the
large prostate cancer TCGA database39. Among the 143 protein-coding genes residing
within the 40 core PMDs, 68 (48%) are consistently and significantly differentially
expressed between normal prostate and primary prostate tumors, with 61 (90%) suppressed
and 7 (10%) induced in cancer. Remarkably, 12/61 (20%) silenced genes within core
prostate PMDs are immune-related. GSEA analysis reveals lipid antigen processing and
presentation (P<1.96E-13) and cellular response to interferon (P<2.74E-5) as the two
most highly enriched pathways (Figure 3E, see Methods). Conversely, of the 255 protein-
encoding genes within the 44 core PMIs, 161 (63.1%) are comparably expressed in prostate
cancer and normal prostate tissues in the same TCGA database. The top GSEA pathways all
relate to cell proliferation, including E2F targets (P<0.000975) and DNA repair (P<0.00116)
(Figures 3F, S4I–J). As control, GSEA pathway analysis does not identify statistically
significant enrichment among core PMD-derived genes that are not expressed or not silenced
in prostate cancer, or among core PMI-derived genes without preserved expression. Thus,
identifying early and consistent changes in DNA methylation in prostate cancer cells points
to silencing of immune-related genes, with selective sparing of genes encoding proliferative
drivers, as initial steps in prostate tumorigenesis.
PMD-associated silencing of the CD1A-IFI16 gene cluster
A remarkable feature of core PMD-associated gene silencing is targeting of the entire CD1
family of lipid antigen presentation genes (CD1A, CD1B, CD1C, CD1D and CD1E) and
four interferon inducible genes of the Pyrin and HIN domain (PYHIN) family involved in
immune sensing of non-self DNA (IFI16, AIM2, PYHIN1 and MNDA). These genes are
clustered within the same core hypomethylation block at chromosome 1q23.1 (hereafter,
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 9
CD1A-IFI16 block), consistent with a single genomic locus playing a major role in
integrating these two immune recognition pathways (Figure 4A). The CD1 gene family
encodes MHC class I-like molecules that exclusively present non-peptides (e.g. glycolipids)
to Natural Killer-T (NKT) cells, a rare subset of T cells implicated in both innate and
adaptive immunity43–45. The CD1 pathway is primarily implicated in innate immunity to
infectious agents, although a possible role for lipid antigens in anti-tumor immunity is
also postulated46,47. Among interferon-inducible genes, IFI16 is highly expressed in normal
prostate cells: it is reported to bind non-self dsDNA in both nucleus and cytoplasm in a DNA
length-dependent manner, recruiting STING and further activating interferon signaling48.
DNA methylation of the CD1A-IFI16 locus declines early and rapidly, scoring as the
14th earliest across all genome-wide PMDs measured at GS6 (Figure 4B). Heterogeneity
in hypomethylation at CD1A-IFI16 is evident within single prostate cancer cells at early
stage GS6 tumors, progressively increasing in both fraction of tumor cells and degree of
hypomethylation within individual tumor cells as they evolve to GS8 and ultimately to
metastatic CTCs (Figures 4C and S5A). This early and progressive loss of DNA methylation
at the CD1A-IFI16 locus, compared with the slower rate of demethylation genomewide, is
also evident in analysis of public databases of primary and metastatic prostate cancer34,42
(Figure S5B). Analysis of TCGA prostate cancer data stratified by Gleason Score further
confirms early progressive loss of methylation within the CD1A-IFI16 locus (Figure S5C),
and the associated transcriptional downregulation of the encoded genes as early as GS6
tumors (Figures 4D). The accelerated decline in DNA methylation at CD1A-IFI16 is not
driven by gene copy number changes, as confirmed by comparing single nuclei with or
without CNV at this locus (Figure S5D).
Early DNA hypomethylation at the CD1A-IFI16 locus is not restricted to prostate cancer.
DNA methylation datasets at defined stages of cancer progression are available for both
colon and thyroid cancers9, both of which demonstrate earlier and more progressive
demethylation of CD1A-IFI16, when compared to other core PMDs (Figure S5E).
Furthermore, analysis of methylation profiles in a TCGA cohort including more than 1,000
samples spanning 33 cancer types (https://portal.gdc.cancer.gov) identifies the CD1A-IFI16
locus as consistently hypomethylated in 23 different cancers (Figures S5F–G). Across all
33 cancer types, CD1A-IFI16 demonstrates the greatest degree of DNA hypomethylation
compared with all other core PMDs (Figure 4E), and 19 of the 33 cancers show a significant
correlation between hypomethylation of this locus and reduced RNA expression of CD1A-
IFI16 resident genes (Figure S5H). Early and profound DNA hypomethylation at CD1A-
IFI16 is thus a consistent feature across multiple cancers.
Along with DNA hypomethylation of the CD1A-IFI16 locus, we observed the expected
enrichment for H3K27me3 chromatin marks, comparing prostate cancer versus normal
prostate cell lines, together with suppression of the encoded genes within that locus (Figures
S6A, B and S6C, Table S3). Extending this analysis to nuclei from microdissected GS6
and GS8 tumors using ultra-low-input native ChIP-seq (ULI-NChIP), we observe marked
progressive enrichment of H3K27me3 at the CD1A-IFI16 locus in early GS6 tumors
compared with normal prostate epithelium (Figures S6D–E), whereas other PMDs show
increased H3K27me3 only at GS8 (Figure S6F). Finally, within high purity TCGA prostate
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 10
samples (tumor purity >0.5 inferred by ABSOLUTE algorithm), all five lipid antigen
presentation genes and three of the four PYHIN interferon inducible genes are suppressed in
primary prostate tumors (n=188) compared with normal prostate (n=14) (Figure S6G). The
suppression of CD1A-IFI16 gene expression is observed at the earliest timepoint of DNA
hypomethylation (GS6), and it persists as DNA hypomethylation progresses, suggesting a
potential threshold effect. Thus, immune-related genes within the CD1A-IFI16 cluster are
among the earliest targets of cancer hypomethylation-induced transcriptional silencing.
Functional recapitulation of hypomethylation-associated silencing at CD1A-IFI16 locus
To investigate the functional relationship between DNA methylation, repressive chromatin
marks and expression of PMD-resident genes, we applied the DNA demethylating agent
5-azacytidine (5 μM) to the human prostate epithelial cells (BPH-1), in which the CD1A-
IFI16 locus shows normal DNA methylation levels (Figure 4A). Global DNA methylation
declines by 4.9 % after 24 hrs of 5-azacytidine, and by 37.7% after 5 days of drug
exposure, compared with DMSO controls (Figure 4F), with the CD1A-IFI16 locus showing
progressive DNA demethylation upon 5-azacytidine treatment (Figure S7A). Bisulfite
treatment and Sanger sequencing confirms gradual demethylation at CD1A-IFI16 (DMSO:
75.9%, day5: 40.8%) (Figure S7B). Ectopically-induced demethylation is accompanied by
marked increase of the chromatin silencing mark H3K27me3, as shown by quantitative
imaging of nuclei (7.18-fold increase after 5 days) (Figures 4G and S7C), along with
H3K9me3 (Figures S7D–E), and associated with reduced expression of CD1 (Figure 4H).
Thus, DNA hypomethylation appears to trigger the recruitment of chromatin suppressive
marks at the CD1A-IFI16 locus, along with repression of the resident genes.
We then tested the converse model, using an inhibitor of the EZH2 methyltransferase,
GSK126, to suppress H3K27me3 in prostate cancer cells, in which the CD1A-IFI16 locus is
hypomethylated and silenced. Treatment of three prostate cancer cell lines (22Rv1, LNCaP
and VCaP) with GSK126 results in loss of global H3K27 trimethylation, associated with a
dramatic increase in expression of all the genes within the CD1A-IFI16 locus (Figures 4I–J
and S7F–G). Together, these observations further support the role of chromatin silencing
marks in repressing coding genes within the CD1A-IFI16 locus and other PMDs.
Re-expression of lipid antigen presentation or interferon-inducible genes restores anti-
tumor immunity in a mouse model
To explore the potential significance of CD1A-IFI16 silencing, we tested the consequences
of restored expression in a murine model of early prostate tumorigenesis. The mouse
prostate cancer cell line Myc-CaP is derived from a genetically engineered model
with prostate-specific expression of a c-Myc transgene driving androgen-dependent
tumorigenesis49. Single-cell methylation sequencing of Myc-CaP cells shows uniform
hypomethylation of two chromosomal loci syntenic with the single human CD1A-IFI16
locus, and encompassing the two murine lipid antigen presentation genes (Cd1d1 and
Cd1d2) and the orthologous PYHIN interferon inducible genes (Ifi204, Aim2, Pyhin1 and
Mnda), respectively (Figures S8A and S8B). Repressive H3K27me3 and H3K9me3 marks
are enriched at the Cd1d and interferon inducible genes (Figures S8A and S8B). The
major CD1 murine ortholog Cd1d1 and the IFI16 murine ortholog Ifi204 are repressed
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 11
in Myc-CaP tumor cells, compared with normal prostate tissues dissected from isogenic
FVB mice (Figure 5A). We ectopically expressed Cd1d1 (16.1-fold) or Ifi204 (4.1-fold) in
Myc-CaP cells by lentiviral transduction, achieving levels comparable to those of normal
mouse prostate (Figure 5A and Table S3, see Methods). Cell surface localization of restored
Cd1d1 is evident using both flow cytometry and confocal microscopy (Figures S8C and
S8D).
Ectopic expression of Cd1d1 in Myc-CaP cells does not alter proliferation in vitro,
but these cells fail to produce tumors in isogenic immune competent FVB mice, when
inoculated either subcutaneously or by direct intraprostatic injection (Figures 5B, 5C and
S8E). This effect is dependent upon immune cell activation, since inoculation of the
same Cd1d1-expressing Myc-CaP cells into immunodeficient NSG mice does not suppress
their ability to give rise to primary tumors (Figure 5D). Cd1d specifically mediates the
presentation and activation of lipogenic antigens to NKT cells, a rare T cell subpopulation
expressing Cd40lg and Icos (http://rstats.immgen.org/Skyline/skyline.html)50, and tumors
from Cd1d1-restored Myc-CaP cells in FVB immune competent mice show increased
expression of Cd40lg (2.8-fold; P=0.0063) and Icos (3.2-fold; P=0.00023) compared with
controls (Figure S8F and Table S3). Flow cytometric analysis of tumor immune infiltrates
in Cd1d1-restored tumors indicates more abundant Cd1d-restricted NKT cells (P=0.0042),
along with increased binding to the high affinity synthetic NKT cell ligand alpha-Galactosyl
Ceramide (α-GalCer) tetramer and an increase in the CD69 marker of NKT cell activation
(P=0.0099) (Figures 5E and S9A–C). To test the consequences of restored Cd1d1 expression
in another mouse isogenic tumor model, we restored its expression in the LLC-1 lung
epidermoid carcinoma model, which does not express Cd1d1. Ectopic expression of Cd1d1
in LLC-1 reduces tumor growth upon subcutaneous inoculation into immune competent
isogenic C57BL/6 mice, despite unaltered in vitro proliferation (Figures S8G–J).
We then tested the effect of restored expression in Myc-CaP cells of Ifi204, the murine
ortholog of the interferon inducible gene IFI16. Re-expression Ifi204 also suppresses Myc-
CaP tumorigenesis in immune competent FVB mice, without any anti-proliferative effect
in vitro (Figures 5B and 5C). This effect is not evident in immune deficient NSG mice,
pointing to an immunological effect (Figure 5D). Tumors derived from Ifi204-expressing
Myc-CaP cells in FVB mice show no difference in the total number of CD4+, CD8+ T
cells or in the expression of general marker of T cell activation (Figures 5F, S9D and
S9E). However, compared to parental controls, Ifi204-reconstituted tumors have a dramatic
reduction in expression of the co-inhibitory receptor PD-1 within CD8+ T cells (P=0.00042),
along with an increase in the functional intracellular cytokine TNFα (P=0.0374), all
consistent with activated CD8+ T cell cytotoxic function (Figures 5F, S9F and S9G). No
change is evident in expression of other co-inhibitory receptors (TIGIT, LAG3 or TIM3) or
cytokine (IFNγ) in CD8+ T cells (Figures S9H–S9K).
Thus, ectopically restored expression of either CD1 or IFI16 murine orthologs in cancer
cells with DNA hypomethylation-induced silencing suppresses tumor formation, a finding
only evident in immune competent mice, and associated with evidence of selectively
increased anti-tumor activity. Our results indicate that at least two distinct immune
populations are impaired by silencing of the CD1A-IFI16 locus (NKT cells modulated
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 12
by Cd1d1 and cytotoxic CD8+ T cells affected by Ifi204), suggesting a complex immune-
modulatory function of this multigene locus in tumorigenesis.
Detection of CTC-derived DNA hypomethylation in blood specimens using Nanopore
sequencing
While our study was focused on the characterization of early methylation changes in
prostate tumorigenesis and their potential biological consequences, we also note the recent
application of CpG island hypermethylation as a blood-based diagnostic assay for early
cancer detection24,25. Genome-wide screening for changes in DNA methylation may be
more sensitive than mutation-based assays, particularly in tumors like prostate cancers,
which do not harbor well defined recurrent driver mutations. Nonetheless among all
cancers tested, early prostate cancer shows one of the lowest detection rates (11.2%), using
screening for CpG island hypermethylation25. CTCs are shed into the blood by invasive
localized prostate cancers long before they establish metastases51–53, raising the possibility
that they may provide an orthogonal assay for early cancer detection. Given the specificity
of DNA hypomethylation domains in cancer cells and their large genomic size, we reasoned
that they may provide high sensitivity and quantitative signal for cancer detection, following
CTC enrichment in blood specimens. For such blood-based rare cell signal detection studies,
we applied a screen for all prostate PMDs, rather than the much smaller number of core
PMDs, so as to increase coverage to a large fraction of the prostate cancer genome.
Oxford Nanopore long-read native sequencing typically produces sequencing reads up to
100 kb, and directly identifies methylated CpG residues (5mC), without requiring bisulfite
conversion in library preparation54,55. In its current configuration, Nanopore signal analysis
does not readily identify 5-hydroxymethyl cytosines (5hmC), which are considerably less
abundant than 5mC, and are also not distinguished from 5mC in conventional bisulfite
sequencing. Indeed, Nanopore sequencing of the VCaP prostate cancer cell line clearly
defines DNA hypomethylation domains, which faithfully recapitulate those identified in
these cells using standard bisulfite sequencing (Figures 6A and 6B).In contrast to the
short Illumina sequencing reads (usually harboring <5 CpG sites per read), mathematical
modeling indicates that the long reads generated by Nanopore sequencing would empower
detection with significantly higher precision for rare signal (Figures 6C and 6D, see
Methods). We therefore processed 10 ml blood specimens from patients with either localized
or metastatic prostate cancer, using microfluidic enrichment to deplete leukocytes (104-fold
depletion), but without further CTC purification or individual CTC micromanipulation
(Figure 6E, see Methods). While 23 age-matched healthy donors (HDs) show minimal
DNA hypomethylation signal (<0.6%), 6 out of 7 (86%) patients with metastatic prostate
cancer have significant signal (from 0.62% to 11.08% of sequencing reads, P=0.00011), as
do 6/16 (37.5%) patients with localized prostate cancer (from 0.62% to 2.29% of sequencing
reads, P=0.004) (Figures 6F and 6G,Tables S4 and S5). Thus, long-range hypomethylated
domains are universal characteristics of prostate cancer and they are detectable from rare
CTCs in patient-derived blood specimens. The simplicity and cost effectiveness of Nanopore
sequencing raises the possibility of hypomethylation-based cancer detection.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Discussion
Page 13
Using single-cell DNA methylation analysis, ranging from indolent low grade localized
prostate cancer to metastatic CTCs, we annotated at high resolution the shared
hypomethylation domains that constitute core PMDs, along with interspersed islands with
preserved methylation, that we identify here as PMIs. PMDs are known to be associated
with the peripheral and transcriptionally silenced B compartment of the nucleus13,56, raising
the possibility that PMIs loop into the active A compartment regions, and hence are spatially
distinct from the surrounding silenced chromatin. Given intercellular heterogeneity, the
denotation of core PMDs was derived from the intersection of PMDs across many single
cells from multiple independent prostate cancers. However, these core PMDs also stand
out by virtue of their detection in the earliest low grade prostate cancers (GS6), leading
to the suggestion that they are driven by early selective pressures in tumorigenesis, and
explaining their universal silencing in advanced prostate cancers. Indeed, silencing within
core hypomethylation domains appear to target immune-related genes, including a single
chromosomal locus containing the entire family of CD1 genes and a cluster of interferon-
inducible genes. PMIs, in contrast, preserve expression of proliferation-associated genes
implicated in cell-cycle and DNA damage repair pathways. DNA methylation changes may
thus convey a selective advantage in prostate cancer development, suppressing expression of
genes contributing to immune surveillance of nascent tumors, while shielding neighboring
genes that enhance cell proliferation. Such selective pressures could drive the very early
targeting of the immune-rich CD1A-IFI16 locus, as demonstrated by in vivo reconstitution
experiments in mouse models. While early PMDs, like the CD1A-IFI16 locus, may emerge
solely from selection pressures favoring proliferating prostate cells that escape immune
surveillance, it is also possible that such loci have intrinsic properties favoring early loss of
DNA methylation.
Early hypomethylation of core PMDs
The model that hypomethylation-associated gene silencing occurs early and favors
tumorigenesis differs conceptually from a hypothesis proposed from a study of advanced
colon cancers, whereby hypomethylation might serve an intrinsic tumor suppressor
mechanism, restraining uncontrolled cell proliferation13. Of note, the colon cancer study
analyzed bulk tumor material, encompassing cancer cells together with reactive stroma and
immune cells, and it therefore excluded from analysis immune-related genes, whose cell-of-
origin is confounded by whole-tumor sequencing. Single-cell level analysis thus allows
assignment of all changes in DNA methylation to the appropriate cell type. Most important,
however, is our definition of a small subset of PMDs, annotated as core PMDs (2.7% of
all PMDs), that appear early in tumorigenesis and are shared uniformly across multiple
independent tumors. The identification of early cancer drivers targeted by epigenetic
silencing is likely to differ from the contribution of additional PMD-encoded genes that
are silenced during subsequent cancer progression, as DNA hypomethylation extends across
major portions of the genome. Compared with the small number of core PMDs identified
in early cancers, the very large fraction of the cancer genome that is hypomethylated
in advanced tumors may thus reflect distinct selection pressures, as well as bystander
effects affecting gene-poor PMDs and the derepression of repetitive elements. While our
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 14
study was centered on prostate cancer, the relevance of core PMDs extends to other
cancers, as illustrated by TCGA analyses showing their consistent early hypomethylation
across multiple tumors, in contrast to most PMDs which show considerable inter-tumor
heterogeneity. Indeed, TCGA methylation data shows that the CD1A-IFI16 locus to have
the strongest difference in DNA methylation between 33 different cancers and their normal
tissue counterparts. This specific locus, encoding immune-related genes that have not been
previously nominated as critical cancer genes, thus appears to be a consistent target of
epigenetic silencing in the early stages of tumorigenesis. Our functional assays using the
demethylating agent 5-azacytidine and the EZH2 inhibitor GSK126 support the recruitment
of chromatin silencing marks to hypomethylated PMDs as a mechanism of transcriptional
silencing. However, further studies will be required to better understand the selectivity
of PMD hypomethylation across the genome, and both genomic structure and selection
pressures that distinguish core PMDs from more global demethylation.
The CD1A-IFI16 immune gene cluster
The CD1A-IFI16 locus is unique in encompassing the entire gene family of CD1 genes,
which together mediate lipid antigen presentation, together with the IFI16 class of
interferon-inducible genes. It is well established that genes that are co-located within
a single genomic locus may be targeted during tumorigenesis by either chromosomal
deletions or amplification events, a single genetic event that may mediate simultaneous
loss-of-function or gain-of-function among physically clustered genes. Conceptually, the
hypomethylation silencing of the CD1A-IFI16 locus during early prostate tumorigenesis
may accomplish a similar function, suppressing T cell recognition of lipid antigens as
well as double stranded DNA sensing, as part of a single epigenetic event affecting both
alleles. Such a potent selective pressure could explain the early and frequent targeting of
this locus in cancer. The 1q23.1 genomic locus has been linked in germline association
studies to neurodegenerative disease and autoimmune diseases57,58, and immunological
pathways regulated by its resident genes have been linked to innate immunity against
infectious pathogens. The potential roles of these genes in immune surveillance of early
cancers will require further functional analyses. Alterations in antigen presentation pathways
constitute the most critical mechanisms by which tumors evade both innate and therapeutic
immune activation59,60. In this respect, the presentation of lipid antigens to NKT cells,
a highly specialized subpopulation of T cells, is of particular interest, given potential
therapeutic implications. Within prostate cancer, the silencing of CD1A-IFI16 genes is also
noteworthy in that it points to tumor cell-intrinsic factors contributing to the escape from
immune surveillance, in addition to the proposed immunosuppressive effects of the tumor
microenvironment.
Diagnostic implications
Finally, from a cancer diagnostic standpoint, blood-based detection of early invasive cancers
remains a major technological challenge. For prostate cancer, it requires the ability to
distinguish between indolent lesions associated with non-specific elevations in serum PSA
and more aggressive cancers that may have similar serum PSA levels but warrant therapeutic
intervention. Early invasive prostate cancers shed CTCs into the circulation long before
metastases are established51–53, and while these rare early CTCs may not be sufficient to
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 15
cause dissemination, they can serve as potential biomarkers of invasive disease. Microscopic
imaging of very rare CTCs in the bloodstream is challenging, hence there is a need for
sensitive and quantitative molecular readouts applied to CTC-enriched blood specimens.
While this study was not designed to formally test Nanopore sequencing of PMDs as a
quantitative molecular surrogate of CTCs, it suggests that such long-range DNA sequencing
strategies may complement current approaches that rely on hypermethylation of CpG
islands within short ctDNA fragments. Such approaches may also enhance tissue-of-origin
determinations, given the information content inherent in such long-range genomic analyses.
Limitations of the study
Our study suggests that early hypomethylation of core PMDs in prostate cancer
differentially silences immune surveillance-associated genes, while sparing genes that
mediate cell proliferation. While we find shared patterns of core PMDs across multiple
different cancers, it is also possible that distinct tumor types will target alternative
biologically relevant pathways. Additional studies in different early stage cancers will be
required to distinguish shared hypomethylation targets from those showing tissue-specific
patterns, and additional patient-derived samples will need to be analyzed within each tumor
type. The potential roles in immune surveillance of lipid antigen presentation genes and
IFI16-related double stranded DNA sensing genes deserves further functional analyses using
additional experimental systems to define their relevance in early tumorigenesis, as well
as their potential relevance for anti-cancer therapy. Finally, the potential utility of PMD
detection in blood-based cancer diagnostics will require further validation in larger numbers
of diverse clinical specimens.
STAR Methods
Resource availability
Lead Contact—Further information required to reanalyze the data reported in this paper
and requests for resources and reagents should be directed to and will be fulfilled by the lead
contact, Daniel A. Haber ([email protected]).
Material Availability—Plasmids generated in this study are available upon written request.
Data and Code availability
•
•
All raw and processed sequencing data in this study, including single-cell DNA
methylation sequencing, single-cell RNA-seq, ChIP-seq, Cut and Run assay
and Nanopore sequencing, have been deposited to the NCBI Gene Expression
Omnibus (GEO) database under accession GSE208449. All data are publicly
available as of the date of publication.
This paper analyses existing, publicly available data or available upon request
to the authors. These accession numbers for the datasets are listed in the key
resources table.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 16
•
•
•
This paper does not report original code. All the scripts and mathematical
algorithms used in this study will be available from the corresponding authors
upon request.
All the versions of software packages used in this study are listed in the key
resource table and noted in the data analysis method accordingly.
Any additional information required to reanalyze the data reported in this paper
is available from the lead contact upon request.
Experimental model
Clinical Specimens—All patient samples were collected in this study after written
informed consent, in accordance with Institutional Review Board (IRB) protocols (DF/HCC
05–300, 11–497, 13–217 or 14–375). For the CTC cohort, 10–20 ml of blood was drawn
from patients with a diagnosis of metastatic prostate cancer, localized prostate cancer, or
age-matched males without a diagnosis of cancer at Massachusetts General hospital (MGH).
For the localized tumor tissue cohort, all samples were acquired from either core biopsies
or surgical resection of untreated localized prostatic adenocarcinoma (Gleason scores 6
and 8) from patients at MGH. In cases with the lowest grade tumors (Gleason score 6),
normal prostate tissue was also identified in the tissue specimen by a Genito-Urinary (GU)
specialized pathologist and used as a source of matched normal prostate cells. Both normal
and tumor tissue samples were de-identified, snap frozen and sectioned. Only tumor sections
with >80% tumor content, as assessed by a specialized GU pathologist were used in this
study. The clinical data of the patients with metastatic prostate cancer enrolled in the
single-cell CTC analysis and patients with resected localized prostate cancer used for single
nucleus analysis are described in Table S1. The clinical data of the patients with localized
prostate cancer and metastatic prostate cancer enrolled in Nanopore sequencing anlysis of
CTC-enriched blood are described respectively in Table S4 and Table S5.
Cell culture—Human prostate cancer cell lines (LNCaP, VCaP, PC3 and 22Rv1), murine
prostate cancer line (Myc-CaP), normal cultured prostate epithelial cells (HPrEC), benign
prostatic hypertrophy cells (BPH-1) and murine Lewis lung carcinoma cells (LLC-1) were
all obtained from ATCC, after authentication by short tandem repeat (STR) profiling. All
cell lines used in the paper were derived from male mice or male human patients. They
were cultured in the following media at 37°: RPMI-1640 (ATCC) medium supplemented
with 10% FBS (Gibco) and 1X Pen/Strep (Gibco) (for LNCaP, VCaP, PC3, 22Rv1 and
BPH-1 cells); Prostate Epithelial Cell medium (ATCC) with 6 nM L-glutamine (ATCC),
0.4% Extract P (ATCC), 1.0 mM Epinephrine (ATCC), 0.5 ng/ml rh-TGFα (ATCC),
100ng/ml hydrocortisone hemisuccinate (ATCC), 5 mg/ml rh-Insulin (ATCC), 5 mg/ml
Apo-transferrin (ATCC), 33 μM Phenol red (ATCC) and 1X Pen/Strep/Ampho Solution
(ATCC) (for HPrEC cells); DMEM high glucose medium (Gibco) with 10% FBS (Gibco)
and 1X Pen/Strep (Gibco) (for Myc-CaP cells and LLC-1 cells). All the cell lines used in
this study were checked for mycoplasma every 4 months using Mycoalert kit (Lonza).
Mouse xenograft assays—All animal experiments were carried out in accordance with
approved protocols by the MGH Subcommittee on Research Animal Care (IACUC). All
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 17
the mice used in this study were maintained under a 12/12 h light/dark cycle in MGH
animal facility. 6–8 weeks old FVB male mice (Jackson Laboratory, Strain#001800) or 6–8
weeks old male immunodeficient NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) mice (Jackson
Laboratory, Strain#005557) were used for intraprostatic injection or subcutaneous injection
of Myc-CaP cells stably expressing luciferase and mCherry. 6–8 weeks old C57BL/6 female
mice (Jackson Laboratory, Strain#000664) were used for subcutaneous injection of LLC-1
cells stably expressing luciferase. Littermates of the same sex were randomly assigned
to experimental groups. For intraprostatic inoculation, mice were first anesthetized using
isoflurane, and a 1 cm skin incision was performed along the midline of the abdomen
to expose the inner muscle layer, which was then separated. The tip of seminal vesicle
was raised gently with forceps to expose the anterior lobe of the prostate gland. 50,000
Myc-CaP cells 1:1 mixed with Matrigel (v/v) (total volume: 30 μl) were slowly injected
into the prostate lobe. All the tissues were then returned into the abdomen, and continuous
sutures were used to close the inner muscle layer, followed by separate skin closure. For
subcutaneous injections, mice were anesthetized, and 50,000 Myc-CaP cells or 1,000,000
LLC-1 cells 1:1 mixed with Matrigel (v/v) (total volume: 100 μl) were injected into the
flank. Tumor cell-derived bioluminescent signal was quantified every other day for the
Myc-CaP cells and 3 times a week for the LLC-1 for mice after either orthotopic injection
or subcutaneous injection. At 2–3 weeks after inoculation, mice were sacrificed and tumors
were harvested for flow cytometry and RNA extraction for the Myc-CaP experiments.
Method Details
CTC isolation—CTCs were isolated from fresh blood specimens drawn from patients
with prostate cancer, following negative depletion of leukocytes using the microfluidic
CTC-iChip as reported previously26,27. Briefly, 10–20 ml of whole blood specimens were
incubated with biotinylated antibody cocktails against CD45 (R&D Systems, clone 2D1),
CD66b (AbD Serotec, clone 80H3), and CD16 (BD Biosciences), followed by incubation
with Dynabeads MyOne Streptavidin T1 (Invitrogen) for magnetic labeling and depletion of
leukocytes. After CTC-iChip processing, the CTC-enriched product was further stained with
FITC-conjugated antibody against EpCAM (Cell Signaling Technology, clone VU1D9) and
PE-conjugated antibody against CD45 (BD Biosciences, clone HI30). Single CTCs (FITC
positive and PE negative) or white blood cells (WBCs, FITC negative and PE positive)
were individually picked into PCR tubes containing 5 μl RNA/DNA lysis buffer using
micromanipulator (Eppendorf TransferMan NK 2) and snap-frozen in liquid nitrogen. In
total, 38 CTCs from 5 different patients (GU114, GU169, GU181, GU216 and GURa15)
with metastatic prostate cancer were individually picked, sequenced and lineage-confirmed
based on transcriptome and DNA copy number. One patient sample (GU169) had only one
CTC, and it was therefore excluded from some downstream analyses focused on the four
patients with multiple CTCs.
Nuclei isolation from frozen tumor sections—Tumor tissue sections with high tumor
content (>80%) and adjacent normal tissue section were micro-dissected and transferred
into a pre-chilled Dounce homogenizer containing ice-cold 1 ml 1X HB buffer (0.26
M sucrose, 30 mM KCl, 10 mM MgCl2, 20 mM Tricine-KOH, 1 mM DTT, 0.5 mM
Spermidine, 0.15 mM Spermine, 0.3% NP-40 and 1X complete protease inhibitor). Tissue
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 18
was homogenized with ~10 strokes of “A” loose pestle, followed by another ~10 strokes
of “B” tight pestle. The tissue homogenate was then filtered using a 70 μm strainer and
pelleted by centrifugation. Nuclear pellets were resuspended and purified by density gradient
centrifugation (top layer: 25% Iodixanol solution; middle layer: 30% Iodixanol solution;
bottom layer: 40% Iodixanol solution). The nuclear band at the interface of 30% and 40%
Iodixanol solutions was collected into a new Eppendorf tube and washed twice with ice-cold
1X PBS. 20% of the purified nuclei were used to isolate single nuclei using fluorescence-
activated cell sorting (FACS) for single-cell DNA methylation analysis, while the remaining
80% of the nuclei were subjected to ChIP-seq analysis.
Western Blot—Cells or tumor tissues were lysed in Laemmli buffer (Sigma) and cleared.
Protein concentration was determined using DC protein assay (Bio-rad). Proteins (25
μg) were separated on precast NuPAGE 4–12% Bis-Tris protein gels (ThermoFisher),
and transferred onto nitrocellulose membranes (Bio-Rad). After blocking with 5% BSA
buffer for 1 hour at room temperature, membranes were incubated with primary antibodies
overnight at the recommended concentrations. HRP conjugated secondary antibodies
(1:10,000; Bio-rad; Cat#5196–2504) were applied, and ultra-sensitive autoradiography film
(Amersham) was used to detect the chemiluminescence signal. Primary antibodies used
are H3K27me3 (1:1,000, Invitrogen Cat#MA5–11198) and H3 total (1:1,000, Abcam
Cat#1791).
5-Azacytidine treatment, bisulfite sequencing and staining of chromatin marks
—The human prostate epithelial cell line BPH-1 was cultured in the presence of 5 μM 5-
azacitidine (Selleck, #S1782). At serial time points (days 0, 1, 4 and 5), cells were collected
for DNA extraction, confocal microscopy, or flow cytometric analysis. DMSO-treated cells
were used as control at each time point. To quantify 5-azacitidine-induced demethylation at
the genomewide level, we used the whole genome bisulfite sequencing (WGBS). Briefly,
DNA ws extracted from BPH-1 cells upon 5-azacitidine treatment, 1 μg genomic DNA was
used to sonicate into 300–500 bp fragments, DNA was end-polished, A-tailed and ligated
with pre-methylated adaters before bisulfite conversion using EZ DNA methylation kit
(Zymo, #D5001), bisulfite-converted DNA was amplified and sample index was introduced
during amplification. To quantify 5-azacytidine-induced demethylation at the CD1A-IFI16
locus, DNA extracted from BPH-1 cells treated with 5-azacitidine was subjected to bisulfite
conversion using EZ DNA methylation kit (Zymo, #D5001), and bisulfite-converted DNA
was used for PCR amplification, applying bisulfite-specific PCR primers covering the
human CD1A-IFI16 locus (see Table S3). PCR products were purified by 1% agarose
gel and cloned using the Zero blunt PCR cloning kit (ThermoFisher, #K270020). 10
individual bacterial clones were randomly picked for Sanger sequencing. Sequencing data
were analyzed and shown using online tool QUMA (http://quma.cdb.riken.jp/)61. Nuclear
accumulation of H3K27me3 was stained with H3K27me3 antibody (1:1000 dilution;
CST#9733), in 5-azacytidine-treated cells. Images were acquired using a Zeiss LSM710
Lase Scanning Confocal and were quantified by quantitative image analysis of cells
(ImageJ). Flow cytometry was also performed at serial time points on BD LSRFortessa
machine to assess CD1d expression using human CD1d-APC antibody (1:100 dilution;
BioLegend#350308, clone: 51.1).
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 19
EZH2 inhibitor treatment—Human prostate cancer cell lines (22Rv1, LNCaP and VCaP)
were cultured in the presence of the small molecule EZH2 inhibitor GSK126 (Selleckchem,
#S7061) at the indicated concentration (0, 5 or 10μM). After 6 days of treatment, protein and
RNA were harvested, for quantitation of H3K27me3 and total H3, using Western blotting
and expression of individual genes within the CD1A-IFI16 locus by real time qPCR.
Paired single-cell DNA methylation and RNA-seq—For these experiments, we used
either single CTCs or WBCs individually picked from fresh blood specimens after CTC
enrichment, and single cells from cultured prostate cell lines (either picked or FACS-sorted).
These were subjected to paired single-cell DNA methylation and RNA-seq analysis to
obtain the transcriptomes and DNA methylomes from the same single cells33,62. Briefly,
single cells were first lysed in 5 μl DNA/RNA lysis buffer; 0.5 μl Magnetic MyOne
Carboxylic Acid Beads (Invitrogen, Cat#65011) were then added to each single cell lysate
to facilitate segregation of nucleus versus cytoplasm. After centrifugation and magnetic
separation, the supernatant (containing cytoplasmic RNA) was transferred into a new tube
for single-cell RNA-seq amplification using the SMART-seq2 protocol63, while the pellet
(aggregated beads with the intact nucleus) was resuspended in DNA methylation lysis buffer
and subjected to single-cell whole genome methylation sequencing using the scBS-seq
protocol64. Single nuclei sorted from the frozen primary prostate tumor sections were also
subjected to the scBS-seq procedure.
MNase native ChIP-seq—Purified nuclei from frozen tissue sections were subjected to
MNase native ChIP-seq following the ULI NChIP procedure, as published elsewhere65.
Briefly, nuclei were suspended in Nuclear Isolation Buffer (Sigma) supplemented with
1% TritonX 100, 1% Deoxycholate and 1X complete protease inhibitor. Chromatin was
digested by MNase enzyme (NEB, 1:10 diluted) at 21°C for 7.5 min, and further diluted
in Complete Immunoprecipitation Buffer, with 1X complete protease inhibitor. 2 μl ChIP-
grade H3K27me3 (Active motif, Cat#39155) or H3K9me3 (Abcam, Cat#ab8898) antibody
was incubated with the digested chromatin overnight at 4°C. DNA was then purified
using protease K digestion followed by phenol-chloroform extraction. ChIP-seq sequencing
libraries were prepared using NEBNext Ultra II DNA Library Prep Kit (NEB, Cat#E7645L).
Cut and Run Assay—H3K27me3 and H3K9me3 Cut and Run assays were performed
with cultured prostate cell lines (LNCaP, 22Rv1, BPH-1, HPrEC and Myc-CaP), using
the CUT&RUN Assay kit (CST, Cat#86652S). Briefly, 100,000 freshly cultured prostate
cells were collected and incubated with Concanavalin A Magnetic Beads. 2 μl ChIP-grade
H3K27me3 (Active motif, Cat#39155) or H3K9me3 (Abcam, Cat#ab8898) or IgG (CST,
Cat#66362S) antibody was added to the cell: bead suspension and incubated overnight at
4°C. 1.5 μl pAG-MNase enzyme was then added to the tube, which was rotated for 1 h at
4°C, followed by activation of pAG-MNase using 3 μl cold Calcium Chloride. The activation
reaction was stopped and DNA was further diluted and collected for phenol-chloroform
extraction. Cut and Run sequencing libraries were constructed using NEBNext Ultra II DNA
Library Prep Kit (NEB, Cat#E7645L).
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 20
Next generation sequencing—All the single-cell RNA-seq, single-cell DNA
methylation, MNase ChIP-seq, Cut and Run samples and WGBS samples were molecularly
barcoded, pooled together and sequenced on a HiSeq X sequencer to obtain 150 bp pair-
ended reads (Novogene).
RNA extraction, reverse transcription and quantitative PCR (qPCR)—RNA
extracted from cultured prostate cells was prepared using the RNeasy Mini kit (QIAGEN)
with DNase I digestion on the column. To extract RNA from mouse tumor tissues, these
were first dissected to remove connective tissue and fat, and washed extensively with 1X
PBS to remove excessive blood or necrotic tissues. Tumors were then homogenized in RLT
RNA lysis buffer using a Dounce homogenizer, and passed through a QIAshredder column
(QIAGEN). RNA from normal prostate of FVB mice were prepared following a similar
method. RNA from tissue homogenate was extracted using the RNeasy Mini kit (QIAGEN)
with DNase I digestion on the column. cDNA was synthesized from 50–200 ng RNA using
SuperScript III One-Step qRT-PCR kit (Invitrogen). qPCR was performed using the primers
listed in Table S3.
CD1d expression measurement by flow cytometry—Cell surface protein expression
of CD1d in human and mouse prostate cells was assessed by flow cytometry. Cells were
first trypsinized, and 500,000 cells were used for staining with antibody against CD1d at
4°C for 20 min, followed by washing and quantitation using a BD LSRFortessa machine,
and data were analyzed using FlowJo software (v10.4; https://www.flowjo.com/). Antibodies
used were as follows: for human prostate cell lines, APC conjugated anti-human CD1d
(BD#563505, clone: CD1d42) and APC-conjugated isotype control (BD#555751); for Myc-
CaP cells, anti-mouse CD1d (Bio X Cell #BE0179, clone 20H2) and the isotype control (Bio
X Cell #BE0088), and secondary antibody anti-rat IgG conjugated with APC (Invitrogen
#A10540).
Plasmid construction—A lentiviral murine Cd1d1 expression construct (pLenti-
Cd1d1-mGFP, Cat#MR226027L4) and its matched control construct (pLenti-C-mGFP,
Cat#PS100093) were obtained from Origene. Murine Ifi204 expression vector (pLenti-
Ifi204-Myc-DDK-Puro, Cat#MR222527L3), together with its control vector (pLenti-C-
Myc-DDK-Puro, Cat#PS100092) were also purchased from Origene, and the puromycin
selection cassette of these two Origene plasmids were replaced by blasticidin from
lentiCRISPRv2-blast plasmid (Addgene#98293) using NEBuilder HiFi DNA Assembly
Cloning kit (NEB, Cat#E5520S). For the LLC-1 experiment, the murine Cd1d1 was cloned
into the receiving vector N174-MCS (Addgene#81061) with the restriction enzymes EcoR1
and Mlu1, using the FastDigest protocol of Thermo Scientific. All final construct sequences
were confirmed by Sanger sequencing. Plasmids generated in this study are available upon
written request.
Lentiviral transduction—Early passage 293T cells were transfected with Cd1d1
or Ifi204 lentiviral constructs, together with pMD2.G (Addgene#12259) and psPAX2
(Addgene#12260) packaging plasmids using Lipofectamine 2000 reagent (Invitrogen). 48–
72 h after transfection, culture medium (containing lentiviral particles) was collected,
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 21
filtered and concentrated using LentiX concentrator (Clontech). Concentrated virus was
added to the Myc-CaP cells in presence of polybrene (Santa Cruz, 8 μg/ml as final
concentration) overnight. FACS was used to select GFP positive cells as marker of Cd1d1
construct transduction in the Myc-CaP cells. The LLC-1 cells transduced with the Cd1d1
cloned in the the N174-MCS vector were selected using G418 (Sigma Aldrich #G8168)
at 400 μg/mL for 4–6 days. To obtain stable Ifi204 overexpression, 10 μg/ml blasticidin
(InvivoGen) was added to the medium for 5–7 days selection.
Tumor immune infiltration assayed by flow cytometry—Mouse tumors generated
by intraprostatic injection of control or Cd1d1-expressing Myc-CaP cells were dissected
and washed to remove blood, fat and connective tissues. Tumor tissues were further
mashed and digested in 5 ml digestion buffer (RPMI1640, 2.5 mg/ml collagenase D,
0.1 mg/ml DNase I) at 37°C for 30 min. Tissue digestion was stopped by adding
another 5 ml RPMI1640 with 2% FBS, and then filtered through 70 μm strainers.
The tissue cell suspension was obtained in the same way for tumors generated by
subcutaneous injection of control or Ifi204 expressing Myc-CaP cells. To stain for
NKT cell infiltration in prostate tumors with control or Cd1d1 expression, the single-
cell suspension was first blocked with rat anti-mouse CD16/CD32 blocking reagent
(BD#553142, Clone: 2.4G2) at 4°C for 30 min, followed by mouse NKT surface antibody
cocktail staining at 4°C for another 30 min. The mouse NKT surface antibodies used
in this study were: BV510-viability dye (BD#564406), APC-α-GalCer-mCD1d Tetramer
(TetramerShop#MCD1d-001), BV711-CD69 (BioLegend#104537, clone: H1.2F3), PerCP-
Cy5.5-TCRβ (BioLegend#109228, clone: H57–597), BV605-CD3e (BioLegend#100351,
clone: 145–2C11) and BUV395-NK1.1 (BD#564144, clone: PK136). Cells obtained from
mouse tumors with control or Ifi204 expression were split into two fractions, with the first
fraction stained using a panel of mouse T cell surface antibody cocktails: BV510-viability
dye (BD#564406), PerCP-Cy5.5-TCRβ (Biolegend#109228, clone: H57–597), BV711-CD8
(Biolegend#100759, clone: 53–6.7), BV650-CD4 (Biolegend#100546, clone: RM4–5),
FITC-CD44 (Biolegend#103006, clone: IM7), PE-Cy7-PD-1 (Biolegend#109110, clone:
RMP1–30), BV421-TIM3 (BD#747626, clone: 5D12), APC-TIGIT (Biolegend#156106,
clone: 4D4/mTIGIT) and BV785-LAG3 (Biolegend#125219, clone:C9B7W). The second
fraction was used to stain for surface and intracellular cytokines by first activating
cells with Cell Stimulation Cocktail (eBioscience#00–4970-93) together with Protein
Transport Inhibitor Cocktail (eBioscience#00–4980) in 37°C cell culture incubator for
4 h. The cells were then stained for surface antigens before fixation, and subsequently
processed for intracellular cytokine staining using BD Fixation/Permeabilization Solution
Kit (BD#554714). Antibody cocktails used for surface and intracellular cytokine staining
were: BV510-viability dye (BD#564406), PerCP-Cy5.5-TCRβ (Biolegend#109228, clone:
H57–597), FITC-CD44 (Biolegend#103006, clone: IM7), PE-TNFα (Biolegend#506306,
clone: MP6-XT22), BV650-CD4 (Biolegend#100546, clone: RM4–5), BV711-CD8
(Biolegend#100759, clone: 53–6.7) and BV605-IFNγ (Biolegend#505840, clone: XMG1.2).
All flow cytometry was done on the BD LSRFortessa machine, and data were analyzed
using FlowJo software (v10.4; https://www.flowjo.com/).
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 22
Multiplex Oxford Nanopore native sequencing—Blood samples from either healthy
donors or patients with localized or metastatic prostate cancer were subjected to CTC-ichip
enrichment (104-fold leukocyte depletion)26,27. The enriched CTCs (ranging from 0.1%
to 1% purity, admixed with residual leukocytes) were subjected to high molecule weight
(HMW) DNA extraction using the HMW DNA extraction kit (QIAGEN), and then prepared
for Oxford Nanopore sequencing using the rapid barcoding kit (Nanopore#SQK-RBK004).
For each sequencing run, 11 blood samples (either from healthy donors or cancer patients),
together with 1 lambda DNA (unmethylated control), were uniquely barcoded and pooled
together. Sequencing was performed using a Nanopore MinION device with R9.4 flowcell
for 48 h, per manufacturer instructions.
Single-cell and bulk RNA-seq data analysis—Raw fastq reads generated from HiSeq
X sequencer were first cleaned using TrimGalore (v0.4.3) (https://github.com/FelixKrueger/
TrimGalore) to remove the adapter-polluted reads and reads with low sequencing quality.
Cleaned reads were aligned to the human (hg19) or mouse (mm9) genome using Tophat
(v2.1.1)66. PCR duplicates were further removed using samtools (v1.3.1)67, gene counts
were computed using HTseq (v0.6.1)68, gene expression level (FPKM) was further
calculated using cufflinks (v2.1.1)66. Gene expression matrix was subjected to R (v3.1.2)
or Prism9 for graphics.
Single-cell and bulk DNA methylation sequencing data analysis—Raw fastq
reads from both the single-cell and bulk DNA methylation sequencing were first trimmed
using TrimGalore (v0.4.3) (https://github.com/FelixKrueger/TrimGalore), and cleaned reads
were aligned to the human hg19 or mouse mm9 genome (in silico bisulfite converted) using
Bismark tool (v0.17.0)69. Samtools (v1.3.1)67 was used to remove PCR duplicates, and CpG
methylation calls were extracted using the Bismark methylation extractor69. 0.1% lambda
DNA was spiked in, prior to bisulfite treatment, for each sample to assess the bisulfite
conversion efficiency. Only samples with more than 4 million unique CpG sites covered at
least once and with a bisulfite conversion rate > 98% were used in this study.
TCGA methylation array data reanalysis—Prostate DNA methylation datasets
from TCGA analyzed by Illumina Infinium Human Methylation 450 K BeadChip
were downloaded from the National Cancer Institute’s GDC Data Portal (https://
portal.gdc.cancer.gov) for 502 tumor samples and 50 normal samples. CpG site-level
methylation files (beta value, txt format) were first converted to hg19 coordinates
using UCSC lift-over tool (https://genome.ucsc.edu/cgi-bin/hgLiftOver) for the downstream
analysis. The data were binned to a fixed set of 10 kb nonoverlapping genomic windows
by computing the average fraction methylation within each bin in each sample. Bins
were excluded if they lacked coverage (i.e., had no probes on the Illumina Infinium
Human Methylation 450 K BeadChip array) or had a mean normal-tissue methylation level,
averaged across all the normal samples, of <70%. For each sample, the global methylation
level was calculated as the fraction of bins having methylation >50%. The methylation level
at the CD1A-IFI16 locus for each sample was calculated as the fraction of bins in the range
chr1:158,130,000–158,340,000 (hg19) having methylation >50%. The gene expression data
and clinical information of TCGA PRAD samples, including Gleason score, tumor stage
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 23
and others, were all downloaded from cbioportal (https://www.cbioportal.org/). Tumor purity
was calculated using ABSOLUTE algorithm70. DNA Methylation 450 K BeadChip datasets
for other cancer types were also downloaded from the National Cancer Institute’s GDC
Data Portal (https://portal.gdc.cancer.gov) and CpG site-level methylation files (beta value,
txt format) were also converted to hg19 coordinates using UCSC lift-over tool (https://
genome.ucsc.edu/cgi-bin/hgLiftOver) for the downstream analysis.
Genomic element enrichment analysis—For analytical purposes, a promoter region
was defined based on the relative position to a transcription start site (TSS): 1,500 bp
upstream and 500 bp downstream. The annotations of TSS, exon, intron, intragenic regions,
CpG islands (CGIs), repetitive elements and UCSC gap regions were all downloaded from
UCSC genome table browser (https://genome.ucsc.edu/cgi-bin/hgTables)71. Enrichment
analysis on different genomic elements was calculated using the Bioconductor package
regioneR (v1.18.1) with overlapPermTest function72.
DNA copy number analysis inferred by single-cell DNA methylation
sequencing data—Single-cell DNA methylation sequencing reads were first aligned to
the genome using Bismark. Uniquely aligned reads were extracted into a bed file and
subsequently submitted to Ginkgo online tool73, http://qb.cshl.edu/ginkgo) to infer the DNA
copy number, using 5 Mb as the bin size. The processed integer copy number data from
the Ginkgo website (SegCopy.tsv) was used to calculate the DNA Copy Number Variation
(CNV) score. Given an assignment of a copy number to all the locations in a diploid
genome, we define a CNV score for any given single cells as follows. Let ci be the copy
number at the ith location of the genome. CNV score is then defined to be the average over
all i in the genome of the absolute value of (ci-2).
DNA copy number analysis inferred by single-cell RNA-seq data—Single-cell
RNA-seq reads were aligned to human genome using TopHat, and large-scale chromosomal
copy number alterations were determined by InferCNV (https://github.com/broadinstitute/
infercnv).
MNase ChIP-seq and Cut and Run data analysis—ChIP-seq and Cut and Run reads
were first trimmed by Trim Galore (v0.4.3) (https://github.com/FelixKrueger/TrimGalore)
and then mapped to the human or mouse genome using BWA men74. Duplicated reads were
marked by sambamba75 and further removed using samtools67. MACS2 (v2.0.10)76 was
used to call the peaks and deepTools77 were used to compute the ChIP-seq or Cut and Run
signal around prostate PMDs.
Determination of Partially Methylated Domains (PMDs)—The human genome was
first binned into 100 kb windows placed at 200 bp offsets. Windows that intersected CGIs
or UCSC gap regions were discarded. For each source (i.e., single CTCs from patients with
prostate cancer, single WBCs from healthy donors, single cells from normal prostate or
prostate cancer cell lines or normal prostate tissues42, the per-source methylation level of
each window was calculated by taking the average over all cells from that source of the
methylation level of the CpG sites within the given window. For each source the distribution
of the per-source methylation level of the 100 kb windows was plotted. Normal cells
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 24
showed a unimodal distribution, while prostate cancer cells showed a bimodal distribution.
A threshold for hypomethylation determination was set at the lowest point of the valley
in the histogram of the bimodal distribution for each prostate cancer patient or prostate
cell line; if the distribution was unimodal, the threshold was set to 60%. The windows
with methylation level lower than threshold were defined as hypomethylation windows
and overlapping hypomethylation windows were merged into per-source PMDs. The 250
kb minimal length threshold was then applied to the per-source PMDs. The union of the
per-source PMDs for all single CTCs from four prostate cancer patients (GU114, GU216,
GURa15 and GU181) and for all single cells from four prostate cancer cell lines (LNCaP,
VCaP, 22Rv1 and PC3) was defined as the total prostate PMDs (1,496 in total). Chromatin
mark and genome element enrichment analyses were performed on these PMDs. To identify
the genes that reside in the most consistently hypomethylated PMDs across all prostate
cancer specimens analyzed (i.e., intersection), we quantile-normalized the DNA methylation
levels for all PMDs among all CTCs from four prostate cancer patients (GU114, GU216,
GURa15 and GU181) and all single cells from four prostate cancer cell lines (LNCaP, VCaP,
22Rv1 and PC3) and only used the PMDs (annotated as core prostate PMDs) with their
averaged quantile-normalized DNA methylation level less than 25% across these 8 sources
to extract the genes.
Determination of Preserved Methylation Islands (PMIs)—After identification of
PMDs for each of the eight sample sources [CTCs from four prostate cancer patients
(GU114, GU216, GURa15 and GU181) and single cells from four prostate cancer cell
lines (LNCaP, VCaP, 22Rv1 and PC3)], we defined small interspersed islands (“gaps”) with
preserved methylation (sample source PMIs) using the following criteria: (1) every PMI is
flanked by defined PMDs in each given source; (2) length of each PMI should be >30 kb and
<3 Mb. Total prostate PMIs were defined by taking the union of sample source PMIs across
8 sources (1,412 in total), while core prostate PMIs (44 in total) were defined by requiring
the uniformity across sample sources: the genomic location of given PMI is overlapped in all
8 sample sources.
Differential gene expression and hypergeometric gene set enrichment
analysis (hGSEA)—Differential gene expression between TCGA prostate normal tissue
and primary tumors was determined as follows: We started by considering the genes that
reside in the most hypomethylated PMDs [as described in the section titled “Determination
of partially methylated domains (PMDs)”]. Of those, genes with 95th percentile of
normalized FPKM values less than 1 were discarded. A two-tailed variance-equal t-test
was performed on each of the remaining genes. The p-values from those t-tests were
used to generate a false-discovery rate (FDR) estimate for each gene by the Benjamini-
Hochberg method. We considered genes for which the FDR estimate was less than 0.1 to be
differentially expressed between normal prostate and prostate tumor samples. hGSEA was
performed to determine the gene set and pathway enrichment using the phyper R function as
reported elsewhere26. All gene sets and pathways evaluated in this study were obtained from
MSigDB (v7.2) from the Broad Institute. Differential gene expression and hGSEA for genes
in PMIs was performed in the same way.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 25
Heterogeneity assessment—Consistent with a previous publication26, means of
correlation coefficients and jackknife estimates were used to assess the heterogeneity within
and between subsets of samples.
Nanopore data analysis—Nanopore sequencing reads (format: fast5) generated by
Nanopore MinION device were first converted into fastq files using ONT Albacore software
(v2.3.1) (https://nanoporetech.com/community). Demultiplexing was also performed during
fast5 to fastq conversion. DNA methylation information was extracted from both fast5
and fastq files using Nanopolish software (v0.10.2) (https://github.com/nanoporetech/
nanopolish). Nanopolish output files (albacore_output.sorted.bam and methylation_calls.tsv)
were used for downstream analysis. Every nanopore run was spiked in with lambda DNA,
which was used as the negative control to assess the fidelity of Nanopore sequencing.
To estimate CTC-derived hypomethylation signal in each Nanopore sequencing sample,
stringent criteria were applied: (1) each Nanopore read should be long enough to harbor
at least 30 CpG sites with confident methylation calls after Nanopolish; (2) the number
of Nanopore reads aligned to prostate PMDs (pre-determined among CTCs isolated from
4 prostate cancer patients and 4 prostate cancer cell lines using single-cell whole genome
bisulfite sequencing) should be no fewer than 300 for metastatic patients or no fewer
than 400 for localized patients; (3) methylation level of spike-in lambda DNA in each run
should be <1%. Following application of these criteria, microfluidic processed (leukocyte-
depleted) blood samples from seven patients with metastatic prostate cancer, six patients
with localized prostate cancer. Since we required different number of Nanopore reads in the
prostate PMDs for metastatic patients and localized patients, 23 age-matched healthy donors
were validated for analysis in the metastatic cohort, and 21 were validated for localized
cohort.
In-silico mathematical modeling of Nanopore sequencing in detecting rare
signal—To assess the ability to detect large hypomethylated domains in rare circulating
tumor cells, we performed an analysis using Nanopore reads from a normally methylated
non-cancer cell line (HUES64) with 1% in-silico spiked-in reads from a cancer cell
line (HCT116). We assessed the ability to determine the correct cell line of origin for
reads that aligned to predefined HCT116 PMDs based on their average methylation level
by quantifying the precision and sensitivity of read classification using the PRROC78.
Methylation was averaged across each read, considering only CpG sites that fall within
PMDs and excluding those within CpG islands.
Illustration—Illustrations were created with BioRender.com.
Quantification and Statistical Analysis
Statistical analyses for all experiments are described in the figure legends and the method
details. Statistical analyses were performed using R (version 3.1.2) and GraphPad Prism 9.0.
Supplementary Material
Refer to Web version on PubMed Central for supplementary material.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 26
Acknowledgments
We thank L. Libby for technical support; J. Fung for flow cytometry assistance. We thank R. Manguso, D. Sen
and all lab members in Haber/Maheswaran lab for discussions. This work was supported by grants from National
Institute of Health (2RO1CA129933 to D.A.H, U01EB012493 to M.T., D.A.H., S.M., U01CA268933 to M.T.,
R01CA259007 to D.T.M., 5P41EB002503 to M.T.), Howard Hughes Medical Institute (to D.A.H.), ESSCO Breast
Cancer Research Fund (to S.M.), Prostate Cancer Foundation (to D.T.M., R.J.L., and J.A.E.), Cygnus Montanus
Foundation founded by the Svanberg Family (to D.T.M.), Breast Cancer Research Foundation (to D.A.H.), National
Foundation for Cancer Research (to D.A.H.) and Max Planck Institute (to A.M.)
References
1. Feinberg AP, and Vogelstein B (1983). Hypomethylation distinguishes genes of some human
cancers from their normal counterparts. Nature 301, 89–92. 10.1038/301089a0. [PubMed: 6185846]
2. Goelz SE, Vogelstein B, Hamilton SR, and Feinberg AP (1985). Hypomethylation of DNA from
benign and malignant human colon neoplasms. Science 228, 187–190. 10.1126/science.2579435.
[PubMed: 2579435]
3. Baylin S, and Bestor TH (2002). Altered methylation patterns in cancer cell genomes: cause or
consequence? Cancer Cell 1, 299–305. 10.1016/s1535-6108(02)00061-2. [PubMed: 12086841]
4. Baylin SB, and Jones PA (2016). Epigenetic Determinants of Cancer. Cold Spring Harb Perspect
Biol 8. 10.1101/cshperspect.a019505.
5. Esteller M (2002). CpG island hypermethylation and tumor suppressor genes: a booming present, a
brighter future. Oncogene 21, 5427–5440. 10.1038/sj.onc.1205600. [PubMed: 12154405]
6. Yegnasubramanian S, Kowalski J, Gonzalgo ML, Zahurak M, Piantadosi S, Walsh PC, Bova GS,
De Marzo AM, Isaacs WB, and Nelson WG (2004). Hypermethylation of CpG islands in primary
and metastatic human prostate cancer. Cancer Res 64, 1975–1986. 10.1158/0008-5472.can-03-3972.
[PubMed: 15026333]
7. Esteller M, Corn PG, Baylin SB, and Herman JG (2001). A gene hypermethylation profile of human
cancer. Cancer Res 61, 3225–3229. [PubMed: 11309270]
8. Berman BP, Weisenberger DJ, Aman JF, Hinoue T, Ramjan Z, Liu Y, Noushmehr H, Lange CP,
van Dijk CM, Tollenaar RA, et al. (2011). Regions of focal DNA hypermethylation and long-range
hypomethylation in colorectal cancer coincide with nuclear lamina-associated domains. Nat Genet
44, 40–46. 10.1038/ng.969. [PubMed: 22120008]
9. Timp W, Bravo HC, McDonald OG, Goggins M, Umbricht C, Zeiger M, Feinberg AP, and Irizarry
RA (2014). Large hypomethylated blocks as a universal defining epigenetic alteration in human
solid tumors. Genome Med 6, 61. 10.1186/s13073-014-0061-y. [PubMed: 25191524]
10. Hansen KD, Timp W, Bravo HC, Sabunciyan S, Langmead B, McDonald OG, Wen B, Wu H,
Liu Y, Diep D, et al. (2011). Increased methylation variation in epigenetic domains across cancer
types. Nat Genet 43, 768–775. 10.1038/ng.865. [PubMed: 21706001]
11. Yegnasubramanian S, Haffner MC, Zhang Y, Gurel B, Cornish TC, Wu Z, Irizarry RA, Morgan
J, Hicks J, DeWeese TL, et al. (2008). DNA hypomethylation arises later in prostate cancer
progression than CpG island hypermethylation and contributes to metastatic tumor heterogeneity.
Cancer Res 68, 8954–8967. 10.1158/0008-5472.CAN-07-6088. [PubMed: 18974140]
12. Hon GC, Hawkins RD, Caballero OL, Lo C, Lister R, Pelizzola M, Valsesia A, Ye Z, Kuan S,
Edsall LE, et al. (2012). Global DNA hypomethylation coupled to repressive chromatin domain
formation and gene silencing in breast cancer. Genome Res 22, 246–258. 10.1101/gr.125872.111.
[PubMed: 22156296]
13. Johnstone SE, Reyes A, Qi Y, Adriaens C, Hegazi E, Pelka K, Chen JH, Zou LS, Drier Y, Hecht
V, et al. (2020). Large-Scale Topological Changes Restrain Malignant Progression in Colorectal
Cancer. Cell 182, 1474–1489 e1423. 10.1016/j.cell.2020.07.030. [PubMed: 32841603]
14. Hansen KD, Sabunciyan S, Langmead B, Nagy N, Curley R, Klein G, Klein E, Salamon D,
and Feinberg AP (2014). Large-scale hypomethylated blocks associated with Epstein-Barr virus-
induced B-cell immortalization. Genome Res 24, 177–184. 10.1101/gr.157743.113. [PubMed:
24068705]
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 27
15. Zhou W, Dinh HQ, Ramjan Z, Weisenberger DJ, Nicolet CM, Shen H, Laird PW, and Berman BP
(2018). DNA methylation loss in late-replicating domains is linked to mitotic cell division. Nat
Genet 50, 591–602. 10.1038/s41588-018-0073-4. [PubMed: 29610480]
16. Epstein JI, Egevad L, Amin MB, Delahunt B, Srigley JR, Humphrey PA, and Grading C (2016).
The 2014 International Society of Urological Pathology (ISUP) Consensus Conference on Gleason
Grading of Prostatic Carcinoma: Definition of Grading Patterns and Proposal for a New Grading
System. Am J Surg Pathol 40, 244–252. 10.1097/PAS.0000000000000530. [PubMed: 26492179]
17. Epstein JI, Zelefsky MJ, Sjoberg DD, Nelson JB, Egevad L, Magi-Galluzzi C, Vickers AJ, Parwani
AV, Reuter VE, Fine SW, et al. (2016). A Contemporary Prostate Cancer Grading System: A
Validated Alternative to the Gleason Score. Eur Urol 69, 428–435. 10.1016/j.eururo.2015.06.046.
[PubMed: 26166626]
18. Fay EK, and Graff JN (2020). Immunotherapy in Prostate Cancer. Cancers (Basel) 12. 10.3390/
cancers12071752.
19. Sfanos KS, Bruno TC, Maris CH, Xu L, Thoburn CJ, DeMarzo AM, Meeker AK, Isaacs WB,
and Drake CG (2008). Phenotypic analysis of prostate-infiltrating lymphocytes reveals TH17
and Treg skewing. Clin Cancer Res 14, 3254–3261. 10.1158/1078-0432.CCR-07-5164. [PubMed:
18519750]
20. Wang S, He Z, Wang X, Li H, and Liu XS (2019). Antigen presentation and tumor immunogenicity
in cancer immunotherapy response prediction. Elife 8. 10.7554/eLife.49020.
21. Venturini NJ, and Drake CG (2019). Immunotherapy for Prostate Cancer. Cold Spring Harb
Perspect Med 9. 10.1101/cshperspect.a030627.
22. Liu L, Toung JM, Jassowicz AF, Vijayaraghavan R, Kang H, Zhang R, Kruglyak KM, Huang
HJ, Hinoue T, Shen H, et al. (2018). Targeted methylation sequencing of plasma cell-free DNA
for cancer detection and classification. Ann Oncol 29, 1445–1453. 10.1093/annonc/mdy119.
[PubMed: 29635542]
23. Haber DA, and Velculescu VE (2014). Blood-based analyses of cancer: circulating tumor cells and
circulating tumor DNA. Cancer Discov 4, 650–661. 10.1158/2159-8290.CD-13-1014. [PubMed:
24801577]
24. Liu MC, Oxnard GR, Klein EA, Swanton C, Seiden MV, and Consortium C (2020). Sensitive
and specific multi-cancer detection and localization using methylation signatures in cell-free DNA.
Ann Oncol 31, 745–759. 10.1016/j.annonc.2020.02.011. [PubMed: 33506766]
25. Klein EA, Richards D, Cohn A, Tummala M, Lapham R, Cosgrove D, Chung G, Clement J, Gao
J, Hunkapiller N, et al. (2021). Clinical validation of a targeted methylation-based multi-cancer
early detection test using an independent validation set. Ann Oncol 32, 1167–1177. 10.1016/
j.annonc.2021.05.806. [PubMed: 34176681]
26. Miyamoto DT, Zheng Y, Wittner BS, Lee RJ, Zhu H, Broderick KT, Desai R, Fox DB, Brannigan
BW, Trautwein J, et al. (2015). RNA-Seq of single prostate CTCs implicates noncanonical
Wnt signaling in antiandrogen resistance. Science 349, 1351–1356. 10.1126/science.aab0917.
[PubMed: 26383955]
27. Ozkumur E, Shah AM, Ciciliano JC, Emmink BL, Miyamoto DT, Brachtel E, Yu M, Chen
PI, Morgan B, Trautwein J, et al. (2013). Inertial focusing for tumor antigen-dependent and
-independent sorting of rare circulating tumor cells. Sci Transl Med 5, 179ra147. 5/179/179ra47
[pii] 10.1126/scitranslmed.3005616.
28. Taylor RA, Toivanen R, Frydenberg M, Pedersen J, Harewood L, Australian Prostate Cancer
B, Collins AT, Maitland NJ, and Risbridger GP (2012). Human epithelial basal cells are cells
of origin of prostate cancer, independent of CD133 status. Stem Cells 30, 1087–1096. 10.1002/
stem.1094. [PubMed: 22593016]
29. Stoyanova T, Cooper AR, Drake JM, Liu X, Armstrong AJ, Pienta KJ, Zhang H, Kohn DB, Huang
J, Witte ON, and Goldstein AS (2013). Prostate cancer originating in basal cells progresses to
adenocarcinoma propagated by luminal-like cells. Proc Natl Acad Sci U S A 110, 20111–20116.
10.1073/pnas.1320565110. [PubMed: 24282295]
30. Lee SO, Tian J, Huang CK, Ma Z, Lai KP, Hsiao H, Jiang M, Yeh S, and Chang C (2012).
Suppressor role of androgen receptor in proliferation of prostate basal epithelial and progenitor
cells. J Endocrinol 213, 173–182. 10.1530/JOE-11-0474. [PubMed: 22393245]
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 28
31. Hayward SW, Dahiya R, Cunha GR, Bartek J, Deshpande N, and Narayan P (1995). Establishment
and characterization of an immortalized but non-transformed human prostate epithelial cell line:
BPH-1. In Vitro Cell Dev Biol Anim 31, 14–24. 10.1007/BF02631333. [PubMed: 7535634]
32. Smallwood SA, Lee HJ, Angermueller C, Krueger F, Saadeh H, Peat J, Andrews SR, Stegle
O, Reik W, and Kelsey G (2014). Single-cell genome-wide bisulfite sequencing for assessing
epigenetic heterogeneity. Nat Methods 11, 817–820. 10.1038/nmeth.3035. [PubMed: 25042786]
33. Bian S, Hou Y, Zhou X, Li X, Yong J, Wang Y, Wang W, Yan J, Hu B, Guo H, et al. (2018).
Single-cell multiomics sequencing and analyses of human colorectal cancer. Science 362, 1060–
1063. 10.1126/science.aao3791. [PubMed: 30498128]
34. Zhao SG, Chen WS, Li H, Foye A, Zhang M, Sjostrom M, Aggarwal R, Playdle D, Liao A,
Alumkal JJ, et al. (2020). The DNA methylation landscape of advanced prostate cancer. Nat Genet
52, 778–789. 10.1038/s41588-020-0648-8. [PubMed: 32661416]
35. Aran D, Toperoff G, Rosenberg M, and Hellman A (2011). Replication timing-related and gene
body-specific methylation of active human genes. Hum Mol Genet 20, 670–680. 10.1093/hmg/
ddq513. [PubMed: 21112978]
36. Hermann A, Goyal R, and Jeltsch A (2004). The Dnmt1 DNA-(cytosine-C5)-methyltransferase
methylates DNA processively with high preference for hemimethylated target sites. J Biol Chem
279, 48350–48359. 10.1074/jbc.M403427200. [PubMed: 15339928]
37. Johnstone SE, Gladyshev VN, Aryee MJ, and Bernstein BE (2022). Epigenetic clocks, aging, and
cancer. Science 378, 1276–1277. 10.1126/science.abn4009. [PubMed: 36548410]
38. Bucay N, Sekhon K, Majid S, Yamamura S, Shahryari V, Tabatabai ZL, Greene K, Tanaka
Y, Dahiya R, Deng G, and Saini S (2016). Novel tumor suppressor microRNA at frequently
deleted chromosomal region 8p21 regulates epidermal growth factor receptor in prostate cancer.
Oncotarget 7, 70388–70403. 10.18632/oncotarget.11865. [PubMed: 27611943]
39. Cancer Genome Atlas Research, N. (2015). The Molecular Taxonomy of Primary Prostate Cancer.
Cell 163, 1011–1025. 10.1016/j.cell.2015.10.025. [PubMed: 26544944]
40. Bethel CR, Faith D, Li X, Guan B, Hicks JL, Lan F, Jenkins RB, Bieberich CJ, and De Marzo AM
(2006). Decreased NKX3.1 protein expression in focal prostatic atrophy, prostatic intraepithelial
neoplasia, and adenocarcinoma: association with gleason score and chromosome 8p deletion.
Cancer Res 66, 10683–10690. 10.1158/0008-5472.CAN-06-0963. [PubMed: 17108105]
41. Bova GS, Carter BS, Bussemakers MJ, Emi M, Fujiwara Y, Kyprianou N, Jacobs SC, Robinson
JC, Epstein JI, Walsh PC, and et al. (1993). Homozygous deletion and frequent allelic loss of
chromosome 8p22 loci in human prostate cancer. Cancer Res 53, 3869–3873. [PubMed: 7689419]
42. Yu YP, Ding Y, Chen R, Liao SG, Ren BG, Michalopoulos A, Michalopoulos G, Nelson J,
Tseng GC, and Luo JH (2013). Whole-genome methylation sequencing reveals distinct impact of
differential methylations on gene transcription in prostate cancer. Am J Pathol 183, 1960–1970.
10.1016/j.ajpath.2013.08.018. [PubMed: 24113458]
43. Van Kaer L, Wu L, and Joyce S (2016). Mechanisms and Consequences of Antigen Presentation by
CD1. Trends Immunol 37, 738–754. 10.1016/j.it.2016.08.011. [PubMed: 27623113]
44. Brennan PJ, Brigl M, and Brenner MB (2013). Invariant natural killer T cells: an innate activation
scheme linked to diverse effector functions. Nat Rev Immunol 13, 101–117. 10.1038/nri3369.
[PubMed: 23334244]
45. Godfrey DI, MacDonald HR, Kronenberg M, Smyth MJ, and Van Kaer L (2004). NKT cells:
what’s in a name? Nat Rev Immunol 4, 231–237. 10.1038/nri1309. [PubMed: 15039760]
46. Hix LM, Shi YH, Brutkiewicz RR, Stein PL, Wang CR, and Zhang M (2011). CD1d-expressing
breast cancer cells modulate NKT cell-mediated antitumor immunity in a murine model of breast
cancer metastasis. PLoS One 6, e20702. 10.1371/journal.pone.0020702. [PubMed: 21695190]
47. McEwen-Smith RM, Salio M, and Cerundolo V (2015). The regulatory role of invariant NKT
cells in tumor immunity. Cancer Immunol Res 3, 425–435. 10.1158/2326-6066.CIR-15-0062.
[PubMed: 25941354]
48. Connolly DJ, and Bowie AG (2014). The emerging role of human PYHIN proteins in innate
immunity: implications for health and disease. Biochem Pharmacol 92, 405–414. 10.1016/
j.bcp.2014.08.031. [PubMed: 25199457]
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 29
49. Watson PA, Ellwood-Yen K, King JC, Wongvipat J, Lebeau MM, and Sawyers CL (2005).
Context-dependent hormone-refractory progression revealed through characterization of a novel
murine prostate cancer cell line. Cancer Res 65, 11565–11571. 10.1158/0008-5472.CAN-05-3441.
[PubMed: 16357166]
50. Shen H, Gu C, Liang T, Liu H, Guo F, and Liu X (2020). Unveiling the heterogeneity of NKT cells
in the liver through single cell RNA sequencing. Sci Rep 10, 19453. 10.1038/s41598-020-76659-1.
[PubMed: 33173202]
51. Stott SL, Lee RJ, Nagrath S, Yu M, Miyamoto DT, Ulkus L, Inserra EJ, Ulman M, Springer S,
Nakamura Z, et al. (2010). Isolation and characterization of circulating tumor cells from patients
with localized and metastatic prostate cancer. Sci Transl Med 2, 25ra23. 2/25/25ra23 [pii] 10.1126/
scitranslmed.3000403.
52. Seiden MV, Kantoff PW, Krithivas K, Propert K, Bryant M, Haltom E, Gaynes L, Kaplan I, Bubley
G, DeWolf W, and et al. (1994). Detection of circulating tumor cells in men with localized prostate
cancer. J Clin Oncol 12, 2634–2639. [PubMed: 7527455]
53. Miyamoto DT, Lee RJ, Kalinich M, LiCausi JA, Zheng Y, Chen T, Milner JD, Emmons E,
Ho U, Broderick K, et al. (2018). An RNA-Based Digital Circulating Tumor Cell Signature Is
Predictive of Drug Response and Early Dissemination in Prostate Cancer. Cancer Discov 8, 288–
303. 10.1158/2159-8290.CD-16-1406. [PubMed: 29301747]
54. Simpson JT, Workman RE, Zuzarte PC, David M, Dursi LJ, and Timp W (2017). Detecting DNA
cytosine methylation using nanopore sequencing. Nat Methods 14, 407–410. 10.1038/nmeth.4184.
[PubMed: 28218898]
55. Rand AC, Jain M, Eizenga JM, Musselman-Brown A, Olsen HE, Akeson M, and Paten B (2017).
Mapping DNA methylation with high-throughput nanopore sequencing. Nat Methods 14, 411–
413. 10.1038/nmeth.4189. [PubMed: 28218897]
56. Lieberman-Aiden E, van Berkum NL, Williams L, Imakaev M, Ragoczy T, Telling A, Amit
I, Lajoie BR, Sabo PJ, Dorschner MO, et al. (2009). Comprehensive mapping of long-range
interactions reveals folding principles of the human genome. Science 326, 289–293. 10.1126/
science.1181369. [PubMed: 19815776]
57. Wang ZT, Chen SD, Xu W, Chen KL, Wang HF, Tan CC, Cui M, Dong Q, Tan L, Yu JT, and
Alzheimer’s Disease Neuroimaging I (2019). Genome-wide association study identifies CD1A
associated with rate of increase in plasma neurofilament light in non-demented elders. Aging
(Albany NY) 11, 4521–4535. 10.18632/aging.102066. [PubMed: 31295725]
58. Liu H, Xing Y, Guo Y, Liu P, Zhang H, Xue B, Shou J, Qian J, Peng J, Wang R, et al. (2016).
Polymorphisms in exon 2 of CD1 genes are associated with susceptibility to Guillain-Barre
syndrome. J Neurol Sci 369, 39–42. 10.1016/j.jns.2016.07.029. [PubMed: 27653862]
59. Schreiber RD, Old LJ, and Smyth MJ (2011). Cancer immunoediting: integrating immunity’s
roles in cancer suppression and promotion. Science 331, 1565–1570. 10.1126/science.1203486.
[PubMed: 21436444]
60. Waldman AD, Fritz JM, and Lenardo MJ (2020). A guide to cancer immunotherapy: from T cell
basic science to clinical practice. Nat Rev Immunol 20, 651–668. 10.1038/s41577-020-0306-5.
[PubMed: 32433532]
61. Kumaki Y, Oda M, and Okano M (2008). QUMA: quantification tool for methylation analysis.
Nucleic Acids Res 36, W170–175. 10.1093/nar/gkn294. [PubMed: 18487274]
62. Hou Y, Guo H, Cao C, Li X, Hu B, Zhu P, Wu X, Wen L, Tang F, Huang Y, and Peng J (2016).
Single-cell triple omics sequencing reveals genetic, epigenetic, and transcriptomic heterogeneity in
hepatocellular carcinomas. Cell Res 26, 304–319. 10.1038/cr.2016.23. [PubMed: 26902283]
63. Picelli S, Faridani OR, Bjorklund AK, Winberg G, Sagasser S, and Sandberg R (2014). Full-length
RNA-seq from single cells using Smart-seq2. Nat Protoc 9, 171–181. 10.1038/nprot.2014.006.
[PubMed: 24385147]
64. Clark SJ, Smallwood SA, Lee HJ, Krueger F, Reik W, and Kelsey G (2017). Genome-wide
base-resolution mapping of DNA methylation in single cells using single-cell bisulfite sequencing
(scBS-seq). Nat Protoc 12, 534–547. 10.1038/nprot.2016.187. [PubMed: 28182018]
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 30
65. Brind’Amour J, Liu S, Hudson M, Chen C, Karimi MM, and Lorincz MC (2015). An ultra-low-
input native ChIP-seq protocol for genome-wide profiling of rare cell populations. Nat Commun 6,
6033. 10.1038/ncomms7033. [PubMed: 25607992]
66. Trapnell C, Roberts A, Goff L, Pertea G, Kim D, Kelley DR, Pimentel H, Salzberg SL, Rinn JL,
and Pachter L (2012). Differential gene and transcript expression analysis of RNA-seq experiments
with TopHat and Cufflinks. Nat Protoc 7, 562–578. 10.1038/nprot.2012.016. [PubMed: 22383036]
67. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Abecasis G, Durbin R, and
Genome Project Data Processing, S. (2009). The Sequence Alignment/Map format and SAMtools.
Bioinformatics 25, 2078–2079. 10.1093/bioinformatics/btp352. [PubMed: 19505943]
68. Anders S, Pyl PT, and Huber W (2015). HTSeq--a Python framework to work with
high-throughput sequencing data. Bioinformatics 31, 166–169. 10.1093/bioinformatics/btu638.
[PubMed: 25260700]
69. Krueger F, and Andrews SR (2011). Bismark: a flexible aligner and methylation caller
for Bisulfite-Seq applications. Bioinformatics 27, 1571–1572. 10.1093/bioinformatics/btr167.
[PubMed: 21493656]
70. Carter SL, Cibulskis K, Helman E, McKenna A, Shen H, Zack T, Laird PW, Onofrio RC, Winckler
W, Weir BA, et al. (2012). Absolute quantification of somatic DNA alterations in human cancer.
Nat Biotechnol 30, 413–421. nbt.2203 [pii] 10.1038/nbt.2203. [PubMed: 22544022]
71. Karolchik D, Hinrichs AS, Furey TS, Roskin KM, Sugnet CW, Haussler D, and Kent WJ (2004).
The UCSC Table Browser data retrieval tool. Nucleic Acids Res 32, D493–496. 10.1093/nar/
gkh103. [PubMed: 14681465]
72. Gel B, Diez-Villanueva A, Serra E, Buschbeck M, Peinado MA, and Malinverni R (2016).
regioneR: an R/Bioconductor package for the association analysis of genomic regions based
on permutation tests. Bioinformatics 32, 289–291. 10.1093/bioinformatics/btv562. [PubMed:
26424858]
73. Garvin T, Aboukhalil R, Kendall J, Baslan T, Atwal GS, Hicks J, Wigler M, and Schatz MC
(2015). Interactive analysis and assessment of single-cell copy-number variations. Nat Methods
12, 1058–1060. 10.1038/nmeth.3578. [PubMed: 26344043]
74. Li H, and Durbin R (2009). Fast and accurate short read alignment with Burrows-Wheeler
transform. Bioinformatics 25, 1754–1760. 10.1093/bioinformatics/btp324. [PubMed: 19451168]
75. Tarasov A, Vilella AJ, Cuppen E, Nijman IJ, and Prins P (2015). Sambamba: fast processing
of NGS alignment formats. Bioinformatics 31, 2032–2034. 10.1093/bioinformatics/btv098.
[PubMed: 25697820]
76. Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein BE, Nusbaum C, Myers RM,
Brown M, Li W, and Liu XS (2008). Model-based analysis of ChIP-Seq (MACS). Genome Biol 9,
R137. 10.1186/gb-2008-9-9-r137. [PubMed: 18798982]
77. Ramirez F, Ryan DP, Gruning B, Bhardwaj V, Kilpert F, Richter AS, Heyne S, Dundar F, and
Manke T (2016). deepTools2: a next generation web server for deep-sequencing data analysis.
Nucleic Acids Res 44, W160–165. 10.1093/nar/gkw257. [PubMed: 27079975]
78. Grau J, Grosse I, and Keilwagen J (2015). PRROC: computing and visualizing precision-
recall and receiver operating characteristic curves in R. Bioinformatics 31, 2595–2597. 10.1093/
bioinformatics/btv153. [PubMed: 25810428]
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 31
Highlights
1.
2.
3.
4.
40 core hypomethylated domains across prostate CTCs arise early in
tumorigenesis.
Hypomethylation silences immune-related genes, sparing adjacent
proliferation genes.
The CD1A-IFI16 immune locus is consistently silenced by hypomethylation
in cancer.
Hypomethylated domains detected in CTC-enriched blood in localized
prostate cancer.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 32
Figure 1. Partially Methylated Domains (PMDs) and Preserved Methylation Islands (PMIs) in
single metastatic prostate cancer cells.
(A) Schematic of CTC enrichment (104-fold leukocyte depletion), and paired DNA
methylation sequencing (nucleus) and RNA-seq (cytoplasm) from individual prostate CTCs.
(B) Confirmation of CTC identity using stringent RNA expression thresholding of prostatic
lineage and epithelial versus leukocyte markers. Maximum log10 (RPM) expression of
epithelial (KRT7, KRT8, KRT18, KRT19, EPCAM) and prostatic markers (AR, KLK3,
FOLH1, AMACR) are plotted against leukocyte markers (CD45, CD16, CD37, CD53, CD7,
CD66b). Only confirmed CTCs without WBC contamination (red crosses) were used in
analyses.
(C) Representative DNA copy number variation (CNV) analysis in individual CTCs from
two patients, compared with a diploid normal prostate epithelial cell (HPrEC) and a healthy
donor-derived leukocyte. Single-cell DNA methylation sequencing data was used to infer
DNA copy number.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 33
(D) IGV representation (hg19) of DNA methylation spanning chromosome 8, showing
extensive PMDs (yellow) across 37 individual CTCs from four patients (GU114, GU216,
GU181 and GURa15), and 17 cells from prostate cancer cell lines (LNCaP, PC3, VCaP,
22Rv1). As controls, 4 normal bulk prostate tissues (N.P.), 36 cells from two prostate
epithelial cell lines (HPrEC, BPH-1) and normal leukocytes (WBCs) are shown. Normal
methylation level (blue).
(E-F) Higher resolution of chromosome 8 in IGV, showing precise PMD boundaries shared
across individual CTCs and prostate cancer cell lines (panel E), with magnified view of
the nested PMI, bracketing a few genes, with precise boundaries of preserved methylation
flanked by profound hypomethylation (panel F).
(G-H) Components of coding genes and classes of repeats differentially enriched in PMDs
versus PMIs (panel G), with differences among subtypes of repeats (panel H). ns, not
significant; *P<0.05; **P<0.01, assessed by permutation test.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 34
Figure 2. Acquired chromatin marks in prostate cancer PMDs and nomination of shared core
PMDs.
(A) Differential enrichment of chromatin marks within prostate cancer PMDs and PMIs.
Annotated chromatin marks from ChIP-seq dataset of PC3 cells in ENCODE (https://
www.encodeproject.org/). ns, not significant; *P<0.05; **P<0.01, assessed by permutation
test.
(B) Line plots showing differential enrichment of silencing chromatin marks at PMDs across
the genome in prostate cancer cells (LNCaP; 3 biological replicates, red lines), compared
with cultured benign prostatic hyperplasia cells (BPH-1; 2 biological replicates, green lines)
and normal prostate epithelial cells (HPrEC; 2 biological replicates, blue lines). Across the
genome, prostate cancer cells acquire H3K27me3, with highest levels at the boundaries of
PMDs (left panel), whereas H3K9me3 enrichment towards the center of PMDs is not altered
between cancer and non-transformed prostate cells (right panel).
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 35
(C) Boxplot showing enrichment of Cut and Run signal for H3K27me3, but not H3K9me3,
across prostate cancer PMDs between LNCaP cells and non-transformed cell lines (HPrEC
and BPH-1). Pvalue, one-tailed Student’s t-test.
(D) IGV track showing representative cancer-associated PMD (DNA hypomethylation:
yellow), with pronounced enrichment of H3K27me3 at PMD borders in cancer cells
(LNCaP: red) versus non-transformed cells (HPrEC: blue, BPH-1: green), whereas PMD-
centered H3K9me3 occupancy is unaltered.
(E) Inter- and intra-patient heterogeneity of PMDs among single CTCs from four prostate
cancer patients (red) and single cells from prostate cancer cell lines. Mean Jaccard
index indicates heterogeneity, with higher mean score indicating less heterogeneity among
samples. Error bar, mean with 95% confidence interval (CI).
(F-G) IGV representation of total PMDs and core PMDs at chromosome 3 locus, across
8 sample sources (4 patients and 4 prostate cancer cell lines). Total PMDs (blue) are the
union of PMDs defined in each sample source, while core PMDs (black) are shared across
all 8 sample sources (panel F); representation of PMDs from the single-cell components of
an individual sample source (22 CTCs from patient GU181) showing a core PMD shared
across all sample sources (black) and neighboring non-core PMDs that are shared by >90%
CTCs in this patient, but not across different sample sources (panel G). See Figure S2D and
Methods for criteria in core PMD and PMI designation.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 36
Figure 3. Demethylation of core PMDs during early prostate tumorigenesis suppresses immune-
related genes, while core PMIs spare proliferation genes.
(A) Schematic showing prostate tumor microdissection, single nucleus isolation and single-
cell DNA methylation sequencing.
(B) Ranking of methylation level at 40 core PMDs (red dots) among all 1,496 total PMDs,
as a function of timeline from normal prostate, to localized (GS6; GS8) and metastatic
cancer (CTCs), showing early demethylation of core PMDs. Within normal prostate, all 40
core PMDs have methylation level >75%, and 31 are hypomethylated as early as GS6.
(C) Quantitation of demethylation as a function of Gleason Score (GS). Demethylation
of core PMDs (red curve) precedes that of other PMDs (magenta) within microdissected
prostate tumor cells and in CTCs. In contrast, core PMIs nested between PMDs (blue)
show minimal DNA methylation changes during tumorigenesis. Error bar, mean with SEM.
Statistical analysis of DNA methylation curves utilizing longitudinal linear mixed effects
model, by which tumor progression x methylation domains was tested.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 37
(D) Quantitation of demethylation as a function of GS in TCGA prostate cancer methylation
array data, showing early and progressive loss of methylation of core PMDs (red curve),
with an attenuated trend for other PMDs (magenta). The core PMIs (blue) display stable
DNA methylation pattern during prostate tumorigenesis. Statistical analysis as for panel C.
(E-F) Gene set enrichment analysis (GSEA) of genes residing within core PMDs and
downregulated in primary prostate cancer (E), and of genes residing within core PMIs with
gene expression preserved (up-regulated and not significantly changed) in primary prostate
cancer (F), compared with normal prostate. (FDR <0.1; two-tailed Student’s t-test with FDR
correction).
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 38
Figure 4. Correlation of DNA demethylation at the CD1A-IFI16 locus with accumulation of
chromatin silencing marks and reduced gene expression.
(A) IGV of single-cell DNA methylation at the CD1A-IF16 genomic locus, including five
lipid antigen presentation and four interferon inducible genes. Tumor cells (37 single CTCs
from four prostate cancer patients (red) and 17 single cells from four prostate cancer cell
lines (green)) exhibit marked hypomethylation at this locus (shaded yellow), while normal
samples (4 bulk normal prostate tissues, 37 single cells from normal prostate cell lines and
leukocytes (blue)) show a preserved DNA methylation (shaded blue).
(B) Heatmap (upper panel; hypomethylation shaded yellow) and matched quantitative scatter
plots (lower panel) of single-cell DNA methylation levels within all 1,496 prostate cancer
PMDs, showing progression from normal prostate to localized prostate cancer (GS6, GS8)
and metastatic CTCs. The CD1A-IFI16 locus (dashed vertical red line) shows early and
profound demethylation, starting at GS6, with its rank number across all PMDs at each
tumor stage shown in parentheses (red).
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 39
(C) IGV screenshot of single-cell DNA methylation data showing progressive demethylation
of CD1A-IFI16 locus (box with red dashed line) from normal prostate cells to localized
(GS6 and GS8) and metastatic prostate cancer (CTCs). Heterogeneity of hypomethylation
(shaded yellow) across single cells is evident at GS6, becoming more prevalent at GS8, and
uniform in CTCs .
(D) Plots showing suppressed expression of lipid antigen presentation and interferon
inducible genes within the CD1A-IFI16 locus, during transition from normal prostate to
low-grade GS6, with persistent silencing in higher grade GS7, 8 and 9 cancers (TCGA
dataset). Error bar, mean with SEM.
(E) Analysis of 33 different tumor types (TCGA) for DNA methylation differences at
core prostate cancer PMDs, compared with corresponding normal tissues. 30 of 35 (86%)
evaluable PMDs are hypomethylated across all tumor types (red circles), with the CD1A-
IFI16 locus having the strongest hypomethylation.
(F) Histograms of DNA methylation level within 100kb windows (200bp offsets) across the
genome in normal prostate cells (BPH-1), following 5-azacytidine treatment (days 1 and 5),
compared with DMSO control.
(G) Quantitation of H3K27me3-related fluorescence intensity within single-cell nuclei
(confocal microscopy). Error bar, mean with SEM. P-value, two-tailed Student’s t-test.
(H) Sequential reduction in CD1d protein expression in normal prostate cells (BPH-1)
treated with 5-azacytidine, compared with DMSO control. Representative flow cytometry
(left panel); median fluorescence intensity (right panel). Error bar, mean with SEM. P-value,
two tailed Student’s t-test.
(I-J) Western blot showing reduced H3K27 trimethylation in 22Rv1 cells treated with EZH2
inhibitor GSK126 for 6 days (panel H); qPCR of genes within the CD1A-IFI16 cluster
show induced expression (panel I), while non-PMD resident control genes (PP1A, HPRT
and β-actin) remain unchanged. P-value, Tukey’s multiple comparison tests, where GSK126
treatment conditions (red bars) were compared to controls (blue bar). n.s. not significant;
****P<0.0001.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 40
Figure 5. Restoring expression of genes within CD1A-IFI16 syntenic locus abrogates
tumorigenesis in an immunocompetent mouse prostate cancer model.
(A) Plots quantifying Cd1d1 and Ifi204 mRNA in the murine prostate tumor cell line
Myc-CaP, which have silenced the syntenic genes (blue), compared to normal prostate cells
from 4 isogenic mice FVB (orange). Ectopic expression of murine Cd1d1 (CD1D ortholog,
green) and Ifi204 (IFI16 ortholog, red) is comparable to that of normal prostate. Error bar,
mean with SEM.
(B) Overexpression (OE) of Cd1d1 or Ifi204 in Myc-CaP cells does not alter in vitro
proliferation compared with controls. Error bar, mean with SD.
(C) Overexpression of either Cd1d1 (green) or Ifi204 (red) in Myc-CaP cells (mCherry-
luciferase tagged) suppresses tumorigenesis in isogenic immunocompetent FVB mice.
Mock-transfected control tumors are shown as control (blue). Tumor size quantified by
luciferase imaging (representative images). Error bar, mean with SEM.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 41
(D) Myc-CaP cells engineered as in (C) show no difference in tumor growth in immune-
deficient NSG mice. Error bar, mean with SEM.
(E) Flow cytometry of Cd1d-restored Myc-CaP tumors in FVB mice, showing recruitment
of CD1d-restricted NKT cells (marked by α-GalCer CD1d Tetramer) and activated NKT
cells (marked by CD69), compared with controls. Error bar, mean with SD.
(F) Flow cytometry of Ifi204-restored Myc-CaP tumors in FVB mice, showing unaltered
infiltration of total CD4+ and CD8+ T cells, but reduced immune infiltration by PD-1+ CD8+
T cells and increased presence of TNFα+ CD8+ T cells, compared with controls. Error bar,
mean with SD. P-values, two-tailed Student’s t-test; ns, not significant.
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 42
Figure 6. Detection of CTC-derived DNA hypomethylation in blood specimens using Nanopore
sequencing.
(A) IGV screenshot showing concordance of DNA hypomethylation measurements between
Oxford Nanopore native sequencing of bulk VCaP cells [B], compared with Illumina
bisulfite sequencing of three single VCaP cells (#1, #2, #3). DNA methylation across entire
chromosome 4 is shown (hypomethylation in shaded yellow).
(B) Scatter plot showing high Pearson correlation (r=0.81) between Nanopore native
sequencing and Illumina bisulfite sequencing.
(C-D) Mathematical modeling showing minimal precision using short reads (average 5 CpG
sites per read) for detection of hypomethylated DNA domains. Modest improvement in
detection is provided by interrogating predetermined PMDs, instead of whole genome (panel
C). Significantly improved precision is predicted using Nanopore long read sequencing (10
or 50 CpGs per read). Highest predicted accuracy by combining Nanopore long reads (>10
CpG sites per read) with selected analysis-predetermined PMD regions (panel D).
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 43
(E) Schematic of microfluidic CTC enrichment (followed by direct Nanopore sequencing of
bulk cells (approximatly 0.1% CTC purity). HMW, high molecular weight.
(F-G) Scatter plot quantitation of hypomethylation signal by Nanopore sequencing,
comparing leukocyte-depleted blood samples from patients with either metastatic (panel
F) or localized prostate cancer before surgical resection or radiation therapy (panel G),
versus healthy age-matched male donors (HDs). Error bar denotes mean with SEM. P-value
assessed by two-tailed Student’s t-test. Dotted lines indicate thresholds of hypomethylation
signal that encompass all healthy donors tested, with the fraction of cancer patients with
hypomethylation signal above that threshold considered positive.
Cell. Author manuscript; available in PMC 2023 August 18.
Guo et al.
Page 44
Key resources table
REAGENT or RESOURCE
SOURCE
IDENTIFIER
Antibodies
Mouse anti-CD45 biotinylated (clone 2D1)
R&D Systems
Cat# BAM1430; RRID:AB_356874
Mouse anti-CD66b (clone 80H3)
Bio-Rad
MCA216T; RRID:AB_2291565
Mouse anti-human CD16 biotinylated (clone 3G8)
BD Biosciences
Cat#555405; RRID:AB_395805
AF488-conjugated mouse anti-human EpCAM (clone VU1D9)
Cell Signaling Technology
Cat#5198; RRID:AB_10692105
PE-conjugated mouse antibody anti-CD45 (clone HI30)
BD Biosciences
Cat#560975; RRID:AB_2033960
Rabbit anti-histone H3K27me3 (Western blot)
Thermo Fisher Scientific
Cat#MA5–11198; RRID:AB_2899176
Rabbit anti-histone H3K27me3 (ChIP and CUT&RUN)
Active motif
Cat#39155; RRID:AB_2561020
Rabbit anti-histone H3K9me3 (ChIP)
Abcam
Cat#ab8898; RRID:AB_306848
Rabbit anti-IgG control (clone DA1E) (CUT&RUN)
Cell Signaling Technology
Cat#66362; RRID:AB_2924329
Rabbit anti-histone H3 total
Abcam
Cat#1791; RRID:AB_302613
Rabbit anti-H3K27me3 (clone C36B11) (immunofluorescence)
Cell Signaling Technology
CST#9733; RRID:AB_2616029
APC conjugated mouse anti-human CD1d (FACS)
BioLegend
Cat#350308; RRID:AB_10642829
APC conjugated mouse anti-human CD1d (clone CD1d42)
BD Biosciences
BD#563505; RRID:AB_2738246
APC-conjugated isotype control
BD Biosciences
BD#555751
Rat inVivoMab anti-mouse CD1d (clone 20H2) (FACS for Myc-
CaP cells)
Rat InVivoPlus anti-mouse isotype control (clone HRPN) (FACS
for Myc-CaP cells)
Bio X Cell
#BE0179; RRID:AB_10949293
Bio X Cell
#BE0088; RRID:AB_1107775
APC conjugated goat anti-rat IgG (H+L)
Thermo Fisher Scientific
Cat#A10540
Rat anti-mouse CD16/CD32 blocking reagent (Clone: 2.4G2)
BD Biosciences
Cat#553142; RRID:AB_394657
BV510-viability dye
APC-α-GalCer-mCD1d Tetramer
BV711-conjugated anti-mouse CD69 (clone: H1.2F3)
BD Biosciences
BD#564406; RRID:AB_2869572
TetramerShop
BioLegend
Cat#MCD1d–001
Cat#104537; RRID:AB_2566120
PerCP-Cy5.5-conjugated anti-mouse TCRβ (clone: H57–597)
Biolegend
Cat#109228; RRID:AB_1575173
BV605-conjugated anti-mouse CD3e (clone: 145–2C11)
BUV395-conjugated anti-mouse NK1.1 (clone: PK136)
BV711- conjugated anti-mouse CD8a (clone: 53–6.7)
BV650- conjugated anti-mouse CD4 (clone: RM4–5)
FITC- conjugated anti-mouse CD44 (clone: IM7)
PE-Cy7- conjugated anti-mouse PD-1 (clone: RMP1–30)
BV421-conjugated anti-mouse TIM3 (clone: 5D12)
APC- conjugated anti-mouse TIGIT (clone: 4D4/mTIGIT)
BV785- conjugated anti-mouse LAG3 (clone:C9B7W)
PE- conjugated anti-mouse TNFα (clone: MP6-XT22)
BV650- conjugated anti-mouse CD4 (clone: RM4–5)
BV605-conjugated anti-mouse IFNγ (clone: XMG1.2)
Biological samples
BioLegend
BioLegend
BioLegend
BioLegend
BioLegend
BioLegend
BioLegend
BioLegend
BioLegend
BioLegend
BioLegend
BioLegend
Cat#100351; RRID:AB_2565842
Cat#564144
Cat#100759; RRID:AB_2563510
Cat#100546; RRID:AB_2562098
Cat#103006; RRID:AB_312957
Cat#109110; RRID:AB_572017
Cat#747626
Cat#156106; RRID:AB_2750515
Cat#125219; RRID:AB_2566571
Cat#506306; RRID:AB_315427
Cat#100546; RRID:AB_2562098
Cat#505840; RRID:AB_2734493
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 45
REAGENT or RESOURCE
Healthy donors for blood samples
Blood samples from patients with a diagnosis of localized of
metastatic prostate cancer
SOURCE
This paper
This paper
Localized tumor tissue cohort (core biopsies or surgical resection)
This paper
Chemicals, peptides, and recombinant proteins
5-azacitidine
GSK126
MNase enzyme (micrococcal nuclease)
G418
Blasticidin
Cell Stimulation Cocktail
Protein Transport Inhibitor Cocktail
Dynabeads MyOne Streptavidin T1
Critical commercial assays
EZ DNA methylation kit
Zero blunt PCR cloning kit
Magnetic MyOne Carboxylic Acid Beads
NEBNext Ultra II DNA Library Prep Kit
CUT&RUN Assay kit
RNeasy Mini kit
SuperScript III One-Step qRT-PCR kit
NEBuilder HiFi DNA Assembly Cloning kit
IDENTIFIER
N/A
N/A
N/A
Cat#S1782
Cat#S7061
Cat# M0247S
Cat#G8168
Cat#ant–bl–05
Cat#00–4970–93
Cat#00–4980
Cat#65–601
Cat#D5001
Cat#K270020
Cat#65011
Cat#E7645L
Cat#74104
Cat#11732020
Cat#E5520S
Cat#554714
Cat#67563
Cat#SQK–RBK004
Selleck
Selleckchem
NEB
Sigma Aldrich
InvivoGen
eBioscience
eBioscience
Invitrogen
Zymo
ThermoFisher
Invitrogen
NEB
QIAGEN
Invitrogen
NEB
QIAGEN
Nanopore
Cell Signaling Technology
Cat#86652S
BD Fixation/Permeabilization Solution Kit
BD Biosciences
HMW DNA extraction kit
Rapid Barcoding Kit
Deposited data
Raw and analyzed data
Human reference genome NCBI build 37, GRCh37 (hg19)
Illumina Infinium Human Methylation 450 K BeadChip
DNA Methylation 450 K BeadChip datasets
This paper
GEO: GSE208449
Genome Reference
Consortium
https://www.ncbi.nlm.nih.gov/
assembly/GCF_000001405.13/
National Cancer
Institute’s GDC Data Portal
National Cancer
Institute’s GDC Data Portal
https://portal.gdc.cancer.gov
https://portal.gdc.cancer.gov
TCGA (PRAD samples)
CBioPortal
https://www.cbioportal.org/
Methylation profiles (TCGA cohorts, 33 cancer types)
TCGA Research Network
https://portal.gdc.cancer.gov
Genome annotations (TSS, exon, intron, intragenic regions, CpG
islands (CGIs), repetitive elements and UCSC gap regions) - UCSC
genome table browser
Karolchik et al, 2004
https://genome.ucsc.edu/cgi-bin/
hgTables
DNA methylation datasets (colon and thyroid)
Timp et al, 2014
GEO: GSE53051
DNA methylation of normal prostate tissues and primary prostate
tumors
Yu et al., 2013
Obtained from authors. https://doi.org/
10.1016/j.ajpath.2013.08.018
DNA methylation of metastatic prostate tumors
Zhao et al., 2020
dbGAP: phs001648
Experimental models: Cell lines
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 46
REAGENT or RESOURCE
SOURCE
IDENTIFIER
Human prostate cancer cell line (LNCaP, clone FGC)
Human prostate cancer cell line (VCaP)
Human prostate cancer cell line (PC3)
Human prostate cancer cell line (22Rv1 )
Murine prostate cancer line (Myc-CaP)
Normal cultured prostate epithelial cells (HPrEC)
Benign prostatic hypertrophy cells (BPH-1)
Murine Lewis lung carcinoma cells (LLC-1)
Experimental models: Organisms/strains
ATCC
ATCC
ATCC
ATCC
ATCC
ATCC
Sigma-Aldrich
ATCC
CRL–1740
CRL–2876
CRL–1435
CRL–2505
CRL–3255
PCS–440–010
SCC256
CRL–1642
Mouse: FVB mice
Jackson Laboratory
Strain#001800
Mouse: NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ
Jackson Laboratory
Strain#005557
Mouse: C57BL/6
Oligonucleotides
Primers for qRT-PCR
Primers for Bisulfite PCR
Recombinant DNA
pLenti-murine Cd1d1-mGFP
pLenti-C-mGFP
pLenti-Ifi204-Myc-DDK-Puro
pLenti-C-Myc-DDK-Puro
lentiCRISPRv2-blast
N174-MCS
pMD2.G
psPAX2
Software and algorithms
QUMA
ImageJ
FlowJo software (v10.4)
Trim Galore (v0.4.3)
Tophat (v2.1.1)
Samtools (v1.3.1)
HTseq (v0.6.1)
Cufflinks (v2.1.1)
R (v3.1.2)
Graph Prism 9
Jackson Laboratory
Strain#000664
This paper
This paper
Origene
Origene
Origene
Origene
Addgene
Addgene
Addgene
Addgene
Table S3
Table S3
Cat#MR226027L4
Cat#PS100093
Cat#MR222527L3
Cat#PS100092
Cat#98293; RRID:Addgene_98293
Cat#81061; RRID:Addgene_81061
Cat#12259; RRID:Addgene_12259
Cat#12260; RRID:Addgene_12260
Kumaki et al., 2008
http://quma.cdb.riken.jp/
https://imagej.nih.gov/ij/
BD Bioscience
https://www.flowjo.com/
Babraham Bioinformatics
https://github.com/FelixKrueger/
TrimGalore
Trapnell et al., 2012
https://github.com/infphilo/tophat
Li et al., 2009
http://samtools.sourceforge.net/
Anders et al., 2015
https://htseq.readthedocs.io/en/master/
Trapnell et al., 2012
https://github.com/cole-trapnell-lab/
cufflinks
R Core Team, 2021
https://www.R-project.org/
GraphPad
https://www.graphpad.com/
Bismark tool (v0.17.0)
Krueger and Andrews, 2011
UCSC lift-over tool
Hinrichs et al., 2006
https://github.com/FelixKrueger/
Bismark
https://genome.ucsc.edu/cgi-bin/
hgLiftOver
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
Guo et al.
Page 47
REAGENT or RESOURCE
ABSOLUTE algorithm
Molecular Signatures Database (MSigDB) (v7.2)
SOURCE
IDENTIFIER
Carter et al., 2012
Broad Institute & UC San
Diego
Subramanian, Tamayo et al.,
2005
Liberzon et al., 2011
http://software.broadinstitute.org/
cancer/cga/absolute_download
https://www.gsea-msigdb.org/gsea/
msigdb/index.jsp
Bioconductor package regioneR (v1.18.1) with overlapPermTest
function
Gel et al., 2016
https://www.bioconductor.org/
packages/release/bioc/html/
regioneR.html
Ginkgo
InferCNV (V 1.10.1)
BWA men
Sambamba
MACS2 (v2.0.10)
DeepTools
phyper R function
ONT Albacore software (v2.3.1)
Nanopolish software (v0.10.2)
PRROC R-package
BioRender
Other
Lipofectamine 2000 reagent
LentiX concentrator
Polybrene
Nanopore MinION device with R9.4 flowcell
HRP conjugated secondary antibodies
Laemmli buffer
Garvin et al., 2015
http://qb.cshl.edu/ginkgo
Tickle et al., 2019
https://github.com/broadinstitute/
infercnv
Li and Durbin, 2009
https://github.com/lh3/bwa
Tarasov et al., 2015
https://github.com/biod/sambamba
Zhang et al., 2008
https://github.com/macs3-project/
MACS
Ramirez et al., 2016
https://github.com/deeptools/deepTools
Johnson et al., 1992
https://www.R-project.org/
Oxford Nanopore
Technologies
Simpson et al., 2017
Grau et al., 2015
https://nanoporetech.com/community
https://github.com/nanoporetech/
nanopolish
https://cran.r-project.org/web/
packages/PRROC/index.html
BioRender
https://www.biorender.com/
Invitrogen
Clontech Labs
Santa Cruz
Oxford Nanopore
Technologies
Bio-rad
Sigma
Cat#1668019
NC0448638
sc–134220
FLO–MIN106D
Cat#5196–2504
S3401–10VL
Cell. Author manuscript; available in PMC 2023 August 18.
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
A
u
t
h
o
r
M
a
n
u
s
c
r
i
p
t
| null |
10.1016_j.enpol.2022.113277.pdf
|
Data availability
No data was used for the research described in the article.
|
Data availability No data was used for the research described in the article.
|
Community wealth building in an age of just transitions: exploring civil
Community wealth building in an age of just transitions: exploring civil
society approaches to net zero and future research synergies
society approaches to net zero and future research synergies
Max Lacey-Barnacle, Adrian Smith, Tim Foxon
Publication date
Publication date
01-01-2023
Licence
Licence
This work is made available under the CC BY 4.0 licence and should only be used in accordance with that
licence. For more information on the specific terms, consult the repository record for this item.
Document Version
Document Version
Published version
Citation for this work (American Psychological Association 7th edition)
Citation for this work (American Psychological Association 7th edition)
Lacey-Barnacle, M., Smith, A., & Foxon, T. (2023). Community wealth building in an age of just transitions:
exploring civil society approaches to net zero and future research synergies (Version 1). University of Sussex.
https://hdl.handle.net/10779/uos.23492954.v1
Published in
Published in
Energy Policy
Link to external publisher version
Link to external publisher version
https://doi.org/10.1016/j.enpol.2022.113277
Copyright and reuse:
Copyright and reuse:
This work was downloaded from Sussex Research Open (SRO). This document is made available in line with publisher policy
and may differ from the published version. Please cite the published version where possible. Copyright and all moral rights to the
version of the paper presented here belong to the individual author(s) and/or other copyright owners unless otherwise stated. For
more information on this work, SRO or to report an issue, you can contact the repository administrators at [email protected].
Discover more of the University’s research at https://sussex.figshare.com/
Contents lists available at ScienceDirect
Energy Policy
journal homepage: www.elsevier.com/locate/enpol
Community wealth building in an age of just transitions: Exploring civil
society approaches to net zero and future research synergies
M. Lacey-Barnacle *, A. Smith , T.J. Foxon
Science Policy Research Unit, University of Sussex, UK
A R T I C L E I N F O
A B S T R A C T
Keywords:
Community wealth building
Grassroots innovations
Transition pathways
Community energy
Just transitions
Economic democracy
Community Wealth Building (CWB) is a burgeoning international policy agenda for local economic development
that seeks to enhance democratic ownership, retain the benefits of local economic activity and empower place-
based economies and workers. Parallel to this, in the context of net zero transitions, there has been increasing
interest in approaches to enhancing civil society and community ownership over local energy provision. How-
ever, in academic and practitioner debates, there has been very little interaction between these two strands of
thinking and action on the need for radical change in current energy provision, particularly as part of a wider
transformative change away from dominant neoliberal economic thinking, policies and structures.
In this Perspective, we explore the various ways in which synergies exist between CWB and energy transitions
by considering two civil society approaches to transitions; namely, the Thousand Flowers transition pathway and
research in Grassroots Innovations. We examine how community energy could be strengthened through CWB, by
showing how the ideas within these two approaches respond to the five core principles of CWB. Promising future
directions for research and practice are identified, including linking up CWB and just transitions strategies, a
renewed focus on local financial innovation and the growing role of anchor institutions in supporting net zero
transitions, particularly where CWB supports economic democracy transformations in new net zero economies.
1. Introduction
Community Wealth Building (CWB) is a burgeoning international
policy agenda for local economic development. CWB seeks to transform
local-scale economies by repurposing and redirecting the procurement
power of ‘anchor institutions’ towards local businesses and supply
chains. Five principles for economic democratisation guide these de-
velopments, diversifying ownership forms, retaining capital within lo-
calities and strengthening worker involvement, security and rights. CWB
arose as a counter to the dominance of neoliberal economic approaches
that prioritise the privatisation, mobility and extraction of local wealth
producing activities. Given growing interests in ‘just transitions’, we
propose that CWB might offer practical ideas for decarbonising energy
systems in which local economic empowerment, the democratisation of
ownership and long-term social sustainability become more central. This
is particularly important in a post-crisis ‘green recovery’.
The current administrations in both the US and the UK are
committed to ‘Build Back Better’, using responses to the global economic
downturn caused by the Coronavirus pandemic to address long-standing
(and worsened) social and economic challenges. This includes commit-
ments to addressing regional economic inequalities (e.g. by ‘levelling
up’ in the UK), as well as action to address climate change and promote a
net zero transition. However, many scholars, practitioners and re-
searchers are sceptical of mainstream approaches to addressing these
challenges (e.g. Alperovitz and Dubb, 2014; Kelly and Howard, 2019;
Guinan & O’Neill, 2020; Paul and Cumbers, 2021). There is criticism
towards the track record of mainstream approaches and scepticism to-
wards their future potential, which typically rely on measures to pro-
mote inward investment to economically deprived areas and regions,
alongside a focus on large-scale technology deployment and interna-
tional competitiveness, often dominated by large multinational
interests.
As a result, communities across the US, Canada, Australia, the UK
and Europe have been undertaking action to develop more bottom-up
alternatives for both local economic development and civic and com-
munity energy innovation. Burgeoning networks of policy, practice and
research have grown in these two areas, though they have largely
developed independently and therefore potential synergies may be
Abbreviations: Community Wealth Building, CWB.
* Corresponding author.
E-mail address: [email protected] (M. Lacey-Barnacle).
https://doi.org/10.1016/j.enpol.2022.113277
Received 1 March 2022; Received in revised form 27 July 2022; Accepted 24 September 2022
EnergyPolicy172(2023)113277Availableonline1November20220301-4215/©2022TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/). M. Lacey-Barnacle et al.
being missed. In this Perspective piece, we examine links and synergies
between (1) Community Wealth Building as a new international move-
ment for local economic development, and (2) civic/community energy
and grassroots innovation approaches to energy decarbonisation.
Through bringing these two fields together for the first time, we also aim
to encourage others to advance research at this interface.
1.1. The rise of community wealth building
CWB is a progressive policy, action and research movement that has
grown in prominence and stature over the past decade. Having first
emerged from the US before spreading to the UK (Hanna and Kelly,
2021; Guinan and O’Neill, 2020; Guinan and O’Neill, 2019), its trans-
atlantic origins are now beginning to be transcended as projects circu-
late across the globe, in locations as far flung as Australia (Fensham,
2020), Italy (Kohn, 2020), Tanzania (Collord, 2019) and closer to the US
in Canada (Hanna, 2019).
In both the UK and US, the locations for experimenting with CWB are
numerous; with both Cleveland (US) and Preston (UK) being celebrated
pioneers, whilst Oakland, Burlington, New York, Denver, Chicago and
Detroit are just some of the US cities adopting a CWB approach, and
local and regional governments in North Ayrshire, Newham, Islington,
Sunderland, Stevenage, Oldham, Wigan, the North of Tyne, Sandwell,
the Liverpool City Region, Lewes and Brighton & Hove are all adopting
CWB in the UK. This is occurring alongside new commitments from the
devolved governments of Wales and Scotland, with the Welsh govern-
ment reforming national procurement policies and strategies and the
Scottish government appointing a minister for community wealth and
introducing a Community Wealth Building bill into the Scottish Parlia-
ment (CLES, 2021). Given this substantial growth, we expect many more
cities, regions and nations to emerge as key CWB actors internationally.
CWB has to be understood through its origins as a direct response to
the dominance of a global neoliberal political economy over the past
four decades, which (under the guise of competition for inward invest-
ment by capital) has seen privatisation, deregulation and liberalisation
policies dominate the economic and political order of advanced liberal
democracies, particularly the US and UK (Harvey, 2007). However,
neoliberal approaches have largely failed to bring renewed prosperity to
deindustrialised cities and regions in the US and the UK. It is therefore
unsurprising that prominent CWB examples are emerging from these
two countries, in numerous place-based economies. In resisting the
neoliberal order, CWB proponents advocate five core principles in their
innovative approach to local economic development: [1] Plural
ownership of the economy [2] Making financial power work for local
places [3] Fair employment and just labour markets [4] Progressive
procurement of goods and services and [5] Socially productive use of
land and property (Manley and Whyman, 2021). In addition, CWB has
been defined as:
‘A local economic development strategy focused on building collaborative,
inclusive, sustainable, and democratically controlled local economies.
[…] these include worker cooperatives, community land trusts, commu-
nity development financial institutions, so-called ‘anchor institution’
procurement strategies, municipal and local public enterprises, and public
and community banking’ (Guinan & O’Neill, 2020 p13-14.)
CWB principles should be understood as resisting a globalised
neoliberal economic system that has increasingly contracted out public
services to (multinational) private companies, thereby reducing di-
versity and ownership over local economic activity and empowering
private financial and commercial institutions to own, manage and
govern key public goods (Peters, 2012; Williams et al., 2014). This in-
cludes energy and transport - key sectors for net zero transitions - that
have also been privatised and liberalised in this way (Bayliss et al.,
2021). Further, CWB principles can be connected to research advancing
new conceptualisations of more progressive local economies, such as
work on the ‘Foundational Economy’ (Heslop et al., 2019; Hansen,
2021), ’New Municpalism’ (Thompson, 2021) ‘Re-municipalisation’
(Cumbers, 2016; Paul and Cumbers, 2021), local policy responses to
globalisation (Imbroscio et al., 2003) and critical work on the regressive
impact of neoliberalism on localism (Catney et al., 2014; Davoudi and
Madanipour, 2015).
Finally, the intellectual influence of key US scholars is important
here, particularly the influence of scholars such as Gar Alperovitz and
Marjorie Kelly, who are both core members of the ‘Democracy Collab-
orative’ think tank in the United States – a key advocate for CWB and a
core actor at the heart of the successful ‘Cleveland model’ in the US
(Lenihan, 2014). Going back half a century, Alperovltz (1972) coined
the concept of a ‘pluralist commonwealth’ and continues to publish on
the relevance of this concept for progressive economic reform today
(Alperovitz and Dubb, 2014; Alperovitz, 2020). Seen as a precursor to
CWB, the pluralist commonwealth is defined by four principles; the
‘democratisation of wealth’, ‘community as a guiding theme’, ‘decen-
tralization’ and ‘democratic planning’ (Alperovitz, 2020). These prin-
ciples demonstrate strong overlaps with Alperovitz and Dubb (2017),
who drew upon these when mobilising for revitalising and regenerating
Detroit, moving from theory to practice – a key hallmark of the CWB
policy community.
With reference to democratising ownership and the first principle of
CWB, Kelly (2012), in ‘Owning our Future’, distinguishes between
‘extractive’ and ‘generative’ ownership forms. Extractive forms of
ownership cater to an international shareholder class or ‘absentee
membership’, where organisations - embedded in global capital markets
and financialised networks - seek to move between a series of profit
maximising opportunities in the short-term, above all other interests.
The business generated by these investments are assumed to
trickle-down to local actors. Generative ownership, in contrast, sees
‘rooted membership’ in local, public and civil society forms of organi-
sation. Governance is controlled by those dedicated to a ‘social mission’
and organisations are constructed around both
long-term and
sustainability-oriented goals (Kelly, 2012). This generative/extractive
distinction has influenced work on democratic economies (Kelly and
Howard, 2019; Hanna and Kelly, 2021), whilst calls for more ‘genera-
tive’ economies now appear amongst UK advocates for CWB (McInroy,
2020). Additionally, this distinction is also vital for civil society ap-
proaches to net zero transitions and civic and community energy
structures of organisation.
Whilst clearly principled, at its core CWB is nevertheless deliberately
pragmatic. Anchor institutions, such as universities, hospitals, schools,
prisons, local government, housing associations, trade unions or large
local companies/social enterprises, are all fixed in place and rooted to a
locality or region by virtue of their organisational design, ’anchoring’
them to their local economies. Through pursuing a CWB approach, these
anchor institutions seek to work in partnership with CWB organisations
to switch their service contracts from multinational to local supply
chains. Simultaneously, capacity is built up in local supply cooperatives
and partnerships through coordinated facilitation and anchor networks.
In Preston (UK), the promotion of a CWB approach by the local council
has led to the percentage of total procurement spending in the city going
up from 5% in 2013 to 18% in 2017, and from 39% in 2013 to 79% in
2017 across the Lancashire region (Jackson and McInroy, 2017). Un-
employment has fallen from 6.5% in 2014 to 3.1% (O’Neill and Guinan,
2020). Preston has been named the most improved city in the UK ‘Good
Growth for Cities 2018’ index and has moved from 143rd to 130th in the
Social Mobility Commission Index. A further 4000 employees –
including all council workers – now receive the Living Wage (Hadfield,
2019). In addition to this, a complex network of mutually supportive
co-operatives and social enterprises has developed in Preston, under-
pinned by the establishment of the Preston Cooperative Development
Network, with the support of anchor institutions in the form of the local
university and the local city council (Manley and Whyman, 2021).
CWB thus positions itself as a pragmatically progressive form of
bottom-up, locally-led social and economic development. And yet, there
EnergyPolicy172(2023)1132772 M. Lacey-Barnacle et al.
has seldom been investigations into how such achievements might
connect to, complement or support the challenge confronting all local-
ities – net zero transitions. We therefore ask the following two questions
in this Perspective piece:
1. What synergies are there between CWB and civil society approaches
to local net zero & sustainability transitions?
2. How can future research and policy support these synergies in
practice?
These questions are intended to open up lines of inquiry (and ac-
tivity) into how CWB can engage in net zero transitions. In the next
section, we argue that many of the CWB principles align well with
emerging calls for just transitions in energy systems.
1.2. New frontiers: community wealth building and the green economy
Whilst recent academic research into CWB analyses its potential in
local economic development (Barnes et al., 2020; Manley and Whyman,
2021; Eder, 2021; Webster et al., 2021; Dubb, 2016), few of these studies
explicitly address how CWB can engage with the green economy. Suc-
cinct reviews of CWB for a general audience (O’Neill and Guinan, 2020),
detailed essays on the history and future of CWB (Hanna and Kelly,
2021) and books devoted to engaging a wider audience in the history of
the Preston Model (Brown and Jones, 2021) and deepening our aca-
demic understanding of both the Preston Model and CWB (Manley and
Whyman, 2021), all contain very little acknowledgement of its potential
relevance to local net zero transitions. Furthermore, a practical 51-page
toolkit designed to assist local councillors in implementing CWB con-
tains only one mention of ‘net zero’ (Democracy Collaborative and
Momentum, 2022). This is surprising, given that CWB is a trans-
formational economic project and net zero transitions imply significant
economic reorientations. Moreover, calls for just transitions to a net zero
economy open an opportunity for CWB to enter into this terrain (Wang
and Lo, 2021), alongside challenges to established notions of what
constitutes a just transition, moving beyond a core focus on providing
‘green jobs’ in the face of a retracting fossil fuel industry (McCauley and
Heffron, 2018), to understanding how justice, equity and inequality are
constituted in new net zero economies (Morena et al., 2020) and how
civil society and grassroots mobilisations for a just transition can be
supported by the state (Routledge et al., 2018), specifically at the local
level.
There are signs that this disconnect between CWB and net zero
transitions is beginning to be bridged, particularly in policy research.
The think tank CommonWealth’s work on ‘Community Wealth Building
for Economic and Environmental Justice’ (Brown et al., 2019) shows
early signs of bringing together the two disconnected fields, arguing for
anchor institutions to play pivotal roles in supporting local Green New
Deals, green jobs and ‘green procurement policies’, whilst the Centre for
Local Economic Strategies (CLES) report & toolkit on a ‘just energy
transition through community wealth building’ (Radcliffe and Williams,
2021), alongside the Democracy Collaborative’s report on ‘Publicly
owned and cooperative electric utilities as anchors for community wealth
building and a just energy transition’ (Hanna et al., 2022) demonstrate the
emergence of a new policy-research field.
Turning briefly to these last two outputs, Radcliffe and Williams
(2021) note a vital role for local authorities to intervene in energy
transitions to advance CWB, where they play key roles in; (1) Acting as
convenor (2) Creating demand (3) Direct delivery of transition projects
(4) Encouraging the early adoption of zero carbon technology and (5)
Funding the energy transition. The authors connect these roles to the
five principles of CWB throughout the report, whilst also noting that
anchor institutions have ‘a critical role in enabling cross-sector approaches
to energy transition which build community wealth’ (Radcliffe and Wil-
liams, 2021 p.14). This connects well to Hanna et al. (2022), who see
‘community utilities’ that are co-operatively and publicly owned as
fundamental additions to the plethora of possible anchor institutions.
The authors advance nine key policy recommendations for building
community wealth in energy markets and transitions; (1) Block Priva-
tisation (2) Deeper Democratic Governance (3) Renewable Energy
Mandates (4) Renewable Energy Financial Incentives (5) Public
Distributed Renewable Energy & Electrification (6) Procurement Pro-
grams (7) Public Banking & Finance (8) Supporting Local Innovation
and (9) Public finance for shifting ‘Investor Owned Utilities’ into Public
and Co-operative Ownership. Both reports demonstrate renewed atten-
tion being paid to critical connections between CWB and civil society-led
energy transitions.
Interestingly, older outputs from the Democracy Collaborative, such as
Warren’s (2010) report entitled ‘Growing a Green Economy for All: from
Green Jobs to Green Ownership’, pay attention to this juncture between
CWB and green economy transitions, whilst the Cleveland model’s
network of different organisations known as the ‘Evergreen Co-
operatives’ supported local food growing, sustainable laundry and local
solar PV deployment (Lenihan, 2014). As Sheffield (2017) reports, many
of the Evergreen Cooperatives are now profitable, employing over 150
people locally, with plans to increase this number to 1000. In the case of
the Evergreen Cooperative Laundry, for example:
‘After a six-month initial “probationary” period, employees begin to
buy into the company through payroll deductions of 50 cents an hour
over three years (for a total of $3,000). Employee-owners are likely
to build up a $65,000 equity stake in the business over eight to nine
years, a substantial amount of money in one of the hardest-hit urban
neighbourhoods in the nation’ (Alperovitz et al., 2010 p.1).
Indeed, this novel form of democratic ownership and governance –
facilitated through laundry service contracts with anchor institutions
(local hospitals and universities) - led Lenihan (2014) to describe the
Cleveland model as;
‘The most robust ongoing American effort to enjoin the economic power of
anchor institutions (and their growing ecological sensitivity) with the
development goal of creating widely shared and more democratic asset
and capital building in low-income neighborhoods’ (Lenihan, 2014 p.18
p.18)
Despite this promising connection with sustainable transitions - both
past and present - academic research seems to be severely lagging
behind. We argue that the introduction of CWB research into the energy
transitions terrain presents scope for facilitative links with established
civil society approaches and theories in bottom-up and local energy
transitions, such as civic energy sector transition pathways (Foxon,
2013) and grassroots innovations that seek to directly tackle the chal-
lenges of sustainability transitions from the bottom-up (Smith and
Seyfang, 2013; Smith et al., 2016). In the following section, we analyse
more closely the links between CWB and these relevant approaches to
civil society-led energy transitions.
2. Community wealth building and energy transitions:
theoretical links
In this section, we first explore the ways in which synergies already
exist between CWB and two widely cited civil society approaches for
local sustainability transitions: The Thousand Flowers transition pathway
and its associated concept of a ‘Civic Energy Sector’ and the theory of
Grassroots Innovations. A variety of related research fields could addi-
tionally be explored, such as research on decentralised ownership and
control over energy systems (Brisbois, 2019), polycentric governance
(Bauwens, 2017), local community power (Kaye, 2020) and a rich his-
tory of community energy research (Lacey-Barnacle, 2020; Creamer,
2018; Smith et al., 2016; Seyfang et al., 2013; Walker et al., 2007). All of
these fields connect to both the Thousand Flowers and Grassroots In-
novations literature; however, a review of more comprehensive links to
CWB is beyond the scope of our paper. Our examination here, of the
EnergyPolicy172(2023)1132773 M. Lacey-Barnacle et al.
different ways in which two illustrative approaches in community-based
energy developments and CWB share similar goals, values and ap-
proaches, can inform future bridge-building research endeavours.
ownership forms that advance direct community control, alongside
providing inclusive energy tariff offers and energy efficiency services to
vulnerable groups (Hoicka et al., 2021; Hanke et al., 2021).
2.1. Transition pathways and the civic energy sector
Influenced by work on socio-technical transitions and the multi-level
perspective on systems transformations (Verbong & Loorbach 2012;
Geels 2002), the ‘Realising Transition Pathways’ research consortium, an
8-year multi-institution project spanning 2008–2016, produced
considerable material and research outputs to assist UK government
policymakers and academic research communities in grappling with the
complexities of transitioning to a low-carbon energy system by 2050.
The Pathways project developed detailed potential paths that would
achieve this momentous transition. Associated outputs analysed how to
‘bring social structures and agency, including institutions and politics, into
scenario […] studies of sustainable energy futures’ (Foxon, 2013 p.12).
These scenarios enhanced understanding of the political and economic
challenges and opportunities in UK low-carbon futures (compared to the
technology-dominant scenarios in many energy scenarios and pathways
studies). Different institutional and socio-technical configurations were
explored for meeting the UK’s legally binding commitment (in the
Climate Change Act 2008) to reduce GHG emissions by 80% by 2050
against a 1990 baseline. Three different transition pathways were con-
trasted; Market Rules, Central Co-ordination and Thousand Flowers. Each
pathway adheres to different governance logics in which power relations
between market, state and civil society actors are varied (Foxon et al.,
2010; Foxon, 2013; Barnacle et al., 2013; Chilvers et al., 2017).
The Thousand Flowers pathway provides one of the few detailed ex-
plorations of the greater role that civil society can play within future UK
energy transitions. The pathway sees a ‘growing dominance of civil society
in the governance of UK energy systems, which leads to an increase in di-
versity of local bottom-up solutions for providing decentralised generation
and energy conservation options’ (Barnacle et al., 2013 p.60). One
outcome of this growing role for civil society in municipal and com-
munity governance of energy, is the development of a ‘civic energy
sector’, a scenario which delivers 50% of final electricity demand by
2050 (Hall et al., 2016).
Central to this vision is a vibrant community energy sector, where
community organisations take a leading role in purchasing, managing
and governing local energy projects and infrastructures. A heavily
researched field of both policy and practice, community energy has very
often been seen by many researchers as particularly competent in
meeting varying social, environmental and economic objectives at the
local level (Zoellner et al., 2008; Warren, 2010; Musall & Kuik 2011;
Seyfang et al., 2013). For example, community energy projects have
encouraged and enabled the active participation of members of the local
community in energy transition processes, while introducing behaviour
change schemes and energy demand reduction into local communities.
Secondly, many schemes have drawn upon local investment and tapped
into local expertise and enthusiasm for renewable energy installations,
raising the necessary capital and increasing local acceptance through
direct community ownership. The wealth generated by newly-valuable
renewable resources thereby circulates and multiplies more locally.
Civic initiatives cultivate multi-actor partnerships working across mul-
tiple scales to engage in and support transitions, and, using this
multi-scalar collaboration, have been able to appropriately tailor local
renewable energy deployment to the technological, political and eco-
nomic specificities of a locality (Walker et al., 2007; Walker et al., 2007;
Seyfang et al., 2013; Hargreaves et al., 2013; Bauwens et al., 2016).
Importantly, the emergence of community and civic energy schemes
is now influencing policy. For example, as part of the EU’s Clean Energy
Package, ‘Energy Communities’ are now formally recognised as essential
civil society entities which will aid the EU’s broader decarbonisation
plans. Recent research also points towards their potential to contribute
to a more just and democratic transition, particularly through novel
Many of the above elements of local and community energy effec-
tively align with CWB approaches, whilst also encouraging the broader
empowerment of civil society actors. In a review of community energy
projects in Europe, Hewitt et al. (2019) note that four aspects of com-
munity energy projects underpin their potential for contributing to-
wards social innovation; (1) Crises and opportunities; (2) the agency of
civil society; (3) reconfiguration of social practices, institutions and
networks; (4) new ways of working. All four of these aspects of com-
munity energy schemes connect closely to CWB. The trigger for CWB in
Preston, for example, was the collapse of a £700m inward-investment
regeneration project in the wake of the global financial crisis, and
therefore, the search for locally-resilient opportunities to develop the
local economy resulted in a CWB approach (and inspired by the US
Cleveland model) (Manley and Whyman, 2021). CWB typically seeks to
enhance and empower the agency of civil society within multi-actor
partnerships and to reconfigure institutions and networks, whilst the
five principles of CWB foster new ways of forging those relations at the
local scale. Importantly, community energy connects with CWB by
seeking to localise and retain wealth and surplus revenue creation
(Lacey-Barnacle, 2019; Stewart, 2021), democratise governance and
engagement in local economies (Van Veelen, 2018) and experiment with
novel social enterprise models and organisational structures (Becker
et al., 2017). Forming a core part of the civic energy sector as outlined in
the Thousand Flowers pathway, community energy schemes can be
considered a vital part of local strategies to build ‘community wealth’.
However, as we explore in subsection 2.2, this wealth is not always
equitably shared and CWB may offer a point of strategic intervention to
address more equitably some historic shortcomings in civic energy
approaches.
Whilst civil society is crucial, this does not negate roles for the state
or market. Barton et al. (2015) note, through the prism of back-casting,
that the Thousand Flowers pathway shifts the role of local government,
as:
‘Local energy ownership became a focus of local government economic
development […] as the scale of the opportunity became clear in terms of
local value capture, net employment creation, and energy security’
(Barton et al., 2015 p.5 p.5)
This ‘local value capture’ connects well to the redirection of pro-
curement processes in CWB advocacy; whilst the focus on local energy
ownership also demonstrates synergies with the ‘plural ownership of the
economy’ principle. Indeed, the authors note that the ‘expansion of this
sector would capture much of the value from energy production and con-
sumption that currently leaks out of the local economy’ (Barton et al., 2015
p.27), demonstrating strong support for wealth retention within local
economies. Furthermore, when anticipating how the Thousand Flowers
pathway is achieved, the authors note that ‘local energy schemes devel-
oped stable and familiar financial relationships with the local banking sector,
which viewed civic power generation as a safe asset’ (Barton et al., 2015
p.5), connecting strongly to the CWB principle centred on making
financial power work for local places.
Drawing on the example of Germany as a ‘co-ordinated market
economy’ (Hall and Soskice, 2001), Hall et al. (2016) show the impor-
tance of the German local banking sector in facilitating civic ownership
structures. This is in contrast to the UK neo-liberal economic model, in
which financial institutions have a national and international focus and
arguably are more motivated by short-term shareholder returns than
long-term stable investment relationships with local partners. Interest-
ingly, new bottom-up financial innovations, in the form of local
municipal energy bonds, are now being developed in the UK (Davis,
2021; Green Finance Institute, 2021). These provide a simple, low-risk
way to enable members of local communities to invest in local renew-
able energy developments, by making use of the financial security of
EnergyPolicy172(2023)1132774 M. Lacey-Barnacle et al.
local municipal authorities. This approach could thus contribute to the
second principle of CWB, whilst also allowing local financial innovation
to be governed and managed by public institutions. Indeed, many of
these crossovers between CWB and the Thousand Flowers pathway show
that new local energy supply models have the potential to incorporate
more complex value propositions, including economic, social and
environmental values (Hall and Roelich, 2016).
Intriguingly, the role of anchor institutions in leveraging procure-
ment spending in support of local net zero innovation, projects and
goals, has been understudied in civic energy research (Uyarra et al.,
2016). The Thousand Flowers pathway does not conceive of anchor in-
stitutions in its detailed scenarios. In identifying key anchor institutions,
such as local hospitals, universities and local government, CWB brings
another mechanism to civic and community energy that can facilitate
novel contractual arrangements to support the growth of local net zero
energy projects and supply chains: contracting energy co-operatives to
provide energy consulting services, supplies of clean electricity, effi-
ciency measures, and supporting community flexibility arrangements in
smart local energy systems. Energy transitions could also form a more
explicit part of what CWB scholars call the ‘anchor mission’ (Kelly et al.,
2016), where their local economic power is used to strengthen local
enterprise, with a focus on socially inclusive organisations. Here,
through aligning anchor missions with net zero transitions, anchor in-
stitutions can be used to offer preferential treatment to organisations
that simultaneously pursue inclusive decarbonisation.
2.2. Grassroots innovations, local sustainability transitions and CWB
In contrast to future scenario conditions under which empowered
civic energy generation might become more widespread, research into
grassroots innovation was borne of historical and contemporary analysis
into innovative local sustainability initiatives. These often develop
despite existing realities being unconducive to such initiatives (Seyfang
and Smith, 2007; Fressoli et al., 2014). Local environmental initiative
was reframed as grassroots innovation, in which networks of neigh-
bours, activists, social entrepreneurs, community organisations, co-
operatives, and others worked creatively and innovatively in generating
and circulating bottom-up solutions for sustainability appropriate to the
needs, aspirations and situations of those involved.
In conceiving local environmental activity as innovative and gener-
ative of wider change, so studies were able to adapt analytical resources
in innovation studies and sustainability transitions. This enabled better
understanding of how grassroots movements produce knowledge,
reframe problems, form networks and attract resources, govern them-
selves and challenge institutions, and thereby develop and diffuse ap-
proaches and solutions for sustainability across localities in ways quite
different to conventional market- and state-based institutions for inno-
vation (Hess, 2007; Jamison, 2001; Smith and Stirling, 2018). Early
research (in the 2010s) included studies of community energy, analysis
of grassroots innovation in food, housing, manufacturing, mobility; as
well as historical research into earlier movements for alternative tech-
nology, socially useful production; and initiatives in the global South as
well as global North (Smith et al., 2017; Pansera and Owen, 2017;
Gupta, 2016). Theories about the development of ‘niche spaces’ for
alternative innovation within the context of unfavourable incumbent
energy regimes were used to explain the achievements and challenges
confronting grassroots action (Smith, 2007).
Grassroots Innovations can seek to change markets and prevailing
market systems, despite sometimes being framed as an alternative to the
market or as a more radical response to the failure of dominant and
mainstream institutions on environmental issues (Feola and Nunes,
2014; Seyfang and Smith, 2007). They do this through the utilisation of
a set of unique characteristics that set them apart from market and
technology-oriented niche innovations (Fressoli et al., 2014). In the
context of community energy, Hargreaves et al. (2013) identify these
unique characteristics as: ‘Distinct organisational forms’; ‘Different
resource bases’; ‘Divergent contextual situations’; ‘Alternative driving
motivations’; and ‘the pursuit of qualitatively different kinds of sus-
tainable development’ (Hargreaves et al., 2013). Indeed, prominent
theorists of Grassroots Innovations suggest that, whilst it is particularly
hard to correlate similarities across cases of local innovation that are by
definition tailored to the specificities of a locality, many grassroots in-
novations will draw upon social enterprise models or function more
broadly within the social economy (Seyfang and Smith, 2007; Har-
greaves et al., 2013; Smith, 2014). Thus, it is important to note that:
‘Grassroots innovation processes share a broadly similar vision and
shared set of principles, regarding local inclusion and control in processes
of technology development and innovative social organisation […]
grassroots innovation movements confront similar fundamental chal-
lenges, even though manifesting in particular ways in contrasting settings’
(Smith, 2014 p.115)
Here, we can already see some strong connections to the five prin-
ciples of CWB. Firstly, the use of ‘distinct organisational forms’ to sup-
port grassroots innovations opens bridges to the demand for more ‘plural
ownership of the economy’ by CWB advocates. Arguably, grassroots
innovation has tended to gloss over questions of ownership and attended
more to participation, so more explicit engagement with diversifying
ownership, in line with CWB, can provide more depth here. Secondly,
the reliance of CWB approaches on anchor institutions and the redi-
rection of procurement processes to support local economies connects
well to the reliance of grassroots innovations on ‘different resource
bases’, which is supported further by the local financial innovation
sought by CWB actors. Lastly, the desire that CWB advocates have for
new models of local economic development that cater to the needs of
different localities are reflective of the ‘alternative driving motivations’
and
that underpin grassroots
innovations.
‘divergent contextual situations’
While these multiple connections are important, there is one incon-
sistency. Differences between CWB and grassroots innovations are found
in the limited engagement of CWB literature in sustainability transitions
and the importance of path-breaking innovations for future net zero
economies, whilst grassroots innovations are often explicitly framed
around contributing towards ‘different kinds of sustainable develop-
ment’. Innovation and transformation as a goal and topic is not so
prominent in CWB practice, where activity rests in carving out oppor-
tunities within the given local economy. And yet, the five principles
imply considerable organisational, business, process and product inno-
vation, and even some changes to the contexts and purposes for tech-
nological change which is the conventional focus of innovation. If CWB
succeeds in bringing in a diversity of actors into local economic devel-
opment (e.g. via anchor networks), the insights from grassroots inno-
vation concerning how these alternative constellations can better
approach innovation and transition could prove helpful.
For example, a dilemma typical for many grassroots innovation
movements seeking to scale-up, circulate more widely, and generally
expand their niche innovations, is whether to align more closely with the
logics of incumbent institutions for innovation (such as through
commodification, intellectual property, and standardisation, thereby
blunting their transformational potential) or to remain radical and
continue pressing for radical reforms to powerful institutions. Such
radical reforms ensure that innovation is conceived and supported using
the participatory democratic norms and commons-based ownership
models favoured by grassroots innovations.
Dynamic tensions exist between ‘fit-and-conform’ versus ‘stretch-
and-transform’ strategies for developing niche spaces: making them
more palatable to prevailing institutions, or building power to transform
those institutions (Smith and Raven, 2012). Analogous dilemmas might
EnergyPolicy172(2023)1132775 M. Lacey-Barnacle et al.
be evident in relations between local economic enterprises and anchor
institutions who, no matter how sympathetic to worker control, say, or
cooperative ownership, might be structurally constrained as to how far
they can depart from norms of supply-chain and service-provision under
capitalism as currently instituted (Smith, 2014). Without a better
appreciation of the complexities of transformative innovation, there is a
risk that CWB measures will tend towards safe, conservative economic
activities or privilege the experimental designs of organisations with the
resources to instigate them. That said, an enduring challenge is moving
beyond creative prototypes and start-up organisational forms, to build
enduring structures and institutions capable of enabling these novelties
to succeed over the long-term (as seen with civic energy generation in
the Thousand Flowers pathway). It might be that moving from innovation
to diffusion, in ways that remain transformational and resist falling into
conformity, is a challenge where insights from CWB can be helpful. CWB
can help cultivate the capabilities, investment and work to develop
innovation more consistently with motivating ideals: so, for example,
community energy schemes remain locally democratic and accountable,
rather than becoming increasingly utility-like. This is a goal that Hanna
et al. (2022) say is fundamental to a CWB approach to energy
transitions.
However, accountability is not the only issue facing community en-
ergy schemes. Community energy risks replicating issues around social
inclusion. For example, researchers such as Catney et al. (2014) and
Seyfang et al. (2013), when offering critical perspectives on community
energy projects, note that much of the literature surrounding commu-
nity energy focuses explicitly on the success stories of the sector, with
little attention given to understanding which communities are unable to
engage in these initiatives and why, leaving out considerations of how to
bring about a more socially ‘just’ transition. Furthermore, Johnson et al.
(2014) find that a decentralised energy system could risk reproducing,
or even worsening, existing socio-economic inequalities within society.
It is important to ask, therefore, whether CWB may encounter similar
risks. Given the primacy of the local state and the need for representative
political leadership to support CWB, we feel a CWB approach could
avoid the pitfalls of a socially exclusive local economic development
approach. Social justice and inclusion considerations are of vital
importance to emerging CWB policy programmes and approaches,
which we feel could be used to address and rectify some of the existing
inequalities in access to community energy schemes and to advance a
more inclusive just transition.
2.3. Comparing CWB, grassroots innovation and thousand flowers
pathway approaches
The above discussions suggest that CWB may be able to offer a
normative direction to the kinds of transformation that many have
argued are necessary for net zero transitions (e.g. more democratic, just,
community-based, socially inclusive etc), addressing areas where soci-
otechnical energy transitions research has been agnostic and lacking.
CWB, aided by its five principles, also has the potential to inject local,
municipal and community energy with stronger elements of democratic
directionality, underpinned by a strong social justice ethos (Kelly et al.,
2016). CWB might thus be a counter to the financialisation and
extraction of local energy initiatives that comes with private institu-
tional investment and corporate control, whilst – with the support of
anchor institutions - offering stability and finance to develop more
democratic and plural economic organisations. Linking back to our
discussion of the Thousand Flowers pathway, there is clearly a key role
for anchor institutions to play in a future where CWB becomes more
closely aligned with the transition to a net zero economy. Drawing on
our analysis above, we further summarise the critical overlaps and
synergies between CWB and grassroots innovations and Thousand
Flowers transition pathway in our table below (Table 1):
Table 1
Overlaps between CWB principles and civil society approaches to transitions.
CWB Principles
Grassroots innovations
Thousand Flowers pathway
[1] Plural ownership of
the economy
Distinct organisational
forms
[2] Making financial
power work for local
places
[3] Fair employment
and just labour
markets
[4] Progressive
procurement of
goods and services
Divergent contextual
situations/Different
resource bases
Alternative driving
motivations
Different resource bases
[5] Socially productive
use of land and
property
Different kinds of
sustainable
development
Dominance of civil society
in the governance of UK
energy systems
Key financial relationships
between the local banking
sector & civic energy sector
Net employment creation
Local value capture/capture
local value of energy
production and
consumption
Local energy ownership a
focus of local government
economic development
3. Conclusion – The future of community wealth building and
civil society-led just transitions
CWB is emerging at a timely and critical juncture; given its recent
expansion over the past decade, it is already demonstrating more dem-
ocratic forms of local economic development, with potential for making
more just a rapidly expanding net zero economy. This opens up new
pathways and future scenarios for radical, diverse and pragmatic ap-
plications of CWB to net zero economies. Our Perspective piece has
outlined the need for CWB to see the transition to a green economy as a
novel opportunity to expand its activities and scope, particularly as
global trends towards decentralised net zero transitions and devolved
governance continues apace (Rodríguez-Pose and Gill, 2003; Burger
et al., 2020). This is particularly true for local and decentralised energy
markets, where new innovations, technologies and organisations are
hastening the shift from centralised to decentralised energy systems.
Smart local energy systems, locational pricing, bespoke tariffs,
peer-to-peer trading of local surplus electricity, flexibility markets and
community opportunities for energy storage, alongside engagement in
vehicle-to-grid markets, are just some of the net zero innovations that
can be integrated into CWB strategies for engagement in energy markets.
In this space, public and community ownership of novel technologies
and platforms will be key to contributing to CWB synergies with net
zero. Research on the financial benefits of community ownership shows
that community-owned wind farms pay their communities 34 times
more than their commercial (private) counterparts (Aquatera, 2021),
whilst co-operative and community energy schemes are more effective
in connecting the benefits of low-carbon technologies to deprived
communities than individual, household models of deployment (Stew-
art, 2021).
Our Perspective thus has strong normative underpinnings; we see the
presence of more plural, democratic, public and civil society forms of
ownership and governance in net zero economies as constituting more
‘just’ forms of organisation, particularly when citizens and workers at
the heart of local communities and economies are given greater auton-
omy and agency in the face of historic corporate and state control over
the expansion of the green economy. Understandings of a ‘just transi-
tion’ cannot, therefore, be divorced from broader questions of owner-
ship and governance in our economy and further explorations of civil
society-led pathways to a just transition are vital. As we have high-
lighted, anchor institutions and emerging anchor networks, supported
by the local state, will be key in designing justice interventions and
advancing social justice aims, where marginalised and deprived com-
munities are placed at the forefront of local green recovery and regen-
eration strategies. This is a vital area of future research for both the CWB
and just transitions research communities.
Moreover, CWB opens up wider discussions on the role of local
EnergyPolicy172(2023)1132776 M. Lacey-Barnacle et al.
economic democracy, its potential, its limits, and how it may effectively
engage with net zero transitions and reshape our conception of a just
transition. While both grassroots innovations and transition pathway
literatures have acknowledged democratic ownership forms in sustain-
ability transitions, often through the guise of a civic energy sector and
community energy schemes, CWB has the potential to highlight how a
local-state-backed form of economic democratisation can strengthen
these endeavours, by drawing on successful examples and empirical
analysis emerging across the world. However, cautioned by analysis of
grassroots innovation alluded to above, CWB may, without due atten-
tion, succumb to a ‘fit and conform’ strategy, where CWB is ultimately
used to reinforce existing market-oriented power structures through
using procurement to support local businesses and supply chains, which
do not attend to decarbonisation and sustainability goals. In contrast, if
CWB pursues a ‘stretch and transform’ approach, it can be used to move
economic democratisation towards the heart of sustainable energy
transitions, incorporating stronger social justice goals as outlined above.
Despite this positive outlook, we do, however, exercise caution here;
it is vital to not advocate for CWB uncritically. As Manley and Whyman
(2021) point out, CWB has top-down tendencies whenever its goals and
visions are set by local political leaders, rather than through local citi-
zens and civil society deliberation. Ensuring economic democratisation
in CWB is supported by appropriate development networks and educa-
tional schemes has the potential to counter any technocratic tendency.
This is vitally important, as research points towards a key role for eco-
nomic democracy in enhancing both equality and sustainability in so-
ciety (Power et al., 2016).
Thus, whilst our paper has championed the possibilities of bridging
with just sustainability transitions, we conclude by acknowledging three
key challenges for CWB and just transitions that future action-research
must address to assess its feasibility and potential to promote trans-
formative change:
(1) Linking up CWB and just transitions policies and strategies -
Although CWB is receiving local and regional policy support
across the globe, support for just transitions appears at multiple
levels of governance globally and plays host to political support
and buy-in at a broader scale than CWB. With broader top-down
policy support and financing, CWB could arguably be more
transformative and impactful. There is significant space in both
research and policy to explore linkages, alignments and com-
plementarities between CWB and just transitions to a net zero
economy, particularly in ‘left behind’ areas, regions and com-
munities that are seeking bold regeneration strategies after de-
cades of deindustrialisation.
(2) Local financial innovation - The role of finance is critical in
decarbonising the economy and many of the financial mecha-
nisms and innovations required for current net zero targets are
beyond the reach of CWB, particularly in highly centralised
financial systems. However, through the redirection of procure-
ment practices that CWB advocates for, there is potential to
redirect local spending towards climate goals and experiment
with local financial innovation, such as Community Municipal
Bonds (Davis, 2021), to support local decarbonisation and local
wealth retention in net zero transitions. Research into how to
unlock and access finance locally to support CWB approaches to
new net zero economies will be vital in coming years, alongside
exploring further how local financial innovations can support and
link up to broader just transition concerns.
(3) Anchor institutions supporting just net zero transitions – it is
clear that anchor institutions, with their associated procurement
power and natural embeddedness within place-based economies,
will have a vital role to play in ensuring they use their local
economic power to support grassroots innovations and civic en-
ergy projects. The expansion of an ‘anchor mission’ – to include
inclusive and sustainable local enterprise – will be fundamental
to this challenge. Anchor institutions should ideally give prefer-
ential treatment to democratic organisational structures in their
economic developments. This would contribute towards the
transformational potential that CWB promises to local economies
across the world.
These three challenges overlap. Local just transitions strategies,
emboldened by CWB agendas, will need to ensure that key anchor in-
stitutions support local financial innovation and leverage procurement
spending to advance CWB approaches to net zero economies, whilst also
supporting economic democratisation as part of reconceptualised just
transitions. The novel insights offered in this Perspective suggest such an
endeavour is worth embarking upon, in research, policy and practice.
CRediT authorship contribution statement
M. Lacey-Barnacle: Conceptualization, Supervision, Project
administration, Writing – original draft, Writing – review & editing,
Funding acquisition. A. Smith: Conceptualization, Writing – original
draft, Writing – review & editing. T.J. Foxon: Conceptualization,
Writing – original draft, Writing – review & editing.
Declaration of competing interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Data availability
No data was used for the research described in the article.
Acknowledgements
The authors wish to thank the reviewers for their time and their
insightful comments which helped improve the paper. This research was
supported by a Leverhulme Trust Early Career Research Fellowship ECF-
2021-191, of which the lead author is a recipient.
References
Alperovitz, G., 2020. A pluralist commonwealth and a community sustaining system. In:
The New Systems Reader. Routledge, pp. 226–252.
Alperovitz, G., Williamson, T., Howard, T., 2010. The Cleveland model. Nation 1 (1),
21–24.
Alperovitz, G., Dubb, S., 2014. The possibility of a pluralist commonwealth and a
community sustaining economy. In: From Capitalistic to Humanistic Business.
Palgrave Macmillan, London, pp. 73–96.
Alperovltz, G., 1972. Notes towards a pluralist commonwealth. Rev. Radic. Polit. Econ. 4
(2), 28–48.
Alperowitz, G., Dubb, S., 2017. A community wealth-building vision for Detroit—and
beyond. In: Reinventing Detroit. Routledge, pp. 183–200.
Aquatera, 2021. A Comparison of the Financial Benefits Arising from Private and
Community Owned Wind Farms. Report by Aquatera Ltd. Report for Point and
Sandwick Trust.
Barnes, M.C., Walker, C.D., Williamson, T., 2020. Community Wealth Building and the
Reconstruction of American Democracy: Can We Make American Democracy Work?
Edward Elgar Publishing.
Barnacle, M., Robertson, E., Galloway, S., Barton, J., Ault, G., 2013. Modelling
generation and infrastructure requirements for transition pathways. Energy Pol. 52,
60–75.
Barton, J., Emmanuel-Yusuf, D., Hall, S., Johnson, V., Longhurst, N., O’Grady, A., Sherry-
Brennan, F., 2015. Distributing Power: A Transition to a Civic Energy Future.
Bauwens, T., 2017. Polycentric governance approaches for a low-carbon transition: the
roles of community-based energy initiatives in enhancing the resilience of future
energy systems. In: Complex Systems and Social Practices in Energy Transitions.
Springer, Cham, pp. 119–145.
Bauwens, T., Gotchev, B., Holstenkamp, L., 2016. What drives the development of
community energy in Europe? The case of wind power cooperatives. Energy Res.
Soc. Sci. 13, 136–147.
Bayliss, K., Mattioli, G., Steinberger, J., 2021. Inequality, poverty and the privatization of
essential services: a ‘systems of provision’ study of water, energy and local buses in
the UK. Compet. Change 25 (3–4), 478–500.
EnergyPolicy172(2023)1132777 M. Lacey-Barnacle et al.
Becker, S., Kunze, C., Vancea, M., 2017. Community energy and social entrepreneurship:
addressing purpose, organisation and embeddedness of renewable energy projects.
J. Clean. Prod. 147, 25–36.
Brisbois, M.C., 2019. Powershifts: a framework for assessing the growing impact of
decentralized ownership of energy transitions on political decision-making. Energy
Res. Social Sci. 50, 151–161.
Harvey, D., 2007. A Brief History of Neoliberalism. Oxford University Press, USA.
Hewitt, R.J., Bradley, N., Baggio Compagnucci, A., Barlagne, C., Ceglarz, A.,
Cremades, R., et al., 2019. Social innovation in community energy in Europe: a
review of the evidence. Front. Energy Res. 7, 31.
Heslop, J., Morgan, K., Tomaney, J., 2019. Debating the foundational economy.
Renewal: J. Social Democracy 27.
Brown, M., Jones, R., 2021. Paint Your Town Red: How Preston Took Back Control and
Hess, D., 2007. Alternative Pathways in Science and Industry. The MIT Press, Cambridge,
Your Town Can Too. Repeater Books.
MA.
Brown, G., Leibowitz, J., McInroy, N., 2019. Community Wealth Building for Economic
Hoicka, C.E., Lowitzsch, J., Brisbois, M.C., Kumar, A., Camargo, L.R., 2021.
and Environmental Justice. Centre for Local Economic Strategies.
Burger, C., Froggatt, A., Mitchell, C., Weinmann, J., 2020. Decentralised Energy.
Ubiquity press.
Catney, P., MacGregor, S., Dobson, A., Hall, S.M., Royston, S., Robinson, Z., et al., 2014.
Big society, little justice? Community renewable energy and the politics of localism.
Local Environ. 19 (7), 715–730.
Chilvers, J., Foxon, T.J., Galloway, S., Hammond, G.P., Infield, D., Leach, M., et al.,
2017. Realising transition pathways for a more electric, low-carbon energy system in
the United Kingdom: challenges, insights and opportunities. Proc. IME J. Power
Energy 231 (6), 440–477.
Collord, 2019. Building community wealth globally: the Kigoma-Preston collaboration.
https://thenextsystem.org/learn/stories/building-community-wealth-globally
-kigoma-preston-collaboration.
Creamer, et al., 2018. Community Energy: Entanglements of Community, State & Private
Implementing a just renewable energy transition: policy advice for transposing the
new European rules for renewable energy communities. Energy Pol. 156, 112435.
Imbroscio, D.L., Williamson, T., Alperovitz, G., 2003. Local policy responses to
globalization: place-based ownership models of economic enterprise. Pol. Stud. J. 31
(1), 31–52.
Jackson, M., McInroy, N., 2017. Community Wealth Building through Anchor
Institutions. CLES. https://cles.org.uk/our-work/publications/community-wealth
-building-throughanchor-institutions.
Jamison, A., 2001. The Making of Green Knowledge. Environmental Politics and Cultural
Transformation. Cambridge University Press, Cambridge.
Johnson, V.C., Hall, S., Barton, J., Emanuel-Yusuf, D., Longhurst, N., O’Grady,
´
A.,
Robinson, E., 2014. Community energy and equity: The distributional implications
of a transition to a decentralised electricity system. People Place Pol. 8 (3), 149–167.
Kaye, S., 2020. Think Big, Act Small: Elinor Ostrom’s Radical Vision for Community
Sector Geography Compass.
Power (New Local).
CLES, 2021. The Brave. Centre for Local Economic Strategies, UK. https://cles.org.
Kelly, M., 2012. Owning Our Future: the Emerging Ownership Revolution. Berrett-
uk/blog/the-brave/.
Koehler Publishers.
Cumbers, A., 2016. Remunicipalization, the low-carbon transition, and energy
democracy. In: State of the World. Island Press, Washington, DC, pp. 275–289.
Davis, M., 2021. Community municipal investments: accelerating the potential of local
net zero strategies. https://eprints.whiterose.ac.uk/175824/1/2021_07_01_CMB-
Innovate-UK-Report-Web-2.pdf.
Davoudi, S., Madanipour, A. (Eds.), 2015. Reconsidering Localism. Routledge.
Democracy Collaborative & Momentum, 2022. Community wealth building: a toolkit for
local councillors. Prepared for People’s Momentum by The Democracy
Collaborative. https://peoplesmomentum.com/wp-content/uploads/2022/04/C
WB_MTM_11_April.pdf.
Kelly, M., McKinley, S., Duncan, V., 2016. Politics of place/politics for places:
community wealth building: America’s emerging asset-based approach to city
economic development. Renewal: J. Labour Politics 24 (2), 51.
Kelly, M., Howard, T., 2019. The Making of a Democratic Economy: Building Prosperity
for the Many, Not Just the Few. Berrett-Koehler Publishers.
Kohn, M., 2020. Community wealth building: lessons from Italy. In: Community Wealth
Building and the Reconstruction of American Democracy. Edward Elgar Publishing.
Lacey-Barnacle, M., 2019. Exploring Local Energy Justice in Times of Austerity: Civic Energy
Sector Low-Carbon Transitions in Bristol City (Doctoral Dissertation. Cardiff University.
Lacey-Barnacle, M., 2020. Proximities of energy justice: contesting community energy
Dubb, S., 2016. Community wealth building forms: what they are and how to use them at
and austerity in England. Energy Res. Social Sci. 69, 101713.
the local level. Acad. Manag. Perspect. 30 (2), 141–152.
Lenihan, R., 2014. Rust-belt recovery: the Cleveland model as economic development in
Eder, J., 2021. Decreasing Dependency through Self-Reliance: Strengthening Local Economies
an age of economic stagnation and climate change. Pepperdine Pol. Rev. 7 (1), 6.
through Community Wealth Building (No. 124). ICAE Working Paper Series.
Fensham, P., 2020. Community wealth building in Australia. https://cles.org.uk/publi
cations/community-wealth-building-in-australia/.
Manley, J., Whyman, P.B. (Eds.), 2021. The Preston Model and Community Wealth
Building: Creating a Socio-Economic Democracy for the Future. Routledge.
Morena, E., Krause, D., Stevis, D., 2020. Just Transitions. Social Justice in a Low-Carbon
Feola, G., Nunes, R., 2014. Success and failure of grassroots innovations for addressing
World. Pluto Press, London.
climate change: The case of the Transition Movement. Glob. Environ. Change 24,
232–250.
Foxon, T.J., 2013. Transition pathways for a UK low carbon electricity future. Energy
Pol. 52, 10–24.
Foxon, T.J., Hammond, G.P., Pearson, P.J., 2010. Developing transition pathways for a
low carbon electricity system in the UK. Technol. Forecast. Soc. Change 77 (8),
1203–1213.
Fressoli, M., et al., 2014. When grassroots innovation movements encounter mainstream
institutions: implications for models of inclusive innovation. Innovat. Dev. 4 (2),
277–292. https://doi.org/10.1080/2157930X.2014.921354.
Green Finance Institute, 2021. Local climate bonds. https://www.greenfinanceinstitute.
co.uk/programmes/ceeb/lcbs/.
Guinan, J., O’Neill, M., 2020. The Case for Community Wealth Building. Polity Press,
Cambridge.
McCauley, D., Heffron, R., 2018. Just transition: integrating climate, energy and
environmental justice. Energy Pol. 119, 1–7.
McInroy, N., 2020. Community is economic energy we must use it. https://www.lgcplus.
com/politics/coronavirus/neil-mcinroy-community-is-economic-energy-we-must-us
e-it-29-06-2020/.
Paul, F.C., Cumbers, A., 2021. The return of the local state? Failing neoliberalism,
remunicipalisation, and the role of the state in advanced capitalism. Environ. Plann.:
Econ. Space, 0308518X211050407.
Pansera, M., Owen, R., 2017. ‘Framing Inclusive Innovation within the Discourse of
Development: Insights from Case Studies in India’, Research Policy, vol. 47. Elsevier,
pp. 23–34. https://doi.org/10.1016/j.respol.2017.09.007 (1).
Peters, J., 2012. Neoliberal convergence in North America and Western Europe: fiscal
austerity, privatization, and public sector reform. Rev. Int. Polit. Econ. 19 (2),
208–235.
Guinan, J., O’Neill, M., 2019. From community wealth-building to system change. IPPR
Power, M., Wilkinson, R., Pickett, K., 2016. Inequality, Economic Democracy and
Progressive Rev. 25 (4), 382–392.
Gupta, A.K., 2016. Grassroots Innovation. Penguin, Gurgaon.
Hadfield, M., 2019. What has the Preston Model achieved so far. Coop News. http
Sustainability, vol. 37. World social science report, p. 160.
Radcliffe, E., Williams, L., 2021. Toolkit: a Community Wealth Building Energy
Transition. Centre for Local Economic Strategies.
s://www.thenews.coop/143245/topic/economy/what-has-the-preston-model-achie
ved-so-far/.
Rodríguez-Pose, A., Gill, N., 2003. The global trend towards devolution and its
implications. Environ. Plann. C Govern. Pol. 21 (3), 333–351.
Hall, P.A., Soskice, D., 2001. Varieties of Capitalism: the Institutional Foundations of
Routledge, P., Cumbers, A., Derickson, K.D., 2018. States of just transition: realising
Comparative Advantage. Oxford University Press.
Hall, S., Foxon, T.J., Bolton, R., 2016. Financing the civic energy sector: how financial
institutions affect ownership models in Germany and the United Kingdom. Energy
Res. Social Sci. 12, 5–15.
Hall, S., Roelich, K., 2016. Business model innovation in electricity supply markets: the
role of complex value in the United Kingdom. Energy Pol. 92, 286–298.
Hanna, T.M., 2019. Community Wealth Building and Resilient Local Economies: the Role
of Anchor Institutions. Public Finance for the Future We Want. Transnational
Institute.
Hanna, T.M., Kelly, M., 2021. Community Wealth Building: the Path towards a
Democratic and Reparative Political Economic System.
climate justice through and against the state. Geoforum 88, 78–86.
Seyfang, G., Park, J.J., Smith, A., 2013. A thousand flowers blooming? An examination of
community energy in the UK. Energy Pol. 61, 977–989.
Seyfang, G., Smith, A., 2007. Grassroots innovations for sustainable development:
towards a new research and policy agenda. Environ. Polit. 16 (4), 584–603.
Available at: http://www.informaworld.com/10.1080/09644010701419121.
Sheffield, H., 2017. The Preston model: UK takes lessons in recovery from rust-belt
Cleveland. Guardian 11.
Smith, A., 2014. Technology networks for socially useful production. J. Peer Product. (5)
Available at: http://peerproduction.net/issues/issue-5-shared-machine-shops/
peer-reviewed-articles/technology-networks-for-socially-useful-production/.
Hanna, T.M., Bozuwa, J., Rao, R., 2022. The Power of Community Utilities: Publicly
Smith, A., 2007. ‘Translating Sustainabilities between Green Niches and Socio-Technical
Owned and Cooperative Electric Utilities as Anchors for Community Wealth Building
and a Just Energy Transition. The Democracy Collaborative and the Climate and
Community Project, Washington, D.C. and Philadelphia.
Hansen, T., 2021. The foundational economy and regional development. Reg. Stud. 1–10.
Hanke, F., Guyet, R., Feenstra, M., 2021. Do renewable energy communities deliver
energy justice? Exploring insights from 71 European cases. Energy Res. Social Sci.
80, 102244.
Hargreaves, T., Hielscher, S., Seyfang, G., Smith, A., 2013. Grassroots innovations in
community energy: The role of intermediaries in niche development. Glob. Environ.
Change 23 (5), 868–880.
regimes’, Technology Analysis & Strategic Management, vol. 19. Routledge,
pp. 427–450 (4).
Smith, A., et al., 2017. Grassroots Innovation Movements. Routledge, London.
Smith, A., Seyfang, G., 2013. Constructing grassroots innovations for sustainability.
Global Environ. Change 23 (5), 827–829.
Smith, A., Hargreaves, T., Hielscher, S., Martiskainen, M., Seyfang, G., 2016. Making the
most of community energies: three perspectives on grassroots innovation. Environ.
Plann. 48 (2), 407–432.
Smith, A., Stirling, A., 2018. Innovation, sustainability and democracy: an analysis of
grassroots contributions. J. Self Govern. Manag. Econ. 6 (1), 64–97.
EnergyPolicy172(2023)1132778 M. Lacey-Barnacle et al.
Smith, A., Raven, R., 2012. What is protective space? Reconsidering niches in transitions
to sustainability. Res. Pol. 41 (6) https://doi.org/10.1016/j.respol.2011.12.012.
Stewart, F., 2021. All for sun, sun for all: can community energy help to overcome
socioeconomic inequalities in low-carbon technology subsidies? Energy Pol. 157,
112512.
Thompson, M., 2021. What’s so new about New Municipalism? Prog. Hum. Geogr. 45
(2), 317–342.
Wang, X., Lo, K., 2021. ‘Just Transition: A Conceptual Review’, Energy Research & Social
Science, vol. 82. Elsevier Ltd, 102291. https://doi.org/10.1016/j.erss.2021.102291.
March.
Walker, G., Hunter, S., Devine-Wright, P., Evans, B., Fay, H., 2007. Harnessing
community energies: explaining and evaluating community-based localism in
renewable energy policy in the UK. Global Environ. Polit. 7 (2), 64–82.
Warren, D.B., 2010. Growing a Green Economy for All: from Green Jobs to Green
Uyarra, E., Shapira, P., Harding, A., 2016. Low carbon innovation and enterprise growth
Ownership. Democracy Collaborative at the University of Maryland.
in the UK: challenges of a place-blind policy mix. Technol. Forecast. Soc. Change
103, 264–272.
Van Veelen, B., 2018. Negotiating energy democracy in practice: governance processes in
community energy projects. Environ. Polit. 27 (4), 644–665.
Webster, A., Kuznetsova, O., Ross, C., Berranger, C., Booth, M., Eseonu, T., Golan, Y.,
2021. Local regeneration and community wealth building–place making: co-
operatives as agents of change. J. Place Manag. Dev.
Williams, A., Goodwin, M., Cloke, P., 2014. Neoliberalism, big society, and progressive
localism. Environ. Plann. 46 (12), 2798–2815.
EnergyPolicy172(2023)1132779
| null |
10.1371_journal.pgen.1008460.pdf
|
Data Availability Statement: The RNA sequencing
data of human cell lines and zebrafish tissues are
available from NCBI Sequence Read Archive (SRA)
(accession numbers PRJNA542249 and
PRJNA543385). All other relevant data are
available within the manuscript and its Supporting
Information files.
|
The RNA sequencing data of human cell lines and zebrafish tissues are available from NCBI Sequence Read Archive (SRA) (accession numbers PRJNA542249 and PRJNA543385). All other relevant data are available within the manuscript and its Supporting Information files.
|
RESEARCH ARTICLE
A missense mutation in SNRPE linked to non-
syndromal microcephaly interferes with U
snRNP assembly and pre-mRNA splicing
Tao Chen1☯, Bin ZhangID
Sebastian Fro¨ hler1, Clemens GrimmID
Min Zhang2, Nadine KraemerID
2,3☯, Thomas Ziegenhals4☯, Archana B. Prusty4☯,
4, Yuhui Hu2, Bernhard Schaefke2,5, Liang FangID
6,7,8*, Utz Fischer4*, Wei Chen2,5*
6,7, Angela M. KaindlID
2,5,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 Laboratory for Functional Genomics and Systems Biology, Berlin Institute for Medical System Biology,
Max-Delbru¨ ck-Center for Molecular Medicine, Berlin, Germany, 2 Department of Biology, Southern
University of Science and Technology (SUSTech), Shenzhen, China, 3 Cancer Science Institute of
Singapore, National University of Singapore, Singapore, 4 Department of Biochemistry, Theodor-Boveri-
Institute, University of Wu¨ rzburg, Wu¨ rzburg, Germany, 5 Academy for Advanced Interdisciplinary Studies,
Southern University of Science and Technology (SUSTech), Shenzhen, China, 6 Charite´ -Universita¨ tsmedizin
Berlin, Institute of Cell Biology and Neurobiology, Berlin, Germany, 7 Charite´ -Universita¨ tsmedizin Berlin,
Department of Pediatric Neurology, Berlin, Germany, 8 Charite´ -Universita¨ tsmedizin Berlin, Center for
Chronically Sick Children, Berlin, Germany
OPEN ACCESS
[email protected] (WC)
☯ These authors contributed equally to this work.
* [email protected] (UF); [email protected] (AK);
Citation: Chen T, Zhang B, Ziegenhals T, Prusty
AB, Fro¨hler S, Grimm C, et al. (2019) A missense
mutation in SNRPE linked to non-syndromal
microcephaly interferes with U snRNP assembly
and pre-mRNA splicing. PLoS Genet 15(10):
e1008460. https://doi.org/10.1371/journal.
pgen.1008460
Editor: A. Gregory Matera, University of North
Carolina at Chapel Hill, UNITED STATES
Received: March 5, 2019
Accepted: October 4, 2019
Published: October 31, 2019
Copyright: © 2019 Chen et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The RNA sequencing
data of human cell lines and zebrafish tissues are
available from NCBI Sequence Read Archive (SRA)
(accession numbers PRJNA542249 and
PRJNA543385). All other relevant data are
available within the manuscript and its Supporting
Information files.
Funding: Tao Chen was funded by China
Scholarship Council (CSC). This work was
supported by Sino-German (NSFC-DFG)
Abstract
Malfunction of pre-mRNA processing factors are linked to several human diseases including
cancer and neurodegeneration. Here we report the identification of a de novo heterozygous
missense mutation in the SNRPE gene (c.65T>C (p.Phe22Ser)) in a patient with non-syn-
dromal primary (congenital) microcephaly and intellectual disability. SNRPE encodes SmE,
a basal component of pre-mRNA processing U snRNPs. We show that the microcephaly-
linked SmE variant is unable to interact with the SMN complex and as a consequence fails
to assemble into U snRNPs. This results in widespread mRNA splicing alterations in fibro-
blast cells derived from this patient. Similar alterations were observed in HEK293 cells upon
SmE depletion that could be rescued by the expression of wild type but not mutant SmE.
Importantly, the depletion of SmE in zebrafish causes aberrant mRNA splicing alterations
and reduced brain size, reminiscent of the patient microcephaly phenotype. We identify the
EMX2 mRNA, which encodes a protein required for proper brain development, as a major
mis-spliced down stream target. Together, our study links defects in the SNRPE gene to
microcephaly and suggests that alterations of cellular splicing of specific mRNAs such as
EMX2 results in the neurological phenotype of the disease.
Author summary
In higher eukaryotes, the protein coding genes are first transcribed as precursor mRNAs
(pre-mRNAs) and further processed by the spliceosome to form the mature mRNA for
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
1 / 23
Cooperative Research Project (No. 31861133013),
National Natural Science Foundation of China (No.
31771443), the Basic Research Grant from Science
and Technology Innovation Commission of
Shenzhen Municipal Government (No.
KQTD20180411143432337 and No.
JCYJ20170307105752508). The receiver of these
three funds is Wei Chen. NSFC: http://www.nsfc.
gov.cn/english/site_1/index.html CSC: https://
www.chinesescholarshipcouncil.com Science and
Technology Innovation Commission of Shenzhen
Municipal Government: http://english.sz.gov.cn/
govt/agencies/s/201811/t20181122_14604925.
htm The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
A missense mutation in SNRPE linked to microcephaly
translation. Malfunction of pre-mRNA processing factors are linked to several human dis-
eases including cancer and neurodegeneration. Here we report the identification of a de
novo heterozygous missense mutation in the SNRPE/SmE gene in a patient with non-syn-
dromal primary (congenital) microcephaly and intellectual disability. The effect of identi-
fied de novo mutation on SNRPE/SmE was characterized in vitro. The zebrafish was used
as in vivo model to further dissect the physiological consequence and pathomechanism.
Finally, the EMX2 gene was identified as one of the major down stream target genes
responsible for the phenotype. Our study links defects in the SNRPE/SmE gene to micro-
cephaly and provides the new pathogenic mechanism for microcephaly.
Introduction
In higher eukaryotes, the vast majority of protein-coding genes are transcribed as precursors
(pre-mRNA) containing non-coding intronic and coding exonic sequences. These pre-
mRNAs need to be extensively processed by splicing to generate the mature mRNA with an
open reading frame. Splicing is mediated by macromolecular machines termed spliceosomes,
which consist of five different small nuclear ribonucleoprotein (snRNP) subunits and a large
number of additional protein cofactors [1–4]. The major spliceosome, containing U1, U2, U4,
U5 and U6 snRNPs, is responsible for splicing of almost 99% of human pre-mRNAs whereas
the minor spliceosome is required to excise a special class of very rare (ATAC) introns from
certain mRNAs [5]. To generate mRNA variants with different coding potential, the splice
sites (SSs) within pre-mRNAs are differentially utilized through alternative splicing (AS). This
process occurs in >95% of human multi-exon genes, thus leading to a large increase of protein
diversity [6–9]. The decision of AS is regulated through the cooperative interplay between cis-
elements, including constitutive splicing elements (such as 5’ SSs, branch point (BP), polypyri-
midine tract (PPT) and 3’ SSs) and optional cis-regulatory elements (exonic and intronic
splicing enhancer/silencer called ESE, ESS, ISE, ISS), and trans-acting factors, such as core
splicing machinery and splicing regulators (SR proteins and heterogenous ribonucleoproteins
(hnRNPs)) [9–11]. It has been shown that AS plays critical roles in the specification of cell
fates [12], tissue types [6,9], developmental process [13], sex determination [14] and stimula-
tion response [15].
Due to the important role in regulation of gene expression and protein diversity, mRNA
splicing is particularly sensitive to mutations and its dysregulation could lead to human dis-
eases [16,17]. The most common type of mutations leading to aberrant splicing, are cis-acting
mutations located in either constitutive splicing elements (5’ SS, 3’ SS and BP) or cis-regulatory
elements (ESE, ESS, ISE and ISS) modulating spliceosome assembly [16]. For instance, ESE,
ESS and 5’ SS mutations in the exon 10 of the MAPT gene, encoding the microtubule-associ-
ated protein Tau, have been identified as the cause of frontotemporal dementia with parkin-
sonism linked to chromosome 17 (FTDP-17) [18].
In addition to mutations affecting cis-elements, mutations in trans-acting splicing factors
are also implicated in a set of human diseases. Since defects in these factors typically affect the
splicing machinery as a whole, they affect the processing of many transcripts and hence often
cause more complex etiologies than mutations in cis elements. An interesting example of this
class are mutations in several protein components of U4/U6.U5 tri-small nuclear ribonucleo-
protein (tri-snRNP) such as pre-mRNA processing factor 3 (PRPF3) [19], PRPF4 [20], PRPF6
[21], PRPF8 [22], PRPF31 [23,24] and SNRNP200 (also called BRR2) [25], that cause the auto-
somal dominant eye disease retinitis pigmentosa (adRP) [26]. In addition, mutations
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
2 / 23
A missense mutation in SNRPE linked to microcephaly
preventing the production of functional SMN protein cause spinal muscular atrophy (SMA)
[27]. This protein is part of the SMN complex, which mediates the assembly of spliceosomal
U snRNPs and hence determines the abundance of active spliceosomes. Although the SMN
protein is ubiquitously expressed, the effect of SMN deficiency on the repertoire of snRNAs
and aberrant splicing shows tissue specific dependence in a SMA mouse model [28]. In addi-
tion, mutations within SmB/B’ and SmE have been reported to be linked to cerebro-costo-
mandibular syndrome (CCMS) [29,30] and hypotrichosis simplex (HS) [31], respectively.
Although these mutations are identified as the genetic cause of these diseases, the disease etiol-
ogies are still unknown. Importantly, mutations in RNU4ATAC have been shown to affect the
formation of minor spliceosome and cause Taybi-Linder syndrome/microcephalic osteodys-
plastic primordial dwarfism type 1 (TALS/MOPD1) [32,33], illustrating that not only malfunc-
tioning of proteins but also of U snRNAs can cause disease.
Using whole exome sequencing, we report here a de novo heterozygous missense mutation
within the SNRPE/SmE gene from a patient with non-syndromal primary (congenital) micro-
cephaly and intellectual disability. This mutation generates a protein product that fails to inter-
act with the SMN-complex and thus cannot become properly assembled into spliceosomal
U snRNPs. Our results further reveal that the mutation in SmE causes aberrant mRNA splicing
in both human cell lines (fibroblast and HEK293) and zebrafish. Furthermore, specific deple-
tion of endogenous SmE protein in zebrafish causes similar brain defect as in the patient. Of
note, we find that one of the affected transcripts in the zebrafish model encodes for the protein
EMX2, which is required for proper early brain development. Our study suggest that the iden-
tified missense mutation in SNRPE disturbs appropriate spatiotemporal gene expression in the
brain through aberrant mRNA splicing, which is likely to cause the microcephaly phenotype.
Results
Identification of a missense mutation within SNRPE/SmE in a
microcephaly patient
To identify the molecular genetic basis of a patient afflicted with non-syndromal microcephaly
in a two-generation pedigree, whole exome sequencing (WES) was performed for the patient
and its unaffected parents (Fig 1A). On average, 180 million reads were obtained for each indi-
vidual and more than 90 fold coverage of exome were achieved for each individual. A de novo
heterozygous missense mutation (c.65T>C (p.Phe22Ser)) was identified in the second exon of
the SNRPE/SmE gene from the patient (Fig 1B). This gene and in particular the mutated resi-
due is highly conserved among different species including zebrafish and the more distant yeast
S. pombe (Fig 1C). It encodes the SNRPE/SmE protein [34], which constitutes a basal compo-
nent of spliceosome. This factor, together with six additional Sm proteins termed SmB/B’,
SmD1, SmD2, SmD3, SmF and SmG, form the common Sm core of spliceosomal U snRNPs.
This raised the possibility that the pathological mutation in SmE affects U snRNP biogenesis
and/or splicing.
Impaired binding of SmE mutant to the SMN complex causes defects in Sm
core assembly
We first investigated whether the identified missense mutation in SmE affects its incorporation
into U snRNPs. Incorporation of newly translated SmE starts with the formation of the hetero-
trimeric complex composed of SmE, SmF and SmG [35]. Subsequently, this heterooligomer
is transferred onto the PRMT5 complex, which assembles together with SmD1/D2 and the
assembly chaperone pICln a closed ring termed the 6S complex [36,37]. The next step of U
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
3 / 23
A missense mutation in SNRPE linked to microcephaly
Fig 1. Identification of potential causative mutation by whole exome sequencing. (A), Family pedigree. Filled symbol indicates individual
suffering from non-syndromal primary microcephaly and intellectual disability. (B), Traditional Sanger sequencing validated the identified
SNRPE/SmE mutation (c.65T>C (p.Phe22Ser)). The red box labels the de novo heterozygous mutation. (C), Alignment of SNRPE/SmE protein
sequences across different species. The red rectangle indicates the mutated residue.
https://doi.org/10.1371/journal.pgen.1008460.g001
snRNP biogenesis is dependent on the SMN complex, consisting of SMN, Gemins 2–8 and
UNRIP [38]. This unit catalyzes the release of pICln from the 6S complex and the transfer of
Sm proteins onto the U snRNA [36,37]. After hypermethylation of the m7G cap to m2,2,7
3G
(m3G/TMG) cap, the assembled U snRNPs are imported into the nucleus and after further
maturation in Cajal bodies (CBs), targeted to splicing speckles [39,40].
To follow the path of SmE into U snRNPs, FLAG-tagged wild type or mutant proteins were
overexpressed in HEK293 cells. The tagged proteins were then immunoprecipitated using
anti-FLAG antibodies and co-precipitated factors indicative for defined U snRNP biogenesis
intermediates were detected by western blotting (Fig 2A and 2C). Interestingly, no significant
change in the interaction of mutant SmE with either SmF, SmD1 or pICln was observed when
compared to the wild type protein. This suggests that the pathogenic missense mutation did
not interfere with the early phase of U snRNP biogenesis, including formation of SmE/F/G
heterooligomer and the 6S complex at the PRMT5 complex. However, only the wild type but
not the mutant SmE protein interacted efficiently with SmD3 as well as the SMN complex (Fig
2A and 2C), suggesting that the SmE mutant was defective in the transfer from the PRMT5
complex onto the SMN complex, which is in turn a pre-requisite for the subsequent loading
onto U snRNA. In agreement with this notion, 3’-end labeling of the RNA co-precipitated
with the SmE-FLAG immunoprecipitations revealed that only wild type SmE was able to effi-
ciently interact with U snRNAs (Fig 2B and 2C). Together these data show that the mutant
SmE is unable to be incorporated into U snRNPs (Fig 2B, quantification in Fig 2C).
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
4 / 23
A missense mutation in SNRPE linked to microcephaly
Fig 2. The missense mutation impairs the biogenesis of spliceosomal U snRNPs during Sm core assembly. (A-E), The NSM mutation in
SmE impairs its interaction with U snRNP assembly machinery and incorporation into U snRNPs. (A), Anti-FLAG immunoprecipitation after
transient transfection in HEK293T cells and western blotting analysis for co-precipitated U snRNP intermediates. Mock immunoprecipitations
were performed with untransfected lysates. (B), 3’-end labeling of co-precipitated RNA and autoradiography. RNA immunoprecipitated using
the H20 antibody against m3G/m7G cap of U snRNAs was used as reference. (C), Quantification of the data shown in A and B from two
independent biological replicates, with black bars representing wild type and gray bars representing mutant SmE. (D), Predicted structural
model for interference of the SmE mutation in its interaction with Gemin2, based on the PDB structure 4V98. (E), Immunoprecipitation using
antibodies specific to Sm proteins, SMN, pICn and U snRNA cap, with lysates from HEK293T cells transfected with dual expression plasmid
encoding 2A-tagged mutant SmE and HA-tagged wild type SmE and western blotting to analyze the integration of the wild type and mutant
SmE into U snRNP biogenesis pathway.
https://doi.org/10.1371/journal.pgen.1008460.g002
Since the interaction of mutant SmE with the SMN complex is affected, we used the previ-
ously published structure of the 8S U snRNP assembly intermediate (Gemin2-SMNΔC bound
to 6S, PDB entry 4V98) [37], to in silico model the effect of the mutation. As evident from
structural data of Gemin2 in association with Sm proteins, Phe22 of SmE is part of a binding
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
5 / 23
A missense mutation in SNRPE linked to microcephaly
module that interacts with Pro49 and Tyr52 of Gemin2 (Fig 2D). The identified SmE mutation
(c.65T>C) changes the polarity of the amino acid residue from hydrophobic (Phe) to hydro-
philic (Ser), which is incompatible with the detected mode of interaction.
To recapitulate the disease condition where both wild type and mutant SmE are expressed
within the cell, we co-expressed HA-tagged wild type SmE and 2A-tagged mutant SmE in
HEK293 cells from a dual expression plasmid and tested how they are processed by the U
snRNP assembly pipeline (Fig 2E). The dual expression construct was designed with a post-
translational self-cleaving 2A tag between the mutant and wild type SmE (Fig 2E), giving raise
to equal amounts of exogenous 2A-tagged mutant and HA-tagged wild type SmE in each
transfected cell. We then performed immunoprecipitations using antibodies specific to endog-
enous U snRNPs (Y12 which predominantly immunoprecipitates U snRNPs and not Sm inter-
mediates), pICln and SMN. As expected, while the wild type SmE was able to efficiently
interact with the U snRNP assembly machinery and hence was incorporated into U snRNPs,
the mutant was not enriched in any of the immunoprecipitations (note that due to the pres-
ence of the highly abundant endogenous Sm protein pool, the efficiency of immunoprecipita-
tion of the tagged proteins was low as compared to those shown in Fig 2A–2C).
We also performed immunostaining of HeLa cells transiently transfected with either the
FLAG-tagged wild type or mutant SmE and studied the co-localization of the exogenously
expressed SmE to the CBs (the subnuclear structures for U snRNPs maturation) and to U
snRNPs [41]. As expected, the wild type SmE co-localized to CBs as confirmed by a strong co-
localization with the CB marker protein coilin (Fig 3A, top panel) and were also efficiently tar-
geted to nuclear speckles as can be seen with co-localization with SmD3 (Fig 3B, top panel).
However, in keeping with our immunoprecipitation results, the SmE mutant was localized to
the cytoplasm, at times even forming very small foci, or non-specifically dispersed in the
nucleus (Fig 3A and 3B, middle panel), showing that the mutant fails to be incorporated into
U snRNPs. We conclude that the non-specific nuclear distribution of SmE results from excess
of overexpressed exogenous SmE that likely diffuses into the nucleus in the absence of cognate
interactors. Together, these results demonstrate that the mutation (c.65T>C(p.Phe22Ser)) in
SmE impairs its incorporation into U snRNPs due to its inability to interact with the SMN
complex. The early assembly phase, however, appears to be unaffected by this mutation.
The SNRPE/SmE deficiency results in reduced levels of U snRNPs in
patient
Taking into account our biochemical data, we hypothesized that the U snRNP levels in the
patient are likely reduced. To this end, we first performed immunostaining and confocal
microscopy analysis of control primary fibroblasts and patient fibroblast (S1A and S1B Fig).
We found a clear difference in the distribution of U snRNAs (m3G/m7G cap) in patient cells.
While in control fibroblasts U snRNAs were found predominantly within the nuclei (S1A Fig,
top panel), there was a marked increase in U snRNAs in the cytoplasm of the patient fibro-
blasts (S1A Fig, bottom panel). Additionally, levels of Sm proteins in the nuclei of patient
fibroblasts was down-regulated (S1A and S1B Fig). CBs are however absent in control as well
as patient fibroblasts (S1A and S1B Fig) since CBs are known to be absent in primary cells
[42]. Since free U snRNAs that are not assembled into U snRNPs are prone to degradation
[43], we proposed that the decrease in U snRNP assembly might result in a reduction in the
total U snRNA pool within the patient fibroblasts. We analyzed the U snRNA transcript levels
in patient and control fibroblasts using qRT-PCR and the SmE expression level in fibroblasts
by RT-qPCR and Western blotting (S1C and S1D Fig). Interestingly, among the U snRNAs
tested, we found a clear reduction in the U1 snRNA abundance and a modest decrease in U2
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
6 / 23
A missense mutation in SNRPE linked to microcephaly
Fig 3. NSM mutation causes mis-localization of the SNRPE/SmE protein. (A-B), Indirect immunofluorescence and
confocal microscopy of HeLa cells transfected with either FLAG-tagged wild type or mutant SmE (WT/Mut) or left
untransfected (negative control). Empty white arrowheads indicate localization pattern observed and filled white
arrowheads indicate zoomed in region shown in the overly inset. (A), Top panel shows clear co-localization of wild
type SmE (green) and coilin (magenta) in CBs and middle panel shows most of the mutant SmE (green) distributed in
the cytoplasm with a minor fraction in the nucleus and co-localizing with coilin (magenta) in CBs. (B), While wild type
SmE (green, top panel) co-localizes with SmD3 (magenta) in CBs and splicing speckles, the mutant SmE (green,
middle panel) is predominantly cytoplasmic with marginal co-localization with SmD3 in CBs or in nuclear speckles.
https://doi.org/10.1371/journal.pgen.1008460.g003
and U4 snRNAs in patient fibroblasts (S1C Fig). We then performed anti-Sm immunoprecipi-
tation from control and patient cells and analyzed the co-precipitated RNA by 3’-end labeling
(S1E and S1F Fig). We found a distinct difference in the amount of co-precipitated U snRNAs,
with the U1 snRNA levels being the most affected. We conclude that the effects are enhanced
specifically in the case of U1 snRNP since the U1–70K protein is known to interact with SMN
complex to increase U1 snRNP assembly in cells [44] and thus the strongest effect would be
observed for the most abundantly assembled U snRNP.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
7 / 23
A missense mutation in SNRPE linked to microcephaly
The SNRPE/SmE deficiency causes widespread splicing alterations
The results above suggest that the identified mutation (c.65T>C (p.Phe22Ser)) in SmE leads to
reduced levels of Sm-class snRNPs. As these are the major trans-acting factors in pre-mRNA
processing, we next asked whether the mutant SmE impacts on the splicing profile of cells. To
address this, the RNA was extracted from fibroblast cells derived from the patient and three
healthy individuals, and subjected to RNA sequencing. Indeed we observed tremendous
altered splicing events between the patient cell and controls, with intron retention (RI) being
the most frequently impacted splicing event. As shown in Fig 4A, more than 2084 introns
showed significantly increased intron retention (p < 0.001, fdr < 0.05, ΔPercentage of Intron
Retention (ΔPIR: mutant—control) > 0.1) in the patient cells while only less than 112 introns
showed significant decreased retention (p < 0.001, fdr < 0.05, ΔPIR < -0.1). Intron retention
often introduces premature termination codon (PTC) into the affected mRNAs, which triggers
nonsense mediated decay (NMD) and potentially also other mRNA decay pathways. We there-
fore examined the changes in the expression levels of transcripts displaying increased intron
retention. Consistent with our assumption, these transcripts show significantly decreased
expression between the patient and control comparing to those genes without any introns with
increased retention (Mann-Whitney test, p = 8.8e-44) (Fig 4B).
To check whether the splicing defects observed in the patient fibroblast cells could be res-
cued by the presence of exogenous wild type SmE protein, we exogenously overexpressed wild
type SmE in the patient fibroblast cells and performed the RNA-seq. In total, more than 350
million reads were obtained for triplicate experiments and around 93% of them could be
uniquely mapped to human reference genome. Given that the splicing defect observed in the
patient fibroblast cells was predominantly manifested as increased intron retention, we
focused our analysis here on intron retention. By applying the same approach as described
above, we firstly compared between the patient fibroblast cells with exogenous wild type SmE
and those without. As shown in S2 Fig, after overexpression of wild type SmE (S2A Fig), 2201
introns were less retained while only 414 introns were more retained (S2B Fig). Moreover,
when we compared the patient fibroblast cells with exogenous wild type SmE to fibroblast cells
from healthy control, as shown in S2C Fig, much less splicing changes were detected and the
direction of changes was more symmetric, in contrast to the comparison between the patient
and healthy control fibroblast cells (Fig 4A). To further evaluate the rescue efficiency, we plot-
ted the splicing changes in two comparisons, i.e. healthy control vs patient (X axis), and patient
with overexpressing SmE vs patient without SmE overexpression (Y axis) (S2D Fig). Among
2084 significant aberrant splicing events that were detected in the patient (Fig 4A), 1130 of
them were significantly rescued (S2D Fig). These results, together demonstrated that overex-
pression of wild type SmE in patient fibroblasts could indeed reduce the predominant splicing
defect, i.e. intron retention, observed in the patient fibroblast cells.
To further analyze the functionality of mutant SmE in mRNA splicing and gene expression,
the expression level of endogenous SmE was knocked down (KD) by siRNA targeting the 3’
UTR region in HEK293 cell, resulting in reduction of the expression level of SmE by approxi-
mately 80% (S3 Fig). Within this background, either wild type or mutant SmE was expressed
and RNA was then prepared for mRNA sequencing. In total, more than 30 million high quality
reads were obtained for each sample and around 93% of them could be uniquely aligned to the
human reference genome (hg19). Among 11670 expressed genes (average RPKM>1), 1060
showed significant alterations in the KD group as compared to the control (BH-adjusted P
value < 0.01, |log2 fold change| > 1). Importantly, these dramatic changes in the gene expres-
sion profile could be reversed by overexpression of wild type SmE, whereas the mutant was
much less effective (Fig 4C). A same pattern was also observed for the alteration of mRNA
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
8 / 23
A missense mutation in SNRPE linked to microcephaly
Fig 4. The identified mutation impairs the functionality of SNRPE/SmE in mRNA splicing. (A), The mRNA splicing
in patient derived fibroblast cells is impaired. MA plot shows the intron retention was dramatically increased in patient
derived fibroblast cells. X axis: the sum of log2 transformed splicing in and splicing out reads number for each intron. Y
axis: difference in percentage of intron retention (PIR) between fibroblast derived from the patient (mutant) and healthy
control. (B), The intron retention leads to decreased gene expression. Backgrounds are those genes without any intron
showing significantly increased retention. (C), Heatmap illustrating expression of 11670 protein coding genes (average
RPKM>1) in HEK293 cells among different experimental conditions. Control, control siRNA; KD, SmE siRNA; KD
+WT, SmE siRNA+wild type SmE; KD+Mut, SmE siRNA+mutant SmE. (D), Number of aberrant splicing events
induced by SNRPE/SmE knockdown (KD) comparing with control. RI, retained intron; SE, skipped exon; MXE,
mutually exclusive exon; ASS, alternative splice site. (E), Numbers of aberrant splicing events in KD, KD+Mut and KD
+WT comparing to control. (F), The intron retention leads to decreased gene expression. Backgrounds are those genes
without any intron showing significant increased retention.
https://doi.org/10.1371/journal.pgen.1008460.g004
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
9 / 23
A missense mutation in SNRPE linked to microcephaly
splicing: the massive aberrant splicing defect caused by SmE deficiency could be dramatically
reduced by overexpression of wild type, but not mutant SmE (Fig 4D and 4E). As already
observed in the patient-derived fibroblasts, mRNA transcripts with increased intron retention
were often down-regulated in KD HEK293 cells. Taken together, these results reveal that the
identified mutation impairs the functionality of SmE protein leading to extensive abnormal
gene expression and aberrant mRNA splicing.
Furthermore, to examine whether the retained introns, either in the patient fibroblast cells
or in HEK293 cells upon SmE knockdown, shared any characteristics, we analyzed 136 fea-
tures using the method as described by Braunschweig et al. [45]. As shown in S4 Fig, the fea-
tures that are sensitive to SmE dysfunction in both the patient fibroblast cells and HEK293, are
quite similar, with the GC content is the most significant one.
The SmE deficiency disturbs brain development of zebrafish
To explore the functional consequence of the identified SmE defect in vivo, we used zebrafish
as a model to dissect the effect of SmE deficiency on animal development. By injecting a mor-
pholino (E-MO) targeting the translation initiation site of zebrafish SmE (zSmE) into fertilized
zebrafish embryos at 1-cell stage, the endogenous zSmE levels were decreased after 48h injec-
tion (S5 Fig). To analyze the impact of zSmE on head development, the head size of embryos
injected with E-MO or a control morpholino (CO-MO) was measured after 48 hours post fer-
tilization. The head size of zebrafish injected with E-MO was significantly decreased (25%
reduction) compared to CO-MO injected embryos (Fig 5A and 5B). This phenotype is unlikely
to be the consequence of a general developmental delay, since the swim bladder and pigmenta-
tion of morphants were phenotypically normal. Although we observed a statistically significant
difference in the body length between E-MO and CO-MO, the magnitude of the change is
only marginal (Fig 5A and 5B).
To validate that this phenotype is caused by reduced zSmE, rescue experiments were per-
formed. The E-MO was co-injected with in vitro transcribed mRNA encoding 2A-mCherry
coupled with wild type zSmE (zSmE(WT)-2A-mCherry) lacking the binding site for E-MO.
Importantly, the co-injection of E-MO and zSmE(WT)-2A-mCherry could successfully
rescue the head-size phenotype. Therefore, the observed phenotype in E-MO injected zebra-
fish is specifically caused by depletion of zSmE (Fig 5A and 5B). However, co-injection of
E-MO and the in vitro transcribed mutant zSmE mRNA (zSmE (Mut)-2A-mCherry) failed to
rescue the defect (Fig 5A and 5B). Furthermore, overexpression of either wild type or mutant
zSmE (WT or Mut)-2A-mCherry alone did not show any phenotype (Fig 5A and 5B). Thus,
SmE is required for proper brain development in zebrafish and its deficiency causes a patient-
like phenotype.
Molecular mechanisms underlying zebrafish phenotypic changes induced
by SmE deficiency
The results in the patient-derived fibroblasts and in the HEK293 cells revealed that, when car-
rying the identified mutation, SmE fails to enter the biogenesis pathway of spliceosomal U
snRNP, resulting in aberrant mRNA splicing and alteration of the gene expression program.
We hence investigated whether the head phenotype in zebrafish is likewise caused by splicing
defects culminating in aberrant gene expression patterns. To explore this, RNA from the head
and tail regions of untreated zebrafish controls were compared with RNA from the same
region isolated from morpholino-injected zebrafish (CO-MO, E-MO alone, and E-MO+WT
as well as E-MO+Mut combinations were analyzed). In total, ~680 million reads were obtained
and 92.3% of them could be uniquely aligned to the zebrafish reference genome (danRer10).
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
10 / 23
A missense mutation in SNRPE linked to microcephaly
Fig 5. The SNPRE/SmE deficiency interferes with zebrafish brain development. (A), Measurement of zebrafish head size across
different experimental conditions. CO-MO, control morpholino; E-MO, SmE morpholino. The morpholino and/or in vitro transcribed
RNA are injected into embryo at 1-cell stage. Yellow line marked the region for quantification. (B), Quantification of zebrafish head size.
Left, head size; middle, body length; right, head size normalized by body length. UN, un-injected; CO-MO, control morpholino; E-MO,
SmE morpholino; WT, Wide type SmE gene in vitro transcript; Mut, mutant SmE gene in vitro transcript; The UN is normalized to 1.
� P<0.05; �� P<0.01. (C), PCA analysis of the expression of 16067 protein coding genes (RPKM > 1) in zebrafish head and tail samples
under different conditions. (D), Numbers of differentially expressed genes comparing to CO-MO. (E), Proportion of DEGs in 14
significant enriched dysregulated GO terms (biologic process, E-MO versus CO-MO, BH-adjusted p < 0.001). (F), Overlap of enriched
dysregulated KEGG pathways (E-MO versus CO-MO, BH-adjusted p < 0.001) between zebrafish head and tail samples. G, Numbers of
aberrant splicing events comparing to CO-MO.
https://doi.org/10.1371/journal.pgen.1008460.g005
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
11 / 23
A missense mutation in SNRPE linked to microcephaly
As expected, zebrafish head and tail have distinct expression profiles as evident from their
divergent transcript profiles (Fig 5C). In addition, the overall PCA clusters of embryos injected
with E-MO and E-MO+Mut significantly differed from untreated and CO-MO injected sam-
ples, while the rescue E-MO+WT represented an intermediate state between these two groups
in both head and tail (Fig 5C).
By comparing each fish treatment to the CO-MO control, thousands of differentially
expressed genes (DEG) were identified in each comparison (Fig 5D). To test whether these
alterations are a direct consequence of zSmE deficiency, we next attempted to rescue the
wild type transcriptome by the co-expression of zSmE variants. Indeed, upon co-expression of
wild type zSmE the number of DEG was drastically reduced, while DEG numbers in fish co-
expressing mutant zSmE was comparable to the zSmE knockdown (Fig 5D). Of note, the num-
ber of DEGs in the tail of E-MO zebrafish was much lower than that in the head (Fig 5D), sug-
gesting that the latter was more sensitive to zSmE deficiency.
Consistent with the observed phenotypic changes in zSmE deficient fish, down-regulated
DEGs in head are enriched for factors implicated in head development, central nervous system
development and cell fate commitment (Fig 5E). Importantly, the proportion of DEGs cluster-
ing in these GO terms was dramatically reduced by co-expressing of wild type zSmE but not
its pathogenic mutant (Fig 5E). The KEGG pathway analysis showed that the zSmE knock-
down affected some pathways such as the Notch signaling pathway in both head and tail (Fig
5F). In contrast, other pathways such as apoptosis were only activated in zebrafish head by
E-MO (Fig 5F) and may explain the death of neurons and reduced brain size. Not only alter-
ations in gene expression but also aberrant splicing induced by zSmE deficiency could be res-
cued by expressing wild type but not mutant zSmE (Fig 5G). Interestingly, the introns more
retained due to zSmE deficiency shared similar features as those due to SmE dysfunction in
the patient fibroblast and HEK293 cells (S4 Fig). Taken together, these results suggest that the
small brain size caused by zSmE deficiency is, likely, a consequence of altered gene expression
and aberrant splicing.
The EMX2 aberrant splicing is a target of defects in constitutive splicing
machinery and causes the microcephaly phenotype
Our RNA-seq data raised the possibility that the phenotype of zSmE deficient zebrafish might
be a consequence of disturbed transcription factor networks controlling neuron differentiation
as well as apoptosis (Fig 5E and 5F).
EMX genes (also known as empty spiracles homeobox) are vertebrate cognates of Drosoph-
ila head gap gene, empty spiracles (ems). EMX2, a homeobox-containing transcription factor,
plays critical roles in controlling patterning and proliferation of dorsal telencephalic progeni-
tors [46,47]. Yoshida et al. [48] reported that Emx2 defective mice lose the dentate gyrus and
display greatly reduced hippocampus and medial limbic cortex size. Emx2 has also been associ-
ated with the diseases of schizencephaly [49]. Importantly, our zSmE deficient zebrafish dis-
played reduced gene expression of the Emx2 gene (log2 fold change = -2.42, BH-adjusted
p = 1.28e-55) and increased intron retention in EMX2 mRNA (ΔPIR = 0.39, p = 9.6e-5,
fdr < 0.05) (Fig 6A and 6B).
This effect is strictly dependent on zSmE deficiency, as both intron retention and gene
expression change can be partially rescued by WT but less well by mutant zSmE (Fig 6A and
6B). Due to the critical role of EMX2 in controlling patterning and proliferation of dorsal tel-
encephalic progenitors, we explored whether alterations of the EMX2 transcript is causative
for the zebrafish phenotype. For this, we tried to rescue the head size phenotype in zSmE
depleted zebrafish by co-injection of in vitro transcribed EMX2 transcripts. Indeed, the co-
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
12 / 23
A missense mutation in SNRPE linked to microcephaly
Fig 6. The brain defect caused by SNPRE/SmE deficiency can be partially rescued by transcription factor EMX2.
(A), The expression of EMX2 gene is specifically disturbed in head of zebrafish after zSmE deficiency. (B), The zSmE
deficiency leads to increased intron retention of EMX2. (C), The head defect caused by zSmE deficiency can be
partially rescued by transcription factor EMX2. The morpholino and/or in vitro transcribed RNA is injected into
embryo at 1-cell stage. Yellow line marked the region for quantification. (D), Quantification of zebrafish head size.
Left, head size; middle, body length; right, head size normalized by body length. UN, un-injected; CO-MO, control
morpholino; E-MO, SmE morpholino, EMX2, EMX2 gene in vitro transcript. The UN is normalized to 1. � P<0.05;
�� P<0.01.
https://doi.org/10.1371/journal.pgen.1008460.g006
injection of EMX2 in vitro transcript with E-MO can partially rescue the brain defect (Fig 6C
and 6D). Of note, we observed only a partial rescue, which is likely due to the fact that zSmE
deficiency also affects the splicing of many other functional relevant genes. Furthermore,
application of EMX2 mRNA alone shows no phenotype (Fig 6C and 6D). These results reveal
that EMX2, as a downstream target, might act as a key factor as its splicing defects further
amplifies the consequence caused by zSmE deficiency.
Discussion
In higher eukaryotes, the specific morphology and physiological capacities of different cell
types is achieved through coordinated precise spatio-temporal expression of lineage specific
genes. Alternative splicing (AS), through differential selection of alternative splice sites in pre-
mRNA, is not only used to increase the coding capacity of the genome, but also extensively
applied to guide the developmental regulation [7]. Defects of mRNA splicing are frequently
related to human disease [17,50,51].
Here, we demonstrate that a heterozygous missense mutation (c.65T>C (p.Phe22Ser)) in
SNRPE/SmE gene causes aberrant mRNA splicing and abnormal gene expression, leading to a
severe brain defect through SNRPE/SmE deficiency (Figs 4 and 5). Saltzman et al. [52] previ-
ously showed that the SmB/B’ protein, another basal component of the spliceosome, self-
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
13 / 23
A missense mutation in SNRPE linked to microcephaly
regulates its expression by inclusion of a highly conserved cassette exon to regulate alternative
splicing through affecting the availability of spliceosomal U snRNPs. Although the SmE pro-
tein is also a basal component of spliceosome, the effect of SNRPE/SmE on mRNA splicing
and its physiological role has never been investigated. Our results revealed that, similar to
down-regulation of core spliceosomal proteins [53,54], the SNRPE/SmE (c.65T>C (p.Phe22-
Ser)) mutation impairs the biogenesis of spliceosomal U snRNPs (Figs 2 and 3), leading to
aberrant mRNA splicing in in vitro HEK293 cells (Fig 4) and in vivo zebrafish samples (Fig 5).
In zebrafish, the specific depletion of endogenous SNPRE/SmE mediated by translation initia-
tion blocking morpholino, leads to decreased head size (Fig 5A and 5B), which successfully
recapitulate the patient phenotype. Similar phenomena were also observed in previous studies
[53,54]. Bezzi et al. [54] showed that conditional knockout of PRMT5 in the central nervous
system (CNS) of mice leads to smaller brain, early postnatal death and aberrant mRNA
splicing. As a type II arginine methyltransferase [55], PRMT5 acts together with pICln and
WDR77/WD40 to symmetrically methylate the arginine residues in SmB/B’, SmD1 and SmD3
proteins to increase their affinity to SMN complex for promoting the spliceosomal U snRNPs
assembly [38,56]. Jia et al. [53] reported that mutation of a U2 snRNA gene in mice causes the
global disruption of alternative splicing and neurodegeneration. In U2 mutant mice, the size
of the cerebellum decreases through progressive neuron loss. No matter whether cells face a
conditional knockout of PRMT5 or a depletion of U2 snRNA or SNRPE/SmE, the direct con-
sequence is the reduced availability of spliceosomal U snRNPs. The CNS, as the most complex
structure, has the highest degree of alternative splicing to keep the diversity of transcriptome
and proteome to guide correct developmental fates [57,58]. Therefore, it is reasonable to
assume that the CNS is most sensitive to aberrant mRNA splicing and similar phenotypes can
be observed under these conditions.
Among the different classes of alternative splicing (AS) events, intron retention (IR) is the
least studied and usually regarded as the consequence of mis-splicing. However, an increasing
number of studies have shown that regulated IR is widely used as a physiological mechanism
to functionally tune the transcriptomes [59–61]. Wong et al. [60] showed that, during granulo-
cyte differentiation, IR coupled with NMD is applied as an energetically favorable way to pre-
cisely control gene expression. Yap et al. [59] demonstrated that IR is applied to coordinated
regulation of neuronal steady-state mRNA levels to guide the neuron differentiation. There-
fore, aberrant IR can be related to diseases as Bezzi et al. [54] and Jia et al. [53] reported that
the homeostasis of IR is disrupted after PRMT5 depletion or U2 snRNA mutation. IR is also
observed as the most abundant aberrant splicing type in the patient-derived fibroblast cells,
SNRPE/SmE depleted HEK293 cells and zebrafish zSmE knockdown head samples (Figs 4 and
5). Molecular analysis demonstrates that the extent of aberrant IR is negatively correlated with
gene expression, which might be mediated through NMD or nuclear sequestration (Fig 4B and
4F). Further KEGG pathway and GO term analyses of expression modulated genes in zebrafish
head with SNRPE/SmE deficiency show that the p53 signaling pathway is enriched in the up-
regulated genes whereas the down-regulated genes are significantly enriched in neuron devel-
opment (Fig 5E and 5F). The up-regulation of p53 signaling pathway was also reported by Jia
et al. [53] and Bezzi et al. [54] and considered to contribute to neuronal death. Therefore, like
with the PRMT5 depletion or U2 snRNA mutation, the p53 signaling pathway activation
might contribute similarly to the SNRPE/SmE deficiency phenotype.
Among those down-regulated genes related to neuron differentiation and brain develop-
ment, LHX5 promotes the forebrain development through inhibiting Wnt signaling [62].
LHX2 and LHX9 guide the neuronal differentiation and compartmentalization in the caudal
forebrain through regulating Wnt signaling [63]. EMX2 functions in the development of dor-
sal telencephalon, the EMX2 mutant shows defect of dentate gyrus and significantly reduced
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
14 / 23
A missense mutation in SNRPE linked to microcephaly
size of the hippocampus and medial limbic cortex [48,64]. Due to the phenotype similarity
between EMX2 mutant and SNRPE/SmE mutant, it is tempting to speculate that the pheno-
type of SNRPE/SmE mutant might be mediated through disrupting the expression of tran-
scription factors responsible for early brain development, such as EMX2. The result that
injection of an in vitro generated transcript encoding EMX2 can partially rescue the phenotype
of reduced SNRPE/SmE (Fig 6), is consistent with this hypothesis. The data are consistent with
the idea that during early development, the SNRPE/SmE deficiency disturbs the brain develop-
ment through interfering with the splicing of transcription factors, which are responsible for
guiding the early brain development.
In addition to the mutation we reported in this study, Pasternack et al. demonstrated that
the mutations of SNRPE/SmE (c.1A>G (p.M1?) and c.133G>A (p.G45S)) can cause the auto-
somal-dominant hypotrichosis simplex [31]. These mutations affect the solubility of proteins,
however, the soluble part can still efficiently integrate into functional spliceosomal U snRNPs.
Moreover, Weiss et al. identified a dominant mutation (c.153T>A (p.E51D)) in SmE from a
hypogonadism mouse strain [65]. Due to the different position of mutations, the effect of
mutations on the functionality of SNRPE might be very different. As the basal component of
spliceosomal U snRNPs, the consequence of such different effects from different mutations
could be further magnified through altered mRNA splicing and stability, especially the splic-
ing/expression of different transcription factors.
Finally, although we identified the SNRPE mutation (c.65T>C (p.Phe22Ser)) from only one
patient, the biochemical and zebrafish data provide strong evidence to link this mutation to
the microcephaly phenotype manifested in this patient. Therefore, this study expands on our
understanding of the effects of core spliceosomal machinery defects on early brain develop-
ment, and provides insight into the etiology of microcephaly.
Material and methods
Ethics statement
The study and use of human samples were approved by the Charite´ Ethics Committee (EA1/
212/08), and the patient’s parents provided written informed consent. For the animal research,
all experiments in the manuscript were performed with embryos of less than 5 days of age.
According to German and EU rule, those experiments need to only to be approved by the local
government and not considered to be animal experiments that need special permission. Zebra-
fish (Danio rerio) were bred and maintained as preciously established [66]. All experimental
procedures were performed according to the guidelines of the German animal welfare law and
approved by the local government (Government of Lower Franconia; Tierschugtzgesetz §11,
Abs. 1, Nr. 1 husbandry permit number 568/300-1870/13). All zebrafish experiments have
been performed at embryonic stage prior to independent feeding. Used zebrafish strains: TL
(Tüpfel long fin; leot1/lofdt2; ZFIN ID: ZDB-GENO-990623-2).
Exome sequencing
All family members were subjected to exome sequencing. In brief, DNA was extracted from
the patient and parents’ blood samples. According to the manufacture’s protocol, the genomic
DNA was enriched by Agilent Human All Exon V4 Kit (Agilent Technologies, Santa Clara,
CA, USA). The whole exome libraries were subjected to Illumina HiSeq2000 system for 100
cycles single end sequencing. After sequencing, the data analysis for exome sequencing was
performed as described before by Fro¨hler et al. [67].
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
15 / 23
A missense mutation in SNRPE linked to microcephaly
Cell lines and antibodies
Fibroblasts from the forearm of the patient and age-matched control were established accord-
ing to a standard protocol and cultured in DMEM with 4.5g/l D-glucose and pyruvate (Invi-
trogen, Darmstadt, Germany) supplemented with 15% fetal bovine serum (FBS) and 1%
penicillin-streptomycin.
SmE lentiviral overexpression plasmid was constructed by replacing the Cas9 cassette on
lentiCas9-Blast (Addgene, #52962) with the SmE sequence followed by HA-tag, T2A and
mCherry cassettes. For each virus package, HEK293T cells (3×105) were seeded in one well of
6-well plate, and were transfected with plasmid after 24 hours. For the transfection, 10.5ul PEI
(1μg/μl, Polysciences, #23966–2) and 3.5μg total plasmid (1μg lentiviral plasmid, 1.5μg pMD2.
G (Addgene, #12259), and 1μg psPAX2 (Addgene, #12260) were added to the 200μl Opti-
MEM (Thermo, #31985075). After 20 minutes, the mix was added to the cells. 12 hours after
transfection, the medium was replaced by the fresh medium. After 48 hours, the supernatant
were collected, and clarified by centrifugation (2000g), and filtrated through a 0.45μm filter
(Millex, #SLHV033RB). The transduction was done by incubating the viral particles contain-
ing supernatant with the patient fibroblast cells overnight in the presence of polybrene (8 μg/
μl, Sigma, H9268).
Stable HEK293 T-Rex Flp-In cell lines, inducibly expressing the HA-tagged wild type or
mutant SmE protein were constructed and maintained as previously described [68]. For tran-
sient transfection, HeLa and HEK293T cells were cultured in DMEM media supplemented
with 10% FBS.
The following antibodies were used in this study: anti-SMN (clone 7B10; purified from
hybridoma supernatant) [69], rabbit anti-pICln [36], mouse anti-m3G/m7G cap (H-20, a kind
gift from Prof. R. Lu¨hrmann) [70], mouse anti-Sm (Y12, a kind gift from Prof. J.A. Steitz) [71],
rabbit anti-coilin (H-300, Santa Cruz Biotechnology, sc-32860), rabbit anti-SmD3 (Pierce,
PA5–26288), rabbit anti-SmD1 (Pierce, PA5–12459), rabbit anti-SmF (Abcam, ab66895),
mouse anti-FLAG (Sigma, F1804 and F3165) and anti-HA (Covance). For western blotting, we
used secondary goat antibodies conjugated with horse raddish peroxidase; anti-mouse (Sigma,
A4416) and anti-rabbit (Sigma, A6154). For indirect immunostaining we used Cy5-conjugated
goat secondary antibody (red channel), anti-rabbit IgG (Jackson ImmunoResearch Laborato-
ries, 111-175-144) and Alexa488-conjugated goat secondary antibody (green channel), anti-
mouse (Thermo Scientific, A11017).
Immunoprecipitation (IP) of proteins and RNA-protein complexes from
stable cell lines or transient transfections, 3’-end labeling of RNA
HEK293T cells were seeded in 150mm cell culture dishes and transfected at 80% confluency
using Mirus Transit-X2 system as per manufacturer’s protocol for immunoprecipitations with
20μg of SmE wild type or mutant construct or dual-expression plasmid or left untransfected
for mock immunoprecipitations. Lysate were prepared 48 hours after transient transfection or
after 24 hours of induction of stable cell lines with 100ng/ml doxycycline.
All IP experiments were performed as previously described [43]. Briefly, the cells were
homogenized in lysis buffer (50mM HEPES pH7.5, 150mM NaCl, 2.5mM MgCl2, 1% NP-40
substitute, RNasin and proteinase inhibitors) and insoluble debris was removed by centrifuga-
tion. The supernatant was then collected, concentration estimated using Bradford assay and
incubated with Protein-G Dynabeads (Thermo Scientific) coupled with corresponding anti-
bodies or with anti-FLAG agarose M2 affinity gel (Sigma) for 3h at 4˚C. After incubation, the
beads were washed three times with ice-cold wash buffer (50mM HEPES pH7.5, 300mM
NaCl, 2.5mM MgCl2) and once with 1×PBS with 2.5mM MgCl2. The immunoprecipitate was
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
16 / 23
A missense mutation in SNRPE linked to microcephaly
subsequently dissociated from the beads using 1×La¨mmli SDS dye, separated on a SDS-PAGE
and analyzed by western blotting or directly treated with TRIzol (Thermo Scientific) for RNA
extractions as per manufacturer’s protocol. The precipitated RNA was resuspended in nuclease
free water and incubated with 32P-pCp and T4 RNA ligase in an overnight reaction at 4˚C.
The RNA was precipitated after Phenol-chloroform extraction and separated on 8% polyacryl-
amide-Urea denaturing gel and exposed for autoradiography.
Immunostaining and confocal microcopy
For immunostaining, HeLa cells were grown on coverslips and transfected with FLAG-tagged
wild type or mutant SmE constructs respectively at 70% confluency using Mirus Transit-X2 or
left un-transfected (control). After 48 hours of transfection, the coverslips were processed for
immunostaining. Control primary human fibroblasts and patient fibroblast were seeded on
coverslips and grown to 70% confluency before immunostaining. The cells were washed and
fixed with 4% para-formaldehyde and permeabilized with 0.2% Triton X-100 in 1×PBS and
blocked with 10% FCS. Primary and secondary antibodies were diluted in 2% FBS. After pri-
mary and secondary antibody binding and washes, the coverslips were mounted using Mowiol
4–88 mounting medium. Confocal imaging was carried out using Leica SP5 confocal micro-
scope with photomultiplier and the images were processed using ImageJ software.
Injection and analysis of zebrafish embryos
The zebrafish (Danio rerio) embryos were maintained and harvested as previous described
[66]. The translation-blocking morpholino against zebrafish SmE was designed and obtained
from Gene tools (SmE MO: 5’-TGTCCTTGTCCTCTGTACGCCATTC-3’) targeting the
translation initiation site. Control morpholino was a scrambled nucleotide sequence provided
by Gene tools (5’-TGTCGTTCTGCTCTCTACCCCATTC -3’). 1nl of morpholino solution
(final concentration 20nM) was injected into zebrafish embryos at the 1–2 cell stage. For RNA
rescue and over-expression experiments, in vitro transcribed RNA (final concentration of
150pg) encoding the CDS of zebrafish SmE with/without point mutation was fused with
mCherry and separated from each other by 2A-tag. To avoid the targeting by SmE morpho-
lino, synonymous codons were used to substitute the 4th-7th amino acid positions. The coding
sequence was changed from AGAGGACAAGGA to CGTGGCCAGGGT. To quantify the
phenotype, the images of embryos were taken at 48 hours post fertilization (hpf), and the size
of the heads and length of the body were quantified. All experiments were repeated for three
times and the significance of the morphant phenotype was determined by Student’s t-test.
RNA sequencing
Total RNAs were extracted from the patient derived fibroblast cells, HEK293 cell lines, zebra-
fish heads and tails using TRIzol reagent (Life Technologies) following manufacturer’s instruc-
tion. Stranded mRNA sequencing libraries were prepared with 500 ng total RNA according to
manufacturer’s protocol (Illumina). The libraries were subjected to Illumina HiSeq 2000 sys-
tem for 100 cycles single end sequencing.
RNA-seq data analysis
All RNA-seq reads were aligned to a reference genome (human: hg19; zebrafish: danRer10) by
using STAR with transcriptome annotation (human: Gencode v18; zebrafish: ensemble 82).
HTseq-Count was further utilized to calculate gene expression by counting uniquely mapped
reads within each gene. DEseq2 was then applied to identify differentially expressed genes
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
17 / 23
A missense mutation in SNRPE linked to microcephaly
between different conditions. Based on transcriptome annotation, splicing events including
alternative splicing sites (ASS), skipped exon (SE), retained intron (RI) and mutually exclusive
exons (MXE) were constructed. Especially for SE and RI, all middle exons and introns were
considered potentially to be skipped or retained. Using reads aligned to exon-exon junction
and exon-intron boundaries, expression of each splicing event was quantified and further
compared between each two different conditions. We used a rank-product based method as
described in a previous study [67], to estimate significance (p < 0.001, fdr < 0.05 were defined
as significant) by checking consistence among different biological replicates. For zebrafish
RNA-seq data analysis, we examined GO and KEGG pathway enrichment (BH-adjusted P
value < 0.001) for genes, which were differentially expressed (BH-adjusted P value < 0.001,
|log2 fold change| > 2, RPKM > 1) between E-MO and Control-MO, using WEB-based Gene
SeT AnaLysis Toolkit (WebGestalt). In brief, we estimated significance of the overrepresenta-
tion of up and down regulated genes in each GO term and KEGG pathway, comparing with
background genes respectively (all expressed genes, i.e. RPKM > 1). Next, in each significant
enriched GO-term, proportions of differentially expressed genes among all genes in the GO
term across different comparisons, including E-MO versus Control-MO, E-MO+WT versus
Control-MO, and E-MO+MT versus Control-MO, were estimated separately. For enriched
KEGG pathways, we also checked the overlap between the results from head and tail RNA-seq
data.
Supporting information
S1 Text. Support information-Clinical information of the patient.
(DOCX)
S1 Fig. U snRNP levels are reduced in the patient due to the SmE mutation. (A-B), Indirect
immunofluorescence and confocal microscopy of control and patient fibroblasts. Empty
white arrowheads indicate localization pattern observed and filled white arrowheads indicate
zoomed in region shown in the overlay inset. (A), Co-staining with DAPI (blue), m3G/m7G
cap of U snRNA (green) and SmD1 (magenta). Control fibroblasts (top panel) show abundant
U snRNPs in nuclear speckles and both SmD1 and U snRNAs are predominantly absent
from the cytoplasm. In patient fibroblasts (bottom panel) though there is an excellent co-loca-
lizaiton of U snRNAs and SmD1, there is a decrease in their nuclear abundance and there is an
increase in their cytoplasmic localization. (B), Indirect immunofluorescence and confocal
microscopy of DAPI (blue), symmetrically dimethylated (sDMA)-Sm proteins (green) and
coilin (magenta). In comparison to the control fibroblasts (top panel), the patient cells (bottom
panel) have reduced Sm proteins in the nucleus and an increased cytoplasmic retention. Coilin
foci is not present in the images as primary cells lacks CBs. (C), Quantitative real-time PCR
analysis of snRNAs and SmE in control (black bars) and patient (gray bars) fibroblasts from
two independent biological replicates. (D), The SmE protein expression level in patient and
control fibroblasts was checked by western blotting. The tubulin was used as loading control.
(E), Immunoprecipitation of Sm proteins from control and patient fibroblasts (bottom panel,
western blotting) and autoradiography (top panel) after 3’-end labeling of coprecipitated RNA.
Mock indicates immunoprecipitation control without any antibody coupled to the beads. (F),
Quantification of autoradiography in E; control in black and patient in gray, from two inde-
pendent biological replicates.
(TIF)
S2 Fig. The impaired mRNA splicing in patient fibroblasts can be rescued by overexpres-
sion wild type SmE protein. (A), Wild type SmE protein was successfully overexpressed in the
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
18 / 23
A missense mutation in SNRPE linked to microcephaly
patient fibroblast cells. The expression level was estimated based on RNA-seq data. (B), The
MA plot compares the intron retention in the patient fibroblast cells with to those without
overexpression of wild type SmE protein; X axis, log2 transformed the product of splicing in
and splicing out reads number for each intron; Y axis, difference in percentage of intron reten-
tion (PIR) between the patient fibroblast cells with overexpression of wild type SmE protein
(OE) and those without (mutant). (C), The MA plot compares the intron retention between
the patient fibroblast cells with overexpression of wild type SmE to fibroblast cells from healthy
control (control). (D), The scatter plot illustrates the PIR changes between healthy control vs
mutant (X axis) and OE vs mutant (Y axis).
(TIF)
S3 Fig. The endogenous SmE can be successfully knocked down by siRNA. Western blot
analysis shows that the endogenous SmE can be specifically depleted by SmE siRNA, targeting
to the 3’ UTR region, and the exogenous HA-tagged SmE protein can be efficiently induced.
The β-tubulin is used as loading control.
(TIF)
S4 Fig. The KS-statistics for the 18 most representative features among 136 features across
different comparisons. The features were compared between group 1 and group 2 (left panel);
between group 3 and group 4 (middle panel); between group 5 and group 6 (right panel). The
GC content is the most significantly enriched feature among all the three comparisons. Group
1: introns with increased retention in the patient fibroblast cells vs healthy control fibroblast
cells (adjusted p < 0.05, delta PIR > 0.1); Group 2: introns without increased retention in the
patient fibroblast cells vs healthy control fibroblast (delta PIR < 0.05, p > 0.05), this group
serves as background for group 1; Group 3: introns with increased retention in HEK293 upon
SmE knockdown vs control HEK293 (adjusted p < 0.05, delta PIR > 0.1); Group 4: introns
without increased retention in HEK293 upon SmE knockdown vs control HEK293 (delta
PIR < 0.05, p > 0.05), this group serves as background for group 3; Group 5: introns with
increased retention in zebrafish upon SmE knockdown vs control (adjusted p < 0.05, delta
PIR > 0.1); Group 6: introns without increased retention in zebrafish upon SmE knockdown
vs control (delta PIR < 0.05, p > 0.05), this group serves as background for group 5.
(TIF)
S5 Fig. The endogenous SmE in zebrafish can be successfully knocked down by SmE mor-
pholino. Western blot analysis shows that the endogenous zSmE can be specifically depleted
by SmE morpholino, targeting to the translation initiation site. The β-tubulin is used as load-
ing control. UN, un-injection; CO-MO, control morpholino; E-MO, SmE morpholino.
(TIF)
Acknowledgments
Bioinformatic analysis was supported by the Center for Computational Science and Engineer-
ing of Southern University of Science and Technology. We thank Prof. R. Lu¨hrmann and Prof.
J.A. Steitz for their kind gift of antibodies. We thank Mirjam Feldkamp, Claudia Langnick,
Madlen Sohn and Claudia Quedenau from Berlin Institute of Medical Systems Biology
(BIMSB) for their excellent technical assistance.
Author Contributions
Conceptualization: Angela M. Kaindl, Utz Fischer, Wei Chen.
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
19 / 23
A missense mutation in SNRPE linked to microcephaly
Data curation: Tao Chen, Bin Zhang, Thomas Ziegenhals.
Formal analysis: Thomas Ziegenhals, Sebastian Fro¨hler.
Investigation: Tao Chen, Archana B. Prusty, Clemens Grimm, Yuhui Hu, Bernhard Schaefke,
Liang Fang, Min Zhang, Nadine Kraemer, Angela M. Kaindl.
Methodology: Tao Chen, Bin Zhang, Thomas Ziegenhals.
Resources: Tao Chen, Angela M. Kaindl.
Software: Bin Zhang.
Supervision: Utz Fischer, Wei Chen.
Writing – original draft: Tao Chen, Bin Zhang, Utz Fischer, Wei Chen.
Writing – review & editing: Tao Chen, Bin Zhang, Thomas Ziegenhals, Archana B. Prusty,
Min Zhang, Utz Fischer, Wei Chen.
References
1. Berget SM, Moore C, Sharp PA. Spliced segments at the 5’ terminus of adenovirus 2 late mRNA.
PNAS. 1977; 74: 3171–3175. https://doi.org/10.1073/pnas.74.8.3171 PMID: 269380
2. Chow LT, Gelinas RE, Broker TR, Roberts RJ. An amazing sequence arrangement at the 5’ ends of
adenovirus 2 messenger RNA. Cell. 1977; 12(1): 1–8. https://doi.org/10.1016/0092-8674(77)90180-5
PMID: 902310
3.
Lerner MR, Boyle JA, Mount SM, Wolin SL, Steitz JA. Are snRNPs involved in splicing? Nature. 1980;
283(5743): 220–224. https://doi.org/10.1038/283220a0 PMID: 7350545
4. Wahl MC, Will CL, Lu¨ hrmann R. The Spliceosome: design principles of a dynamic RNP machine. Cell.
2009; 136(4): 701–718. https://doi.org/10.1016/j.cell.2009.02.009 PMID: 19239890
5. Wickramasinghe VO, Gonzàlez-Porta M, Perera D, Bartolozzi AR, Sibley CR, Hallegger M, et al. Regu-
lation of constitutive and alternative mRNA splicing across the human transcriptome by PRPF8 is deter-
mined by 50 splice site strength. Genome Biol. 2015; 16(201): 1–21.
6. Pan Q, Shai O, Lee LJ, Frey BJ, Blencowe BJ. Deep surveying of alternative splicing complexity in the
human transcriptome by high-throughput sequencing. Nat Genet. 2008; 40(12): 1413–1415. https://
doi.org/10.1038/ng.259 PMID: 18978789
7. Nilsen TW, Graveley BR. Expansion of the eukaryotic proteome by alternative splicing. Nature. 2010;
463(7280): 457–463. https://doi.org/10.1038/nature08909 PMID: 20110989
8. Blencowe BJ. Alternative splicing: new insights from global analyses. Cell. 2006; 126(1): 37–47. https://
doi.org/10.1016/j.cell.2006.06.023 PMID: 16839875
9. Wang ET, Sandberg R, Luo S, Khrebtukova I, Zhang L, Mayr C, et al. Alternative isoform regulation in
human tissue transcriptomes. Nature. 2008; 456(7221): 470–476. https://doi.org/10.1038/nature07509
PMID: 18978772
10. Witten JT, Ule J. Understanding splicing regulation through RNA splicing maps. Trends Genet. 2011;
27(3): 89–97. https://doi.org/10.1016/j.tig.2010.12.001 PMID: 21232811
11. Barash Y, Calarco JA, Gao W, Pan Q, Wang X, Shai O, et al. Deciphering the splicing code. Nature.
2010; 465(7294): 53–59. https://doi.org/10.1038/nature09000 PMID: 20445623
12. Boutz PL, Stoilov P, Li Q, Lin C-H, Chawla G, Ostrow K, et al. A post-transcriptional regulatory switch in
polypyrimidine tract-binding proteins reprograms alternative splicing in developing neurons. Genes
Dev. 2007; 21(13): 1636–1652. https://doi.org/10.1101/gad.1558107 PMID: 17606642
13. Graveley BR, Brooks AN, Carlson JW, Duff MO, Landolin JM, Yang L, et al. The developmental tran-
scriptome of Drosophila melanogaster. Nature. 2011; 471(7339): 473–479. https://doi.org/10.1038/
nature09715 PMID: 21179090
14. Baker BS. Sex in flies: the splice of life. Nature. 1989; 340(6234): 521–524. https://doi.org/10.1038/
340521a0 PMID: 2505080
15. Xie J, Black DL. A CaMK IV responsive RNA element mediates depolarization-induced alternative splic-
ing of ion channels. Nature. 2001; 410(6831): 936–939. https://doi.org/10.1038/35073593 PMID:
11309619
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
20 / 23
A missense mutation in SNRPE linked to microcephaly
16. Singh RK, Cooper TA. Pre-mRNA splicing in disease and therapeutics. Trends Mol Med. 2012; 18:
472–482. https://doi.org/10.1016/j.molmed.2012.06.006 PMID: 22819011
17. Cooper TA, Wan L, Dreyfuss G. RNA and Disease. Cell. 2009; 136(4): 777–793. https://doi.org/10.
1016/j.cell.2009.02.011 PMID: 19239895
18. Niblock M, Gallo J-M. Tau alternative splicing in familial and sporadic tauopathies. Biochemical Soc
Trans. 2012; 40(4): 677–680.
19. Comitato A, Spampanato C, Chakarova C, Sanges D, Bhattacharya SS, Marigo V. Mutations in splicing
factor PRPF3, causing retinal degeneration, form detrimental aggregates in photoreceptor cells. Hum
Mol Genet. 2007; 16(14): 1699–1707. https://doi.org/10.1093/hmg/ddm118 PMID: 17517693
20.
21.
Linder B, Hirmer A, Gal A, Ru¨ ther K, Bolz HJ, Winkler C, et al. Identification of a PRPF4 loss-of-function
variant that abrogates U4/U6.U5 Tri-snRNP integration and is associated with retinitis pigmentosa.
PLoS One. 2014; 9(11): e111754. https://doi.org/10.1371/journal.pone.0111754 PMID: 25383878
Tanackovic G, Ransijn A, Ayuso C, Harper S, Berson Eliot L, Rivolta C. A missense mutation in PRPF6
causes impairment of pre-mRNA splicing and autosomal-dominant retinitis pigmentosa. Am J Hum
Genet. 2011; 88(5): 643–649. https://doi.org/10.1016/j.ajhg.2011.04.008 PMID: 21549338
22. Maubaret CG, Vaclavik V, Mukhopadhyay R, Waseem NH, Churchill A, Holder GE, et al. Autosomal
dominant retinitis pigmentosa with intrafamilial variability and incomplete penetrance in two families car-
rying mutations in PRPF8. Invest Ophthalmol Vis Sci. 2011; 52: 9304–9309. https://doi.org/10.1167/
iovs.11-8372 PMID: 22039234
23. Venturini G, Rose AM, Shah AZ, Bhattacharya SS, Rivolta C. CNOT3 Is a modifier of PRPF31 muta-
tions in retinitis pigmentosa with incomplete penetrance. PLoS Genet. 2012; 8(11): e1003040. https://
doi.org/10.1371/journal.pgen.1003040 PMID: 23144630
24.
Linder B, Dill H, Hirmer A, Brocher J, GLee GP, Mathavan S, et al. Systemic splicing factor deficiency
causes tissue-specific defects: a zebrafish model for retinitis pigmentosa. Hum Mol Genet. 2011; 20(2):
368–377. https://doi.org/10.1093/hmg/ddq473 PMID: 21051334
25. Cvačkova´ Z, Matějů D, Staněk D. Retinitis pigmentosa mutations of SNRNP200 enhance cryptic splice-
site recognition. Hum Mutat. 2014; 35(3): 308–317. https://doi.org/10.1002/humu.22481 PMID:
24302620
26. Mordes D, Luo X, Kar A, Kuo D, Xu L, Fushimi K, et al. Pre-mRNA splicing and retinitis pigmentosa. Mol
Vis. 2006; 12: 1259–1271. PMID: 17110909
27.
28.
29.
Lefebvre S, Bu¨rglen L, Reboullet S, Clermont O, Burlet P, Viollet L, et al. Identification and characteriza-
tion of a spinal muscular atrophy-determining gene. Cell. 1995; 80(1): 155–165. https://doi.org/10.
1016/0092-8674(95)90460-3 PMID: 7813012
Zhang Z, Lotti F, Dittmar K, Younis I, Wan L, Kasim M, et al. SMN deficiency causes tissue-specific per-
turbations in the repertoire of snRNAs and widespread defects in splicing. Cell. 2008; 133(4): 585–600.
https://doi.org/10.1016/j.cell.2008.03.031 PMID: 18485868
Lynch DC, Revil T, Schwartzentruber J, Bhoj EJ, Innes AM, Lamont RE, et al. Disrupted auto-regulation
of the spliceosomal gene SNRPB causes cerebro–costo–mandibular syndrome. Nat Commun. 2014;
5: 4483–4488. https://doi.org/10.1038/ncomms5483 PMID: 25047197
30. Bacrot S, Doyard M, Huber C, Alibeu O, Feldhahn N, Lehalle D, et al. Mutations in SNRPB, encoding
components of the core splicing machinery, cause cerebro-costo-mandibular syndrome. Hum Mut.
2015; 36(2): 187–190. https://doi.org/10.1002/humu.22729 PMID: 25504470
31. Pasternack Sandra M, Refke M, Paknia E, Hennies Hans C, Franz T, Schafer N, et al. Mutations in
SNRPE, which encodes a core protein of the spliceosome, cause autosomal-dominant hypotrichosis
simplex. Am J Hum Genet. 2013; 92(1): 81–87. https://doi.org/10.1016/j.ajhg.2012.10.022 PMID:
23246290
32. Edery P, Marcaillou C, Sahbatou M, Labalme A, Chastang J, Touraine R, et al. Association of TALS
developmental disorder with defect in minor splicing component U4atac snRNA. Science. 2011; 332
(6026): 240–243. https://doi.org/10.1126/science.1202205 PMID: 21474761
33. He H, Liyanarachchi S, Akagi K, Nagy R, Li J, Dietrich RC, et al. Mutations in U4atac snRNA, a compo-
nent of the minor spliceosome, in the developmental disorder MOPD I. Science. 2011; 332(6026):
238–240. https://doi.org/10.1126/science.1200587 PMID: 21474760
34. Neiswanger K, Stanford DR, Sparkes RS, Nishimura D, Mohandas T, Klisak I, et al. Assignment of the
gene for the small nuclear ribonucleoprotein E (SNRPE) to human chromosome 1q25–q43. Genomics.
1990; 7(4): 503–508. https://doi.org/10.1016/0888-7543(90)90192-w PMID: 2143747
35. Raker VA, Plessel G, Lu¨ hrmann R. The snRNP core assembly pathway: identification of stable core
protein heteromeric complexes and an snRNP subcore particle in vitro. EMBO J. 1996; 15(9): 2256–
2269. PMID: 8641291
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
21 / 23
A missense mutation in SNRPE linked to microcephaly
36. Chari A, Golas MM, Klingenha¨ger M, Neuenkirchen N, Sander B, Englbrecht C, et al. An assembly
chaperone collaborates with the SMN complex to generate spliceosomal snRNPs. Cell. 2008; 135(3):
497–509. https://doi.org/10.1016/j.cell.2008.09.020 PMID: 18984161
37. Grimm C, Chari A, Pelz J-P, Kuper J, Kisker C, Diederichs K, et al. Structural basis of assembly chaper-
one- mediated snRNP formation. Mol Cell. 2013; 49(4): 692–703. https://doi.org/10.1016/j.molcel.
2012.12.009 PMID: 23333303
38. Meister G, Eggert C, Bu¨ hler D, Brahms H, Kambach C, Fischer U. Methylation of Sm proteins by a com-
plex containing PRMT5 and the putative U snRNP assembly factor pICln. Curr Biol. 2001; 11(24):
1990–1994. https://doi.org/10.1016/s0960-9822(01)00592-9 PMID: 11747828
39.
Fischer U, Luhrmann R. (1990) An essential signaling role for the m3G cap in the transport of U1
snRNP to the nucleus. Science. 1990; 249(4970): 786–790. https://doi.org/10.1126/science.2143847
PMID: 2143847
40. Mouaikel J, Verheggen C, Bertrand E, Tazi J, Bordonne´ R. Hypermethylation of the cap structure of
both yeast snRNAs and snoRNAs requires a conserved methyltransferase that is localized to the nucle-
olus. Mol Cell. 2002; 9(4): 891–901. https://doi.org/10.1016/s1097-2765(02)00484-7 PMID: 11983179
41.
Lemm I, Girard C, Kuhn AN, Watkins NJ, Schneider M, Bordonne R, et al. Ongoing U snRNP biogene-
sis is required for the integrity of cajal bodies. Mol Biol Cell. 2006; 17(7): 3221–3231. https://doi.org/10.
1091/mbc.E06-03-0247 PMID: 16687569
42. Hearst SM, Gilder AS, Negi SS, Davis MD, George EM, Whittom AA, et al. Cajal-body formation corre-
lates with differential coilin phosphorylation in primary and transformed cell lines. J Cell Sci. 2009; 122:
1872–1881. https://doi.org/10.1242/jcs.044040 PMID: 19435804
43. Prusty AB, Meduri R, Prusty BK, Vanselow J, Schlosser A, Fischer U. Impaired spliceosomal U snRNP
assembly leads to Sm mRNA down-regulation and Sm protein degradation. J Cell Biol. 2017; 216(8):
2391–2407. https://doi.org/10.1083/jcb.201611108 PMID: 28637748
44. So BR, Wan L, Zhang Z, Li P, Babiash E, Duan J, et al. A U1 snRNP–specific assembly pathway
reveals the SMN complex as a versatile hub for RNP exchange. Nat Struct Mol Biol. 2016; 23(3): 225–
230. https://doi.org/10.1038/nsmb.3167 PMID: 26828962
45. Braunschweig U, Barbosa-Morais NL, Pan Q, Nachman EN, Alipanahi B, Gonatopoulos-Pournatzis T,
et al. Widespread intron retention in mammals functionally tunes transcriptomes. Genome Res. 2014;
24(11): 1774–1786. https://doi.org/10.1101/gr.177790.114 PMID: 25258385
46. Heins N, Cremisi F, Malatesta P, Gangemi RMR, Corte G, Price J, et al. Emx2 promotes symmetric
cell divisions and a multipotential fate in precursors from the cerebral cortex. Mol Cell Neurosci. 2001;
18(5): 485–502. https://doi.org/10.1006/mcne.2001.1046 PMID: 11922140
47. O’Leary DDM, Chou S-J, Sahara S. Area patterning of the mammalian cortex. neuron. 2007; 56(2):
252–269. https://doi.org/10.1016/j.neuron.2007.10.010 PMID: 17964244
48. Yoshida M, Suda Y, Matsuo I, Miyamoto N, Takeda N, Kuratani S, et al. Emx1 and Emx2 functions in
development of dorsal telencephalon. Development. 1997; 124(1): 101–111. PMID: 9006071
49. Brunelli S, Faiella A, Capra V, Nigro V, Simeone A, Cama A, et al. Germline mutations in the
homeobox gene EMX2 in patients with severe schizencephaly. Nat Genet. 1996; 12(1): 94–96. https://
doi.org/10.1038/ng0196-94 PMID: 8528262
50. Bonnal S, Vigevani L, Valca´rcel J. The spliceosome as a target of novel antitumour drugs. Nat Rev
Drug Discov. 2012; 11(11): 847–859. https://doi.org/10.1038/nrd3823 PMID: 23123942
51.
Licatalosi DD, Darnell RB. Splicing regulation in neurologic disease. Neuron. 2006; 52(1): 93–101.
https://doi.org/10.1016/j.neuron.2006.09.017 PMID: 17015229
52. Saltzman AL, Pan Q, Blencowe BJ. Regulation of alternative splicing by the core spliceosomal machin-
ery. Genes Dev. 2011; 25(4): 373–384. https://doi.org/10.1101/gad.2004811 PMID: 21325135
53.
Jia Y, Mu John C, Ackerman Susan L. Mutation of a U2 snRNA gene causes global disruption of alter-
native splicing and neurodegeneration. Cell. 2012; 148(1–2): 296–308. https://doi.org/10.1016/j.cell.
2011.11.057 PMID: 22265417
54. Bezzi M, Teo SX, Muller J, Mok WC, Sahu SK, Vardy LA, et al. Regulation of constitutive and alternative
splicing by PRMT5 reveals a role for Mdm4 pre-mRNA in sensing defects in the spliceosomal machin-
ery. Genes Dev. 2013; 27(17): 1903–1916. https://doi.org/10.1101/gad.219899.113 PMID: 24013503
55. Bedford MT, Clarke SG. Protein Arginine Methylation in Mammals: Who, What, and Why. Mol Cell.
2009; 33(1): 1–13. https://doi.org/10.1016/j.molcel.2008.12.013 PMID: 19150423
56.
Friesen WJ, Massenet S, Paushkin S, Wyce A, Dreyfuss G. SMN, the product of the spinal muscular
atrophy gene, binds preferentially to dimethylarginine-containing protein targets. Mol Cell. 2001; 7(5):
1111–1117. https://doi.org/10.1016/s1097-2765(01)00244-1 PMID: 11389857
57. Yeo G, Holste D, Kreiman G, Burge CB. Variation in alternative splicing across human tissues. Genome
Biol. 2004; 5(10): R74. https://doi.org/10.1186/gb-2004-5-10-r74 PMID: 15461793
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
22 / 23
A missense mutation in SNRPE linked to microcephaly
58. Grosso AR, Gomes AQ, Barbosa-Morais NL, Caldeira S, Thorne NP, Grech G, et al. Tissue-specific
splicing factor gene expression signatures. Nucleic Acids Res. 2008; 36(15): 4823–4832. https://doi.
org/10.1093/nar/gkn463 PMID: 18653532
59. Yap K, Lim ZQ, Khandelia P, Friedman B, Makeyev EV. Coordinated regulation of neuronal mRNA
steady-state levels through developmentally controlled intron retention. Genes Dev. 2012; 26(11):
1209–1223. https://doi.org/10.1101/gad.188037.112 PMID: 22661231
60. Wong Justin JL, Ritchie W, Ebner Olivia A, Selbach M, Wong Jason WH, Huang Y, et al. Orchestrated
intron retention regulates normal granulocyte differentiation. Cell. 2013; 154(3): 583–595. https://doi.
org/10.1016/j.cell.2013.06.052 PMID: 23911323
61. Ge Y, Porse BT. The functional consequences of intron retention: alternative splicing coupled to NMD
as a regulator of gene expression. BioEssays. 2014; 36(3): 236–243. https://doi.org/10.1002/bies.
201300156 PMID: 24352796
62. Peng G, Westerfield M. Lhx5 promotes forebrain development and activates transcription of secreted
Wnt antagonists. Development. 2006; 133(16): 3191. https://doi.org/10.1242/dev.02485 PMID:
16854974
63. Peukert D, Weber S, Lumsden A, Scholpp S. Lhx2 and Lhx9 determine neuronal differentiation and
compartition in the caudal forebrain by regulating Wnt signaling. PLoS Biol. 2011; 9(12): e1001218.
https://doi.org/10.1371/journal.pbio.1001218 PMID: 22180728
64. Suda Y, Hossain ZM, Kobayashi C, Hatano O, Yoshida M, Matsuo I, et al. Emx2 directs the develop-
ment of diencephalon in cooperation with Otx2. Development. 2001; 128(13): 2433–2450. PMID:
11493561
65. Weiss J, Hurley LA, Harris RM, Finlayson C, Tong M, Fisher LA, et al. ENU mutagenesis in mice identi-
fies candidate genes for hypogonadism. Mamm Genome. 2012; 23(5–6): 346–355. https://doi.org/10.
1007/s00335-011-9388-5 PMID: 22258617
66. Westerfield M. The zebrafish Book: a guide for the laboratory use of zebrafish (Danio Rerio). University
of Oregon Press; 1995.
67.
Fro¨ hler S, Kieslich M, Langnick C, Feldkamp M, Opgen-Rhein B, Berger F, et al. Exome sequencing
helped the fine diagnosis of two siblings afflicted with atypical Timothy syndrome (TS2). BMC Med
Genet. 2014; 15: 1–6.
68. Wang Y, Gogol-Do¨ring A, Hu H, Fro¨ hler S, Ma Y, Jens M, et al. Integrative analysis revealed the molec-
ular mechanism underlying RBM10-mediated splicing regulation. EMBO Mol Med. 2013; 5(9): 1431–
1442. https://doi.org/10.1002/emmm.201302663 PMID: 24000153
69. Meister G, Hannus S, Plo¨ttner O, Baars T, Hartmann E, Fakan S, et al. SMNrp is an essential pre-
mRNA splicing factor required for the formation of the mature spliceosome. EMBO J. 2001; 20(9):
2304–2314. https://doi.org/10.1093/emboj/20.9.2304 PMID: 11331595
70. BOCHNIG P, REUTER R, BRINGMANN P, LU¨ HRMANN R. A monoclonal antibody against 2,2,7-tri-
methylguanosine that reacts with intact, class U, small nuclear ribonucleoproteins as well as with 7-
methylguanosine-capped RNAs. Eur J Biochem. 1987; 168(2): 461–467. https://doi.org/10.1111/j.
1432-1033.1987.tb13439.x PMID: 2959477
71.
Lerner EA, Lerner MR, Janeway CA Jr., Steitz JA. Monoclonal antibodies to nucleic acid-containing cel-
lular constituents: probes for molecular biology and autoimmune disease. PNAS. 1981; 78(5): 2737–
2741. https://doi.org/10.1073/pnas.78.5.2737 PMID: 6789322
PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008460 October 31, 2019
23 / 23
| null |
10.1371_journal.ppat.1009995.pdf
|
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
|
All relevant data are within the manuscript and its Supporting Information files.
|
RESEARCH ARTICLE
Acquisition of yersinia murine toxin enabled
Yersinia pestis to expand the range of
mammalian hosts that sustain flea-borne
plague
David M. BlandID*, Ade´ laïde MiarinjaraID
Joseph Hinnebusch
¤a, Christopher F. Bosio, Jeanette CalarcoID
¤b, B.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
Laboratory of Bacteriology, Rocky Mountain Laboratories, National Institute of Allergy and Infectious
Diseases, NIH, Hamilton, Montana, United State of America
¤a Current address: Department of Environmental Sciences, Emory University, Atlanta, Georgia, United
State of America
¤b Current address: Department of Integrative Biology, University of South Florida, Tampa, Florida, United
State of America
* [email protected]
OPEN ACCESS
Citation: Bland DM, Miarinjara A, Bosio CF, Calarco
J, Hinnebusch BJ (2021) Acquisition of yersinia
murine toxin enabled Yersinia pestis to expand the
range of mammalian hosts that sustain flea-borne
plague. PLoS Pathog 17(10): e1009995. https://
doi.org/10.1371/journal.ppat.1009995
Editor: Deborah M. Anderson, University of
Missouri, UNITED STATES
Received: July 14, 2021
Accepted: September 30, 2021
Published: October 14, 2021
Copyright: This is an open access article, free of all
copyright, and may be freely reproduced,
distributed, transmitted, modified, built upon, or
otherwise used by anyone for any lawful purpose.
The work is made available under the Creative
Commons CC0 public domain dedication.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: This research was funded by the
Intramural Research Program of the NIH, NIAID.
The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Abstract
Yersinia murine toxin (Ymt) is a phospholipase D encoded on a plasmid acquired by Yersi-
nia pestis after its recent divergence from a Yersinia pseudotuberculosis progenitor. Despite
its name, Ymt is not required for virulence but acts to enhance bacterial survival in the flea
digestive tract. Certain Y. pestis strains circulating in the Bronze Age lacked Ymt, suggest-
ing that they were not transmitted by fleas. However, we show that the importance of Ymt
varies with host blood source. In accordance with the original description, Ymt greatly
enhanced Y. pestis survival in fleas infected with bacteremic mouse, human, or black rat
blood. In contrast, Ymt was much less important when fleas were infected using brown rat
blood. A Y. pestis Ymt− mutant infected fleas nearly as well as the Ymt+ parent strain after
feeding on bacteremic brown rat blood, and the mutant was transmitted efficiently by flea
bite during the first weeks after infection. The protective function of Ymt correlated with red
blood cell digestion kinetics in the flea gut. Thus, early Y. pestis strains that lacked Ymt
could have been maintained in flea-brown rat transmission cycles, and perhaps in other
hosts with similar blood characteristics. Acquisition of Ymt, however, served to greatly
expand the range of hosts that could support flea-borne plague.
Author summary
The bacterium Yersinia pestis causes highly lethal bubonic plague in a wide variety of
mammals and is transmitted primarily by the bites of infected fleas. During its recent evo-
lutionary divergence from Yersinia pseudotuberculosis, a mild pathogen incapable of flea-
borne transmission, Y. pestis acquired a new gene that encodes a phospholipase enzyme
called Yersinia murine toxin (Ymt). This was a critical step in the transition to an insect-
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
1 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
borne life cycle as it was reported that Ymt activity greatly enhances bacterial survival in
the flea gut. Recent genomic sequencing of ancient Y. pestis strains revealed that some
lacked Ymt, leading to the conclusion that these strains were not transmitted by flea bite.
Here, we report that the importance of Ymt for survival in the flea is greatly dependent on
host blood source. Ymt is required if fleas take up Y. pestis in mouse, human, or black rat
blood, but is not required if brown rat blood is used. We conclude that ancient Y. pestis
strains lacking Ymt could have circulated in certain flea-rodent transmission cycles.
Acquisition of Ymt, however, enabled Y. pestis to greatly expand its host range to an eco-
logically broad range of mammals and their fleas.
Introduction
Yersinia pestis evolved from the closely related Yersinia pseudotuberculosis, a food-borne path-
ogen that generally causes self-limiting enteric disease, within the last 6,000 years [1,2]. Making
only 5 specific genetic changes to Y. pseudotuberculosis results in a strain able to produce a
transmissible infection in the flea [3]. One key gene acquired during transition to the flea-
borne life cycle encodes Yersinia murine toxin (Ymt), a phospholipase D enzyme that has an
important role in the ability of Y. pestis to colonize the flea midgut [4]. Ymt is encoded on the
Y. pestis-specific pMT1 plasmid, which was acquired through horizontal gene transfer [5].
Ymt was once believed to be an important virulence factor in the mammalian host, as Ymt-
enriched protein fractions are highly lethal to mice and rats [6,7]. However, Ymt is not
required for typical plague disease progression and virulence and the LD50 of a Ymt-negative
strain in mice is equivalent to that of wild-type Y. pestis [8]. Murine toxicity of Ymt is likely
related to its ability to act as a β-andrenergic-blocking agonist in mice and rats [9,10], but tox-
icity is not observed in other mammals such as guinea pigs, rabbits, dogs, and primates [11].
Application of molecular Koch’s postulates to a standardized flea model of Y. pestis infection
revealed that Ymt’s true biological function is to enhance bacterial survival in the flea midgut,
significantly improving the ability of the plague bacillus to stably infect and be transmitted by
its vector [4,12].
In the original characterization, a Ymt mutant was rapidly eliminated from ~90% of Xenop-
sylla cheopis fleas, and those few fleas with chronic infections had reduced bacterial burdens in
which only the proventricular valve in the foregut (and not the midgut) was colonized. The
incidence of transmission-enhancing proventricular blockage due to Y. pestis biofilm accumu-
lation was correspondingly rare, indicating low potential for Ymt− strains to be vectored by
fleas [4]. The Y. pestis Ymt mutant was eliminated from fleas within the first 24h following
uptake in a blood meal, preceded by conversion of the bacilli to an atypical spheroplast mor-
phology in the midgut [4]. Bacterial spheroplast formation usually indicates damage to, or loss
of, the bacterial outer membrane and a reduction in osmotolerance. Addition of recombinant
Ymt protein to the infectious blood meal did not protect mutant bacilli from clearance, and in
fleas coinfected with Ymt− and Ymt+ Y. pestis, Ymt− bacteria persisted in the midgut only if
they were embedded within a biofilm of Ymt+ bacilli. Immunohistochemistry and immunoas-
says of culture supernatants indicate that Ymt is not secreted and is released only upon cell
lysis [8]. Collectively, current data indicate that the Ymt phospholipase exerts its protective
function intracellularly and that Ymt mutant bacteria are better able to survive in the flea gut if
protected from the surrounding digestive and/or immunological milieu of the midgut [4].
In seeming contradiction to the rapid clearance phenotype observed for Ymt mutant bacte-
ria [4], a separate study showed that Ymt− Y. pestis could survive in and be transmitted by fleas
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
2 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
up to 3 days after infection nearly as efficiently as the parental strain [13]. Notably, the study
indicating that Ymt was dispensable for this early-phase transmission used brown rat (Rattus
norvegicus) blood for the infectious blood meal [13], whereas the study demonstrating rapid
clearance of the Ymt mutant from fleas used mouse blood [4]. Recently, we have shown that
the source of infectious host blood alters the nature of the Y. pestis infection in the flea foregut
[14]. Specifically, the slow digestion rate of brown rat blood and the relative insolubility of its
hemoglobin promotes more rapid and extensive foregut infection (proventriculus and esopha-
gus) in which partially digested blood meal contents mixed with Y. pestis are refluxed from the
midgut into the esophagus; a phenomenon we have termed post-infection esophageal reflux
(PIER) [14]. PIER-inducing blood sources reduce the time required for some rodent fleas to
become infectious; increasing the number of bacilli transmitted during the first few days fol-
lowing an infectious blood meal. Because brown rat blood promotes infection of the esophagus
and the bactericidal agent of Ymt− strains is believed to be generated during blood digestion in
the midgut, we thought Ymt− Y. pestis might be better able to survive in fleas if PIER-inducing
blood sources were used for the infection. To test this hypothesis and evaluate the permissive-
ness of different host blood sources to flea colonization, we infected rodent fleas with either
wild-type or Ymt− Y. pestis suspended in blood collected from mice, brown rats, black “roof”
rats (Rattus rattus), or humans.
In resolution of the seemingly contradictory results, we found that Ymt mutant Y. pestis
can chronically infect and be transmitted by the rodent fleas X. cheopis and Oropsylla montana
at much higher levels if brown rat blood is used for the infectious blood meal than if mouse,
human, or black rat blood is used. Our results suggest that ancestral Y. pestis strains lacking
Ymt could have been maintained in flea-borne transmission cycles involving brown rats and
perhaps other mammals with similarly permissive blood biochemistry. Acquisition of Ymt,
however, fortified that ability and allowed Y. pestis to greatly expand its host range to involve
many other mammals and their fleas, resulting in strong positive selective pressure for the
Ymt+ lineage.
Results
The Y. pestis Ymt mutant induces PIER in fleas following an infectious
brown rat blood meal
When fleas ingest Y. pestis suspended in blood that is digested relatively slowly and is charac-
terized by a poorly soluble hemoglobin molecule, many of them exhibit post-infection esoph-
ageal reflux (PIER) [14]. The foregut of these fleas contains a mixture of partially digested
blood components and Y. pestis aggregates that extends from the proventriculus forward into
the esophagus within 24 h after an infectious blood meal. This phenomenon is seen following
infections using brown rat and guinea pig blood, but not when mouse or gerbil blood is used
[14]. Because digestive enzymes are likely not present at high concentration in the foregut, we
hypothesized that bacteria aggregated there would be protected from bactericidal agents gener-
ated in the midgut, and that the foregut thus might provide a niche for Ymt− Y. pestis to tem-
porarily colonize if PIER-inducing blood is used for the infection.
To determine if PIER induction occurs and could provide protection to Ymt− strains, we
infected X. cheopis fleas using one of four blood sources (brown rat, black rat, mouse, and
human) and screened them 24 h later for PIER (Fig 1). Consistent with our previous study,
PIER was evident in ~20% of fleas infected with wild-type Y. pestis KIM6+ using brown rat
blood, but not when mouse blood was used. PIER was also induced in fleas infected using
black rat blood, but at lower incidence (~5%) than for brown rat blood. Fleas infected using
human blood did not develop PIER (Fig 1A). Notably, PIER was also observed in fleas
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
3 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
Fig 1. A Y. pestis Ymt mutant induces PIER in X. cheopis fleas when brown rat blood is used for the infectious
blood meal. A) Incidence of post-infection esophageal reflux (PIER) in groups of 25 to 220 X. cheopis fleas 24 h after
feeding on brown rat (Rn), black rat (Rr), mouse (M), or human blood (H) containing 1.5 x 108–1.1 x 109 CFU/ml
KIM6+ or KIM6+ymtH188N Y. pestis. Bars show the mean and standard error of 3 independent experiments
(n = 164–438 mixed sex fleas). �p < 0.005 by chi-square test. B) Female X. cheopis with PIER 24 h after feeding on
black rat blood containing GFP-positive Y. pestis KIM6+; blue arrow indicates where blood and Y. pestis has been
refluxed from the proventriculus and/or midgut into the esophagus. C) light and D) fluorescence microscopy images
of the digestive tract dissected from this flea showing the presence of partially digested blood components and bacteria
in the proventriculus (PV) and esophagus (E). Scale bar = 50 μm.
https://doi.org/10.1371/journal.ppat.1009995.g001
following infection with Ymt− Y. pestis in brown rat blood (but not black rat blood), but only
about half as often as in fleas infected with the parental KIM6+ strain (Fig 1A).
As in our previous study, PIER correlated with the presence of hemoglobin crystals, par-
tially digested red blood cell stroma, and Y. pestis in the proventriculus and esophagus of fleas
(Fig 1B–1D) [14]. Hemoglobin crystals were commonly observed in the midgut of infected
fleas when black rat blood was used for the infectious blood meal but appeared to be more sol-
uble than brown rat hemoglobin crystals. Black rat hemoglobin crystals typically had a long
rod-like shape and rapidly dissolved in the PBS we used to prepare wet mounts of infected flea
digestive tracts, making them difficult to image and possibly causing us to underestimate their
prevalence. In addition, unlike brown rat blood [14], hemolysis of black rat red blood cells in
water did not result in hemoglobin crystallization. Hemoglobin crystals were not observed in
the gut of fleas infected using mouse blood [14], and rarely observed in fleas infected using
human blood.
Collectively, these data suggest that Ymt− Y. pestis can colonize the flea foregut, induce
PIER, and potentially be protected from elimination when brown rat blood is used for the
infectious blood meal.
Blood source affects colonization of rodent fleas by Ymt-deficient Y. pestis
To determine whether blood source and PIER affect the overall ability of the Ymt mutant to
colonize the flea, X. cheopis were fed mouse, human, black rat, or brown rat blood containing
~ 5x108 CFU/ml Y. pestis KIM6+, KIM6+ymtH188N, or KIM6+ymtH118N (pYmt). Infected
fleas subsequently received two sterile maintenance blood meals over the course of 1 week to
evaluate their potential to become blocked. Replicating previously published results [4], 80–
90% of female fleas infected with the Ymt mutant in mouse blood cleared the infection within
24 h, whereas strains that produce the functional Ymt enzyme were rarely cleared by fleas dur-
ing the first week (Fig 2A). The foregut of the few fleas that remained infected with the Ymt
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
4 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
Fig 2. Ymt− and Ymt+ Y. pestis colonize female X. cheopis similarly when brown rat blood is used for the
infectious blood meal, but not if mouse, human, or black rat blood are used. Groups of female X. cheopis fleas that
fed on mouse (blue), black rat (black), human (orange), or brown rat (red) blood containing 1.5x108–1.1x109 CFU/ml
Y. pestis KIM6+, KIM6+ymtH188N, or KIM6+ymtH188N (pYmt) were scored for 1 week for A) the percentage of
fleas that remained infected; B) the percentage that developed obstruction of the foregut (partial or complete blockage)
that interfered with normal blood-feeding; and C) bacterial burden. Data are cumulative from 3 (KIM6+ and KIM6
+ymtH188N groups) or 1 (KIM6+ymtH188N(pYmt) groups) independent experiments. Samples consisted of 7–20
female (A and C) or 25–220 fleas (roughly equal numbers of males and females; B) per experiment. The mean and
standard error (A, B) or median (C) are indicated. �p < 0.05 by chi-square (A, B) or by Kruskal-Wallis test with
Dunn’s post-test (mouse, human, and brown rat groups) or Mann-Whitney test (black rat group) (C). Dotted lines
indicate the limit of detection (40 CFU). KIM6+ymtH188N(pYmt) was not used for black rat blood infections due to
the limited availability of this blood.
https://doi.org/10.1371/journal.ppat.1009995.g002
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
5 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
mutant in mouse blood rarely became obstructed by a bacterial mass (partially or fully
blocked) during the first week of the infection and had reduced bacterial burdens (Fig 2B and
2C). Comparable results were observed for fleas infected with the Ymt mutant in human or
black rat blood (Fig 2). In contrast, the average infection rate (73%) and median bacterial bur-
den (1.2 x106 CFU) after 1 week for fleas infected with the Ymt mutant in brown rat blood
were only modestly lower than the infection rate (mean 95%) and bacterial burden (median
1.5 x106 CFU) of fleas infected with the wild-type parent strain. Furthermore, fleas infected
with the Ymt mutant in brown rat blood developed proventricular blockage at a rate similar to
that of fleas infected with the parent strain at the first feeding following infection (12% vs 13%)
and at a slightly reduced rate (6% vs 8%) after the second feeding (Fig 2B). To verify that these
results were not unique to X. cheopis rat fleas, we replicated the experiments using mouse and
brown rat blood with Oropsylla montana, a North American ground squirrel flea. The results
mirrored those seen for X. cheopis: the Ymt mutant was rapidly cleared from O. montana fleas
infected using mouse blood, but those infected using brown rat blood had infection rates, bac-
terial burdens, and proventricular obstruction rates that were equivalent to or only slightly
reduced from wild-type levels (S1 Fig).
These results were surprising, because although a minority of the fleas infected using brown
rat blood developed PIER (10–20%), much higher proportions (30–100%) remained infected
for up to 1 week. Thus, it seems unlikely that PIER alone accounted for the high rates of flea
colonization observed for the Ymt mutant-brown rat blood infections. However, the data pro-
vide insight into a previous report that the Ymt mutant can be as efficiently transmitted as its
wild-type parent during the early phase when brown rat blood is used for the infectious blood
meal [13]. In sum, our results show that the previously reported lability of Ymt− Y. pestis in the
flea gut varies depending on the infectious blood source. This mutant fares poorly after infec-
tious mouse, black rat, or human blood meals, but survives much better after brown rat infec-
tious blood meals. This effect is conserved in two rodent flea species from distinct taxonomic
families.
The protective role of Ymt is more pronounced in female fleas than in male
fleas
In the original characterization of the Ymt mutant strain in fleas, infection rates were deter-
mined only for female X. cheopis fleas infected using mouse blood [4], and the rates in Fig 2
were also based on female fleas. Because the metabolism and physiology of insects is not identi-
cal between sexes, we evaluated infection rates separately for male and female fleas infected
with Ymt mutant Y. pestis using either mouse or brown rat blood. Unexpectedly, when mouse
blood was used for the flea infection, 61% of male fleas remained infected after 24 h, whereas
only a single female (4%) had evidence of GFP+ bacteria in the digestive tract (Fig 3A and
Table 1). In contrast, when brown rat blood was used for the flea infections, male and female
fleas had equivalently high rates of Y. pestis colonization and 25% (including examples of both
sexes) had more severe bacterial infections in the proventriculus (Table 1). These data show an
enhanced capacity for Ymt− bacteria to survive in the male flea midgut.
The infection status and bacterial load of fleas 1 and 7 days after infection was determined
to assess whether the Ymt mutant persisted in male fleas infected using mouse blood. Signifi-
cantly more males (40–70%) than females (0–25%; Fig 3B) remained infected for up to 1 week
with Ymt mutant Y. pestis. The mean bacterial load of fleas infected with Ymt− Y. pestis was
higher for males at both 1 and 7 days after infection, but the difference was not statistically sig-
nificant (Fig 3C). In contrast, infection rates were identical between sexes when infected with
the wild-type parent strain (Fig 3A and 3B). Regardless of sex, the few fleas that remained
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
6 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
Fig 3. The Y. pestis Ymt mutant colonizes male fleas more efficiently than females following infection using
mouse blood. Infection rates for groups of female or male X. cheopis infected using mouse blood (blue symbols) or
brown rat blood (red symbols) containing 1x108–5.7x108 CFU/ml GFP-positive KIM6+ or KIM6+ymtH188N Y. pestis
were determined 1 day after infection by fluorescence microscopy of dissected flea digestive tracts A); or 0, 1, and 7
days after infection by CFU counts from individual triturated fleas (B, C). For A, each symbol represents the
percentage of fleas containing GFP+ bacteria in their digestive tract. n = 4–10 fleas of each sex in 3 independent
experiments (Table 1). For B and C, the mean and standard error (B) or median (C) of pooled data from 3
independent experiments for groups of 5–20 fleas infected using mouse blood are shown. �p < 0.05 by chi-square test
(B) or two-way ANOVA with Tukey’s post-test (C). D) Examples of the foregut infection in female or male X. cheopis 1
day after ingesting KIM6+ymtH188N Y. pestis suspended in mouse blood (Left) or brown rat blood (Right). Scale
bar = 50 μm.
https://doi.org/10.1371/journal.ppat.1009995.g003
Table 1. Flea Dissection Summary.
Blood Source /Experiment
% Fleas Infected (KIM6
+ymtH188N)
Bacteria Present In:
PV Infection Severity
X. cheopis
Mouse Blood #1
Mouse Blood #2
Mouse Blood #3
Total/Average
Rat Blood #1
Rat Blood #2
Rat Blood #3
Male
67% (9)
100% (4)
40% (10)
61% (14/23)
70% (10)
83% (6)
100% (10)
Female
10% (10)
0% (5)
0% (9)
4% (1/24)
100% (10)
100% (6)
90% (10)
PV+ MG
PV only
MG only
66%
100%
50%
0%
0%
0%
33%
0%
50%
Light
67%
100%
50%
Moderate
Heavy
0%
0%
0%
0%
0%
0%
67% (10/15)
0% (0/15)
33% (5/15)
67% (10/15)
0% (0/15)
0% (0/15)
82%
100%
100%
6%
0%
0%
12%
0%
0%
47%
73%
100%
35%
27%
11%
6%
0%
0%
Total/Average
85% (22/26)
96% (25/26)
94% (44/47)
2% (1/47)
4% (2/47)
70% (33/47)
23% (11/47)
2% (1/47)
PV = proventriculus, MG = midgut. Numbers in parentheses indicate flea sample sizes.
https://doi.org/10.1371/journal.ppat.1009995.t001
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
7 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
infected had a lightly colonized proventriculus (� 25% coverage of the proventricular spines
by a bacterial mass; (Fig 3D and Table 1).
Clearance of Ymt− Y. pestis correlates with the rate of RBC digestion
To confirm that the differential survival of the Ymt mutant in the flea was not due to inhibitory
components in certain blood sources, in vitro growth of wild-type and mutant Y. pestis in defi-
brinated mouse blood, rat blood, or BHI-hemin broth was monitored during 24 h of incuba-
tion at 21˚ or 37˚C. As expected, the Ymt mutant and the parental strain grew equally well in
all three substrates (S2A and S2B Fig). Additionally, exposure to hemolyzed mouse RBCs or
defibrinated mouse plasma did not affect bacterial viability (S2C Fig). These data indicate that
the Ymt mutant phenotype observed in the flea gut is unrelated to differential growth charac-
teristics in mouse blood and that the clearance of the Ymt mutant may require processing of
the blood meal by flea digestive enzymes [4].
To determine the fraction of blood responsible for clearance of the Ymt mutant in the flea
gut, we infected X. cheopis using reconstituted, plasma-swapped mouse or brown rat blood
(rat plasma mixed with mouse RBCs or vice versa) containing KIM6+ymtH188N Y. pestis.
One day after infection, only 7% of fleas infected using brown rat plasma with mouse RBCs
remained infected compared to 80% of those infected using mouse plasma and brown rat
RBCs (Fig 4A). In sum, addition of brown rat plasma to mouse RBCs did not rescue the Ymt
mutant in the flea gut, and addition of mouse plasma to rat RBCs did not result in impaired
bacterial infectivity. These results indicate that the bactericidal agent is primarily produced as
a consequence of digestion of RBCs, such as those from a mouse, and that the contribution of
plasma to the Ymt− strain phenotype is likely modest or inconsequential.
Given that Ymt appeared to protect against a bactericidal product of RBC digestion, we
decided to test whether female and male X. cheopis digested mouse and rat RBCs at equivalent
rates. First, to get a better understanding of flea digestion kinetics between flea sexes, we deter-
mined that female fleas ingest, on average, roughly twice as much blood as male fleas (Fig 4B).
Next, we found that the RBC concentration in the flea gut was similar, regardless of blood
source or flea sex, within the first 30 minutes after the bloodmeal (Fig 4C). However, by two
hours after feeding, female fleas that ingested mouse blood had the largest reduction in red cell
counts. The majority of mouse RBCs had lysed by 2 h in ~1/3rd of female fleas, whereas all other
flea sex-blood source combinations showed a lower RBC digestion rate (Fig 4C). Beyond 2 h,
both mouse and rat RBCs frequently aggregated in large clusters in the flea digestive tract, ren-
dering hemocytometer counts unfeasible. To address this, we imaged digestive tracts excised
from fleas every 2 h for the first 8 h after an uninfected blood meal. We found that most female
fleas completely digest and liquify mouse RBCs within 4–6 h (Figs 4D and S3). At 6 h, the gut of
70% of the female fleas (9 of 13) contained only a moderately viscous pink fluid, devoid of cellu-
lar material. Male fleas took longer to digest mouse blood; after 6–8 h of digestion, only ~30%
had completely liquified their blood meal (Figs 4D and S3). In contrast, brown rat blood took
considerably longer for both male and female fleas to digest. By 6 h after feeding, the digestive
tract always contained a thick, viscid, brownish-red slurry of aggregated RBC stroma in various
stages of breakdown, distinctly more viscous than what was present in fleas fed mouse blood
(S3 Fig). By 8 h, no female fleas (0 of 11) and only 7% of males (1 of 15) had completely liquified
their brown rat blood meal (Fig 4D). The relative amount of solid material in the fleas fed
brown rat blood remained fairly constant over the first 8 h of digestion, indicating that both
male and female fleas typically require more than 8 h to liquify brown rat blood.
The identical temporal patterns of RBC digestion were also observed in fleas infected with
Ymt− Y. pestis. By 24 h after infection, fleas infected using brown rat blood routinely contained
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
8 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
Fig 4. Survival of the Ymt mutant in the flea correlates with slower RBC digestion. A) Bacterial titers and infection
rates for groups of female X. cheopis infected using reconstituted, plasma-swapped mouse blood (brown rat plasma
mixed with mouse RBCs; blue) or brown rat blood (mouse plasma mixed with rat RBCs; red) containing 1.3x108–
2.8x108 CFU/ml KIM6+ymtH188N. Data are the pooled results from 3 independent experiments (n = 10); bars
represent the median. �p < 0.0001 by Mann-Whitney test. B) Blood meal volumes of individual female or male X.
cheopis allowed to feed for 1 h on a neonatal mouse. Mean blood meal volumes are indicated, �p <0.0001 by Student’s
t-test. C) The RBC concentration in individual X. cheopis female or male digestive tracts 0.5 or 2 h after ingestion of
sterile mouse or rat blood. Bars represent the mean of 3 independent assays using n = 3–6 (0.5 h) or n = 6–10 (2 h)
fleas. �p <0.05 by two-way ANOVA with Tukey’s post-test. D) The mean proportion and range of male or female X.
cheopis that completely liquified sterile mouse or brown rat blood during the first 8 h of digestion. Data are from
groups of 3–6 digestive tracts excised from fleas at each timepoint and condition from 3 independent experiments;
n = 9–15. �p <0.05 by Fisher’s exact test compared to rat blood group. A representative image series of these data is
shown in S3 Fig.
https://doi.org/10.1371/journal.ppat.1009995.g004
significant quantities of undigested midgut material (Fig 3D). In contrast, fleas infected using
mouse blood typically contained only the viscous pink or red liquid, with little to no solid
material (Fig 3D). Collectively, these data indicate a correlation between the rate of RBC diges-
tion and the clearance of Ymt− Y. pestis from the flea gut.
Ymt− Y. pestis can be transmitted beyond the early phase when fleas are
infected using brown rat blood
To assess transmission of the mutant strain, groups of O. montana or X. cheopis fleas were
infected with the Ymt mutant or the parent strain in either brown rat or mouse blood and
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
9 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
Fig 5. Rodent fleas can transmit Ymt mutant Y. pestis for at least 3 weeks when infected using brown rat blood. Y.
pestis transmission dynamics were monitored for 3 to 4 weeks for groups of 150–267 X. cheopis (A) or O. montana
fleas infected using 3.4 x 108−1.9 x109 CFU/ml KIM6+ or KIM6+ymtH188N Y. pestis (B) in either mouse (blue) or
brown rat (red) blood and subsequently maintained on sterile blood of the same type. Numbers in parentheses indicate
the total number of fleas that fed followed by the number of fleas with evidence of foregut obstruction (partially or fully
blocked). Roughly equivalent numbers of male and female fleas were used for transmission assays. Infection rate was
determined for groups of 10–20 female C) X. cheopis or D) O. montana at various times following infection.
https://doi.org/10.1371/journal.ppat.1009995.g005
were fed periodically on sterile blood of the same source. After each maintenance feed, the
blood was collected from the feeding device and plated to determine the number of CFUs
transmitted. Early-phase transmission (3 days post-infection) of Ymt mutant Y. pestis was
detected for O. montana infected using brown rat blood, but not if mouse blood was used (Fig
5A and 5B). Furthermore, X. cheopis and O. montana infected using brown rat blood transmit-
ted moderate to high levels of the Ymt mutant for at least 2 or 3 weeks, respectively, during the
biofilm-dependent phase of transmission. Transmission of the Ymt mutant and the parental
Ymt+ Y. pestis strains by X. cheopis infected using brown rat blood was roughly comparable
(Fig 5A). In contrast, fleas infected using mouse blood rarely became blocked and only a single
instance of transmission was observed (X. cheopis, day 17), in which very few CFU were trans-
mitted (Fig 5A). Infection rates of fleas used for transmission tests were similar to those shown
in Figs 2 and S1 (Fig 5C and 5D). Reduced transmission by fleas infected using brown rat
blood during the later weeks of infection may be partially attributable to the higher mortality
rate of these fleas. Greater than 80% of both flea species had died by 3 weeks after infection.
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
10 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
The elevated mortality may be attributable to the more severe Y. pestis infection in the foregut
of fleas infected using brown rat blood (Figs 1 and 5) [14]. The overall incidence of foregut
obstruction (fully and partially blocked fleas) was 39–45% for rodent fleas infected and main-
tained on brown rat blood but only 0–2.5% for those infected using mouse blood (Fig 5). Con-
sistent with our finding that male X. cheopis are more susceptible to infection by Ymt− Y. pestis
in mouse blood (Fig 3), all 5 fleas in this group that became blocked were males (Fig 5A).
Discussion
Gene gain and gene loss were both major drivers of the recent evolutionary emergence of Y.
pestis, and acquisition of Ymt was critical for the transition to a flea-borne life cycle because it
greatly enhanced survival of Y. pestis in the flea gut [15]. The original report of the protective
effect of Ymt hypothesized that the Phospholipase D activity of Ymt directly or indirectly pro-
tects Y. pestis against a bactericidal byproduct of blood digestion [4], and was based mainly on
female fleas infected using mouse blood. We extend that original characterization here, show-
ing that the protective function of Ymt is much less important when fleas feed on bacteremic
brown rat blood than on bacteremic mouse, human, or black rat blood. With brown rat blood,
flea infection and proventricular blockage rates were not significantly different for the first few
days of infection and only slightly reduced after 1 week for Ymt− compared to Ymt+ Y. pestis,
whereas these rates were greatly reduced in the first 24 h of infection for the Ymt mutant using
the other blood sources. Correspondingly, both X. cheopis and O. montana fleas infected using
brown rat blood transmitted Ymt− Y. pestis relatively efficiently, whereas as predicted by the
previous study [4] fleas infected using mouse blood rarely transmitted, and few CFU were
transmitted.
Other major findings of this study are that Ymt likely protects Y. pestis from a product of
RBC digestion, and not a plasma digestion product as hypothesized previously [4]; and that
the importance of Ymt correlates with RBC digestion kinetics. Previous studies of RBC diges-
tion by female X. cheopis demonstrated that fleas digest their blood meals more rapidly than
many other blood-feeding arthropods, such as mosquitos and ticks, and that the digestive tract
expresses a number of trypsin-like transcripts within the first hours following feeding [16,17].
Electron microscopic analysis of the X. cheopis midgut epithelium indicates that secretory vesi-
cles, likely containing digestive enzymes, are produced in advance of feeding and are released
more or less immediately following ingestion of blood [18]. However, we found that the diges-
tion rate can vary depending on the host blood source. Brown rat RBCs were digested more
slowly and incompletely than mouse RBCs. Fleas infected using brown rat blood routinely had
large quantities of undigested material in their gut 24 h after infection, whereas fleas infected
using mouse blood were essentially devoid or had greatly reduced amounts of solid blood
material. In addition, our results show that female X. cheopis digest mouse blood more rapidly
than males do, despite ingesting roughly twice as much blood. Digestion kinetics often differ
between insects of the opposite sex, as oviposition and egg maturation are physiologically asso-
ciated with digestion, and females typically have greater energetic demands due to these bio-
logical imperatives [19,20]. The digestion patterns we observed for female fleas and mouse
blood, in which most erythrocytes are digested during the first few hours, are consistent with
previous estimates [16]. Other microscopic analyses of flea gut contents also have indicated
that X. cheopis females digest mouse blood more rapidly than males [21]. Overall, the surviv-
ability of Ymt− Y. pestis after being ingested by a flea correlated well with RBC digestion rate:
good, nearly normal survival in both sexes with rat blood infections; and intermediate survival
in male fleas but poor survival in female fleas with mouse blood infections.
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
11 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
In addition to digestive enzymes, RNA-Seq analysis of the X. cheopis digestive tract tran-
scriptome revealed that antimicrobial peptides are rapidly produced in response to ingestion
of Y. pestis [17]. However, Ymt does not appear to have a role in protection against the flea
immune response. Drosophila, an insect that has been developed as a surrogate Y. pestis infec-
tion model, produces a diverse array of antimicrobials in response to Gram-negative bacteria,
yet Ymt mutant strains show no defect in fruit fly colonization or bacterial burden [22]. In this
model, Drosophila larvae ingest Y. pestis-laden cornmeal agar, rather than blood, to initiate
infection [23], suggesting that the Ymt mutant colonization defect in the flea is uniquely
related to blood digestion rather than to insect innate immunity. Furthermore, the Ymt
mutant shows no enhanced susceptibility to common antimicrobials that target the outer
membrane (polymyxin B, SDS, lysozyme, etc.) or to other potentially bacteriolytic enzymes
and environmental stressors that would be encountered in the arthropod gut environment
(proteases, lipases, osmotic and oxidative stress) [24]. It was previously hypothesized that the
Phospholipase D activity of Ymt provides protection against a bacteriolytic byproduct of blood
digestion by either modifying the bacterial envelope to make the bacteria resistant to lysis (pro-
phylaxis model), or by direct or indirect neutralization of the lytic agent (antidote model) [24].
While we have not resolved the mechanism by which Ymt provides protection to the bacteria,
our data further indicate that it is a byproduct of RBC digestion that induces the abnormal
spheroplast morphology indicative of cell envelope damage to the Ymt− mutant in the flea.
Notably, Ymt− Y. pestis grows normally in all blood sources during in vitro growth assays, and
can produce septicemic plague in mice [8].
Based on these findings, we propose a nuanced model for the role of Ymt in flea-borne
transmission. When fleas are infected from a host (e.g. mouse) whose RBCs are digested rap-
idly, the bactericidal byproduct generated reaches cytotoxic levels within the first few hours
and eradicates Ymt− Y. pestis from the midgut. This is more pronounced in female fleas, which
take larger blood meals and digest them more rapidly than males. A significantly higher per-
centage of male X. cheopis become infected after feeding on mouse blood containing Ymt− Y.
pestis, and the infection can involve the midgut and the proventriculus. In females, only those
few fleas in which the mutant localizes to the foregut, sequestered from the digestive milieu of
the midgut, remain colonized [4]. In contrast, when fleas are infected from a host (e.g. brown
rat) whose RBCs are digested slowly, we hypothesize that the bactericidal byproduct does not
reach lethal levels before the bacteria have time to coalesce into large dense aggregates. These
aggregates develop in the midgut and proventriculus within a few hours after ingestion and
appear to be surrounded by a viscous matrix [25,26], suggesting that they may be protected
from exposure to bactericidal factors in the midgut. Supporting this idea is the observation
that providing fleas infected using brown rat blood with two maintenance mouse blood meals
2–7 days later did not significantly reduce their high infection rates, which remained compara-
ble to those seen for fleas provided brown rat blood maintenance meals. This model could also
account for the disparity in infection rates between male and female fleas infected with Ymt−
Y. pestis using mouse blood. Our results suggest that the bactericidal agent is produced in the
flea gut regardless of the host blood ingested, but the digestion kinetics of the various blood
sources dictate the frequency and rate at which an absolute lethal concentration is achieved rel-
ative to the time it takes for Y. pestis to coalesce into large dense masses. Although the model
emphasizes RBC digestion kinetics, it is also possible that biochemical differences between
mouse and rat RBCs contribute to the much greater sensitivity of the Ymt− mutant to mouse
RBC digestion.
These results suggest a revision to the evolutionary history of Y. pestis (Fig 6). Ancestral
strains that had not yet acquired ymt, such as those circulating in the Neolithic and Bronze
Age, have been thought to be fully virulent for mammals but incompetent for flea-borne
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
12 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
Fig 6. Model of the adaptive function of Yersinia murine toxin (Ymt) during the evolution of the flea-borne life
cycle of Y. pestis. In this model, ancestral Y. pestis strains lacking the ymt gene (Ymt−; left) could cycle between fleas
and certain species of rodent with flea-colonization-permissive host blood, such as brown rats (Rattus norvegicus), but
not those with non-permissive blood, such as mice (Mus spp). Following acquisition of ymt on the pMT1 plasmid
(Ymt+ strains; right), the progenitor of modern, extant strains of Y. pestis was able to stably colonize fleas that fed on
bacteremic hosts with a blood chemistry that is not permissive for Ymt-negative strains. Thus, acquisition of ymt
effectively greatly expanded the range of mammalian hosts that could support a flea-mammal transmission cycle.
Although male fleas become infected at a moderate rate with Ymt− Y. pestis infected with non-permissive mouse blood
(Fig 3), their potential to transmit is likely not sufficient to maintain a stable transmission cycle (Fig 5).
https://doi.org/10.1371/journal.ppat.1009995.g006
transmission [27–30]. However, our data indicate that they could have been maintained in stable
flea-borne transmission cycles among brown rats and other hosts with similarly permissive blood
characteristics. In addition to early-phase transmission, which does not require Ymt following rat
blood infection [13], transmission by the later, proventricular blockage mechanism would be
robust within these host populations. Acquisition of ymt, however, would have been adaptive for
two reasons. First, it modestly augments flea infectivity even when the bacteremic host blood has
a largely permissive biochemical profile, such as the blood of the brown rat. More consequen-
tially, Ymt enzymatic activity greatly enhances the percentage of fleas that develop a chronic,
transmissible infection when Y. pestis is acquired from a host with blood biochemistry that was
(originally) poorly permissive for flea infection (e.g., mice, humans, black rats) (Fig 6). The ances-
tral Y. pestis lineages that lacked ymt are extinct, suggesting that host restriction and reduced flea
transmissibility of these strains contributed to reduced Darwinian fitness and their eventual dis-
appearance [31]. It’s tempting to hypothesize that rodents involved in ancestral plague transmis-
sion cycles have blood biochemistry similar to that of the brown rat. For example, Tarbagan
marmots (Marmota siberica) have been proposed as host to the original Y. pestis clone, if so, their
blood would likely support chronic flea infection by Ymt− Y. pestis strains [32].
Collectively, our results indicate that acquisition of Ymt did not allow Y. pestis to colonize
fleas per se, but significantly improved Y. pestis survival in the flea gut in the context of RBC
digestion and processing kinetics of blood meals from different mammals. In effect, acquisi-
tion of Ymt greatly expanded the range of hosts that could support a stable mammal-flea trans-
mission cycle (Fig 6). The antibacterial product of RBC digestion and the mechanism of Ymt-
mediated resistance to it remain to be determined. However, this study provides an important
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
13 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
update and revision to Ymt’s adaptive function during the recent evolutionary transition of Y.
pestis to a flea-borne pathogen involving the ecologically broad range of mammals that charac-
terizes modern strains.
Methods
Ethics statement
Experiments involving animals were approved by the Rocky Mountain Laboratories, National
Institute of Allergy and Infectious Diseases, National Institutes of Health Animal Care and
Use Committee (Animal Protocol #2019–011E) and were conducted in accordance with all
National Institutes of Health guidelines.
Bacterial strains and plasmids
Y. pestis strains, plasmids, and primers used in this study are listed in Table 2. Y. pestis KIM6+
ymtH188N expresses a non-functional Ymt with a point mutation in one of the two HKD cata-
lytic domains that typify this class of phospholipase D enzymes [33] and is referred to here as
the Ymt− or Ymt mutant strain. pCH16 (referred to hereafter as pYmt) contains the wild type
ymt gene expressed by its native promoter and was used for complementation of the
ymtH188N mutant [8]. All Y. pestis strains were transformed with pAcGFP1 (Clontech/Takara
Bio) or pGFP-Kmr (this study), respectively. pGFP-Kmr was used for strains complemented
with pCH16 to maintain selection for both plasmids prior to use in flea infections.
Flea infection and host blood
Prior to infection, X. cheopis or O. montana fleas were randomly pulled from colonies estab-
lished at the Rocky Mountain Laboratories and starved for three days. Y. pestis strains were
Table 2. Strain and Plasmid List.
Strain/Plasmid
Y. pestis strains
KIM6+
pCD1-, pMT1+, pPCP1+, Pgm+
Key Properties
KIM6+ymtH188N KIM6+ modified to express a non-functional Ymt with a point mutation in one of the two HKD catalytic
Plasmids
pAcGFP1
pGFP-Kmr
pCH16 (pYmt)
Primers
pGFP-Kmr (Inverse
PCR)
pGFP-Kmr (Kmr
Casette)
domains.
Apr, constitutively expresses GFP
pAcGFP1 was amplified by inverse PCR to selectively exclude the bla gene and replace it with a SacI site. A
kanamycin resistance cassette with terminal SacI sites was inserted into the linear inverse PCR product of
pAcGFP1, which was then religated.
Apr, expresses ymt from its native promoter.
F: CGTCGAGCTCTTCGTTCCACTGAGCGTCA
R: CGTAGAGCTCGTACAATCTGCTCTGATGCCG
F: CGTAGAGCTCTCCAGCCAGAAAGTGAGGGAG
R: GCATGAGCTCGGGAAAGCCACGTTGTGTCTC
(Amplified from pKD4; [36])
pCD1 virulence plasmid, encodes type 3 secretion system
pMT1 plasmid, encodes the phospholipase D Yersinia murine toxin (Ymt) and capsule antigen (F1)
pPCP1 plasmid, encodes plasminogen activator/protease (Pla), the bacteriocin pesticin (Pst), and pesticin immunity protein (Pim)
Pgm pigmentation locus and pathogenicity island, encodes the hemin storage locus (hmsHFRS operon) and iron acquisition genes
Apr ampicillin resistance; Kmr kanamycin resistance
https://doi.org/10.1371/journal.ppat.1009995.t002
Reference
[34,35]
[33]
Clontech/Takara Bio
(Mountain View, CA)
This Study
[8]
This Study
This Study
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
14 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
grown in brain-heart infusion (BHI) broth with appropriate antibiotic selection as described
previously [37]. Briefly, 100 ml Y. pestis cultures, grown at 37˚C overnight, were centrifuged
and the bacterial pellet resuspended in 1 ml sterile PBS. Bacterial suspensions were added to a
final concentration of ~5x108 CFU/ml to 5 ml of heparinized Swiss-Webster mouse blood, defi-
brinated Sprague-Dawley brown rat (R. norvegicus) or human blood (both from BioIVT, New
York), or to heparinized wild black rat (R. rattus) blood collected and shipped overnight by Ala-
meda County, CA, Vector Control personnel. Prior to use in flea infections, black rat blood was
treated with carbenicillin (100 μg/ml) and plated on 5% sheep blood agar to ensure sterility. The
blood and bacterial mixture was added to a membrane feeding apparatus and groups of fleas
were allowed to feed for 1 h [4]. Fleas (approximately equal numbers of males and females) that
took an infectious blood meal were collected and kept at 21˚C in a humidified chamber (75%
RH). These fleas were provided maintenance feedings on neonatal mice 2 to 3 days after infec-
tion and again 6 to 7 days after infection. Following each maintenance feed, fleas were screened
for the presence of fresh red blood in the esophagus, a condition of fleas with partial or complete
blockage or PIER. At 0, 1, and 7 days following infection, 10 to 20 infected fleas were frozen at
-80˚C for later determination of infection status and bacterial load per flea by plating individual
triturated fleas in BHI soft agar overlays as previously described [38].
For flea infections using plasma-swapped blood, the plasma fraction was separated from
mouse or rat red blood cells (RBC) following centrifugation at 3000 rpm, the RBCs were
washed 3 times with an equivalent volume of sterile PBS, and whole blood was reconstituted
with heterologous plasma from the other rodent.
Dissection and imaging of flea digestive tracts
Fleas infected with KIM6+ymtH188N suspended in mouse or brown rat blood were dissected
one day after infection to determine the localization of bacteria in the digestive tract and their
phenotype. The severity of proventricular infection was scored as light, moderate, or heavy as
described previously [38]. Images of flea digestive tracts and bacterial biofilms were taken with
a Nikon Eclipse E800 microscope equipped with a DP72 Olympus camera (Center Valley, PA)
and a G-2E/C (540/25 EX) fluorescent filter (Nikon), and were processed using Olympus cell-
Sens software.
Blood meal volume and red blood cell digestion rate
For blood meal volume determination, individual adult X. cheopis not fed for 5 days prior,
were weighed using a Sartorius SC 2 Microbalance (Goettingen, Germany) before and imme-
diately after feeding on a neonatal mouse. Fleas were anesthetized with CO2 and placed in a
microcentrifuge tube prior to each weighing. Bloodmeal weight was determined by subtracting
the pre-feed weight from the post-feed weight and then converted to volume based on the spe-
cific gravity of mouse blood [39].
To assess digestion kinetics, X. cheopis were allowed to feed on sterile Swiss Webster mouse
or Sprague-Dawley rat blood for 30 minutes. Digestive tracts were dissected from groups of
male or female fleas immediately (0.5 h) following feeding or 2 h after feeding. Excised diges-
tive tracts were placed in 20 μl sterile PBS on a microscope slide and expressed with forceps to
release midgut contents. Expelled gut contents were diluted 1:5 in PBS containing 0.4% (w/v)
trypan blue, mixed, and the number of RBCs determined using a hemocytometer.
For the digestion image series, fleas were fed as described above and digestive tracts were
imaged using a Nikon SMZ1500 dissection microscope with a DP72 Olympus camera. Diges-
tive tracts were visually scored for the presence or absence of cellular material to determine if
they had completely liquified the blood meal.
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
15 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
Transmission assays
Transmission by O. montana or X. cheopis fleas was assayed as described previously with
minor modifications [40]. Fleas infected as described above were refed using the artificial feed-
ing system every 3 days following infection on sterile host blood of the same source they were
initially infected with. After each feed (except for the X. cheopis-mouse blood experiment in
which transmission was assessed every 6 days following the initial assay on day 3) the entirety
of the blood from the feeding reservoir was collected and distributively spread onto blood
agar-carbenicillin plates. In addition, the feeding reservoir was washed 6 times with 5 ml of
sterile PBS; these washes were combined, centrifuged, and the resulting pellet resuspended in 2
to 3 ml of PBS and plated. Blood agar plates were incubated for 48 h at 28˚C and GFP positive
colonies counted to determine the number of CFUs transmitted.
Bacterial in vitro growth and susceptibility assays
Y. pestis strains were grown in BHI containing 10 μg/ml hemin. After 18 h incubation at 28˚C
without shaking, cultures were diluted to an OD600 of 0.1, centrifuged at 6000 rpm for 10
min, and the bacterial pellets resuspended in an equal volume of sterile PBS. Y. pestis was then
added to 10 ml of BHI-hemin, defibrinated brown rat blood, or defibrinated mouse blood to a
final concentration of ~1x106 CFU/ml and cultures were incubated in 50 ml conical tubes at
either 21˚ or 37˚C without shaking. After 0, 2, 4, 6, 8, and 24 h incubation the cultures were
mixed well and a 100 μl sample was removed, serially diluted, and plated on blood agar for
CFU determination.
For susceptibility assays, ~1x106 Y. pestis, prepared as above, were added to 1 ml of defibrin-
ated mouse plasma, a suspension of lysed mouse RBCs, or BHI broth in an 8-well culture dish.
Lysed mouse RBCs were prepared by mixing washed cells 1:1 with sterile PBS followed by
three freeze-thaw cycles. After 1 h at 25˚C, 10-fold serial dilutions of the suspensions were
plated on blood agar. The percentage of CFU recovered from each medium relative to the BHI
control was calculated to assess antibacterial activity.
Supporting information
S1 Fig. The Y. pestis Ymt mutant phenotype in female Oropsylla montana fleas is identical
to that in X. cheopis. Groups of O. montana fleas that fed on mouse or brown rat blood con-
taining 2.8 x 108–7.1 x 108 CFU/ml Y. pestis KIM6+, KIM6+ymtH188N, or KIM6+ymtH188N
(pYmt) were screened for 1 week for A) the percentage of fleas that remained infected; B)
development of a foregut obstruction that interfered with normal blood-feeding; and C) bacte-
rial burden. Data are the results from 3 (KIM6+ymtH188N groups) or 1–2 (KIM6+ and KIM6
+ymtH188N(pYmt) groups) independent experiments. Samples consisted of 9–20 female fleas
(A, C) or 40 to 112 fleas (approximately equal numbers of males and females; B) per experi-
ment. The mean and standard error (A, B) or median (C) are indicated. �p <0.05 by chi-
square (A, B) or by Kruskal-Wallis test with Dunn’s post-test (C).
(TIF)
S2 Fig. The Ymt mutant has no growth or survival defects under in vitro conditions.
Growth kinetics of Y. pestis KIM6+ and KIM6+ymtH188N grown in mouse blood, brown rat
blood, or BHI broth supplemented with hemin and incubated for 24 h at A) 21˚C or B) 37˚C.
The mean and standard error of 3 independent experiments are shown. C) Bacterial survival
assay in which 1x106 CFU KIM6+ymtH188N were added to BHI broth, defibrinated mouse
plasma, lysed mouse red blood cells and incubated for 1 h at 25˚C. Dilutions of each medium
were then plated to determine CFU concentrations. The mean and standard error of 3
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
16 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
independent experiments are shown and expressed as the percent CFU recovered relative to
the BHI control.
(TIF)
S3 Fig. X. cheopis blood source- and sex-related differences in red blood cell digestion
rates. Representative image series of the data shown in Fig 4D. Digestive tract preparations
were scored for the presence or absence of particulates that exuded from the flea midgut into
the surrounding saline. Mouse blood meals were completely liquified by most female fleas in
4–6 h (far left) but partially digested RBC stroma were still present in most males for 6–8 h
(middle left). With rare exception, fleas that ingested sterile rat blood, regardless of sex, con-
tained a fairly stable amount of partially digested RBCs for at least 8 h following feeding
(right). Scale bar = 100 μm.
(TIF)
S1 Data. File containing numerical data used for Figs 1–5, S1 and S2.
(XLSX)
Acknowledgments
We thank David K. James and colleagues at Alameda County Vector Control (Alameda, CA)
for generously collecting black rat blood; Ryan Kissinger for assistance with graphic design;
and Clayton Jarrett, Jeff Shannon, and Phil Stewart for critical review of the manuscript.
Author Contributions
Conceptualization: David M. Bland, B. Joseph Hinnebusch.
Data curation: David M. Bland.
Formal analysis: David M. Bland, Ade´laïde Miarinjara, B. Joseph Hinnebusch.
Funding acquisition: B. Joseph Hinnebusch.
Investigation: David M. Bland, Ade´laïde Miarinjara, Christopher F. Bosio, Jeanette Calarco,
B. Joseph Hinnebusch.
Methodology: David M. Bland, Ade´laïde Miarinjara, Christopher F. Bosio, B. Joseph
Hinnebusch.
Project administration: B. Joseph Hinnebusch.
Resources: B. Joseph Hinnebusch.
Supervision: B. Joseph Hinnebusch.
Validation: David M. Bland, B. Joseph Hinnebusch.
Visualization: David M. Bland, B. Joseph Hinnebusch.
Writing – original draft: David M. Bland.
Writing – review & editing: David M. Bland, Ade´laïde Miarinjara, Christopher F. Bosio, Jean-
ette Calarco, B. Joseph Hinnebusch.
References
1. Achtman M, Zurth K, Morelli G, Torrea G, Guiyoule A, Carniel E. Yersinia pestis, the cause of plague, is
a recently emerged clone of Yersinia pseudotuberculosis. Proc. Natl. Acad. Sci. USA. 1999; 96
(24):14043–8. https://doi.org/10.1073/pnas.96.24.14043 PMID: 10570195.
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
17 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
2. Cui Y, Yu C, Yan Y, Li D, Li Y, Jombart T, et al. Historical variations in mutation rate in an epidemic path-
ogen, Yersinia pestis. Proc. Natl. Acad. Sci. USA. 2013; 110(2):577–82. https://doi.org/10.1073/pnas.
1205750110 PMID: 23271803.
3. Sun YC, Jarrett CO, Bosio CF, Hinnebusch BJ. Retracing the evolutionary path that led to flea-borne
transmission of Yersinia pestis. Cell Host Microbe. 2014; 15(5):578–86. https://doi.org/10.1016/j.chom.
2014.04.003 PMID: 24832452.
4. Hinnebusch BJ, Rudolph AE, Cherepanov P, Dixon JE, Schwan TG, Forsberg A. Role of Yersinia
murine toxin in survival of Yersinia pestis in the midgut of the flea vector. Science. 2002; 296
(5568):733–5. https://doi.org/10.1126/science.1069972 PMID: 11976454.
5.
Lindler LE, Plano GV, Burland V, Mayhew GF, Blattner FR. Complete DNA sequence and detailed anal-
ysis of the Yersinia pestis KIM5 plasmid encoding murine toxin and capsular antigen. Infect. Immun.
1998; 66(12):5731–42. https://doi.org/10.1128/IAI.66.12.5731-5742.1998 PMID: 9826348.
6. Hildebrand GJ, Ng J, Von Metz EK, Eisler DM. Studies on the mechanism of circulatory failure induced
in rats by Pasteurella pestis murine toxin. J. Infect. Dis. 1966; 116(5):615–29. https://doi.org/10.1093/
infdis/116.5.615 PMID: 5957268.
7. Ajl SJ, Reedal JS, Durrum EL, Warren J. Studies on plague. I. Purification and properties of the toxin of
Pasteurella pestis. J. Bacteriol. 1955; 70(2):158–69. https://doi.org/10.1128/jb.70.2.158-169.1955
PMID: 13251981
8. Hinnebusch BJ, Cherepanov P, Du Y, Rudolph A, Dixon JD, Schwan T, et al. Murine toxin of Yersinia
pestis shows phospholipase D activity but is not required for virulence in mice. Int. J. Med. Microbiol.
2000; 290(4–5):483–7. https://doi.org/10.1016/S1438-4221(00)80070-3 PMID: 11111930.
9. Brown SD, Montie TC. Beta-adrenergic blocking activity of Yersinia pestis murine toxin. Infect. Immun.
1977; 18(1):85–93. https://doi.org/10.1128/iai.18.1.85-93.1977 PMID: 198377.
10.
Fan Y, Zhou Y, Feng N, Wang Q, Tian G, Wu X, et al. Recombinant murine toxin from Yersinia pestis
shows high toxicity and beta-adrenergic blocking activity in mice. Microbes Infect. 2016; 18(5):329–35.
https://doi.org/10.1016/j.micinf.2016.01.001 PMID: 26774329.
11. Montie TC, Ajl SJ. Microbial Toxins. Montie TC, Kadis S, Ajl SJ, editors. New York: Academic Press;
1970. 1–37 p.
12.
13.
Falkow S. Molecular Koch’s postulates applied to microbial pathogenicity. Rev. Infect. Dis. 1988; 10
Suppl 2:S274–6. https://doi.org/10.1093/cid/10.supplement_2.s274 PMID: 3055197.
Johnson TL, Hinnebusch BJ, Boegler KA, Graham CB, MacMillan K, Montenieri JA, et al. Yersinia
murine toxin is not required for early-phase transmission of Yersinia pestis by Oropsylla montana
(Siphonaptera: Ceratophyllidae) or Xenopsylla cheopis (Siphonaptera: Pulicidae). Microbiology. 2014;
160(Pt 11):2517–25. https://doi.org/10.1099/mic.0.082123-0 PMID: 25187626.
14. Bland DM, Jarrett CO, Bosio CF, Hinnebusch BJ. Infectious blood source alters early foregut infection
and regurgitative transmission of Yersinia pestis by rodent fleas. PLoS Path. 2018; 14(1):e1006859.
https://doi.org/10.1371/journal.ppat.1006859 PMID: 29357385.
15. Hinnebusch BJ, Chouikha I, Sun YC. Ecological opportunity, evolution, and the emergence of flea-
borne plague. Infect. Immun. 2016; 84(7):1932–40. https://doi.org/10.1128/IAI.00188-16 PMID:
27160296.
16. Vaughan JA, Azad AF. Patterns of erythrocyte digestion by bloodsucking insects: constraints on vector
competence. J. Med. Entomol. 1993; 30(1):214–6. https://doi.org/10.1093/jmedent/30.1.214 PMID:
8094460.
17. Bland DM, Martens CA, Virtaneva K, Kanakabandi K, Long D, Rosenke R, et al. Transcriptomic profiling
of the digestive tract of the rat flea, Xenopsylla cheopis, following blood feeding and infection with Yersi-
nia pestis. PLoS Negl. Trop. Dis. 2020; 14(9):e0008688. https://doi.org/10.1371/journal.pntd.0008688
PMID: 32946437.
18. Reinhardt CA. Ultrastructural comparison of the midgut epithelia of fleas with different feeding behavior
patterns (Xenopsylla cheopis, Echidnophaga gallinacea, Tunga penetrans, Siphonaptera, Pulicidae).
Act. Trop. 1976; 33(2):105–32. 8972.
19. Sarfati M, Krasnov BR, Ghazaryan L, Khokhlova IS, Fielden LJ, Degen AA. Energy costs of blood diges-
tion in a host-specific haematophagous parasite. J. Exp. Biol. 2005; 208(Pt 13):2489–96. https://doi.
org/10.1242/jeb.01676 PMID: 15961735.
20.
Lehane MJ. The Biology of Blood Sucking Insects. 2nd ed. Cambridge: Cambridge University Press;
2005.
21. Vashchenok VS, Solina LT, Zhirnov AE. Digestion of blood of different animals by fleas Xenopsylla
cheopis [In Russian]. Parazitologiya. 1976; 10(6):544–9. PMID: 1023147.
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
18 / 19
PLOS PATHOGENS Protective effect of Yersinia murine toxin in the flea gut is dependent on source of host blood
22. Earl SC, Rogers MT, Keen J, Bland DM, Houppert AS, Miller C, et al. Resistance to innate immunity
contributes to colonization of the insect gut by Yersinia pestis. PloS One. 2015; 10(7):e0133318. https://
doi.org/10.1371/journal.pone.0133318 PMID: 26177454.
23.
Ludlow AB, Pauling CD, Marketon MM, Anderson DM. Drosophila as a model for understanding the
insect host of Yersinia pestis. Methods Mol. Biol. 2019; 2010:167–78. https://doi.org/10.1007/978-1-
4939-9541-7_12 PMID: 31177438.
24. Hinnebusch BJ. The evolution of flea-borne transmission in Yersinia pestis. Curr. Issues Mol. Biol.
2005; 7(2):197–212. PMID: 16053250.
25. Hinnebusch BJ, Jarrett CO, Bland DM. "Fleaing" the plague: adaptations of Yersinia pestis to its insect
vector that lead to transmission. Annu. Rev. Microbiol. 2017; 71:215–32. https://doi.org/10.1146/
annurev-micro-090816-093521 PMID: 28886687.
26. Dewitte A, Bouvenot T, Pierre F, Ricard I, Pradel E, Barois N, et al. A refined model of how Yersinia pes-
tis produces a transmissible infection in its flea vector. PLoS Path. 2020; 16(4):e1008440. https://doi.
org/10.1371/journal.ppat.1008440 PMID: 32294143.
27. Rasmussen S, Allentoft ME, Nielsen K, Orlando L, Sikora M, Sjogren KG, et al. Early divergent strains
of Yersinia pestis in Eurasia 5,000 years ago. Cell. 2015; 163(3):571–82. https://doi.org/10.1016/j.cell.
2015.10.009 PMID: 26496604.
28. Rascovan N, Sjogren KG, Kristiansen K, Nielsen R, Willerslev E, Desnues C, et al. Emergence and
spread of basal lineages of Yersinia pestis during the Neolithic decline. Cell. 2019; 176(1–2):295–305
e10. https://doi.org/10.1016/j.cell.2018.11.005 PMID: 30528431.
29. Spyrou MA, Tukhbatova RI, Wang CC, Valtuena AA, Lankapalli AK, Kondrashin VV, et al. Analysis of
3800-year-old Yersinia pestis genomes suggests Bronze Age origin for bubonic plague. Nat. Commun.
2018; 9(1):2234. https://doi.org/10.1038/s41467-018-04550-9 PMID: 29884871.
30. Andrades Valtueña A, Mittnik A, Key FM, Haak W, Allmae R, Belinskij A, et al. The Stone Age plague
and its persistence in Eurasia. Curr. Biol. 2017; 27(23):3683–91 e8. https://doi.org/10.1016/j.cub.2017.
10.025 PMID: 29174893.
31. Demeure C, Dussurget O, Fiol GM, Le Guern AS, Savin C, Pizarro-Cerda J. Yersinia pestis and plague:
an updated view on evolution, virulence determinants, immune subversion, vaccination and diagnos-
tics. Microbes. Infect. 2019; 21(5–6):202–12. https://doi.org/10.1016/j.micinf.2019.06.007 PMID:
31252217.
32. Suntsov VV. Sympatric speciation of the plague microbe Yersinia pestis: monohostal specialization in
the host-parasite marmot-flea (Marmota sibirica-Oropsylla silantiewi) system. Izv. Akad. Nauk. Ser.
Biol. 2016;(2):117–27. PMID: 27396172.
33. Rudolph AE, Stuckey JA, Zhao Y, Matthews HR, Patton WA, Moss J, et al. Expression, characteriza-
tion, and mutagenesis of the Yersinia pestis murine toxin, a phospholipase D superfamily member. J.
Biol. Chem. 1999; 274(17):11824–31. https://doi.org/10.1074/jbc.274.17.11824 PMID: 10207000.
34. Perry RD, Fetherston JD. Yersinia pestis—etiologic agent of plague. Clin. Microbiol. Rev. 1997; 10
(1):35–66. https://doi.org/10.1128/CMR.10.1.35 PMID: 8993858.
35. Sikkema DJ, Brubaker RR. Resistance to pesticin, storage of iron, and invasion of HeLa cells by Yersi-
niae. Infect. Immun. 1987; 55(3):572–8. https://doi.org/10.1128/iai.55.3.572-578.1987 PMID: 3818085.
36. Datsenko KA, Wanner BL. One-step inactivation of chromosomal genes in Escherichia coli K-12 using
PCR products. Proc. Natl. Acad. Sci. USA. 2000; 97(12):6640–5. https://doi.org/10.1073/pnas.
120163297 PMID: 10829079.
37. Bland DM, Brown LD, Jarrett CO, Hinnebusch BJ, Macaluso KR. Methods in Flea Research. 2017.
https://www.beiresources.org/Catalog/VectorResources.aspx
38. Bland DM, Hinnebusch BJ. Feeding behavior modulates biofilm-mediated transmission of Yersinia pes-
tis by the cat flea, Ctenocephalides felis. PLoS Negl. Trop. Dis. 2016; 10(2):e0004413. https://doi.org/
10.1371/journal.pntd.0004413 PMID: 26829486.
39. Altman PL. Blood and Other Body Fluids. 1 ed. Washington, D.C.: Federation of American Societies
for Experimental Biology; 1961. 15 p.
40. Hinnebusch BJ, Bland DM, Bosio CF, Jarrett CO. Comparative ability of Oropsylla montana and Xenop-
sylla cheopis fleas to transmit Yersinia pestis by two different mechanisms. PLoS Negl. Trop. Dis. 2017;
11(1):e0005276. https://doi.org/10.1371/journal.pntd.0005276 PMID: 28081130.
PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1009995 October 14, 2021
19 / 19
PLOS PATHOGENS
| null |
10.1038_s41467-023-39627-7.pdf
|
Data availability
The sequencing data generated in this study have been deposited in
the NCBI Gene Expression Omnibus (GEO) database under accession
code GSE192447. The publicly available whole blood bulk short-read
RNA-Seq data from healthy samples and samples infected with Makona
Ebola Virus data used in this study are available in the NCBI Gene
Expression Omnibus (GEO) database under accession code GSE115785.
The single-cell RNA-Seq data used in this study are available in the NCBI
Gene Expression Omnibus (GEO) database under accession code
GSE158390. Raw Seurat Objects for both single-cell datasets used in
this study are available at Zenodo. The full co-expression network file is
also provided (https://doi.org/10.5281/zenodo.7997135). The refer-
ence genome of EBOV used in this study is available in the GenBank
database under accession code KU182905.1. The assembly and refer-
ence genome of Macaca Mulatta used in this study are available in the
Ensembl database (Mmul_10) (https://ftp.ensembl.org/pub/release-100/
fasta/macaca_mulatta/dna/Macaca_mulatta.Mmul_10.dna.toplevel.fa.gz,
https://ftp.ensembl.org/pub/release-100/gtf/macaca_mulatta/Macaca_
mulatta.Mmul_10.100.gtf.gz). The assembly and reference genome of
human used in this study are available in the Gencode database
(https://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_
(release_23)
human/release_23/gencode.v23.annotation.gtf.gz, https://ftp.ebi.ac.uk/
pub/databases/gencode/Gencode_human/release_23/GRCh38.primary_
assembly.genome.fa.gz). Source data are provided with this paper.
Code availability
The code used for this study is available at: https://github.com/Mele-
Lab/2023_SingleCellEbolaLncRNAs_NatComms.
|
Data availability The sequencing data generated in this study have been deposited in the NCBI Gene Expression Omnibus (GEO) database under accession code GSE192447 . The publicly available whole blood bulk short-read RNA-Seq data from healthy samples and samples infected with Makona Ebola Virus data used in this study are available in the NCBI Gene Expression Omnibus (GEO) database under accession code GSE115785 . The single-cell RNA-Seq data used in this study are available in the NCBI Gene Expression Omnibus (GEO) database under accession code GSE158390 . Raw Seurat Objects for both single-cell datasets used in this study are available at Zenodo. The full co-expression network file is also provided ( https://doi.org/10.5281/zenodo.7997135 ). The reference genome of EBOV used in this study is available in the GenBank database under accession code KU182905.1 . The assembly and reference genome of Macaca Mulatta used in this study are available in the Ensembl database (Mmul_10) ( https://ftp.ensembl.org/pub/release-100/ fasta/macaca_mulatta/dna/Macaca_mulatta.Mmul_10.dna.toplevel.fa.gz , https://ftp.ensembl.org/pub/release-100/gtf/macaca_mulatta/Macaca_ mulatta.Mmul_10.100.gtf.gz ). The assembly and reference genome of human used in this study are available in the Gencode database (release_23) ( https://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_ human/release_23/gencode.v23.annotation.gtf.gz , https://ftp.ebi.ac.uk/ pub/databases/gencode/Gencode_human/release_23/GRCh38.primary_ assembly.genome.fa.gz ). Source data are provided with this paper. Code availability The code used for this study is available at: https://github.com/Mele- Lab/2023_SingleCellEbolaLncRNAs_NatComms .
|
Article
https://doi.org/10.1038/s41467-023-39627-7
Single-cell profiling of lncRNA expression
during Ebola virus infection in rhesus
macaques
Received: 25 August 2022
Accepted: 19 June 2023
Check for updates
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
1,2,13, Maria Sopena-Rios1,13, Raquel García-Pérez1,14,
Luisa Santus
Aaron E. Lin 3,4,5,14, Gordon C. Adams
Katherine J. Siddle 3,4, Shirlee Wohl3,4,8, Ferran Reverter
Richard S. Bennett11, Lisa E. Hensley11
Marta Melé 1
, Pardis C. Sabeti
3,4, Kayla G. Barnes
4,6,7,
9, John L. Rinn 10,
3,4,5,12 &
Long non-coding RNAs (lncRNAs) are involved in numerous biological pro-
cesses and are pivotal mediators of the immune response, yet little is known
about their properties at the single-cell level. Here, we generate a multi-tissue
bulk RNAseq dataset from Ebola virus (EBOV) infected and not-infected rhesus
macaques and identified 3979 novel lncRNAs. To profile lncRNA expression
dynamics in immune circulating single-cells during EBOV infection, we design
a metric, Upsilon, to estimate cell-type specificity. Our analysis reveals that
lncRNAs are expressed in fewer cells than protein-coding genes, but they are
not expressed at lower levels nor are they more cell-type specific when
expressed in the same number of cells. In addition, we observe that lncRNAs
exhibit similar changes in expression patterns to those of protein-coding
genes during EBOV infection, and are often co-expressed with known immune
regulators. A few lncRNAs change expression specifically upon EBOV entry in
the cell. This study sheds light on the differential features of lncRNAs and
protein-coding genes and paves the way for future single-cell lncRNA studies.
Long non-coding RNAs (lncRNAs) are transcripts longer than 200 bp
that lack protein-coding potential. LncRNAs play important roles in a
myriad of processes, such as development1, evolutionary innovation2,
and disease3. LncRNAs often regulate gene expression by acting as
signaling molecules4–6, decoys7, molecular guides8, or
through
scaffolding9. Importantly, many lncRNAs are important host immune
response regulators10,11. Specifically, they regulate the maturation and
development of lymphoid and myeloid cells12, mediate pathogen-
induced monocyte and macrophage activation, and the subsequent
release of inflammatory factors such as cytokines and chemokines11,13,14.
Despite lncRNAs sharing similar biogenesis with protein-coding
genes15,16, they are distinguishable by a variety of features, such as
1Life Sciences Department, Barcelona Supercomputing Center, Barcelona, Catalonia 08034, Spain. 2Centre for Genomic Regulation (CRG), The Barcelona
Institute for Science and Technology, Barcelona, Spain. 3FAS Center for Systems Biology, Department of Organismic and Evolutionary Biology, Harvard
University, Cambridge, MA 02138, USA. 4Broad Institute of MIT and Harvard, Cambridge, MA 02142, USA. 5Harvard Program in Virology, Harvard Medical
School, Boston, MA 02115, USA. 6Department of Immunology and Infectious Diseases, Harvard T.H. Chan School of Public Health, Harvard University, Boston,
MA 02115, USA. 7Liverpool School of Tropical Medicine, Liverpool L3 5QA, UK. 8The Scripps Research Institute, Department of Immunology and Microbiology,
La Jolla, CA, USA. 9Department of Genetics, Microbiology and Statistics University of Barcelona, Barcelona, Spain. 10Department of Biochemistry, University of
Colorado Boulder, Boulder 80303, USA. 11Integrated Research Facility, Division of Clinical Research, National Institute of Allergy and Infectious Diseases,
National Institutes of Health, Frederick, MD 21702, USA. 12Howard Hughes Medical Institute, Chevy Chase, MD 20815, USA. 13These authors contributed
equally: Luisa Santus, Maria Sopena-Rios. 14These authors jointly supervised this work: Raquel García-Pérez, Aaron E. Lin.
[email protected]; [email protected]
e-mail: [email protected];
Nature Communications |
(2023) 14:3866
1
Article
lower expression levels16–18, higher tissue specificity16,17,19,20, lower spli-
cing efficiency21,22, and differences in their promoter regulation23.
However, most of these observations arise from bulk tissue analyses;
therefore, whether their expression kinetics are driven by overall low
expression levels across many cells or by high expression levels in
specific cell populations remains unclear. This lack of knowledge at
single-cell resolution hampers our understanding of how lncRNAs
function and whether their regulation and response upon infection are
intrinsically different from that of protein-coding genes.
EBOV is one of the most lethal pathogens to humans, and it is
infamously notorious for its high infectiousness and severe case
fatality rates24,25. In the past, EBOV caused alarming outbreaks; up to
the present day, it represents a major global health threat26. Previously,
bulk tissue transcriptomic analyses improved our understanding of
EBOV’s evoked host immune response27,28. Now, emerging single-cell
RNA-sequencing (scRNA-Seq) technologies are refining our under-
standing of the systemic immune response mounted upon viral
infections28–30 by allowing the dissection of gene expression dynamics
in multiple cell populations simultaneously. More importantly, in the
case of organisms infected with a virus, scRNA-Seq can identify and
profile infected cells separately from uninfected bystander cells and
thus, distinguish the host cellular transcriptional response triggered by
viral replication versus the inflammatory cytokine milieu. However,
previous studies have focused on the host protein-coding gene
response and have ignored the role that non-coding genes such as
lncRNAs may play in the host response to EBOV infection. This is
mostly due to poor lncRNA annotations in non-human primates, the
main species of EBOV research.
In this work, we generate multi-tissue bulk RNAseq data from
EBOV-infected and not-infected rhesus macaque tissues to expand the
lncRNA annotation in this model organism. We then study circulating
immune single-cells infected with EBOV in vivo to address the question
of how lncRNAs differentially respond to viral infection at single-cell
resolution compared to protein-coding genes. Our results question the
long-assumed differences between lncRNA and protein-coding genes
and identify lncRNAs involved in the transcriptional response elicited
upon EBOV infection.
Results
De novo annotation largely expands the rhesus macaque
non-coding transcriptome
Bulk and single-cell transcriptomic studies in rhesus macaque have
reported widespread host gene expression changes upon EBOV
infection30–32. However, most
lncRNAs have been systematically
neglected in such studies due to incomplete annotations, especially in
rhesus macaque, where the number of annotated lncRNAs is only 28% of
that in humans (Supplementary Fig. 1A). To improve the current lncRNA
annotation, we generated short-read RNA-sequencing data from 13 tis-
sues (Fig. 1a) of not infected (16 samples) and EBOV-infected (43 sam-
ples) macaques. We further combined this data with publicly available
blood RNA-sequencing of not infected (21 samples) and EBOV infected
(39 samples) macaques33, adding up to a total of 119 samples and almost
4 billion reads (Supplementary Data 1). To identify novel lncRNAs, we
implemented a computational pipeline that performs de novo tran-
scriptome assembly, extensive quality controls, and non-coding tran-
script selection based on concordance between three different tools
(Fig. 1b, Supplementary Fig. 1B) (see “Methods”). Our approach had high
accuracy (82%) and specificity (86%) when predicting Ensembl anno-
tated macaque lncRNAs (Supplementary Fig. 1C). In total, we discovered
3979 novel lncRNA genes (5299 transcripts) (Fig. 1b, c), of which 3191
(80%) were intergenic and 788 (20%) were antisense. Consistent with
previous work34, we identified a human lncRNA ortholog for a relatively
low number of lncRNAs (528 lncRNAs (14%)) (Supplementary Fig. 1D).
Novel and annotated lncRNA transcripts were shorter, with longer and
fewer exons compared to protein-coding genes (Mann–Whitney U test,
https://doi.org/10.1038/s41467-023-39627-7
all P-values < 2.2 × 10−16) (Fig. 1d, e). We also observed differences in
intron length (Supplementary Fig. 2A). All these observations hold true
when we analyze intergenic and antisense lncRNAs separately (Sup-
plementary Fig. 2B–E).
In line with previous studies in bulk
samples17,19,20,35, both annotated and novel lncRNAs had lower expres-
(cid:1)16) and were
sion levels (Mann–Whitney U test, all P-values < 2.2 × 10
expressed in fewer tissues (two-sided Kolmogorov–Smirnov test,
P-values < 2.2 × 10−16) compared to protein-coding genes (Supplemen-
tary Fig. 3A, B).
genes
To further assess the expression profile of lncRNAs, we calculated
Tau tissue-specificity scores36. Tau is a widely-used metric that mea-
sures the level of tissue-specific expression of a gene. It ranges from 0
for housekeeping genes to 1 for tissue-specific genes. As expected21,22,
lncRNAs were more tissue-specific than protein-coding genes
(cid:1)16) (Fig. 1f). We used Tau to
(Mann–Whitney U test, P-values < 2.2 × 10
classify
intermediate
(Tau > 0.7),
into
(0.3 ≤ Tau ≤ 0.7), and ubiquitous (Tau < 0.3) (Fig. 1g). We found a total
of 5203 tissue-specific lncRNAs from which 2429 were novel and 2774
were annotated (Fig. 1g, Supplementary Fig. 3C). Then, for each
lncRNA, we identified the tissue in which it presented the highest
average expression (see “Methods”). Within such tissues, ubiquitous
novel and annotated lncRNAs had similar average expression levels,
whereas novel tissue-specific and intermediate lncRNAs were more
expressed than annotated lncRNAs (Fig. 1h).
tissue-specific
In summary, using de novo bulk sequencing of multi-tissue not
infected and EBOV-infected samples, we identified lncRNAs that
resemble lncRNA reference annotation and double the current lncRNA
rhesus macaque gene annotation.
LncRNAs are systematically expressed in fewer cells compared
to protein-coding genes
Bulk tissue studies have established that lncRNAs are more lowly
expressed, more tissue-specific, and often have a more time and context-
dependent expression compared to protein-coding genes16,17,19,20,22.
However, whether this signal arises from lncRNAs being lowly expressed
across individual cells or from their expression being restricted to only a
few cells remains elusive37. To address this, we used single-cell tran-
scriptomics data from macaque’s peripheral blood mononuclear cells
(PBMCs) from Kotliar et al.30,38. After quality control (see “Methods”), we
selected 38,067 cells and classified them into four major cell types:
monocytes, neutrophils, B cells, and T cells (Fig. 2a, Supplementary
Fig. 4A, B). Whereas lncRNAs were slightly less expressed on average
than protein-coding genes (Mann–Whitney U test, P-value = 0.017)
(Fig. 2b), differences in the number of cells in which they were expressed
were much larger with lncRNAs being expressed in fewer cells
(cid:1)16) (Fig. 2c–e). In addition,
(Mann–Whitney U test, P-value < 2.2 × 10
lncRNAs are consistently expressed in a lower proportion of cells than
protein-coding genes when we inspected the different cell types sepa-
(cid:1)14) (Supplementary
rately (Mann–Whitney U test, all P-values < 4 × 10
Fig. 4C). The proportion of cells expressing a gene and its gene
expression levels are tightly correlated (Supplementary Fig. 4D). Thus,
we tested whether lncRNA expression levels were lower than those of
protein-coding genes when expressed in a comparable number of cells.
We found no significant differences in the expression levels of lncRNAs
and protein-coding genes when they were matched by the proportion of
cells in which they were expressed (one-side Wilcoxon signed-rank test,
P-value > 0.05) (Fig. 2f). Conversely, lncRNAs were expressed in fewer
cells compared to protein-coding genes when controlling for median
expression levels (one-side Wilcoxon signed-rank test, P-value < 2.2 ×
(cid:1)16) (Fig. 2g). These results indicate that a main distinctive feature of
10
lncRNAs is the low number of cells they are expressed in. We wanted to
see if we could reproduce these results in humans, where lncRNA
annotation is more complete, and by using an independent platform
such as 10X Genomics which has a higher yield than Seq-Well39. We used
Nature Communications |
(2023) 14:3866
2
Article
a
https://doi.org/10.1038/s41467-023-39627-7
b
c
lncRNA
macaque
4,769
3,979
annotated novel
human
16,887
10,000
# of genes
g
lncRNA
protein-coding
d
6,000
)
p
b
(
h
t
g
n
e
l
t
p
i
r
c
s
n
a
r
t
4,000
2,000
0
21,591
19,970
20,000
e
e
)
p
b
(
h
t
g
n
e
l
n
o
x
e
800
400
0
lncRNA
protein-coding
macaque
novel
n=3,979
macaque
annotated
n=4,769
human
n=16,887
macaque
n=21,591
human
n=19,970
macaque
novel
n=9,126
macaque
annotated
n=12,370
human
n=47,020
macaque
n=195,853
human
n=199,097
ubiqutous
intermediate
tissue-specific
100
75
50
s
e
n
e
g
f
o
%
5,064
2,429
2,774
7,014
25
1,328
7,012
0
222
749
90
novel
annotated
lncRNA
protein
coding
h
)
M
P
T
0
1
g
o
l
(
i
n
o
s
s
e
r
p
x
e
e
g
a
r
e
v
a
4
2
0
-2
tissue−specific
tissue-specific
specific
ubiquitous
intermediate
intermediate
ubiquitous
ubiquitous
p < 2.2e-16
p =1.76e-60
p = 0.24
novel
n=2,429
annotated
n=2,774
lncRNA
protein
coding
n=5,064
novel
n=1,328
annotated
n=749
lncRNA
protein
coding
n=7,014
novel
n=222
annotated
n=90
lncRNA
protein
coding
n=7,012
protein-coding
macaque
human
0
f
1.00
u
a
T
0.50
0.00
p < 2.2e-16
p < 2.2e-16
p=1.59e-24
novel
n=3,979
annotated
n=4,769
lncRNA
protein
coding
n=21,591
Fig. 1 | Novel lncRNAs resemble annotated lncRNAs and significantly expand
the current macaque lncRNA annotation. a Samples used for de novo tran-
scriptome assembly. CSF: cerebrospinal fluid. b LncRNA discovery pipeline.
N corresponds to the number of transcripts. c Number of novel and annotated
lncRNAs and protein-coding genes in the macaque and human annotation
(Ensembl release 100). d Distribution of transcript length and e exon length.
f Distribution of Tau specificity scores of macaque novel and annotated lncRNA
(red) and protein-coding genes (blue). Mann–Whitney U test. g Percentage of
ubiquitous (Tau < 0.3), intermediate (0.3 ≤ Tau ≤ 0.7), and tissue-specific (Tau > 0.7)
lncRNAs and protein-coding genes. Labels indicate the number of genes within
each category. h Distribution of average expression (log10TPM) in the tissue with
the highest expression of tissue-specific, intermediate, and ubiquitous lncRNAs and
protein-coding genes. Mann–Whitney U test. N corresponds to the sample size of
each category. All boxplots display the median and the first and third quartiles of
the data. The whiskers extend to the highest and lowest values within 1.5 times the
interquartile range (IQR) of the data.
publicly available single-cell RNA-sequencing data from healthy human
PBMCs generated with 10X Genomics40 and replicated our findings
(Supplementary Fig. 4E–G). Thus, our observations are consistent
regardless of single-cell technology, species, gene annotation or infec-
tion status. Overall, our results indicate that in circulating immune cells,
the lower expression levels of lncRNA previously reported in bulk stu-
dies may be driven by lncRNA being expressed in fewer cells compared
to protein-coding genes rather than having less expression across indi-
vidual cells.
Upsilon, a metric to measure cell-type specificity in single-cell
expression data
Tau is a metric routinely used to measure tissue specificity36. However,
to our knowledge, no metric to estimate cell-type specificity has been
Nature Communications |
(2023) 14:3866
3
Article
a
d
Neutrophil
Monocyte
T cell
B cell
b
y
t
i
s
n
e
d
7
1
0
.
0
=
p
5
4
3
2
1
0
e
1.5
ENSMMUG00000045507
NUDT9
https://doi.org/10.1038/s41467-023-39627-7
lncRNA (n=925)
protein-coding (n=11,321)
lncRNA (n=925)
protein-coding (n=11,321)
c
6
7
1
-
e
3
6
.
1
=
p
y
t
i
s
n
e
d
0.8
0.6
0.4
0.2
0.0
2.0
2.5
median expression (logCP10K + 1)
3.0
3.5
0.1
10.0
1.0
% of cells (log10)
100.0
matched by
median expression
p = 3.91e-80
matched by # of cells
p=1
f
)
1
+
K
0
1
P
C
g
o
l
(
i
n
o
s
s
e
r
p
x
e
i
n
a
d
e
m
3.0
2.5
2.0
1.5
g
100
75
s
l
l
50
e
c
%
0.25
0.00
lncRNA
n=925
protein coding
n=925
lncRNA
n=925
protein coding
n=925
Fig. 2 | Expression patterns of lncRNAs and protein-coding genes at single-cell
resolution. a UMAP embedding of 38,067 cells. Cell types are indicated by the
different colors. b Distribution of median gene expression levels (log(CP10K + 1))
and c percentage of cells (log10) in which lncRNA (red) and protein-coding (blue)
genes are expressed. Mann–Whitney U test. d UMAP embedding showing the
expression levels of a lncRNA (ENSMMUG00000045507) and e a protein-coding
gene (NUDT9), with the same median expression level but expressed in a different
number of cells. f Distribution of median expression levels of lncRNA (red) and
protein-coding genes (blue) when matched by the percentage of cells in which they
were expressed. g Percentage of cells in which lncRNA (red) and protein-coding
(blue) genes were expressed when matched by median expression levels. One-side
Wilcoxon signed-rank test. All boxplots display the median and the first and third
quartiles (the 25th and 75th percentiles) of the data. The whiskers extend to the
highest and lowest values within 1.5 times the interquartile range (IQR) of the data.
established. To address this issue, we designed a metric that estimates
cell-type specificity based on single-cell data. We named it Upsilon,
which is the next letter in the Greek alphabet after Tau. Whereas Tau
relies mostly on differences in expression levels41, Upsilon relies on the
proportion of cells expressing a gene (see “Methods”). Similarly to Tau,
Upsilon scores range from 0, for ubiquitous genes to 1, for cell-type
specific genes.
In order to evaluate the ability of both metrics to estimate cell-
type specificity, we repurposed the Tau calculation. Instead of using
the mean expression levels per tissue41, we used the mean expression
levels per cell type (see “Methods”). Both Tau and Upsilon could
accurately classify genes as ubiquitous, intermediate, or cell-type
specific in simulated scenarios although Upsilon was better at classi-
fying different degrees of intermediate cell-type specificity (Supple-
mentary Fig. 5). In addition, we selected a set of housekeeping and
marker genes (see “Methods”) and compared their Tau and Upsilon
scores. While both metrics assigned low values to housekeeping genes,
our metric better identified marker genes as tissue-specific (Fig. 3a).
We then computed Upsilon scores to characterize the cell-type
specificity of lncRNAs. Both novel and annotated lncRNA showed
similar values (Mann–Whitney U test, P-value > 0.05) (Supplementary
Fig. 6A). Using Upsilon, we classified lncRNAs as cell-type specific
intermediate (0.3 ≤ Upsilon ≤ 0.7), and ubiquitous
(Upsilon > 0.7),
(Upsilon < 0.3) (Fig. 3b). We identified 153 cell-type specific lncRNAs, of
which 67 (44%) were annotated and 86 (56%) were novel (Fig. 3b,
Supplementary Fig. 6B). Previously reported disease biomarkers, such
as MIAT42 and DIO3OS43, were among the set of cell-type specific
lncRNAs highlighting the utility of our novel metric in identifying
candidate genes for diseases. Also, we found that cell-type specific
lncRNAs have slightly shorter transcript lengths and slightly fewer and
shorter exons as compared to ubiquitous genes (Mann–Whitney U
test, all P-values < 3.5 × 10−3) (Supplementary Fig. 6C–E).
Tissue-specific expression of protein-coding genes mainly occurs
due to restricted expression at specific cell types44. We sought to
identify whether this held true for lncRNAs as well. We selected genes
that were expressed in both whole blood bulk RNA-seq data and
PBMCs single-cell RNA-seq data (see “Methods”) (Supplementary
Fig. 6F) and compared Tau scores computed in bulk with Upsilon
scores computed in single-cell. Tissue specificity was significantly
correlated with cell-type specificity both in lncRNAs (Spearman
ρ = 0.31, P-value < 2.2 × 10−16) (Fig. 3c) and in protein-coding genes
(Spearman ρ = 0.46, P-value < 2.2 × 10−16) (Supplementary Fig. 6G)
indicating that similar to protein-coding genes, tissue-specific lncRNAs
are more likely expressed in particular cell types.
In summary, we developed a metric called Upsilon, which uses
single-cell data, to identify and characterize cell-type specific lncRNAs,
including known disease biomarkers, demonstrating its potential to
pinpoint candidate disease-associated genes.
The higher specificity of lncRNAs can be attributed to their
expression in fewer cells
LncRNAs are known to be more tissue-specific than protein-coding
genes21,22. We thus wondered whether lncRNA’s higher tissue specifi-
city was due to lncRNAs being expressed in fewer cells or to lncRNAs
being more cell-type specific. To address this, we compared cell-type
specificity values between lncRNAs and protein-coding genes and
found that lncRNAs were more cell-type specific (Mann–Whitney U
(cid:1)10) (Fig. 3d) and could separate cell types in a
test, P-value < 2 × 10
UMAP visualization (Supplementary Fig. 7A). However, when matched
by the number of cells in which they were expressed, protein-coding
Nature Communications |
(2023) 14:3866
4
Article
a
1.00
0.75
e
r
o
c
s
0.50
0.25
0.00
housekeeping
p=0.123
marker
p < 2.2e-16
Tau
n=15
Upsilon
n=15
Tau
n=14
Upsilon
n=14
https://doi.org/10.1038/s41467-023-39627-7
b
600
600
A
N
R
c
n
l
f
o
#
400
400
200
200
0
0
726
620
novel
annotated
351
290
86
67
c
)
u
a
T
(
i
y
t
i
c
fi
c
e
p
s
e
u
s
s
i
t
1.00
0.75
0.50
0.25
0.00
ubiquitous intermediate cell-type
specific
ubiquitous
n=969
intermediate
n=453
cell-type
specific
n=110
d
CTD-2325A15.5
p = 1.07e-33
CD19
matched by # of cells
p=0.41
MSTGR.205441
e
f
n
o
l
i
s
p
U
1.00
0.75
0.50
0.25
0.00
ENSMMUG00000056793
marker gene
housekeeping gene
B2M
1.00
0.75
0.50
n
o
l
i
s
p
U
0.25
0.00
RORA
lncRNA
n=925
protein coding
n=11,332
lncRNA
n=925
protein coding
n=925
Fig. 3 | Identification of cell-type specific lncRNAs. a Distribution of Tau (green)
and Upsilon (yellow) cell-type specificity scores for housekeeping (left) and maker
(right) genes. Wilcoxon signed-rank test. b Bar plot showing the number of ubi-
quitous (Upsilon < 0.3), intermediate (0.3 ≤ Upsilon ≤ 0.7), and cell-type specific
lncRNAs (Upsilon > 0.7). c Distribution of tissue-specificity Tau scores of ubiqui-
tous, intermediate, and specific lncRNAs. d Distribution of cell-type specificity
scores of lncRNA (red) and protein-coding (blue) genes. Cell-type marker genes are
highlighted in green, housekeeping genes in purple. UMAP embeddings of cell-type
specific and ubiquitously expressed lncRNA (red) and protein-coding (blue) genes
are shown as examples. Mann–Whitney U test. e Distribution of Upsilon cell-type
specificity scores of lncRNA and protein-coding genes when matched by the per-
centage of cells in which they were expressed. Wilcoxon signed-rank test. f UMAP
embedding shows the expression pattern of the cell-type-specific lncRNA
MSTRG.205441 (Upsilon = 0.9) (top) and the protein-coding gene RORA (Upsi-
lon = 0.89) (bottom) which were matched by the percentage of cells in which they
are expressed. All boxplots display the median and the first and third quartiles (the
25th and 75th percentiles) of the data. The whiskers extend to the highest and
lowest values within 1.5 times the interquartile range (IQR) of the data.
and lncRNA had comparable cell-type specificity scores (Wilcoxon
signed-rank test, P-value > 0.05) (Fig. 3e, f). On the contrary, when
lncRNA and protein-coding genes were matched by their cell-type
specificity, lncRNAs were expressed in fewer cells (Supplementary
Fig. 7B). To assess whether these observations were independent of
species, completeness of lncRNA annotation,
infection status, or
sequencing platform, we analyzed healthy human PBMC single-cell
data. With this dataset, we also observe that lncRNAs are as cell-type
specific as protein-coding genes when expressed in the same number
of cells (Supplementary Fig. 7C, D).
Overall our observations indicate that the long-assumed higher
tissue specificity of lncRNAs derived from bulk studies might be the
result of their expression in fewer cells rather than overall higher cell-
type specificity.
LncRNAs are dynamically regulated upon EBOV infection
LncRNAs play crucial roles in the host response to viral infections45–48.
However, previous studies mostly relied on bulk tissue data, which
hinders the detection of expression differences at the cellular level. To
investigate the cell-type-specific dynamics of lncRNAs upon immune
stimulation, we use single-cell data from in vivo EBOV-infected maca-
que PBMCs30. We sought to identify lncRNAs with immune regulatory
roles during viral infections in specific cell types. We performed a
differential gene expression analysis separately in each cell type
(monocytes, T, and B cells), comparing each stage of the infection
(early, middle, late) to the baseline (see “Methods”).
We detected 186 differentially expressed (DE) lncRNAs in at least
one cell type (Benjamini–Hochberg’s correction, false discovery rate
(FDR) < 0.05, fold change >10%) (Fig. 4a–c, Supplementary Fig. 8A–D)
(Supplementary Data 2), the majority of which (124 lncRNA, 66%) were
novel, underscoring the importance of refining the annotation of
lncRNAs in model organisms such as rhesus macaque. The largest
number of DE lncRNAs were found in monocytes (142 lncRNAs)
(Fig. 4c, Supplementary Fig. 8A–D), consistent with monocytes being
the main EBOV target49,50 as well as the most abundant cell type in our
dataset. We then used our cell-type specificity metric, Upsilon, to
investigate the cell-type specificity of DE genes. We found that most DE
genes were not cell-type specific (Fig. 4d). Of all DE lncRNAs, 34 had a
human ortholog, and, 28 of those have been previously reported to
change expression during immune response in humans51 (Supple-
mentary Fig. 8E). Consistent with previous studies of immune response
upon infection, SNHG6 and LINC00861 were upregulated52–54. Inter-
estingly, the most transcriptionally repressed lncRNA was the nuclear-
enriched abundant transcript 1 (NEAT1) (Fig. 4c, e). NEAT1 is a well-
studied lncRNA known to play important anti-viral roles55,56. In most
studies, however, NEAT1 is upregulated upon viral infection57 and
Nature Communications |
(2023) 14:3866
5
Article
a
c
Monocyte
M
E
L
PBMCs
b
T cell
https://doi.org/10.1038/s41467-023-39627-7
Neutrophil
day post
infection
B cell
i
E
D
g
n
d
o
c
-
n
e
i
t
o
r
p
f
o
#
7500
5000
2500
0
Monocyte
e
NEAT1
monocytes
2.6
2.4
2.2
)
1
+
K
0
1
P
C
(
g
o
l
B
n=6,770
E
n=9,537
M
n=7,057
L
n=14,703
Infection stages
% of cells
40
50
60
24
15
14
13
10
5
d
ubiquitous
intermediate
specific
lncRNA
lncRNA
protein-coding
E
D
A
N
R
c
n
l
f
o
#
200
150
100
50
0
SNHG6
RP11−320M2.1
AC008079.10
LINC00102
RP11−348N5.9
PVT1
KB−1507C5.4
DANCR
PSMB8−AS1
Fold change
1
0.5
0
−0.5
−1
Novel
Human ortholog
Reported in immLnc
B cell
M
E
L
n = 52
T cell
M
E
L
onoMB
T
onoMB
T
105
120
90
60
30
A
N
R
c
n
l
f
o
r
e
b
m
u
n
0
T
B
Monocyte
001
05
0
f
g
n = 142
n = 44
Fig. 4 | LncRNA expression changes upon EBOV infection are cell-type specific.
a Schematic overview of the in vivo experiment design. b UMAP embedding of
38,067 cells from the in vivo dataset, colored by day post-infection (DPI).
c Heatmaps display lncRNAs DE in monocytes, T cells, and B cells in at least one
infection stage—early (E), middle (M), or late (L)—as compared to baseline (b). Cells
are colored according to the fold changes (log2) in expression values between
baseline and the corresponding infection stage. Only lncRNAs with a human
ortholog have the name displayed. The numbers of DE lncRNAs in each cell type are
depicted at the bottom of the heatmap. d Number of DE lncRNAs (left) and protein-
coding genes (right) ubiquitously expressed (Upsilon <0.3), with intermediate cell-
type specificity score (0.3 ≤ Upsilon ≤ 0.7) or cell-type specific (Upsilon > 0.7) in B
cells, Monocytes and T cells. e NEAT1 expression pattern at different stages of
infection in monocytes. N corresponds to the number of cells in each reported
infection stage. Dots’ sizes represent the percentage of cells in which the gene was
expressed. Dots’ centers represent the mean. Error bars indicate the 95% con-
fidence interval around the mean, calculated using the standard error of the mean
(SEM). f Upset plots showing the overlap of DE lncRNAs across cell types (g) and
infection stages.
downregulation has only been described in dengue and Crimean
Congo hemorrhagic fever58,59. Our results suggest that NEAT1 deple-
tion may be specific to severe hemorrhagic fevers and in the case of
EBOV at least, downregulation occurs specifically in monocytes.
We then wanted to compare the expression dynamics of lncRNAs
to that of protein-coding genes upon immune stimulation. Most
lncRNAs (144 lncRNAs, ~78%) were DE in exclusively one cell type
(Fig. 4f) which was a significantly larger proportion than the one
observed for protein-coding genes (Fisher’s exact test, OR = 2.06,
(cid:1)5) (see “Methods”). However, when matched by
P-value = 1.945 × 10
the number of cells in which they were expressed, the two gene classes
had comparable proportions of cell-type specific DE genes (Fisher’s
exact test; OR = 0.90, P-value = 0.69). Similarly, the majority of
lncRNAs (109 lncRNAs, ~60%) were DE in only one stage of the infection
(Fig. 4g) which is a significantly larger proportion than that of protein-
(cid:1)3)
coding genes (Fisher’s exact test, OR = 1.49, P-value = 8.83 × 10
Nature Communications |
(2023) 14:3866
6
Article
https://doi.org/10.1038/s41467-023-39627-7
(see “Methods”). This difference disappeared when comparing lncRNA
and protein-coding genes matched by the number of cells in which
they were expressed (Fisher’s exact test; OR = 1.03, P-value = 0.57).
Overall, our results indicate that upon EBOV immune stimulation the
transcriptional response of lncRNAs is stage and cell-type specific
similar to that of protein-coding genes.
Functional characterization of lncRNAs differentially expressed
upon EBOV infection
Although we detected many lncRNAs that change their expression
upon EBOV infection, most of them remain functionally uncharacter-
ized. Some lncRNAs are known to exert their modulatory role in cis60.
To identify possible cis-regulatory lncRNAs, we first identified 327
lncRNA protein-coding gene pairs that were both DE in the same cell
type and in close physical proximity (<1 Mbp). DE lncRNA and protein-
coding genes were not significantly co-located more often than
expected by chance (Fisher’s exact test, OR = 0.91, P-value > 0.05) (see
“Methods”). However, we found 41 gene pairs that were co-located and
co-expressed at cell-type resolution (Spearman correlation test, P-
value < 0.05, Supplementary Fig. 9).
interferon-stimulated genes
To explore further the pathways and putative functions of our DE
lncRNAs, we built a cell-type-specific co-expression network in
monocytes using both lncRNA and protein-coding genes (see “Meth-
ods”). The network had 8 modules with an average of 7 lncRNAs and 15
protein-coding genes (Fig. 5a). Three modules displayed significant
functional enrichments, primarily related to immune stimulation
(Supplementary Fig. 10A–C, Supplementary Data 3). One of these
modules contained several
(ISGs),
including MX1, IFIT2, and ISG15, and was enriched in genes that
increased expression at early and mid stages of infection30 (Fig. 5a).
Interestingly, we identified a lncRNA, ENSMMUG00000064224,
directly connected to ISGs, that exhibited a similar expression profile
as ISG with an upregulation in all three cell types at early infection
(Supplementary Fig. 10D, E). We also found one module with a
remarkable number of enriched terms related to cell proliferation and
migration. Most of the genes in this module were downregulated with
the strongest expression changes at the late stages of infection
(Fig. 5a), suggesting a late host
response to prevent EBOV
replication61,62. Although the remaining five modules did not have
significant enrichments, all of them included between 1 and 8 central
regulators or downstream effectors of the innate immune response63
(Supplementary Data 3).
Previous work based on PBMCs infected with EBOV ex vivo
showed that EBOV hijacks infected cells’ defenses by downregulating
anti-viral genes and upregulating pro-viral genes30. Using an ex vivo
experimental setup allows for higher viral exposure to EBOV and
consequently a higher number of infected cells with higher viral loads
compared with the same cell type bystander cells. We sought to
investigate if lncRNAs were up or downregulated upon viral cellular
entry and proliferation compared to bystander cells. To do this, we
identified lncRNAs whose expression significantly correlated with viral
load in EBOV-infected monocytes ex vivo (Fig. 5b, Supplementary
Fig. 11A–E). We identified 16 lncRNAs significantly correlated with viral
load (Spearman correlation test, P-value < 0.05)
(Supplementary
Data 4), the majority of which (12) were positively correlated (Fig. 5c).
Importantly, ENSMMUG00000058644 and MSTRG.15458, which had
the strongest correlations, were also significantly correlated at nom-
inal P-values in the in vivo dataset (Spearman ρ = 0.10, P-value = 0.03
and Spearman ρ = −0.12, P-value = 0.01, respectively), suggesting that
the in vivo dataset might not have enough infected cells, and thus
power, to identify significant correlations. In line with this, lncRNAs
correlated with viral load were expressed in significantly fewer cells
in vivo compared to ex vivo (Mann–Whitney U test, P-value < 2 ×
(cid:1)10). 10 out of the 16 identified lncRNAs were not detected as DE with
10
EBOV infection in monocytes in vivo (Fig. 4c, Supplementary Data 2
and 4), suggesting that most of these lncRNAs change their expression
exclusively in infected cells. This observation highlights the power of
the single-cell analysis to discern between expression changes in
bystanders and infected cells. Interestingly, the remaining five lncRNAs
were DE upon infection in monocytes in the in vivo dataset but in
opposite directions: two lncRNAs were upregulated during EBOV
infection in the general in vivo monocyte population but were nega-
tively correlated with the viral load in ex vivo infected cells; three
lncRNAs were downregulated during EBOV infection in the general
in vivo monocyte population but increased their expression with viral
load in ex vivo infected monocytes (Fig. 5d–g, Supplementary
Fig. 12A–F).
Overall, our functional analyses revealed that lncRNAs whose
expression varies upon EBOV infection are involved in the same
pathways as DE protein-coding genes, suggesting that these lncRNAs
might be important immune regulators. In addition, our ex vivo results
indicate that EBOV entry in the cell can alter the expression of lncRNA
exclusively in infected cells and that in some cases, the expression
changes differ between infected and bystander cells. This would be
consistent with previous studies that reported that EBOV hijacks par-
ticular pathways in infected cells to promote viral entry and
replication30.
Discussion
Long non-coding RNAs play critical roles in immune regulation10,11.
However, studies that require working with non-human animal mod-
els, such as Ebola virus infection, are constrained by an incomplete
lncRNAs’ annotation. To address this issue, we generated a multi-tissue
bulk RNA sequencing dataset from both EBOV-infected and uninfected
samples and annotated nearly 4000 novel lncRNAs. This effort resul-
ted in nearly doubling the current annotation of lncRNA in rhesus
macaque. Importantly, we found that 66% of all lncRNAs changing
expression upon EBOV infection in single cells were novel. These
findings underscore the importance of expanding current non-coding
transcriptome annotations with datasets that sample different phy-
siological conditions, especially in model species widely used in bio-
medical research64. Future work using emerging long-read sequencing
technologies65 will further improve the discovery and annotation of
lncRNAs in model species in the context of infection.
LncRNAs are generally assumed to be more lowly expressed and
more tissue-specific than protein-coding genes16. These observations
arise from bulk studies that measure average expression levels across
cell populations. Single-cell data allows both detecting gene expres-
sion levels in individual cells and determining how many cells in a given
population express a gene. Exploiting this unique feature, we found
that, when controlling for the number of cells in which lncRNA and
protein-coding genes are expressed, lncRNAs are not less expressed,
neither are more cell-type specific. Liu et al.66 made a similar obser-
vation in brain tissue although their study was heavily constrained by
the number of cells analyzed (<250 cells). This result raises the intri-
guing question of why lncRNAs’ expression is systematically restricted
to fewer cells but when transcribed they reach similar expression levels
to protein-coding genes. In a recent study, Johnsson et al.67 use allele-
sensitive single-cell RNA sequencing to assess the transcriptional
dynamics of lncRNAs. Their results show that lncRNAs have lowered
transcriptional burst frequencies and longer duration between those
bursts. Consistent with this, our previous work showed that lncRNAs
harbor fewer transcription factor binding sites and higher chromatin
repressive marks in their promoter regions compared to equally
expressed protein-coding genes22. In addition, transcription factor
binding sites in lncRNAs’ promoters are less complex than those in
protein-coding genes, suggesting that fewer transcription factors can
bind to lncRNAs’ promoters23. Overall, these results are consistent with
a model in which the promoters of lncRNAs differ from those of
equally expressed protein-coding genes in the probability of engaging
Nature Communications |
(2023) 14:3866
7
Article
a
https://doi.org/10.1038/s41467-023-39627-7
immune response
immune response
response
to LPS
cell proliferation
response
to LPS
cell proliferation
b
viral load
(log10)
1.6
1.2
0.8
0.4
0.0
c
negatively
correlated
positively
correlated
annotated
novel
0
2
4
6
8
10
12
number of lncRNAs
d
)
1
+
K
0
1
P
C
(
g
o
l
1.5
1.4
1.3
1.2
1.1
f
)
1
+
K
0
1
P
C
(
g
o
l
0.10
0.08
0.06
0.04
0.02
ENSMMUG00000064224
baseline
n=981
bystander
n=679
infected
n=1,843
MSTRG.181870
% cells
62
64
66
68
e
1.5
1.0
0.5
0
)
1
+
K
0
1
P
C
(
g
o
l
-0.5
-1.0
0
g
% cells
2
4
6
1.0
0.5
)
1
+
K
0
1
P
C
(
g
o
l
ENSMMUG00000064224
IFIT2
ISG15
MX1
10
20
30
40
% viral load
MSTRG.181870
baseline
n=981
bystander
n=679
infected
n=1,843
0
0
10
20
30
40
% viral load
Fig. 5 | In silico functional characterization of lncRNAs and protein-coding
genes upon EBOV infection in monocytes. a Regulatory network of lncRNAs
(circles) and protein-coding (squares) DE. Vertices’ colors represent up- or down-
regulated genes (left) or whether a gene has the strongest fold-change compared to
baseline in early (yellow), middle (orange), or late (red) stages of infection (right).
Modules with significant enrichments are circled in gray and their description
summarizes top enriched terms. b UMAP embedding of 56,317 cells from the
ex vivo dataset. The magnified UMAP shows the viral load in monocytes. c Number
of lncRNAs correlated with viral load in monocytes. d Expression of the lncRNA
ENSMMUG00000064224 in monocytes in baseline, bystander, and infected cells
(24 h). Dots’ centers represent the mean. Error bars indicate the 95% confidence
interval around the mean, calculated using the standard error of the mean. Dots’-
sizes represent the percentage of cells expressing the gene. N corresponds to the
number of cells in each infection stage. e Expression of ENSMMUG00000064224
and ISGs versus viral load. The shaded area around the smoothed line represents
the 95% confidence interval (loess smoothing method). f, g Same as (d, e) for the
lncRNA MSTRG.181870.
in active transcription rather than in the strength of the transcriptional
response.
LncRNAs whose expression is condition or cell-type specific are
candidate disease biomarkers and potential therapeutic targets68.
Multiple metrics have been developed to measure tissue specificity in
bulk data41, but none of those has been specially designed to measure
cell-type specificity. In this study, we introduce Upsilon, a metric that
leverages the unique feature of single-cell technologies to know the
number of cells expressing a gene to estimate cell-type specificity. We
have identified 153 cell-type specific lncRNAs in PMBCs, including
Nature Communications |
(2023) 14:3866
8
Article
https://doi.org/10.1038/s41467-023-39627-7
some disease biomarkers supporting the utility of this metric to
identify disease-related genes. We anticipate that, with the growing
availability of single-cell transcriptomics data69, Upsilon will be
extensively used in advancing our understanding of cell-type specific
processes in the context of health and disease. Furthermore, our work
has consistently shown that lncRNAs have cell-type and stage-specific
regulation upon EBOV infection, to a similar extent to that of protein-
including a differential response when comparing
coding genes,
infected versus bystander monocytes. Further studies with larger
sample sizes will increase our understanding of lncRNA regulation
upon viral entry and immune stimulation.
Collectively, this study elucidates the roles of lncRNAs in response
to EBOV infection and paves the way for future studies on how to
systematically analyze lncRNAs at single-cell resolution.
Methods
Animal sampling
No animal handling was involved in this study. Samples from Rhesus
macaques (Macaca Mulatta, 43 samples across 12 tissues) were
obtained from ref. 70. Animal handling was performed in accordance
with the Guide for the Care and Use of Laboratory Animals of the
National Institute of Health, the Office of Animal Welfare, and the US
Department of Agriculture. In addition, some other samples were
obtained from commercially available samples of 2 Rhesus macaques
(Macaca Mulatta, 16 samples across 10 tissues)
(Zyagen, San
Diego, CA, USA).
RNA sample processing
For de novo annotation, we generated paired-end, strand-specific
bulk short-read RNA-sequencing (RNA-Seq) on high-quality,
commercially available rhesus macaque (Macaca mulatta) total
RNA (Zyagen, San Diego, CA, USA; hereafter referred to as Zya-
gen) of non-infected samples from 10 different tissues (Supple-
mentary Data 1). Briefly, we depleted ribosomal RNA and
performed random-primed cDNA synthesis71, followed by second
strand marking and DNA ligation72 with adapters containing
unique molecular identifiers (UMIs)73 (IDT, Coralville, IA, USA).
We performed the identical bulk RNA-Seq protocol but without
UMIs on rhesus macaque RNA samples from 12 different tissues
from the study by Luke et al.70. In addition, we downloaded whole
blood bulk short-read RNA-Seq data from healthy samples and
samples infected with Makona Ebola Virus from the NCBI Gene
Expression Omnibus (GEO; accession number GSE115785). For the
single-cell RNA-Seq analysis, we downloaded the PBMCs dataset
from the NCBI Gene Expression Omnibus (GEO) with accession
number GSE158390.
QC and mapping
First, we merged Ensembl Mmul_10 release 100 assembly and Ensembl
release 100 gene annotation with the Ebola virus/H. sapiens-tc/COD/
1995/Kikwit-9510621 (GenBank #KU182905.1; Filoviridae: Zaire ebola-
virus) assembly and annotation,
respectively, and used them
throughout all downstream analyses. We used Hisat v2.1.074 to com-
pute assembly indexes and known splice sites and mapped each
sample’s reads to the merged assembly. We ran Hisat2 with default
parameters, except for RNA-strandness, which we set according to the
experiments’ strandness (Supplementary Data 1), previously inferred
with InferExperiment.py from RSeQCc v3.0.075. We sorted mapped
bam files with samtools sort v1.976 with default parameters. We
retained only paired and uniquely mapped reads using samtools view
with parameters -f3 -q 60. In addition, we removed duplicates from the
samples tagged with UMIs (Zyagen) (Supplementary Data 1) with
umi_tools dedup v1.0.077. We excluded all samples with less than 10 M
sequenced reads, a mapping rate lower than 0.3, or a genic mapping
rate lower than 0.7. We defined the genic mapping rate as the
proportion of exonic and intronic reads, as computed by read_-
distribution.py from RSeQCc v3.0.075 (see Supplementary Data 1).
LncRNA discovery pipeline
We ran de novo transcriptome assembly separately on each
sample with Stringtie v1.3.678, with default parameters except for
strand information that was set depending on the dataset (Sup-
plementary Data 1). We used Stringtie to merge all the de novo
assemblies using the parameter “--merge”. To identify novel
transcripts absent
from the reference annotation, we used
Gffcompare v0.10.6 and retained exclusively the transcripts with
class codes “u” and “x”, corresponding to intergenic and anti-
sense transcripts. We removed mono-exonic transcripts, tran-
scripts shorter than 200 bp, and kept only transcripts abundantly
expressed (log(TPM) > 0.5) in at least three samples. To assess the
coding potential of the newly assembled transcripts, we used
three sequence-based lncRNAs prediction tools: Coding Potential
Assessment Tool v3.0.0 (CPAT)79, Coding Potential Calculator
v2.0 (CPC2)80, and Coding-Non-Coding Identifying Tool v2
(CNIT)81 with default parameters. For each independent predic-
tion tool, we removed genes with at least one isoform predicted
as non-coding and one as protein-coding. We considered a gene
to be a long non-coding RNA if the three tools classified it as non-
coding. We then merged the obtained list of novel lncRNAs to the
reference annotation and used it in downstream analyses. To
benchmark our lncRNAs discovery pipeline, we predicted the
biotype of annotated genes (Ensembl v100) (coding or non-cod-
ing) and compared our predictions to their annotated biotype. To
compare lncRNA and protein-coding transcript length, number of
exons and exon length, we considered the longest transcript per
gene. To identify lncRNAs orthologs to human, we used the
synteny-based lncRNAs detection tool slncky v1.0 on human hg38
assembly and gencode hg38 v23 annotation82. For the sake of
reproducibility, the lncRNAs discovery pipeline is implemented in
Nextflow83 and combined with Singularity software containers.
Tissue-specificity estimates
We calculated gene tissue-specificity scores using Tau36 based on
average tissue TPM gene expression values. Tau ranges from 0 to 1:
genes with a score close to 1 are more specifically expressed in one
tissue, while genes with a score closer to 0 are equally expressed across
all tissues. We classified genes as tissue-specific (Tau > 0.7), inter-
mediate (0.3 ≤ Tau ≤ 0.7), or ubiquitous (Tau <0.3). For tissue-specific
genes, we determined the tissue in which they exhibited the highest
average expression (log10TPM value) and considered them to be
specific for that tissue. To compare the expression levels between
tissue-specificity groups, we selected the expression value of the tissue
with the highest average expression for each gene.
Single-cell RNA sequencing data and processing
We used two publicly available single-cell RNA-Seq datasets of Rhesus
Macaque peripheral mononuclear cells (PBMCs) infected with EBOV
in vivo and ex vivo30. The in vivo dataset comprised samples from 21
individuals, collected before and at several days post-infection (DPI)
with EBOV, and contained 38,067 cells. We performed the gene quan-
tification using the Drop-seq analysis pipeline (https://github.com/
broadinstitute/Drop-seq), with the scripts executed using Nextflow83
and Singularity containers for better reproducibility (https://github.
com/Mele-Lab/2023_SingleCellEbolaLncRNAs_NatComms). We used
Scrublet v.0.2.184 for doublet detection and applied the IntegrateData
method of Seurat v3.085 for fresh versus frozen batch effect correction.
To select suitable filtering thresholds, we followed the best practices for
single-cell analyses86, including the selection of cells with at least 1000
and a maximum of 10,000 UMIs, at least 600 and a maximum of 2000
detected genes, and the exclusion of cells with more than 5% of
Nature Communications |
(2023) 14:3866
9
Article
https://doi.org/10.1038/s41467-023-39627-7
mitochondrial reads. The counts were normalized to log(CP10K + 1)
after removing viral transcripts to avoid library size normalization
biases.
The ex vivo dataset included PBMCs from healthy macaques,
either inoculated, irradiated, or incubated with the virus, that were
sequenced at 4 or 24 h post-infection (Supplementary Fig. 11) and it
contained 56,317 cells. We followed the same processing steps as the
in vivo dataset but increased the upper thresholds to ensure we did not
exclude highly infected cells or cells with particularly increased
expression of host genes, keeping those with less than 15,000 UMIs
and less than 4000 detected genes per cell.
To replicate some of our observations in human data, we used
available gene counts of human healthy PBMCs from 10x Genomics40
(32,738 available cells) and human Ensembl version 100 gene annota-
tion. We applied the same QC and filtering protocols.
Single-cell clustering and cell type identification
To cluster cells, we used the Louvain algorithm as implemented in the
Seurat package85. To identify cluster-specific genes, we ran a differ-
ential expression analysis between each cluster and all the remaining
ones using the Seurat function FindAllMarkers. Based on the expres-
sion levels of known marker genes, we classified clusters into the four
major PBMCs cell types (T cell, B cell, Monocytes, and Natural Kill-
ers) (Fig. 2a).
which we detected the gene as expressed, so that, per gene, the pro-
portions assigned to the different cell types sum up to one.
– Ei,j is the expected proportion of cells in which gene i would be
expressed in cell-type j if it was not cell-type specific. The expected
proportion of cells for cell type j is equal for all the genes and corre-
sponds to the proportion of cells of cell type j in the dataset.
Then, we divided the difference between the observed and
expected proportions by the maximum value this difference could
reach. The maximum value is reached when the gene is expressed in all
cells of one cell type, which is the difference between 1 and the
expected proportion. The value, therefore, ranges from 0 to 1. We then
calculated the specificity of each gene to each of the cell types and
these values as the gene’s global
reported the maximum of
specificity score.
Cell-type specificity simulations
To explore the performance of cell-type specificity metrics, we
designed different hypothetical scenarios with genes presenting three
degrees of cell-type specificity (highly, intermediate, or lowly cell-type
specific genes) in a cell population of three cell types. To do this, we
kept a fixed expression value for expressed genes (TPM = 2) and zero
for non-expressed and modified the proportion of cells of a particular
cell type where the gene was expressed (50%, 30%, and 20% of the total
number of cells) (Supplementary Fig. 5).
LncRNA and protein-coding gene comparisons
We used Seurat’s normalization values (log(CP10K + 1)) to com-
pare expression levels between lncRNAs and protein-coding
genes. We considered a gene to be expressed in a cell when its
normalized expression value was larger than 1. This generated a
total of 2037 lncRNA and 13,718 protein-coding genes. To com-
pare the properties (i.e., expression or number of cells in which a
gene is expressed or cell-type specificity), we only used genes
expressed in more than 60 cells, leaving a total of 925 lncRNA and
11,321 protein-coding genes. Median expression values were cal-
culated exclusively across cells in which the gene was expressed.
We used the MatchIt R package
v4.0.0 (https://www.
rdocumentation.org/packages/MatchIt/) to obtain the pairs of
lncRNA and protein-coding genes matched either by median
expression or by the percentage of cells in which they were
expressed.
Cell-type specificity estimates
We considered two distinct cell-type specificity measurements. First,
we leveraged Tau41, a metric originally designed to assess tissue-
specificity. Instead of calculating the mean expression per tissue for
each gene, we calculated the mean expression per cell type, including
zeros. Tau was calculated as follows:
P
τ =
Þ
ð1 (cid:1) ^xi
i
n (cid:1) 1
, i = 1,2 . . . n; ^xi =
xi
(cid:2) (cid:3)
maxi = 1...n xi
ð1Þ
where xi is the mean expression of a gene in cell type i and n is the total
number of cell types. In addition, we designed a score (Upsilon, υ) that
relies purely on the proportion of cells in which each gene is expressed,
which was calculated as follows:
υ = maxj = 1...n
(cid:1) Ej
Oi,j
1 (cid:1) Ej
ð2Þ
where:
– Oi,j is the observed proportion of cells in which gene i is found
expressed in cell type j. To calculate the proportions of cells in which
each gene is expressed per cell type, we considered only the cells in
Marker and housekeeping genes selection
We obtained the list of PBMC marker genes with the Seurat85 function
FindAllMarkers. As housekeeping genes, we selected RRN18S, RPLP0,
GAPDH, ACTB, PGK1, RPL13A, ARBP, B2M, YWHAZ, SDHA, TFRC, GUSB,
HMBS, HPRT1, and TBP87–89. We used the cell-type specificity score of
the collected marker and housekeeping genes to compare the ability
of Upsilon and the repurposed Tau to distinguish established cell-type
specific and ubiquitous genes.
Correlation tissue and cell-type specificity
To determine the correlation between tissue specificity and cell-
type specificity, we selected genes expressed in both whole blood
samples from the bulk RNA-seq dataset (average TPM > 0.1) and in
the single cell in vivo PBMC dataset (log(CP10K + 1) > 1 in at least 10
cells). A total of 1532 lncRNAs and 11,501 protein-coding genes were
obtained (Supplementary Fig. 6F). We then conducted a Fisher
exact test to confirm that the overlap was significant. The variables
tested included genes expressed in both datasets, genes expressed
only in whole blood, genes expressed in PBMCs, and macaque-
annotated genes not expressed. Using the resulting set of expressed
genes in both datasets, we calculated the Spearman correlation
coefficient separately for lncRNAs and protein-coding genes, to
determine the correlation between tissue specificity Tau and cell-
type specificity Upsilon.
Differential expression analysis
We grouped samples of the in vivo dataset based on their day post-
infection: baseline (0 DPI) (13 individuals), early (3 DPI) (3 individuals),
middle (4–5 DPI) (4 individuals), and late stages (6–8 DPI) (8 indivi-
duals). We ran differential expression analysis using MAST v1.12.090 in
each cell type separately. We excluded neutrophils as they were
detected exclusively at later stages of infection. As input, we used the
log-normalized and scaled expression counts (logCP10K + 1) from
those genes expressed in at least 10% of the cells within each cell type.
We performed pairwise comparisons between each stage of infection
(early, middle, late) and baseline within each cell type separately. We fit
a hurdle model that included as covariates the number of genes
detected per cell and a binary variable corresponding to the proces-
sing of the sample, whether it was fresh or frozen. The resulting model
was the following:
Nature Communications |
(2023) 14:3866
10
Article
https://doi.org/10.1038/s41467-023-39627-7
Expression (logCP10K + 1) ~ InfectionStage + NumDetectedGenes +
SampleProcessing
Differential expression P-values were corrected with Benjamin
and Hochberg multiple testing91. Genes were considered to be DE if
they had a logFC > 0.1 and adjusted P-value < 0.05.
Expression dynamics differences between lncRNA and protein-
coding genes
We used Fisher’s exact test to investigate whether lncRNAs have dif-
ferential expression patterns more cell-type-specific or stage-specific
than protein-coding genes. The two tested variables are gene biotype
and whether the gene is DE in one or more cell types or the stage.
Gene colocation analysis
We used the GenomicRanges package v1.38.0 (https://bioconductor.
org/packages/release/bioc/html/GenomicRanges.html)
to calculate
the genomic distance between genes in the macaque Ensembl v100
annotation. We considered a pair to be co-located if they are less than
1 Mbp. To test whether DE lncRNAs were closer to DE protein-coding
genes more often than not DE lncRNAs, we set up Fishers’ exact test.
The two tested variables were whether the lncRNA is DE and whether it
is in cis to a DE protein-coding gene.
Co-expression network
We built a co-expression network using all differentially expressed
genes in monocytes with GrnBoost292. To focus on the co-regulatory
network involving lncRNAs, we only retained edges connected to at
least one lncRNA. Also, we retained the top 0.5% edges when sorted by
weight. We identified communities with the Louvain algorithm93 and
reported those with at least 7 edges. For the functional enrichment of
the modules, we used the R package clusterProfiler v4.2.094.
Correlation with viral load
To determine the correlation between viral transcript changes and
gene expression in infected cells, we focused solely on monocytes at a
late stage of infection (24 h post-infection ex vivo and 6–8 days post-
infection in vivo). We obtained the viral load by dividing the number of
viral counts by the total number of counts and then computed the
Spearman correlation coefficient between the viral load (log10) and
the normalized expression of each gene (log(CP10K + 1)). The resulting
P-values were corrected for multiple testing using the Benjamin and
Hochberg method91.
Ethics
The study was performed in accordance with the Guide for the Care
and Use of Laboratory Animals of the National Institute of Health, the
Office of Animal Welfare, and the US Department of Agriculture38.
Reporting summary
Further information on research design is available in the Nature
Portfolio Reporting Summary linked to this article.
Data availability
The sequencing data generated in this study have been deposited in
the NCBI Gene Expression Omnibus (GEO) database under accession
code GSE192447. The publicly available whole blood bulk short-read
RNA-Seq data from healthy samples and samples infected with Makona
Ebola Virus data used in this study are available in the NCBI Gene
Expression Omnibus (GEO) database under accession code GSE115785.
The single-cell RNA-Seq data used in this study are available in the NCBI
Gene Expression Omnibus (GEO) database under accession code
GSE158390. Raw Seurat Objects for both single-cell datasets used in
this study are available at Zenodo. The full co-expression network file is
also provided (https://doi.org/10.5281/zenodo.7997135). The refer-
ence genome of EBOV used in this study is available in the GenBank
database under accession code KU182905.1. The assembly and refer-
ence genome of Macaca Mulatta used in this study are available in the
Ensembl database (Mmul_10) (https://ftp.ensembl.org/pub/release-100/
fasta/macaca_mulatta/dna/Macaca_mulatta.Mmul_10.dna.toplevel.fa.gz,
https://ftp.ensembl.org/pub/release-100/gtf/macaca_mulatta/Macaca_
mulatta.Mmul_10.100.gtf.gz). The assembly and reference genome of
human used in this study are available in the Gencode database
(https://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_
(release_23)
human/release_23/gencode.v23.annotation.gtf.gz, https://ftp.ebi.ac.uk/
pub/databases/gencode/Gencode_human/release_23/GRCh38.primary_
assembly.genome.fa.gz). Source data are provided with this paper.
Code availability
The code used for this study is available at: https://github.com/Mele-
Lab/2023_SingleCellEbolaLncRNAs_NatComms.
References
1.
Delás, M. J. & Hannon, G. J. lncRNAs in development and disease:
from functions to mechanisms. Open Biol. 7, 170121 (2017).
2. Necsulea, A. et al. The evolution of lncRNA repertoires and
3.
expression patterns in tetrapods. Nature 505, 635–640 (2014).
de Goede, O. M. et al. Population-scale tissue transcriptomics maps
long non-coding RNAs to complex disease. Cell 184,
2633–2648.e19 (2021).
4. Nicholas, W. & Mathy, X.-M. C. Long non-coding RNAs (lncRNAs)
and their transcriptional control of inflammatory responses. J. Biol.
Chem. 292, 12375–12382 (2017).
5. Wang, K. C. & Chang, H. Y. Molecular mechanisms of long non-
6.
7.
coding RNAs. Mol. Cell 43, 904–914 (2011).
Pandey, R. R. et al. Kcnq1ot1 antisense noncoding RNA mediates
lineage-specific transcriptional silencing through chromatin-level
regulation. Mol. Cell 32, 232–246 (2008).
Kallen, A. N. et al. The imprinted H19 lncRNA antagonizes let-7
microRNAs. Mol. Cell 52, 101–112 (2013).
8. Grote, P. et al. The tissue-specific lncRNA Fendrr is an essential
9.
regulator of heart and body wall development in the mouse. Dev.
Cell 24, 206–214 (2013).
Yang, L., Froberg, J. E. & Lee, J. T. Long noncoding RNAs: fresh
perspectives into the RNA world. Trends Biochem. Sci. 39,
35–43 (2014).
10. Heward, J. A. & Lindsay, M. A. Long non-coding RNAs in the reg-
ulation of the immune response. Trends Immunol. 35,
408–419 (2014).
11. Marina, R. & Hadjicharalambous, M. A. L. Long non-coding
RNAs and the innate immune response. Non-Coding RNA 5, 34
(2019).
12. Ahmad, I., Valverde, A., Ahmad, F. & Naqvi, A. R. Long noncoding
RNA in myeloid and lymphoid cell differentiation, polarization and
function. Cells 9, 269 (2020).
13. Mariotti, B. et al. The long non-coding RNA NRIR drives IFN-
response in monocytes: implication for systemic sclerosis. Front.
Immunol. 10, 100 (2019).
14. Cui, H. et al. Long noncoding RNA Malat1 regulates differential
activation of macrophages and response to lung injury. JCI Insight
4, e124522 (2019).
15. Quinn, J. J. & Chang, H. Y. Unique features of long non-coding RNA
biogenesis and function. Nat. Rev. Genet. 17, 47–62 (2016).
16. Derrien, T. et al. The GENCODE v7 catalog of human long non-
coding RNAs: analysis of their gene structure, evolution, and
expression. Genome Res. 22, 1775–1789 (2012).
17. Cabili, M. N. et al. Integrative annotation of human large intergenic
noncoding RNAs reveals global properties and specific subclasses.
Genes Dev. 25, 1915–1927 (2011).
Igor Ulitsky, D. P. B. lincRNAs: genomics, evolution, and mechan-
isms. Cell 154, 26–46 (2013).
18.
Nature Communications |
(2023) 14:3866
11
Article
https://doi.org/10.1038/s41467-023-39627-7
19. Cabili, M. N. et al. Localization and abundance analysis of human
lncRNAs at single-cell and single-molecule resolution. Genome
Biol. 16, 20 (2015).
42. Ye, T. et al. LncRNA MIAT services as a noninvasive biomarker for
diagnosis and correlated with immune infiltrates in breast cancer.
Int. J. Women’s Health 13, 991–1004 (2021).
20. Hezroni, H. et al. Principles of long noncoding RNA evolution
43. Wang, Y., Wang, J., Wang, C., Chen, Y. & Chen, J. DIO3OS as a
derived from direct comparison of transcriptomes in 17 species.
Cell Rep. 11, 1110–1122 (2015).
potential biomarker of papillary thyroid cancer. Pathol. Res. Pract.
229, 153695 (2022).
21. Tilgner, H. et al. Deep sequencing of subcellular RNA fractions
44. Karlsson, M. et al. A single-cell type transcriptomics map of human
shows splicing to be predominantly co-transcriptional in the human
genome but inefficient for lncRNAs. Genome Res. 22,
1616–1625 (2012).
22. Melé, M. et al. Chromatin environment, transcriptional regulation,
and splicing distinguish lincRNAs and mRNAs. Genome Res. 27,
27–37 (2017).
tissues. Sci. Adv. 7, eabh2169 (2021).
45. Ginn, L., La Montagna, M., Wu, Q. & Shi, L. Diverse roles of long non‐
coding RNAs in viral diseases. Rev. Med. Virol. 31, https://doi.org/
10.1002/rmv.2198 (2021).
46. Kesheh, M. M., Mahmoudvand, S. & Shokri, S. Long noncoding RNAs
in respiratory viruses: a review. Rev. Med. Virol. 32, 2275 (2022).
23. Mattioli, K. et al. High-throughput functional analysis of lncRNA
47. Fortes, P. & Morris, K. V. Long noncoding RNAs in viral infections.
core promoters elucidates rules governing tissue specificity. Gen-
ome Res. 29, 344–355 (2019).
Virus Res. 212, 1–11 (2016).
48. Liu, W. & Ding, C. Roles of LncRNAs in viral infections. Front. Cell.
24. Malvy, D., McElroy, A. K., de Clerck, H., Günther, S. & van Griensven,
Infect. Microbiol. 7, 205 (2017).
J. Ebola virus disease. Lancet 393, 936–948 (2019).
25. Jacob, S. T. et al. Ebola virus disease. Nat. Rev. Dis. Prim. 6,
26.
1–31 (2020).
Ilunga Kalenga, O. et al. The ongoing Ebola epidemic in the
Democratic Republic of Congo, 2018–2019. N. Engl. J. Med. 381,
373–383 (2019).
27. Caballero, I. S. et al. In vivo Ebola virus infection leads to a strong
innate response in circulating immune cells. BMC Genom. 17,
1–13 (2016).
28. Jain, S., Khaiboullina, S. F. & Baranwal, M. Immunological per-
spective for Ebola virus infection and various treatment measures
taken to fight the disease. Pathogens 9, 850 (2020).
29. Kazer, S. W. et al. Integrated single-cell analysis of multicellular
immune dynamics during hyperacute HIV-1 infection. Nat. Med. 26,
511–518 (2020).
30. Kotliar, D. et al. Single-cell profiling of Ebola virus disease in vivo
reveals viral and host dynamics. Cell 183, 1383–1401 (2020).
31. Siragam, V., Wong, G. & Qiu, X. G. Animal models for filovirus
infections. Zool. Res. 39, 15–24 (2018).
32. Nakayama, E. & Saijo, M. Animal models for Ebola and Marburg virus
infections. Front. Microbiol. 4, 267 (2013).
33. Cross, R. W. et al. Comparative transcriptomics in Ebola Makona-
infected ferrets, nonhuman primates, and humans. J. Infect. Dis.
218, 486 (2018).
34. Bryzghalov, O., Szcześniak, M. W. & Makałowska, I. SyntDB: defining
orthologues of human long noncoding RNAs across primates.
Nucleic Acids Res. 48, D238–D245 (2019).
35. Chen, L. et al. Tissue expression difference between mRNAs and
lncRNAs. Int. J. Mol. Sci. 19, 3416 (2018).
36. Yanai, I. et al. Genome-wide midrange transcription profiles reveal
expression level relationships in human tissue specification. Bioin-
formatics 21, 650–659 (2005).
37. Gawronski, K. A. & Kim, J. Single cell transcriptomics of non-coding
RNAs and their cell-specificity. Wiley Interdiscip. Rev. RNA 8,
e1433 (2017).
49. Geisbert, T. W. et al. Mechanisms underlying coagulation abnorm-
alities in ebola hemorrhagic fever: overexpression of tissue factor in
primate monocytes/macrophages is a key event. J. Infect. Dis. 188,
1618–1629 (2003).
50. Geisbert, T. W. et al. Pathogenesis of Ebola hemorrhagic fever in
cynomolgus macaques: evidence that dendritic cells are early and
sustained targets of infection. Am. J. Pathol. 163, 2347–2370 (2003).
51. Li, Y. et al. Pan-cancer characterization of immune-related lncRNAs
identifies potential oncogenic biomarkers. Nat. Commun. 11,
1–13 (2020).
52. Zhao, H., Maoshan, C. & Ulf, P. Distinct temporal changes in host
cell lncRNA expression during the course of an adenovirus infec-
tion. Virology 492, 242–250 (2016).
53. Waickman, A. T. et al. Dissecting the heterogeneity of DENV
vaccine-elicited cellular immunity using single-cell RNA sequen-
cing and metabolic profiling. Nat. Commun. 10, 1–16 (2019).
54. Cheng, Y. et al. Dysregulated lncRNAs are involved in the progress
of sepsis by constructing regulatory networks in whole blood cells.
Front. Pharmacol. 12, 678256 (2021).
55. Ma, H. et al. The long noncoding RNA NEAT1 exerts antihantaviral
effects by acting as positive feedback for RIG-I signaling. J. Virol. 91,
e02250-16 (2017).
56. Zhang, Q., Chen, C.-Y., Yedavalli, V. S. R. K. & Jeang, K.-T. NEAT1
long noncoding RNA and paraspeckle bodies modulate HIV-1
posttranscriptional expression. MBio 4, e00596–12 (2013).
57. Prinz, F., Kapeller, A., Pichler, M. & Klec, C. The implications of the
long non-coding RNA NEAT1 in non-cancerous diseases. Int. J. Mol.
Sci. 20, 627 (2019).
58. Pandey, A. D. et al. Correlation of altered expression of a long non-
coding RNA, NEAT1, in peripheral blood mononuclear cells with
dengue disease progression. J. Infect. 75, 541–554 (2017).
59. Bayyurt, B., Bakir, M., Engin, A., Oksuz, C. & Arslan, S. Investigation
of NEAT1, IFNG-AS1, and NRIR expression in Crimean-Congo
hemorrhagic fever. J. Med. Virol. 93, 3300–3304 (2021).
60. Gil, N. & Ulitsky, I. Regulation of gene expression by cis-acting long
38. Bennett, R. S. et al. Kikwit Ebola virus disease progression in the
non-coding RNAs. Nat. Rev. Genet. 21, 102–117 (2019).
rhesus monkey animal model. Viruses 12, 753 (2020).
39. Gierahn, T. M. et al. Seq-Well: portable, low-cost RNA sequencing of
single cells at high throughput. Nat. Methods 14, 395–398
(2017).
40. Datasets. 10x Genomics. https://www.10xgenomics.com/
resources/datasets?query=33k&page=1&configure%5Bfacets%5D
%5B0%5D=chemistryVersionAndThroughput&configure%5Bfacets
%5D%5B1%5D=pipeline.version&configure%5BhitsPerPage%5D=
500&configure%5BmaxValuesPerFacet%5D=1000 (2016).
61. Bagga, S. & Bouchard, M. J. in Cell Cycle Control: Mechanisms and
Protocols (eds. Noguchi, E. & Gadaleta, M. C.) 165–227 (Springer
New York, 2014).
62. Kota, K. P. et al. High content image based analysis identifies cell
cycle inhibitors as regulators of Ebola virus infection. Viruses 4,
1865–1877 (2012).
63. Breuer, K. et al. InnateDB: systems biology of innate immunity and
beyond–recent updates and continuing curation. Nucleic Acids
Res. 41, D1228–D1233 (2013).
41. Kryuchkova-Mostacci, N. & Robinson-Rechavi, M. A benchmark of
gene expression tissue-specificity metrics. Brief. Bioinform. 18,
205–214 (2017).
64. Chiou, K. L. et al. Rhesus macaques as a tractable physiological
model of human ageing. Philos. Trans. R. Soc. Lond. B Biol. Sci. 375,
20190612 (2020).
Nature Communications |
(2023) 14:3866
12
Article
https://doi.org/10.1038/s41467-023-39627-7
65. Logsdon, G. A., Vollger, M. R. & Eichler, E. E. Long-read human
90. Finak, G. et al. MAST: a flexible statistical framework for
genome sequencing and its applications. Nat. Rev. Genet. 21,
597–614 (2020).
66. Liu, S. J. et al. Single-cell analysis of long non-coding RNAs in the
developing human neocortex. Genome Biol. 17, 67 (2016).
67. Johnsson, P. et al. Transcriptional kinetics and molecular functions
of long noncoding RNAs. Nat. Genet. 54, 306–317 (2022).
68. Statello, L., Guo, C.-J., Chen, L.-L. & Huarte, M. Gene regulation by
long non-coding RNAs and its biological functions. Nat. Rev. Mol.
Cell Biol. 22, 96–118 (2021).
69. Regev, A. et al. The human cell atlas. Elife 6, e27041 (2017).
70. Luke, T. et al. Fully human immunoglobulin G from transchromo-
somic bovines treats nonhuman primates infected with Ebola virus
Makona isolate. J. Infect. Dis. 218, S636–S648 (2018).
71. Matranga, C. B. et al. Enhanced methods for unbiased deep
sequencing of Lassa and Ebola RNA viruses from clinical and bio-
logical samples. Genome Biol. 15, 519 (2014).
72. Levin, J. Z. et al. Comprehensive comparative analysis of strand-
specific RNA sequencing methods. Nat. Methods 7, 709–715 (2010).
73. MacConaill, L. E. et al. Unique, dual-indexed sequencing adapters
with UMIs effectively eliminate index cross-talk and significantly
improve sensitivity of massively parallel sequencing. BMC Genom.
19, 30 (2018).
74. Kim, D., Paggi, J. M., Park, C., Bennett, C. & Salzberg, S. L. Graph-
based genome alignment and genotyping with HISAT2 and HISAT-
genotype. Nat. Biotechnol. 37, 907–915 (2019).
75. Wang, L., Wang, S. & Li, W. RSeQC: quality control of RNA-seq
experiments. Bioinformatics 28, 2184–2185 (2012).
76. Li, H. et al. The sequence alignment/Map format and SAMtools.
Bioinformatics 25, 2078–2079 (2009).
77. Smith, T., Heger, A. & Sudbery, I. UMI-tools: modeling sequencing
errors in Unique Molecular Identifiers to improve quantification
accuracy. Genome Res. 27, 491–499 (2017).
78. Pertea, M. et al. StringTie enables improved reconstruction of a
transcriptome from RNA-seq reads. Nat. Biotechnol. 33,
290–295 (2015).
79. Wang, L. et al. CPAT: Coding-Potential Assessment Tool using an
alignment-free logistic regression model. Nucleic Acids Res. 41,
e74–e74 (2013).
80. Kang, Y.-J. et al. CPC2: a fast and accurate coding potential calcu-
lator based on sequence intrinsic features. Nucleic Acids Res. 45,
W12–W16 (2017).
81. Guo, J.-C. et al. CNIT: a fast and accurate web tool for identifying
protein-coding and long non-coding transcripts based on intrinsic
sequence composition. Nucleic Acids Res. 47, W516–W522 (2019).
82. Chen, J. et al. Evolutionary analysis across mammals reveals distinct
classes of long non-coding RNAs. Genome Biol. 17, 1–17 (2016).
83. Di Tommaso, P. et al. Nextflow enables reproducible computational
workflows. Nat. Biotechnol. 35, 316–319 (2017).
84. Wolock, S. L., Lopez, R. & Klein, A. M. Scrublet: computational
identification of cell doublets in single-cell transcriptomic data. Cell
Syst. 8, 281–291.e9 (2019).
85. Stuart, T. et al. Comprehensive integration of single-cell data. Cell
177, 1888–1902.e21 (2019).
86. Luecken, M. D. & Theis, F. J. Current best practices in single-cell
RNA-seq analysis: a tutorial. Mol. Syst. Biol. 15, e8746 (2019).
87. Eisenberg, E. & Levanon, E. Y. Human housekeeping genes, revis-
ited. Trends Genet. 29, 569–574 (2013).
88. Moein, S., Javanmard, S. H., Abedi, M., Izadpanahi, M. H. & Gheisari,
Y. Identification of appropriate housekeeping genes for gene
expression analysis in long-term hypoxia-treated kidney cells. Adv.
Biomed. Res. 6, 15 (2017).
89. Curina, A. et al. High constitutive activity of a broad panel of
assessing transcriptional changes and characterizing hetero-
geneity in single-cell RNA sequencing data. Genome Biol. 16,
278 (2015).
91. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a
practical and powerful approach to multiple testing. J. R. Stat. Soc.
57, 289–300 (1995).
92. Moerman, T. et al. GRNBoost2 and Arboreto: efficient and scalable
inference of gene regulatory networks. Bioinformatics 35,
2159–2161 (2019).
93. Blondel, V. D., Guillaume, J.-L., Lambiotte, R. & Lefebvre, E. Fast
unfolding of communities in large networks. J. Stat. Mech. 77,
P10008 (2008).
94. Wu, T. et al. clusterProfiler 4.0: a universal enrichment tool for
interpreting omics data. Innovation 2, 100141 (2021).
Acknowledgements
We thank Kaia Mattioli for her thoughtful comments on the manuscript
and Aida Ripoll-Cladellas for feedback and helpful discussions. This
material was based upon work supported by Grant RYC-2017-22249
funded by MCIN/AEI/10.13039/501100011033 and Grant PID2019-
107937GA-I00 funded by MCIN/AEI/10.13039/501100011033 (M.M.), the
Howard Hughes Medical Institute Investigator Award (P.C.S.), the
National Institute of Allergy and Infectious Diseases (NIAID) U19AI110818,
the US Food and Drug Administration (FDA) contract
HHSF223201810172C. We acknowledge SAB Biotherapeutics as partners
for providing study materials from the study by Luke et al.70 and for their
collaborative support that allowed the study’s success. Figures 1a, 1c, 4a
and Supplementary Fig. 11A were created with BioRender.com.
Author contributions
L.S. and M.S.R. performed the computational analysis. M.M. designed
the project. J.L.R. contributed to the study design. M.M. and R.G.P.
supervised the analysis. L.S., M.S.R., M.M., and R.G.P. wrote the manu-
script. A.E.L., G.C.A., K.G.B., K.J.S., and S.W. did all the experimental
work. F.R. contributed to the design of the cell-type specificity score.
L.E.H., R.S.B., and P.C.S. designed and led all experimental work. All
authors have read and approved the manuscript for publication.
Competing interests
SAB Biotherapeutics, Inc. provided the study materials from the study by
Luke et al. None of the authors of this study has financial interest in SAB
Biotherapeutics, Inc. company. P.C.S. is a co-founder of, shareholder in,
and advisor to Sherlock Biosciences, Inc.; a board member of and
shareholder in the Danaher Corporation; and a co-founder of and
shareholder in Delve Bio. The other authors declare no competing
interests.
Additional information
Supplementary information The online version contains
supplementary material available at
https://doi.org/10.1038/s41467-023-39627-7.
Correspondence and requests for materials should be addressed to Lisa
E. Hensley, Pardis C. Sabeti or Marta Melé.
Peer review information Nature Communications thanks the anon-
ymous reviewer(s) for their contribution to the peer review of this work.
A peer review file is available.
Reprints and permissions information is available at
http://www.nature.com/reprints
housekeeping and tissue-specific cis-regulatory elements depends
on a subset of ETS proteins. Genes Dev. 31, 399–412 (2017).
Publisher’s note Springer Nature remains neutral with regard to jur-
isdictional claims in published maps and institutional affiliations.
Nature Communications |
(2023) 14:3866
13
Article
https://doi.org/10.1038/s41467-023-39627-7
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as
long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate if
changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons license and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.
© The Author(s) 2023
Nature Communications |
(2023) 14:3866
14
| null |
10.15252_embj.2022112987.pdf
|
Data availability
Original high resolution Z stacks for all images used in figures have
been deposited in the BioImage Archive: accession number
S-BIAD651 (https://www.ebi.ac.uk/biostudies/BioImages/studies/
S-BIAD651?query=S-BIAD651).
|
Data information: ****P < 0.0001; ***P < 0.001; *P < 0.05; ns, not significant. For panel B significance was determined using a one-way ANOVA; for all other panels significance was determined using an unpaired t-test. All images in this figure are maximum intensity projections. Scale bars = 10 lm. Source data are available online for this figure. Data information: ****P < 0.0001; ns, not significant. For panel A (right graph) significance was determined using a one-way ANOVA; for all other panels significance was determined using an unpaired t-test. All images in this figure are maximum intensity projections. Scale bars = 10 lm. Source data are available online for this figure. Data information: ****P < 0.0001; ns, not significant. For the ogt and oga mutants significance was determined using a one-way ANOVA; for all other conditions significance was determined using an unpaired t-test. All the images in this figure are maximum intensity projections. Scale bars = 10 lm. Source data are available online for this figure. Data availability Original high resolution Z stacks for all images used in figures have been deposited in the BioImage Archive: accession number S-BIAD651 ( https://www.ebi.ac.uk/biostudies/BioImages/studies/ S-BIAD651?query=S-BIAD651 ). Expanded View for this article is available online .
|
Article
Nucleoporin foci are stress-sensitive condensates
dispensable for C. elegans nuclear pore assembly
, Basma Taleb Ismail1, Peter Askjaer2 & Geraldine Seydoux1,*
Laura Thomas1
Abstract
Nucleoporins (Nups) assemble nuclear pores that form the perme-
ability barrier between nucleoplasm and cytoplasm. Nucleoporins
also localize in cytoplasmic foci proposed to function as pore pre-
assembly intermediates. Here, we characterize the composition
and incidence of cytoplasmic Nup foci
in an intact animal, C.
elegans. We find that, in young non-stressed animals, Nup foci only
appear in developing sperm, oocytes and embryos, tissues that
express high levels of nucleoporins. The foci are condensates of
highly cohesive FG repeat-containing nucleoporins (FG-Nups),
which are maintained near their solubility limit in the cytoplasm
by posttranslational modifications and chaperone activity. Only a
minor fraction of FG-Nup molecules concentrate in Nup foci, which
dissolve during M phase and are dispensable for nuclear pore
assembly. Nucleoporin condensation is enhanced by stress and
advancing age, and overexpression of a single FG-Nup in post-
mitotic neurons is sufficient to induce ectopic condensation and
organismal paralysis. We speculate that Nup foci are non-essential
and potentially toxic condensates whose assembly is actively
suppressed in healthy cells.
Keywords aging; C. elegans; condensate; nucleoporin; oocyte
Subject Categories Organelles; Translation & Protein Quality
DOI 10.15252/embj.2022112987 | Received 2 November 2022 | Revised 2 May
2023 | Accepted 10 May 2023 | Published online 31 May 2023
The EMBO Journal (2023) 42: e112987
Introduction
In all eukaryotes, the double-membraned nuclear envelope (NE)
partitions the nucleoplasm from the cytoplasm and material
is
exchanged between the two compartments by way of nuclear
pore complexes. Nuclear pore complexes are composed of at least
30 distinct nucleoporins (Nups) arranged in biochemically stable
subcomplexes (Fig 1A; Cohen-Fix & Askjaer, 2017; Hampoelz
et al, 2019a). Approximately, two-thirds of Nups are essential to
scaffold and anchor pore complexes to the NE. The remaining
one-third contain large phenylalanine/glycine (FG) rich domains
that are highly intrinsically disordered. FG-Nups are enriched in
the central channel of the pore and form multivalent interactions
in vivo and in vitro (Frey et al, 2006; Patel et al, 2007; Labokha
et al, 2012; Xu & Powers, 2013). In the ‘selective phase’ model of
transport selectivity,
the permeability barrier is established by
cohesive interactions among FG-Nups that form a phase separated
network (Ribbeck & Go¨rlich, 2001; Schmidt & Go¨rlich, 2016). In
support of this model, interactions among FG-Nups are critical for
the formation of the permeability barrier and FG-Nup hydrogels
recapitulate nuclear pore selectivity in vitro (Strawn et al, 2004;
Frey & Go¨rlich, 2007; H€ulsmann et al, 2012; Schmidt &
Go¨rlich, 2015; Ng et al, 2021).
In addition to their localization at the NE, Nups have been
observed in discrete cytoplasmic foci in yeast, oocytes, and animal
cell types cultured in vitro (Cordes et al, 1996; Colombi et al, 2013;
Raghunayakula et al, 2015; Ren et al, 2019). Cytoplasmic Nup foci
have been implicated in miRNA-mediated mRNA repression (Sahoo
et al, 2017), nuclear pore inheritance (Colombi et al, 2013), and
pore assembly by a condensate-based, non-canonical mechanism
that generates annulate lamellae (Hampoelz et al, 2019b). Annulate
lamellae are a specialized subdomain of the endoplasmic reticulum
(Kessel, 1989) proposed to function as a source of ready-made pore
complexes in rapidly dividing cells (Hampoelz et al, 2016; Ren
et al, 2019). Although some have argued against a stockpiling func-
tion (Stafstrom & Staehelin, 1984; Onischenko et al, 2004), annulate
lamellae generated in oocytes have been proposed to fuel the expan-
sion of nuclear membranes in Drosophila embryos (Hampoelz
et al, 2016, 2019b).
Nups are also frequently enriched in pathological cytoplasmic
inclusions
that are hallmarks of neurodegenerative disease
(reviewed in Fallini et al, 2020; Hutten & Dormann, 2020; Chandra
& Lusk, 2022), leading to the proposal that Nups become seques-
tered and depleted from nuclear pores under disease conditions
(Zhang et al, 2018; Gasset-Rosa et al, 2019). Given the inherent pro-
pensity of FG-Nups to form multivalent networks, it is possible that
Nup condensation directly contributes to protein aggregation in dis-
ease. In support of this hypothesis, condensation of FG-Nup fusion
oncogenes drives certain cancers (Zhou & Yang, 2014; Terlecki-
Zaniewicz et al, 2021; Chandra et al, 2022), cytoplasmic Nup gran-
ules form upon loss of
fragile X-related proteins (Agote-Aran
et al, 2020), and cytoplasmic FG-Nups drive aggregation of TDP-43
in ALS/FTLD and following traumatic brain injury (Anderson
1 HHMI and Department of Molecular Biology and Genetics, Johns Hopkins University School of Medicine, Baltimore, MD, USA
2 Andalusian Center for Developmental Biology (CABD), CSIC/JA/Universidad Pablo de Olavide, Seville, Spain
*Corresponding author. Tel: +1 410 614 4622; E-mail: [email protected]
(cid:1) 2023 The Authors. Published under the terms of the CC BY 4.0 license
The EMBO Journal 42: e112987 | 2023
1 of 19
The EMBO Journal
Laura Thomas et al
et al, 2021; Gleixner et al, 2022). These observations suggest that
Nups are not passive clients of cytoplasmic inclusions but rather
active promoters of protein aggregation and disease progression.
Cytoplasmic Nup foci were reported previously in C. elegans
oocytes and embryos (Pitt et al, 2000; Jud et al, 2007; Sheth
et al, 2010; Patterson et al, 2011). In this study, we use the C.
elegans model to systematically investigate the origin, regulation,
and function of Nup foci. We find that, in addition to oocytes and
embryos, Nup foci form in developing sperm and in the somatic tis-
sues of aged animals. We find that the majority of Nup foci are con-
densates of FG-Nups, which are maintained in a mostly soluble
cytoplasmic pool by posttranslational modifications and the chaper-
one activity of nuclear transport receptors (NTRs). Condensation is
enhanced by heat stress and FG-Nup overexpression, which when
induced in neurons can disrupt nuclear pore assembly and lead to
organismal paralysis. Our findings suggest that Nup foci are inciden-
tal byproducts of the natural tendency of FG-Nups to undergo con-
densation, which is required to generate the permeability barrier of
nuclear pores but must be suppressed in the cytoplasm to avoid pre-
mature and potentially toxic condensation.
Results
In young animals Nup foci assemble only in growing oocytes,
developing sperm, and early embryos
Cytoplasmic Nup foci have been observed in C. elegans oocytes and
early embryos using the mAb414 antibody (Davis & Blobel, 1986;
Pitt et al, 2000; Jud et al, 2007). To characterize the distribution of
Nup foci across all C. elegans tissues, we used two Nups, Nup358
and Nup88, which have been reported in cytoplasmic foci in Dro-
sophila oocytes, yeast, and a range of cultured cell types from differ-
ent organisms (Cordes et al, 1996; Wu et al, 2001; Colombi
et al, 2013; Raghunayakula et al, 2015; Sahoo et al, 2017; Hampoelz
et al, 2019b). We used CRISPR genome engineering to tag Nup358
and Nup88 at their endogenous loci and examined their distribution
in all
tissues across hermaphrodite development (Fig 1A). As
expected, both Nups localized to the NE in all cell types, including
muscle, hypodermis, intestine, neurons, and germ cells (Fig 1B,
Appendix Fig S1A). In the germline of hermaphrodites, germ cell
nuclei proliferate in a syncytial cytoplasm before individualizing to
produce sperm during the fourth larval (L4) stage and oocytes in
adults (Fig 1A, Appendix Fig S1B). We detected Nups in cytoplasmic
foci (“Nup foci”) in the residual body of spermatocytes, a transient
structure that accumulates components discarded during spermato-
genesis (Appendix Fig S1B). We also detected Nup358 and Nup88 in
Nup foci in growing oocytes and in early embryos (<~80-cell stage;
Fig 1B and C, Appendix Fig S1A and C). The intensity of Nup foci in
oocytes increased between days 1 and 2 of adulthood (Appendix
Fig S1D). In contrast, we did not detect Nup foci in somatic cells at
any stage through Day 2 of adulthood. The cytoplasmic concentra-
tion of Nups in germ cells and early embryos was ~3-5-fold higher
than that observed in somatic cells (Fig 1D). We conclude that, in
developing animals and young adults, Nup foci only form in gam-
etes and early embryos, which accumulate higher levels of cytoplas-
mic Nups compared with somatic tissues.
Nup foci in growing oocytes contain FG-Nups and their binding
partners, but not transmembrane, inner ring complex, or
nucleoplasmic Nups
Nup foci have been proposed to correspond to (i) condensates
containing pore assembly intermediates, or (ii) mature pore com-
plexes in membranous annulate lamellae (Raghunayakula et al,
2015; Hampoelz et al, 2016, 2019b; Ren et al, 2019). To systemati-
cally compare the composition and stoichiometry of Nup foci to that
of mature nuclear pore complexes at the NE, we used a collection of
genomically-encoded tags, transgenes, and antibodies against 16
Nups (including representatives of each nuclear pore subcomplex)
as well as the Nup358 binding partners RanGAP and NXF1 (Fig 2A,
Appendix Table S1). We examined Nup distribution in growing
oocytes of Day 2 adult wild-type hermaphrodites where Nup foci are
prominent.
As expected, all Nups tested localized to the NE (Fig 2B, Appen-
dix Fig S2A). Nuclear basket and Y complex Nups additionally local-
ized to the nucleoplasm and meiotic chromosomes, respectively, as
previously described (Gómez-Saldivar et al, 2016; Hattersley
et al, 2016). Only a subset of Nups localized to cytoplasmic foci,
including FG-Nups of the central channel and cytoplasmic filaments
Figure 1. Cytoplasmic Nup foci are not present in somatic cells of young animals.
A Left: Schematic depicting the structure of a nuclear pore complex, which consists of ~30 Nup proteins arranged in distinct subcomplexes. Blue subcomplexes are
structural elements of the pore and include transmembrane Nups, the inner ring complex, and two copies of the Y complex. FG domain Nups are designated in
orange and generate the permeability barrier of the central channel, and additionally localize to cytoplasmic filaments and the nuclear basket. Right: Schematic
depicting the tissues and germline organization of a C. elegans adult hermaphrodite. Germ cell nuclei (designated in blue) proliferate in a syncytial cytoplasm before
becoming enclosed by membrane to form individual oocytes. Oocytes arrest in meiosis I and grow in an assembly line-like fashion until induced by sperm signaling
to re-enter the cell cycle in preparation for fertilization. NEBD: nuclear envelope breakdown.
B Representative confocal micrographs of CRISPR-tagged mNeonGreen::Nup358 in the intestine, hypodermis, body wall muscle, head, and oocytes of Day 2 adult C.
elegans. Nuclei are marked by a mCherry::histone transgene. White arrows denote cytoplasmic foci in oocytes.
C Representative confocal micrographs showing mNeonGreen::Nup358 in interphase 4-cell versus ~80-cell embryos. White arrowheads denote polar bodies (meiotic
products).
D Left: Representative confocal micrograph of mNeonGreen::Nup358 in -2 and -3 oocytes and intestinal cells of a Day 2 adult. Red dashed lines denote intestinal cells,
gray dashed lines outline oocytes. Right: Quantification of cytoplasmic (soluble) mNeonGreen::Nup358 signal in intestinal, hypodermal, muscle, head (pharyngeal), or
early (4-cell) embryonic cells as compared to that of the -1 oocyte. Values are normalized within the same animal so that the measurement for the -1 oocyte = 1.0.
Error bars represent 95% CI for n > 7 animals (biological replicates).
Data information: ****P < 0.0001; ***P < 0.001. Significance was determined using a one-way ANOVA. All images in this figure are maximum intensity projections. Scale
bars = 10 lm.
Source data are available online for this figure.
▸
2 of 19
The EMBO Journal 42: e112987 | 2023
(cid:1) 2023 The Authors
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Laura Thomas et al
The EMBO Journal
A
Y complex
cytosol
cytoplasmic
/ filament
inner
ring
complex
channel
trans-
membrane
nucleus
nucleoplasmic
/ basket
Germline
Body wall
muscle
Intestine
Hypodermis
Head
(neurons)
syncytial cytoplasm
oocytes
distal germline
- Nuclei
4-cell embryo
fertilization
-6
-4
-5
oocyte growth
-3
-1
-2
re-entry to cell
cycle and NEBD
B
mNeon::Nup358; mCherry::histone (Day 2 adults)
Intestine
Hypodermis
Body wall muscle
Head (neurons)
-6
-5
Oocytes
-4
-3
C
mNeon::Nup358; mCherry::histone
D
mNeon::Nup358 (Day 2 adult)
Intestine
****
****
****
i
8
5
3
p
u
N
c
m
s
a
p
o
t
y
C
l
****
***
1.5
1.0
0.5
0.0
-3
Oocytes
-2
embryo
hypodermis
pharynx
body wall muscle
-1 oocyte
intestine
4-cell embryo
~80-cell embryo
Figure 1.
(cid:1) 2023 The Authors
The EMBO Journal 42: e112987 | 2023
3 of 19
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
The EMBO Journal
A
Y complex
Nup96 (NPP-10C)
Nup85 (NPP-2)
Nup107 (NPP-5)
cytosol
cytoplasmic / filament
Nup358 (NPP-9)
Nup214 (NPP-14)
Nup88 (NPP-24)
inner ring
complex
Nup35 (NPP-19)
channel
Nup98 (NPP-10N)
Nup62 (NPP-11)
Nup54 (NPP-1)
transmembrane
gp210 (NPP-12)
NDC1 (NPP-22)
nucleus
nucleoplasmic / basket
Nup153 (NPP-7)
ELYS (MEL-28)
TPR (NPP-21)
Nup50 (NPP-16)
mNeon Nup
Day 2 adults
Nup62::wrmScarlet
Merge
Laura Thomas et al
B
Day 2 adults
mNeon::Nup358
Nup214::OLLAS
GFP::Nup88
mNeon::Nup98
Nup62::wrmScarlet
mCherry::Nup54
anti-Nup96
GFP::Nup85
GFP::Nup107
GFP::Nup35
gp210::mNeon
NDC1::wrmScarlet
anti-Nup153
GFP::ELYS
TPR::GFP
anti-Nup50
D
GFP::Nup88
Day 2 adult
HaloTag::HDEL
Merge
NE
Foci
**
***
ns
ns
****
****
5
4
3
2
1
0
/
p
u
N
n
e
e
r
G
t
e
l
r
a
c
S
m
w
r
:
:
2
6
p
u
N
-1
Nup88
Nup358
Nup98
Nup85
gp210
ELYS
C
8
5
3
p
u
N
8
9
p
u
N
0
1
2
p
g
Figure 2. Cytoplasmic Nup foci primarily contain FG-Nups and their binding partners.
A Schematic depicting the nuclear pore location of the Nups examined in this study. Nups are designated using human nomenclature, followed by the C. elegans
homolog in parentheses. Nups listed in orange localize to cytoplasmic foci in growing oocytes, whereas those denoted in black do not.
B Representative confocal micrographs of the -3 and -4 oocytes from Day 2 adult C. elegans expressing tagged versions of each indicated Nup or stained with anti-Nup
antibodies. All images are maximum intensity projections, with the exception of gp210 and NDC1 which are single imaging planes. Orange labels designate Nups
enriched in cytoplasmic foci.
C Left: Representative confocal micrographs of Day 2 adult oocytes comparing the localization of CRISPR-tagged Nup62::wrmScarlet to mNeonGreen-tagged Nup358,
Nup98 and gp210. Right: Quantification of the overlap between Nup62::wrmScarlet and each indicated Nup at the nuclear envelope (NE) versus cytoplasmic foci. Each
point designates an individual nucleus or focus. Values are normalized so that the average ratio at the nuclear envelope = 1.0. Error bars represent 95% CI for n > 7
(nuclei) or n > 59 (foci).
D Representative confocal micrographs showing partial overlap of CRISPR-tagged GFP::Nup88 with the luminal endoplasmic reticulum/nuclear envelope marker
HaloTag::HDEL in a Day 2 adult oocyte. 20% of foci completely overlapped with HaloTag::HDEL, 64% partially overlapped, and 16% showed no overlap with HaloTag::
HDEL (n = 118, see Materials and Methods). Areas indicated by white boxes are magnified below; white arrows indicate foci that do not completely overlap with the
endoplasmic reticulum.
Data information: ****P < 0.0001; ***P < 0.001; **P < 0.01; ns, not significant. Significance was determined using an unpaired t-test. Scale bars = 10 lm (panel B) or
5 lm (panels C and D).
Source data are available online for this figure.
(Nup62, Nup98, Nup214, and Nup358) and their binding partners
(Y complex Nups, Nup88, RanGAP, and NXF1; Fig 2B, Appendix
Fig S2A and B). The transmembrane Nups gp210 and NCD1 could
be detected throughout the endoplasmic reticulum as described pre-
viously (Galy et al, 2008; Huelgas-Morales et al, 2020; Mauro
et al, 2022), but did not enrich in foci, nor did Nup35, an inner ring
complex Nup. All nucleoplasmic-facing Nups (Nup153, Nup50,
TPR, and ELYS) were enriched in the nucleoplasm and absent from
cytoplasmic foci. The assembly-line arrangement of the C. elegans
germline allowed us to visualize Nup distribution throughout oocyte
growth and maturation. We found that nucleoplasmic Nups and the
inner ring complex component Nup35 never became incorporated
into Nup foci during oocyte growth (Appendix Fig S2C). We also
analyzed the distribution of Nups in 4-cell stage early embryos and
obtained the same results except for Nup35, which did not localize
to foci in oocytes but did in embryos (Appendix Fig S2D and E). We
conclude that Nup foci
in growing oocytes primarily enrich
cytoplasm-facing FG-Nups and their binding partners (Fig 2A).
Co-staining experiments using the mAb414 antibody, which in
vertebrates recognizes Nup62, Nup153, Nup214, and Nup358,
suggested that Nup foci contain multiple Nups (Appendix Fig S2A
and E). To examine Nup stoichiometry in the foci, we crossed a sub-
set of GFP-tagged Nups pairwise with Nup62::wrmScarlet. As
expected, all Nups tested colocalized with Nup62::wrmScarlet at the
4 of 19
The EMBO Journal 42: e112987 | 2023
(cid:1) 2023 The Authors
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Laura Thomas et al
The EMBO Journal
NE (Fig 2C). Nups that localize to cytoplasmic foci (Nup85, Nup88,
Nup98, and Nup358) additionally colocalized with Nup62::
wrmScarlet in all foci. Quantification of the ratio of the GFP-tagged
Nup to Nup62::wrmScarlet revealed that each Nup accumulates in
fixed stoichiometry relative to Nup62 at the NE. In contrast, Nups
exhibited variable stoichiometry in the cytoplasmic foci (Fig 2C).
We also found that only 20% of GFP::Nup88 foci in growing oocytes
fully overlapped with a marker for endoplasmic reticulum mem-
branes (Fig 2D). This is consistent with our observation that gp210
and NDC1, which both localize to the endoplasmic reticulum (Galy
et al, 2008; Mauro et al, 2022), are not enriched in Nup foci. We
conclude that the majority of Nup foci
in growing oocytes are
unlikely to correspond to stockpiled mature pores, as they lack criti-
cal pore scaffolds, exhibit variable Nup stoichiometry, and rarely
associate with endoplasmic reticulum membranes.
Nup foci are condensates scaffolded by excess FG-Nups
In vitro, FG-Nups condense into hydrogels (Frey et al, 2006;
Labokha et al, 2012; Schmidt & Go¨rlich, 2015) raising the possibility
that cytoplasmic Nup foci might form by spontaneous condensation
of FG-Nups in the saturated environment of the oocyte. Condensa-
tion is highly sensitive to concentration: proteins de-mix into dense
and dilute phases when their concentration exceeds the saturation
concentration (csat), the maximum concentration allowed in the sol-
uble, dilute phase (Alberti et al, 2019).
To estimate the percent of Nup molecules that undergo conden-
sation, we used Imaris software to quantify Nup fluorescence in
nuclei, the cytoplasm, and cytoplasmic foci (Appendix Fig S3A and
see Materials and Methods). Remarkably, we found that the vast
majority of Nups distribute between a nuclear pool (~30–40%) and
a diffuse cytoplasmic pool (~60–70%), with less than 3% of Nup
molecules in foci (Fig 3A). The soluble cytoplasmic pool is the least
concentrated but largest by volume and is readily visualized in sum
projection micrographs (Appendix Fig S3B). These observations
suggest that FG-Nups are maintained in oocytes at concentrations
just in excess of saturation, such that most molecules are soluble
and only a minority condense in the foci. If so, we predicted that
removal of individual FG-Nups may be sufficient to drop below the
threshold for condensation and reduce foci formation.
We used RNAi and mutagenesis to systematically deplete individ-
ual Nups and examined the effect on Nup foci
formation. As
expected, depletion of non-FG or nucleoplasmic Nups, which are not
present in foci, had no effect on foci formation (Fig 3B, Appendix
Fig S3C–E). In contrast, depletion of individual cytoplasm-facing FG-
Nups (Nup62, Nup98, Nup214, or Nup358) reduced the formation of
Nup foci by >95% without affecting Nup levels at the NE. Depletion
of Nup88, which is structured but interacts with multiple subcom-
plexes containing FG-Nups (Fornerod et al, 1997; Griffis et al, 2003;
Xylourgidis et al, 2006; Yoshida et al, 2011), reduced Nup foci by
~70%, suggesting that interactions among FG-Nup subcomplexes
contribute to foci
formation. As expected for structural Nups
(Mansfeld et al, 2006; Stavru et al, 2006; Onischenko et al, 2009;
R(cid:2)odenas et al, 2009; Mauro et al, 2022), loss of Nup35 or the trans-
membrane Nups NDC1 or gp210 decreased Nup levels at the NE
(Fig 3B and C, Appendix Fig S3C and F), and enhanced foci forma-
tion, presumably because impaired pore assembly liberates FG-Nups
to the cytoplasm. We conclude that Nup foci assembly in oocytes
depends primarily on the cumulative effect of high concentrations of
the FG-Nups Nup62, Nup98, Nup214, and Nup358 in the cytoplasm.
To directly test whether high levels of FG-Nups are sufficient to
drive foci formation, we generated a transgenic strain with an extra
copy of nup214::wrmScarlet expressed under the control of the
germline-specific mex-5 promoter (Fan et al, 2020). We found that
overexpression of Nup214::wrmScarlet was sufficient to increase the
proportion of endogenous mNeonGreen::Nup358 in Nup foci by 4-
fold (Fig 3D, Appendix Fig S3G).
In vitro, some Nup98 FG-domain hydrogels have been shown to be
dissolved by the aliphatic alcohol 1,6-hexanediol
(Schmidt &
Go¨rlich, 2015), which disrupts hydrophobic interactions and has been
reported to dissolve Nup foci in yeast, Drosophila, and HeLa cells
(Patel et al, 2007; Hampoelz et al, 2019b; Agote-Aran et al, 2020). As
expected, we found that hexanediol treatment reduced the intensity of
Nup foci in embryos (Appendix Fig S3H). We conclude that Nup foci
are FG-Nup condensates that arise when the cytoplasmic concentra-
tion of FG-Nups exceeds the saturation concentration.
Nup foci assembly is enhanced by oocyte arrest, heat stress,
and aging
Oocyte production occurs continuously in young hermaphrodites
such that fully grown oocytes are immediately ovulated and fertil-
ized. In contrast, in animals lacking sperm, fully grown oocytes
arrest and are stored in the oviduct. Electron microscopy studies
have reported annulate lamellae in ~10% of arrested oocytes in
C. elegans (and 42% of arrested oocytes in the related nematode
C. remanei), but not in the growing oocytes of hermaphrodites or in
embryos (Pitt et al, 2000; Patterson et al, 2011; Langerak et al,
2019). To examine Nup foci in arrested oocytes, we used unmated
fog-2(q71) females which do not produce sperm and accumulate
fully grown arrested oocytes in the oviduct (Schedl & Kimble, 1988).
We observed a 14-fold increase in the percent of GFP::Nup88 in foci
in the arrested oocytes of Day 1 adult fog-2(q71) females compared
to growing oocytes of age-matched wild-type hermaphrodites
(Fig 4A, Appendix Fig S4A). The Nup foci
in arrested oocytes
enriched additional Nups at low levels including Nup35 and ELYS
(Appendix Fig S4B). Furthermore, 42% of Nup foci
in arrested
oocytes overlapped with a marker for endoplasmic reticulum mem-
branes (Appendix Fig S4C), raising the possibility that a subset of
Nup foci in arrested oocytes could correspond to annulate lamellae.
We also observed the formation of Nup-rich ‘blebs’ at the NE of
arrested oocytes (Fig 4A, Appendix Fig S4A). These findings are
consistent with prior studies which found that the abundance of
cytoplasmic Nup foci and nuclear blebs were significantly increased
in arrested oocytes of C. elegans and related nematodes (Jud
et al, 2007; Patterson et al, 2011).
Previous studies have reported parallels between oocyte arrest
and environmental stresses in inducing the formation of conden-
sates in C. elegans oocytes (Jud et al, 2008; Patterson et al, 2011;
Elaswad et al, 2022). In agreement with these findings, we found
that a 20 min shift from 20°C to 30°C was sufficient to increase the
intensity of Nup foci in growing oocytes by 16-fold (Fig 4A, Appen-
dix Fig S4A and D) and significantly increase Nup condensation in
embryos (Fig 4B). In contrast, the same conditions of oocyte arrest
and mild heat stress did not change the distribution of the stress
to higher
granule protein G3BP, which requires
exposure
(cid:1) 2023 The Authors
The EMBO Journal 42: e112987 | 2023
5 of 19
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License The EMBO Journal
Laura Thomas et al
A
B
f
o
e
g
n
a
h
c
d
o
F
l
100
p
u
N
l
a
t
o
t
f
o
%
75
50
25
3
2
1
0
Cytoplasmic
NE/Nucleoplasm
Foci
C
mNeon::Nup358 (Day 1 adults)
wild-type
Nup88
Nup358
Nup98
Nup62
Nup85
Nup35
gp210
NDC1
gp210(cid:2)
ELYS
TPR
GFP::Nup85 (Day 2 adults)
GFP::Nup85 (Day 1 adults)
D
mNeon::Nup358 (Day 1 adults)
control
control
wild-type
nup214 (RNAi)
ndc1 (RNAi)
Nup214 overexpression
f
o
%
ns
****
ns
*
***
****
ns
12
8
4
1.4
0.7
0.0
f
o
e
g
n
a
h
c
d
o
F
l
E
N
t
a
5
8
p
u
N
1.2
0.6
0.0
i
c
o
f
n
i
5
8
p
u
N
Cytoplasmic
NE/Nucleoplasm
Foci
****
****
100
75
50
25
8
5
3
p
u
N
l
a
t
o
t
f
o
%
1.0
0.5
0.0
wt
gp210
Cytoplasmic
NE/Nucleoplasm
Foci
****
*
100
8
5
3
p
u
N
l
a
o
t
t
75
50
25
1.2
0.6
0.0
wt
Nup214
RNAi:
control
nup98
nup88
nup358
nup35
nup214
nup62
ndc1tpr
nup50
nup54
nup107
RNAi:
control
nup88
nup98
nup358
nup62
nup214
nup35
ndc1tpr
nup50
nup54
nup107
Figure 3. Nup foci are condensates scaffolded by cytoplasmic facing FG-Nups.
A Quantification of the distribution of CRISPR-tagged Nups between the cytoplasm (soluble), nuclear envelope (NE)/nucleoplasm, and cytoplasmic foci (see Materials
and Methods). Measurements were made using the -3 and -4 oocytes of Day 2 adults. Error bars represent 95% CI for n > 5 germlines (biological replicates).
B Top: Representative confocal micrographs showing -3 and -4 oocytes of Day 1 or Day 2 adults with CRISPR-tagged GFP::Nup85. nup214 RNAi is representative of a
treatment that largely abolishes Nup foci, whereas ndc1 RNAi enhanced Nup foci. Left graph: Quantification of the total percent of GFP::Nup85 in foci following each
RNAi treatment. Values are normalized so that the average control measurement = 1.0. Error bars represent 95% CI for n > 7 germlines (biological replicates). Right
graph: Line-scan quantification measuring GFP::Nup85 signal at the NE following each RNAi treatment. Values are normalized so that the average control
measurement = 1.0. Error bars represent 95% CI for n > 13 nuclei (biological replicates).
C Left: Representative confocal micrographs showing CRISPR-tagged mNeonGreen::Nup358 in -3 and -4 oocytes of Day 1 wild-type versus gp210Δ adults. Right: Quanti-
fication of the distribution of mNeonGreen::Nup358 between the cytoplasm, NE/nucleoplasm, and cytoplasmic foci in wild-type versus gp210Δ oocytes. Error bars rep-
resent 95% CI for n > 6 germlines (biological replicates).
D Left: Representative confocal micrographs showing mNeonGreen::Nup358 in -3 and -4 oocytes of Day 1 adults with or without overexpression of Nup214::wrmScarlet.
Right: Quantification of the distribution of mNeonGreen::Nup358 between the cytoplasm, NE/nucleoplasm, and cytoplasmic foci in wild-type oocytes versus those
with Nup214 overexpression. Error bars represent 95% CI for n > 9 germlines (biological replicates).
Data information: ****P < 0.0001; ***P < 0.001; *P < 0.05; ns, not significant. For panel B significance was determined using a one-way ANOVA; for all other panels sig-
nificance was determined using an unpaired t-test. All images in this figure are maximum intensity projections. Scale bars = 10 lm.
Source data are available online for this figure.
temperatures to condense (Abbatemarco et al, 2021; Fig 4A).
Together these observations indicate that Nup foci assembly is read-
ily enhanced by mild stress conditions. Consistent with this view,
we also found that Nup foci accumulate in somatic cells in > 90%
of Day 7 adult hermaphrodites, compared to 0% at Day 1 of adult-
hood (Fig 4C and D). We conclude that Nup foci are stress-sensitive
structures that accumulate with age.
Nup foci disassemble during the oocyte-to-embryo transition
and are not required for nuclear pore assembly in embryos
In the presence of sperm, oocytes at the -1 position in the oviduct
initiate meiotic M phase in preparation for ovulation and fertiliza-
tion (Huelgas-Morales & Greenstein, 2018). We found that oocytes
in M phase, whether in hermaphrodites or mated females, lack Nup
6 of 19
The EMBO Journal 42: e112987 | 2023
(cid:1) 2023 The Authors
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Laura Thomas et al
A
Day 1 adults
Day 1 adults
GFP::Nup88
G3BP::RFP
mNeon::Nup358
G3BP::mCherry
Growing
Growing
20°C
20°C
Arrested
Arrested
30°C
30°C
The EMBO Journal
****
****
20
15
10
5
0
growing
arrested
20°C
30°C
i
c
o
f
n
i
p
u
N
f
o
e
g
n
a
h
c
d
o
F
l
B
20°C
anti-Nup358
30°C
C
l
e
c
s
u
m
l
l
a
w
y
d
o
b
Day 1 adult
Day 7 adult
GFP::Nup88; mCherry::histone
D
GFP::Nup88; mCherry::histone
Day 1 adult
d
a
e
h
11/11
Day 7 adult
11/12
Figure 4. Cytoplasmic Nup foci increase with oocyte arrest, heat stress, and age.
A Left: Representative confocal micrographs of growing -3 and -4 oocytes or -3, -4, and -5 arrested oocytes in Day 1 adults expressing GFP::Nup88 or the stress granule
marker G3BP::RFP. GFP::Nup88 expressing oocytes are in wild-type hermaphrodites (growing oocytes) or fog-2(q71) unmated females (arrested oocytes). G3BP::RFP
expressing oocytes are in mated (growing oocytes) or unmated (arrested oocytes) fog-2(q71) females. Middle: Representative confocal micrographs showing CRISPR-
tagged mNeonGreen::Nup358 and G3BP::mCherry in the -3 and -4 growing oocytes of Day 1 adults maintained at 20°C or after a 20 min shift to 30°C. Right: Quantifi-
cation of the percent of GFP::Nup88 in foci in growing versus arrested oocytes or mNeonGreen::Nup358 in foci at 20°C versus 30°C. Error bars represent 95% CI for
n > 7 germlines (growing vs. arrested; biological replicates) or n = 6 germlines (20°C versus 30°C; biological replicates). Values are normalized so that the average
control condition (growing oocytes or 20°C) measurement = 1.0. See Appendix Fig S4A for raw (non-normalized) values of the distribution of Nup between the cyto-
plasm (soluble), nuclear envelope (NE)/nucleoplasm, and cytoplasmic foci for each condition.
B Representative confocal micrographs of endogenous Nup358 in 2-cell interphase embryos grown at 20°C or after shifting to 30°C for 20 min.
C Representative confocal micrographs showing GFP::Nup88 in body wall muscle cells of Day 1 versus Day 7 adults. Nuclei are marked by a mCherry::histone transgene.
D Representative confocal micrographs showing GFP::Nup88 in the head of a Day 1 versus Day 7 adult. Nuclei are marked by a mCherry::histone transgene. Areas
indicated by white boxes are magnified at right. 100% (n = 11) of Day 1 adults lacked foci in somatic cells whereas cytoplasmic foci were observed outside of the
germline in 92% (n = 12) of Day 7 adults.
Data information: ****P < 0.0001. Significance was determined using an unpaired t-test. All images in this figure are maximum intensity projections, with the exception
of G3BP (panel A) which are single focal planes. Scale bars = 10 lm.
Source data are available online for this figure.
foci (Fig 5A, Appendix Fig S5A). In both hermaphrodites and mated
females, Nup levels did not decrease in M phase oocytes lacking foci
(Fig 5A, Appendix Fig S5A), indicating that the absence of visible
foci is not due to Nup degradation. Similarly, Nup foci were absent
in zygotes undergoing meiosis (Fig 5A) and blastomeres undergoing
mitosis, and reappeared during interphase with no change in overall
Nup levels (Fig 5B, Movies EV1 and EV2). Moreover, the concentra-
tion of soluble cytoplasmic Nup increased in M phase oocytes of
hermaphrodites as well as mitotic embryos (Appendix Fig S5B).
These observations suggest that FG-Nup solubility oscillates with
the cell cycle, peaking during M phase, consistent with prior studies
(Pitt et al, 2000; Onischenko et al, 2005). We conclude that Nup foci
are transient structures that are not maintained during the oocyte-
to-embryo transition.
To test whether Nup foci contribute to pore assembly in
embryos, we used CRISPR genome engineering to generate a
complete deletion of the nup214 locus. Consistent with our RNAi
results, using four independent markers (mNeonGreen::Nup358,
GFP::Nup85, RanGAP::wrmScarlet, and mAb414), we found that
Nup foci were greatly reduced in growing oocytes and early
embryos of nup214Δ hermaphrodites (Fig 5C and D, Appendix
Fig S5C–E). We also found that GFP::Nup88 was largely localized to
the cytoplasm in the nup214Δ mutant
(Appendix Fig S5F),
supporting a role for Nup214 in stabilizing and targeting Nup88 to
pore complexes (Xylourgidis et al, 2006). Despite lacking robust
Nup foci, nup214Δ embryos were viable (Appendix Fig S5G). Fur-
thermore, nuclear pore formation was not disrupted in nup214Δ
mutant embryos (Appendix Fig S5H), as evidenced by normal parti-
tioning of cargos between the nucleus and cytoplasm, including the
nuclear RNA-binding protein TDP-43 and an importin b binding
(IBB) domain reporter (Lott & Cingolani, 2011; Fig 5E). We obtained
similar results in a nup88Δ mutant, which also reduces the
(cid:1) 2023 The Authors
The EMBO Journal 42: e112987 | 2023
7 of 19
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
The EMBO Journal
Laura Thomas et al
incidence of Nup foci: nup88Δ mutants were viable and had normal
partitioning of cargo between the nucleus and cytoplasm (Appendix
Fig S5C and I–K). We conclude that robust Nup foci are not essential
for viability or nuclear pore assembly during embryogenesis.
As the number and size of Nup foci increase significantly during
oocyte arrest (Fig 4), we considered whether Nup foci might serve
function specifically in arrested oocytes. nup214Δ
an essential
mutant females exhibited a ~40% reduction in Nup foci in arrested
oocytes compared with wild-type (Appendix Fig S5L). Remarkably,
we also observed an ~26% increase in Nup levels at the NE, pre-
sumably because reduced condensation in foci liberates Nups to
associate with the NE (Appendix Fig S5L). Using timed matings of
nup214Δ and wild-type females (see Materials and Methods), we
found that embryos produced from 1-day old arrested nup214Δ or
wild-type oocytes were equally viable (Appendix Fig S5M). We
conclude that robust assembly of Nup foci in arrested oocytes is not
essential to support embryonic development.
Multiple mechanisms enhance Nup solubility in the cytoplasm
The observation that Nup foci disassemble during M phase suggests
that cell cycle regulators modulate Nup solubility. PLK1 and CDK1
are two kinases that are active in oocytes and known to drive
nuclear pore disassembly during NE breakdown in M phase (Chase
et al, 2000; De Souza et al, 2004; Onischenko et al, 2005; Laurell
et al, 2011; Rahman et al, 2015; Linder et al, 2017; Martino et al,
2017; Huelgas-Morales & Greenstein, 2018; Kutay et al, 2021). CDK1
enriches in Nup foci in C. elegans oocytes (Appendix Fig S6A) con-
sistent with prior observations in Xenopus oocytes (Beckhelling
et al, 2003). We found that RNAi depletion of PLK1 and CDK1
GFP::Nup88; mCherry::histone (mated for 1.5 hr)
Cytoplasmic
NE/Nucleoplasm
Foci
****
ns
ns
100
75
50
25
8
8
p
u
N
l
a
t
o
t
1.2
0.8
0.4
e
c
n
e
s
e
r
o
u
l
f
8
8
p
u
N
l
a
t
o
T
0.0
-3 -2 -1
Oocyte
Cytoplasmic
NE/Nucleoplasm
Foci
****ns
100
75
50
25
8
5
3
p
u
N
l
a
t
o
t
f
o
%
1.2
0.6
0.0
wt
nup214
-3
-2
-1
f
o
%
zygote
zygote
C
10
5
0
unmated
mated
GFP::Nup88; mCherry::histone
3-cell (mitosis)
mNeonGreen::Nup358 (Day 2 adults)
wild-type
Cytoplasmic
NE/Nucleoplasm
Foci
****
1.2
ns
e
c
n
e
s
e
r
o
u
l
f
8
8
p
u
N
l
t
a
o
T
nup214
0.8
0.4
0.0
3-cell
4-cell
4-cell (interphase)
100
8
8
p
u
N
l
a
t
o
t
f
o
%
75
50
25
5.0
2.5
0.0
3-cell
4-cell
D
wild-type
mNeonGreen::Nup358
nup214
A
B
Cytoplasmic
NE/Nucleoplasm
Foci
****
E
100
75
50
25
8
5
3
p
u
N
l
a
t
o
t
f
o
%
0.8
0.4
0.0
wt
nup214
l
i
c
m
s
a
p
o
t
y
c
/
r
a
e
c
u
N
l
y
t
i
s
n
e
n
t
i
IBB
ns
TDP-43
CRM1
G3BP
ns
ns
ns
8
6
4
2
1.0
0.5
0.0
wt
nup214 wt
nup214 wt
nup214 wt
nup214
Figure 5.
8 of 19
The EMBO Journal 42: e112987 | 2023
(cid:1) 2023 The Authors
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Laura Thomas et al
The EMBO Journal
◀ Figure 5. Nup foci are transient condensates that are not required for nuclear pore biogenesis.
A Left: Representative confocal micrograph showing CRISPR-tagged GFP::Nup88 in oocytes and newly fertilized zygotes of a fog-2(q71) female 1.5 h post-mating. The
fluorescent foci outside of the germline and zygotes are autofluorescent intestinal gut granules. Middle: Quantification of the distribution of GFP::Nup88 between the
cytoplasm (soluble), nuclear envelope (NE)/nucleoplasm, and cytoplasmic foci in -1 oocytes of fog-2(q71) females unmated or mated for 1.5 h. Error bars represent
95% CI for n = 6 oocytes (biological replicates). Right: Total GFP::Nup88 fluorescence in -1, -2, and -3 position oocytes following mating for 1.5 h. Values are normal-
ized within the same germline so that the -1 oocyte measurement = 1.0. Error bars represent 95% CI for n = 6 germlines (biological replicates).
B Left: Representative confocal micrographs showing GFP::Nup88 in 3-cell (mitosis) versus 4-cell (interphase) embryos. Middle: Quantification of the distribution of GFP::
Nup88 between the cytoplasm, NE/nucleoplasm, and cytoplasmic foci. Error bars represent 95% CI for n > 6 embryos (biological replicates). Right: Quantification of
total GFP::Nup88 fluorescence in 3-cell (mitosis) versus 4-cell (interphase) embryos. Values are normalized so that the average fluorescence of 3-cell embryos = 1.0.
Error bars represent 95% CI for n > 6 embryos (biological replicates).
C Left: Representative confocal micrographs showing CRISPR-tagged mNeonGreen::Nup358 in -3 and -4 oocytes of wild-type versus nup214Δ Day 2 adults. Right: Quan-
tification of the distribution of mNeonGreen::Nup358 between the cytoplasm (soluble), NE/nucleoplasm, and cytoplasmic foci in wild-type versus nup214Δ oocytes.
Error bars represent 95% CI for n > 8 germlines (biological replicates).
D Left: Representative confocal micrographs showing mNeonGreen::Nup358 in wild-type versus nup214Δ interphase 4-cell embryos. Right: Quantification of the distri-
bution of mNeonGreen::Nup358 between the cytoplasm (soluble), NE/nucleoplasm, and cytoplasmic foci in wild-type versus nup214Δ embryos. Error bars represent
95% CI for n = 5 embryos (biological replicates).
E Quantification of the nuclear/cytoplasmic ratio of an IBBdomain::mNeonGreen reporter or CRISPR-tagged TDP-43::wrmScarlet, CRM1::mNeonGreen, and G3BP::mCherry
in 28-cell stage embryos. Values are normalized so that the average wild-type measurement = 1.0. Error bars represent 95% CI for n > 9 embryos (IBBdomain::mNeon-
Green), n > 11 embryos (TDP-43::wrmScarlet), n > 11 embryos (CRM1::mNeonGreen), or n = 11 embryos (G3BP::mCherry; biological replicates).
Data information: ****P < 0.0001; ns, not significant. For panel A (right graph) significance was determined using a one-way ANOVA; for all other panels significance was
determined using an unpaired t-test. All images in this figure are maximum intensity projections. Scale bars = 10 lm.
Source data are available online for this figure.
increased the proportion of GFP::Nup88 in foci and at the NE in
growing oocytes (Fig 6A and B, Appendix Fig S6B and C). Inhibition
of the phosphatase PP2A blocks nuclear pore complex and Nup foci
assembly in Drosophila embryos (Onischenko et al, 2005). Consis-
tently, RNAi depletion of the scaffolding subunit of PP2A led to a
striking loss of GFP::Nup88 from foci and the NE (Fig 6A and B,
Appendix Fig S6B and C). These observations suggest that, in addi-
tion to regulating nuclear pore assembly, Nup phosphorylation by
cell cycle kinases increases the solubility of Nups in the cytoplasm
and the phosphatase PP2A counteracts this effect.
FG domains are heavily modified by O-GlcNAcylation, a modifi-
cation proposed to limit FG domain interactions within the nuclear
pore central channel (Ruba & Yang, 2016; Yoo & Mitchison, 2021).
O-GlcNAcylation is catalyzed by the enzyme O-GlcNAc transferase
(OGT), which enriches in Nup foci in oocytes (Appendix Fig S6D).
ogt mutant animals lack Nup O-GlcNAcylation as previously
described (Appendix Fig S6E and F; Hanover et al, 2005) and,
remarkably, exhibit enhanced Nup foci (Fig 6A and C, Appendix
Fig S6B and G). We also visualized Nup foci in a loss of function
allele of the C. elegans O-GlcNAcase (OGA) reported to exhibit
higher
levels of Nup O-GlcNAcylation in embryos (Forsythe
et al, 2006). We did not detect a significant change in Nup foci in
the oga mutant, suggesting that, in oocytes, Nups may be suffi-
ciently O-GlcNAcylated such that loss of OGA activity does not
affect Nup solubility.
To test whether O-GlcNAcylation contributes to Nup solubility
outside of the germline, we examined ogt mutant animals for Nup
foci in somatic tissues. We found that, by Day 4 of adulthood, ogt
mutant animals had a higher incidence of Nup foci in somatic cells
compared to wild-type (Fig 6D),
suggesting a role for O-
GlcNAcylation in promoting FG-Nup solubility in both soma and
germline tissues.
Recent studies have suggested that NTRs function as “chaper-
ones” to prevent aggregation of intrinsically disordered proteins,
including Nups (Milles et al, 2013; Guo et al, 2018; Hofweber
et al, 2018; Hutten et al, 2020; Khalil et al, 2022). We found that
two endogenously tagged NTRs (CRM1 and transportin) are
enriched in cytoplasmic Nup foci in C. elegans oocytes (Appendix
Fig S7A and B). The exportin CRM1 makes high affinity interactions
with the FG-Nups Nup214 and Nup358 (Port et al, 2015; Ritterhoff
et al, 2016; Tan et al, 2018). As we found Nup214 and Nup358 to be
key scaffolds for Nup foci (see Fig 3B), we next tested whether
CRM1-binding promotes the solubility of cytoplasmic Nups. Consis-
tent with a solubilizing effect for CRM1 interaction, RNAi depletion
of CRM1 led to an increase in Nup foci formation (Fig 6A and E,
Appendix Figs S6B and S7C and D). This effect is unlikely to be due
to impaired nuclear export, as Nup foci were not altered in worms
treated for 4 h with the CRM1 inhibitor leptomycin B (LMB; Fig 6A
and E, Appendix Figs S6B and S7D and E). Despite enrichment of
transportin at Nup foci, RNAi depletion of transportin did not affect
Nup solubility (Appendix Fig S7F and G). In summary, these results
suggest that Nup solubility is enhanced by multiple mechanisms
including phosphorylation, O-GlcNAcylation, and CRM1 binding.
Ectopic Nup condensation in neurons is toxic
We only detected Nup foci in somatic cells of aged hermaphrodites
and ogt mutants (see Figs 4C and D, and 6D). To determine whether
Nup condensation might be detrimental in somatic cells, we used
the neuron-specific rab-3 promoter to overexpress Nup98::mNeon-
Green, a highly cohesive FG-Nup that interacts with multiple struc-
tured Nups (Schmidt & Go¨rlich, 2016; Onischenko et al, 2017).
Unlike Nup98 expressed from its endogenous locus (Appendix
Fig S8A), overexpressed Nup98 readily formed abundant cytoplas-
mic foci and localized to the NE at low levels (Appendix Fig S8B).
Remarkably, the ectopic Nup98 foci recruited endogenous Nup62,
resulting in partial depletion of Nup62 from the NE (Fig 7A). In con-
trol, non-neuronal cells that did not express the Nup98 transgene,
Nup62 localized to the NE as in wild-type animals (Fig 7B). Consis-
tent with disrupted nuclear transport in the Nup98 overexpressing
neurons, the nuclear protein TDP-43 was mislocalized to the cyto-
plasm (Appendix Fig S8C). Strikingly, rab-3p::Nup98 animals had
shorter lifespans (Appendix Fig S8D) and appeared uncoordinated
(barely moving) on plates or in liquid media (Fig 7C, Movies EV3
(cid:1) 2023 The Authors
The EMBO Journal 42: e112987 | 2023
9 of 19
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License The EMBO Journal
Laura Thomas et al
A
i
c
o
f
n
i
p
u
N
f
o
e
g
n
a
h
c
d
o
F
l
40
30
20
10
2
1
0
****
****
****
ns
****
B
Day 1 adults
Day 1 adults
Day 2 adults
****
ns
control
control
control
control (RNAi)
control (RNAi)
cdk1 (RNAi)
plk1 (RNAi)
control (RNAi)
pp2a (RNAi) wt
ogt
oga
control (RNAi)
crm1 (RNAi)
8
8
p
u
N
:
:
P
F
G
control
LMB
plk1 (RNAi)
cdk1 (RNAi)
pp2a (RNAi)
C
GFP::Nup88 (Day 1 adults)
wild-type
D
GFP::Nup88; mCherry::histone (Day 4 adults)
ogt
oga
wild-type
ogt
E
control
None
Mild
Moderate
Day 1 adults
control
1.00
0.75
0.50
5
8
p
u
N
:
:
P
F
G
crm1 (RNAi)
LMB
i
c
o
f
h
t
i
w
s
m
r
o
w
f
o
%
0.25
0.00
wt
ogt
Figure 6. Phosphorylation, GlcNAcylation, and CRM1 promote Nup solubility.
A Quantification of the relative Nup intensity in foci in each indicated condition compared to control. Values are normalized so that the average control condition
measurement = 1.0. Error bars represent 95% CI (biological replicates) and data correspond to micrographs in Fig 6B (plk1 RNAi, n > 8 germlines; cdk1 RNAi, n > 6
germlines; pp2A RNAi, n > 6 germlines), Fig 6C (ogt and oga mutants, n > 6 germlines), and Fig 6E (crm1 RNAi, n > 7 germlines; LMB treatment, n > 8 germlines). See
Appendix Fig S6B for raw (non-normalized) values for the distribution of each Nup between the cytoplasm (soluble), nuclear envelope (NE)/nucleoplasm, and cyto-
plasmic foci for each condition.
B Representative confocal micrographs showing CRISPR-tagged GFP::Nup88 in -3 and -4 oocytes depleted of PLK1, CDK1, or the PP2A scaffolding subunit PAA-1. Day 1
adults were used for kinase depletion, and Day 2 adults were used for phosphatase depletion.
C Representative confocal micrographs showing GFP::Nup88 in -3 and -4 oocytes of wild-type, ogt, or oga mutant Day 1 adults.
D Left: Representative confocal micrographs showing GFP::Nup88 in the head of wild-type versus ogt mutant Day 4 adults. Right: Quantification of the number of wild-
type versus ogt mutant Day 4 adults lacking foci (none), or with mild or moderate cytoplasmic foci in somatic tissues. n > 30 animals for each genotype.
E Left: Representative confocal micrographs showing CRISPR-tagged GFP::Nup85 in -3 and -4 control oocytes or oocytes depleted of CRM1. Right: Representative confo-
cal micrographs showing GFP::Nup85 in -3 and -4 oocytes of control animals or following treatment with the CRM1 inhibitor leptomycin B (LMB). All images are from
Day 1 adults.
Data information: ****P < 0.0001; ns, not significant. For the ogt and oga mutants significance was determined using a one-way ANOVA; for all other conditions signifi-
cance was determined using an unpaired t-test. All the images in this figure are maximum intensity projections. Scale bars = 10 lm.
Source data are available online for this figure.
and EV4), consistent with neuronal dysfunction and paralysis
(Dimitriadi & Hart, 2010). We obtained similar results with an inde-
pendently generated transgenic animal with N-terminally tagged
mNeonGreen::Nup98 expressed using the rab-3 promoter (Appendix
Fig S8E–G, Movies EV5 and EV6). We conclude that uncontrolled
Nup condensation in post-mitotic neurons is toxic and leads to cellu-
lar dysfunction.
Discussion
Cytoplasmic Nup foci have been observed in oocytes, yeast, and in
many cell types in culture (Cordes et al, 1996; Colombi et al, 2013;
Raghunayakula et al, 2015; Ren et al, 2019). In this study, we report
the systematic examination of the incidence of Nup foci across all
tissues in an intact animal. We find that Nup foci are rare in healthy
animals and arise only in cells where cytoplasmic Nup concentra-
tion is highest: gametes and early embryos. Although Nup conden-
sates appear prominent when observed by fluorescence microscopy,
in growing oocytes and embryos they account for less than 3% of
total cellular Nups and consist primarily of highly cohesive FG-
Nups. The vast majority of FG-Nups are stored as soluble molecules
in the cytoplasm whose condensation is actively suppressed by mul-
tiple mechanisms. Stress and aging promote FG-Nup condensation
which can be toxic in post-mitotic cells if uncontrolled. Our findings
do not support an essential role for Nup foci in pore assembly in C.
elegans and instead we propose that Nup foci are non-functional
byproducts of the natural tendency of FG-Nups to condense.
10 of 19
The EMBO Journal 42: e112987 | 2023
(cid:1) 2023 The Authors
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Laura Thomas et al
The EMBO Journal
A
t
l
u
d
a
1
y
a
D
B
t
l
u
d
a
1
y
a
D
rab-3p::Nup98::mNeon
endogenous Nup62::wrmScarlet
Merge
rab-3p::Nup98
Nup62
Hoechst
Merge
endogenous mNeon::Nup98
endogenous Nup62::wrmScarlet
Merge
Nup98
Nup62
Hoechst
Merge
****
6
4
2
E
N
t
a
2
6
p
u
N
0
rab-3p::Nup98
control
Day 1 adults
C
l
o
r
t
n
o
c
8
9
p
u
N
:
:
p
3
-
b
a
r
100
****
e
t
u
n
m
i
/
s
d
n
e
b
80
60
40
20
0
rab-3p::Nup98
control
Figure 7. Ectopic Nup98 foci in neurons deplete an endogenous Nup from the nuclear envelope and cause paralysis.
A Top left: Representative confocal micrographs showing localization of endogenous CRISPR-tagged Nup62::wrmScarlet relative to transgenic rab-3p::Nup98::mNeon-
Green in the tail of a Day 1 adult. Gray dashed lines indicate the boundary of the tail. Transgenic rab-3p::Nup98::mNeonGreen is only expressed in neurons, which are
designated by the red dashed outline. White arrows indicate enrichment of endogenous Nup62 in ectopic Nup98 foci. Bottom left: Representative confocal micro-
graphs showing endogenous Nup62 depletion from the nuclear envelope. White dashed lines indicate the boundary of the nucleus. Right: Line-scan quantification of
the nuclear envelope (NE) to nucleoplasm ratio of endogenous Nup62 in control (non-neuronal) cells, versus neurons with ectopically expressed rab-3p::Nup98::
mNeonGreen. Error bars represent 95% CI for n > 12 nuclei (biological replicates).
B Top: Representative confocal micrographs showing localization of Nup62::wrmScarlet versus CRISPR-tagged endogenous mNeonGreen::Nup98 in the tail of a Day 1
adult. Gray dashed lines indicate the boundary of the tail. Bottom: Representative confocal micrographs showing localization of Nup62 versus Nup98 at a single
nucleus.
C Left: Representative images of Day 1 adults expressing no transgene (control) or the Nup98::mNeonGreen transgene driven by the pan-neuronal rab-3 promoter. The
control animal shows the wild-type sinusoidal posture, whereas the transgenic animal exhibits an uncoordinated posture. Right: Graph showing the number of
bends/minute during a swim test for Day 1 adults expressing no transgene (control) or the rab-3p::Nup98::mNeonGreen transgene. Error bars represent 95% CI for
n > 11 worms (biological replicates).
Data information: ****P < 0.0001. Significance was determined using an unpaired t-test. All the images in this figure are maximum intensity projections, with the excep-
tion of panels A and B (bottom) which are single focal planes. Scale bars = 10 lm (panels A and B, top) or 2 lm (panels A and B, bottom).
Source data are available online for this figure.
Cytoplasmic Nup foci arise by condensation of FG-Nups and their
binding partners
Several lines of evidence suggest that condensation of FG-Nups
underlies Nup foci assembly. First, Nup foci in growing oocytes pri-
marily consist of FG-Nups and their binding partners and lack nucle-
oplasmic Nups as well as Nups essential for pore assembly (inner
ring complex and transmembrane Nups). Second, Nup foci display
heterogeneous Nup stoichiometry and rarely colocalize with mem-
branes. Finally, consistent with condensation, a concentration-
dependent assembly mechanism, depletion and overexpression of
foci
individual FG-Nups eliminate and enhance,
respectively,
formation. Together these observations suggest that Nup foci are
not structured pre-pore assemblies, but are condensates scaffolded
by cohesive FG-Nups,
including Nup62, Nup98, Nup214, and
Nup358, and their binding partner Nup88.
Consistent with our findings, a recent systematic survey in
HEK293T cells revealed that cytoplasm-facing FG-Nups and their
binding partners accumulate in cytoplasmic foci, but Nup153, which
faces the nucleoplasm, does not (Cho et al, 2022). Other studies in
HeLa and Cos7 cells have also documented that most Nup foci do
not colocalize with membranes (Ren et al, 2019; Agote-Aran
et al, 2020). Similarly, Nup foci
in yeast cells contain multiple
lack transmembrane or inner ring complex Nups
FG-Nups but
(cid:1) 2023 The Authors
The EMBO Journal 42: e112987 | 2023
11 of 19
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
The EMBO Journal
Laura Thomas et al
(Colombi et al, 2013). In agreement with this study, we found that
the FG-Nup Nup214 forms hexanediol-sensitive foci in yeast cells,
but
the nucleoplasm-facing Nups Nup50 and Nup153 do not
(Appendix Fig S8H). Together, these studies suggest that, as we pro-
pose here for C. elegans, many reported Nup foci likely correspond
to FG-Nup condensates rather than pre-assembled pore complexes.
We suggest that Nup condensates arise whenever the concentra-
tion of FG-Nups exceeds the solubility threshold in the cytoplasm.
Consistent with this hypothesis, depletion of scaffold nucleoporins
that liberate FG-Nups enhance foci formation in C. elegans oocytes
(Fig 3, Appendix Fig S3), yeast (Makio et al, 2009), and HeLa cells
(Raghunayakula et al, 2015). Similarly, intranuclear Nup assemblies
termed GLFG bodies were reported in HeLa cell lines with elevated
levels of Nup98, a highly cohesive FG-Nup (Griffis et al, 2002;
Morchoisne-Bolhy et al, 2015). Our findings indicate that, even
under conditions where FG-Nups exceed their solubility limit by a
foci easily visible by standard
small margin,
microscopy techniques.
they from bright
Phosphorylation, GlcNAcylation, and CRM1-mediated
chaperoning activity suppress Nup condensation
that
What keeps most FG-Nups soluble in the cytoplasm? Our findings
suggest
the solubility limit of FG-Nups in the cytoplasm
depends on several factors and oscillates with the cell cycle, peaking
during M phase. The same kinases that drive nuclear pore complex
disassembly during M phase (PLK1 and CDK1) appear to also pro-
mote Nup solubility in the cytoplasm during interphase. Although
we did not directly monitor Nup phosphorylation, PLK1 and CDK1
have been well-characterized as kinases that directly phosphorylate
Nups to drive nuclear pore disassembly (Chase et al, 2000; De Souza
et al, 2004; Onischenko et al, 2005; Laurell et al, 2011; Rahman
et al, 2015; Linder et al, 2017; Martino et al, 2017). Consistent with
phosphorylation promoting Nup solubility, cellular fractionation
experiments have shown that soluble Nups are more highly phos-
phorylated than Nups in pore complexes (Onischenko et al, 2004).
Other kinases implicated in Nup phosphorylation and pore complex
disassembly,
including NIMA and DYRK kinases (De Souza
et al, 2004; Laurell et al, 2011; Wippich et al, 2013), could also con-
tribute to Nup solubility.
Consistent with prior findings showing that O-GlcNAcylated FG-
Nups are less prone to condensation in vitro (Labokha et al, 2012;
Schmidt & Go¨rlich, 2015), our observations also suggest that O-
GlcNAcylation contributes to Nup solubility in oocytes, embryos,
and somatic cells. Numerous studies have reported a protective role
for O-GlcNAcylation in neurodegenerative disease (reviewed in Lee
et al, 2021), raising the possibility that this modification plays a key
role in solubilizing certain aggregation-prone proteins. A separate
study found that O-GlcNAcylation promotes condensation of stress
granules and P-bodies (Ohn et al, 2008), indicating that the solubi-
lizing effect of O-GlcNAcylation is likely protein- and context-
dependent.
Finally, we find that the nuclear export factor CRM1 also contrib-
utes to Nup solubility. Structural analyses of CRM1/Nup214 com-
plexes reveal that hydrophobic patches on the surface of CRM1
make
(Port
et al, 2015). CRM1 generates high-affinity interactions with both
Nup214 and Nup358 that are significantly stronger than the weak,
contacts with Nup214 FG domains
extensive
transient interactions characteristic of most Nup/NTR pairs (Port
et al, 2015; Ritterhoff et al, 2016; Tan et al, 2018). Both Nup214 and
Nup358 are required for Nup foci formation and therefore neutrali-
zation of these proteins by CRM1 is predicted to reduce Nup foci for-
mation. Interestingly, loss of another transport factor, transportin,
did not affect Nup solubility in C. elegans oocytes, suggesting that
not all NTRs play a significant role in promoting Nup solubility.
Nup foci are not required for nuclear pore function or viability in
C. elegans
During Drosophila oogenesis, Nup condensates mature into annu-
late lamellae (Hampoelz et al, 2019b), endoplasmic reticulum-
derived membranous structures that contain pore-like complexes
(Cordes et al, 1996; Miller & Forbes, 2000). Annulate lamellae have
been observed in arrested oocytes of unmated C. elegans females
(Patterson et al, 2011), where we find that Nup foci associate with
the endoplasmic reticulum at a higher frequency, accumulate a
greater proportion of FG-Nups, and recruit additional Nups not pre-
sent in foci of growing oocytes. It is possible, therefore, that, as
reported for Drosophila, Nup foci have the potential to evolve into
annulate lamellae in C. elegans, but this possibility will require fur-
ther investigation.
In Drosophila, annulate lamellae assembled in oocytes have
been proposed to fuel the rapid expansion of nuclear membranes
in embryos by directly inserting into nuclear membranes during
interphase (Hampoelz et al, 2016). Annulate lamellae have not
been observed in C. elegans embryos (Pitt et al, 2000) and we find
that all Nup assemblies dissolve during the oocyte-to-embryo tran-
sition. In embryos, Nup foci re-appear and dissolve again with
every M phase, in synchrony with the disassembly of nuclear pores
at the NE. Annulate lamellae assembled in oocytes, therefore, are
unlikely to be a source of pre-assembled pores for embryos in C.
elegans. Furthermore, we have identified two mutants, nup214Δ
and nup88Δ, that severely reduce the incidence of Nup foci in
oocytes and embryos yet assemble functional nuclear pores during
embryogenesis and are viable. We consider it unlikely, therefore,
that Nup foci contribute significantly to nuclear pore assembly in
C. elegans.
If so, why do Nup foci assemble in C. elegans oocytes? We con-
sidered the possibility that Nup foci sequester damaged Nups that
must be removed from the soluble pool before embryogenesis,
which may be particularly important during oocyte arrest (Bohnert
& Kenyon, 2017). This possibility, however, appears unlikely as the
~10% of Nups in foci in arrested oocytes return to the soluble pool
without any loss upon oocyte maturation. Additionally, nup214Δ
arrested oocytes, which have reduced Nup foci, have higher Nup
levels at the nuclear rim (compared with wild-type) and produce
fully viable embryos, suggesting that, when not induced to form
foci, excess Nups can assemble into nuclear pores at the nuclear
periphery. Another possibility is that Nup foci serve a role unrelated
to nuclear pore formation that becomes essential under conditions
not tested in this study. Prior studies have noted that Nup foci
assemble near RNP granules (Pitt et al, 2000; Jud et al, 2007; Sheth
et al, 2010; Patterson et al, 2011; Sahoo et al, 2017), suggesting that
Nup foci may contribute to RNA homeostasis, but this possibility
remains to be tested. A final possibility is that Nup foci serve no
function and arise simply as the inevitable consequence of the high
12 of 19
The EMBO Journal 42: e112987 | 2023
(cid:1) 2023 The Authors
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Laura Thomas et al
The EMBO Journal
concentration of FG-Nups needed for embryogenesis, which tran-
siently saturates the cytoplasm of oocytes and early embryos.
Uncontrolled Nup condensation can be toxic in post-mitotic cells
Overexpression of Nup98 in neurons was sufficient to assemble
ectopic Nup foci and cause paralysis, suggesting that uncontrolled
Nup condensation in somatic cells is potentially toxic. We speculate
impaired
that neuronal dysfunction arose as a consequence of
nucleocytoplasmic transport due to recruitment of endogenous Nups
to the ectopic foci. Our findings are consistent with recent studies
reporting that cytoplasmic FG-Nups drive aggregation of TDP-43 in
both ALS/FTLD and following traumatic brain injury (Anderson
et al, 2021; Gleixner et al, 2022). Several other studies have linked
Nup condensation to stress and disease, including: (i) Nup accumu-
lation in stress granules (Zhang et al, 2018; Agote-Aran et al, 2020),
(ii) aberrant condensation of Nup98 and Nup214 fusion proteins
driving oncogenic transformation in certain leukemias (Zhou &
Yang, 2014; Terlecki-Zaniewicz et al, 2021; Chandra et al, 2022),
(iii) the formation of NE associated Nup condensates in models of
DYT1 dystonia (Prophet et al, 2022), and (iv) the presence of Nups
in pathological inclusions in primary patient samples and models of
neurodegenerative disease (reviewed in Fallini et al, 2020; Hutten &
Dormann, 2020; Chandra & Lusk, 2022).
The deleterious effects of Nup condensation are likely context
dependent. In arrested oocytes, Nup condensation increases ~14-
fold over growing oocytes, yet is not damaging as the majority of
arrested oocytes go on to form viable embryos when fertilized (Jud
et al, 2008; Patterson et al, 2011). Pore complexes and Nup conden-
sates in oocytes and embryos fully disassemble during M phase,
allowing for a cycle of “renewal” with each cell division. Nup con-
densation may only be dangerous in post-mitotic cells that lack M
phase-specific Nup solubilizers and where certain Nups are natu-
rally long-lived (D’Angelo et al, 2009; Toyama et al, 2013). We sug-
gest that post-mitotic cells avoid Nup condensation by maintaining
low levels of cytoplasmic Nups and high levels of solubilizing modi-
fications. Indeed, we observed that Nup foci in the somatic tissues
of aged adults become more prominent in ogt mutants lacking O-
GlcNAcylation. We do not know the origin of Nup foci in aged cells,
but they may be linked to the progressive decline in proteostasis
and nuclear ‘leakiness’ that initiates during C. elegans adulthood
(Herndon et al, 2002; Ben-Zvi et al, 2009; D’Angelo et al, 2009). C.
elegans oocytes and embryos, which naturally accumulate and clear
Nup condensates, offer a powerful model system to explore possible
mechanisms to prevent or reverse Nup condensation during aging.
Materials and Methods
C. elegans and yeast strains and culture
C. elegans were cultured using standard methods (Brenner, 1974).
Briefly, worms were maintained at 20°C on normal nematode
growth media (NNGM) plates (IPM Scientific Inc. cat # 11006-548)
seeded with OP50 bacteria. We have found that Nup solubility is
highly influenced by multiple factors including animal age: for all
in growing oocytes
Nups tested the number and size of
increased significantly between Days 1 and 2 of adulthood
foci
(see Appendix Fig S1D). Therefore, for all experiments worms were
synchronized as Day 1 or 2 adults using vulval morphology to stage
L4 larvae. The age of animals used for each experiment is indicated
in figures and legends.
Endogenous npp-21 (TPR) was tagged with GFP using CRISPR/
Cas9-mediated genome editing as previously described (Arribere
et al, 2014). Endogenous npp-24 (Nup88) and npp-2 (Nup85) were
tagged with G>F>P using SapTrap CRISPR/Cas9 gene modification as
previously described (Schwartz & Jorgensen, 2016). G>F>P contains
Frt sites in introns 1 and 2 of GFP that enable FLP-mediated, condi-
tional knockout; in the absence of FLP, the construct behaves as nor-
mal GFP. To generate a permanent npp-24 knockout, recombination
was induced in the germline (Mac(cid:2)ıas-Le(cid:2)on & Askjaer, 2018) followed
by selection of progeny in which the second GFP exon was excised
from both alleles. This strategy phenocopies complete gene removal
(Fragoso-Luna et al, 2023). Endogenous npp-19 (Nup35) was tagged
with G>F>P based on protocols for nested CRISPR (Vicencio
et al, 2019) and ‘hybrid’ partially single-stranded DNA donors
(Dokshin et al, 2018). All other endogenous edits were performed
using CRISPR/Cas9-mediated genome editing as described previously
(Paix et al, 2017). Transgenic Nup214 and Nup98 strains (JH4119,
JH4204, JH4205, and JH4395) were generated using SapTrap cloned
vectors as previously described (Fan et al, 2020). Standard crosses
were used to generate strains with multiple genomic edits. All strains
used or generated in this study are described in Appendix Table S1.
Yeast strains were generated using homologous recombination of
PCR-amplified cassettes (Longtine et al, 1998). Endogenous NUP159
(Nup214), NUP60 (Nup153), and NUP2 (Nup50) were tagged by
amplifying the mNeonGreen::HIS3 cassette from pFA6a-mNeon-
Green::HIS3 (Thomas et al, 2019) using primers with homology to
the C-termini (without the stop codon) and downstream regions of
the genes. Yeast strains generated in this study are described in
Appendix Table S1.
RNAi
RNAi was performed by feeding (Timmons & Fire, 1998). RNAi vec-
tors were obtained from the Ahringer or Open Biosystems libraries
and sequence verified, or alternatively cloned from C. elegans cDNA
and inserted into the T777T enhanced RNAi vector (Addgene cat #
113082). RNAi feeding vectors were freshly transformed into HT115
bacteria, grown to log phase in LB + 100 lg/ml ampicillin at 37°C,
induced with 5 mM IPTG for 45 min, and plated on RNAi plates (50
lg/ml Carb, 1 mM IPTG; IPM Scientific Inc. cat # 11006-529). Seeded
plates were allowed to dry overnight at RT before adding L4 larvae or
Day 1 adults. For depletion of Nup98 (Fig 3B, Appendix Fig S3C),
RNAi feeding was performed for 6 h at 25°C; partial depletion was
used to minimize cytological defects caused by loss of Nup98. For all
other experiments, RNAi feeding was performed for 18–24 h at 25°C.
For all experiments, control worms were fed HT115 bacteria trans-
formed with the corresponding L4440 or T777T empty vector.
Immunofluorescence
For immunostaining of embryos, gravid adults were placed into 7 ll
of M9 media on a poly-L-lysine coated slide and compressed with a
coverslip to extrude embryos. For immunostaining of oocytes, staged
adults were dissected on poly-L-lysine slides to extrude the germline,
(cid:1) 2023 The Authors
The EMBO Journal 42: e112987 | 2023
13 of 19
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License The EMBO Journal
Laura Thomas et al
and a coverslip was placed gently on top. In both cases, slides were
immediately frozen on aluminum blocks pre-chilled with dry ice.
After > 5 min, coverslips were removed to permeabilize embryos
(freeze-cracking), and slides were fixed > 24 h in pre-chilled MeOH
at (cid:1)20°C. Slides were then incubated in pre-chilled acetone for
10 min at (cid:1)20°C, and blocked in PBS–T (PBS, 0.1% Triton X-100,
0.1% BSA) for > 30 min at RT. Slides were then incubated overnight
in primary antibody in a humid chamber at 4°C. Slides were washed
3 × 10 min in PBS-T at RT, incubated in secondary antibody for 2 h
in a humid chamber at RT, and washed 3 × 10 min in PBS–T at RT.
Slides were then washed 1x in PBS before being mounted using Pro-
long Glass Antifade Mountant with NucBlue (Thermo Fisher cat #
P36981). Primary antibodies were diluted as follows: mAb414
(1:1,000; Biolegend cat # 902907), anti-Nup358 (1:250; Novus Biolog-
icals cat # 48610002), anti-Nup50 (1:250, Novus Biologicals cat #
48590002), anti-GlcNAc RL2 (1:100; Invitrogen cat # MA1-072), anti-
Nup96 (1:250, R(cid:2)odenas et al, 2012), anti-Nup153 (1:250, Galy
et al, 2003), anti-OLLAS-L2 (1:50, Novus Biologicals cat # NBP1-
06713). Secondary antibodies were diluted as follows: Cy3 Donkey
anti-Mouse IgG (1:200; Jackson cat # 715-165-151), AlexaFluor 488
Goat anti-Rabbit IgG (1:200; Invitrogen cat # A-11034), AlexaFluor
568 Goat anti-Rabbit IgG (1:200; Invitrogen cat # A-11011), Alexa
Fluor 488 Goat anti-Rat IgG (1:200; Invitrogen cat # A-11006), Alexa-
Fluor 488 anti-GFP (1:500; Invitrogen cat # A-21311).
LMB and HXD treatment, HaloTag and Hoechst labeling, and
heat stress
For CRM1 inhibition, leptomycin B (LMB; Sigma cat # L2913) was
diluted in OP50 bacteria to a final concentration of 500 ng/ml and
seeded on NNGM plates. 10-20 Day 1 adults were transferred to
LMB or control vehicle plates and incubated at 20°C for 4 h prior to
imaging. For treatment of embryos with 1,6-hexanediol (HXD, Acros
Organics cat # 629-11-8), L4 larvae were fed ptr-2 RNAi for 18–24 h
at 20°C. Embryos depleted of PTR-2, which permeabilizes the egg-
shell to allow for HXD treatment, were dissected into L-15 media
(Thermo Fisher cat # 21083027) containing 2% HXD and immedi-
ately imaged. For HXD treatment of yeast, log-phase yeast were
pelleted, re-suspended in media containing 5% HXD, and allowed to
grow for 10 min at 30°C prior to imaging.
For HaloTag labeling,
Janelia Fluor 646 HaloTag Ligand
(Promega cat # GA1121) was diluted in OP50 bacteria to a final con-
centration of 25 lg/ml and seeded on NNGM plates. 10–20 L4 lar-
vae or Day 1 adults were added and incubated without light at 20°C
for 16–20 h prior to imaging. For Hoechst staining, Hoechst 33342
dye (Thermo Fisher cat # 62249) was diluted in OP50 bacteria to a
final concentration of 200 lM and seeded on NNGM plates. 10–20
L4 larvae were added and incubated without light at 20°C for 16–20
h prior to imaging.
To induce heat stress, animals were grown at 20°C then trans-
ferred to pre-warmed NNGM plates at 30°C for 20 min prior to imag-
ing at room temperature or processing for immunofluorescence.
Embryonic viability and lifespan analysis
To measure embryonic viability of the nup214D mutant (Appendix
Fig S5G), six Day 1 adults were transferred to six NNGM plates (36
animals total) and allowed to lay embryos for 1 h at 20°C. To
the nup88D mutant (Appendix
measure embryonic viability of
Fig S5J), two Day 1 adults were transferred to six plates (12 animals
total) and allowed to lay embryos for 5 h at 20°C. Adults were then
removed and the number of embryos on each plate was counted. For
the nup214D mutant, embryos were allowed to hatch and the number
of adults on each plate was counted after 3 days at 20°C. For the
nup88D mutant, the number of unhatched embryos was counted after
1 day at 20°C. Viability counts were repeated in at least two indepen-
dent experiments, and embryonic viability was measured as the num-
ber of surviving adults or hatched larvae divided by the original
number of embryos counted in each experiment.
To measure embryonic viability of arrested oocytes following
mating (Appendix Fig S5M), fog-2(q71) female L4 larvae were incu-
bated overnight at 20°C in the absence of males. Individual Day 1
adult females with arrested oocytes were then mated with 5–7 males
on a small patch of bacteria. Matings were monitored closely, and
the adults removed once 3–10 embryos were laid. The number of
embryos on each plate was counted, embryos were allowed to
hatch, and the number of adults on each plate was counted after
3 days at 20°C. Viability counts were repeated in three independent
experiments, and embryonic viability was measured as the number
of surviving adults divided by the original number of embryos laid.
Prior studies have reported that > 90% of arrested oocytes produce
viable embryos (Jud et al, 2008; Patterson et al, 2011), whereas we
found that ~70% of arrested oocytes gave rise to viable embryos
(Appendix Fig S5M). These prior studies measured the viability of
all oocytes accumulated prior to mating (~20 per gonad arm),
whereas we measured the viability of the first few oocytes ovulated
following mating (~2–5 per gonad arm). This difference may explain
the comparatively lower viability observed in our experiments.
To measure adult lifespan, 75 Day 1 adults were transferred to
five NNGM plates (15 animals per plate) and incubated at 20°C.
Worms were scored daily and considered to be dead if they failed to
move when prodded. Worms were transferred every 2 days to avoid
progeny, and any worms that crawled off the plates were censored
from analysis.
Swimming assay
To measure swimming behavior, 5–10 Day 1 adults were transferred
to a 33 mm culture dish (MatTek cat # P35G-1.5-14-C) containing
400 ll M9 media and immediately filmed using an Axiocam 208
color camera (Zeiss) mounted on a Stemi 508 Stereo Microscope
(Zeiss). Swimming assays were performed at RT (~22°C). Movies
were exported to ImageJ, and the number of body bends per minute
was counted manually.
Imaging
For live imaging of germlines and somatic tissues, 5 staged adults
were transferred to the middle well of a 3-chambered slide (Thermo
Fisher cat # 30-2066A) in 10 ll of L-15 media with 1 mM levamisole.
20 lm polystyrene beads (Bangs Laboratories Inc. cat # PS07003)
were then added to support a coverslip (Marienfeld cat # 0107052).
Germlines were imaged using an inverted Zeiss Axio Observer with
CSU-W1 SoRa spinning disk scan head (Yokogawa), 1×/2.8×/4× relay
lens (Yokogawa), and an iXon Life 888 EMCCD camera (Andor) con-
trolled by Slidebook 6.0 software (Intelligent Imaging Innovations).
14 of 19
The EMBO Journal 42: e112987 | 2023
(cid:1) 2023 The Authors
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Laura Thomas et al
The EMBO Journal
To image germlines or somatic cells, a 20 lm Z stack (1 lm step size)
was captured using a 63× objective (Zeiss) with the 1× relay lens. For
high resolution images of oocytes, 3 lm Z stacks (0.1 lm step size)
were acquired using the 63× objective with the 2.8× relay lens. As
germline condensates are highly sensitive to imaging-induced stress
(Elaswad et al, 2022), care was taken to avoid compression of germ-
lines, and all animals were imaged only once and maintained on the
slide for <5 min. To image entire worms (Appendix Fig S8A, B, and
E), an 80 lm Z stack (1 lm step size) was captured using a 10× objec-
tive (Zeiss) with the 1x relay lens.
For imaging arrested oocytes and newly fertilized zygotes follow-
ing mating (Fig 5A), fog-2(q71) female L4 larvae were incubated
overnight at 20°C in the absence of males. Day 1 adult females with
arrested oocytes were then mated or not with an abundance of
males at 20°C for 1.5 h prior to imaging.
For live imaging of embryos, five young adults were transferred
to 10 ll of L-15 media on a coverslip and dissected to release
embryos. 20 lm polystyrene beads were then added to prevent com-
pression, and the coverslip was inverted onto a microscope slide
(Thermo Fisher cat # 12-550-403). Embryos were imaged as 15 lm
Z stacks (1 lm step size), captured using the 63× objective with the
2.8× relay lens. For imaging fixed germlines and embryos, prepared
slides were imaged as 15 lm Z stacks (0.5 lm step size), captured
using the 63× objective with the 2.8× relay lens.
For live imaging of yeast, cells were grown overnight in synthetic
dropout media (Thermo Fisher cat # DF0919-15-3) at 30°C and
imaged in log-phase (OD600 of ~0.5) at room temperature. Yeast
were imaged as 6 lm Z stacks (0.5 lm step size), captured using the
63× objective with the 2.8× relay lens.
Images were exported from Slidebook software and further ana-
lyzed using ImageJ or Imaris image analysis software. For presenta-
tion in figures, images were processed using ImageJ, adjusting only
the minimum/maximum brightness levels for clarity with identical
leveling between all images in a figure panel. Images presented in
figures are maximum intensity projections (10 lm for germlines, 15
lm for embryos, 6 lm for yeast) or single focal planes as indicated
in the legends.
Image quantification
The overlap of GFP or mNeonGreen-tagged Nups with Nup62::
wrmScarlet (Fig 2C) was measured using single focal planes
exported to ImageJ. The Nup62::wrmScarlet micrograph was used
to create a mask defining the NE as well as cytoplasmic foci as indi-
vidual regions of interest (ROIs). This mask was then applied to
both the GFP/mNeonGreen Nup micrograph as well as the Nup62::
wrmScarlet micrograph and the integrated density was measured
within each ROI. To control for cytoplasmic background, the aver-
age cytoplasmic signal for the GFP/mNeonGreen Nup was multi-
plied by the area of each ROI, and the resulting value subtracted
from integrated density for the GFP/mNeonGreen Nup. Background
normalized GFP/mNeonGreen Nup values were divided by Nup62::
wrmScarlet values to obtain the ratio of GFP/mNeonGreen Nup to
Nup62::wrmScalet at each ROI.
To quantify the overlap of GFP::Nup88 with membranes (Fig 2D,
Appendix Fig S4C), Z stacks of oocytes expressing GFP::Nup88 and
the HaloTag::HDEL reporter were manually scored into three catego-
ries: 1. Complete overlap (the entire Nup88 focus overlapped with
HaloTag::HDEL); 2. Partial overlap (the Nup88 focus partially over-
lapped or was directly adjacent to HaloTag::HDEL); 3. No overlap
(the Nup88 focus did not directly contract membranes marked by
HaloTag::HDEL).
To quantify the distribution of Nups in oocytes as well as total
expression, Z stacks were exported to Imaris image analysis soft-
ware. The ‘Surface’ tool was first used to isolate the -3 and -4
oocytes from each germline (Appendix Fig S3A). For each pair of -3
and -4 oocytes, the Surface tool was then used to isolate both nuclei
and the “Spot” tool was used to isolate cytoplasmic foci. The per-
cent of Nup present at the NE/nucleoplasm was measured as the
intensity sum for both nuclei divided by the total intensity sum of
the oocytes. Similarly, the percent of Nup present in foci was mea-
sured as the intensity sum for all foci divided by the total intensity
sum of the oocytes. Finally, the percent soluble Nup was defined as
100% minus the percentage of Nup in both nuclei and foci. Total
Nup expression was measured as the intensity sum of the -3 and -4
oocytes normalized to volume. To control for autofluorescent back-
ground in all measurements, staged animals lacking fluorescent tags
were imaged using identical imaging settings. The average intensity
sum per volume was calculated for the -3 and -4 oocytes of germ-
lines lacking fluorescent tags and subtracted from the intensity sum
measured for oocytes with tagged Nups. To measure the intensity of
GFP::Nup35 per nuclear volume (Appendix Fig S5H), the intensity
value for the -3 and -4 oocyte nuclei or 3 embryonic nuclei were
divided by nuclear volume and the resulting values were averaged
for each germline or embryo.
To quantify the distribution of Nups in embryos, Z stacks were
exported to Imaris software. The Surface tool was used to isolate the
entire embryo as well as all nuclei, and the Spot tool was used to
isolate cytoplasmic foci. The percent of Nup at the NE/nucleoplasm
or foci was measured as the intensity sum of all nuclei or foci
divided by the total intensity sum of the embryo, respectively. The
percent soluble Nup was defined as 100% minus the percentage of
Nup in nuclei and foci. For all measurements, embryos lacking fluo-
rescent tags were used to control for autofluorescent background as
described for oocytes.
The Y complex component Nup85 localizes to meiotic chromo-
somes and a high percentage of Nup85 is present in the nucleo-
line-scan analysis was used to measure the
plasm. Therefore,
amount of GFP::Nup85 at the NE (Fig 3B). Z stacks were exported
to ImageJ and line traces were drawn to pass through the central
plane of -3 and -4 oocyte nuclei as well as the image background.
For each nucleus, the two peak values of the NE rim were averaged
and normalized to the image background. Line-scan analysis was
also used to quantify depletion of endogenous Nup62::wrmScarlet
from the NE in neurons expressing rab-3p::Nup98::mNeonGreen
(Fig 7A). Line traces were drawn to pass through the central plane
of nuclei identified by Hoechst staining. For each nucleus, the two
peak values for the NE rim were averaged and normalized to the
average value of Nup62 in the nucleoplasm. To quantify the parti-
tioning of TFEB::GFP, IBBdomain::mNeonGreen, TDP-43::wrmScarlet,
CRM1::mNeonGreen, and G3BP::mCherry between the nucleus and
cytoplasm (Fig 5E, Appendix Figs S5K and S7E) line traces were
drawn to pass through the cytoplasm as well as the nucleoplasm
and image background. Average intensity values for the nucleus
and cytoplasm were background subtracted, then the value for the
nucleus was divided by that of the cytoplasm.
(cid:1) 2023 The Authors
The EMBO Journal 42: e112987 | 2023
15 of 19
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License The EMBO Journal
Laura Thomas et al
To quantify cytoplasmic levels of mNeonGreen::Nup358 in
oocytes versus somatic cells and embryos (Fig 1D), single focal
planes captured from the same animal were exported to ImageJ. For
each animal 3 ROIs in the -1 oocyte, somatic cell type of interest, or
4-cell embryo cytoplasm were measured, averaged, and normalized
to the image background. Values for the somatic cell or embryo
cytoplasm were then normalized to that of the oocyte within the
same animal.
To quantify foci formation in somatic tissues of aged animals
(Fig 6D), Z stacks of Day 4 adult heads expressing GFP::Nup88 were
manually scored into three categories: 1. None (no ectopic GFP::
Nup88 foci were present); 2. Mild (several small GFP::Nup88 foci were
observed); 3. Moderate (many large GFP::Nup88 foci were present).
Statistical analysis
All the statistical tests were performed using GraphPad Prism 9.2.0
software. For comparison of three or more groups, significance was
determined using a one-way ANOVA. For comparison of
two
groups, significance was determined using an unpaired t-test. In all
figures, error bars represent 95% confidence intervals (CIs). For all
**P < 0.01;
figures, ns
***P < 0.001; ****P < 0.0001.
indicates not
*P < 0.05;
significant;
Data availability
Original high resolution Z stacks for all images used in figures have
been deposited in the BioImage Archive: accession number
S-BIAD651 (https://www.ebi.ac.uk/biostudies/BioImages/studies/
S-BIAD651?query=S-BIAD651).
Expanded View for this article is available online.
Investigation: LT, BTI; Formal analysis: LT, BTI; Validation: LT, BTI; Visualization:
LT, BTI; Conceptualization: LT, GS; Writing – original draft: LT, GS; Writing –
review and editing: PA; Funding acquisition: GS, PA; Supervision: GS, PA.
Disclosure and competing interests statement
The authors declare that they have no conflict of interest.
References
Abbatemarco S, Bondaz A, Schwager F, Wang J, Hammell CM, Gotta M (2021)
PQN-59 and GTBP-1 contribute to stress granule formation but are not
essential for their assembly in C. elegans embryos. J Cell Sci 134: jcs258834
Agote-Aran A, Schmucker S, Jerabkova K, Jmel Boyer I, Berto A, Pacini L,
Ronchi P, Kleiss C, Guerard L, Schwab Y et al (2020) Spatial control of
nucleoporin condensation by fragile X-related proteins. EMBO J 39:
e104467
Alberti S, Gladfelter A, Mittag T (2019) Considerations and challenges in
studying liquid-liquid phase separation and biomolecular condensates.
Cell 176: 419 – 434
Anderson EN, Morera AA, Kour S, Cherry JD, Ramesh N, Gleixner A, Schwartz
JC, Ebmeier C, Old W, Donnelly CJ et al (2021) Traumatic injury
compromises nucleocytoplasmic transport and leads to TDP-43 pathology.
Elife 10: e67587
Arribere JA, Bell RT, Fu BXH, Artiles KL, Hartman PS, Fire AZ (2014) Efficient
marker-free recovery of custom genetic modifications with CRISPR/Cas9 in
Caenorhabditis elegans. Genetics 198: 837 – 846
Beckhelling C, Chang P, Chevalier S, Ford C, Houliston E (2003) Pre-M phase-
promoting factor associates with annulate lamellae in Xenopus oocytes
and egg extracts. Mol Biol Cell 14: 1125 – 1137
Ben-Zvi A, Miller EA, Morimoto RI (2009) Collapse of proteostasis represents
an early molecular event in Caenorhabditis elegans aging. Proc Natl Acad
Sci U S A 106: 14914 – 14919
Bohnert KA, Kenyon C (2017) A lysosomal switch triggers proteostasis renewal
Acknowledgements
We thank all members of the Seydoux and Cochella Labs, Orna Cohen-Fix, and
in the immortal C. elegans germ lineage. Nature 551: 629 – 633
Chandra S, Lusk CP (2022) Emerging connections between nuclear pore
the Baltimore Worm Club for support and many helpful discussions. We thank
complex homeostasis and ALS. Int J Mol Sci 23: 1329
Cristina Ayuso and Skyler Lemmon for assistance with C. elegans genome
engineering, Madeline Cassani for generating strain JH3656, and the Fromme
Lab for generously sharing yeast strains and plasmids. We thank the Chuang
Lab for sharing the transportin::mNeonGreen strain and the Greenstein Lab for
sharing the GFP::NDC1 strain. Several C. elegans strains were provided by the
Caenorhabditis Genetics Center (CGC), which is supported by the National
Institutes of Health Office of Research Infrastructure Programs (P40
OD010440). This work was funded by the Agencia Estatal de Investigaci(cid:2)on
(PID2019-105069GB-I00) and the National Institutes of Health (R37HD037047).
LT is a postdoctoral fellow of the Life Sciences Research Foundation supported
by the Howard Hughes Medical Institute. GS is an investigator of the Howard
Hughes Medical Institute.
Author contributions
Laura Thomas: Conceptualization; formal analysis; validation; investigation;
visualization; writing – original draft. Basma Taleb Ismail: Formal analysis;
validation; investigation; visualization. Peter Askjaer: Supervision; funding
acquisition; writing – review and editing. Geraldine Seydoux:
Conceptualization; supervision; funding acquisition; writing – original draft.
In addition to the CRediT author contributions listed above, the contributions
in detail are:
Brenner S (1974) The genetics of Caenorhabditis elegans. Genetics 77: 71 – 94
Chandra B, Michmerhuizen NL, Shirnekhi HK, Tripathi S, Pioso BJ, Baggett
DW, Mitrea DM, Iacobucci I, White MR, Chen J et al (2022) Phase
separation mediates NUP98 fusion oncoprotein leukemic transformation.
Cancer Discov 12: 1152 – 1169
Chase D, Serafinas C, Ashcroft N, Kosinski M, Longo D, Ferris DK, Golden A
(2000) The polo-like kinase PLK-1 is required for nuclear envelope
breakdown and the completion of meiosis in Caenorhabditis elegans.
Genesis 26: 26 – 41
Cho NH, Cheveralls KC, Brunner A-D, Kim K, Michaelis AC, Raghavan P,
Kobayashi H, Savy L, Li JY, Canaj H et al (2022) OpenCell: endogenous
tagging for the cartography of human cellular organization. Science 375:
eabi6983
Cohen-Fix O, Askjaer P (2017) Cell biology of the Caenorhabditis elegans
nucleus. Genetics 205: 25 – 59
Colombi P, Webster BM, Fröhlich F, Lusk CP (2013) The transmission of
nuclear pore complexes to daughter cells requires a cytoplasmic pool of
Nsp1. J Cell Biol 203: 215 – 232
Cordes VC, Reidenbach S, Franke WW (1996) Cytoplasmic annulate lamellae
in cultured cells: composition, distribution, and mitotic behavior. Cell
Tissue Res 284: 177 – 191
16 of 19
The EMBO Journal 42: e112987 | 2023
(cid:1) 2023 The Authors
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Laura Thomas et al
The EMBO Journal
D’Angelo MA, Raices M, Panowski SH, Hetzer MW (2009) Age-dependent
Griffis ER, Altan N, Lippincott-Schwartz J, Powers MA (2002) Nup98 is a
deterioration of nuclear pore complexes causes a loss of nuclear integrity
in postmitotic cells. Cell 136: 284 – 295
mobile nucleoporin with transcription-dependent dynamics. Mol Biol Cell
13: 1282 – 1297
Davis LI, Blobel G (1986) Identification and characterization of a nuclear pore
complex protein. Cell 45: 699 – 709
De Souza CPC, Osmani AH, Hashmi SB, Osmani SA (2004) Partial nuclear
pore complex disassembly during closed mitosis in Aspergillus nidulans.
Curr Biol 14: 1973 – 1984
Dimitriadi M, Hart AC (2010) Neurodegenerative disorders: insights from the
nematode Caenorhabditis elegans. Neurobiol Dis 40: 4 – 11
Dokshin GA, Ghanta KS, Piscopo KM, Mello CC (2018) Robust genome editing
with short single-stranded and long, partially single-stranded DNA donors
in Caenorhabditis elegans. Genetics 210: 781 – 787
Elaswad MT, Munderloh C, Watkins BM, Sharp KG, Breton E, Schisa JA (2022)
Imaging-associated stress causes divergent phase transitions of RNA-binding
proteins in the Caenorhabditis elegans germ line. G3 (Bethesda) 12: jkac172
Fallini C, Khalil B, Smith CL, Rossoll W (2020) Traffic jam at the nuclear pore:
all roads lead to nucleocytoplasmic transport defects in ALS/FTD. Neurobiol
Dis 140: 104835
Fan X, De Henau S, Feinstein J, Miller SI, Han B, Frøkjær-Jensen C, Griffin EE
(2020) SapTrap assembly of Caenorhabditis elegans MosSCI transgene
vectors. G3 (Bethesda) 10: 635 – 644
Fornerod M, van Deursen J, van Baal S, Reynolds A, Davis D, Murti KG,
Fransen J, Grosveld G (1997) The human homologue of yeast CRM1 is in a
dynamic subcomplex with CAN/Nup214 and a novel nuclear pore
component Nup88. EMBO J 16: 807 – 816
Forsythe ME, Love DC, Lazarus BD, Kim EJ, Prinz WA, Ashwell G, Krause MW,
Hanover JA (2006) Caenorhabditis elegans ortholog of a diabetes susceptibility
locus: oga-1 (O-GlcNAcase) knockout impacts O-GlcNAc cycling, metabolism,
and dauer. Proc Natl Acad Sci U S A 103: 11952 – 11957
Fragoso-Luna A, Romero-Bueno R, Eibl M, Ayuso C, Mu~noz-Jim(cid:2)enez C, Benes
V, Cases I, Askjaer P (2023) Expanded FLP toolbox for spatiotemporal
protein degradation and transcriptomic profiling in Caenorhabditis elegans.
Genetics 223: iyac166
Frey S, Görlich D (2007) A saturated FG-repeat hydrogel can reproduce the
permeability properties of nuclear pore complexes. Cell 130: 512 – 523
Frey S, Richter RP, Görlich D (2006) FG-rich repeats of nuclear pore proteins
form a three-dimensional meshwork with hydrogel-like properties. Science
314: 815 – 817
Galy V, Mattaj IW, Askjaer P (2003) Caenorhabditis elegans nucleoporins
Nup93 and Nup205 determine the limit of nuclear pore complex size
exclusion in vivo. Mol Biol Cell 14: 12
Galy V, Antonin W, Jaedicke A, Sachse M, Santarella R, Haselmann U, Mattaj I
(2008) A role for gp210 in mitotic nuclear-envelope breakdown. J Cell Sci
121: 317 – 328
Gasset-Rosa F, Lu S, Yu H, Chen C, Melamed Z, Guo L, Shorter J, Da Cruz S,
Cleveland DW (2019) Cytoplasmic TDP-43 de-mixing independent of stress
granules drives inhibition of nuclear import, loss of nuclear TDP-43, and
cell death. Neuron 102: 339 – 357
Gleixner AM, Verdone BM, Otte CG, Anderson EN, Ramesh N, Shapiro OR,
Gale JR, Mauna JC, Mann JR, Copley KE et al (2022) NUP62 localizes to
ALS/FTLD pathological assemblies and contributes to TDP-43 insolubility.
Nat Commun 13: 3380
G(cid:2)omez-Saldivar G, Fernandez A, Hirano Y, Mauro M, Lai A, Ayuso C,
Haraguchi T, Hiraoka Y, Piano F, Askjaer P (2016) Identification of
conserved MEL-28/ELYS domains with essential roles in nuclear assembly
and chromosome segregation. PLoS Genet 12: e1006131
Griffis ER, Xu S, Powers MA (2003) Nup98 localizes to both nuclear and
cytoplasmic sides of the nuclear pore and binds to two distinct
nucleoporin subcomplexes. Mol Biol Cell 14: 600 – 610
Guo L, Kim HJ, Wang H, Monaghan J, Freyermuth F, Sung JC, O’Donovan K,
Fare CM, Diaz Z, Singh N et al (2018) Nuclear-import receptors reverse
aberrant phase transitions of RNA-binding proteins with prion-like
domains. Cell 173: 677 – 692
Hampoelz B, Mackmull M-T, Machado P, Ronchi P, Bui KH, Schieber N,
Santarella-Mellwig R, Necakov A, Andr(cid:2)es-Pons A, Philippe JM et al (2016)
Pre-assembled nuclear pores insert into the nuclear envelope during early
development. Cell 166: 664 – 678
Hampoelz B, Andres-Pons A, Kastritis P, Beck M (2019a) Structure and
assembly of the nuclear pore complex. Annu Rev Biophys 48: 515 – 536
Hampoelz B, Schwarz A, Ronchi P, Bragulat-Teixidor H, Tischer C, Gaspar I,
Ephrussi A, Schwab Y, Beck M (2019b) Nuclear pores assemble from
nucleoporin condensates during oogenesis. Cell 179: 671 – 686
Hanover JA, Forsythe ME, Hennessey PT, Brodigan TM, Love DC, Ashwell G,
Krause M (2005) A Caenorhabditis elegans model of insulin resistance:
Altered macronutrient storage and dauer formation in an OGT-1 knockout.
Proc Natl Acad Sci U S A 102: 11266 – 11271
Hattersley N, Cheerambathur D, Moyle M, Stefanutti M, Richardson A, Lee K-
Y, Dumont J, Oegema K, Desai A (2016) A nucleoporin docks protein
phosphatase 1 to direct meiotic chromosome segregation and nuclear
assembly. Dev Cell 38: 463 – 477
Herndon LA, Schmeissner PJ, Dudaronek JM, Brown PA, Listner KM, Sakano Y,
Paupard MC, Hall DH, Driscoll M (2002) Stochastic and genetic factors
influence tissue-specific decline in ageing C. elegans. Nature 419: 808 – 814
Hofweber M, Hutten S, Bourgeois B, Spreitzer E, Niedner-Boblenz A, Schifferer
M, Ruepp M-D, Simons M, Niessing D, Madl T et al (2018) Phase
separation of FUS is suppressed by its nuclear import receptor and
arginine methylation. Cell 173: 706 – 719
Huelgas-Morales G, Greenstein D (2018) Control of oocyte meiotic maturation
in C. elegans. Semin Cell Dev Biol 84: 90 – 99
Huelgas-Morales G, Sanders M, Mekonnen G, Tsukamoto T, Greenstein D
(2020) Decreased mechanotransduction prevents nuclear collapse in a
Caenorhabditis elegans laminopathy. Proc Natl Acad Sci 117: 31301 – 31308
H€ulsmann BB, Labokha AA, Görlich D (2012) The permeability of reconstituted
nuclear pores provides direct evidence for the selective phase model. Cell
150: 738 – 751
Hutten S, Dormann D (2020) Nucleocytoplasmic transport defects in
neurodegeneration — cause or consequence? Semin Cell Dev Biol 99:
151 – 162
Hutten S, Usluer S, Bourgeois B, Simonetti F, Odeh HM, Fare CM, Czuppa M,
Hruska-Plochan M, Hofweber M, Polymenidou M et al (2020) Nuclear
import receptors directly bind to arginine-rich dipeptide repeat proteins
and suppress their pathological interactions. Cell Rep 33: 108538
Jud M, Razelun J, Bickel J, Czerwinski M, Schisa JA (2007) Conservation of
large foci formation in arrested oocytes of Caenorhabditis nematodes. Dev
Genes Evol 217: 221 – 226
Jud MC, Czerwinski MJ, Wood MP, Young RA, Gallo CM, Bickel JS, Petty EL,
Mason JM, Little BA, Padilla PA et al (2008) Large P body-like RNPs form
in C. elegans oocytes in response to arrested ovulation, heat shock,
osmotic stress, and anoxia and are regulated by the major sperm protein
pathway. Dev Biol 318: 38 – 51
(cid:1) 2023 The Authors
The EMBO Journal 42: e112987 | 2023
17 of 19
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License The EMBO Journal
Laura Thomas et al
Kessel G (1989) The annulate lamellae–from obscurity to spotlight. Electron
Microsc Rev 2: 257 – 348
Khalil B, Chhangani D, Wren MC, Smith CL, Lee JH, Li X, Puttinger C, Tsai C-
W, Fortin G, Morderer D et al (2022) Nuclear import receptors are
recruited by FG-nucleoporins to rescue hallmarks of TDP-43
proteinopathy. Mol Neurodegener 17: 80
Kutay U, J€uhlen R, Antonin W (2021) Mitotic disassembly and reassembly of
Ng SC, G€uttler T, Görlich D (2021) Recapitulation of selective nuclear import
and export with a perfectly repeated 12mer GLFG peptide. Nat Commun
12: 4047
Ohn T, Kedersha N, Hickman T, Tisdale S, Anderson P (2008) A functional
RNAi screen links O-GlcNAc modification of ribosomal proteins to
stress granule and processing body assembly. Nat Cell Biol 10:
1224 – 1231
nuclear pore complexes. Trends Cell Biol 31: 1019 – 1033
Labokha AA, Gradmann S, Frey S, H€ulsmann BB, Urlaub H, Baldus M, Görlich
Onischenko EA, Gubanova NV, Kieselbach T, Kiseleva EV, Hallberg E (2004)
Annulate lamellae play only a minor role in the storage of excess
D (2012) Systematic analysis of barrier-forming FG hydrogels from
Xenopus nuclear pore complexes. EMBO J 32: 204 – 218
Langerak S, Trombley A, Patterson JR, Leroux D, Couch A, Wood MP, Schisa JA
(2019) Remodeling of the endoplasmic reticulum in Caenorhabditis elegans
oocytes is regulated by CGH-1. Genesis 57: e23267
Laurell E, Beck K, Krupina K, Theerthagiri G, Bodenmiller B, Horvath P,
Aebersold R, Antonin W, Kutay U (2011) Phosphorylation of Nup98 by
multiple kinases is crucial for NPC disassembly during mitotic entry. Cell
144: 539 – 550
Lee BE, Suh P-G, Kim J-I (2021) O-GlcNAcylation in health and
neurodegenerative diseases. Exp Mol Med 53: 1674 – 1682
Linder MI, Köhler M, Boersema P, Weberruss M, Wandke C, Marino J, Ashiono
C, Picotti P, Antonin W, Kutay U (2017) Mitotic disassembly of nuclear
pore complexes involves CDK1- and PLK1-mediated phosphorylation of
key interconnecting nucleoporins. Dev Cell 43: 141 – 156
Longtine MS, Mckenzie A III, Demarini DJ, Shah NG, Wach A, Brachat A,
Philippsen P, Pringle JR (1998) Additional modules for versatile and
economical PCR-based gene deletion and modification in Saccharomyces
cerevisiae. Yeast 14: 953 – 961
Lott K, Cingolani G (2011) The importin b binding domain as a master
regulator of nucleocytoplasmic transport. Biochim Biophys Acta 1813:
1578 – 1592
Mac(cid:2)ıas-Le(cid:2)on J, Askjaer P (2018) Efficient FLP-mediated germ-line
recombination in C. elegans. MicroPubl Biol 2018: W2G66S
Makio T, Stanton LH, Lin C-C, Goldfarb DS, Weis K, Wozniak RW (2009) The
nucleoporins Nup170p and Nup157p are essential for nuclear pore
complex assembly. J Cell Biol 185: 459 – 473
Mansfeld J, G€uttinger S, Hawryluk-Gara LA, Pant(cid:2)e N, Mall M, Galy V,
Haselmann U, M€uhlh€ausser P, Wozniak RW, Mattaj IW et al (2006)
The conserved transmembrane nucleoporin NDC1 is required for
nuclear pore complex assembly in vertebrate cells. Mol Cell 22:
93 – 103
Martino L, Morchoisne-Bolhy S, Cheerambathur DK, Van Hove L, Dumont J,
Joly N, Desai A, Doye V, Pintard L (2017) Channel nucleoporins recruit PLK-
1 to nuclear pore complexes to direct nuclear envelope breakdown in C.
elegans. Dev Cell 43: 157 – 171
Mauro MS, Celma G, Zimyanin V, Magaj MM, Gibson KH, Redemann S,
Bahmanyar S (2022) Ndc1 drives nuclear pore complex assembly
independent of membrane biogenesis to promote nuclear formation and
growth. Elife 11: e75513
nucleoporins in Drosophila embryos: annulate lamellae assembly in
Drosophila. Traffic 5: 152 – 164
Onischenko EA, Gubanova NV, Kiseleva EV, Hallberg E (2005) Cdk1 and
okadaic acid-sensitive phosphatases control assembly of nuclear pore
complexes in Drosophila embryos. Mol Biol Cell 16: 5152 – 5162
Onischenko E, Stanton LH, Madrid AS, Kieselbach T, Weis K (2009) Role of the
Ndc1 interaction network in yeast nuclear pore complex assembly and
maintenance. J Cell Biol 185: 475 – 491
Onischenko E, Tang JH, Andersen KR, Knockenhauer KE, Vallotton P, Derrer
CP, Kralt A, Mugler CF, Chan LY, Schwartz TU et al (2017) Natively
unfolded FG repeats stabilize the structure of the nuclear pore complex.
Cell 171: 904 – 917
Paix A, Folkmann A, Seydoux G (2017) Precision genome editing using
CRISPR-Cas9 and linear repair templates in C. elegans. Methods 121–122:
86 – 93
Patel SS, Belmont BJ, Sante JM, Rexach MF (2007) Natively unfolded
nucleoporins gate protein diffusion across the nuclear pore complex. Cell
129: 83 – 96
Patterson JR, Wood MP, Schisa JA (2011) Assembly of RNP granules in
stressed and aging oocytes requires nucleoporins and is coordinated with
nuclear membrane blebbing. Dev Biol 353: 173 – 185
Pitt JN, Schisa JA, Priess JR (2000) P granules in the germ cells of
Caenorhabditis elegans adults are associated with clusters of nuclear pores
and contain RNA. Dev Biol 219: 315 – 333
Port SA, Monecke T, Dickmanns A, Spillner C, Hofele R, Urlaub H, Ficner R,
Kehlenbach RH (2015) Structural and functional characterization of
CRM1-Nup214 interactions reveals multiple FG-binding sites involved in
nuclear export. Cell Rep 13: 690 – 702
Prophet SM, Rampello AJ, Niescier RF, Gentile JE, Mallik S, Koleske AJ,
Schlieker C (2022) Atypical nuclear envelope condensates linked to
neurological disorders reveal nucleoporin-directed chaperone activities.
Nat Cell Biol 24: 1630 – 1641
Raghunayakula S, Subramonian D, Dasso M, Kumar R, Zhang X-D (2015)
Molecular characterization and functional analysis of annulate lamellae
pore complexes in nuclear transport in mammalian cells. PLoS One 10:
e0144508
Rahman MM, Munzig M, Kaneshiro K, Lee B, Strome S, M€uller-Reichert T,
Cohen-Fix O (2015) Caenorhabditis elegans polo-like kinase PLK-1 is
required for merging parental genomes into a single nucleus. Mol Biol Cell
26: 4718 – 4735
Miller BR, Forbes DJ (2000) Purification of the vertebrate nuclear pore
Ren H, Xin G, Jia M, Zhu S, Lin Q, Wang X, Jiang Q, Zhang C (2019)
complex by biochemical criteria. Traffic 12: 941 – 951
Milles S, Huy Bui K, Koehler C, Eltsov M, Beck M, Lemke EA (2013) Facilitated
aggregation of FG nucleoporins under molecular crowding conditions.
EMBO Rep 14: 178 – 183
Morchoisne-Bolhy S, Geoffroy M-C, Bouhlel IB, Alves A, Audug(cid:2)e N, Baudin X,
Van Bortle K, Powers MA, Doye V (2015) Intranuclear dynamics of the
Nup107-160 complex. Mol Biol Cell 26: 2343 – 2356
Postmitotic annulate lamellae assembly contributes to nuclear envelope
reconstitution in daughter cells. J Biol Chem 294: 10383 – 10391
Ribbeck K, Görlich D (2001) Kinetic analysis of translocation through nuclear
pore complexes. EMBO J 20: 1320 – 1330
Ritterhoff T, Das H, Hofhaus G, Schröder RR, Flotho A, Melchior F (2016) The
RanBP2/RanGAP1*SUMO1/Ubc9 SUMO E3 ligase is a disassembly machine
for Crm1-dependent nuclear export complexes. Nat Commun 7: 11482
18 of 19
The EMBO Journal 42: e112987 | 2023
(cid:1) 2023 The Authors
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Laura Thomas et al
The EMBO Journal
R(cid:2)odenas E, Klerkx EP, Ayuso C, Audhya A, Askjaer P (2009) Early embryonic
requirement for nucleoporin Nup35/NPP-19 in nuclear assembly. Dev Biol
327: 399 – 409
R(cid:2)odenas E, Gonz(cid:2)alez-Aguilera C, Ayuso C, Askjaer P (2012) Dissection of the
NUP107 nuclear pore subcomplex reveals a novel interaction with spindle
assembly checkpoint protein MAD1 in Caenorhabditis elegans. Mol Biol Cell
23: 930 – 944
Ruba A, Yang W (2016) O-GlcNAc-ylation in the nuclear pore complex. Cell
Mol Bioeng 9: 227 – 233
Sahoo MR, Gaikwad S, Khuperkar D, Ashok M, Helen M, Yadav SK, Singh A,
Magre I, Deshmukh P, Dhanvijay S et al (2017) Nup358 binds to AGO
proteins through its SUMO -interacting motifs and promotes the
association of target MRNA with miRISC. EMBO Rep 18: 241 – 263
Schedl T, Kimble J (1988) fog-2, a germ-line-specific sex determination gene
required for hermaphrodite spermatogenesis in Caenorhabditis elegans.
Genetics 119: 43 – 61
Schmidt HB, Görlich D (2015) Nup98 FG domains from diverse species
spontaneously phase-separate into particles with nuclear pore-like
permselectivity. Elife 4: e04251
Biomolecular condensation of NUP98 fusion proteins drives leukemogenic
gene expression. Nat Struct Mol Biol 28: 190 – 201
Thomas LL, van der Vegt SA, Fromme JC (2019) A steric gating mechanism
dictates the substrate specificity of a Rab-GEF. Dev Cell 48: 100 – 114
Timmons L, Fire A (1998) Specific interference by ingested dsRNA. Nature 395:
854
Toyama BH, Savas JN, Park SK, Harris MS, Ingolia NT, Yates JR, Hetzer MW
(2013) Identification of long-lived proteins reveals exceptional stability of
essential cellular structures. Cell 154: 971 – 982
Vicencio J, Mart(cid:2)ınez-Fern(cid:2)andez C, Serrat X, Cer(cid:2)on J (2019) Efficient generation
of endogenous fluorescent reporters by nested CRISPR in Caenorhabditis
elegans. Genetics 211: 1143 – 1154
Wippich F, Bodenmiller B, Trajkovska MG, Wanka S, Aebersold R, Pelkmans L
(2013) Dual specificity kinase DYRK3 couples stress granule condensation/
dissolution to mTORC1 signaling. Cell 152: 791 – 805
Wu X, Kasper LH, Mantcheva RT, Mantchev GT, Springett MJ, van Deursen
JMA (2001) Disruption of the FG nucleoporin NUP98 causes selective
changes in nuclear pore complex stoichiometry and function. Proc Natl
Acad Sci U S A 98: 3191 – 3196
Schmidt HB, Görlich D (2016) Transport selectivity of nuclear pores, phase
Xu S, Powers MA (2013) In vivo analysis of human nucleoporin repeat
separation, and membraneless organelles. Trends Biochem Sci 41: 46 – 61
domain interactions. Mol Biol Cell 24: 1222 – 1231
Schwartz ML, Jorgensen EM (2016) SapTrap, a toolkit for high-throughput
CRISPR/Cas9 gene modification in Caenorhabditis elegans. Genetics 202:
1277 – 1288
Xylourgidis N, Roth P, Sabri N, Tsarouhas V, Samakovlis C (2006) The
nucleoporin Nup214 sequesters CRM1 at the nuclear rim and modulates
NFjB activation in Drosophila. J Cell Sci 119: 4409 – 4419
Sheth U, Pitt J, Dennis S, Priess JR (2010) Perinuclear P granules are the
Yoo TY, Mitchison TJ (2021) O-GlcNAc modification of nuclear pore complexes
principal sites of mRNA export in adult C. elegans germ cells. Development
137: 1305 – 1314
Stafstrom JP, Staehelin LA (1984) Are annulate lamellae in the Drosophila
accelerates bidirectional transport. J Cell Biol 220: e202010141
Yoshida K, Seo H-S, Debler EW, Blobel G, Hoelz A (2011) Structural and
functional analysis of an essential nucleoporin heterotrimer on the
embryo the result of overproduction of nuclear pore components? J Cell
Biol 98: 699 – 708
Stavru F, H€ulsmann BB, Spang A, Hartmann E, Cordes VC, Görlich D (2006)
NDC1: a crucial membrane-integral nucleoporin of metazoan nuclear pore
complexes. J Cell Biol 173: 509 – 519
Strawn LA, Shen T, Shulga N, Goldfarb DS, Wente SR (2004) Minimal nuclear
pore complexes define FG repeat domains essential for transport. Nat Cell
Biol 6: 197 – 206
Tan PS, Aramburu IV, Mercadante D, Tyagi S, Chowdhury A, Spitz D, Shammas
SL, Gr€ater F, Lemke EA (2018) Two differential binding mechanisms of FG-
nucleoporins and nuclear transport receptors. Cell Rep 22: 3660 – 3671
Terlecki-Zaniewicz S, Humer T, Eder T, Schmoellerl J, Heyes E, Manhart G,
Kuchynka N, Parapatics K, Liberante FG, M€uller AC et al (2021)
cytoplasmic face of the nuclear pore complex. Proc Natl Acad Sci U S A
108: 16571 – 16576
Zhang K, Daigle JG, Cunningham KM, Coyne AN, Ruan K, Grima JC, Bowen KE,
Wadhwa H, Yang P, Rigo F et al (2018) Stress granule assembly disrupts
nucleocytoplasmic transport. Cell 173: 958 – 971
Zhou M-H, Yang Q-M (2014) NUP214 fusion genes in acute leukemia
(Review). Oncol Lett 8: 959 – 962
License: This is an open access article under the
terms of the Creative Commons Attribution License,
which permits use, distribution and reproduction in
any medium, provided the original work is properly
cited.
(cid:1) 2023 The Authors
The EMBO Journal 42: e112987 | 2023
19 of 19
14602075, 2023, 13, Downloaded from https://www.embopress.org/doi/10.15252/embj.2022112987 by Cochrane France, Wiley Online Library on [13/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
| null |
10.2196_20920.pdf
| null | null |
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
Original Paper
Evaluating Behavioral and Linguistic Changes During Drug
Treatment for Depression Using Tweets in Spanish: Pairwise
Comparison Study
Angela Leis*, PsyM; Francesco Ronzano*, PhD; Miguel Angel Mayer, MD, MPH, PhD; Laura I Furlong, PhD; Ferran
Sanz, Prof Dr
Research Programme on Biomedical Informatics, Hospital del Mar Medical Research Institute, Department of Experimental and Health Sciences,
Pompeu Fabra University, Barcelona, Spain
*these authors contributed equally
Corresponding Author:
Ferran Sanz, Prof Dr
Research Programme on Biomedical Informatics
Hospital del Mar Medical Research Institute
Department of Experimental and Health Sciences, Pompeu Fabra University
Barcelona Biomedical Research Park
Carrer Dr Aiguader 88
Barcelona, 08003
Spain
Phone: 34 933160540
Fax: 34 933160550
Email: [email protected]
Abstract
Background: Depressive disorders are the most common mental illnesses, and they constitute the leading cause of disability
worldwide. Selective serotonin reuptake inhibitors (SSRIs) are the most commonly prescribed drugs for the treatment of depressive
disorders. Some people share information about their experiences with antidepressants on social media platforms such as Twitter.
Analysis of the messages posted by Twitter users under SSRI treatment can yield useful information on how these antidepressants
affect users’ behavior.
Objective: This study aims to compare the behavioral and linguistic characteristics of the tweets posted while users were likely
to be under SSRI treatment, in comparison to the tweets posted by the same users when they were less likely to be taking this
medication.
Methods: In the first step, the timelines of Twitter users mentioning SSRI antidepressants in their tweets were selected using
a list of 128 generic and brand names of SSRIs. In the second step, two datasets of tweets were created, the in-treatment dataset
(made up of the tweets posted throughout the 30 days after mentioning an SSRI) and the unknown-treatment dataset (made up
of tweets posted more than 90 days before or more than 90 days after any tweet mentioning an SSRI). For each user, the changes
in behavioral and linguistic features between the tweets classified in these two datasets were analyzed. 186 users and their timelines
with 668,842 tweets were finally included in the study.
Results: The number of tweets generated per day by the users when they were in treatment was higher than it was when they
were in the unknown-treatment period (P=.001). When the users were in treatment, the mean percentage of tweets posted during
the daytime (from 8 AM to midnight) increased in comparison to the unknown-treatment period (P=.002). The number of characters
and words per tweet was higher when the users were in treatment (P=.03 and P=.02, respectively). Regarding linguistic features,
the percentage of pronouns that were first-person singular was higher when users were in treatment (P=.008).
Conclusions: Behavioral and linguistic changes have been detected when users with depression are taking antidepressant
medication. These features can provide interesting insights for monitoring the evolution of this disease, as well as offering
additional information related to treatment adherence. This information may be especially useful in patients who are receiving
long-term treatments such as people suffering from depression.
(J Med Internet Res 2020;22(12):e20920) doi: 10.2196/20920
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 1
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
KEYWORDS
depression; antidepressant drugs; serotonin uptake inhibitors; mental health; social media; infodemiology; data mining
Introduction
Background
Depression is one of the most common mental disorders [1].
According to the World Health Organization, depression affects
more than 322 million people of all ages globally, being a
leading cause of disability worldwide [2]. The proportion of
people with depression went up by around 18% between 2005
and 2015 [3]. This mental disorder constitutes a challenge for
society and health care systems due to devastating personal and
social consequences and the associated economic costs [4-13].
In spite of the high prevalence of depression and the efforts of
health care services to improve its management, this health
condition remains underdiagnosed and undertreated [14].
In the case of moderate and severe forms of depression,
pharmacological treatment can improve the quality of life of
these patients [4]. There are several types of antidepressant
drugs, and among them, selective serotonin reuptake inhibitors
(SSRIs) are currently the most prescribed antidepressants around
the world. For instance, according to the Spanish Agency for
Medicines and Health Products [15], SSRIs constitute more
than 70% of all antidepressants prescribed in Spain. They have
fewer side effects than other antidepressants [16], show a good
risk-benefit ratio [17,18], are safer and better tolerated [19], and
exhibit a reduced risk of toxicity in overdose in comparison to
tricyclic antidepressants [20]. They are commonly used as
first-line treatment for depression [21-23] and are usually
prescribed as maintenance therapy to prevent relapse [4,23-26].
SSRIs include the following drugs: fluvoxamine, fluoxetine,
paroxetine, sertraline, citalopram, and escitalopram [17].
Furthermore, although social media platforms have typically
not been created with health-related purposes in mind [27,28],
millions of people publicly share personal health information
on social media platforms every day [29,30]. For this reason,
these platforms represent an important source of health
information that is faster and more broadly available than other
sources of health information, being unsolicited, spontaneous,
and up to date. Infodemiology approaches have been developed
and applied to better understand the dynamics of these platforms
when used as a health information source [31-33]. In this
context, social media users share health-related information,
such as experiences with prescribed drugs [34], cancer patients’
sentiments [35], opinions on vaccines [36], or online
conversations on epidemic outbreaks [37]. The massive data
from social media can be monitored and analyzed by using
natural language processing and machine learning technologies,
providing new possibilities to better understand users’ behavior
[30], including automatic identification of early signs of mental
disorders [38-40]. In particular, it is typical for people suffering
from depression to talk about their illness and the drugs they
are taking [41-43].
amount of data that can collected in real time [28,30,33,45-48].
Twitter users post short messages about facts, feelings, and
opinions, including about health conditions [49].
in
Mining of drug-related information from Twitter has been
applied
the pharmacovigilance field [27,50]. Some
pharmacovigilance studies carried out on Twitter studied specific
cohorts by identifying users’ mentions of drug intake [37,51-53].
Other studies focused on adverse drug reactions, analyzing
users’ tweets regarding adverse events and side effects
associated with drug use, which were identified by means of
generic or brand names [29,47,54,55]. In our previous study
[49], we observed that Twitter users who are potentially
suffering from depression show particular behavioral and
linguistic features in their tweets. These features were related
to an increase in their activity during the night, a different style
of writing with increased use of the first-person singular
pronoun, fewer characters in their tweets, an increase in the
frequency of words related to sadness and disgust emotions,
and more frequent presence of negation words and negative
polarity. This information can be used as a complementary tool
to detect signals of depression and for monitoring and supporting
patients using Twitter.
Objectives
In this paper, we aim to enrich our previous study [49] by
focusing on analysis of the changes in behavioral and linguistic
features of Twitter users in Spanish language, which may be
associated with the antidepressant medication these users are
taking. It is worth mentioning that users from Spanish-speaking
countries are among the most active on Twitter in the world
[56]. The study is focused on Twitter users who mention
treatment with SSRIs, which are the most frequently prescribed
antidepressants [15]. In particular, this study compares the
characteristics of the tweets posted while users were probably
taking SSRIs versus the tweets posted by the same users when
they have a lower probability of taking this antidepressant
medication. This analysis can contribute to better understanding
how these drugs affect users’ mood. Although we found two
additional studies describing changes in Twitter users’ language
in some mental disorders [57,58], to the best of our knowledge,
there are no other studies that analyze Twitter posts in Spanish
language to detect behavioral and linguistic changes when the
users are taking antidepressant medication.
Methods
Study Design
This study was designed with the aim of analyzing the
behavioral patterns and linguistic features of users who mention
SSRIs in their Twitter timeline. The study was developed in
several steps and focused on tweets written in Spanish. The
flow diagram of the study is depicted in Figure 1.
Twitter is a very popular microblogging platform with more
than 330 million active users worldwide [44]. Tweets, freely
available in almost 90% of users’ accounts, provide a huge
As shown in Figure 1, two nonoverlapping datasets of tweets
from users mentioning treatment with SSRIs were obtained: (1)
The in-treatment tweets dataset was made up of the tweets
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 2
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
posted throughout the 30 days after the publication date of any
tweet mentioning SSRI intake. We assumed that these tweets
were posted while the users had a high probability of being in
treatment with an SSRI. (2) The unknown-treatment tweets
dataset was made up of the tweets that were posted more than
90 days before or more than 90 days after the publication date
of any tweet mentioning SSRI intake. We assumed that these
tweets were posted while users had a lower probability of being
in treatment with an SSRI than in the previous dataset.
These datasets were designed in a way that made it possible to
carry out intrasubject comparisons, since the in-treatment tweets
and unknown-treatment tweets datasets were obtained from the
same Twitter users.
The strategy for the selection of the tweets included in the two
datasets is depicted in Figure 2.
Figure 1. Flow diagram of the study process. SSRI: selective serotonin reuptake inhibitor.
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 3
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
Figure 2. The in-treatment and unknown-treatment dataset selection strategy. SSRI: selective serotonin reuptake inhibitor.
Data Collection and User Selection
The selection of the tweets and their users was based on the
filtered real-time streaming support provided by the Twitter
application programming interface [59]. In the first step, we
selected tweets in Spanish that mention any of the SSRI generic
and brand names used around the world. To obtain the generic
and brand names, we performed searches on the following
databases and resources: DrugBank [60], the Anatomical
Therapeutic Chemical Classification System and the Defined
Daily Dose of the World Health Organization [61], Wikipedia
[62], and the Database for Pharmacoepidemiological Research
in Primary Care [63]. The list of 135 generic and brand names
obtained is shown in Table 1.
Table 1. Selective serotonin reuptake inhibitors (SSRIs) used in the study.
Generic name
Brand names
Fluvoxamina (fluvoxamine)
Dumirox, Faverin, Floxyfral, Fluvoxin, Luvox, Uvox
Fluoxetina (fluoxetine)
Paroxetina (paroxetine)
Prozac, Reneuron, Adofen, Luramon, Sarafem
Seroxat, Motivan, Frosinor, Praxil, Daparox, Xetin,
Sertralina (sertraline)
Citalopram (citalopram)
Apo-oxpar, Appoxar, Aropax, Aroxat, Aroxat CR, Bectam, Benepax, Casbol, Cebrilin, Deroxat,
Hemtrixil, Ixicrol, Loxamine, Meplar, Olane, Optipar, Oxetine, Pamax, ParadiseCR, Paradox,
Paraxyle, Parexis, Paroxat, Paroxet, Paxan, Paxera, Paxil, Paxil CR, Pexot, Plasare, Pondera,
Posivyl, Psicoasten, Rexetin, Seretran, Sereupin, Tiarix, Tamcere, Traviata, Xerenex, Xetroran
Aremis, Besitran, Zoloft, Altisben, Aserin, Altruline, Ariale, Asertral, Atenix, Eleval, Emergen,
Dominium, Inosert, Irradial, Sedora, Serolux, Sertex
Seropram, Celexa, Akarin, C Pram S, Celapram, Celica, Ciazil, Cilate, Cilift, Cimal, Cipralex,
Cipram, Cipramil, Cipraned, Cinapen, Ciprapine, Ciprotan, Citabax, Citaxin, Citalec, Citalex,
Citalo, Citalopram, Citol, Citox, Citrol, Citta, Dalsan, Denyl, Elopram, Estar, Humorup, Humorap,
Oropram, Opra, Pram, Pramcit, Procimax, Recital, Sepram, Szetalo, Talam, Temperax, Vodelax,
Zentius, Zetalo, Cipratal, Zylotex
Escitalopram (escitalopram)
Cipralex, Diprex, Esertia, Essential, Heipram, Lexapro
The following 7 brand names of medicines have been excluded
due to their semantic ambiguity: Essential, Motivan, Estar,
Traviata, Pondera, Recital, and Emergen. These commercial
names are, at the same time, very common words used with
different meanings in Spanish, as we verified after reviewing a
random sample of 200 tweets with mentions of these words.
The number of tweets excluded because of their semantic
ambiguity was 21,104. In the manual check of a random sample
of 200 tweets, the mentions of SSRIs when using these words
were 0% (0/200) in some cases, such as for Motivan and Estar,
and 0.5% (1/200) for Recital. The final list of words included
128 generic and brand names of SSRIs.
Using the aforementioned 128 SSRI names, we collected 3651
tweets in Spanish posted during November 2019 with
occurrences of the words listed in Table 1. These tweets were
posted by 3138 different Twitter users and mentioned 33
different words from the list. The frequencies of these 32 words
are shown in Table 2.
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 4
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
Table 2. Frequencies of SSRI names mentioned in Spanish tweets during November 2019.
SSRI mentions
Prozac
Fluoxetina
Sertralina
Escitalopram
Citta
Citalo
Paroxetina
Pram
Fluvoxamina
Citalopram
Seroxat
Eleval
Lexapro
Opra
Casbol
Ariale
Zoloft
Altruline
Paxil
Akarin
Heipram
Aremis
Cimal
Tiarix
Seretran
Dominium
Citox
Atenix
Aserin
Talam
Dalsan
Celexa
Frequency
998
756
542
248
210
109
69
49
40
33
22
21
20
18
14
11
9
9
7
7
4
4
3
2
2
2
2
2
2
1
1
1
In a second step, we crawled the public Twitter timelines of the
3138 users (until the 3200 most recent tweets for each user were
retrieved). Given that retweets are not useful for analyzing the
linguistic behavior of a particular user, the third step consisted
of excluding the retweets and checking if the remaining tweets
from each timeline included the mention of at least one SSRI.
1800 users were excluded by this filter, leaving a total of 1338
Twitter users. We obtained 3,791,609 tweets after compiling
the timelines from these 1338 users. From these timelines, 4872
tweets mentioning at least one of the SSRIs from the list were
automatically detected. These 4872 tweets were independently
reviewed by two experts, a psychologist and a family physician,
both with clinical experience. These experts manually selected
the tweets that suggested that the user who posted the tweet was
taking an SSRI on the date of posting. Examples of these tweets
are shown in Textbox 1.
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 5
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
Textbox 1. Examples of tweets that positively or negatively suggest whether the user is taking an SSRI.
Positive examples:
•
•
“Eso de tener sueños raros debido a la fluoxetina se está saliendo de control.” (“Having odd dreams due to fluoxetine is getting out of control.”)
“Yo tomo sertralina, como me lo receta el doctor y aún así a veces siento que el mundo donde estoy no es para mi. Ese susto esa angustia esas
ganas de correr es algo que sólo el que lo padece lo entiende” (“I take sertraline as my doctor prescribes it to me and, even so, sometimes I feel
that the world I’m living in is not for me. This fear this anxiety this desire to run out is something that only one who suffers from it can understand”)
Negative examples:
•
•
“Ella debería tomar prozac, como Tic Tac” (“She should take prozac, like Tic Tac” [a candy brand])
“La Paroxetina es un medicamento que pertenece a la familia de los antidepresivos inhibidores de la recaptación de la serotonina ¡Conoce más
sobre él!” (“Paroxetine is a drug that belongs to the antidepressant family of serotonin reuptake inhibitors. Find out more about it!”)
The agreement between reviewers was 93.1% (4537/4872) with
a Cohen kappa score of 0.68, indicating that there was
substantial agreement between raters. The reviewers discussed
and reached a consensus on the classification of the 335 tweets
they classified differently. Finally, we obtained a total number
of 518 tweets with one or more SSRI mentions, suggesting that
the users who posted these tweets were taking an SSRI at the
moment of posting. These tweets corresponded to 279 different
users. Therefore, these users had two characteristics: first, the
tweets on their timeline included at least one mention of SSRIs,
and second, the text of tweets mentioning SSRIs suggested that
the user was taking the antidepressant. In addition, we analyzed
the tweets posted by each user that belonged to the two datasets
(in-treatment and unknown-treatment; see Figure 1) by trying
different minimum numbers of tweets per dataset (10, 30, 60,
and 100 tweets) in order to include a user in the study. 10 tweets
contained little information in terms of number of words or
posting characteristics. In the cases of 60 and 100 tweets, the
number of users included dropped dramatically. For this reason,
we applied a requirement of a minimum of 30 tweets in both
in-treatment and unknown-treatment datasets to keep the balance
between the number of tweets and the number of users to be
included in the study. After applying this requirement, 187 users
were finally included in the study. The complete timelines of
these users were compiled, totaling 668,842 tweets, which were
reduced to 482,338 once retweets were removed. Out of these,
168,369 more tweets were excluded because they were posted
on dates located outside the periods that qualified a tweet for
being included in the in-treatment or the unknown-treatment
datasets. Finally, 57,525 tweets were included in the in-treatment
dataset and 256,444 in the unknown-treatment dataset.
Data Analysis
The two datasets of tweets, in-treatment and unknown-treatment,
were compared in order to determine the existence of behavioral
and linguistic differences between the tweets generated by the
users in each period. The features that were analyzed are listed
in Table 3.
Table 3. Features of the tweets analyzed.
Features
Analyses performed
Distribution over time
Tweets per hour, tweets during daytime vs night, tweets per day, tweets during weekdays vs weekend
Length
Number of characters, number of words
Part-of-speech (POS)
Number of words by grammatical categories (part-of-speech tags)
Emotion analysis
Frequencies of emotion types
Negations
Polarity
Frequencies of negation words
Polarity of tweets on the basis of Spanish Sentiment Lexicon
Paired data statistical significance tests (paired t tests) were
carried out whenever possible. The Benjamini-Hochberg false
discovery rate was applied for multiple testing correction
analysis [64]. The P values provided incorporate it.
The textual content of each tweet was analyzed using the same
methodology and tools used in our previous study [49]. The
textual content of each tweet was analyzed by means of the
following steps: tokenization performed based on a customized
Twitter tokenizer included in the Natural Language Toolkit
[65]; part-of-speech (POS) tagging performed by means of the
FreeLing Natural Language Processing tool in order to analyze
the usage patterns of grammatical categories, such as verbs,
nouns, pronouns, adverbs, and adjectives, in the text of tweets
[66]; identification of negations performed by building upon a
customized list of Spanish negation expressions, such as nada
(nothing), nadie (nobody), no (no), nunca (never), and similar;
identification of positive and negative words inside the text of
each tweet using the Spanish Sentiment Lexicon [67]; and
identification of words and expressions associated with emotions
such as happiness, anger, fear, disgust, surprise, and sadness
[68] by using the Spanish Emotion Lexicon [69].
The statistical analyses were carried out using Python 3.7, the
Tweepy, SciPy, and Natural Language Toolkit libraries, and R
version 3.6.2 (R Development Core Team), including the R
“psych” package 1.9.12.31. All the aforementioned software
tools are publicly available.
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 6
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
Ethical Approval
The protocol used in this study was reviewed and approved by
the Ethics Committee of Parc Salut Mar (approval number
2017/7234/1).
in-treatment periods; the percentage went down to 74.40% (SD
5.31) in unknown-treatment periods, with a mean percentage
difference of 1.56% (SD 8.9) that implies statistically significant
differences between the two periods (t186=2.39; P=.02).
Results
Distribution Over Time
Several types of distribution-over-time analysis were performed
in order to study the potential influence of being in in-treatment
periods in comparison to unknown-treatment ones. The tweet
hours were adjusted by the users’ time zone.
The mean duration of the time period analyzed of all the users
was 28.2 months (SD 24.7); the mean of the total number of
tweets analyzed was 307.6 (SD 336.0) for in-treatment periods
and 1371.4 (SD 748.2) in the case of unknown-treatment
periods. The mean number of tweets per day generated by users
during in-treatment periods was 11.44 (SD 10.05); this number
dropped to 9.07 (SD 7.21) in the unknown-treatment dataset
with a mean difference of 2.37 (SD 9.72) between periods,
which shows statistically significant differences between the
two datasets (t186=3.33; P<.001).
The mean percentage of tweets posted during daytime (between
8 AM and midnight) was 64.30% (SD 14.83) when the users
were in-treatment periods; this percentage fell to 61.78% (SD
13.69) during the unknown-treatment periods, with a mean
percentage difference of 2.52% (SD 11.81), which implies
statistically significant differences (t186=3.07; P=.004).
in
periods
(SD 6.70)
The mean number of tweets generated during the weekdays
(from Monday to Friday) was 12.28 (SD 11.05) during
in-treatment
the
and 9.33
unknown-treatment periods, with a mean difference of 2.95 (SD
10.23) and statistically significant differences between the
datasets (t186=3.93; P<.001). For the mean number of tweets
generated during the weekends (Saturday and Sunday), it was
9.35 (SD 9.31) in the in-treatment period and 8.41 (SD 9.82)
in the unknown-treatment period, with a mean difference of
0.94 (SD 10.92) that implies statistically significant differences
between the datasets (t186=1.18; P=.23). The mean percentage
of tweets posted on weekdays was 75.95% (SD 9.17) during
Length
The average number of characters per tweet was 88.03 (SD
30.74) and 85.19 (SD 28.82) in the in-treatment and
unknown-treatment datasets, respectively, with a mean
difference of 2.84 (SD 17.70) and statistically significant
differences between the periods (t186=2.19; P=.03). As for the
number of words per tweet, the mean was 15.68 (SD 5.75) in
the
the
unknown-treatment dataset, with a mean difference of 0.59 (SD
3.54) and statistically significant differences (t186=2.28; P=.02).
in-treatment dataset and 15.09 (SD 5.20)
in
Links and Mentions to Other Users
The mean percentages of tweets that include at least one link
were 23.10% (SD 16.16) and 23.27% (SD 15.29) in the
in-treatment and unknown-treatment datasets, respectively, with
a mean difference of −0.17 (SD 10.94), which is not statistically
significant (t186=−0.23; P=.82).The mean percentages of tweets
that include at least one mention of another Twitter user were
45.79% (SD 24.77) and 43.52% (SD 24.71) in the in-treatment
and unknown-treatment datasets, respectively, with a mean
difference of 2.27% (SD 12.13), which is statistically significant
(t186=2.56; P=.01).
Part-of-Speech
As for the analysis of the number of words by grammatical
category (ie, part-of-speech) in each tweet, we also compared
the in-treatment and unknown-treatment datasets. The mean
percentage of words per grammatical category over the total
number of words in each dataset is shown in Table 4. We
considered the most relevant lexical POS such as verbs, nouns,
pronouns, adverbs, and adjectives, excluding conjunctions,
interjections, punctuations, determiners, adpositions, numbers,
and dates.
Regarding the different types of pronouns, the mean percentages
of personal pronouns in each dataset are shown and compared
in Table 5.
Table 4. Percentages of part-of-speech words compared between in-treatment and unknown-treatment datasets.
POSa
Verbs
Nouns
Pronouns
Adverbs
Adjectives
in-treatment (%), mean
unknown-treatment (%), mean
Difference (%), mean (SD)
Paired t test
P value
18.50
19.50
9.19
6.42
6.05
18.20
19.94
8.93
6.36
6.21
0.3 (1.28)
−0.44 (2.57)
0.26 (1.33)
0.06 (0.84)
−0.16 (0.95)
3.15
−2.35
2.61
0.97
−2.34
.002
.02
.01
.34
.02
aPOS: part-of-speech.
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 7
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
Table 5. Mean percentages of personal pronouns compared between in-treatment and unknown-treatment datasets.
Personal pronouns
in-treatment (%), mean
unknown-treatment (%), mean
Difference (%), mean (SD)
Paired t test
P value
1st person singular
2nd person singular
3rd person singular
1st person plural
2nd person plural
3rd person plural
49.50
14.77
22.13
3.44
1.00
5.60
47.80
16.07
22.86
3.43
1.00
5.39
Emotion Analysis
The mean percentages of the different emotions, obtained using
the Spanish Sentiment Lexicon on the tweets posted in the two
periods, are shown in Table 6.
1.7 (8.68)
−1.3 (6.17)
−0.73 (5.79)
0.01 (3.43)
0 (1.22)
0.21 (3.68)
2.67
−2.88
−1.72
0.04
−0.01
0.77
.008
.004
.08
.96
.98
.44
Table 6. Mean percentages of different emotions compared between in-treatment and unknown-treatment datasets.
Emotion
in-treatment (%), mean
unknown-treatment (%), mean
Difference (%), mean (SD)
Paired t test
P value
Happiness
Sadness
Fear
Anger
Disgust
Surprise
26.93
10.01
3.20
5.52
3.11
5.59
25.94
9.76
3.02
5.20
3.06
5.06
Negation Analysis
The mean percentages of tweets, among all users, that included
one or more negation words were 27.66% (SD 10.54) and
26.59% (SD 9.87) for the in-treatment and unknown-treatment
datasets, respectively, with a mean difference of 1.07% (SD
6.99), which is statistically significant (t186=2.10; P=.04).
Polarity Analysis
As for the polarity of tweets, the percentage of tweets, among
all users, with one or more positive words inside the text was
15.13% (SD 6.56) in the in-treatment dataset and 14.50% (SD
5.43) in the unknown-treatment dataset, with a mean percentage
difference of 0.63% (SD 5.22; t186=1.66; P=.09). The percentage
of tweets with one or more negative words was 7.97% (SD 4.40)
in the in-treatment dataset and 7.54% (SD 3.52) in the
unknown-treatment dataset, with a mean percentage difference
of 0.43% (SD 3.58) (t186=1.64; P=.10). No statistically
significant differences were detected in this analysis.
Discussion
Principal Findings
Social media platforms in general, and Twitter in particular,
may provide useful information on how patients respond when
they receive a pharmacological treatment, as has been shown
in several studies in which social media has been used as a
complementary source of pharmacovigilance and monitoring
[34,70]. In this study, we analyzed the tweets of users who
mentioned they were taking antidepressant drugs, in particular
SSRIs, with the aim of detecting behavioral changes when they
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
0.99 (5.82)
0.25 (4.20)
0.18 (1.94)
0.32 (2.71)
0.05 (1.97)
0.53 (2.42)
2.32
0.81
1.23
1.62
0.38
2.98
.02
.41
.21
.11
.69
.003
are more likely to be in treatment in comparison to periods in
which they are less likely to be in treatment (“in-treatment” vs
“unknown-treatment” periods).
The results of this study show that, in general, Twitter users
significantly increased their activity of posting tweets during
the in-treatment periods. This increase was more pronounced
during weekdays than during weekends. We also observed a
significantly greater proportion of tweets posted during the
daytime during the in-treatment periods. These results are
consistent with the results of our previous paper [49], in which
we observed that the control group without signs of depression
showed more tweet posting activity than the group of users with
signs of depression, especially during the daytime and the
weekdays. These results are also consistent with another paper
that described the behavior in social media of people with
self-reported depression [41], as well as with a study on the
diurnal mood variation of patients suffering from major
depressive disorder [71]. In summary, we can state that when
considering tweet posting activity, the behavior of individuals
suffering from depression becomes more similar to that of the
general population when they are in treatment with SSRIs.
Likewise, the average number of characters and words per tweet
were significantly higher when the Twitter users were in
treatment with SSRIs, a finding that again points toward an
increase in the activity of these treated users. In addition, the
increase in the number of mentions per tweet can reflect a
greater interest in interacting with other people. All these
changes may be due to some improvement in their anhedonic
symptoms because of the medication.
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 8
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
the
and
in-treatment
changes between
Regarding the linguistic analysis, we observed quantitatively
slight
the
unknown-treatment periods, although in some cases they are
statistically significant. These slight findings are not easily
interpretable. In general, given that the style of writing of people
suffering from depression is characterized by self-focus
attention, which is associated with negative emotional states
and psychological distancing in order to connect with others
[72], we can conclude that when the studied subjects were in
treatment, they improved some traits related to their posting
activity as previously mentioned, but at the same time, their
language maintained the features of people suffering from
depression without a clear influence of the medication.
Emotion is another important aspect that characterizes people
suffering from depression, and it was consequently analyzed.
When the users were in treatment, they showed small but
statistically significant increases in the happiness and surprise
emotions, but not in sadness or other emotions (ie, anger, fear
and disgust). As for the number of negations, the users slightly
increased their use of these types of words during the
in-treatment period. However, the polarity analysis did not show
differences between the periods.
The increased activity observed on Twitter when the users were
likely to be in treatment with SSRIs can be linked to improved
emotional status in their happiness and surprise emotions. These
changes are consistent with our previous observations on mood
states of Twitter users without depression compared to those
with depression [49]. However, the traits that are related to
language, as indicated by the POS analysis and the use of
negations, maintained a similar profile to that of subjects with
depression, independently of the pharmacological treatment
detected. These results denote that users with depression who
are taking SSRIs show some mood improvements while
receiving antidepressant treatment, but at the same time maintain
an altered language pattern, which may be indicative of
incomplete recovery.
On the basis of our statistically significant results, we may state
that Twitter timelines can be used as a complementary tool to
monitor subjects in order to detect adherence to treatment, which
is an important problem in this kind of patient. Adherence to
treatment is essential for disease remission [73-76]. According
to some studies, it is common for patients suffering from
depression to not maintain the duration of antidepressant
treatment that is clinically recommended [4,18,77]. In summary,
the follow-up of behavioral and language changes in users’
Twitter timelines can be useful for monitoring the evolution of
depressive symptoms and the effect of treatments.
Limitations and Future Directions
This type of study in general, and this one in particular, presents
some limitations. For instance, we considered tweets written in
Spanish and from public Twitter users’ timelines, and these
users may be not representative of the general population or
people suffering from depression [33,49,78,79]. Some studies
have shown that Twitter users are often urban people with high
levels of education, and they are generally younger than the
general population [33,49,78,80,81]. We should also take into
account that SSRIs are used in different types of depressive
disorders and in other mental conditions. Moreover, we have
no information about whether these drugs were taken in the
context of a prescribed medical treatment or as a result of an
inappropriate self-medication decision.
Another limitation may be the fact that Twitter users who share
their personal drug intake may use words or expressions not
included in the list of drug names employed in this study for
streaming tweets, even though we tried to be exhaustive in the
list of names used. Twitter texts are informal and limited by the
number of characters, and they commonly include abbreviations,
errors, or slang language [33,45]. All these issues can make it
difficult to automatically extract drug mentions and link them
to a formal lexicon [28,30,50,53,55]. Unlike clinical records
that could be linked to domain resources, the lack of lay
vocabularies related to health concepts and terminologies hinders
the processing of social media texts [55]. In addition, the results
obtained may depend on the particular drugs selected for the
study [33], as well as on the periods of time set up for classifying
the tweets into the in-treatment and unknown-treatment datasets.
On the basis of the strategy applied for defining the groups of
tweets to be compared (tweets generated just after mentions to
SSRI intake vs tweets generated in periods far from any mention
to the SSRI intake), there is some chance of misclassification;
it is likely that not all the tweets in the first group were generated
by users under actual SSRI treatment, and it is probable that
some tweets of the second group have been generated by users
under SSRI treatment.
Furthermore, we must take into account that data from social
media posts contain irrelevant information. Although the
proportion of useful information for the specific research
purpose can be quite limited, it constitutes a useful starting point
[28,30,51,53]. In this scenario, the human curation of tweets is
a necessary step in this kind of analysis [34]. Even so, due to
the different nuances that a tweet can involve, it is not easy to
detect real drug intakes or firsthand experiences [24,46,52].
Conclusions
Social media can be used to monitor the health status of people
and, in particular, to detect symptoms or features related to
diseases or health conditions by means of analysis of the users’
behavior and language on social media platforms. Moreover,
the detection of changes in symptoms or other features when
patients are taking medications can provide interesting insights
for monitoring pharmacological treatments, as well as for
following up on the evolution of the disease, detecting side
effects, or providing information related to treatment adherence.
Changes in some features, such as a decrease in activity on
Twitter or of the length of tweets, an increase of self-focus
through the use of the first-person singular pronoun, and changes
in the happiness and surprise emotions could be used as
the
to detect
complementary
psychological status of users suffering from depression, as well
as to perceive lack of adherence to treatment. This information
may be especially useful in patients suffering from chronic
diseases who are receiving long-term treatments, as is the case
for mental disorders in general and depression in particular.
However, it is not possible to determine the specific reasons
why individuals change their behavior and language on social
the worsening of
tools
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 9
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
media platforms in the framework of a disease and its treatment
without performing a clinical assessment. Overall, this study
shows the relevance of monitoring behavioral and linguistic
changes in the tweets of persons taking antidepressants. These
changes are likely to be influenced by the diverse stages of the
disease and the therapeutic effects of the treatment that these
Twitter users are receiving, opening a new line of research to
better understand and follow up on depression through social
media.
Acknowledgments
We received support from the Agency for Management of University and Research Grants in Catalonia (Spain) for the incorporation
of new research personnel (FI2016) and from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement
number 802750 (FAIRplus) with the support of the European Union’s Horizon 2020 research and innovation programme and
European Federation of Pharmaceutical Industries and Associations Companies.
The Research Programme on Biomedical Informatics is a member of the Spanish National Bioinformatics Institute, funded by
Instituto de Salud Carlos III and the European Regional Development Fund (PRB2-ISCIII), and it is supported by grant
PT17/0009/0014. The Department of Experimental and Health Sciences, Universitat Pompeu Fabra, is a “Unidad de Excelencia
María de Maeztu”, funded by the Ministry of Economy, Spain [MDM-2014-0370]. Funding for the open access charge is from
the Agència de Gestió d’Ajuts Universitaris i de Recerca Generalitat de Catalunya (2017 SGR 00519).
The Database for Pharmacoepidemiological Research in Primary Care, from the Spanish Agency for Medicines and Health
Products of the Ministry of Health and Consumer Affairs and Social Welfare of the Government of Spain, was used to obtain
useful information about the prescription frequency of antidepressants.
Conflicts of Interest
None declared.
References
1. World Health Organization. Depression: Key Facts. 2019. URL: https://www.who.int/news-room/fact-sheets/detail/depression
[accessed 2020-01-09]
2. World Health Organization. Depression and Other Common Mental Disorders: Global Health Estimates. 2017. URL: http:/
3.
4.
/apps.who.int/iris/bitstream/10665/254610/1/WHO-MSD-MER-2017.2-eng.pdf?ua=1 [accessed 2020-01-09]
Ferrari AJ, Charlson FJ, Norman RE, Patten SB, Freedman G, Murray CJL, et al. Burden of depressive disorders by country,
sex, age, and year: findings from the global burden of disease study 2010. PLoS Med 2013 Nov;10(11):e1001547 [FREE
Full text] [doi: 10.1371/journal.pmed.1001547] [Medline: 24223526]
Royal College of Psychiatrists. Position statement on antidepressants and depression. PS04/19. London, United Kingdom:
Royal College of Psychiatrists; May 2019.
6.
5. Marcus M, Yasamy MT, van Ommeren M, Chisholm D, Saxena S, WHO Department of Mental Health and Substance
Abuse. Depression: a global public health concern.: World Health Organization; 2012. URL: http://www.who.int/
mental_health/management/depression/who_paper_depression_wfmh_2012.pdf [accessed 2020-11-12]
Global BODS2C. Global, regional, and national incidence, prevalence, and years lived with disability for 301 acute and
chronic diseases and injuries in 188 countries, 1990-2013: a systematic analysis for the global burden of disease study 2013.
Lancet Aug 22 2015;386(9995):743-800. [Medline: 26063472]
Trautmann S, Rehm J, Wittchen H. The economic costs of mental disorders: Do our societies react appropriately to the
burden of mental disorders? EMBO Rep 2016 Sep;17(9):1245-1249 [FREE Full text] [doi: 10.15252/embr.201642951]
[Medline: 27491723]
Patel V, Chisholm D, Parikh R, Charlson F, Degenhardt L, Dua T. Global priorities for addressing the burden of mental,
neurological, substance use disorders. In: Patel V, Chisholm D, Dua T, Laxminarayan R, Medina-Mora ME, Vos T. eds.
Disease Control Priorities: Mental, Neurological, and Substance Use Disorders, 3rd ed. (Vol 4). Washington, DC: The
World Bank; 2016:1-27.
8.
7.
9. Whiteford HA, Degenhardt L, Rehm J, Baxter AJ, Ferrari AJ, Erskine HE, et al. Global burden of disease attributable to
mental and substance use disorders: findings from the Global Burden of Disease Study 2010. Lancet 2013 Nov
9;382(9904):1575-1586. [doi: 10.1016/S0140-6736(13)61611-6] [Medline: 23993280]
10. Wongkoblap A, Vadillo MA, Curcin V. Researching mental health disorders in the era of social media: Systematic review.
J Med Internet Res 2017 Dec 29;19(6):e228 [FREE Full text] [doi: 10.2196/jmir.7215] [Medline: 28663166]
11. Working Group of the Clinical Practice Guideline on the Management of Depression in adults. Ministry of Health, Social
Services and Equality. Galician Agency for Health Technology Assessment (avalia-t). Clinical practice guideline on the
management of depression in adults. 2014. URL: https://portal.guiasalud.es/wp-content/uploads/2018/12/
GPC_534_Depresion_Adulto_Avaliat_compl_en.pdf [accessed 2020-11-12]
12. Mathers CD, Loncar D. Projections of global mortality and burden of disease from 2002 to 2030. PLoS Med 2006
Nov;3(11):e442 [FREE Full text] [doi: 10.1371/journal.pmed.0030442] [Medline: 17132052]
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 10
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
13. Vigo D, Thornicroft G, Atun R. Estimating the true global burden of mental illness. Lancet Psychiatry 2016 Feb;3(2):171-178.
[doi: 10.1016/S2215-0366(15)00505-2] [Medline: 26851330]
14. Kraus C, Kadriu B, Lanzenberger R, Zarate CA, Kasper S. Prognosis and improved outcomes in major depression: a review.
Transl Psychiatry 2019 Apr 03;9(1):127 [FREE Full text] [doi: 10.1038/s41398-019-0460-3] [Medline: 30944309]
15. Ministerio de Sanidad, Servicios Sociales e Igualdad. Agencia Española de Medicamentos y Productos Sanitarios (AEMPS).
Utilización de medicamentos antidepresivos en España durante el periodo 2000-2013. Informe de utilización de medicamentos
U/AD/V1/14012015. Madrid: Ministerio de Sanidad, Servicios Sociales e Igualdad. Agencia Española de Medicamentos
y Productos Sanitarios (AEMPS); 2015.
16. Luo Y, Kataoka Y, Ostinelli EG, Cipriani A, Furukawa TA. National prescription patterns of antidepressants in the treatment
of adults with major depression in the US between 1996 and 2015: A population representative survey based analysis. Front
Psychiatry 2020;11:35 [FREE Full text] [doi: 10.3389/fpsyt.2020.00035] [Medline: 32116850]
Fasipe OJ. The emergence of new antidepressants for clinical use: Agomelatine paradox versus other novel agents. IBRO
Rep 2019 Jun;9(6):95-110 [FREE Full text] [doi: 10.1016/j.ibror.2019.01.001] [Medline: 31211282]
17.
18. Depression in adults: recognition and management. Clinical guideline CG90. National Institute for Health and Care
Excellence (NICE); 2009 Oct 28. URL: https://www.nice.org.uk/guidance/cg90 [accessed 2020-11-12]
19. Gelenberg A, Freeman M, Markowitz J, Rosenbaum J, Thase M, Trivedi M, et al. Practice guideline for the treatment of
patients with major depressive disorder. 3rd ed. Washington, DC: American Psychiatric Association; 2010.
20. Lane R, Baldwin D, Preskorn S. The SSRIs: advantages, disadvantages and differences. J Psychopharmacol 1995 Jan;9(2
21.
Suppl):163-178. [doi: 10.1177/0269881195009002011] [Medline: 22297235]
Pundiak TM, Case BG, Peselow ED, Mulcare L. Discontinuation of maintenance selective serotonin reuptake inhibitor
monotherapy after 5 years of stable response: a naturalistic study. J Clin Psychiatry 2008 Nov;69(11):1811-1817. [doi:
10.4088/jcp.v69n1117] [Medline: 19026252]
22. Emslie GJ, Mayes TL, Ruberu M. Continuation and maintenance therapy of early-onset major depressive disorder. Paediatr
Drugs 2005;7(4):203-217. [doi: 10.2165/00148581-200507040-00001] [Medline: 16117558]
23. Garnock-Jones KP, McCormack PL. Escitalopram: a review of its use in the management of major depressive disorder in
adults. CNS Drugs 2010 Sep;24(9):769-796. [doi: 10.2165/11204760-000000000-00000] [Medline: 20806989]
24. Clevenger SS, Malhotra D, Dang J, Vanle B, IsHak WW. The role of selective serotonin reuptake inhibitors in preventing
relapse of major depressive disorder. Ther Adv Psychopharmacol 2018 Jan;8(1):49-58 [FREE Full text] [doi:
10.1177/2045125317737264] [Medline: 29344343]
Sim K, Lau WK, Sim J, Sum MY, Baldessarini RJ. Prevention of relapse and recurrence in adults with major depressive
disorder: Systematic review and meta-analyses of controlled trials. Int J Neuropsychopharmacol 2015 Jul 07;19(2):pyv076
[FREE Full text] [doi: 10.1093/ijnp/pyv076] [Medline: 26152228]
Peselow ED, Tobia G, Karamians R, Pizano D, IsHak WW. Prophylactic efficacy of fluoxetine, escitalopram, sertraline,
paroxetine, and concomitant psychotherapy in major depressive disorder: outcome after long-term follow-up. Psychiatry
Res 2015 Feb 28;225(4):680-686. [doi: 10.1016/j.psychres.2014.11.022] [Medline: 25496869]
Salathé M. Digital pharmacovigilance and disease surveillance: Combining traditional and Big-Data systems for better
public health. J Infect Dis 2016 Dec 01;214(suppl_4):S399-S403 [FREE Full text] [doi: 10.1093/infdis/jiw281] [Medline:
28830106]
Sarker A, Ginn R, Nikfarjam A, O'Connor K, Smith K, Jayaraman S, et al. Utilizing social media data for pharmacovigilance:
A review. J Biomed Inform 2015 Apr;54:202-212 [FREE Full text] [doi: 10.1016/j.jbi.2015.02.004] [Medline: 25720841]
25.
26.
27.
28.
29. Adrover C, Bodnar T, Huang Z, Telenti A, Salathé M. Identifying adverse effects of HIV drug treatment and associated
sentiments using Twitter. JMIR Public Health Surveill 2015;1(2):e7 [FREE Full text] [doi: 10.2196/publichealth.4488]
[Medline: 27227141]
30. Nikfarjam A, Sarker A, O'Connor K, Ginn R, Gonzalez G. Pharmacovigilance from social media: mining adverse drug
reaction mentions using sequence labeling with word embedding cluster features. J Am Med Inform Assoc 2015
May;22(3):671-681 [FREE Full text] [doi: 10.1093/jamia/ocu041] [Medline: 25755127]
Salathé M. Digital epidemiology: what is it, and where is it going? Life Sci Soc Policy 2018 Jan 04;14(1):1 [FREE Full
text] [doi: 10.1186/s40504-017-0065-7] [Medline: 29302758]
31.
32. Eysenbach G. Infodemiology and infoveillance: framework for an emerging set of public health informatics methods to
analyze search, communication and publication behavior on the Internet. J Med Internet Res 2009 Mar 27;11(1):e11 [FREE
Full text] [doi: 10.2196/jmir.1157] [Medline: 19329408]
33. Lardon J, Bellet F, Aboukhamis R, Asfari H, Souvignet J, Jaulent M, et al. Evaluating Twitter as a complementary data
source for pharmacovigilance. Expert Opin Drug Saf 2018 Aug;17(8):763-774. [doi: 10.1080/14740338.2018.1499724]
[Medline: 29991282]
Freifeld CC, Brownstein JS, Menone CM, Bao W, Filice R, Kass-Hout T, et al. Digital drug safety surveillance: monitoring
pharmaceutical products in twitter. Drug Saf 2014 May;37(5):343-350 [FREE Full text] [doi: 10.1007/s40264-014-0155-x]
[Medline: 24777653]
34.
35. Crannell WC, Clark E, Jones C, James TA, Moore J. A pattern-matched Twitter analysis of US cancer-patient sentiments.
J Surg Res 2016 Dec;206(2):536-542. [doi: 10.1016/j.jss.2016.06.050] [Medline: 27523257]
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 11
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
36.
Surian D, Nguyen DQ, Kennedy G, Johnson M, Coiera E, Dunn AG. Characterizing Twitter discussions about HPV vaccines
using topic modeling and community detection. J Med Internet Res 2016;18(8):e232 [FREE Full text] [doi: 10.2196/jmir.6045]
[Medline: 27573910]
37. Chen E, Lerman K, Ferrara E. Tracking Social Media Discourse About the COVID-19 Pandemic: Development of a Public
Coronavirus Twitter Data Set. JMIR Public Health Surveill 2020 May 29;6(2):e19273 [FREE Full text] [doi: 10.2196/19273]
[Medline: 32427106]
38. Reece AG, Reagan AJ, Lix KLM, Dodds PS, Danforth CM, Langer EJ. Forecasting the onset and course of mental illness
with Twitter data. Sci Rep 2017 Oct 11;7(1):13006 [FREE Full text] [doi: 10.1038/s41598-017-12961-9] [Medline:
29021528]
Park M, Cha C, Cha M. Depressive moods of users portrayed in Twitter. In: Proceedings of the ACM SIGKDD Workshop
on Health Informatics. 2012 Presented at: HI-KDD'12; August 12-16; Beijing, China p. 1-8.
39.
40. Conway M, O'Connor D. Social Media, Big Data, and Mental Health: Current Advances and Ethical Implications. Curr
Opin Psychol 2016 Jun;9:77-82 [FREE Full text] [doi: 10.1016/j.copsyc.2016.01.004] [Medline: 27042689]
41. De Choudhury C, Gamon M, Counts S, Horvitz E. Predicting depression via social media. In: Proceedings of the Seventh
International Conference on Weblogs and Social Media. 2013 Presented at: AAA'13; July 8-11; Cambridge, MA p. 128-138.
42. Cavazos-Rehg P, Krauss M, Sowles S, Connolly S, Rosas C, Bharadwaj M, et al. A content analysis of depression-related
tweets. Comput Human Behav 2016 Jan 01;54:351-357 [FREE Full text] [doi: 10.1016/j.chb.2015.08.023] [Medline:
26392678]
43. Nguyen T, O’Dea B, Larsen M, Phung D, Venkatesh S, Christensen H. Using linguistic and topic analysis to classify
sub-groups of online depression communities. Multimed Tools Appl 2015 Dec 21;76(8):10653-10676. [doi:
10.1007/s11042-015-3128-x]
Statista. Number of Monthly Active Twitter Users Worldwide From 1st Quarter 2010 to 2nd Quarter 2019 (in Millions).
2019. URL: https://www.statista.com/statistics/282087/number-of-monthly-active-twitter-users/ [accessed 2020-02-03]
45. Audeh B, Calvier F, Bellet F, Beyens M, Pariente A, Lillo-Le Louet A, et al. Pharmacology and social media: Potentials
44.
and biases of web forums for drug mention analysis-case study of France. Health Informatics J 2019 Sep 30;26(2):1253-1272.
[doi: 10.1177/1460458219865128] [Medline: 31566468]
46. Alvaro N, Conway M, Doan S, Lofi C, Overington J, Collier N. Crowdsourcing Twitter annotations to identify first-hand
experiences of prescription drug use. J Biomed Inform 2015 Dec;58:280-287 [FREE Full text] [doi: 10.1016/j.jbi.2015.11.004]
[Medline: 26556646]
Pierce CE, Bouri K, Pamer C, Proestel S, Rodriguez HW, Van LH, et al. Evaluation of Facebook and Twitter monitoring
to detect safety signals for medical products: An analysis of recent FDA safety alerts. Drug Saf 2017 Apr;40(4):317-331
[FREE Full text] [doi: 10.1007/s40264-016-0491-0] [Medline: 28044249]
47.
48. Bian J, Topaloglu U, Yu F. Towards Large-scale Twitter mining for drug-related adverse events. SHB12 2012 Oct
29;2012:25-32 [FREE Full text] [doi: 10.1145/2389707.2389713] [Medline: 28967001]
49. Leis A, Ronzano F, Mayer MA, Furlong LI, Sanz F. Detecting signs of depression in tweets in Spanish: Behavioral and
linguistic analysis. J Med Internet Res 2019 Jun 27;21(6):e14199 [FREE Full text] [doi: 10.2196/14199] [Medline: 31250832]
50. O'Connor K, Pimpalkhute P, Nikfarjam A, Ginn R, Smith KL, Gonzalez G. Pharmacovigilance on twitter? Mining tweets
for adverse drug reactions. AMIA Annu Symp Proc 2014;2014:924-393 [FREE Full text] [Medline: 25954400]
51. Mahata D, Friedrichs J, Shah RR, Jiang J. Detecting personal intake of medicine from Twitter. IEEE Intell Syst 2018
Jul;33(4):87-95. [doi: 10.1109/mis.2018.043741326]
52. Klein A, Sarker A, Rouhizadeh M, O'Connor K, Gonzalez G. Detecting personal medication intake in Twitter: an annotated
corpusbaseline classification system. 2017 Presented at: the 16th Biomedical Natural Language Processing (BioNLP); Aug
4; Vancouver, BC, Canada p. 1-11 URL: https://www.aclweb.org/anthology/W17-2316.pdf
53. Kiritchenko S, Mohammad S, Morin J, de BB. NRC-Canada at SMM4H shared task: classifying tweets mentioning adverse
drug reactions and medication intake. 2017 Presented at: In Proceedings of 2nd Social Media Mining for Health Applications
Workshop co-located with the American Medical Informatics Association Annual Symposium (AMIA); Nov 4, 2017;
Washington DC p. 1-11 URL: http://ceur-ws.org/Vol-1996/paper1.pdf
55.
54. Nikfarjam A, Ransohoff JD, Callahan A, Jones E, Loew B, Kwong BY, et al. Early detection of adverse drug reactions in
social health networks: A natural language processing pipeline for dignal detection. JMIR Public Health Surveill 2019 Jun
03;5(2):e11264 [FREE Full text] [doi: 10.2196/11264] [Medline: 31162134]
Segura-Bedmar I, Martínez P, Revert R, Moreno-Schneider J. Exploring Spanish health social media for detecting drug
effects. BMC Med Inform Decis Mak 2015;15 Suppl 2:S6 [FREE Full text] [doi: 10.1186/1472-6947-15-S2-S6] [Medline:
26100267]
Statista. Leading Countries Based on Number of Twitter Users as of April 2020 (in Millions). 2020. URL: https://www.
statista.com/statistics/242606/number-of-active-twitter-users-in-selected-countries/ [accessed 2020-07-05]
56.
57. Liu J, Weitzman E, Chunara R. Assessing behavioral stages from social media data. In: Proceedings of the 2017 ACM
Conference on Computer Supported Cooperative Work and Social Computing. 2017 Presented at: CSCW 2017; February
25-March 1, 2017; Portland, Oregon p. 1320-1333.
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 12
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
58.
Saha K, Sugar B, Torous J, Abrahao B, Kiciman E, De Choudhury M. A social media study on the effects of psychiatric
medication use. In: Proceedings of the thirteenth International Conference on Web and Social Media. 2019 Presented at:
ICWSM 2019; 2019; Munich, Germany p. 440-451.
59. Twitter Developer. URL: https://developer.twitter.com/en.html [accessed 2019-12-16]
60. Wishart DS, Knox C, Guo AC, Cheng D, Shrivastava S, Tzur D, et al. DrugBank: a knowledgebase for drugs, drug actions
and drug targets. Nucleic Acids Res 2008;36(Database issue):D901-D906 [FREE Full text] [doi: 10.1093/nar/gkm958]
[Medline: 18048412]
61. WHO Collaborating Centre for Drug Statistics Methodology, Norwegian Institute of Public Health. The Anatomical
Therapeutics Chemical Classification System (ATC). Last updated December 16. 2019 Jan 16. URL: https://www.whocc.no/
[accessed 2019-05-06]
Selective serotonin reuptake inhibitor.: Wikipedia; 2019 May 07. URL: https://en.wikipedia.org/wiki/
Selective_serotonin_reuptake_inhibitor [accessed 2019-05-07]
Spanish Agency for Medicines and Health Products (AEMPS), Ministry of Health and Consumer Affairs and Social Welfare.
Multiregional primary care database of Spanish population (BIFAP). 2019. URL: http://www.bifap.org/ [accessed 2019-05-07]
62.
63.
64. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple hypothesis
testing. J R Stat Soc B 1995;57(1):289-300. [doi: 10.1111/j.2517-6161.1995.tb02031.x]
65. Natural Language Tool Kit. URL: https://www.nltk.org/api/nltk.tokenize.html [accessed 2019-12-10]
66.
Padró L, Stanilovsky E. FreeLing 3.0: Towards wider multilinguality. In: Proceedings of the Eighth International Conference
on Language Resources and Evaluation. 2012 Presented at: LREC'12; May 21-27; Istanbul, Turkey p. 2473-2479 URL:
http://www.lrec-conf.org/proceedings/lrec2012/pdf/430_Paper.pdf
Perez-Rosas V, Banea C, Mihalcea R. Learning sentiment lexicons in Spanish. In: Proceedings of the Eighth International
Conference on Language Resources and Evaluation. 2012 Presented at: LREC'12; May 21-27; Istanbul, Turkey p. 3077-3081
URL: http://www.lrec-conf.org/proceedings/lrec2012/pdf/1081_Paper.pdf
67.
68. Ekman P, Friesen WV, O'Sullivan M, Chan A, et al. Universals and cultural differences in the judgments of facial expressions
of emotion. J Pers Soc Psychol 1987 Oct;53(4):712-717. [doi: 10.1037//0022-3514.53.4.712] [Medline: 3681648]
Sidorov G, Miranda-Jiménez S, Viveros-Jiménez F, Gelbukh A, Castro-Sánchez N, Castillo F. Empirical study of opinion
mining in Spanish tweets. In: Proceedings of the 11th Mexican International Conference on Artificial Intelligence. 2012
Presented at: MICAI'12; October 27-November 4, 2012; San Luis Potosí, Mexico p. 1-4. [doi: 10.1007/978-3-642-37798-3_4]
69.
70. Carbonell P, Mayer MA, Bravo A. Exploring brand-name drug mentions on Twitter for pharmacovigilance. Stud Health
Technol Inform 2015;210:55-59. [Medline: 25991101]
71. Morris DW, Rush AJ, Jain S, Fava M, Wisniewski SR, Balasubramani GK, et al. Diurnal mood variation in outpatients
with major depressive disorder: Implications for DSM-V from an analysis of the sequenced treatment alternatives to relieve
depression study data. J Clin Psychiatry 2007 Sep;68(9):1339-1347. [Medline: 17915971]
Pennebaker JW, Mehl MR, Niederhoffer KG. Psychological aspects of natural language use: our words, our selves. Annu
Rev Psychol 2003;54:547-577. [doi: 10.1146/annurev.psych.54.101601.145041] [Medline: 12185209]
72.
73. Anghelescu IG, Kohnen R, Szegedi A, Klement S, Kieser M. Comparison of Hypericum extract WS 5570 and paroxetine
in ongoing treatment after recovery from an episode of moderate to severe depression: results from a randomized multicenter
study. Pharmacopsychiatry 2006 Nov;39(6):213-219. [doi: 10.1055/s-2006-951388] [Medline: 17124643]
Sansone RA, Sansone LA. Antidepressant adherence: are patients taking their medications? Innov Clin Neurosci 2012
May;9(5-6):41-46 [FREE Full text] [Medline: 22808448]
74.
75. Mitchell AJ. Depressed patients and treatment adherence. Lancet 2006 Jun 24;367(9528):2041-2043. [doi:
10.1016/S0140-6736(06)68902-2] [Medline: 16798371]
76. De las Cuevas C, Peñate W, Sanz EJ. Risk factors for non-adherence to antidepressant treatment in patients with mood
disorders. Eur J Clin Pharmacol 2014 Jan;70(1):89-98. [doi: 10.1007/s00228-013-1582-9] [Medline: 24013851]
77. De Choudhury M, De S. Mental health discourse on Reddit: Self-disclosure, social support, and anonymity. In: Proceedings
of the 8th International Conference on Weblogs and Social Media. 2014 Presented at: AAAI'14; June 1-4; Ann Arbor,
Michigan p. 71-80 URL: https://www.aaai.org/ocs/index.php/ICWSM/ICWSM14/paper/view/8075/8107
Sadah SA, Shahbazi M, Wiley MT, Hristidis V. A study of the demographics of web-based health-related social media
users. J Med Internet Res 2015 Aug 06;17(8):e194 [FREE Full text] [doi: 10.2196/jmir.4308] [Medline: 26250986]
78.
79. Eichstaedt JC, Schwartz HA, Kern ML, Park G, Labarthe DR, Merchant RM, et al. Psychological language on Twitter
predicts county-level heart disease mortality. Psychol Sci 2015 Feb;26(2):159-169 [FREE Full text] [doi:
10.1177/0956797614557867] [Medline: 25605707]
80. Yom-Tov E, Johansson-Cox I, Lampos V, Hayward AC. Estimating the secondary attack rate and serial interval of
influenza-like illnesses using social media. Influenza Other Respir Viruses 2015 Jul;9(4):191-199 [FREE Full text] [doi:
10.1111/irv.12321] [Medline: 25962320]
Paul MJ, Dredze M. Discovering health topics in social media using topic models. PLoS One 2014;9(8):e103408 [FREE
Full text] [doi: 10.1371/journal.pone.0103408] [Medline: 25084530]
81.
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 13
(page number not for citation purposes)
JOURNAL OF MEDICAL INTERNET RESEARCH
Leis et al
Abbreviations
POS: part-of-speech
SSRIs: selective serotonin reuptake inhibitors
Edited by G Eysenbach, R Kukafka; submitted 01.06.20; peer-reviewed by F Lopez Segui, E Yom-Tov; comments to author 22.06.20;
revised version received 01.09.20; accepted 12.11.20; published 18.12.20
Please cite as:
Leis A, Ronzano F, Mayer MA, Furlong LI, Sanz F
Evaluating Behavioral and Linguistic Changes During Drug Treatment for Depression Using Tweets in Spanish: Pairwise Comparison
Study
J Med Internet Res 2020;22(12):e20920
URL: http://www.jmir.org/2020/12/e20920/
doi: 10.2196/20920
PMID: 33337338
©Angela Leis, Francesco Ronzano, Miguel Angel Mayer, Laura I Furlong, Ferran Sanz. Originally published in the Journal of
Medical Internet Research (http://www.jmir.org), 18.12.2020. This is an open-access article distributed under the terms of the
Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution,
and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is
properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this
copyright and license information must be included.
http://www.jmir.org/2020/12/e20920/
XSL•FO
RenderX
J Med Internet Res 2020 | vol. 22 | iss. 12 | e20920 | p. 14
(page number not for citation purposes)
| null |
10.1371_journal.pone.0287008.pdf
|
Data Availability Statement: The data underlying
the results presented in the study are available
from Table 1.
|
The data underlying the results presented in the study are available from Table 1 .
|
RESEARCH ARTICLE
Sex differences in hepatitis A incidence rates–
a multi-year pooled-analysis based on
national data from nine high-income
countries
Manfred S. GreenID*, Naama Schwartz, Victoria PeerID
School of Public Health, University of Haifa, Haifa, Israel
* [email protected]
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
Abstract
Background
OPEN ACCESS
Citation: Green MS, Schwartz N, Peer V (2023)
Sex differences in hepatitis A incidence rates–a
multi-year pooled-analysis based on national data
from nine high-income countries. PLoS ONE 18(6):
e0287008. https://doi.org/10.1371/journal.
pone.0287008
Editor: Inge Roggen, Universitair Kinderziekenhuis
Koningin Fabiola: Hopital Universitaire des Enfants
Reine Fabiola, BELGIUM
Received: April 30, 2022
Accepted: May 28, 2023
Published: June 13, 2023
Copyright: © 2023 Green et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The data underlying
the results presented in the study are available
from Table 1.
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist.
Possible sex differences in hepatitis A virus (HAV) incidence rates in different age groups
are not well documented. We aimed to obtain stable pooled estimates of such differences
based on data from a number of high-income countries.
Methods
We obtained data on incident cases of HAV by sex and age group over a period of 6–25
years from nine countries: Australia, Canada, Czech Republic, Finland, Germany, Israel,
Netherland, New Zealand and Spain. Male to female incidence rate ratios (IRR) were com-
puted for each year, by country and age group. For each age group, we used meta-analytic
methods to combine the IRRs. Meta-regression was conducted to estimate the effects of
age, country, and time period on the IRR.
Results
A male excess in incidence rates was consistently observed in all age groups, although in
the youngest and oldest age groups, where the numbers tended to be lower, the lower
bounds of the 95% confidence intervals for the IRRs were less than one. In the age groups
<1, 1–4, 5–9, 10–14, 15–44, 45–64 and 65+, the pooled IRRs (with 95% CI) over countries
and time periods were 1.18 (0.94,1.48), 1.22 (1.16,1.29), 1.07 (1.03,1.11), 1.09 (1.04,1.14),
1.46 (1.30,1.64), 1.32 (1.15,1.51) and 1.10 (0.99,1.23) respectively.
Conclusions
The excess HAV incidence rates in young males, pooled over a number of countries, sug-
gest that the sex differences are likely to be due at least in part to physiological and biologi-
cal differences and not just behavioral factors. At older ages, differential exposure plays an
important role. These findings, seen in the context of the excess incidence rates in young
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
1 / 18
PLOS ONE Male excess in hepatitis A incidence rates
males for many other infectious diseases, can provide further keys to the mechanisms of
the infection.
Introduction
There is an expanding literature on sex differences in the incidence rates of various infectious
diseases [1–3]. The type and extent of the differences frequently vary by disease and age group.
The mechanisms underlying these differences have not been fully elucidated and cannot be
explained entirely by differences in exposure. The pattern of male to female ratios in the inci-
dence rates of different infectious diseases can make an important contribution to understand-
ing the underlying mechanisms of the diseases.
Despite the availability of an effective vaccine, hepatitis A virus (HAV) infection remains a
common disease, particularly in low-income countries with overcrowding and poor sanitation,
where the incidence rates of the disease are particularly high in infancy and childhood [4, 5].
In countries with high hepatitis A vaccine coverage, the incidence of cases and outbreaks have
decreased in children and the infection has shifted significantly to other risk groups, such as
men who have sex with men (MSM) [6–9].
There are reports in the literature on sex differences in the incidence rates of hepatitis A,
but they are inconsistent and poorly documented by age group [10–13]. While some report
higher incidence rates of viral hepatitis A in males [6, 10], there are inconsistencies. For exam-
ple, in a report from Germany, during 2018–2020, no sex differences were observed in the inci-
dence of the disease [11]. One report from South Korea found a change in the sex differences,
possibly due to increased immunization in the military [13].
In this study, we aimed to obtain pooled estimates of the age-specific male to female ratios
in the incidence rates of HAV infection based on data from a number of developed countries
over extended time periods.
Methods
Source of data
National surveillance data on reported cases of HAV infection, by age, sex and year, were
obtained from relevant government institutions for nine countries from Czech Republic, Fin-
land, Germany, Netherland, Spain, Australia, New Zealand, Canada and Israel. The data for
Australia, for years 2001–2016, was extracted from the National Notifiable Diseases Surveil-
lance System (NNDSS), [14] for Canada for the years 1991–2015, from the Public Health
Agency of Canada (PHAC) [15], for the Czech Republic, for 2008–2013, from the Institute of
Health Information and Statistics [16], for Finland, for years1995-2016 from the National
Institute for Health and Welfare (THL) [17], for Germany for the years 2001–2016, from the
German Federal Health Monitoring System [18], for Israel from the Department of Epidemiol-
ogy in the Ministry of Health for years 1998–2016, for the Netherland (2003–2017), directly
from the official representative of RIVM, for New Zealand for years 1997–2015 from the Insti-
tute of Environmental Science and Research (ESR) [19] and for Spain from the Spanish Epide-
miological Surveillance for years 2005–2015 [20].
Information about the population size by age, sex and year was obtained for Australia from
ABS.Stat [21] (Australia’s Bureau of statistics), for Canada from Statistics Canada CANSIM
database [22], for the Czech Republic from the Czech Statistical Office [23], for Finland from
Statistics Finland’s PX-Web databases [24], for Germany from the German Federal Health
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
2 / 18
PLOS ONE Male excess in hepatitis A incidence rates
Monitoring System [25], for Israel from the Central Bureau of Statistics [26], for Netherland
from Netherlands’ database (StatLine) [27], for New Zealand from Statistics New Zealand [28]
and for the Spain from the Department of Economic and Social Affairs, Population Division
[29].
Ethics and informed consent
National, open access, sex-and-age disaggregated, anonymous data were used and there was
no need for ethics committee approval.
Statistical analyses
Data analysis. HAV incidence rates (IR) per 100,000 were calculated by age group and
sex, for each country and calendar year using the number of reported cases divided by the
respective population size and multiplied by 100,000. The age groups considered were <1
years (infants), 1–4 (early childhood), 5–9 (late childhood), 10–14 (puberty), 15–44 (young
adulthood), 45–64 (middle adulthood) and 65+ (senior adulthood). Surveillance systems in
Canada and New Zealand used similar age-groups except for 15–39, 40–59 and 60+. For Aus-
tralia, data for infants and age 1–4, disaggregated by sex and age, are missing. The male to
female incidence rate ratio (IRR) was calculated by dividing the incidence rate in males by that
of females, by age group, country and time period.
Pooled analysis. As in previous studies of sex differences in infectious diseases [1–3], we
used meta-analytic methods to establish the magnitude of the pooled sex differences in the
incidence of HAV infection, by age group, across different countries and over a number of
years. The outcome variable was the male to female IRR. For each age group, the IRRs for each
country were pooled over time periods and then the pooled IRRs for each country were com-
bined. Forest plots with the pooled IRRs, over countries and years of reporting, were prepared
separately for the seven age groups. Heterogeneity was evaluated using the Q statistic and I2
was calculated as an estimate of the percentage of between-study variance. If the p-value for
the Q statistic was less than 0.05, or I2 exceeded 50%, the random effects models was used to
estimate pooled IRRs and 95% confidence intervals (CI). Otherwise, the fixed effects model
was considered, although due to the low power of the Q statistic, the more conservative ran-
dom effects model was preferred. In order to explore the contribution of countries and the
reported years to the variability in the IRRs, meta-regression analyses were performed. To eval-
uate the effect of individual countries and years on the male to female incidence risk ratio, we
performed leave-one-out sensitivity analysis and recomputed the pooled IRRs. The meta-ana-
lytic methods and meta-regressions were carried out using STATA software version 12.1
(Stata Corp., College Station, TX).
Results
Descriptive statistics
The summary of the male to female IRRs per 100,000 populations in different countries for
each age group is presented in Table 1.
Significant differences in incidence rates were observed between the countries, with the high-
est incidence rates in all ages and both sexes in Czech Republic. Higher incidence rates were
observed in Israel and Spain up to age 44 and in Germany in the group of adults (age 45–64).
Forest plots. The forest plots for the IRRs by age group, are shown in Figs 1–7.
The forest plot for infants is shown in Fig 1. The pooled male to female IRR was 1.18 (95%
CI 0.94–1.48) with I2 = 0.0% and varied between 0.86 in Canada and 2.26 in Spain.
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
3 / 18
PLOS ONE Table 1. Details of the countries included in the study, by sex and age group—descriptive data.
Male excess in hepatitis A incidence rates
Age
<1
1–4
5–9
10–14
15–44
45–64
Country
Canada
Czech Republic
Germany
Israel
Netherland
New Zealand
Spain
Canada
Czech Republic
Germany
Israel
Netherland
New Zealand
Spain
Australia
Canada
Czech Republic
Finland
Germany
Israel
Netherlands
New Zealand
Spain
Australia
Canada
Czech Republic
Finland
Germany
Israel
Netherlands
New Zealand
Spain
Australia
Canada
Czech Republic
Finland
Germany
Israel
Netherlands
New Zealand
Spain
Australia
Canada
Czech Republic
Finland
Germany
Years
1991–2015
2008–2013
2001–2016
1998–2016
2003–2017
1997–2015
2005–2015
1991–2015
2008–2013
2001–2016
1998–2016
2003–2017
1997–2015
2005–2015
2001–2016
1991–2015
2008–2013
1995–2016
2001–2016
1998–2016
2003–2017
1997–2015
2005–2005
2001–2016
1991–2015
2008–2013
1995–2016
2001–2016
1998–2016
2003–2017
1997–2015
2005–2005
2001–2016
1991–2015
2008–2013
1995–2016
2001–2016
1998–2016
2003–2017
1997–2015
2005–2005
2001–2016
1991–2015
2008–2013
1995–2016
2001–2016
Males
n/N
29/4682619
25/349195
27/5740478
38/1486100
3/1616870
3/576900
41/2679186
679/19156418
267/1410748
615/23509315
778/5731500
94/5811264
80/2308880
587/10880587
258/11398585
1556/24668602
256/1532669
42/3440956
1361/30760941
1263/6616300
208/7478265
126/2899540
983/13017097
190/11377822
954/25685783
157/1416001
45/3522497
908/33455166
522/6106400
182/7658243
72/2919850
535/12301238
1462/73591102
7148/143987472
1428/13725818
562/18898064
3840/257895408
1103/29586200
774/39930903
401/13546700
4962/110542308
420/41988401
2419/110461323
356/8403729
220/16513241
Females
IR
0.62
7.16
0.47
2.56
0.19
0.52
1.53
3.54
18.93
2.62
13.57
1.62
3.46
5.39
2.26
6.31
16.70
1.22
4.42
19.09
2.78
4.35
7.55
1.67
3.71
11.09
1.28
2.71
8.55
2.38
2.47
4.35
1.99
4.96
10.40
2.97
1.49
3.73
1.94
2.96
4.49
1.00
2.19
4.24
1.33
n/N
32/4446799
23/332712
23/5448550
34/1410400
3/1540059
1/548520
17/2514548
560/18225737
222/1343670
511/22311030
550/5443300
88/5543773
55/2191980
419/10233932
212/10814642
1506/23469919
220/1450621
42/3297629
1196/29187252
1012/6287700
243/7139402
139/2752910
821/12287011
144/10797396
854/24391864
148/1339518
42/3375446
846/31724889
430/5807300
151/7312854
80/2776650
409/11627137
990/72741755
3620/140453550
918/12978912
332/18050351
2753/247590330
878/29264100
414/39138712
295/13976900
2425/105413400
330/42573071
1390/109655649
347/8624880
177/16307550
IR
0.72
6.91
0.42
2.41
0.19
0.18
0.68
3.07
16.52
2.92
10.10
1.59
2.51
4.09
1.96
6.42
15.17
1.27
4.10
16.09
3.40
5.05
6.68
1.33
3.50
11.05
1.24
2.67
7.40
2.06
2.88
3.52
1.36
2.58
7.07
1.84
1.11
3.00
1.06
2.11
2.30
0.78
1.27
4.02
1.09
1742/181698132
15.55
1694/181849520
15.08
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
RR
0.86
1.04
1.11
1.06
0.95
2.85
2.26
1.15
1.15
1.14
1.34
1.02
1.38
1.32
1.15
0.98
1.10
0.96
1.08
1.19
0.82
0.86
1.13
1.25
1.06
1.00
1.03
1.02
1.15
1.15
0.86
1.24
1.46
1.93
1.47
1.62
1.34
1.24
1.83
1.40
1.95
1.29
1.73
1.05
1.23
1.03
(Continued )
4 / 18
PLOS ONE Male excess in hepatitis A incidence rates
Table 1. (Continued)
Age
65+
Country
Israel
Netherlands
New Zealand
Spain
Australia
Canada
Czech Republic
Finland
Germany
Israel
Netherlands
New Zealand
Spain
Years
1998–2016
2003–2017
1997–2015
2005–2005
2001–2016
1991–2015
2008–2013
1995–2016
2001–2016
1998–2016
2003–2017
1997–2015
2005–2005
Males
n/N
140/12368500
400/36361477
186/10201030
601/63103755
120/21417772
693/64590224
50/4087584
75/11159619
948/108019284
65/6010700
95/24551738
75/6302700
131/37127234
Females
IR
1.13
1.10
1.82
0.95
0.56
1.07
1.22
0.67
0.88
1.08
0.39
1.19
0.35
n/N
129/13327000
211/35859166
126/10685350
432/64340310
151/25538457
816/78346403
93/5999018
74/15066114
1435/149862231
54/7903600
89/29339281
70/7386000
133/49879431
IR
0.97
0.59
1.18
0.67
0.59
1.04
1.55
0.49
0.96
0.68
0.30
0.95
0.27
RR
1.17
1.87
1.55
1.42
0.95
1.03
0.79
1.37
0.92
1.58
1.28
1.26
1.32
IR = incidence rate, IR per 100 000 Male or Female population, incidence RR = female: male incidence Rate Ratio
n- Cumulative total of cases for given years.
N- Cumulative total of the population for given years.
Infants = age<1 year; early childhood = 1–4 years; late childhood = 5–9 years; puberty = 10–14 years; young adulthood = 15–44 or 15–39 years; middle adulthood = 40–
59 or 45–64 years; senior adulthood = 60+ or 65+ years.
https://doi.org/10.1371/journal.pone.0287008.t001
The forest plot for the age 1–4 is shown in Fig 2. The pooled IRR was 1.22 (95% CI 1.16–
1.29) with I2 = 23.3% and varied from 1.02 in Netherland 1.38 in New Zealand.
The forest plot for age 5–9 is given in Fig 3. The pooled IRR was 1.07 (95% CI 1.03–1.11)
with I2 = 30.3% and varied from 0.82 in the Netherlands to 1.19 in Israel.
The forest plot for age 10–14 is given in Fig 4. The pooled IRR was 1.09 (95% CI 1.04–1.14)
with I2 = 0.0% and varied between 0.86 in New Zealand to 1.25 in Australia.
The forest plot for age 15–44 is given in Fig 5. The pooled IRR was 1.46 (95% CI 1.30–1.64),
I2 = 95.9% and varied between 1.33 in Israel to 1.86 in the Netherlands.
The forest plot for age 45–64 is shown in Fig 6. The pooled IRR = 1.32 (95% CI 1.15–1.51),
I2 = 90.8%, and varied from 1.03 in Germany to 1.87 in the Netherlands.
The forest plot for age 65+ is given in Fig 7. The pooled IRR was 1.10 (95% CI 0.99–1.23) I2
= 57.0% and varied from 0.80 in Czech Republic to 1.50 in Israel.
Other analyses. Meta-regression analysis showed that almost all the variance in the inci-
dence RRs was contributed by the age groups, with small differences between countries and
time periods. To evaluate the effect of individual countries on the male to female incidence
ratios, we performed leave-one-out sensitivity analysis and recomputed the pooled IRRs (pre-
sented in Tables 2 and 3).
After omitting each country (one country at a time, Table 2) or a group of years at a time
(Table 3), the pooled IRR’s remained very similar.
Thus, no single country or group of years substantially affected the pooled IRRs. This con-
firms that the results of this pooled analysis are stable and robust.
Discussion
In this study, we found that the incidence rates of clinically manifested HAV, pooled over a
number of years, for various high-income countries, are consistently higher in males in all age
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
5 / 18
PLOS ONE Male excess in hepatitis A incidence rates
Fig 1. Forest plot of the male to female hepatitis A incidence rate ratios (IRR) in infancy for different years in Canada, Czech Republic, Germany, Israel,
Netherland, New Zealand, and Spain.
https://doi.org/10.1371/journal.pone.0287008.g001
groups. In the youngest and oldest age groups, where the numbers were small, the confidence
intervals included unity. Based on the pooled analysis of national data from nine countries,
over a period of 6–25 years, we found that the incidence rates of clinical hepatitis A were
higher in males by 22%, 7%, 9%, 46%, 32%, and 10% in the age groups 1–4, 5–9, 10–14, 15–44,
45–64 and 65+ respectively.
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
6 / 18
PLOS ONE Male excess in hepatitis A incidence rates
Fig 2. Forest plot of the male to female hepatitis A incidence rate ratios (IRR) at age 1–4, for different years in Canada, Czech Republic,
Germany, Israel, Netherland, New Zealand, and Spain.
https://doi.org/10.1371/journal.pone.0287008.g002
While sex differences in the incidence of HAV have been examined in a number of studies,
they have usually been conducted in individual countries or selected groups of patients. For
example, in a national study in Israel in 1992, there was a male predominance of HAV inci-
dence rates [30]. This sex differential was especially pronounced among infants. In a 15-year
nationwide epidemiological study in Taiwan, there were higher hospitalization rates in males
while male sex and age over 40 years were significant factors associated with mortality [31]. In
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
7 / 18
PLOS ONE Male excess in hepatitis A incidence rates
Fig 3. Forest plot of the male to female hepatitis A incidence rate ratios (IRR) at age 5–9, for different years in Australia, Canada,
Czech Republic, Finland, Germany, Israel, Netherland, New Zealand, and Spain.
https://doi.org/10.1371/journal.pone.0287008.g003
the study of HAV patients in Saudi Arabia, no sex differences were among hospitalized
patients [12]. In a hepatitis A outbreak in Chiba, Japan, in 2011, 40.7% of the 27 patients were
male [32], and in another, 65% of the 60 patients were male [33]. However, these figures may
simply represent gender differences in exposure to the virus. In addition, the impact of vac-
cines on sex differences in HAV incidence rates is not clear. There is evidence that females
may respond with up to 2–3 times higher anti-HAV antibody levels than males after the prim-
ing and after the booster dose and has been observed at different ages [34–37].
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
8 / 18
PLOS ONE Male excess in hepatitis A incidence rates
Fig 4. Forest plot of the male to female hepatitis A incidence rate ratios (IRR) at age 10–14, for different years in Australia, Canada,
Czech Republic, Finland, Germany, Israel, Netherland, New Zealand, and Spain.
https://doi.org/10.1371/journal.pone.0287008.g004
The incidence of both viral and bacterial diseases have frequently been reported to be
higher in males [1–3]. In addition, there are reported sex differences in the severity of different
infections, suggesting that males are more prone to suffer from clinical manifestations of infec-
tions than females [38, 39]. While in excess morbidity in males is most common for infectious
diseases [1–3], pertussis is a prominent exception, where there is a female excess in morbidity
[40]. It is of interest that in the COVID-19 pandemic, there has been no clear evidence of sex
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
9 / 18
PLOS ONE Male excess in hepatitis A incidence rates
Fig 5. Forest plot of the male to female hepatitis A incidence rate ratios (IRR) at age 15–44 (15–39), for different years in
Australia, Canada, Czech Republic, Finland, Germany, Israel, Netherland, New Zealand, and Spain.
https://doi.org/10.1371/journal.pone.0287008.g005
differences in incidence rates, although case-fatality rates have consistently been reported to be
higher in males [41, 42], even after controlling for other variables.
It has been shown that the male to female IRRs differential will be most evident where there
is a low proportion of clinical disease [30]. Since children more commonly suffer from asymp-
tomatic HAV infection [43, 44] and the clinical to subclinical ratio for HAV increases with
age, one might expect that the male excess in disease would be less evident at older ages.
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
10 / 18
PLOS ONE Male excess in hepatitis A incidence rates
Fig 6. Forest plot of the male to female hepatitis A incidence rate ratios (IRR) at age 45-64(40–59), for different years in Australia,
Canada, Czech Republic, Finland, Germany, Israel, Netherland, New Zealand, and Spain.
https://doi.org/10.1371/journal.pone.0287008.g006
However, the higher male to female IRRs in the older age groups is most likely due to larger
differences in exposure in high risk groups such as in the men who have sex with men (MSM)
or people who are HIV positive [7, 8, 45–49]. Thus, behavioral factors can partially explain sex
differences in HAV incidence rates in the older age groups. For the youngets age group, there
may be protection from maternal HAV antibodies on short-term immunity [50]. However, we
have not found evidence that it impacts male and female infants differently.
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
11 / 18
PLOS ONE Male excess in hepatitis A incidence rates
Fig 7. Forest plot of the male to female hepatitis A incidence rate ratios (IRR) at age 65+ (60+), for different years in Australia,
Canada, Czech Republic, Finland, Germany, Israel, Netherland, New Zealand, and Spain.
https://doi.org/10.1371/journal.pone.0287008.g007
The exact mechanisms underlying the excess HAV incidence rates in males found in this
study are not clear and probably multi-factorial. This study was not designed to address the
mechanisms. In addition to behavioral differences, genetic and hormonal factors could be
important. In infants and early childhood, and based on the seroprevalence studies, it is
unlikely that the sex differences in incidence rates are due to differences in exposure [51]. A
study of kindergarten children showed that females had higher anti-HAV antibodies than
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
12 / 18
PLOS ONE Male excess in hepatitis A incidence rates
Table 2. Sensitivity analysis, by age group and country.
Sensitivity by country
Countries Removed
Infants
Early childhood
Late childhood
Puberty
Young adulthood Middle adulthood
Senior adulthood
Australia
Canada
-
-
1.04 (0.96–1.13)
1.08 (1.03–1.14)
1.58 (1.37–1.82)
1.35 (1.11–1.64)
1.13 (0.98–1.3)
1.29 (0.99–1.66)
1.24 (1.17–1.32)
1.06 (0.98–1.15)
1.1 (1.04–1.16)
1.52 (1.33–1.74)
1.3 (1.11–1.51)
1.13 (0.96–1.34)
Czech Republic
1.21 (0.95–1.56)
1.23 (1.16–1.3)
1.04 (0.96–1.13)
1.09 (1.04–1.15)
1.58 (1.37–1.81)
1.39 (1.15–1.67)
1.14 (1–1.29)
Finland
Germany
Israel
Netherland
New Zealand
Spain
-
-
1.05 (0.97–1.14)
1.09 (1.04–1.14)
1.56 (1.36–1.79)
1.36 (1.12–1.64)
1.08 (0.95–1.23)
1.2 (0.93–1.54)
1.24 (1.17–1.32)
1.04 (0.95–1.14)
1.12 (1.06–1.18)
1.6 (1.41–1.81)
1.4 (1.2–1.62)
1.15 (1–1.32)
1.23 (0.94–1.59)
1.19 (1.12–1.26)
1.03 (0.95–1.11)
1.08 (1.02–1.14)
1.61 (1.42–1.83)
1.36 (1.13–1.65)
1.07 (0.95–1.2)
1.19 (0.94–1.5)
1.23 (1.17–1.3)
1.08 (1.01–1.15)
1.09 (1.03–1.14)
1.54 (1.34–1.76)
1.29 (1.07–1.55)
1.09 (0.95–1.24)
1.17 (0.93–1.47)
1.22 (1.16–1.29)
1.08 (1.01–1.15)
1.1 (1.04–1.15)
1.59 (1.38–1.81)
1.32 (1.09–1.6)
1.09 (0.96–1.25)
1.02 (0.8–1.32)
1.2 (1.13–1.28)
1.08 (1.01–1.15)
1.07 (1.01–1.12)
1.52 (1.33–1.74)
1.33 (1.09–1.63)
1.08 (0.95–1.22)
IRR = Incidence rate ratio; CI = confidence interval
https://doi.org/10.1371/journal.pone.0287008.t002
males [52]. In adults, the results are varied. In a study of blood donors in the US in 2015, [53]
no sex differences were observed in the prevalence of anti-HAV IgG antibodies (61% and 60%
for males and females, respectively). In a study of ambulatory patients in Portugal between
2002 and 2012, no significant differences between sexes were observed [54]. In a study of refu-
gees and asylum seekers in Germany, HAV seroprevalence rates were higher in adult males
than females [55].
Although liver injury in hepatitis A is known to be caused by immune-mediated events, the
exact biological mechanisms are not clarified. It is plausible that immune-related mechanisms
of liver injury are common to the pathogenesis of all types of hepatitis [56]. Virus-specific CD8
+ T cells from hepatitis A patients are considered as a major cause of liver damage. Natural
killer cells are also involved and contribute to liver damage [57, 58]. In hepatitis A patients,
serum levels of cytokines and chemokines, including interleukin (IL)-6, IL-8, IL-18, IL-22,
CXC-chemokine ligand (CXCL)9, and CXCL10 are increased [59] and contribute to liver
injury. Many studies have shown that the overall inflammatory response, innate and adaptive
immune systems are stronger in females than males, with greater CD4+ T-cell counts a higher
CD4+ /CD8+ ratio in females but higher CD8+ T and NK frequencies in males [60].
Sex differences in the clinical expression of hepatitis A may be related to the imbalance in
the expression of genes encoded on the X and Y-chromosomes of a host. X chromosome-asso-
ciated biological processes and X-linked genes are responsible for the immunological advan-
tage of females due to the X-linked microRNAs related processes. The phenomenon of X
chromosome inheritance and expression is a cause of immune disadvantage of males and the
enhanced survival of females following immunological challenges [61].
The increase in sex hormone levels in infancy that mimics sex steroid levels during puberty
(‘minipuberty’) could affect immune cells differently in boys and girls. Testosterone levels
Table 3. Sensitivity analysis, by age group and years.
Sensitivity by years
Years Removed
Infants
Early childhood
Late childhood
Puberty
Young adulthood Middle adulthood
Senior adulthood
1991–1999
2000–2009
2010–2017
1.2 (0.92–1.58)
1.21 (1.13–1.29)
1.08 (1.03–1.13)
1.11 (1.05–1.17)
1.49 (1.3–1.72)
1.19 (1.14–1.24)
0.97 (0.87–1.09)
1.07 (0.79–1.45)
1.21 (1.12–1.29)
1.04 (0.99–1.09)
1.07 (1–1.14)
1.68 (1.16–2.43)
1.51 (0.92–2.48)
1.13 (0.94–1.35)
1.26 (0.96–1.64)
1.25 (1.18–1.33)
1.07 (1.03–1.12)
1.09 (1.03–1.15)
1.8 (1.43–2.27)
1.52 (0.94–2.46)
1.06 (0.79–1.42)
IRR = Incidence rate ratio; CI = confidence interval
https://doi.org/10.1371/journal.pone.0287008.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
13 / 18
PLOS ONE Male excess in hepatitis A incidence rates
predominate in boys at 1–3 months of age and decline at 6–9 months of age, whereas in girls,
estradiol levels remain elevated longer [62]. This phenomenon of ’’mini-puberty’’ with sex dif-
ferences in gonadal hormone levels could influence the maturation of the immune system
[63]. This transient rise in sex steroid levels may also influence immune cells differently
between boys and girls at later ages [64]. Before any physical signs of puberty, girls had higher
levels of estrogens than boys at age 5–9. These higher estradiol levels or lower testosterone lev-
els in young girls may play a part in protection against clinical disease and should be investi-
gated further.
Strengths and limitations
This current study has several strengths and limitations. The inclusion of nine countries, each
evaluated over a number of years, allowed us to evaluate the consistency of the findings over
different populations and many years. The analyses are based on national data where both the
numbers of cases and denominators are large. Selection bias has been minimized by using
national data, which should be representative of each country. However, the countries evalu-
ated in this study are classified as high-income, so the results may not be directly generalizable
to low- and middle-income countries. Differential underreporting between countries is likely
and may contribute to the variability in the incidence of reported cases of HAV. However,
there does not appear to be any reason to believe that the reporting differs between males and
females. In the countries examined, there is no evidence that male infants and children are
more likely to receive health care. Thus any information bias in the underreporting of inci-
dence rates will most likely be non-differential by sex and the IRRs should not be materially
affected. In adults, there could be gender differences in the utilization of medical care, although
reports suggest that females in some countries tend to make greater use of health services [65],
which would operate in the opposite direction of our observations.
Conclusions
This study provides stable estimates of the excess male incidence rates in hepatitis A incidence
rates in most age groups. While much of the excess in older males may be attributed to differ-
ential exposure, the excess in young males, while not large, is remarkably consistent over a
number of high-income countries and for extended periods of time. The mechanism is largely
unknown. A better understanding of the gender differences can help to elucidate genetic and
hormonal determinants of HAV infection and contribute to the role of sex as a biological
variable.
Acknowledgments
We thank the official representative of RIVM, Netherlands, and to all the official institutions
of all other countries for the providing their national data on hepatitis A incidence.
Author Contributions
Conceptualization: Manfred S. Green.
Data curation: Manfred S. Green, Naama Schwartz, Victoria Peer.
Formal analysis: Naama Schwartz.
Methodology: Manfred S. Green, Naama Schwartz, Victoria Peer.
Project administration: Manfred S. Green.
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
14 / 18
PLOS ONE Male excess in hepatitis A incidence rates
Supervision: Manfred S. Green.
Writing – original draft: Manfred S. Green.
Writing – review & editing: Manfred S. Green, Victoria Peer.
References
1. Green MS, Schwartz N, Peer V. Sex differences in campylobacteriosis incidence rates at different ages
—a seven country, multi-year, meta-analysis. A potential mechanism for the infection. BMC Infect Dis.
2020; 20:625. https://doi.org/10.1186/s12879-020-05351-6 PMID: 32842973
2. Peer V, Schwartz N, Green MS. Consistent, excess viral meningitis incidence rates in young males: A
multi-country, multi-year, meta-analysis of national data. The importance of sex as a biological variable.
EClinicalMedicine. 2019; 15:62–71. https://doi.org/10.1016/j.eclinm.2019.08.006 PMID: 31709415
3. Peer V, Schwartz N, Green MS. A pooled analysis of sex differences in rotaviral enteritis incidence
rates in three countries over different time periods. Womens Health Rep (New Rochelle). 2022; 3:228–
237 https://doi.org/10.1089/whr.2021.0096 PMID: 35262061
4.
5.
Jacobsen KH. Globalization and the changing epidemiology of hepatitis A virus. Cold Spring Harb Per-
spect Med. 2018; 8:a031716. https://doi.org/10.1101/cshperspect.a031716 PMID: 29500305
Zeng Dan-Yi, Li Jing-Mao, Lin Su, et al. Global burden of acute viral hepatitis and its association with
socioeconomic development status, 1990–2019. J Hepatol. 2021. 75:547–556. https://doi.org/10.
1016/j.jhep.2021.04.035 PMID: 33961940
6. Green MS, Block C, Slater PE. Rise in the incidence of viral hepatitis in Israel despite improved socio-
economic conditions. Rev Infect Dis. 1989; 11:464–469. https://doi.org/10.1093/clinids/11.3.464 PMID:
2749104
7. Alberts CJ, Boyd A, Bruisten SM, et al. Hepatitis A incidence, seroprevalence, and vaccination decision
among MSM in Amsterdam, the Netherlands. Vaccine. 2019; 37:2849–2856 https://doi.org/10.1016/j.
vaccine.2019.03.048 PMID: 30992222
8. Honda M, Asakura H, Kanda T, et. al. Male-dominant hepatitis A outbreak observed among non-HIV-
infected persons in the northern part of Tokyo, Japan.Viruses. 2021; 13:207.
9. Andani A, Bunge E, Kassianos G, et.al. Hepatitis A occurrence and outbreaks in Europe over the past
two decades: a systematic review. J Viral Hepat. 2023
10. Bauer D, Farthofer A, Chromy D, et al. Recent outbreaks of severe hepatitis A virus infections in Vienna.
Eur J Clin Microbiol Infect Dis. 2021; 40:335–344. https://doi.org/10.1007/s10096-020-04028-x PMID:
32940811
11. Dudareva S, Faber M, Zimmermann R, et.al. Epidemiology of viral hepatitis A to E in Germany. Bundes-
gesundheitsblatt Gesundheitsforschung Gesundheitsschutz. 2022; 65:149–158.
12. Al-Tawfiq JA, Anani A. Profile of viral hepatitis A, B, and C in a Saudi Arabian hospital. Med Sci Monit.
2008; 14:CR52–56. PMID: 18160946
13. Choe YJ, Son H. The changing gender differences in hepatitis A incidence in South Korea. Vaccine.
2020; 38: 712–714. https://doi.org/10.1016/j.vaccine.2019.11.048 PMID: 31787416
14. National Notifiable Diseases Surveillance System (NNDSS), Department of Health. Available at: http://
www9.health.gov.au/cda/source/rpt_5_sel.cfm. Accessed 1 April 2018.
15. Public Health Agency of Canada. Available at: https://www.canada.ca/en/public-health.html. Accessed
on 1 June 2018.
16.
Institute of Health Information and Statistics. Available at: https://www.uzis.cz/en/catalogue/infectious-
diseases. Accessed 1 March 2018.
17. National institute for health and welfare (THL): https://www.thl.fi/ttr/gen/rpt/tilastot.html. Accessed on
May 1, 2018.
18. German Federal Health Monitoring System. Available at: http://www.gbe-bund.de/ gbe10/pkg_isgbe5.
prc_isgbe? p_uid¼gast&p_aid¼0&p_sprache¼D (1 February 2018, date last accessed).
19. Environmental Science and Research (ESR) for the Ministry of Health. Available at: https://surv.esr.cri.
nz/surveillance/annual_surveillance.php. Accessed 30 March 2018.
20.
Instituto de Salud Carlos III. Available: http://www.eng.isciii.es/ISCIII/es/contenidos/fd-servicios-
cientifico-tecnicos/fd-vigilancias-alertas/fd-enfermedades/enfermedades-declaracion-obligatoria-
informes-anuales.shtml. Accessed 1 March 2018.
21. ABS.Stat (Australian Bureau of Statistics). Available at: http://stat.data.abs.gov.au/Index.aspx?
DatasetCode=ABS_ERP_ASGS2016.Accessed 15 May 2018.
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
15 / 18
PLOS ONE Male excess in hepatitis A incidence rates
22. Statistics, Canada, CANSIM database: Available at: https://www150.statcan.gc.ca/t1/tbl1/en/cv.action?
pid=1710010201. Accessed 1 June 2018.
23. Czech Statistical Office. Available at: https://www.czso.cz/csu/czso/population. Accessed 1 March
2018.
24. Statistics Finland’s PX-Web databases. Available at: http://pxnet2.stat.fi/PXWeb/pxweb/en/StatFin/
StatFin__vrm__vaerak/statfin_vaerak_pxt_021.px/?rxid=2f968705-bdaa-48b1-9d5a-d4985ead7d40.
Accessed 15 April 2018.
25. German Federal Health Monitoring System: http://www.gbe-bund.de/gbe10/abrechnung.prc_abr_test_
logon?p_uid=gast&p_aid=46300054&p_knoten=VR&p_sprache=E&p_suchstring=population.
Accessed on February 1, 2018
26. Central Bureau of Statistics: http://www.cbs.gov.il/reader/shnatonhnew_site.htm?sss=%E4%EE%F9%
EA&shnaton_scan=45. Accessed on March 1, 2018.
27. Statistics Netherlands’ database (StatLine). Available at: https://opendata.cbs.nl/statline/#/CBS/en/
dataset/37325eng/table?ts=1528798782913. Accessed 15 May 2018.
28. Stats NZ, Infoshare. Available at: http://archive.stats.govt.nz/infoshare/SelectVariables.aspx?pxID=
b854d8a2-3fdf-402c-af69-604112e80baa. Accessed 15 May 2018.
29. Demographic Statistics Database (United Nations Statistics: Division). Available at: http://data.un.org/
Data.aspx?d=POP&f=tableCode%3A22. Accessed 1 April 2018.
30. Green MS. The male predominance in the incidence of infectious diseases in children: a postulated
explanation for disparities in the literature. Int J Epidemiol. 1992; 21:381–386 https://doi.org/10.1093/
ije/21.2.381 PMID: 1428496
31. Chen CM, Chen SC, Yang HY, Yang ST, Wang CM. Hospitalization and mortality due to hepatitis A in
Taiwan: a 15-year nationwide cohort study. J Viral Hepat. 2016; 23:940–945. https://doi.org/10.1111/
jvh.12564 PMID: 27386835
32.
33.
Tominaga A, Kanda T, Akiike T, et.al. Hepatitis A outbreak associated with a revolving sushi bar in
Chiba, Japan: Application of molecular epidemiology. Hepatol Res. 2012; 42: 828–834. https://doi.org/
10.1111/j.1872-034X.2012.00988.x PMID: 22776552
Takahashi H, Yotsuyanagi H, Yasuda K, et.al. Molecular epidemiology of hepatitis A virus in metropoli-
tan areas in Japan. J Gastroenterol. 2006; 41:981–986. https://doi.org/10.1007/s00535-006-1888-9
PMID: 17096067
34. Bovier PA, Bock J, Loutan L, Farinelli T, Glueck R, Herzog C. Long-term immunogenicity of an inacti-
vated virosome hepatitis A vaccine. J Med Virol. 2002; 68:489–493. https://doi.org/10.1002/jmv.10244
PMID: 12376955
35. Ho¨hler T, Groeger-Bicanic G, Hoet B, Stoffel M. Antibody persistence and immune memory elicited by
combined hepatitis A and B vaccination in older adults. Vaccine. 2007; 25:1503–1508. https://doi.org/
10.1016/j.vaccine.2006.10.024 PMID: 17097774
36. Spradling PR, Bulkow LR, Negus SE, Homan C, Bruce MG, McMahon BJ. Persistence of seropositivity
among persons vaccinated for hepatitis A during infancy by maternal antibody status: 15-year follow-
up. Hepatology. 2016; 63:703–711. https://doi.org/10.1002/hep.28375 PMID: 26637987
37. Van Herck K, Hens A, De Coster I, et.al. Long-term antibody persistence in children after vaccination
with the pediatric formulation of an aluminum-free virosomal hepatitis A vaccine. Pediatr Infect Dis J.
2015; 34:e85–91 https://doi.org/10.1097/INF.0000000000000616 PMID: 25389920
38. Klein SL, Flanagan KL. Sex differences in immune responses. Nat Rev Immunol. 2016; 16:626–638
https://doi.org/10.1038/nri.2016.90 PMID: 27546235
39. Klein SL, Jedlicka A, Pekosz A. The Xs and Y of immune responses to viral vaccines. Lancet Infect Dis.
2010; 10:338–349. https://doi.org/10.1016/S1473-3099(10)70049-9 PMID: 20417416
40. Peer V, Schwartz N, Green MS. A multi-country, multi-year, meta-analytic evaluation of the sex differ-
ences in age-specific pertussis incidence rates. PLoS One. 2020; 15:e0231570 https://doi.org/10.1371/
journal.pone.0231570 PMID: 32324790
41.
Jacobsen H, Klein SL. Sex differences in immunity to viral Infections. Front Immunol. 2021; 12:720952
https://doi.org/10.3389/fimmu.2021.720952 PMID: 34531867
42. Green MS, Nitzan D, Schwartz N, Niv Y, Peer V. Sex differences in the case-fatality rates for COVID-
19-A comparison of the age-related differences and consistency over seven countries. PLoS One.
2021; 16(4):e0250523 https://doi.org/10.1371/journal.pone.0250523 PMID: 33914806
43. Wensley A, Smout E, Ngui SL, et.al. An outbreak of hepatitis A virus infection in a secondary school in
England with no undetected asymptomatic transmission among students. Epidemiol Infect. 2022; 151:
e6 https://doi.org/10.1017/S095026882200190X PMID: 36502811
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
16 / 18
PLOS ONE Male excess in hepatitis A incidence rates
44. Abutaleb A, Kottilil S. Hepatitis A: epidemiology, natural history, unusual clinical manifestations, and
prevention. Gastroenterol Clin North Am. 2020; 49:191–199. https://doi.org/10.1016/j.gtc.2020.01.002
PMID: 32389358
45.
46.
Lin KY, Chen GJ, Lee YL, et.al. Hepatitis A virus infection and hepatitis A vaccination in human immuno-
deficiency virus-positive patients: A review. World J Gastroenterol. 2017; 23: 3589–3606. https://doi.
org/10.3748/wjg.v23.i20.3589 PMID: 28611512
Franco E, Giambi C, Ialacci R, Coppola RC, Zanetti AR. Risk groups for hepatitis A virus infection. Vac-
cine. 2003; 21:2224–2233 https://doi.org/10.1016/s0264-410x(03)00137-3 PMID: 12744847
47. Migueres M, Lhomme S, Izopet J. Hepatitis A: epidemiology, high-risk groups, prevention and research
on antiviral treatment. Viruses. 2021; 13:1900 https://doi.org/10.3390/v13101900 PMID: 34696330
48. Yoshimura Y, Horiuchi H, Sawaki K, et.al. Hepatitis A Outbreak Among Men Who Have Sex With Men,
Yokohama, Japan, January to May 2018. Sex Transm Dis. 2019; 46: e26–e27. https://doi.org/10.1097/
OLQ.0000000000000937 PMID: 30395105
49. Bazzardi R, Dore E, Ciccozzi M, et.al. Outbreak of acute hepatitis A associated with men who have sex
with men (MSM) in North Sardinia 2017–2018. J Infect Dev Ctries. 2020; 14:1065–1070. https://doi.org/
10.3855/jidc.12184 PMID: 33031097
50. Bell BP, Negus S, Fiore AE, et.al. Immunogenicity of an inactivated hepatitis A vaccine in infants and
young children. J.Pediatr Infect Dis J. 2007; 26: 116–122. https://doi.org/10.1097/01.inf.0000253253.
85640.cc PMID: 17259872
51. Walter F, Ott JJ, Claus H, Krause G. Sex- and age patterns in incidence of infectious diseases in Ger-
many: analyses of surveillance records over a 13-year period (2001–2013). Epidemiol Infect. 2018;
146: 372–378. https://doi.org/10.1017/S0950268817002771 PMID: 29357958
52.
53.
Lin DB, Tsai TP, Yang CC, et.al. Association between seropositivity of antibodies against hepatitis a
virus and Helicobacter pylori. Am J Trop Med Hyg. 2000; 63:189–191. https://doi.org/10.4269/ajtmh.
2000.63.189 PMID: 11388513
Tejada-Strop A, Zafrullah M, Kamili S, Stramer SL, Purdy MA. Distribution of hepatitis A antibodies in
US blood donors. Transfusion. 2018; 58: 2761–2765. https://doi.org/10.1111/trf.14916 PMID:
30284286
54. Pereira S, Linhares I, Neves AF, Almeida A. Hepatitis A immunity in the District of Aveiro (Portugal): an
eleven-year surveillance study (2002–2012). Viruses. 2014; 6: 133613–45. https://doi.org/10.3390/
v6031336 PMID: 24638206
55.
Jablonka A, Solbach P, Wo¨bse M, et.al. Seroprevalence of antibodies and antigens against hepatitis A-
E viruses in refugees and asylum seekers in Germany in 2015. Eur J Gastroenterol Hepatol. 2017; 29:
939–945. https://doi.org/10.1097/MEG.0000000000000889 PMID: 28492419
56. Shin EC, Sung PS, Park SH. 2016a. Immune responses and immunopathology in acute and chronic
viral hepatitis. Nat Rev Immunol. 2016: 509–523. https://doi.org/10.1038/nri.2016.69 PMID: 27374637
57. Wang M, Feng Z. Mechanisms of Hepatocellular Injury in Hepatitis A. Viruses. 2021; 13:861. https://doi.
org/10.3390/v13050861 PMID: 34066709
58. Vallbracht A, Maier K, Stierhof YD, Wiedmann KH, Flehmig B, Fleischer B. Liver-derived cytotoxic T
cells in hepatitis A virus infection. J Infect Dis. 1989; 160:209–217. https://doi.org/10.1093/infdis/160.2.
209 PMID: 2503564
59. Shin SY, Jeong SH, Sung PS, et.al. Comparative analysis of liver injury-associated cytokines in acute
hepatitis A and B. Yonsei Med J. 2016; 57:652–657. https://doi.org/10.3349/ymj.2016.57.3.652 PMID:
26996565
60. Abdullah M, Chai P-S, Chong M-Y, et al. Gender effect on in vitro lymphocyte subset levels of healthy
individuals. Cell Immunol. 2012; 272:214–219. https://doi.org/10.1016/j.cellimm.2011.10.009 PMID:
22078320
61. Schurz H, Salie M, Tromp G, Hoal EG, Kinnear CJ, Mo¨ ller M. The X chromosome and sex-specific
effects in infectious disease susceptibility. Hum Genomics. 2019; 13:2. https://doi.org/10.1186/s40246-
018-0185-z PMID: 30621780
62.
Lanciotti L, Cofini M, Leonardi A, Penta L, Esposito S. Up-To-Date review about minipuberty and over-
view on hypothalamic-pituitary-gonadal axis activation in fetal and neonatal life. Front Endocrinol (Lau-
sanne). 2018; 9:410. https://doi.org/10.3389/fendo.2018.00410 PMID: 30093882
63. Moreira-Filho CA, Bando SY, Bertonha FB, et.al Minipuberty and sexual dimorphism in the infant
human thymus. Sci Rep. 2018; 8: 13169. https://doi.org/10.1038/s41598-018-31583-3 PMID:
30177771
64. Courant F, Aksglaede L, Antignac JP, et.al. Assessment of circulating sex steroid levels in prepubertal
and pubertal boys and girls by a novel ultrasensitive gas chromatography-tandem mass spectrometry
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
17 / 18
PLOS ONE Male excess in hepatitis A incidence rates
method. J Clin Endocrinol Metab. 2010; 95:82–92. https://doi.org/10.1210/jc.2009-1140 PMID:
19933393
65. Bertakis KD, Azari R, Helms LJ, Callahan EJ, Robbins JA. Gender differences in the utilization of health
care services. J Fam Pract. 2000; 49:147–152. PMID: 10718692
PLOS ONE | https://doi.org/10.1371/journal.pone.0287008 June 13, 2023
18 / 18
PLOS ONE
| null |
10.1007_s10858-020-00303-3.pdf
| null | null |
Journal of Biomolecular NMR (2020) 74:183–191
https://doi.org/10.1007/s10858-020-00303-3
ARTICLE
Slow ring flips in aromatic cluster of GB1 studied by aromatic 13C
relaxation dispersion methods
Matthias Dreydoppel1 · Heiner N. Raum1 · Ulrich Weininger1
Received: 28 November 2019 / Accepted: 27 January 2020 / Published online: 3 February 2020
© The Author(s) 2020
Abstract
Ring flips of phenylalanine and tyrosine are a hallmark of protein dynamics. They report on transient breathing motions of
proteins. In addition, flip rates also depend on stabilizing interactions in the ground state, like aromatic stacking or cation–π
interaction. So far, experimental studies of ring flips have almost exclusively been performed on aromatic rings without
stabilizing interactions. Here we investigate ring flip dynamics of Phe and Tyr in the aromatic cluster in GB1. We found that
all four residues of the cluster, Y3, F30, Y45 and F52, display slow ring flips. Interestingly, F52, the central residue of the
cluster, which makes aromatic contacts with all three others, is flipping significantly faster, while the other rings are flipping
with the same rates within margin of error. Determined activation enthalpies and activation volumes of these processes are
in the same range of other reported ring flips of single aromatic rings. There is no correlation of the number of aromatic
stacking interactions to the activation enthalpy, and no correlation of the ring’s extent of burying to the activation volume.
Because of these findings, we speculate that F52 is undergoing concerted ring flips with each of the other rings.
Keywords Aromatic interaction · NMR spectroscopy · Protein dynamics · Protein breathing · Protein stability
Introduction
Aromatic residues are overrepresented in protein binding
interfaces where they contribute to a significant part of the
binding free energy. They also contribute to a significant part
(roughly 25% of the volume in average) of the hydrophobic
core where they stabilize proteins in two ways. Firstly, they
are hydrophobic (especially Trp and Phe) and contribute to
the so called hydrophobic effect, where hydrophobic side
chains are excluded from the solvent (water) (Pace et al.
2014; Rose and Wolfenden 1993). Secondly, due to their
quadrupolar electrostatic character, they can be engaged
in specific aromatic-aromatic pair interactions (Burley and
Petsko 1985, 1989) and interact with cations (Mahadevi and
Sastry 2013) or sulfur (Valley et al. 2012).
Electronic supplementary material The online version of this
article (https ://doi.org/10.1007/s1085 8-020-00303 -3) contains
supplementary material, which is available to authorized users.
* Ulrich Weininger
[email protected]
1
Institute of Physics, Biophysics, Martin-Luther-University
Halle-Wittenberg, 06120 Halle (Saale), Germany
Additionally, many Phe and Tyr residues undergo fre-
quent 180° rotations ("ring flips") of the χ2 dihedral angle
(around the imaginary Cβ–Cγ–Cζ axis) (Campbell et al.
1975; Hattori et al. 2004; Hull and Sykes 1975; Wagner
et al. 1976, 1987; Weininger et al. 2013, 2014b; Wüthrich
and Wagner 1975; Yang et al. 2015). The requirement for
a ring flip to occur is that the surrounding undergoes con-
certed "breathing" motions with relatively large activation
volumes (Hattori et al. 2004; Li et al. 1999; Wagner 1980).
Thus, aromatic side chains are the perfect probe for such
transient dynamic processes in proteins. Additionally, a
ring flip directly reports on the energy difference between
the ground state and the transition state (90° tilted ring).
Because of the quadrupolar electrostatic nature, interac-
tions that are stabilizing the ground state are destabilizing
the transition state and thus are leading to slower ring flips.
Comparing ring flips for aromatic residues involved in dif-
ferent interactions should therefore provide an experimental
measure of the energy of these interactions.
Experimental measurements of ring flips, however, have
been limited so far to a handful of cases since their dis-
covery in the 1970s (Campbell et al. 1975; Hull and Sykes
1975; Wagner et al. 1976). Recently, new cases have been
reported (Weininger et al. 2014b; Yang et al. 2015) enabled
Vol.:(0123456789)1 3
184
Journal of Biomolecular NMR (2020) 74:183–191
by methodological advances in site-selective 13C labeling
(Lundström et al. 2007; Teilum et al. 2006), aromatic 13C
relaxation dispersion experiments (Weininger et al. 2012,
2014a) and the understanding of strong 1H–1H couplings
(Weininger et al. 2013). Additionally, ring flips can be stud-
ied by long scale MD simulations (Shaw et al. 2010) and
extremely fast ring flips are shown to affect order param-
eters (Kasinath et al. 2015). So far, ring flips in all but one
system (Nall and Zuniga 1990) are of aromatic residues
without specific interactions, like aromatic-aromatic pair
interactions (Burley and Petsko 1985, 1989) and interactions
with cations (Mahadevi and Sastry 2013) or sulfur (Valley
et al. 2012). They all show a similar activation enthalpy of
83–97 kJ mol−1 (Hattori et al. 2004; Weininger et al. 2014b),
while for Iso-2-cytochrome c higher activation enthalpies
of 117–150 kJ mol−1 have been observed (Nall and Zuniga
1990). Here the rings of Y46 and Y48 pack tightly together
in a typical aromatic-pair interaction, while Y67 packs
against the hem group.
Applying high pressure is an elegant way to slow down
ring flips and to study their activation volumes. So far, acti-
vation volumes have been determined to 27 mL mol−1 (Y6 in
HPr) (Hattori et al. 2004), 51 mL mol−1 (Y35 in BPTI) and
27 mL mol−1 (F45 in BPTI) (Li et al. 1999). A connection
with the energy of a ring flip is not known. For extremely
fast ring flips, that affect order parameters, no sizeable pres-
sure effect was observed (Kasinath et al. 2015).
Here we investigate slow ring flips in the aromatic clus-
ter of GB1 that have been found recently (Dreydoppel et al.
2018), using 13C aromatic relaxation dispersion methods
(Weininger et al. 2012, 2014a) in a temperature and pres-
sure dependent way. We found that all four residues of the
cluster (Y3, F30, Y45, F52) show slow ring flips. Y3, Y45
and F52 displayed nearly identical activation enthalpies and
activation volumes similar to previously determined (Camp-
bell et al. 1975; Hattori et al. 2004; Li et al. 1999; Weininger
et al. 2014b), while F30 did not allow any quantification.
Moreover, ring flip rates are nearly identical for Y3, Y45
(and F30) while ring flips for F52 are significantly faster.
F52 is the central part of the aromatic cluster, in contact with
all the other slow flipping rings. We speculate that standard
activation enthalpies and faster flip rates in the center of the
cluster point to correlated flip motions of F52 with all its
other ring partners, each at a time.
Materials and methods
Protein samples
1-13C and 2-13C glucose labeled GB1 (UniProtKB P06654)
was expressed and purified as described elsewhere (Lind-
man et al. 2006). 1-13C glucose labeling (Teilum et al. 2006)
results in site-selective 13C labeled Phe and Tyr δ positions,
2-13C glucose labeling (Lundström et al. 2007) in site-selec-
tive 13C labeled Phe and Tyr ε positions. It was dissolved to
a concentration of around 5 mM in 20 mM HEPES, 90%
H2O/10% D2O with addition of small amounts of NaN3. The
pH was adjusted to 7.0 in the sample.
NMR spectroscopy
All experiments were performed at Bruker Avance III
spectrometers at a static magnetic field strength of 14.1 T.
Aromatic L-optimized TROSY selected 13C CPMG (Wei-
ninger et al. 2012) and R1ρ (Weininger et al. 2014a) relaxa-
tion dispersion experiments have been acquired between 10
and 40 °C and 0.1 and 100 MPa. R1ρ relaxation dispersion
experiments have been recorded on-resonance. High pres-
sure experiments were performed using a commercial 3 mm
ceramic cell (Peterson and Wand 2005) (Daedalus Innova-
tions LLC), connected to a home-built pressure generator.
An aromatic 1H13C-TROSY-HSQC spectrum at − 5 °C and
200 MPa was recorded by utilizing pre-cooled air from an
external device. Spectra were processed with NMRPipe
(Delaglio et al. 1995) and analyzed with PINT (Ahlner et al.
2013).
Non‑averaged signals at low temperature and high
pressure
At − 5 °C and 200 MPa ring flips become so slow that the
individual sides of the ring could be observed in the spectra
(see Table 1). This enabled us to determine the 13C Δδ for
the two sides of Y3δ (2.11 ppm), Y3ε (1.40 ppm), F30δ
(5.39 ppm), F30ε (0.00 ppm) and F52ε (1.76 ppm). Previ-
ously, it was found that the shift difference Δδ is not chang-
ing with temperature (Weininger et al. 2014b). Therefore,
we used the derived Δδ as fixed parameters in the fitting of
the R1ρ relaxation dispersion experiments, when possible.
Derived 13C Δδ might be slightly too low, because the spec-
trum might still be affected by exchange. However they still
serve as a meaningful restraint of the fit. Furthermore, in
BPTI the potential problem can be estimated to less than 1%.
Data analysis
R1ρ relaxation dispersion data were fitted to the general
equation for symmetric exchange derived by Miloushev
and Palmer (2005) using fixed populations, p1 = p2 = 0.5, and
treating Δδ either as a free parameter (Δδdisp) or fixed at the
value (Δδspectra) measured from HSQC spectra under slow-
exchange conditions. Derived relaxation dispersion data at
different temperatures and pressures were fitted simulta-
neously with the restrictions: kflip (Thigh) > kflip (Tlow), R2,0
(Thigh) ≤ R2,0 (Tlow), and kflip (phigh) < kflip (plow).
1 3 Journal of Biomolecular NMR (2020) 74:183–191
185
Table 1 Effect of slow ring flips
on possible positions of Phe and
Tyr residues
Position
Δδ 1H (ppm)
Δδ 13C (ppm)
LB 1H
LB 13C
Rex 13C
Ring flip
RD method
Y3δ
Y3ε
F30δ
F30ε
Y33δ
Y33ε
Y45δ
Y45ε
F52δ
F52ε
0.40
0.50
0.84
0.56
2.11
1.40
5.39
0.00
0.00
1.76
Yes
Yes
Yes
Yes
No
No
Yes
No
No
Yes
Yes
Yes
No
No
No
Yes
No
Yes
Yes
Yes
Yes
No
No
No
Yes
Yes
Slow
Slow
Slow
Slow
Fast
Fast
Slow
Slow
Slow
Slow
1H/13C
1H/13C
1H/13C
1H
1H/13C
13C
Δδ: chemical shift difference between individual signals of both sides of the ring, detected at − 5 °C and
200 MPa. LB: significant line broadening at lower temperatures. Rex: exchange contribution of R2 at lower
temperatures. RD method: suitable relaxation dispersion method to study slow ring flips on this position
Activation barriers of the ring flips were determined by
non-linear regression of the flip rates, kflip = kex/2, on the
temperature T, using the Eyring equation. The Eyring equa-
tion was parameterized as
kflip =
kBT
h
(
)
× exp
ΔH‡ − TΔS‡
−
(
[
RT
)/
]
(1)
where kB and h are Boltzmann’s and Planck’s constants,
respectively, and ∆H‡ and ∆S‡ are the activation enthalpy
and activation entropy, respectively. Activation volumes ∆V‡
were determined from the pressure dependence of the flip
rates according to
𝜕 ln kflip
(
𝜕p )
= −
ΔV ‡
RT
(2)
Errors in the fitted parameters were estimated using
Monte–Carlo simulations (Press et al. 2002); the reported
errors correspond to one standard deviation.
Volume occupancies from aromatic rings in ground or
transition state were estimated considering them as rota-
tional ellipsoids with half-axes of 3.5 Å and 1.76 Å (Tsai
et al. 1999; Wagner 1980). The intersection volumes of two
rings in aromatic contact were then calculated using their
spatial dispositions from the crystal structure (1pgb.pdb).
Results
Protein GB1 consists of five symmetric aromatic residues
(Fig. 1), three Tyr (3, 33, 45) and two Phe (30, 52). Accord-
ing to their hydrophobicity, the Tyr are located closer to
the surface while the Phe are buried more in the interior.
The accessible surface area of the aromatic side-chains
determined by GETAREA (Fraczkiewicz and Braun 1998)
using 1pgb.pdb ranks the following: Y33 (70 Å2) ≫ Y45
Fig. 1 Three-dimensional structure of GB1 (1pgb.pdb) shown as rib-
bon presentation. Phe and Tyr side-chains are shown colored in stick
representation and are labeled accordingly
(48 Å2) ≫ Y3 (6 Å2) > F30 (4 Å2) ~ F52 (4 Å2). Y33 is not
involved in any particular stabilizing interactions. Y45 is
stacking with the π cloud of F52 from one side, F30 is stack-
ing with it from the other side. F52 itself is stacking with
the π cloud of Y3.
Identification of slow ring flips in GB1
Five averaged signals of the δ positions (δ*) and five aver-
aged signals of the ε positions (ε*) can be observed in the
aromatic 1H13C TROSY-HSQC spectra at higher tempera-
tures. At lower temperatures signals from Y3δ/ε, F30δ/ε,
Y45δ/ε and F52ε are becoming broadened (SI Fig. 1) and
significantly less intense (Fig. 2). In contrast, both signals of
Y33 are unaffected (other than intensity losses from slower
tumbling at lower temperature). A combination of low tem-
perature (− 5 ℃) and applied high pressure (200 MPa) is
slowing down the flip processes so far, that a splitting of
several signals (Y3δ, Y3ε, F30δ, F30ε and F52ε) could be
observed, representing both sides of the ring in different
chemical environments (Fig. 3). This effect was further
1 3 186
Journal of Biomolecular NMR (2020) 74:183–191
elaborated by aromatic 13C CPMG relaxation dispersion
experiments, showing an increase in 13C R2 (at lower tem-
peratures) for the exact same positions where an increase in
the 13C line width was observed (SI Figs. 1, 2). Furthermore,
the kinetic process that is causing the increase in R2 is too
fast to be quenched by CPMG experiments (SI Fig. 2). Taken
all these findings together (Table 1), it could be established
that all rings of the aromatic cluster are undergoing slow
ring flips which causes an effect on 13C R2, line shapes and
consequently signal intensity. The exception is Y33, which
does not show any signs that would point towards a slow ring
flip. Together with its high surface exposure we concluded
that Y33 is undergoing fast ring flips. Five positions are suit-
able for studying slow ring flips by 13C relaxation dispersion
methods over a range of temperature: Y3δ, Y45ε and F52ε,
and to a lesser degree Y3ε and F30δ. In F30ε and F52δ 13C
(and in case of F52 also 1H) is unaffected by ring flips, since
the respective Δδ between both sides of the ring is (close to)
zero. Y45δ is only detectable at 35 °C where ring flips are
too fast to be studied by 13C R1ρ experiments.
Quantification of slow ring flips in Y3, F30, Y45
and F52 by aromatic 13C R1ρ relaxation dispersion
experiments
Over the whole studied range of temperature (10 °C to
40 °C) at ambient pressure only averaged signals could be
observed, or signals have been broadened beyond detection.
The underlying ring flips causing the averaged signals are
too fast to be captured by aromatic 13C CPMG relaxation
dispersion experiments (Weininger et al. 2012) (SI Fig. 2),
in agreement with observations on BPTI (Weininger et al.
2014b). Therefore, aromatic 13C R1ρ relaxation dispersion
Fig. 2 Intensity of aromatic signals that can be affected by ring flips
(Phe and Tyr δ and ε). Y3 is shown in blue, F30 in magenta, Y33 in
grey, Y45 in cyan and F52 in red. Normalized relative intensities of δ
(a) and ε (b) are plotted against the temperature. Intensities of − 5 °C
and 200 MPa are plotted at − 15 °C, since going from 0.1 to 200 MPa
has roughly the same effect on the rate of ring flips than lowering the
temperature by 10 K. Here the intensities of the two individual sig-
nals (δ1 and δ2, or ε1 and ε2) are the same within the symbol size. In
all other cases, only averaged signals δ* and ε* (or no signals) could
be observed
Fig. 3 Region of a Tyr δ* (Y3 and Y33), b Tyr ε* (Y3 and Y33)
and c Phe ε* (F30 and F52) in the aromatic 1H13C-TROSY-HSQC of
GB1 at 30 °C (red), 25 °C (orange), 20 °C (yellow), 10 °C (green)
and ambient pressure. The spectrum at − 5 °C and 200 MPa is shown
in blue, where split signals (δ1 and δ2, or ε1 and ε2, respectively)
can be observed. Signals indicated as # are caused by sample impuri-
ties which can be detected at very high S/N experiments, which were
needed for the − 5 °C and 200 MPa condition, where the split signals
are still severely broadened
1 3 Journal of Biomolecular NMR (2020) 74:183–191
187
Fig. 4 Aromatic 13C L-TROSY-selected R1ρ relaxation dispersions
recorded on-resonance (tilt angle θ > 85°) at a static magnetic field-
strength of 14.1 T. Dispersion profiles for Y3δ at 25 °C (a), F30δ at
35 °C (b), Y45ε at 20 °C (c) and F52ε at 10 °C (d) are shown. Data
were fitted with fixed populations p1 = p2 = 0.5 and free (Y45) or fixed
chemical shift differences ∆δdisp derived from low temperature and
high pressure spectra. The resulting flip rates are: (12 ± 2) × 103 s−1,
(53 ± 4) × 103 s−1, (6 ± 2) × 103 s−1 and (4.8 ± 0.9) × 103 s−1, respec-
tively
experiments (Weininger et al. 2014a) have been applied.
Relaxation dispersion profiles could be recorded for Y3δ,
Y45ε and F52ε (Fig. 4a, c, d), which could be fitted to the
ring flip processes. F30δ at high temperatures displays
increased R1ρ values, which cannot be quenched sufficiently
(Fig. 4b). 13C R1ρ relaxation dispersion experiments are
allowing an accurate quantification of the ring flip processes
by a simultaneous and restricted fit at different tempera-
tures or pressures. Determined flip rates range from 4000 to
38,000 s−1 (75,000 s−1 for F30).
Aromatic 13C R1ρ relaxation dispersion profiles could be
recorded and quantified for Y3δ at 25 °C, 30 °C and 35 °C,
Y45ε at 20 °C, 25 °C and 30 °C, and F52ε at 10 °C, 15 °C
and 20 °C. For F30δ, only two temperatures (35 °C and
40 °C) could be used (SI Figs. 3–6). Plotting the derived
flip rates against temperature (Fig. 5) reveals similar flip
rates for Y3, F30 and Y45 (within margin of error), but sig-
nificantly faster flip rates for F52 (at a given temperature).
The latter are approximately three times higher, as can be
seen from the values at 25 °C, where rates of 12 × 103 s−1
and 11 × 103 s−1 can be measured for Y3 and Y45, respec-
tively, and a value of 37 × 103 s−1 can be extrapolated for
F52. Moreover, Y3 and F52, the two residues studied with
the highest accuracy, display the same temperature depend-
ence. Because of the higher flip rates, F52 had to be studied
at lower temperatures. This finding is somewhat surprising,
since F52 is among the most interior aromatic ring and the
central part of the cluster (Fig. 1).
Fig. 5 Temperature dependence of flip rates. kflip is plotted as a func-
tion of 1/T for F52 (red), Y3 (blue), Y45 (cyan) and F30 (magenta).
The fits are displayed as solid lines, while the uncertainties of the fits
are displayed as shaded areas in the appropriate colors. The data are
represented using a logarithmic y-axis to show the expected linearity,
but the fit was performed using non-linear regression of kflip on T
In order to further validate our results derived by
aromatic 13C R1ρ relaxation dispersion experiments, we
reanalyzed the dispersion profiles for Y3δ and F52ε with-
out the ∆δ fixed from information of the low tempera-
ture and high pressure spectrum. Derived ring flip rates
and activation enthalpies and entropies are the same
(within margin of error) with and without the additionally
fixed Δδ (SI Fig. 7). Furthermore, derived ∆δ of the fits
(2.17 ± 0.20 ppm and 1.84 ± 0.09 ppm, for Y3δ and F52ε,
respectively) are in excellent agreement with the ∆δ from
the spectrum (2.11 ppm and 1.76 ppm).
Y3, Y45 and F52 display similar activation
enthalpies
Ring flip rates at three temperatures for Y3, Y45 and F52
could be used to derive the activation enthalpy (∆H‡)
and activation entropy (∆S‡) for the individual flip pro-
cesses using Eq. 1 (Fig. 5). Activation enthalpies for Y3
(87 ± 14 kJ mol −1) and F52 (88 ± 11 kJ mol−1) are virtu-
ally identical. The activation enthalpy for Y45 appears to
be somewhat higher (129 ± 29 kJ mol−1), but could still be
interpreted to be the same as for Y3 and F52, considering
the significantly higher error. In fact, only the flip rate at
the highest temperature for Y45, which is the least well
covered in the relaxation dispersion profiles, is deviating
from Y3. Activation entropies are 126 ± 46 J mol−1 K−1,
137 ± 38 J mol−1 K−1 and 275 ± 102 J mol−1 K−1, for Y3,
F52 and Y45, respectively. It is not meaningful to derive
activation enthalpy and entropy for F30. However, it is safe
to assume that the activation enthalpy is not higher than for
Y3 and F52, as indicated by the determined flip rates.
1 3 188
Journal of Biomolecular NMR (2020) 74:183–191
Y3, Y45 and F52 display similar activation volumes
Ring flip rates for Y3, Y45 and F52 could also be recorded
and quantified at three different (0.1, 50 and 100 MPa)
hydrostatic pressures (SI Figs. 8–10). This allowed us to
determine the activation volumes (∆V‡) of the individual
ring flip processes using Eq. 2 (Fig. 6). Activation volumes
for Y3 (26 ± 5 mL mol −1) and F52 (29 ± 2 mL mol−1) are
virtually identical. The activation volume for Y45 appears
to be somewhat higher (51 ± 11 mL mol−1), but could still
be interpreted as the same as for Y3 and F52, considering
the errors. The findings for the activation volumes thereby
resemble the same general observation as for the activation
enthalpies.
Again, we validated our results by an analysis without
fixed ∆δ. Derived ring flip rates and activation enthalpies
and entropies are again the same (within margin of error)
(SI Fig. 7) and derived ∆δ of the fits (1.99 ± 0.31 ppm,
1.12 ± 0.09 ppm and 1.89 ± 0.07 ppm, for Y3δ, Y3ε and
F52ε, respectively) are in good agreement with the ∆δ from
the spectrum (2.11 ppm, 1.40 ppm and 1.76 ppm).
Discussion
rings (Phe and Tyr) could be observed, at least not at ambi-
ent pressure and temperatures above 0 °C. They are in the
fast exchange regime, in which only averaged signals for
both sides of symmetric aromatic rings can be observed.
By lowering the temperature these signals become gradu-
ally broadened and less intense until signals are completely
vanished. Since there are surprisingly very limited reports of
slow ring flips in the literature, this might be the case for the
vast majority of proteins. Ring flips are somewhat slow and
can cause a dramatic reduction of signal intensity close to
or in the intermediate exchange regime, but are not as slow
to reach the slow exchange regime. Thus by more thorough
temperature dependent studies of aromatic signals, many
more examples of slow ring flips can be expected, despite
not reaching the slow exchange regime. The aromatic 13C
R1ρ relaxation dispersion experiment is completely eligi-
ble to obtain correct ring flip rates (SI Fig. 7) and chemical
shift differences, even without information from the slow
exchange regime, and therefore allows the quantification of
ring flips in the fast to intermediate NMR exchange regime.
Furthermore, the determination of flip rates is robust to
small variations in the chemical shift difference. In addi-
tion, high-pressure NMR is an important tool that allows
additional changing of the ring flip conditions.
Ring flips in the fast to intermediate NMR exchange
regime
Individual nuclei in aromatic side chains are
affected differently
In contrast to previously reported cases of slow ring flips
investigated by NMR spectroscopy (Hattori et al. 2004;
Wagner et al. 1976, 1987; Weininger et al. 2014b), ring flips
in GB1 do not reach the slow exchange regime, in which
individual signals for both sides of symmetric aromatic
Fig. 6 Pressure dependence of flip rates. kflip is plotted as a function
of pressure for F52 (20 °C, red), Y3 (30 °C, blue) and Y45 (30 °C,
cyan). The fits are displayed as solid lines, while the uncertainties of
the fits are displayed as shaded areas in the appropriate colors. The
data are represented using a logarithmic y-axis to show the expected
linearity, but the fit was performed using non-linear regression of kflip
on p
Four rings in GB1 undergo slow ring flips. In theory, ring
flips could be studied on eight positions (4δ, 4ε). In prac-
tice the number of positions that can be used is significantly
reduced. While some positions display differences in 1H and
13C chemical shifts and therefore can be studied by 1H and
13C methods (Y3δε, F30δ, Y45δε), others just show differ-
ences in 1H (F30ε) or 13C (F52ε), or not at all (F52δ). Simi-
lar behaviour has been observed in BPTI and rapamycin- or
FK506-bound FKBP12 (SI Fig. 11) (Wagner et al. 1987;
Weininger et al. 2014b; Yang et al. 2015). There are also
examples of slow ring flips where both positions (δ and ε)
do not display shift differences in 1H and 13C and thus are
not accessible by relaxation dispersion methods (Weininger
et al. 2013). Since the time scale of exchange is in the limit
of R1ρ and not CPMG relaxation dispersion experiments, and
to date no 1H R1ρ relaxation dispersion methods in aromatic
side chains exist, F30ε is also not accessible. If the size of
the chemical shift difference (for both sides of the ring) is
large in 1H or 13C, the 1H–13C cross signal will be broad-
ened over a large range of temperature, which is the case for
F30δ and presumably Y45δ. Together with the upper rate
limit, that can be studied by 13C R1ρ relaxation dispersion
experiments, and limited protein stability at higher tempera-
tures, the final number of accessible positions is reduced
even more. In case of GB1, three positions can be studied
1 3 Journal of Biomolecular NMR (2020) 74:183–191
189
Fig. 7 Activation enthalpy of
ring flips for certain scenarios.
a Activation enthalpy of a ring
without stabilizing contacts. b
Activation enthalpy of a ring
with stabilizing contacts, in this
case a stacking ring (shown in
red). The stabilization of the
ground state is between 5 and
10 kJ mol−1 (Burley and Petsko
1989). c Activation enthalpy of
a ring with stabilizing contacts
of a stacking ring (shown in
red), both rings are undergoing
concerted ring flips
well (Y3δ, Y45ε and F52ε), while for two others the range
and accuracy is less (Y3ε and F30δ). Taken all together,
it requires a significant amount of screening conditions in
order to conduct a quantitative study of ring flips, if the slow
exchange regime cannot be reached.
Ring flips in an aromatic cluster
The key findings for ring flips in GB1 are the following. F52,
the central part of the aromatic cluster with three aromatic-
aromatic contacts, is flipping at a higher rate (at a given
temperature) than Y3, Y45 and F30, which flip with roughly
the same rate constants (Fig. 5). The activation enthalp-
ies for F52 (88 ± 11 kJ mol−1) and Y3 (87 ± 14 kJ mol−1)
are virtually the same, no activation enthalpy could be
determined for F30, but it is rather safe to conclude that
it is not larger, whereas Y45 (129 ± 29 kJ mol−1) might
also display the same activation enthalpy (within mar-
gin of error) or a slightly higher value. Activation entro-
pies (126 ± 46 J mol −1 K−1, 137 ± 38 J mol −1 K−1 and
275 ± 102 J mol−1 K−1) are somewhat higher than previously
reported ones, which range between 16 and 96 J mol−1 K−1
(Hattori et al. 2004; Weininger et al. 2014b; Yang et al.
2015). This reflects a higher loss in order in the transition
state compared to the ground state of the aromatic cluster of
GB1. This might be a characteristic of aromatic clusters in
general and potentially is reporting a more ordered ground
state. Activation volumes (Fig. 6) for F52 (29 ± 2 mL mol−1)
and Y3 (26 ± 5 mL mol −1) are virtually identical; again,
Y45 might display the same activation volume (within mar-
gin of error) or a slightly higher one (51 ± 11 mL mol−1).
Previously reported activation enthalpies (86, 83, 86 and
89 kJ mol−1 for BPTI Y23, Y35 and F45, and HPr Y6,
respectively (Hattori et al. 2004; Weininger et al. 2014b) and
activation volumes (51, 28 and 27 mL mol−1 for BPTI Y35
and F45 (Li et al. 1999) and HPr Y6 (Hattori et al. 2004)),
that have been derived on isolated aromatic rings, are very
similar. The only difference is in the activation entropy.
Given all this findings, a global breathing (transient
expansion) or unfolding of the aromatic cluster (which
would result in higher activation enthalpies and activation
volumes) can be ruled out. The ring flip process of aromatic
side chains in an aromatic cluster therefore seems to be a
local process, only involving a single ring or two rings in
a concerted flip as will be discussed below. In fact, derived
activation enthalpies and activation volumes are in very
good agreement with the flipping of a single ring in an inde-
pendent event. However, there are two reasons that might
question this. Firstly, the central ring of the cluster which
has a low accessible surface area, is flipping significantly
(around three times) faster (at a given temperature). This
finding is surprising, but clearly supported by the experi-
mental data. Isolated single ring flips do not give an expla-
nation for this. Secondly, one would have to assume that
aromatic interactions (Burley and Petsko 1985, 1989) do
not significantly contribute to ground state stabilization, not
even for F52, the central ring with three such interactions.
Aromatic stacking, however, is believed to provide between
5 and 10 kJ mol−1 (Burley and Petsko 1989) which would
roughly translate to an increased activation enthalpy of 10
to 20 kJ mol−1 (Fig. 7a, b). But it might simply be, that the
aromatic environment of F52 is more homogenous and bet-
ter suitable for dynamic processes like ring flips and this
somehow counters the enthalpic ground state stabilization
by aromatic stacking.
The other possibility would be that the aromatic ring in
aromatic contact with each of the others (F52) could flip in
a concerted event with each one of the other rings. Under
this assumption (F52 has the possibility to flip together
with each Y3, F30 and Y45 in individual events) the flip
1 3 190
Journal of Biomolecular NMR (2020) 74:183–191
rate of F52 would be the sum of all the other flip rates. In
case of a concerted flip, the transition state would not be
destabilized by an aromatic stacking but also stabilized,
resulting in unchanged activation enthalpies (Fig. 7c). Fur-
thermore, activation volumes could then be imagined to be
reduced, because of the rings providing partial space for
their partners to flip into, when rotating into the transition
state. For the spatial configuration in the hydrophobic core
of GB1, one obtains volume advantages of 1.6 mL mol−1
and 1.5 mL mol−1 conceded to F52 by Y45 and F30, respec-
tively. This could partially explain why the activation vol-
ume of F52 is not significantly higher than for the others,
despite being the central aromatic ring of the cluster. While
all these are good reasons to speculate about concerted ring
flips, it should be noted that none of the experiments per-
formed in this study is proof for it. In order to accurately
prove or disprove concerted ring flips, one has to perform
MD simulations or develop challenging multiple quantum
(of two rings) NMR exchange experiments (Lundström et al.
2005) through space.
Conclusions
Here we find that the ring in the center of an aromatic clus-
ter (F52), making aromatic stacking to three other aromatic
rings, is flipping with a faster rate than the other rings, whose
rates are comparable. Activation enthalpies and activation
volumes in the cluster, even in its center are not increased.
The only ring with a possible increase (Y45) is the ring in
the cluster located most on the protein surface. We speculate
that these findings are caused by correlated ring flips of F52
to at least two of its adjacent rings.
Acknowledgements Open access funding provided by Projekt DEAL.
This research was supported by the Deutsche Forschungsgemeinschaft
(Grant No. WE 5587/1-1).
Open Access This article is licensed under a Creative Commons Attri-
bution 4.0 International License, which permits use, sharing, adapta-
tion, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source,
provide a link to the Creative Commons licence, and indicate if changes
were made. The images or other third party material in this article are
included in the article’s Creative Commons licence, unless indicated
otherwise in a credit line to the material. If material is not included in
the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a
copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
References
Ahlner A, Carlsson M, Jonsson BH, Lundström P (2013) PINT: a soft-
ware for integration of peak volumes and extraction of relaxation
rates. J Biomol NMR 56:191–202. https ://doi.org/10.1007/s1085
8-013-9737-7
Burley SK, Petsko GA (1985) Aromatic-aromatic interaction: a mecha-
nism of protein structure stabilization. Science 229:23–28
Burley SK, Petsko GA (1989) Electrostatic interactions in aromatic
oligopeptides contribute to protein stability. Trends Biotechnol
7:354–359
Campbell ID, Dobson CM, Williams RJP (1975) proton magnetic-
resonance studies of tyrosine residues of hen lysozyme-assign-
ment and detection of conformational mobility. Proc Royal Soc B
189:503–509. https ://doi.org/10.1098/rspb.1975.0070
Delaglio F, Grzesiek S, Vuister GW, Zhu G, Pfeifer J, Bax A (1995)
Nmrpipe—a multidimensional spectral processing system based
on unix pipes. J Biomol NMR 6:277–293
Dreydoppel M, Becker P, Raum HN, Gröger S, Balbach J, Weininger
U (2018) Equilibrium and kinetic unfolding of GB1: stabiliza-
tion of the native state by pressure. J Phys Chem B 122:8846–
8852. https ://doi.org/10.1021/acs.jpcb.8b068 88
Fraczkiewicz R, Braun W (1998) Exact and efficient analytical
calculation of the accessible surface areas and their gradi-
ents for macromolecules. J Comput Chem 19:319–333. https
://doi.org/10.1002/(Sici)1096-987x(19980 2)19:3%3c319
:Aid-Jcc6%3e3.3.Co;2-3
Hattori M, Li H, Yamada H, Akasaka K, Hengstenberg W, Gronwald
W, Kalbitzer HR (2004) Infrequent cavity-forming fluctuations
in HPr from Staphylococcus carnosus revealed by pressure- and
temperature-dependent tyrosine ring flips. Protein Sci 13:3104–
3114. https ://doi.org/10.1110/ps.04877 104
Hull WE, Sykes BD (1975) Fluorotyrosine alkaline-phosphatase—
internal mobility of individual tyrosines and role of chemical-
shift anisotropy as a F-19 nuclear spin relaxation mechanism in
proteins. J Mol Biol 98:121–153. https ://doi.org/10.1016/S0022
-2836(75)80105 -7
Kasinath V, Fu YN, Sharp KA, Wand AJ (2015) A sharp thermal
transition of fast aromatic-ring dynamics in ubiquitin. Angew
Chem Int Ed 54:102–107
Li H, Yamada H, Akasaka K (1999) Effect of pressure on the tertiary
structure and dynamics of folded basic pancreatic trypsin inhibi-
tor. Biophys J 77:2801–2812
Lindman S, Xue WF, Szczepankiewicz O, Bauer MC, Nilsson H,
Linse S (2006) Salting the charged surface: pH and salt depend-
ence of protein G B1 stability. Biophys J 90:2911–2921. https
://doi.org/10.1529/bioph ysj.105.07105 0
Lundström P, Mulder FAA, Akke M (2005) Correlated dynamics of
consecutive residues reveal transient and cooperative unfolding
of secondary structure in proteins. Proc Natl Acad Sci USA
102:16984–16989. https ://doi.org/10.1073/pnas.05043 61102
Lundström P et al (2007) Fractional C-13 enrichment of isolated car-
bons using [1-C-13]- or [2-C-13]-glucose facilitates the accu-
rate measurement of dynamics at backbone C-alpha and side-
chain methyl positions in proteins. J Biomol NMR 38:199–212
Mahadevi AS, Sastry GN (2013) Cation-pi interaction: its role and
relevance in chemistry, biology, and material science. Chem
Rev 113:2100–2138
Miloushev VZ, Palmer AG (2005) R(1p) relaxation for two-site
chemical exchange: general approximations and some exact
solutions. J Magn Reson 177:221–227. https ://doi.org/10.1016/j.
jmr.2005.07.023
Nall BT, Zuniga EH (1990) Rates and energetics of tyrosine ring
flips in yeast iso-2-cytochrome-c. Biochemistry 29:7576–7584
1 3 Journal of Biomolecular NMR (2020) 74:183–191
191
Pace CN, Scholtz JM, Grimsley GR (2014) Forces stabilizing pro-
teins. FEBS Lett 588:2177–2184. https ://doi.org/10.1016/j.febsl
et.2014.05.006
Peterson RW, Wand AJ (2005) Self contained high pressure cell,
apparatus and procedure for the preparation of encapsulated
proteins dissolved in low viscosity fluids for NMR spectroscopy.
Rev Sci Instrum 76:1–7. https ://doi.org/10.1063/1.20380 87
Press WH, Teukolsky SA, Vetterling WT, Flannery BP (2002)
Numerical recipes in C++: the art of scientific computing, 2nd
edn. Cambridge University Press, Cambridge
Rose GD, Wolfenden R (1993) Hydrogen-bonding, hydrophobicity,
packing, and protein-folding. Annu Rev Biophys Biomol Struct
22:381–415. https ://doi.org/10.1146/annur ev.bb.22.06019
3.00212 1
Shaw DE et al (2010) Atomic-level characterization of the structural
dynamics of proteins. Science 330:341–346
Teilum K, Brath U, Lundström P, Akke M (2006) Biosynthetic C-13
labeling of aromatic side chains in proteins for NMR relaxation
measurements. J Am Chem Soc 128:2506–2507
Tsai J, Taylor R, Chothia C, Gerstein M (1999) The packing density
in proteins: standard radii and volumes. J Mol Biol 290:253–266.
https ://doi.org/10.1006/jmbi.1999.2829
Valley CC, Cembran A, Perlmutter JD, Lewis AK, Labello NP,
Gao J, Sachs JN (2012) The methionine-aromatic motif plays
a unique role in stabilizing protein structure. J Biol Chem
287:34979–34991
Wagner G (1980) Activation volumes for the rotational motion of
interior aromatic rings in globular-proteins determined by high-
resolution h-1-nmr at variable pressure. FEBS Lett 112:280–284
Wagner G, Demarco A, Wüthrich K (1976) Dynamics of aromatic
amino-acid residues in globular conformation of basic pancreatic
trypsin-inhibitor (Bpti) 1 H-1 NMR-studies. Biophys Struct Mech
2:139–158
Wagner G, Bruhwiler D, Wüthrich K (1987) Reinvestigation of the
aromatic side-chains in the basic pancreatic trypsin-inhibitor by
heteronuclear two-dimensional nuclear-magnetic-resonance. J
Mol Biol 196:227–231
Weininger U, Respondek M, Akke M (2012) Conformational exchange
of aromatic side chains characterized by L-optimized TROSY-
selected C-13 CPMG relaxation dispersion. J Biomol NMR
54:9–14
Weininger U, Respondek M, Löw C, Akke M (2013) Slow aromatic
ring flips detected despite near-degenerate NMR frequencies of
the exchanging nuclei. J Phys Chem B 117:9241–9247
Weininger U, Brath U, Modig K, Teilum K, Akke M (2014a) Off-res-
onance rotating-frame relaxation dispersion experiment for C-13
in aromatic side chains using L-optimized TROSY-selection. J
Biomol NMR 59:23–29
Weininger U, Modig K, Akke M (2014b) Ring flips revisited: C-13
relaxation dispersion measurements of aromatic side chain
dynamics and activation barriers in basic pancreatic trypsin
inhibitor. Biochemistry 53:4519–4525
Wüthrich K, Wagner G (1975) Nmr investigations of dynamics
of aromatic amino-acid residues in basic pancreatic trypsin-
inhibitor. FEBS Lett 50:265–268. https ://doi.org/10.1016/0014-
5793(75)80504 -7
Yang CJ, Takeda M, Terauchi T, Jee J, Kainosho M (2015) Differ-
ential large-amplitude breathing motions in the interface of
FKBP12-drug complexes. Biochemistry 54:6983–6995. https ://
doi.org/10.1021/acs.bioch em.5b008 20
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
1 3
| null |
10.1038_s41598-021-00119-7.pdf
|
Data availability
The data presented in this study are available on request from the corresponding authors.
|
Data availability The data presented in this study are available on request from the corresponding authors. Received: 7 July 2021; Accepted: 6 October 2021
|
OPEN
Elevated serum levels
of methylglyoxal are associated
with impaired liver function
in patients with liver cirrhosis
Maurice Michel1*, Cornelius Hess2, Leonard Kaps1, Wolfgang M. Kremer1, Max Hilscher1,
Peter R. Galle1, Markus Moehler1, Jörn M. Schattenberg1, Marcus‑Alexander Wörns1,
Christian Labenz1 & Michael Nagel1*
Methylglyoxal (MGO) is a highly reactive dicarbonyl species that forms advanced glycation end
products (AGEs). The binding of these AGEs to their receptor (RAGE) causes and sustains severe
inflammation. Systemic inflammation is postulated to be a major driver in the progression of liver
cirrhosis. However, the role of circulating MGO levels in liver cirrhosis remains unknown. In this study,
we investigated the serum levels of two dicarbonyl species, MGO and glyoxal (GO) using tandem
mass spectrometry (HPLC–MS/MS) and evaluated their association with disease severity. A total of
51 inpatients and outpatients with liver cirrhosis of mixed etiology and different disease stages were
included. Elevated MGO levels were seen in an advanced stage of liver cirrhosis (p < 0.001). High MGO
levels remained independently associated with impaired liver function, as assessed by the model
for end‑stage liver disease (MELD) (β = 0.448, p = 0.002) and acute decompensation (AD) (β = 0.345,
p = 0.005) scores. Furthermore, MGO was positively correlated with markers of systemic inflammation
(IL‑6, p = 0.004) and the development of ascites (p = 0.013). In contrast, no changes were seen in GO
serum levels. Circulating levels of MGO are elevated in advanced stages of liver cirrhosis and are
associated with impaired liver function and liver‑related parameters.
Abbreviations
AD
AGE
ALDH
ATP
BMI
CC
CRP
DC
Glo-I
Glo-II
GO
GSH
HPLC–MS/MS
HVPG
IL-6
INR
MELD
MELD-Na
MGO
Acute decompensation
Advanced glycation end product
Aldehyde dehydrogenase
Adenosine triphosphate
Body mass index
Compensated cirrhosis
C-reactive protein
Decompensated cirrhosis
Glyoxalase-I
Glyoxalase-II
Glyoxal
Glutathione
High-performance liquid chromatography tandem mass spectrometry
Hepatic venous pressure gradient
Interleukin-6
International normalized ratio
Model for end-stage liver disease
Model for end-stage liver disease-sodium
Methylglyoxal
1I. Department of Medicine, University Medical Center of the Johannes Gutenberg University Mainz, Mainz,
55131 Mainz, Germany. 2Institute of Forensic Medicine, Forensic Toxicology, University Medical Center of the
Johannes Gutenberg University Mainz, 55131 Mainz, Germany. *email: [email protected];
[email protected]
Scientific Reports | (2021) 11:20506
| https://doi.org/10.1038/s41598-021-00119-7
1
Vol.:(0123456789)www.nature.com/scientificreports NAFLD
NF-κB
OHE
RAGE
RCS
ROS
Nonalcoholic fatty liver disease
Nuclear factor ‘kappa-light-chain-enhancer’ of activated B-cells
Overt hepatic encephalopathy
Receptor for advanced glycation end products
Reactive carbonyl species
Reactive oxygen species
Liver cirrhosis is one of the leading liver diseases worldwide and accounts for more than one million deaths
every year 1. It marks chronic and progressive inflammation of the liver with increasing scarring of liver tissue.
This leads to a loss of function, the development of liver-related complications and a significantly higher risk of
developing liver cancer 2. Liver cirrhosis is roughly classified as either compensated or decompensated. Although
the liver is already scarred in the compensated state, it still retains its basic functions to some extent, and patients
are often asymptomatic 3. In contrast, in the decompensated state, three major complications, namely, ascites,
gastrointestinal hemorrhage and hepatic encephalopathy, impair quality of life and overall survival of patients 4.
More recently, a systemic inflammatory response has been hypothesized to be a key driver of disease progression
and the development of decompensation, even in the absence of bacterial infections 5,6.
Methylglyoxal (MGO) is a highly cytotoxic and reactive dicarbonyl, also termed reactive carbonyl species
(RCS), which leads to so-called dicarbonyl stress. It is a potent glycating agent and a major precursor that
facilitates the formation of advanced glycation end products (AGEs) and reactive oxygen species (ROS) through
mitochondrial dysfunction 7. As a consequence, binding of these so-called MGO-derived AGEs to their receptor
RAGE induces and sustains an inflammatory response through the activation of the transcription factor NF-κB
8. Although glyoxal (GO) is also considered an RCS, it is far less reactive than MGO 9. Only high concentrations
of GO for a prolonged time result in the formation of AGEs 10,11. Because MGO is mainly formed as a byproduct
during glycolysis and hyperglycemia is associated with higher MGO levels, it is thought to be a significant media-
tor in the development and progression of diabetes 12. If present in high concentrations, MGO can modify and
impair the function of albumin 13. Moreover, several other chronic inflammatory conditions, such as rheumatoid
arthritis or chronic kidney disease, have shown elevated MGO blood levels 14,15.
Detoxification of MGO is mainly achieved by means of the glutathione (GSH)-dependent glyoxalase system
constituting glyoxalase-I (Glo-I) and -II (Glo-II). Either elevated energy demands, as seen in inflammation and
cancer cells, or impaired detoxification can cause the accumulation of MGO to a toxic threshold 8,16. Preliminary
findings have indicated a decline in the expression of Glo-I and a subsequent increase in MGO in an animal
model of liver cirrhosis 17. Earlier studies have shown increased levels of AGEs—but not its precursor MGO—in
patients with liver cirrhosis 18, with an amelioration after liver transplantation 19,20. Proteomic profiling identified
decreased expression of Glo-I in hepatocytes in a murine model of nonalcoholic fatty liver disease (NAFLD)
with higher levels of MGO-derived AGEs in patients 21.
Currently, the role of circulating MGO and GO levels in patients with liver cirrhosis remains unknown.
Therefore, the aim of this study was to investigate MGO and GO serum levels in patients with varying stages of
liver cirrhosis and to elucidate their association with disease severity.
Results
Baseline characteristics. A total of 51 patients with liver cirrhosis were prospectively enrolled. The major-
ity of patients were male (n = 30, 58.8%), and the median age was 60 years. In terms of Child–Pugh score, 51%
(n = 26) of the liver cirrhosis patients scored as A, 31.4% (n = 16) scored as B, and 17.6% (n = 9) scored as C. The
cohort was then divided into either compensated (CC, n = 26) or decompensated cirrhosis (DC, n = 25). The
median model for end-stage liver disease (MELD) score was 13 (IQR 10; 18), and the median acute decompensa-
tion (AD) score was 50 (IQR 45; 53) in the entire cohort. Higher scores were seen in patients with DC (p < 0.001).
In line with these findings, INR (p < 0.001) and total bilirubin (p < 0.001) were also higher in DC. No significant
difference in creatinine levels were detected between CC and DC. Interleukin-6 (IL-6) (p < 0.001), a marker of
systemic inflammation, as well as other inflammatory markers, was significantly elevated in patients with DC.
The baseline characteristics and a comparison of patients between CC and DC are summarized in Table 1.
In the entire cohort, the median MGO level was 37.1 ng/mL (IQR 18; 55.4). No difference in MGO levels
between males (m) and females (f) was observed (m: 42.51 ± 24.50 vs. f: 42.06 ± 28.47, p = 0.72 (Supplementary
Fig. S1a). Patients with a comorbidity of type 2 diabetes (n = 18, 35.3%) did not show higher levels of MGO
than patients without diabetes (no diabetes: 46.35 ± 28.24 vs. diabetes: 34.21 ± 27.94, p = 0.085 (Supplementary
Fig. S1b). The blood sugar levels assessed during blood withdrawal did not correlate with the MGO levels
(r = 0.051, p = 0.722) or GO levels (r = 0.082, p = 0.566). The MGO levels were higher in patients with alcohol-
related liver cirrhosis than in patients with NAFLD (46.81 ± 29.06 vs. 28.95 ± 32.44, p = 0.008) (Supplementary
Fig. S1c). The median GO level was 52 ng/mL (IQR 32.9; 59.1). The circulating GO levels in comparison with
several patient characteristics are displayed in Supplementary Fig. S2.
Serum levels of methylglyoxal are elevated with increasing severity of liver disease. Patients
with decompensated cirrhosis (DC) had significantly higher levels of MGO than patients with compensated
cirrhosis (CC) (CC: 26.68 ± 15.62 vs. DC: 54.99 ± 31.95, p < 0.001) (Fig. 1a). In this context, Child–Pugh C
(89.44 ± 27.22, p < 0.001) liver cirrhosis showed the highest levels compared with Child–Pugh A (26.68 ± 15.62)
and Child–Pugh B (38.78 ± 18.12) liver cirrhosis (Fig. 1b). Although the MGO levels were higher in Child–Pugh
B than in Child–Pugh A liver cirrhosis, no significant difference was seen (p = 0.054) (Fig. 1b). Concordantly,
higher MGO levels were seen in patients with a MELD score ≥ 15 than in patients with a MELD score < 15
(< 15: 29.03 ± 16.76 vs. ≥ 15: 55.19 ± 32.55, p < 0.001) (Fig. 1c). According to the AD score, patients with an AD
Scientific Reports | (2021) 11:20506 |
https://doi.org/10.1038/s41598-021-00119-7
2
Vol:.(1234567890)www.nature.com/scientificreports/ Total cohort (n = 51)
Compensated cirrhosis (CC)
(n = 26)
Decompensated cirrhosis
(DC) (n = 25)
n (%) or median (25th; 75th)
n (%) or median (25th; 75th)
n (%) or median (25th; 75th)
p value
Variables
Age (years)
BMI (kg/m2)
Male sex
Type 2 diabetes
Etiology of liver cirrhosis
Alcohol
NAFLD
Hepatitis C
Others
60 (53; 66)
28 (24; 31)
30 (58.8)
18 (35.3)
33 (64.7)
11 (21.6)
2 (3.9)
7 (13.7)
Biochemical parameters
Sodium (mmol/L)
137 (135; 139)
AST (U/L)
ALT (U/L)
55 (41; 77)
23 (17; 37)
Total bilirubin (mg/dL)
1.7 (1.2; 3.8)
Creatinine (mg/dL)
0.82 (0.68; 1.2)
INR
Albumin (g/dL)
CRP (mg/L)
IL-6 (pg/mL)
Leukocytes (/nL)
Hemoglobin (g/dL)
Thrombocytes (/nL)
MELD score
MELD-Na
Child–Pugh score
AD score
Clinical parameters
HVPGa (mmHg)
1.3 (1.2; 1.7)
30 (23; 35)
6.1 (3.9; 16)
18 (9; 31)
5.1 (3.9; 7.6)
11 (9.8; 13.1)
107 (78; 147)
13 (10; 18)
16 (12; 22)
6 (5; 9)
50 (45; 53)
60 (51; 66)
29.5 (24.4; 33.9)
59 (54; 65)
25.8 (23; 30.3)
15 (57.7)
11 (42.3)
10 (38.5)
9 (34.6)
0
7 (26.9)
138 (136; 139)
43 (36.8; 60)
22.5 (17; 37)
1.35 (0.98; 1.5)
0.83 (0.65; 1.1)
1.2 (1.2; 1.3)
33 (30.8; 37)
5.2 (2.7; 7.2)
10 (6; 18.3)
4.78 (3.7; 6.1)
12.8 (10.2; 14.4)
126 (84.8; 161.8)
10 (8.8; 12.3)
12 (10; 14.3)
5 (5; 6)
46 (40.7; 50)
15 (60)
7 (28)
24 (92.3)
2 (3.9)
2 (3.9)
0
137 (131; 138.5)
70 (53; 113.5)
27 (17; 39.5)
3.8 (2.3; 8.9)
0.82 (0.72; 1.3)
1.7 (1.4; 2.0)
26 (20.5; 29)
9.9 (5.8; 22)
30 (18.5; 55.5)
6.79 (4.3; 9.4)
10.7 (9.3; 12.1)
95 (72; 144.5)
18 (14.5; 24)
22 (17.5; 26)
9 (7. doi: 10.5)
52 (49.5; 56.5)
0.990
0.110
0.867
0.285
< 0.001
0.124
0.001
0.720
< 0.001
0.429
< 0.001
0.124
0.001
< 0.001
0.049
0.016
0.127
< 0.001
< 0.001
< 0.001
< 0.001
0.016
0.037
0.006
< 0.001
16.5 (11; 20.3)
14.5 (9; 18.8)
18 (14.8; 22.5)
History of OHE
7 (13.7)
Ascites at study inclusion
15 (29.4)
History of ascites
23 (45.1)
1 (3.8)
3 (11.5)b
5 (19.2)
6 (24)
12 (48)
18 (69.2)
Table 1. Clinical characteristics, demographic data, and differences between compensated and decompensated
liver cirrhosis. Data are expressed as numbers, medians, percentages (%) or interquartile ranges (IQR 25th;
75th). AD acute decompensation, ALT alanine-aminotransaminase, AST aspartate-aminotransaminase, BMI
body mass index, CRP C-reactive protein, OHE overt hepatic encephalopathy, INR international normalized
ratio, HVPG hepatic venous pressure gradient, MELD model for end-stage liver disease, NAFLD nonalcoholic
fatty liver disease. p values refer to the comparison between compensated (CC) and decompensated (DC) liver
cirrhosis. Boldface indicates statistical significance. A p value < 0.05 was considered significant. a Measured
in 46 patients. b Only a small volume of ascites was detected on abdominal ultrasound and not accessible for
paracentesis.
score ≥ 50 showed elevated levels of MGO compared with patients with an AD score < 50 (< 50: 29.92 ± 15.75
vs. ≥ 50: 53.74 ± 33.06, p = 0.0079) (Fig. 1d).
Elevated methylglyoxal levels are associated with liver‑related complications. Next, liver-
related complications were analyzed with regard to MGO serum levels. Patients who presented with ascites had
significantly higher MGO levels (no ascites: 36.44 ± 24.62 vs. ascites: 54.08 ± 28.18, p = 0.008) (Fig. 2a). Patients
with hepatic encephalopathy (Fig. 2b) or a history of gastroesophageal varices (Fig. 2c) did not show higher
MGO levels.
Methylglyoxal is associated with impaired liver function.
In a univariable analysis, higher levels
of MGO were associated with ascites at study inclusion and a history of ascites as well as liver-related scores
(Child–Pugh, MELD (Fig. 3a), MELD-Na and AD scores (Fig. 3b)). Furthermore, markers of liver dysfunction
(albumin, total bilirubin, and INR) and inflammation (IL-6) were also associated with elevated levels of MGO. In
a multivariable linear regression analysis, hepatic dysfunction scores (MELD: standardized β coefficient = 0.448,
95% CI 5.13, 20.3, p = 0.002; AD score: standardized β coefficient = 0.345, 95% CI 0.44, 2.28, p = 0.005), high
blood levels of total bilirubin (standardized β coefficient = 0.401, 95% CI 3.71, 19.1, p = 0.005) and low blood
Scientific Reports | (2021) 11:20506 |
https://doi.org/10.1038/s41598-021-00119-7
3
Vol.:(0123456789)www.nature.com/scientificreports/ a
150
)
l
/
m
g
n
(
O
G
M
100
50
0
c
150
)
l
/
m
g
n
(
O
G
M
100
50
0
***
CC
DC
***
MELD <15
MELD > 15
150
b
)
l
/
m
g
n
(
O
G
M
100
50
0
150
d
)
l
/
m
g
n
(
O
G
M
100
50
0
***
***
ns
A
B
C
Child-Pugh score
**
AD score < 50
AD score > 50
Figure 1. Whisker boxplots showing median (IQR 10th; 90th) MGO serum levels in patients with different
liver disease severities. (a) The MGO levels were higher in patients with decompensated cirrhosis (DC). (b)
Patients with Child–Pugh C showed higher levels than patients with Child–Pugh A and B. (c) MELD ≥ 15
showed increased MGO serum levels. (d) Patients with an acute decompensation (AD) score ≥ 50 presented
with higher MGO levels. Differences between two groups were analyzed using the Mann–Whitney U test. More
than two groups were analyzed by the Kruskal–Wallis test. The dots refer to values beyond the range of the 10th
and 90th percentiles. *p < 0.05; **p < 0.01; ***p < 0.001; ns not significant.
values of albumin (a: standardized β coefficient = − 0.312, 95% CI − 16.6, − 1.18, p = 0.025; b: standardized β
coefficient = − 0.281, 95% CI − 15.6, − 0.37, p = 0.040; c: standardized β coefficient = − 0.450, 95% CI − 19.4, − 6.2,
p < 0.001) remained independently associated with higher MGO levels (a: R2 = 0.396; b: R2 = 0.422; R2 = 0.397)
(Table 2).
Serum levels of glyoxal remain unaltered with increasing severity of liver disease.
In contrast
to MGO, serum levels of glyoxal (GO) were not altered between different stages of liver cirrhosis. The levels
of GO were not higher in patients with DC (CC: 46.51 ± 13.69 vs. DC: 56.88 ± 28.17, p = 0.287) (Fig. 4a) or a
Child–Pugh C cirrhosis (A: 46.51 ± 13.69 vs. B: 56.02 ± 31.37 vs. C: 61.01 ± 23.17, p = 0.262) (Fig. 4b). A MELD
score ≥ 15 (< 15: 49.19 ± 16.62 vs. ≥ 15: 55.32 ± 29.34, p = 0.721) (Fig. 4c) or an AD score ≥ 50 (< 50: 51.27 ± 23.96
vs. ≥ 50: 51.59 ± 21.28, p = 0.692) did not show elevated levels of GO (Fig. 4d).
Circulating levels of glyoxal are not associated with markers of liver dysfunction.
In the uni-
variable analysis, serum levels of GO were not associated with markers of impaired hepatic dysfunction (MELD:
r = 0.151, p = 0.290 (Fig. 5a); AD score: r = 0.073, p = 0.611 (Fig. 5b)), inflammatory markers or other clinical
characteristics (Table 3).
Discussion
The present study shows the association of elevated MGO serum levels with impaired liver function in patients
with liver cirrhosis. In this context, an increasing severity of liver cirrhosis, as indicated by Child–Pugh, MELD
and AD scores was associated with significantly higher levels of MGO. Patients who presented with liver-related
complications, in particular the development of ascites, also had higher MGO serum levels. In contrast to these
findings, circulating GO levels were not significantly altered in these patients.
Systemic inflammation is postulated to be the major driver in the development and progression of liver
cirrhosis 5,6. The activation of inflammatory cells leads to the production of proinflammatory cytokines (IL-1,
IL-6, IL-8 and TNFα), acute phase reactants (CRP) and the induction of intracellular signaling cascades (NF-κB,
RAGE) 22. In this regard, increasing levels of IL-6 have been associated with higher mortality or the development
of complications such as OHE in liver cirrhosis 23,24. In line with previous findings, markers of inflammation
(CRP, leukocytes and IL-6) were significantly elevated in patients presenting with decompensated cirrhosis in
this cohort. Although no association with leukocytes and CRP was observed, MGO accumulation was associ-
ated with upregulated IL-6 concentrations. An interaction of MGO and AGE-RAGE as well as NF-κB is likely
to enhance the production of IL-6 from T cells and macrophages. Moreover, the binding of MGO to proteins,
Scientific Reports | (2021) 11:20506 |
https://doi.org/10.1038/s41598-021-00119-7
4
Vol:.(1234567890)www.nature.com/scientificreports/
ns
yes
no
Hepatic Encephalopathy
150
b
)
l
/
m
g
n
(
O
G
M
100
50
0
ns
a
150
**
l
/
m
g
n
O
G
M
100
50
0
yes
no
Ascites
c
150
)
l
/
m
g
n
(
O
G
M
100
50
0
yes
no
Gastroesophageal Varices
Figure 2. Whisker boxplots showing median (IQR 10th; 90th) MGO serum levels with respect to the presence
of liver-related complications. ‘Yes’ or ‘no’ indicates whether the characteristic was present at study inclusion. (a)
Patients with ascites at study entry showed higher MGO levels. The MGO levels were not elevated in patients
presenting with hepatic encephalopathy (b) or gastroesophageal varices (c). Differences between two groups
were analyzed using the Mann–Whitney U test. The dots refer to values beyond the range of the 10th and 90th
percentiles. *p < 0.05; **p < 0.01; ***p < 0.001; ns not significant.
a
40
e
r
o
c
s
D
L
E
M
30
20
10
0
0
80
b
e
r
o
c
s
D
A
60
40
20
0
0
50
100
150
MGO (ng/ml)
50
100
150
MGO (ng/ml)
Figure 3. Circulating methylglyoxal serum levels correlate with the MELD score (effect size r = 0.529) (a) and
AD score (effect size r = 0.373) (b).
nucleic acids and lipids leads to protein dysfunction and exerts mutagenesis and cell death 7. Therefore, higher
levels of MGO may sustain an inflammatory response with worsening liver cirrhosis. Consequently, this is further
aggravated by cumulative immune dysfunction in more advanced stages of liver cirrhosis, leading to a higher
risk of bacterial infections and mortality 25. Recently, Baumann et al. discovered the inhibitory effects of MGO
accumulation on the effector functions of immune cells 26. In this context, MGO may distract anti-inflammatory
signaling in favor of a proinflammatory environment.
Detoxification of MGO is mainly dependent on Glo-I, which converts MGO with the aid of GSH into unre-
active lactate 27. Therefore, depletion of GSH, as seen in many inflammatory reactions and during oxidative
stress, is likely to impair the enzymatic activity of hepatic Glo-I 28. Preliminary findings in an animal model
revealed decreased expression of Glo-I and an increase in MGO levels with worsening liver cirrhosis 17. Lower
expression of Glo-I in the liver was also seen in a murine model of NAFLD 21. The use of a pharmacological
inducer of Glo-I activity led to lower MGO levels in overweight and obese patients 29. Although analyzing Glo-I
expression and enzymatic activity in human tissues was not the focus of this study, decreased detoxification
of MGO in the liver is likely the result of impaired Glo-I function and may be an explanation for our current
Scientific Reports | (2021) 11:20506 |
https://doi.org/10.1038/s41598-021-00119-7
5
Vol.:(0123456789)www.nature.com/scientificreports/
Methylglyoxal (ng/mL)
Univariable
analysis
Multivariable
analysisa
Multivariable
analysisb
Multivariable
analysisc
Variable
Glyoxal
Age
Sex
BMI
Type 2 Diabetes
Metformin
Ascites at study
inclusion
r
0.238
− 0.189
− 0.079
− 0.151
0.245
0.282
− 0.347
History of ascites
− 0.541
History of OHE
− 0.194
Child–Pugh score
MELD
MELD-Na
AD score
Sodium
Albumin
INR
Total bilirubin
IL-6
CRP
Leukocytes
Creatinine
0.633
0.529
0.586
0.373
− 0.164
− 0.446
0.368
0.474
0.399
0.278
0.212
0.273
Thrombocytes
− 0.178
HVPG*
0.215
p
β (95% CI)
p
β (95% CI)
p
β (95% CI)
p
0.093
0.185
0.584
0.291
0.083
0.045
0.013
< 0.001
0.173
< 0.001
< 0.001
< 0.001
0.007
0.250
0.001
0.009
0.448 (5.13, 20.3) 0.002
0.345 (0.44, 2.28)
0.005
− 0.312
(− 16.6, − 1.18)
0.025
− 0.281
(− 15.6, − 0.37)
0.040
− 0.450
(− 19.4, − 6.2)
< 0.001
0.001
0.401 (3.71, 19.1) 0.005
0.004
0.050
0.135
0.055
0.211
0.151
Table 2. Univariable and multivariable analyses for predictors of higher methylglyoxal levels in patients with
liver cirrhosis. AD acute decompensation, BMI body mass index, CRP C-reactive protein, OHE overt hepatic
encephalopathy, IL-6 interleukin-6, INR international normalized ratio, HVPG hepatic venous pressure
gradient, MELD model for end-stage liver disease, MELD-Na model for end-stage liver disease-sodium.
Univariable and multivariable analyses of the data are shown. With all factors showing a p value < 0.05 and
the clinical parameters age, sex and type 2 diabetes, a multivariable linear regression model was built. Beta (β)
and 95% confidence intervals (CIs) show standardized values. Boldface indicates significance. A p value < 0.05
was considered significant. *Measured in 46 patients. a Linear regression analysis: Age, sex, type 2 diabetes,
metformin, ascites at study inclusion, history of ascites, albumin, INR, bilirubin, IL-6. b Linear regression
analysis: Age, sex, type 2 diabetes, metformin, ascites at study inclusion, history of ascites, MELD, albumin,
IL-6. c Linear regression analysis: Sex, type 2 diabetes, metformin, ascites at study inclusion, history of ascites,
AD score, albumin, IL-6.
findings. Supporting this assumption, GO—which is structurally similar to MGO—is mostly detoxified through
aldehyde dehydrogenase (ALDH) and not Glo-I 30,31. Interestingly, no differences in GO blood concentrations
were seen in this study. Moreover, blood levels of GO-derived AGEs remained unchanged across the liver in
liver cirrhosis, emphasizing the minor role of hepatic Glo-I in GO detoxification 18. Nevertheless, GO-derived
AGE accumulation has also been observed in liver cirrhosis, which may also be due to impaired renal clearance
and kidney dysfunction 32. In this cohort, most patients presented with preserved renal function, as indicated
by normal blood levels of creatinine. Thus, MGO accumulation in patients with liver cirrhosis may be a result
of lower MGO clearance in the liver.
Markers of hepatic dysfunction were associated with elevated MGO serum levels in this study. A decrease
in liver function is reflected by laboratory values and clinical signs that are incorporated into the Child–Pugh,
MELD and AD scores. The highest levels of MGO were particularly seen in patients presenting with Child–Pugh
C cirrhosis, which marks advanced liver failure. Low levels of albumin reflect impaired liver synthesis, and
albumin remained independently associated with higher MGO levels. Albumin has been proposed to possess a
strong antioxidant capacity that acts as a free radical scavenger for ROS 33. However, whether reduced albumin
levels lead to a decline in antioxidant activity against MGO and RCS cannot be proven by our current study
design and is beyond the scope of this study.
Recently, it has been shown that advanced liver cirrhosis drives a shift in cell metabolism to immune cells
with elevated energy demands at the site of inflammation. Metabolomic studies have revealed an accumulation
of several glycolytic metabolites in blood samples from patients with advanced liver cirrhosis 34. In inflammatory
tissues, innate immune cells need adenosine triphosphate (ATP), which is rapidly produced during glycolysis
Scientific Reports | (2021) 11:20506 |
https://doi.org/10.1038/s41598-021-00119-7
6
Vol:.(1234567890)www.nature.com/scientificreports/ a
150
)
l
/
m
g
n
(
O
G
100
50
0
c
150
ns
CC
DC
ns
100
)
l
/
m
g
n
(
O
G
50
0
ns
A
B
C
Child-Pugh score
ns
150
b
)
l
/
m
g
n
(
O
G
100
50
0
150
d
)
l
/
m
g
n
(
O
G
100
50
0
MELD < 15
MELD > 15
AD score < 50
AD score > 50
Figure 4. Whisker boxplots showing median (IQR 10th; 90th) GO serum levels in patients with different liver
disease severities. (a) The GO levels were not higher in patients with decompensated cirrhosis (DC). (b) Patients
with a Child–Pugh score of C did not show higher levels than patients with Child–Pugh A and B patients. (c)
A MELD score ≥ 15 did not increase the GO serum levels. (d) The acute decompensation (AD) score was not
significantly different between the two groups. Differences between two groups were analyzed using the Mann–
Whitney U test. More than two groups were analyzed by the Kruskal–Wallis test. The dots refer to values beyond
the range of the 10th and 90th percentiles. ns not significant.
a
40
e
r
o
c
s
D
L
E
M
30
20
10
0
0
80
b
e
r
o
c
s
D
A
60
40
20
0
0
50
100
150
GO (ng/ml)
50
100
150
GO (ng/ml)
Figure 5. Circulating glyoxal serum levels correlated with neither the MELD score (effect size r = 0.151) (a) nor
the AD score (effect size r = 0.073) (b).
instead of mitochondrial oxidative phosphorylation, a mechanism also seen in cancer cells 35. In this regard, the
accumulation of MGO could be a reflection of the higher glycolytic activity of immune cells, as MGO is the main
byproduct of glycolysis. In addition, higher MGO levels are also seen in patients with diabetes 12. Metformin, a
widely used medication in diabetes, is known to be a strong MGO scavenger 36. Of note, patients on metformin
showed lower levels of MGO than patients without metformin. However, this result must be interpreted with
caution. In patients with more advanced liver cirrhosis, metformin is contraindicated, thus imposing a certain
selection bias. Therefore, higher MGO levels in advanced liver cirrhosis may be a reflection of elevated energy
turnover of inflammatory cells.
This study has several limitations. The metabolism of carbonyl species is highly dynamic, and levels can
fluctuate over time. The levels of MGO and GO were only assessed at one time point in each patient in this
study. However, the blood levels of MGO and GO were assessed using HPLC–MS/MS, which represents the
most accurate method to date 37. Because the presence of ascites or OHE assessed during physical examination
Scientific Reports | (2021) 11:20506 |
https://doi.org/10.1038/s41598-021-00119-7
7
Vol.:(0123456789)www.nature.com/scientificreports/
Glyoxal (ng/mL)
Univariable analysis
Variable
Age
Sex
BMI
Type 2 diabetes
Metformin
r
0.054
0.229
− 0.082
− 0.086
− 0.175
Ascites at study inclusion
0.003
History of ascites
History of OHE
Child–Pugh score
MELD
MELD-Na
AD score
Sodium
Albumin
INR
Bilirubin
IL-6
CRP
Leukocytes
Creatinine
Thrombocytes
HVPG*
− 0.027
− 0.043
0.161
0.151
0.086
0.073
0.195
− 0.043
0.020
0.001
0.184
0.060
0.075
0.149
− 0.055
0.031
p
0.707
0.106
0.568
0.546
0.220
0.984
0.852
0.767
0.260
0.290
0.546
0.611
0.170
0.766
0.887
0.996
0.196
0.674
0.601
0.296
0.701
0.838
Table 3. Univariable analysis for predictors of higher glyoxal levels in patients with liver cirrhosis. AD acute
decompensation, BMI body mass index, CRP C-reactive protein, OHE overt hepatic encephalopathy, IL-6
interleukin-6, INR international normalized ratio, HVPG hepatic venous pressure gradient, MELD model for
end-stage liver disease, MELD-Na model for end-stage liver disease-sodium. A univariable analysis of the data
is shown. Boldface indicates significance. A p value < 0.05 was considered significant. *Measured in 46 patients.
is part of the Child–Pugh score, it may introduce a potential bias. Therefore, we focused on the MELD and
AD scores for further analysis in a linear regression model. These scores only consider laboratory values and
follow a standardized assessment. Additionally, blood glucose levels can potentially alter MGO concentrations
independently of the underlying disease. Therefore, blood was taken from fasting patients, and blood glucose
levels were additionally measured to minimize this potential confounder. Furthermore, this study cohort was
small; therefore, these findings need to be interpreted with caution. However, this study may be the first step in
analyzing MGO in liver cirrhosis, and future studies with larger cohorts are aimed at analyzing the predictive
value of MGO.
In conclusion, higher MGO levels were associated with increasing disease severity in patients with liver cir-
rhosis. The highest MGO levels were seen in patients with advanced liver cirrhosis in particular. The hepatic
dysfunction scores and liver-related parameters were independently associated with higher MGO levels. However,
further research is needed to elucidate the importance of MGO as a diagnostic biomarker and therapeutic target
in patients with liver cirrhosis.
Methods
Study population. A total of 51 patients with liver cirrhosis were prospectively enrolled between 2019 and
2021 in this cross-sectional cohort study after informed consent was obtained. The inclusion criteria were a diag-
nosis of liver cirrhosis according to current European clinical practice guidelines 38. Patients had to be at least
18 years of age. Patients with liver cancer/active malignancy or an active infection were not approached for this
study. Patients were recruited either during outpatient visits or at elective hospitalizations for measurement of
the hepatic venous pressure gradient (HVPG) and screening for esophageal varices. Clinical and laboratory data
were prospectively recorded on the day of study inclusion and available from the electronic health care records.
The presence of ascites was detected during a routine assessment of the abdomen with ultrasound. Esophageal
varices were assessed during routine endoscopy. Overt hepatic encephalopathy (OHE) was clinically evaluated
according to current practice guidelines 38. All biochemical parameters were assessed by the Institute of Clinical
Chemistry and Laboratory Medicine at the University Medical Center Mainz.
Assessment of liver disease severity. The severity of liver disease was stratified according to the Child–
Pugh score, the model for end-stage liver disease (MELD) score and the acute decompensation (AD) score.
Scientific Reports | (2021) 11:20506 |
https://doi.org/10.1038/s41598-021-00119-7
8
Vol:.(1234567890)www.nature.com/scientificreports/ Using the Child–Pugh score, patients were stratified as Child–Pugh A (5–6 points), B (7–9 points) or C (10–15
points) 39. In this context, Child–Pugh A refers to compensated cirrhosis (CC), whereas Child–Pugh B and C
resemble decompensated cirrhosis (DC). The MELD score is used to allocate organs to patients in need of liver
transplantation 40. It comprises the creatinine, total bilirubin and INR blood concentrations with scores ranging
from 5 to 40. A MELD score ≥ 15 has been adopted as a threshold value to list patients for liver transplantation
41. The MELD-Na score additionally contains blood sodium levels and is thought to predict mortality better than
the original MELD score 42. In contrast, the AD score was designed to predict mortality in patients presenting
with decompensated liver cirrhosis. It involves age, leukocyte count and the blood levels of sodium, creatinine
and INR. Patients with an AD score ≥ 60 are termed high risk and show a greater mortality, whereas a score ≤ 45
is associated with a lower mortality risk 43. For the comparison of patients, an AD score cutoff of 50 was chosen
according to the median value in this study population.
Sample collection. Blood samples were taken from patients during a routine measurement of the HVPG.
All patients were fasting for this procedure. After blood withdrawal, samples were incubated for 30 min to allow
clotting and then centrifuged at room temperature at 2000/rotations per minute (rpm) for 10 min 44. Then, the
serum supernatant was transferred into 1.5 mL tubes and immediately stored at − 80 °C until further processing.
Measurement of methylglyoxal (MGO) and glyoxal (GO). MGO and GO were quantified using a
previously published high-performance liquid chromatographic tandem mass spectrometric method (HPLC–
MS/MS) 45. Briefly, serum samples (500 μL) were spiked with 10 μL of the working solution of the internal
standard 3,4-hexanedione (1 μg/mL). After the addition of 250 μL of perchloric acid (7%), the samples were
mixed for 10 s, left for 15 min, and centrifuged for 10 min at 10,000×g. The supernatant was removed and
neutralized by adding 250 μL of saturated sodium hydrogen carbonate solution. Derivatization was performed
with 100 μL of 2,3-diaminonaphthalene (1 mg/mL in methanol) overnight at 4 °C. Afterward, the sample was
extracted with 4 mL ethyl acetate. The organic layer was transferred to another reaction tube, evaporated under
a stream of nitrogen at 40 °C, and reconstituted in 200 μL of methanol. Chromatographic separation was per-
formed with a Phenomenex (Aschaffenburg, Germany) Synergi® MAX-RP C12 analytical column (150 × 2 mm,
4 μm particle size) and a Phenomenex C18 (4 × 2 mm) guard column and a gradient flow (0.4 mL/min). Mol-
ecules were ionized by electrospray ionization (ESI) in positive mode. The following ion transitions in multiple
reaction monitoring (MRM) mode were used: for the derivative of MGO: 195.1–126.7 (collision energy: 49 eV,
target ion transition) and 195.1–77.0 (collision energy: 73 eV, qualifier ion transition); for the derivative of GO:
181.1–154.0 (collision energy: 45 eV, target ion transition) and 181.1–77.0 (collision energy: 80 eV, qualifier ion
transition); for the derivative of IS: 237.1–222.0 (collision energy: 35 eV, target ion transition) and 195.1–125.9
(collision energy: 85 eV, qualifier ion transition).
Ethics. All patients provided written informed consent. The study was conducted according to the ethical
guidelines of the 1975 Declaration of Helsinki (6th revision, 2008). The study was approved by the ethics com-
mittee of Landesärztekammer Rhineland-Palatine (Nr. 837.052.12 (8153)).
Statistical analysis. Descriptive analyses of data are expressed as either the mean with standard deviations
or median with interquartile ranges (IQR 25th; 75th). The Mann–Whitney U rank test was used to compare
groups and to calculate differences between two groups with quantitative values. The Kruskal–Wallis test was
used to compare differences between more than two groups. A chi-squared test was applied for the comparison
of two or more patient groups with categorical values. All tests were two-tailed, and significant values were
defined as p < 0.05. Univariable correlation analyses were used to examine associations between two variables.
All variables with p < 0.05 were then included in a multivariable linear regression model with a stepwise selection
process. To avoid multicollinearity, the MELD score, AD score and total bilirubin were independently analyzed
in multivariable linear regression models. Because the data analysis was exploratory, no adjustment for multiple
testing was performed. Due to the large number of tests, p values should be interpreted with caution and in con-
nection with effect estimates. For all data analysis and statistical tests, IBM SPSS Statistic Version 23.0 (Armonk,
NY: IBM Corp.) was used. For all graphs, GraphPad Prism 5.0 (San Diego, CA: GraphPad Software, LLC) was
used.
Institutional review board statement. The study was conducted according to the guidelines of the Dec-
laration of Helsinki and approved by the Ethics Committee of Landesärztekammer Rhineland-Palatine (Nr.
837.052.12 (8153)).
Informed consent.
Informed consent was obtained from all patients involved in the study.
Data availability
The data presented in this study are available on request from the corresponding authors.
Received: 7 July 2021; Accepted: 6 October 2021
Scientific Reports | (2021) 11:20506 |
https://doi.org/10.1038/s41598-021-00119-7
9
Vol.:(0123456789)www.nature.com/scientificreports/ References
1. Mokdad, A. A. et al. Liver cirrhosis mortality in 187 countries between 1980 and 2010: A systematic analysis. BMC Med. 12, 145.
https:// doi. org/ 10. 1186/ s12916- 014- 0145-y (2014).
2. Tsochatzis, E. A., Bosch, J. & Burroughs, A. K. Liver cirrhosis. The Lancet 383, 1749–1761. https:// doi. org/ 10. 1016/ S0140- 6736(14)
60121-5 (2014).
3. Ginès, P. et al. Liver cirrhosis. The Lancet. https:// doi. org/ 10. 1016/ S0140- 6736(21) 01374-X (2021).
4. D’Amico, G. et al. Clinical states of cirrhosis and competing risks. J. Hepatol. 68, 563–576. https:// doi. org/ 10. 1016/j. jhep. 2017. 10.
020 (2018).
5. Bernardi, M., Moreau, R., Angeli, P., Schnabl, B. & Arroyo, V. Mechanisms of decompensation and organ failure in cirrhosis: From
peripheral arterial vasodilation to systemic inflammation hypothesis. J. Hepatol. 63, 1272–1284. https:// doi. org/ 10. 1016/j. jhep.
2015. 07. 004 (2015).
6. Arroyo, V. et al. The systemic inflammation hypothesis: Towards a new paradigm of acute decompensation and multiorgan failure
in cirrhosis. J. Hepatol. https:// doi. org/ 10. 1016/j. jhep. 2020. 11. 048 (2020).
7. Rabbani, N. & Thornalley, P. J. Methylglyoxal, glyoxalase 1 and the dicarbonyl proteome. Amino Acids 42, 1133–1142. https:// doi.
org/ 10. 1007/ s00726- 010- 0783-0 (2012).
8. Hollenbach, M. The role of glyoxalase-I (Glo-I), advanced glycation endproducts (AGEs), and their receptor (RAGE) in chronic
liver disease and hepatocellular carcinoma (HCC). International J. Mol. Sci. 18, 2466. https:// doi. org/ 10. 3390/ ijms1 81124 66 (2017).
9. Nigro, C. et al. Dicarbonyl stress at the crossroads of healthy and unhealthy aging. Cells 8, 749. https:// doi. org/ 10. 3390/ cells 80707
49 (2019).
10. Shamsi, A., Amani, S., Alam, M. T. & Naeem, A. Aggregation as a consequence of glycation: Insight into the pathogenesis of
arthritis. Eur. Biophys. J. 45, 523–534. https:// doi. org/ 10. 1007/ s00249- 016- 1119-0 (2016).
11. Shamsi, A., Ahmed, A. & Bano, B. Glyoxal induced structural transition of buffalo kidney cystatin to molten globule and aggregates:
Anti-fibrillation potency of quinic acid. IUBMB Life 68, 156–166. https:// doi. org/ 10. 1002/ iub. 1471 (2016).
12. Schalkwijk, C. G. & Stehouwer, C. D. A. Methylglyoxal, a highly reactive dicarbonyl compound, in diabetes, its vascular complica-
tions, and other age-related diseases. Physiol. Rev. 100, 407–461. https:// doi. org/ 10. 1152/ physr ev. 00001. 2019 (2020).
13. Ahmed, A., Shamsi, A., Khan, M. S., Husain, F. M. & Bano, B. Methylglyoxal induced glycation and aggregation of human serum
albumin: Biochemical and biophysical approach. Int. J. Biol. Macromol. 113, 269–276. https:// doi. org/ 10. 1016/j. ijbio mac. 2018. 02.
137 (2018).
14. Knani, I. et al. Methylglyoxal: A relevant marker of disease activity in patients with rheumatoid arthritis. Dis. Mark. 2018, 8735926.
https:// doi. org/ 10. 1155/ 2018/ 87359 26 (2018).
15. Rabbani, N. & Thornalley, P. J. Advanced glycation end products in the pathogenesis of chronic kidney disease. Kidney Int. 93,
803–813. https:// doi. org/ 10. 1016/j. kint. 2017. 11. 034 (2018).
16. Michel, M., Hollenbach, M., Pohl, S., Ripoll, C. & Zipprich, A. Inhibition of glyoxalase-I leads to reduced proliferation, migration
and colony formation, and enhanced susceptibility to sorafenib in hepatocellular carcinoma. Front. Oncol. 9, 785. https:// doi. org/
10. 3389/ fonc. 2019. 00785 (2019).
17. Hollenbach, M. et al. Expression of glyoxalase-I is reduced in cirrhotic livers: A possible mechanism in the development of cir-
rhosis. PLoS ONE 12, e0171260. https:// doi. org/ 10. 1371/ journ al. pone. 01712 60 (2017).
18. Ahmed, N. et al. Processing of protein glycation, oxidation and nitrosation adducts in the liver and the effect of cirrhosis. J. Hepatol.
41, 913–919. https:// doi. org/ 10. 1016/j. jhep. 2004. 08. 007 (2004).
19. Šebeková, K., Kupčová, V., Schinzel, R. & Heidland, A. Markedly elevated levels of plasma advanced glycation end products in
patients with liver cirrhosis—Amelioration by liver transplantation. J. Hepatol. 36, 66–71. https:// doi. org/ 10. 1016/ s0168- 8278(01)
00232-x (2002).
20. Brenner, T. et al. Reactive metabolites and AGE-RAGE-mediated inflammation in patients following liver transplantation. Mediat.
Inflamm. 2013, 501430. https:// doi. org/ 10. 1155/ 2013/ 501430 (2013).
21. Spanos, C. et al. Proteomic identification and characterization of hepatic glyoxalase 1 dysregulation in non-alcoholic fatty liver
disease. Proteome Sci. 16, 4. https:// doi. org/ 10. 1186/ s12953- 018- 0131-y (2018).
22. Luedde, T., Kaplowitz, N. & Schwabe, R. F. Cell death and cell death responses in liver disease: Mechanisms and clinical relevance.
Gastroenterology 147, 765–783. https:// doi. org/ 10. 1053/j. gastro. 2014. 07. 018 (2014).
23. Remmler, J. et al. Increased level of interleukin 6 associates with increased 90-day and 1-year mortality in patients with end-stage
liver disease. Clin. Gastroenterol. Hepatol. 16, 730–737. https:// doi. org/ 10. 1016/j. cgh. 2017. 09. 017 (2018).
24. Labenz, C. et al. Raised serum Interleukin-6 identifies patients with liver cirrhosis at high risk for overt hepatic encephalopathy.
Aliment. Pharmacol. Ther. 50, 1112–1119. https:// doi. org/ 10. 1111/ apt. 15515 (2019).
25. Bernsmeier, C., van der Merwe, S. & Périanin, A. Innate immune cells in cirrhosis. J. Hepatol. 73, 186–201. https:// doi. org/ 10.
1016/j. jhep. 2020. 03. 027 (2020).
26. Baumann, T. et al. Regulatory myeloid cells paralyze T cells through cell-cell transfer of the metabolite methylglyoxal. Nat. Immunol.
21, 555–566. https:// doi. org/ 10. 1038/ s41590- 020- 0666-9 (2020).
27. Thornalley, P. J. The glyoxalase system: New developments towards functional characterization of a metabolic pathway fundamental
to biological life. Biochem. J. 269, 1–11. https:// doi. org/ 10. 1042/ bj269 0001 (1990).
28. Masterjohn, C. et al. Acute glutathione depletion induces hepatic methylglyoxal accumulation by impairing its detoxification to
D-lactate. Exp. Biol. Med. 238, 360–369. https:// doi. org/ 10. 1177/ 15353 70213 477987 (2013).
29. Xue, M. et al. Improved glycemic control and vascular function in overweight and obese subjects by glyoxalase 1 inducer formula-
tion. Diabetes 65, 2282–2294. https:// doi. org/ 10. 2337/ db16- 0153 (2016).
30. Yang, K. et al. Differences in glyoxal and methylglyoxal metabolism determine cellular susceptibility to protein carbonylation and
cytotoxicity. Chem. Biol. Interact. 191, 322–329. https:// doi. org/ 10. 1016/j. cbi. 2011. 02. 012 (2011).
31. Baba, S. P. et al. Reductive metabolism of AGE precursors: A metabolic route for preventing AGE accumulation in cardiovascular
tissue. Diabetes 58, 2486–2497. https:// doi. org/ 10. 2337/ db09- 0375 (2009).
32. Gaens, K. H. J. et al. Endogenous formation of Nε-(carboxymethyl)lysine is increased in fatty livers and induces inflammatory
markers in an in vitro model of hepatic steatosis. J. Hepatol. 56, 647–655. https:// doi. org/ 10. 1016/j. jhep. 2011. 07. 028 (2012).
33. Bernardi, M. et al. Albumin in decompensated cirrhosis: New concepts and perspectives. Gut 69, 1127–1138. https:// doi. org/ 10.
1136/ gutjnl- 2019- 318843 (2020).
34. Moreau, R. et al. Blood metabolomics uncovers inflammation-associated mitochondrial dysfunction as a potential mechanism
underlying ACLF. J. Hepatol. 72, 688–701. https:// doi. org/ 10. 1016/j. jhep. 2019. 11. 009 (2020).
35. Wang, A., Luan, H. H. & Medzhitov, R. An evolutionary perspective on immunometabolism. Science. https:// doi. org/ 10. 1126/
scien ce. aar39 32 (2019).
36. Beisswenger, P. J., Howell, S. K., Touchette, A. D., Lal, S. & Szwergold, B. S. Metformin reduces systemic methylglyoxal levels in
type 2 diabetes. Diabetes 48, 198–202. https:// doi. org/ 10. 2337/ diabe tes. 48.1. 198 (1999).
37. Scheijen, J. L. J. M. & Schalkwijk, C. G. Quantification of glyoxal, methylglyoxal and 3-deoxyglucosone in blood and plasma by
ultra performance liquid chromatography tandem mass spectrometry: Evaluation of blood specimen. Clin. Chem. Lab. Med. 52,
85–91. https:// doi. org/ 10. 1515/ cclm- 2012- 0878 (2014).
Scientific Reports | (2021) 11:20506 |
https://doi.org/10.1038/s41598-021-00119-7
10
Vol:.(1234567890)www.nature.com/scientificreports/ 38. Angeli, P. et al. EASL clinical practice guidelines for the management of patients with decompensated cirrhosis. J. Hepatol. 69,
406–460. https:// doi. org/ 10. 1016/j. jhep. 2018. 03. 024 (2018).
39. Pugh, R. N., Murray-Lyon, I. M., Dawson, J. L., Pietroni, M. C. & Williams, R. Transection of the oesophagus for bleeding oesopha-
geal varices. Br. J. Surg. 60, 646–649. https:// doi. org/ 10. 1002/ bjs. 18006 00817 (1973).
40. Durand, F. & Valla, D. Assessment of the prognosis of cirrhosis: Child-Pugh versus MELD. J. Hepatol. 42(Suppl), S100–S107.
https:// doi. org/ 10. 1016/j. jhep. 2004. 11. 015 (2005).
41. Habib, S. et al. MELD and prediction of post-liver transplantation survival. Liver Transplant. 12, 440–447. https:// doi. org/ 10. 1002/
lt. 20721 (2006).
42. Kim, W. R. et al. Hyponatremia and mortality among patients on the liver-transplant waiting list. N. Engl. J. Med. 359, 1018–1026.
https:// doi. org/ 10. 1056/ NEJMo a0801 209 (2008).
43. Jalan, R. et al. The CLIF consortium acute decompensation score (CLIF-C ADs) for prognosis of hospitalised cirrhotic patients
without acute-on-chronic liver failure. J. Hepatol. 62, 831–840. https:// doi. org/ 10. 1016/j. jhep. 2014. 11. 012 (2015).
44. Rabbani, N. & Thornalley, P. J. Measurement of methylglyoxal by stable isotopic dilution analysis LC-MS/MS with corroborative
prediction in physiological samples. Nat. Protoc. 9, 1969–1979. https:// doi. org/ 10. 1038/ nprot. 2014. 129 (2014).
45. Hess, C. et al. Clinical and forensic examinations of glycaemic marker methylglyoxal by means of high performance liquid
chromatography-tandem mass spectrometry. Int. J. Legal Med. 127, 385–393. https:// doi. org/ 10. 1007/ s00414- 012- 0740-4 (2013).
Acknowledgements
Dr. M. Michel is supported by the Clinician Scientist Fellowship “Else Kröner Research College: 2018_Kolleg.05”.
Author contributions
Performed research: M.Mi. (Michel), C.H., C.L. and M.N. Contributed to acquisition of data: M.Mi., C.H., L.K.,
W.M.K., M.H., C.L. and M.N.; Designed the experiments and analyzed the data: M.Mi., C.H., C.L. and M.N.;
Contributed reagents/materials/analysis tools: M.Mi., C.H., P.R.G., M.Mo. (Moehler), M.A.W., J.M.S., C.L. and
M.N. Wrote the manuscript: M.Mi. Revised and edited the manuscript: C.H., L.K., J.M.S., C.L. and M.N. Statisti-
cal analysis: M.Mi. All authors approved the final version of the manuscript and the authorship list. Guarantor
of the article: M.Mi. and M.N.
Funding
Open Access funding enabled and organized by Projekt DEAL.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary Information The online version contains supplementary material available at https:// doi. org/
10. 1038/ s41598- 021- 00119-7.
Correspondence and requests for materials should be addressed to M.M. or M.N.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
© The Author(s) 2021
Scientific Reports | (2021) 11:20506 |
https://doi.org/10.1038/s41598-021-00119-7
11
Vol.:(0123456789)www.nature.com/scientificreports/
| null |
10.1038_s41593-023-01409-1.pdf
| null |
Data availability All primary data for the figures and extended data figures are available from the corresponding author (K.D.) upon request. Code availability The code used for data processing and analysis is available from the corresponding author (K.D.) upon request. Nature Neuroscience Article https://doi.org/10.1038/s41593-023-01409-1 Article https://doi.org/10.1038/s41593-023-01409-1 Extended Data Fig. 4 | See next page for caption. Nature Neuroscience Article https://doi.org/10.1038/s41593-023-01409-1
|
Orbitofrontal cortex control of striatum
leads economic decision-making
https://doi.org/10.1038/s41593-023-01409-1
Received: 12 February 2023
Accepted: 17 July 2023
Published online: 17 August 2023
Check for updates
Felicity Gore1,2,3, Melissa Hernandez1,2, Charu Ramakrishnan
Ailey K. Crow1,2, Robert C. Malenka
2,3 & Karl Deisseroth
1,2,4
1,2,
Animals must continually evaluate stimuli in their environment to decide
which opportunities to pursue, and in many cases these decisions can be
understood in fundamentally economic terms. Although several brain
regions have been individually implicated in these processes, the brain-wide
mechanisms relating these regions in decision-making are unclear.
Using an economic decision-making task adapted for rats, we find that
neural activity in both of two connected brain regions, the ventrolateral
orbitofrontal cortex (OFC) and the dorsomedial striatum (DMS), was
required for economic decision-making. Relevant neural activity in both
brain regions was strikingly similar, dominated by the spatial features of
the decision-making process. However, the neural encoding of choice
direction in OFC preceded that of DMS, and this temporal relationship was
strongly correlated with choice accuracy. Furthermore, activity specifically
in the OFC projection to the DMS was required for appropriate economic
decision-making. These results demonstrate that choice information in the
OFC is relayed to the DMS to lead accurate economic decision-making.
Economic decision-making, the process of evaluating options in the
environment to inform the best course of action, is critical for a wide
range of behaviors essential for survival and well-being. To make opti-
mal decisions, the neural representation of each option must be inte-
grated with information about the type and scale of outcome it predicts
to provide a representation of the subjective value of each alterna-
tive. Representations of subjective value can then be compared before
engaging neural circuits that generate flexible behavioral responses1–3.
Neural representations of subjective value have been identified
in the orbitofrontal cortex (OFC)4, and electrical microstimulation
of the OFC can bias choice behavior5. These results have supported
a widespread hypothesis that the OFC has a role in economic
decision-making1,6–11. However, lesions and inactivation of the OFC
yielded conflicting results on choice behavior12–16. Furthermore, repre-
sentations of subjective value exist in other brain regions including
the medial prefrontal cortex17, dorsomedial striatum (DMS)18 and
mediodorsal thalamus19; similar manipulations of each of these brain
regions influence decision-making behavior20–24. Thus, multiple brain
regions may have important roles in economic decision-making; how-
ever, surprisingly little is known about if and how these brain regions
may interact to mediate economic choices. One reason for this limited
understanding is that most studies examining the neural correlates
of value-based decision-making have been conducted in nonhuman
primate systems, wherein tools are more restricted for recording and
manipulating activity of precisely defined populations of neurons. To
address this limitation, we adapted an economic decision-making task
for rats, which permits recording and manipulation of neural activity
in multiple defined neural populations while rats make economic
decisions.
Results
Integration of reward quantity and quality information during
economic decision-making
In the initial experiments, we developed and validated an economic
decision-making task in rats (Fig. 1a). On each trial, rats were pre-
sented with two visual cues side by side. The type of stimulus (vertical
1Department of Bioengineering, Stanford University, Stanford, CA, USA. 2Department of Psychiatry and Behavioral Sciences, Stanford University, Stanford,
CA, USA. 3Nancy Pritzker Laboratory, Stanford University, Stanford, CA, USA. 4Howard Hughes Medical Institute, Stanford University, Stanford, CA, USA.
e-mail: [email protected]
Nature Neuroscience | Volume 26 | September 2023 | 1566–1574
1566
nature neuroscienceArticle a
b
)
t
n
a
r
r
u
c
k
c
a
l
b
e
s
o
o
h
c
(
P
d
i
)
s
(
e
c
o
h
c
o
t
y
c
n
e
t
a
l
e
v
i
t
a
l
e
R
1.0
0.8
0.6
0.4
0.2
0
1.5
1.0
0.5
0
5–10-s ITI
2-s cue
Choice
Reward
Blackcurrant
Lemon
***
***
***
***
***
***
***
***
***
***
***
***
Prefer
blackcurrant
Prefer
lemon
*
***
***
***
*
***
e
i
s
l
a
m
n
a
f
o
r
e
b
m
u
N
10
8
6
4
2
0
c
)
t
n
a
r
r
u
c
k
c
a
l
b
e
s
o
o
h
c
(
P
1.0
0.8
0.6
0.4
0.2
0
f
4
h
t
n
o
m
e
r
o
c
s
e
c
n
e
r
e
f
e
r
P
1
0
–1
y
c
a
r
u
c
c
A
1.0
0.8
0.6
0.4
0.2
0
–3
–2
–1
1
2
3
0
–
R = 0.96
***P < 0.001
–3
–2
–1
0
–
1
2
3
–1
0
1
2
–1
0
1
Preference score
Preference score month 0
Fig. 1 | Rats integrate information about reward quantity and reward
identity to make economic decisions. a, Schematic of economic decision-
making task for rats. b, Probability of choosing the blackcurrant-predictive cue
for all cue combinations (n = 42 rats, one-way repeated-measures analysis of
variance (ANOVA)). c, Probability of choosing the blackcurrant-predictive cue
as a function of difference in size of the reward available. Rats were more likely
to choose the larger available reward (n = 42 rats, one-way repeated-measures
ANOVA). Inset, fraction of trials in which the animal chose the larger available
reward (n = 42 rats, 0.82 ± 0.01). d, Latency to choice nosepoke response as a
function of the difference in the size of the reward available. Rats were faster
when the difference in reward volume was high (easy trials) (n = 42 rats, one-way
repeated-measures ANOVA). The center dot represents the median, the bars
represent the first and third quartiles. e, Histogram of preference scores: the
difference in available reward at which the animal was equally likely to choose
the blackcurrant-predictive or lemon-predictive cue (negative values, blue
shading: rats preferring blackcurrant; positive values, yellow shading: rats
preferring lemon; n = 42 rats). f, Correlation of preference scores computed on
sessions performed 4 months apart. Preference scores were highly correlated,
indicating that the juice preferences of individual animals were stable across
time (n = 12 rats, Pearson correlation). *P < 0.05, ***P < 0.001. Unless otherwise
noted, data are presented as the mean ± s.e.m. Full statistical details are shown in
Supplementary Table 1.
or horizontal drifting gratings) indicated the identity of the associated
reward (blackcurrant-flavored or lemon-flavored water), and the size
of the visual stimulus indicated the size of the associated reward. After
2 s, the animals could perform a nosepoke to the side of the chosen
cue to indicate choice, whereupon the chosen reward was delivered.
We found that rats reliably chose visual stimuli that predicted larger
volume rewards (Fig. 1b,c). In addition, animals displayed slower choice
latencies on trials in which the difference in available reward volume
was small (difficult trials), compared with trials in which the difference
in available reward volume was large (easy trials) (Fig. 1d). To confirm
that animals were making decisions based on the value of the stimuli
presented, as opposed to simply detecting larger visual cues more
reliably, we included a subset of animals in which the size of the visual
stimulus was not positively correlated with the size of the reward it
predicted. These animals still reliably chose stimuli that predicted
larger volume rewards, indicating that animals used information about
the available reward volume to make appropriate decisions (Extended
Data Fig. 1b–e).
We next asked whether animals used information about reward
identity, in addition to information about reward volume, to guide
their decision-making. For each animal, we generated a preference
score by calculating the difference in available reward (the num-
ber of drops of blackcurrant-flavored water − number of drops
of lemon-flavored water) at which the animal was equally likely to
choose the blackcurrant-predictive or lemon-predictive cue. We
found that individual animals displayed modest preferences for either
blackcurrant-predictive or lemon-predictive cues (Fig. 1e). Moreover,
we found that the preference scores of individual animals were strongly
correlated across consecutive sessions (Extended Data Fig. 1f) as well
as across sessions separated by approximately 4 months (Fig. 1f).
Thus, juice preferences were stable across both short and long time-
scales in individual animals. Animals therefore integrated individual
Nature Neuroscience | Volume 26 | September 2023 | 1566–1574
1567
Articlehttps://doi.org/10.1038/s41593-023-01409-1
(subjective) internal preferences regarding the available reward quality
with externally accessible (objective) information about available
reward quantity to make economic decisions.
Activity in both OFC and DMS is necessary for economic
decision-making
Electrophysiological recordings identified several brain regions that
appear to encode important features of economic decision-making
tasks4,17–19. To determine which brain regions were critical for this
task, we performed an optogenetic inactivation screen. Specifically,
after rats achieved criterion performance (Methods), we injected an
adeno-associated virus (AAV) encoding the inhibitory stabilized step
function opsin SwiChR++25 under the control of the human synapsin
promoter (AAV8 hSyn:SwiChR++EYFP) bilaterally into the OFC, DMS,
mediodorsal thalamus or prelimbic cortex, and positioned optical fibers
above each of these structures (Fig. 2a and Extended Data Fig. 2).
When animals had reestablished criterion performance, we asked
whether optical inhibition of each of these brain structures altered
decision-making performance (Fig. 2b).
In accordance with previous work in mice16, optogenetic inhibi-
tion of the OFC impaired economic decision-making. We found that
psychometric curves were flatter and latencies for easy choices were
slower (Fig. 2c,d). In addition, we found that preferences computed
on trials in which the OFC was inhibited were not correlated with
preferences computed on trials in which the OFC was not inhibited,
indicating that optogenetic inhibition of the OFC also disrupted juice
preferences (Fig. 2e). Optogenetic inhibition of the DMS also impaired
economic decision-making; psychometric curves were flatter and
latencies for easy choices were slower, but choice preferences were
unchanged (Fig. 2i–k), suggesting that decision-making based on
reward volume was disrupted but juice preferences remained intact.
In contrast, optogenetic inhibition of either the prelimbic cortex
(Fig. 2f–h) or mediodorsal thalamus (Fig. 2l–n) had no discernible effect
on economic decision-making. Decision-making was also unchanged in
animals injected with control virus encoding enhanced yellow fluores-
cent protein (EYFP) and subjected to the same procedures (Extended
Data Fig. 3a–d).
To determine whether the effects we observed were attribut-
able to a specific deficit in economic decision-making or due to an
unanticipated nonspecific effect of intervention (such as impaired
visual perception, action execution or value recall), we placed the
same animals into a control task, in which the choice component of the
economic decision-making task was removed (Extended Data Fig. 3e).
On uninhibited trials, animals were faster to respond to cues that pre-
dicted larger volume rewards, suggesting that animals could perceive
the cues, remember their values and act accordingly. Importantly
this relationship was maintained when either the OFC (Extended Data
Fig. 3f) or DMS (Extended Data Fig. 3g) was inhibited. Thus, inhibi-
tion of the OFC or DMS impairs economic decision-making without
impairing visual perception, action execution or the representation
(or recollection) of cue value.
Choice-related activity in the OFC precedes choice-related
activity in the DMS
To explore in more detail what function these brain areas might have
in economic decision-making, we performed wireless extracellular
electrophysiological recordings in the OFC and DMS in freely moving
rats. A large proportion of task-modulated single units were identified
among all the units resolved in both brain areas (OFC: 1,157 of 1,329
units, n = 6 rats; DMS: 524 of 656 units, n = 6 rats). In both regions,
trial-averaged single-unit activity spanned the trial, and single units
that were modulated by a range of task features were identified
(Fig. 3a,b). We observed striking similarity in neural encoding in the
OFC and DMS, with single-unit responses dominated by the spatial
features of the task (size of the reward offered on the left, size of
the reward offered on the right, and side chosen) in both brain areas
(Fig. 3c and Extended Data Fig. 4a). Interestingly, in agreement with
our inactivation data, we observed that despite a similar proportion of
neurons encoding both the objective value (size) and subjective value
of rewards predicted by cues presented on either side of the animal,
neurons in the OFC were more strongly modulated by the subjective
value of a stimulus than by its objective value, an effect that was not
observed in the DMS (Extended Data Fig. 4b).
To characterize the temporal dynamics of encoding between the
OFC and DMS, we trained a linear support vector machine (SVM) to
decode the choice the animal made on each trial (left or right) from
neural activity data recorded in either the OFC or DMS (Fig. 3d). We
were able to decode choice direction with high accuracy on held-out
neural activity data from both brain regions. Importantly, across all
animals, choice prediction peaked in the OFC before it peaked in the
DMS (Fig. 3e and Extended Data Fig. 4c). We next examined how this
temporal relationship related to choice accuracy. Cross-correlations
of the predicted choice parameter (the perpendicular distance of the
decoded decision value from the support vector, a proxy for deci-
sion confidence) computed on single trials revealed that the OFC
led the DMS more on trials in which animals chose the larger avail-
able reward (‘correct’ trials) than on trials in which animals chose the
smaller available reward (‘incorrect’ trials) (Fig. 3f,g; correct trials
lag = −23.93 ± 22.86 ms (OFC leads), incorrect trials lag 44.82 ± 26.69 ms
(DMS leads), n = 30 sessions from five rats). These data demonstrate
that the encoding of choice-related information in the OFC precedes
the encoding of choice-related information in the DMS, and that this
relationship is correlated with choice accuracy.
To examine how information transmission between the OFC and
DMS might be disrupted on error trials, we first asked whether an
SVM trained on trials where animals chose the larger available reward
(correct trials) could predict choice behavior on trials when animals
chose the smaller available reward (incorrect trials). Strikingly, a model
trained on data recorded from either the OFC or DMS on correct trials
predicted the side the animal would choose equally well on correct
and incorrect trials (Extended Data Fig. 5a), suggesting that both brain
areas encode the chosen side with equivalent accuracy regardless of the
correctness of the choice. We next examined the SVM predicted choice
Fig. 2 | Activity in the OFC and DMS is important for economic decision-
making. a, Left: schematic of surgical preparation. Right: example single units
showing inhibition of spiking activity in response to optical stimulation.
b, Schematic of choice task with optical inhibition restricted to the cue evaluation
period. c–n, OFC and DMS inhibition impairs economic decision-making.
c,f,i,l, Probability of choosing the blackcurrant-predictive cue as a function of
the difference in the volume of available rewards for uninhibited (green) and
inhibited (magenta) trials. Rats were less likely to choose larger volume rewards
when the OFC (c) or DMS (i) was inhibited but not when the prelimbic cortex (f)
or mediodorsal thalamus (l) was inhibited (OFC: n = 12 rats; prelimbic cortex:
n = 7 rats; DMS: n = 6 rats; mediodorsal thalamus: n = 6 rats; two-way repeated-
measures ANOVA). Inset, fraction of trials in which the animal chose the larger
available reward on uninhibited (green) and inhibited (magenta) trials (two-sided
paired t-test). d,g,j,m, Latency to nosepoke choice response as a function of the
absolute difference in the size of rewards available on uninhibited (green) and
inhibited (magenta) trials. Rats were slower to respond when the OFC (d) or DMS
(j) was inhibited in trials wherein the difference in reward volume was high (easy
trials) (OFC: n = 12 rats; prelimbic cortex: n = 7 rats; DMS: n = 6 rats; mediodorsal
thalamus: n = 6 rats, two-way repeated-measures ANOVA). Inhibition of either the
prelimbic cortex (g) or mediodorsal thalamus (m) did not alter response latency.
e,h,k,n, Juice preferences computed on trials in which the OFC (e) was inhibited
were not correlated with juice preferences computed on trials in which the OFC
was not inhibited (Pearson correlation). Inhibition of the prelimbic cortex (h),
DMS (k) or mediodorsal thalamus (n) did not change juice preferences. *P < 0.05,
**P < 0.01, ***P < 0.001. Data are presented as the mean ± s.e.m. Full statistical
details can be found in Supplementary Table 1.
Nature Neuroscience | Volume 26 | September 2023 | 1566–1574
1568
Articlehttps://doi.org/10.1038/s41593-023-01409-1 parameters computed on held-out trials where the animal made either
correct or incorrect choices. As before, on correct trials we observed
that the predicted choice parameter increased in the OFC before the
DMS. However, when animals made an erroneous choice, we observed
that despite the predicted choice parameter reaching similar levels as
seen on correct trials, the predicted choice parameter did not increase
in the OFC before the DMS (Extended Data Fig. 5b–d). Thus, while the
transmission of spatial choice information from the OFC to the DMS
a
AAV8 hSyn:SwiChR++
(bilateral)
473 nm
635 nm
b
5–10-s ITI
2-s cue
Choice
OFC inhibition
***
e
0
4
8
12
Time (s)
No inhibition
Inhibition
***
***
***
***
***
***
***
1.0
0.8
0.6
y
c
a
r
u
c
c
A
–3
–2
–1
1
2
3
0
–
d
i
)
s
(
e
c
o
h
c
o
t
y
c
n
e
t
a
l
e
v
i
t
a
l
e
R
1.0
0.5
0
1
3
2
–
Prelimbic cortex inhibition
g
1.0
0.5
0
1
3
2
–
DMS inhibition
1.0
0.5
**
i
)
s
(
e
c
o
h
c
o
t
y
c
n
e
t
a
l
e
v
i
t
a
l
e
R
j
i
)
s
(
e
c
o
h
c
o
t
y
c
n
e
t
a
l
e
v
i
t
a
l
e
R
1.0
0.8
0.6
y
c
a
r
u
c
c
A
–3
–2
–1
1
2
3
0
–
***
*
**
***
1.0
0.8
0.6
y
c
a
r
u
c
c
A
***
***
*
–3
–2
–1
1
2
3
0
–
0
1
3
2
–
Mediodorsal thalamus inhibition
1.0
0.5
m
i
)
s
(
e
c
o
h
c
o
t
y
c
n
e
t
a
l
e
v
i
t
a
l
e
R
1.0
0.8
0.6
y
c
a
r
u
c
c
A
–3
–2
–1
1
2
3
0
–
0
1
3
2
–
c
)
t
n
a
r
r
u
c
k
c
a
l
b
e
s
o
o
h
c
(
P
f
)
t
n
a
r
r
u
c
k
c
a
l
b
e
s
o
o
h
c
(
P
i
)
t
n
a
r
r
u
c
k
c
a
l
b
e
s
o
o
h
c
(
P
l
)
t
n
a
r
r
u
c
k
c
a
l
b
e
s
o
o
h
c
(
P
1.0
0.8
0.6
0.4
0.2
0
1.0
0.8
0.6
0.4
0.2
0
1.0
0.8
0.6
0.4
0.2
0
1.0
0.8
0.6
0.4
0.2
0
n
o
i
t
i
b
h
n
i
i
e
r
o
c
s
e
c
n
e
r
e
f
e
r
P
h
n
o
i
t
i
b
h
n
i
i
e
r
o
c
s
e
c
n
e
r
e
f
e
r
P
k
n
o
i
t
i
b
h
n
i
i
e
r
o
c
s
e
c
n
e
r
e
f
e
r
P
n
n
o
i
t
i
b
h
n
i
i
e
r
o
c
s
e
c
n
e
r
e
f
e
r
P
1
0
–1
1
0
–1
1
0
–1
1
0
–1
R = –0.25
P = 0.44
–1
0
1
Preference score no inhibition
R = 0.69
P = 0.09
–1
0
1
Preference score no inhibition
R = 0.87
*P = 0.02
–1
0
1
Preference score no inhibition
R = 0.82
*P = 0.04
–1
0
1
Preference score no inhibition
Nature Neuroscience | Volume 26 | September 2023 | 1566–1574
1569
Articlehttps://doi.org/10.1038/s41593-023-01409-1
is necessary to initiate appropriate value-based choice behavior, with-
out this information choice might be initiated by other brain regions
reflecting internal biases relating to habitual behavior.
Activity of the OFC projection to the DMS is necessary for
economic decision-making
The temporal relationship between choice-related information in the
OFC and DMS suggests that choices represented in the OFC could be
relayed to the DMS to guide appropriate choice behavior. To address
this hypothesis, we first examined the axonal projections from the OFC
and confirmed the presence of a robust projection to the DMS26
(Fig. 4a,b). We next specifically inhibited this direct projection by bilate-
rally injecting an AAV encoding a variant of the inhibitory halorhodopsin,
which we optimized for axonal trafficking27, under the control of the
human synapsin promoter (AA8 hSyn:eNpHR3.0-NRN-EYFP) into
the OFC (Fig. 4c,d). We positioned optical fibers bilaterally in either
the DMS or mediodorsal thalamus, another major target of the OFC
projections (Fig. 4b). We found that optogenetic inhibition of OFC
inputs into the DMS selectively impaired decision-making related to
reward volume: psychometric curves were flatter and choice latencies
were disrupted (Fig. 4e,f), while preference scores were unchanged,
indicating that inhibition of the OFC projection to the DMS did not
disrupt juice preferences (Fig. 4g). In contrast, optogenetic inhibition of
the OFC inputs to the mediodorsal thalamus had no effect on economic
decision-making (Fig. 4h–j). In addition, optogenetic inhibition of the
OFC projection to the DMS or mediodorsal thalamus had no effect on
response latencies in the control task in which the choice component
of the economic decision-making task was selectively eliminated,
confirming that this manipulation did not impair visual perception,
action execution or the representation (or recollection) of cue value
(Extended Data Fig. 6a–c). Taken together, the data shown in this study
indicate that information relayed directly from the OFC to the DMS is
important for guiding economic decision-making.
Discussion
Animals must constantly evaluate stimuli in their environment to guide
appropriate approach and avoidance behaviors1–3. To study how neural
activity patterns across the brain may mediate these complex behav-
iors, we adapted an economic decision-making task for rats. Our experi-
ments demonstrate that activity in the OFC and DMS, but surprisingly
not in the prelimbic cortex or mediodorsal thalamus, is important for
economic decision-making. Moreover, neural activity in both brain
areas is dominated by spatial features of the economic decision-making
task. Interestingly, we found that choice-related activity emerges in
the OFC before the DMS, a relationship that correlates with choice
accuracy. Finally, we found that activity of the direct connection from
OFC to DMS is important for appropriate decision-making behavior.
Taken together, these data suggest that spatial choice information is
relayed from the OFC to the DMS to guide economic decision-making
appropriate to the individual.
Several lines of previous evidence have supported a role for the
OFC in economic decision-making1,4–11; however, inactivation and
lesion studies have yielded contradictory results12–16. In this study,
we leveraged the temporal resolution and enhanced the sensitivity of
a designed inhibitory stabilized step function opsin25 to inhibit OFC
selectively during the cue evaluation period, when rats are making
decisions. This optogenetic strategy avoided prolonged tissue heat-
ing (which could modulate neural activity directly) and prevented
OFC disruption during choice execution and reward consumption
(which could have other influences on decision-making behavior28,29).
In addition, we used a new training paradigm in which exposure to
pairs of cues was limited to the testing context, so that animals would
be unlikely to develop unnatural habitual responses to specific cue
combinations (a phenomenon that could underlie the negative
results observed in some previous studies12–14). This training paradigm
resulted in precise psychometric curve functions that allowed us
to detect subtle impairments in economic decision-making.
Finally, we demonstrated that activity in the OFC was not necessary
for performance of a control task in which the choice component was
selectively removed. This experiment excluded the possibility
that effects were driven by sensory, motor or motivational deficits
induced by optical inhibition. Taken together, these data revealed that
OFC inhibition—restricted to the cue evaluation period—specifically
and potently impaired economic decision-making appropriate to
individual preference.
The OFC has been proposed to function as a cognitive map of the
world, that is, an internal model of the associative and predictive rela-
tionships present in the environment30–34. This hypothesis could unify
several contrasting observations regarding the role of the OFC in dis-
tinct tasks, in which the OFC appears to be specifically required when
individuals must use multiple categories of established knowledge to
guide behavior in new scenarios35–37. Consistent with this hypothesis,
we found that OFC activity is necessary when animals must choose
between differently valued options, only previously experienced
in isolation. Importantly, we observed that OFC inhibition does not
appear to preclude the ability to access value information; for example,
OFC-inhibited animals still respond more rapidly to cues that predict
larger-magnitude rewards in a single-cue control condition. Notably,
this is also a task the animals had never seen before.
These data therefore suggest that OFC activity (and associated
cognitive maps) is specifically recruited when animals must resolve
motivational conflict to guide new decision-making. It should be noted
that the OFC is a large, heterogenous structure consisting of the medial,
ventral, ventrolateral, lateral and dorsolateral orbital areas33,38. In this
study, we specifically targeted the ventrolateral orbital area due to its
reported role in supporting flexible behavior39–42. In the future, it will
be important to determine how these results compare to inactivation
of other orbitofrontal subregions and how future results relate to
established differences in anatomical connectivity across mediolateral
and anterior-posterior gradients33,38.
Fig. 3 | Activity in the OFC and DMS encodes spatial features of economic
decision-making. a, Upper: heatmap of z-scored firing rates, averaged across
trials, for each task-modulated unit in the OFC (top) or DMS (bottom). Lower:
population-averaged z-scored firing rates. b, Tuning of example single units
recorded in the OFC (top) or DMS (bottom). Left: trial-averaged peristimulus
time histograms. Right: violin plots of peak z-score. Different shades of blue and
red correspond to different trial types. The center dot represents the median; the
bars represent the first and third quartiles. c, Proportion of units significantly
modulated by different task features in the OFC (top) or DMS (bottom). OFC and
DMS activity properties are similarly dominated by spatial features. d, Linear
decoding approach. e, Chosen side classification accuracy of fourfold cross-
validated SVM trained on single-unit activity in the OFC (blue) or DMS (red).
Increase in decoding accuracy in the OFC precedes increase in decoding accuracy
in the DMS (n = 656 units per brain area from six rats). f, Left: predicted choice
parameter computed on single trials from an example single session in which
activity in the OFC (top) and DMS (bottom) was recorded simultaneously. The
predicted choice parameter was defined as the perpendicular distance of the
decoded decision value from the support vector. Predicted choice parameters
were aligned to choice response and color-coded according to the side chosen by
the animal on each trial. Right: histograms of predicted choice parameters at the
time when choice was indicated (n = 132 trials, n = 11 OFC units, n = 11 DMS units)
g, Single-session mean peak cross-correlation lags of the OFC and DMS predicted
choice parameters on trials in which the animal chose the larger available reward
(correct trials, green) and trials in which the animal chose the smaller available
reward (incorrect trials, gray) (n = 30 sessions from five rats, two-sided paired
t-test; the black lines denote the means). **P < 0.01. Unless otherwise noted,
data are presented as the mean ± s.e.m. Full statistical details are shown in
Supplementary Table 1.
Nature Neuroscience | Volume 26 | September 2023 | 1566–1574
1570
Articlehttps://doi.org/10.1038/s41593-023-01409-1 In contrast to previous observations of nonhuman primates mak-
ing economic decisions1,6, which have consistently demonstrated that
task variables are represented in the OFC in goods (that is, resource)
space, our data suggest that the rodent OFC has a critical role in making
decisions in action space16,43. Consistent with this idea, we observed that
decision-related variables are represented in the rat OFC in a spatially
mapped manner. Moreover, although optogenetic inhibition of the OFC
did not influence behavior in animals presented with a single sensory
a
OFC: 1,157/1,329 tasks
modulated units, n = 6 rats
b
500 ms
Example OFC units
e
r
o
c
s
-
z
k
a
e
P
e
r
o
c
s
-
z
k
a
e
P
14
7
0
10
5
0
1
432
5
6
Total available
reward
10
2
3
Left offer
amount
1
–1
z
-
s
c
o
r
e
Cue
Choice
DMS: 524/656 tasks
modulated units, n = 6 rats
Cue
c
Choice
Reward
OFC summary
Whole trial
Cue
Choice
Reward
Chosen side
0.60
0.16 0.32 0.38
Chosen juice
0.05
0
0 0.05
Choice difficulty
0.03
0.01
0 0.02
Chosen amount
0.12
0.02 0.01 0.09
Rejected amount
0.06
0.02 0.03 0.03
Total amount
0.07
0.03 0.01 0.04
Left offer amount
0.29
0.08 0.11 0.16
Right offer amount
0.24
0.05 0.10 0.14
DMS summary
Whole trial
Cue
Choice
Reward
Chosen side
0.47
0.13 0.25 0.25
Chosen juice
0.06
0
0 0.06
Choice difficulty
0.04
0.01 0.01 0.03
Chosen amount
0.11
0.01 0.02 0.09
Rejected amount
0.06
0.02 0.03 0.02
Total amount
0.08
0.03 0.01 0.03
Left offer amount
0.25
0.08 0.10 0.13
Example DMS units
500 ms
1
–1
z
-
s
c
o
r
e
500 ms
Cue
Choice
e
r
o
c
s
-
z
k
a
e
P
e
r
o
c
s
-
z
k
a
e
P
16
8
0
16
8
0
10
2
3
Right offer
amount
Left Right
Chosen side
SVM:
Train 75%
trials
Test 25%
trials
OFC
Cue
Choice
DMS
Reward
0
1
2
–1
0
0
1
2
3
4
5
Time (s)
d
e
y
c
a
r
u
c
c
a
n
o
i
t
a
c
i
f
i
s
s
a
l
C
)
s
l
a
i
r
t
t
h
g
i
r
s
u
s
r
e
v
t
f
e
l
(
f
1.0
0.9
0.8
0.7
0.6
0.5
OFC
Choice
Left choice
DMS
Right choice
Left choice
i
r
e
t
e
m
a
r
a
p
e
c
o
h
c
d
e
t
c
d
e
r
P
i
**
g
DMS
lead
250
0
–250
)
s
m
(
g
a
L
OFC
lead
Right offer amount
0.20
0.04 0.09 0.11
500 ms
Right choice
Nature Neuroscience | Volume 26 | September 2023 | 1566–1574
1571
Articlehttps://doi.org/10.1038/s41593-023-01409-1
a
c
AAV8 hSyn:NpHR-NRN
(bilateral)
d
OFC
DMS
Mediodorsal
thalamus
1.0
0.5
b
d
e
l
e
b
a
l
n
o
i
t
c
a
r
F
0
Contralateral OFC
Motor cortex
Ipsilateral DMS
Contralateral DMS
Lateral hypothalamus
Substantia nigra
Mediodorsal thalamus
Ventromedial thalamus
5–10-s ITI
2-s cue
Choice
OFC → DMS inhibition
e
)
t
n
a
r
r
u
c
k
c
a
l
b
e
s
o
o
h
c
(
P
h
)
t
n
a
r
r
u
c
k
c
a
l
b
e
s
o
o
h
c
(
P
1.0
0.8
0.6
0.4
0.2
0
1.0
0.8
0.6
0.4
0.2
0
No inhibition
Inhibition
**
***
***
**
***
**
***
1.0
0.8
0.6
y
c
a
r
u
c
c
A
1.0
0.5
f
i
)
s
(
e
c
o
h
c
o
t
y
c
n
e
t
a
l
e
v
i
t
a
l
e
R
g
n
o
i
t
i
b
h
n
i
i
**
e
r
o
c
s
e
c
n
e
r
e
f
e
r
P
–3
–2
–1
1
2
3
0
–
0
1
3
2
–
1
0
–1
R = 0.94
**P = 0.002
–1
0
1
Preference score no inhibition
OFC → mediodorsal thalamus inhibition
i
i
)
s
(
e
c
o
h
c
o
t
y
c
n
e
t
a
l
e
v
i
t
a
l
e
R
1.0
0.5
0
1
3
2
–
j
n
o
i
t
i
b
h
n
i
i
e
r
o
c
s
e
c
n
e
r
e
f
e
r
P
1
0
–1
1.0
0.8
0.6
y
c
a
r
u
c
c
A
–3
–2
–1
1
2
3
0
–
R = 0.90
P = 0.02
–1
0
1
Preference score no inhibition
Fig. 4 | Activity of the projection from the OFC to the DMS is necessary for
economic decision-making. a, Photograph of representative intact rat brain
before (left) and after (middle) clearing. Right: brain-wide axonal projections of
oScarlet-expressing cell bodies located in the OFC. Scale bar, 1 cm. Inset, coronal
section of cell bodies located in the OFC. Scale bar, 1 mm. b, Quantification of
brain-wide axonal projections of oScarlet-expressing cell bodies located in the
OFC, n = 3 rats. c, Schematic of surgical preparation for inhibiting OFC axonal
terminals during economic decision-making. d, Schematic of optical inhibition
during the cue evaluation period of the economic decision-making task.
e–j, Inhibiting the OFC projection to the DMS, but not the OFC projection to the
mediodorsal thalamus, impairs economic decision-making. e,h, Probability of
choosing the blackcurrant-predictive cue as a function of the difference in the
volume of available rewards for uninhibited (green) and inhibited (magenta)
trials. Rats were less likely to choose larger volume rewards when the OFC
projection to the DMS was inhibited (e) but not when the OFC projection to the
mediodorsal thalamus was inhibited (h) (OFC-DMS: n = 7 rats; OFC-mediodorsal
thalamus: n = 6 rats, two-way repeated-measures ANOVA). Inset, fraction of trials
in which the animal chose the larger available reward on uninhibited (green)
and inhibited (magenta) trials (two-sided paired t-test). f,i, Latency to choice
nosepoke response as a function of the absolute difference in the size of rewards
available on uninhibited (green) and inhibited (magenta) trials. Rats were slower
to respond when the difference in reward volume was high (easy trials), when the
OFC projection to the DMS was inhibited (f). Inhibition of the OFC projection to
the mediodorsal thalamus (i) did not alter response latency (OFC-DMS: n = 7 rats;
OFC-MD: n = 6 rats, two-way repeated-measures ANOVA). g,j, Juice preferences
computed on trials in which the OFC projection to the DMS (g) or mediodorsal
thalamus (j) was inhibited were correlated with juice preferences computed on
trials in which the OFC projection to the DMS or mediodorsal thalamus was not
inhibited (Pearson correlation). **P < 0.01, ***P < 0.001. Data are presented as the
mean ± s.e.m. Full statistical details are shown in Supplementary Table 1.
Nature Neuroscience | Volume 26 | September 2023 | 1566–1574
1572
Articlehttps://doi.org/10.1038/s41593-023-01409-1
cue eliciting a single action in the control task, optogenetic inhibition
profoundly impaired behavior when animals were presented with
the same single cue to guide decision-making between two different
actions in the choice task (for example, three drops of blackcurrant
juice reward versus no reward). Taken together, these data suggest
that OFC activity in rodents is specifically recruited when animals must
make choices between differently valued actions. Moving forward, it
will be important to determine whether this reflects a fundamental dif-
ference in processing across species or is due to the different demands
of the specific tasks used44 (for example, the freely moving task used
in this study might necessitate a more detailed representation of the
spatial environment than the head-restrained tasks that have typically
been used in nonhuman primates).
In contrast to the role of the OFC itself, the role of OFC outputs
to other brain regions in value-based decision-making has been less
comprehensively characterized. Previous studies showed that OFC
projections to the ventral tegmental area can mediate aspects of appro-
priate credit assignment45, projections to basolateral amygdala from
lateral or medial OFC can mediate encoding and retrieval of values
respectively46,47, and OFC projections to both the dorsal and ventral
striatum are important for using outcomes to update the value of spe-
cific actions42,48–53. In this study, we expanded on this work and showed
for the first time that the direct transmission of choice information
from the OFC to the DMS, a region implicated in the generation of
goal-directed actions21,23,42,49,54–60, is important for the evaluation of
different reward options before any outcome is delivered. Moreover,
by demonstrating that activity of the same projection is not required
for performance of a control task in which we selectively removed
motivational conflict, we confirm that this deficit in decision-making
behavior is not due to a general failure to recall outcomes that specific
cues predict50.
Surprisingly, while inhibition of the OFC disrupts choices based on
both reward size (objective value) and reward type (subjective value),
inhibition of either the DMS or the projection from the OFC to DMS
only disrupts choices based on reward size (objective value). In addi-
tion, we found that neurons in the OFC are more strongly modulated
by subjective value than objective value, an effect that is not observed
in the DMS. These data suggest that an additional pathway out of the
OFC may also contribute to decision-making about different types of
reward. In the future, it will be important to identify how distinct OFC
projections function in concert to support different components of
decision-making. Taken together, these data provide new insight into
how choices encoded in the OFC engage downstream neural circuits
to generate appropriate behavioral responses.
Economic decision-making requires animals to compare the
subjective value of sensory stimuli to guide appropriate behavior.
To achieve this goal, sensory representations must be imbued with
subjective value information, compared and used to engage neural
circuits that generate appropriate behavioral responses. In this study
we report that the projection from the OFC to the DMS ultimately
connects sensory representations to appropriate behavioral output,
to implement accurate economic decisions. Thus, the OFC projection
to the DMS provides a critical anatomical substrate through which
cortical representations exert dynamic control over ongoing behavior.
Online content
Any methods, additional references, Nature Portfolio reporting sum-
maries, source data, extended data, supplementary information,
acknowledgements, peer review information; details of author contri-
butions and competing interests; and statements of data and code avail-
ability are available at https://doi.org/10.1038/s41593-023-01409-1.
References
1.
Padoa-Schioppa, C. Neurobiology of economic choice: a
good-based model. Annu. Rev. Neurosci. 34, 333–359 (2011).
2. Pearson, J. M., Watson, K. K. & Platt, M. L. Decision-making: the
neuroethological turn. Neuron 82, 950–965 (2014).
3. Sugrue, L. P., Corrado, G. S. & Newsome, W. T. Choosing the
greater of two goods: neural currencies for valuation and
decision-making. Nat. Rev. Neurosci. 6, 363–375 (2005).
4. Padoa-Schioppa, C. & Assad, J. A. Neurons in the orbitofrontal
cortex encode economic value. Nature 441, 223–226 (2006).
5. Ballesta, S., Shi, W., Conen, K. E. & Padoa-Schioppa, C. Values
encoded in orbitofrontal cortex are causally related to economic
choices. Nature 588, 450–453 (2020).
6. Fellows, L. K. Orbitofrontal contributions to value-based
decision-making: evidence from humans with frontal lobe
damage. Ann. N. Y. Acad. Sci. 2139, 51–58 (2011).
7. Kable, J. W. & Glimcher, P. W. The neurobiology of decision:
consensus and controversy. Neuron 63, 733–745 (2009).
8. Levy, D. J. & Glimcher, P. W. The root of all value: a common
currency for choice. Curr. Opin. Neurobiol. 22, 1027–1038 (2012).
9. Murray, E. A. & Rudebeck, P. H. Specializations for reward-guided
decision-making in the primate ventral prefrontal cortex.
Nat. Neurosci. 19, 404–417 (2018).
10. Shadlen, M. N. & Shohamy, D. Decision-making and sequential
sampling from memory. Neuron 90, 927–939 (2016).
11. Wallis, J. D. Cross-species studies of orbitofrontal cortex and
value-based decision-making. Nat. Neurosci. 15, 13–19 (2011).
12. Gardner, M. P. H., Conroy, J. S., Shaham, M. H., Styer, C. V. &
Schoenbaum, G. Lateral orbitofrontal inactivation dissociates
devaluation-sensitive behavior and economic choice. Neuron 96,
1192–1203 (2017).
13. Gardner, M. P. H., Conroy, J. C., Sanchez, D. C., Zhou, J. &
Schoenbaum, G. Real-time value integration during economic
choice is regulated by orbitofrontal cortex. Curr. Biol. 29,
4315–4322 (2019).
14. Gardner, M. P. H. et al. Processing in lateral orbitofrontal cortex is
required to estimate subjective preference during initial, but not
established, economic choice. Neuron 108, 526–537 (2020).
Izquierdo, A., Suda, R. K. & Murray, E. A. Bilateral orbital prefrontal
cortex lesions in rhesus monkeys disrupt choices by both reward
value and reward contingency. J. Neurosci. 24, 7540–7548 (2004).
15.
16. Kuwabara, M., Kang, N., Holy, T. E. & Padoa-Schioppa, C. Neural
mechanisms of economic choices in mice. eLife 9, e49669
(2020).
17. Kable, J. W. & Glimcher, P. W. The neural correlates of subjective
value during intertemporal choice. Nat. Neurosci. 10, 1625–1633
(2007).
18. Cai, X., Kim, S. & Lee, D. Heterogeneous coding of temporally
discounted values in the dorsal and ventral striatum during
intertemporal choice. Neuron 69, 170–182 (2011).
19. Orona, E. & Gabriel, M. Multiple-unit activity of the prefrontal
cortex and mediodorsal thalamic nucleus during reversal learning
of discriminative avoidance behavior in rabbits. Brain Res. 263,
313–329 (1983).
20. Verharen, J. P. H., den Ouden, H. E. M., Adan, R. A. H. &
Vandershchuren, L. J. M. J. Modulation of value-based
decision-making behavior by subregions of the rat prefrontal
cortex. Psychopharmacology 237, 1267–1280 (2020).
21. Friedman, A. et al. A corticostriatal path targeting striosomes
controls decision-making under conflict. Cell 161, 1320–1333
(2015).
22. Schumacher, J. D., van Holstein, M., Bagrodia, V., Le Bouder, H. B.
& Floresco, S. B. Dorsomedial striatal contributions to different
forms of risk/reward decision making. Neurobiol. Learn. Mem. 178,
107369 (2021).
23. Yin, H. H., Ostlund, S. B., Knowlton, B. J. & Balleine, B. W. The role
of the dorsomedial striatum in instrumental conditioning. Eur. J.
Neurosci. 22, 513–523 (2005).
Nature Neuroscience | Volume 26 | September 2023 | 1566–1574
1573
Articlehttps://doi.org/10.1038/s41593-023-01409-1 24. Chakraborty, S., Kolling, N., Walton, M. E. & Mitchell, A. S.
46. Malvaez, M., Shieh, C., Murphy, M. D., Greenfield, V. Y. & Wassum, K. M.
Critical role for the mediodorsal thalamus in permitting rapid
reward-guided updating stochastic reward environments. eLife 5,
e13588 (2016).
25. Berndt, A. et al. Structural foundations of optogenetics:
determinants of channelrhodopsin ion selectivity. Proc. Natl Acad.
Sci. USA 26, 822–829 (2016).
26. Hoover, W. B. & Vertes, R. P. Projections of the medial orbital and
ventral orbital cortex in the rat. J. Comp. Neurol. 519, 3766–3801
(2011).
27. Ye, L. et al. Wiring and molecular features of prefrontal ensembles
representing distinct experiences. Cell 165, 1776–1788 (2016).
28. Constantinople, C. M. et al. Lateral orbitofrontal cortex promotes
trial-by-trial learning of risky, but not spatial biases. eLife 8,
e49744 (2019).
29. Jennings, J. H. et al. Interacting neural ensembles in the
orbitofrontal cortex for social and feeding behaviour. Nature 565,
645–649 (2019).
30. Stalnaker, T. A., Cooch, N. K. & Schoenbaum, G. What the
orbitofrontal cortex does not do. Nat. Neurosci. 18, 620–627 (2015).
Distinct cortical-amygdala projections drive reward value
encoding and retrieval. Nat. Neurosci. 22, 762–769 (2019).
47. Lichtenberg, N. et al. The medial orbitofrontal cortex-basolateral
amygdala circuit regulates the influence of reward cues on
adaptive behavior and choice. J. Neurosci. 41, 7267–7277 (2021).
48. Groman, S. M. et al. Orbitofrontal circuits control multiple
reinforcement-learning processes. Neuron 103, 734–746 (2019).
49. Gremel, C. M. et al. Endocannabinoid modulation of orbitostriatal
circuits gates habit formation. Neuron 90, 1312–1324 (2016).
50. Li, D. C. et al. A molecularly integrated amygdalo-fronto-striatal
network coordinates flexible learning and memory. Nat. Neurosci.
25, 1213–1224 (2022).
51. Jenni, N. L., Rutledge, G. & Floresco, S. B. Distinct medial
orbitofrontal-striatal circuits support dissociable component
processes of risk/reward decision-making. J. Neurosci. 42,
2743–2755 (2022).
52. Hirokawa, J., Vaughan, A., Masset, P., Ott, T. & Kepecs, A. Frontal
cortex neuron types categorically encode single decision
variables. Nature 576, 446–451 (2019).
31. Schuck, N. W., Cai, M. B., Wilson, R. C. & Niv, Y. Human
53. Pascoli, V. et al. Stochastic synaptic plasticity underlying
orbitofrontal cortex represents a cognitive map of state space.
Neuron 91, 1402–1412 (2016).
compulsion in a model of addiction. Nature 564, 366–371 (2018).
54. Bissonette, G. B. & Roesch, M. R. Rule encoding in dorsal striatum
32. Wilson, R. C., Takahashi, Y. K., Schoenbaum, G. & Niv, Y. Orbitofrontal
cortex as a cognitive map of task space. Neuron 81, 267–279 (2014).
33. Bradfield, L. A. & Hart, G. Rodent medial and lateral orbitofrontal
cortices represent unique components of cognitive maps of task
space. Neurosci. Biobehav. Rev. 108, 287–294 (2020).
34. Behrens, T. E. J. et al. What is a cognitive map? Organizing
knowledge for flexible behavior. Neuron 100, 490–509 (2018).
35. Jones, J. J. et al. Orbitofrontal cortex supports behavior and
learning using inferred but not cached values. Science 338,
2743–2770 (2012).
36. Stalnaker, T. A. et al. Orbitofrontal neurons infer the value and
identity of predicted outcomes. Nat. Commun. 5, 3926 (2014).
37. Vertechi, P. et al. Inference-based decisions in a hidden state
foraging task: differential contributions of prefrontal cortical
areas. Neuron 106, 166–176 (2020).
38. Izquierdo, A. Functional heterogeneity within rat orbitofrontal
cortex in reward learning and decision making. J. Neurosci. 37,
10529–10540 (2017).
39. Chase, E. A., Tait, D. S. & Brown, V. J. Lesions of the orbital
prefrontal cortex impair the formation of attentional set in rats.
Eur. J. Neurosci. 36, 2368–2375 (2012).
40. Meyer, H. C. & Bucci, D. J. Imbalanced activity in the orbitofrontal
cortex and nucleus accumbens impairs behavioral inhibition.
Curr. Biol. 26, 2834–2839 (2016).
41. Parkes, S. L. et al. Insular and ventrolateral orbitofrontal cortices
differentially contribute to goal-directed behavior in rodents.
Cereb. Cortex 28, 2313–2325 (2018).
42. Gremel, C. M. & Costa, R. M. Orbitofrontal and striatal circuits
dynamically encode the shift between goal-directed and habitual
actions. Nat. Commun. 4, 2264 (2013).
43. Feierstein, C. E., Quirk, M. C., Uchida, N., Sosulski, D. L. &
Mainen, Z. F. Representation of spatial goals in rat orbitofrontal
cortex. Neuron 51, 495–507 (2006).
44. Yoo, S. B. M., Sleezer, B. J. & Hayden, B. Y. Robust encoding of
spatial information in the orbitofrontal cortex and striatum.
J. Cogn. Neurosci. 30, 898–913 (2018).
45. Namboodiri, V. M. K. et al. Single-cell activity tracking reveals that
orbitofrontal neurons acquire and maintain a long-term memory
to guide behavioral adaptation. Nat. Neurosci. 22, 1110–1121 (2019).
impacts action selection. Eur. J. Neurosci. 42, 2555–2567 (2015).
55. Cromwell, H. C. & Schultz, W. Effects of expectations for different
reward magnitudes on neuronal activity in the primate striatum.
J. Neurophysiol. 22, 2823–2838 (2003).
56. Cui, G. et al. Concurrent activation of striatal direct and indirect
pathways during action initiation. Nature 494, 238–242
(2013).
57. Hikosaka, O., Sakamoto, M. & Usui, S. Functional properties of
monkey caudate neurons. III. Activities related to expectation of
target and reward. J. Neurophysiol. 61, 814–832 (1989).
58. Hollerman, J., Tremblay, L. & Schultz, W. Influence of reward
expectation of behavior-related neuronal activity in the primate
striatum. J. Neurophysiol. 80, 947–963 (1998).
59. Santacruz, S. R., Rich, E. L., Wallis, J. D. & Carmena, J. M. Caudate
microstimulation increases value of specific choices. Curr. Biol.
27, 3375–3383 (2017).
60. Tai, L.-H., Moses Lee, A., Benavidez, N., Bonci, A. & Wilbrecht, L.
Transient stimulation of distinct subpopulations of striatal
neurons mimics changes in action value. Nat. Neurosci. 15,
1281–1289 (2012).
Publisher’s note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format,
as long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate
if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons license and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this license, visit http://creativecommons.
org/licenses/by/4.0/.
© The Author(s) 2023
Nature Neuroscience | Volume 26 | September 2023 | 1566–1574
1574
Articlehttps://doi.org/10.1038/s41593-023-01409-1 Methods
Experimental procedures were approved by the Stanford University
Institutional Animal Care and Use Committee and by the Administrative
Panel on Laboratory Animal Care (protocol no. 32908), according to
the National Institutes of Health (NIH) guidelines for the care and use
of laboratory animals.
Experimental animals and stereotactic surgery
Adult (10–12 weeks) male and female Long–Evans rats (Charles River
Laboratories) were group-housed until surgery. Rats were randomly
assigned to different experimental groups. Animals were anesthetized
with isoflurane (1–5%, Henry Schein) and placed into a stereotactic
frame (Kopf Instruments). Bone screws (Stoelting Co.) were inserted.
For the optogenetic experiments, microinjection needles (WPI) were
then inserted (coordinates from bregma: OFC +4 anteroposterior, ±2
mediolateral, −3 dorsoventral; prelimbic cortex +2.5 anteroposterior,
±0.5 mediolateral, −3.5 dorsoventral; DMS +1 anteroposterior, ±2.5
mediolateral, −4 dorsoventral; mediodorsal thalamus −2.8 anteropos-
terior, ±0.8 mediolateral, −5 dorsoventral; note that the dorsoventral
coordinates reflect the distance from the brain surface) and each struc-
ture was injected with virus at a speed of 0.1 μl min. A 200-μm diameter
optical fiber (Thorlabs) was placed 250 μm above the target sites and
fixed in place using dental cement (RelyX, 3M). For the electrophysio-
logical recordings, 64-channel silicon probes (Cambridge NeuroTech)
were mounted on a microdrive and lowered to 500 μm above the site of
interest. Craniotomies were sealed with Dura-Gel and microdrives were
fixed in placed using dental cement. Molex connectors were attached to
a wireless headstage (White Matter LLC), which was affixed to the skull
with dental cement. Probes were lowered to the recording site 2 days
before recordings. Buprenorphine SR (1 mg kg−1) was administered.
As an exclusion criterion, we only included rats with viral expression
confined to the site of interest and fiber placement above the target
site. (This resulted in the exclusion of one animal.) All experiments were
conducted according to approved protocols at Stanford University.
Rat behavior
Water scheduled rats (1 h of water per day) were placed into a custom
operant chamber equipped with three nosepoke portals mounted on
a screen. The center portal was equipped with a lick spout for reward
delivery. Entries into each nosepoke portal were detected by the break-
age of an infrared beam and licks were detected using a capacitive touch
sensor. (This was omitted for the electrophysiological recordings.)
All events were controlled and recorded using custom MATLAB code
using the MATLAB Support Package for Arduino and the Psychophysics
Toolbox v.3 (ref. 61). For training, animals were placed into the oper-
ant chamber. One second after entering the center portal they were
presented with a visual cue on one side of the center portal. The type
of cue (vertical or horizontal drifting gratings) indicated the type of
reward associated with the cue (zero calorie blackcurrant-flavored or
lemon-flavored water); the number of squares that included the cue
indicated the size of reward associated with the cue. Lemon-predictive
and blackcurrant-predictive cues could be presented on either the left
or right side of the animal, randomized for each trial. After 2 s, animals
had to perform a nosepoke to the side the cue was presented to obtain
the corresponding reward. Reward was delivered in the center portal.
Reward collection was followed by a variable intertrial interval (ITI) of
5–10 s. If animals responded to the wrong side, no reward was delivered
and the screen turned white for a 10-s time-out period. This taught
animals to move to the side of the cue to indicate the response and to
reinforce contingency. Trials in which animals took more than 12 s to
indicate a response, and trials in which the animal took more than 5 s
to collect the reward, were excluded.
When animals had achieved criterion performance (> 90% accu-
racy and response latency inversely proportional to reward magni-
tude on three consecutive sessions; each stimulus was comparably
Nature Neuroscience
learned as shown in Extended Data Fig. 1a), they were placed into a
full choice session. Animals were placed into the operant chamber;
1 s after entering the center portal, animals were presented with two
visual cues side by side. Lemon-predictive or blackcurrant-predictive
cues could be presented on either the left or right side of the animal,
randomized for each trial. After 2 s, animals had to move to the side
of the chosen cue to indicate their choice, and the chosen reward
was delivered in the center portal. Reward collection was followed
by a variable ITI of 5–10 s. Trials in which animals took more than 12 s
to indicate choice, and trials in which animals took more than 5 s to
collect the reward, were excluded. If animals performed at more than
75% accuracy (as animals made choices primarily to maximize the total
volume of liquid consumed, accuracy was defined as the proportion
of trials wherein animals selected the larger available reward), the
following day animals were placed into another full choice session
(for a maximum of three consecutive full choice sessions). For the
choice sessions, a total of 15 cue combinations were used; each ses-
sion was terminated after 600 trials or after 2.5 h, whichever came
first. Otherwise, animals were placed back into training sessions until
reachieving criterion performance. Summary data are presented as a
composite of three consecutive full choice sessions per rat. Behavio-
ral data were fitted by probit regression using the glmfit function in
MATLAB. Preference scores were computed by calculating the differ-
ence in available reward (number of drops of blackcurrant − number
of drops of lemon) for which the animal was equally likely to choose a
blackcurrant-predictive or lemon-predictive cue. Long-term prefer-
ence comparisons were between the preference score from the final
three consecutive full choice sessions before a 4-month university
shutdown, and the preference score from the first three consecutive
full choice sessions after the 4-month shutdown. Comparisons of
short-term preferences were performed on preference scores from
each of three consecutive sessions. Latency to choice was calculated
by finding the mean latency from the end of the mandatory 2-s cue
presentation period, to the time at which the animal made its nosepoke
response for each trial type. For each animal, we then subtracted the
trial type with the fastest mean response time from all other trial types
to obtain a relative latency to choice.
We carried out control behavior to account for the nonspecific
effects of optical inhibition. Animals were placed into an operant
chamber equipped with two nosepoke portals mounted on a screen;
the left portal was equipped with a lick spout for reward delivery. One
second after entering the left portal, animals were presented with
a single visual cue in the center of the portal. (The same visual cues
were also used for training and the full choice task.) After 2 s, animals
had to perform a nosepoke in the second portal to indicate response.
Reward was delivered in the left portal. Reward collection was followed
by a variable ITI of 5–10 s. Trials in which animals took more than 12 s
to indicate response, and trials in which the animal took more than 5 s
to collect the reward, were excluded.
Optogenetic inhibition
Rats were placed into the operant chamber and a top-branch with a
200-μm diameter fiber-optic patch cord (Doric) coupled to either a
473 nm (Omicron) and 635 nm (CNI), or a 594 nm (Cobalt), laser setup
outside of the operant chamber connected to the implanted optical
fibers. Immediately beforehand, power output from the patch cord
was adjusted to 8 mW (473 nm), 5 mW (635 nm) or 10 mW (594 nm).
Animals received randomly interleaved presentations of inhibited and
uninhibited trials. On the SwiChR++ inhibition trials, 1 s of 473-nm light
stimulation to initiate inhibition was delivered when the visual stimuli
were presented; 1 s of 635 nm light stimulation to relieve inhibition
was delivered when the animal exited the center portal to indicate its
choice. On the halorhodopsin inhibition trials, 594-nm light stimulation
was initiated when the visual stimuli were presented and terminated
when the animal exited the center portal to indicate choice.
Articlehttps://doi.org/10.1038/s41593-023-01409-1 Chronic electrophysiology
Animals were implanted with 64-channel silicon probes over the right
DMS and right OFC. On the day of implantation, electrodes were lowered
to 500 μm above the site of interest. Animals were allowed to recover for
2–3 weeks before behavioral training was resumed. Microdrives were
lowered by 250 μm 2 days before each recording session. Electrophysi-
ological data were acquired at 20 kHz using a wireless acquisition system
(White Matter LLC). Recordings were made in freely moving rats, which
may impact the degree of lateralization of the neural responses observed.
Behavioral time stamps were acquired at 30 kHz using an Open Ephys
acquisition system. Clocks were synchronized by sending a signal on
every Open Ephys sample to the White Matter LLC acquisition system.
Acute electrophysiology
Animals expressing SwiChR++ in the OFC were anesthetized with isoflu-
rane and placed into a stereotactic frame. A craniotomy was placed over
the OFC and a custom optrode (200-μm fiber cemented onto a silicone
probe) was inserted into the region of the infected cells. Recordings
were made using an Open Ephys acquisition system applying a bandpass
filter from 300 to 6,000 Hz to the voltage signal. A 1-s pulse of blue light
(473 nm, 8 mW) was delivered to initiate inhibition and a 1-s pulse of red
light (635 nm, 5 mW) was delivered to alleviate inhibition 4 s later. Laser
timing was controlled by a Master-8 pulse generator (AMPI).
Electrophysiology data analysis
Spikes were sorted using Kilosort2 and were manually curated using
Phy2 (ref. 62). Units with less than 1% inter-spike intervals shorter than
2 ms were considered single units for the analysis purposes. Spike
counts were binned in 50-ms bins, stepped at 25-ms increments and
converted into a z-scored firing rate across the whole session. Z-scored
firing rates were aligned to task events (cue presentation, choice nose-
poke and reward delivery) and the mean firing rate in the 500 ms before
cue presentation was subtracted on a per trial basis. Task-modulated
units were identified based on a Wilcoxon rank-sum test of the mean
firing rate within the 500-ms baseline and ten 500-ms epochs span-
ning the trial starting at cue onset. A cell was deemed task-modulated
if any of the task epochs differed significantly from baseline after false
discovery rate correction, with a corrected significance threshold of
P < 0.001. For each neural response, we performed a linear regression
against each of a set of ten predefined variables (separately). For sub-
jective value regression, preference scores were calculated for each
session by finding the difference in the available reward at which the
animal was equally likely to choose blackcurrant and lemonade. This
score was then added to the volume of lemonade available on each trial
to generate subjective value predictors. Units were deemed modulated
by the variable if the regression slope differed significantly from zero
(correct significance threshold of P < 0.001).
Decoding analysis was performed using a fourfold cross-validated
linear SVM63. Classification accuracy was calculated as the fraction
of correct predictions made on held-out data averaged across four
cross-validation splits, repeated five times. For the single-trial analysis,
predicted choice parameters were computed as the perpendicular
distance of decision value from the support vector at each time point,
repeated across four cross-validation splits. Cross-correlations of the
predicted choice parameters were calculated in the 3 s surrounding the
choice nosepoke and averaged across 20 decoding repeats per session.
Single-trial predicted choice parameters were smoothed with a 50-ms
Gaussian filter for analysis and a 250-ms Gaussian filter for visualiza-
tion. For the latency analysis, an arbitrary threshold of 0.33 was set.
For all decoding analysis, the numbers of units across brain areas were
matched to the size of the smallest recorded population.
100-μm sections were cut on a vibratome. Slices were labeled with goat
anti-GFP (1:1,000, Abcam) primary antibody and Alexa Fluor 488 donkey
anti-goat (1:1,000, Invitrogen). For the axon tracing studies, a micro-
injection needle was inserted into the brain (coordinates from bregma:
+4 anteroposterior, ±2 mediolateral, −3 dorsoventral) and 0.5 μl AAV8
hSyn:oScarlet was injected into the OFC at a speed of 0.1 μl min−1. At least
8 weeks later, brains were prepared for histology and axonal projections
were quantified as described previously64. Briefly, 100-μm coronal slices
were imaged on a confocal microscope (ZEISS, Zen software) using a
×20 objective and the resultant images were processed in ImageJ for
quantification. Briefly, the injection site was first manually removed and
background was subtracted. Threshold was set to ×4 the mean of the
local background and pixels above this threshold were interpreted as
positive signal from the OFC axons. Region of interest (ROI) boundaries
were manually defined based on 4,6-diamidino-2-phenylindole staining
and the Paxinos and Watson rat brain atlas65. Axon density was calculated
as the percentage of total ROIs containing pixels above the threshold.
Three sections per ROI were analyzed and those values were averaged to
calculate a single value per ROI per rat. This approach cannot distinguish
between axon terminal and fibers of passage.
Whole-brain clearing
Adult Long–Evans rats were anesthetized with isoflurane and placed
into a stereotactic frame. A microinjection needle was inserted into the
brain (coordinates from bregma: +4 anteroposterior, ±2 mediolateral, −3
dorsoventral) and 0.5 μl AAV8 hSyn:oScarlet was injected into the OFC
at a speed of 0.1 μl min−1. Eight weeks later, brains prepared for imaging
using SHIELD66. Briefly, rats were euthanized by transcardial perfusion
with 150 ml PBS, followed by 100 ml 4% paraformaldehyde, followed by
50 ml 12% epoxide SHIELD perfusion solution. Brains were extracted
and incubated in SHIELD perfusion solution at 4 °C for 48 h. Brains were
removed from SHIELD perfusion solution and transferred to SHIELD OFF
solution and incubated at 4C for 48 h. Brains were then transferred to
SHIELD ON solution and incubated at 37 °C for 24 h. After completion
of the SHIELD reaction, brains were transferred to SDS clearing solution
and cleared passively at 37 °C for 7 days, before being transferred to a
SmartClear system for active clearing for 10–14 days. When brains were
clear, they were washed in 0.1% PBS with Tween 20 at 37 °C for 3 days,
before being equilibrated in exPROTOS at room temperature for 2 days
and imaged using a COLM light sheet microscope67,68.
Statistics and reproducibility
Data are presented as the mean ± s.e.m. unless otherwise indicated.
Raw data were tested for normality of distribution; statistical analyses
were performed using Student’s t-test, Wilcoxon signed-rank test
or ANOVA with Bonferroni correction for multiple comparisons.
Statistical analyses were performed in Prism (GraphPad Software) and
MATLAB (MathWorks). No statistical method was used to predetermine
sample size, but sample sizes were based on previous studies69. For
practical reasons, data collection and analysis could not be performed
blind to the conditions of the experiments (for example, because of
obviously different positions of the fibers), but data were collected
and analyzed in an automated manner to prevent experimenter bias.
Reporting summary
Further information on research design is available in the Nature Port-
folio Reporting Summary linked to this article.
Data availability
All primary data for the figures and extended data figures are available
from the corresponding author (K.D.) upon request.
Histological processing and analysis
Rats were euthanized by transcardiac perfusion with 150 ml PBS, fol-
lowed by 100 ml 4% paraformaldehyde. Brains were extracted and
Code availability
The code used for data processing and analysis is available from the
corresponding author (K.D.) upon request.
Nature Neuroscience
Articlehttps://doi.org/10.1038/s41593-023-01409-1 References
61. Kleiner, M. et al. What’s new in psychtoolbox-3. Perception 36,
Foundation (formerly National Alliance for Research on Schizophrenia
& Depression) Young Investigator Award (to F.G.).
1–16 (2007).
62. Stringer, C., Pachitariu, M., Steinmetz, N., Carandini, M. &
Harris, K. D. High-dimensional geometry of population responses
in visual cortex. Nature 571, 361–365 (2019).
63. Meyers, E. M. The neural decoding toolbox. Front. Neuroinform. 7,
8 (2013).
64. Beier, K. T. et al. Circuit architecture of VTA dopamine neurons
revealed by systematic input-output mapping. Cell 162, 622–634
(2015).
65. Paxinos, G. & Watson, C. The Rat Brain in Stereotactic Coordinates
(Academic Press, 2007).
66. Park, Y. G. et al. Protection of tissue physicochemical properties
using polyfunctional crosslinkers. Nat. Biotechnol. 37, 73–83
(2019).
67. Tomer, R., Ye, L., Hsueh, B. & Deisseroth, K. Advanced CLARITY for
rapid and high-resolution imaging of intact tissues. Nat. Protoc. 9,
1682–1697 (2014).
68. Lerner, T. N. et al. Intact-brain analyses reveal distinct
information carried by SNc dopamine subcircuits. Cell 162,
635–647 (2015).
69. Zalocusky, K. A. et al. Nucleus accumbens D2R cells signal
Author contributions
F.G., R.C.M. and K.D. conceived the project, designed the experiments
and wrote the manuscript with input from all authors. F.G.
performed the behavioral experiments, optogenetic manipulations,
electrophysiological recordings and data analysis. M.H. assisted with
the behavioral experiments. C.R. designed and produced the viral
constructs. A.K.C. performed the light sheet microscopy. R.C.M. and
K.D. supervised all aspects of the work.
Competing interests
R.C.M. is on the scientific advisory boards of MapLight Therapeutics,
Bright Minds Biosciences, MindMed and Aelis Farma. K.D. is on the
scientific advisory boards of MapLight Therapeutics, Stellaromics,
and Bright Minds Biosciences. The other authors declare no
competing interests.
Additional information
Extended data is available for this paper at
https://doi.org/10.1038/s41593-023-01409-1.
prior outcomes and control risky decision-making. Nature 531,
642–646 (2016).
Supplementary information The online version contains supplementary
material available at https://doi.org/10.1038/s41593-023-01409-1.
Acknowledgements
We thank T. Machado, J. Baruni and members of the Deisseroth
and Malenka laboratories for helpful discussions. This work was
supported by grants from the UCSF Dolby Family Center for Mood
Disorders (to R.C.M.), the NIH (no. P50DA042012 to K.D. and R.C.M.;
no. K99DA050662 to F.G.), National Science Foundation, Gatsby,
Fresenius, Wiegers, Grosfeld and NOMIS Foundations (to K.D.),
a Walter V. and Idun Berry award and a Brain & Behavior Research
Correspondence and requests for materials should be addressed to
Karl Deisseroth.
Peer review information Nature Neuroscience thanks the anonymous
reviewers for their contribution to the peer review of this work.
Reprints and permissions information is available at
www.nature.com/reprints.
Nature Neuroscience
Articlehttps://doi.org/10.1038/s41593-023-01409-1 Extended Data Fig. 1 | Behavioral characterization. a. Proportion of correct
responses to each cue in final 3 training sessions before surgery; cues were
comparably learned, n = 36 rats. b. Reinforcement contingencies where size
of stimulus was not proportional to reward size for probing whether animals
perform value-based or perceptual decision-making. c. Probability of choosing
blackcurrant-predictive cue for all cue combinations for animals training using
reversed reinforcement contingencies (n = 8 rats, one-way repeated-measures
ANOVA). d. Probability of choosing blackcurrant-predictive cue as a function
of the difference in the size of rewards available for animals trained using
reversed reinforcement contingencies (n = 8 rats, one-way repeated-measures
ANOVA). Inset, fraction of trials in which animal chose the larger available
reward (n = 8 rats, 0.79±0.01). e. Latency to choice nosepoke response as a
function of difference in the size of rewards available for animals trained using
reversed reinforcement contingencies (n = 8 rats, one-way repeated-measures
ANOVA). Center dot represents median, bars represent first and third quartile.
f. Correlations of preference scores computed on 3 consecutive sessions.
Preference scores are highly correlated, indicating juice preferences of individual
animals are stable across time (Pearson’s correlation). * P < 0.05, ** P < 0.01,
*** P < 0.001, Unless otherwise noted, data are presented as mean ± SEM.
Nature Neuroscience
Articlehttps://doi.org/10.1038/s41593-023-01409-1 Extended Data Fig. 2 | Fiber placements. a-d. Representative images of EYFP expression and fiber placements in animals injected with AAV8 hSyn:SwiChR++EYFP or
AAV8 hSyn:EYFP in the OFC (a), the prelimbic cortex (b), the DMS (c), or the mediodorsal thalamus (d).
Nature Neuroscience
Articlehttps://doi.org/10.1038/s41593-023-01409-1 Extended Data Fig. 3 | Optogenetic inhibition control experiments.
a-d. Optical stimulation does not alter economic decision-making in animals
expressing EYFP in the OFC. a. Schematic of experimental preparation.
b. Probability of choosing blackcurrant-predictive cue as a function of difference
in the volume of available rewards for no illumination (green) and illumination
(magenta) trials (n = 6 rats, two-way repeated-measures). Inset, fraction of trials
in which animal chose the larger available reward on no illumination (green) and
illumination (magenta) trials (n = 6 rats, two-sided paired t-test). c. Latency to
choice nosepoke response as a function of the absolute difference in the size of
rewards available on no illumination (green) and illumination (magenta) trials
(n = 6 rats, two-way repeated-measures ANOVA). d. Juice preferences computed
on trials with OFC illumination are correlated with juice preferences computed
on trials without OFC illumination (Pearson’s correlation). e. Schematic of
control task for probing whether effects of optical inhibition specifically impact
choice behavior. f, g. Latency to nosepoke response for cues predicting different
size rewards in control no-choice task on trials in which the OFC (f) or DMS (g)
was not inhibited (green) or was inhibited (magenta). OFC or DMS inhibition
did not alter latencies to respond in no-choice control task (OFC: n = 12 rats,
DMS< n = 6 rats, two-way repeated-measures ANOVA). * P < 0.05, ** P < 0.01,
*** P < 0.001, Data are presented as mean ± SEM.
Nature Neuroscience
Articlehttps://doi.org/10.1038/s41593-023-01409-1 Extended Data Fig. 4 | See next page for caption.
Nature Neuroscience
Articlehttps://doi.org/10.1038/s41593-023-01409-1 Extended Data Fig. 4 | Electrophysiology supporting data 1. a. Proportion of
units significantly modulated by distinct task features in the OFC (top) or DMS
(bottom) for each individual animal. b. Left, proportions of neurons significantly
modulated by objective (reward size) and subjective value did not differ across
OFC (blue) or DMS (red). Right, coefficients of determination (R2) of each
modulated unit in either OFC (blue) or DMS (red) when either the subjective
or objective value of the stimuli presented on either left or right were used as
predictors. Black lines denote means. OFC units were more strongly modulated
by subjective value than objective value (n = 107 units per condition, three-way
mixed ANOVA). c. Chosen-side classification accuracy of 4-fold cross validated
support vector machine trained on single unit activity in the OFC (blue) or DMS
(red) for each individual animal (R102 n = 86 units per brain area, R109 n = 141
units per brain area, R116 n = 54 units per brain area, R140 n = 103 units per brain
area, R144, n = 145 units per brain area, R145 n = 133 OFC units, R123 n = 71 DMS
units). * P < 0.05, ** P < 0.01, *** P < 0.001, Data are presented as mean ± SEM.
Nature Neuroscience
Articlehttps://doi.org/10.1038/s41593-023-01409-1 Extended Data Fig. 5 | Electrophysiology supporting data 2. a. Chosen-side
classification accuracy of 4-fold cross validated support vector machine trained
on single unit activity recorded in either the OFC (blue) or DMS (red) on correct
trials, and tested on either held-out correct trials (left) or incorrect trials (center).
Note decoding performance is reduced compared to Fig. 3 due to the relatively
small number of incorrect trials performed. Right, classification accuracy
reaches equivalent levels on both correct and incorrect trials in both the OFC and
DMS (n = 20 random samples, two-way repeated-measures ANOVA). b. Average
predicted choice parameters computed on single correct (left) or incorrect
(right) trials aligned to choice response (n = 30 sessions). c. Peak predicted
choice parameters are equivalent on correct and incorrect trials (n = 20 random
samples, two-way repeated-measures ANOVA). d. Latency to predicted choice
parameter threshold relative to choice on correct or incorrect choice trials for
models trained using data recorded simultaneously from OFC (blues) or DMS
(reds). OFC activity does not precede DMS activity when animals make incorrect
choices (n = 20 random samples, two-way repeated-measures ANOVA). * P < 0.05,
** P < 0.01, *** P < 0.001, Data are presented as mean ± SEM.
Nature Neuroscience
Articlehttps://doi.org/10.1038/s41593-023-01409-1 Extended Data Fig. 6 | Optogenetic axon terminal inhibition control
experiments, related to Fig. 4. a. Schematic of control task for probing whether
effects of optically inhibiting OFC axon terminals specifically influence choice
behavior. b, c. Latency to nosepoke response for cues predicting different size
rewards in control no-choice task on trials in which the projection from OFC
to DMS (b) or OFC to mediodorsal thalamus (c) was not inhibited (green) or
was inhibited (magenta). Inhibition of OFC projections to DMS or mediodorsal
thalamus does not alter latencies to respond in no-choice control task (n = 7 rats,
two-way repeated-measures ANOVA). * P < 0.05, ** P < 0.01, *** P < 0.001, Data are
presented as mean ± SEM.
Nature Neuroscience
Articlehttps://doi.org/10.1038/s41593-023-01409-1 Corresponding author(s):
Karl Deisseroth
Last updated by author(s): 06/19/2023
Reporting Summary
Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.
Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested
A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)
For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.
For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.
Software and code
Policy information about availability of computer code
Data collection
Open Ephys and White Matter (in vivo extracellular electrophysiology acquisition), Matlab 2019b and Psychtoolbox3 (behavioral control and
acquisition), Zen (confocal image acquisition).
Data analysis
Matlab 2019b, GraphPad Prism 9, Kilosort2, Phy2, ImageJ.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.
Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A description of any restrictions on data availability
- For clinical datasets or third party data, please ensure that the statement adheres to our policy
The data that support this paper are available upon request to the corresponding author.
n
a
t
u
r
e
p
o
r
t
f
o
l
i
o
|
r
e
p
o
r
t
i
n
g
s
u
m
m
a
r
y
A
p
r
i
l
2
0
2
3
1
Research involving human participants, their data, or biological material
Policy information about studies with human participants or human data. See also policy information about sex, gender (identity/presentation),
and sexual orientation and race, ethnicity and racism.
Reporting on sex and gender
N/A
Reporting on race, ethnicity, or
other socially relevant
groupings
Population characteristics
Recruitment
Ethics oversight
N/A
N/A
N/A
N/A
Note that full information on the approval of the study protocol must also be provided in the manuscript.
Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences
Behavioural & social sciences
Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf
Life sciences study design
All studies must disclose on these points even when the disclosure is negative.
Sample size
No statistical methods were used to determine sample size a priori. Sample sizes were based on prior studies (Zalocusky et al., 2016).
Data exclusions
Animals with unintended experimental error (poor ChR2, fiber, or electrode placement) were excluded.
Replication
Data was collected in 3 or more experiments and all results were pooled. All attempts at replication were successful
Randomization
Rats were randomly assigned to each experimental group.
Blinding
Blinding of group was not possible due to obvious differences in fiber position. However data were collected and analyzed in an automated
manner to prevent experimenter bias.
Reporting for specific materials, systems and methods
We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.
Materials & experimental systems
n/a Involved in the study
Methods
n/a Involved in the study
Antibodies
Eukaryotic cell lines
Palaeontology and archaeology
Animals and other organisms
Clinical data
Dual use research of concern
Plants
ChIP-seq
Flow cytometry
MRI-based neuroimaging
Antibodies
Antibodies used
Antibodies used Goat anti-GFP (Abcam ab6673, 1:1000). Alexa Fluor 488 anti-goat (Life Technologies A11055, 1:1000).
Validation
Validation and references on manufacturer's website:
n
a
t
u
r
e
p
o
r
t
f
o
l
i
o
|
r
e
p
o
r
t
i
n
g
s
u
m
m
a
r
y
A
p
r
i
l
2
0
2
3
2
Validation
ab6673: anti-GFP assayed by ELISA for direct binding of antigen recognizes wild type, recombinant and enhanced forms of GFP. IHC.
Species independent. 483 references.
A11055: IHC tested 1-10ug/ml. 2866 references.
Animals and other research organisms
Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research, and Sex and Gender in
Research
Laboratory animals
Charles River and in-house bred Long Evans rats, 10-12 weeks.
Wild animals
No wild animals were used in this study.
Reporting on sex
Data were collected from male and female rats. No differences were observed between male and female animals so results were
pooled.
Field-collected samples
No field samples were collected in this study.
Ethics oversight
Experimental procedures were approved by the Stanford University Institutional Animal Care and Use Committee (IACUC) and by the
Administrative Panel on Laboratory Animal Care (APLAC), following the National Institutes of Health guidelines for the care and use of
laboratory animals.
Note that full information on the approval of the study protocol must also be provided in the manuscript.
n
a
t
u
r
e
p
o
r
t
f
o
l
i
o
|
r
e
p
o
r
t
i
n
g
s
u
m
m
a
r
y
A
p
r
i
l
2
0
2
3
3
|
Could not heal snippet
|
10.1007_s00406-021-01302-7.pdf
| null | null |
European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
https://doi.org/10.1007/s00406-021-01302-7
ORIGINAL PAPER
P300 and delay‑discounting in obsessive–compulsive disorder
Vera Flasbeck1
· Björn Enzi1 · Christina Andreou2
· Georg Juckel1
· Paraskevi Mavrogiorgou1
Received: 27 October 2020 / Accepted: 4 July 2021 / Published online: 13 July 2021
© The Author(s) 2021
Abstract
Previous research showed that dysfunctions of fronto-striatal neural networks are implicated in the pathophysiology of obses-
sive–compulsive disorder (OCD). Accordingly, patients with OCD showed altered performances during decision-making
tasks. As P300, evoked by oddball paradigms, is suggested to be related to attentional and cognitive processes and generated
in the medial temporal lobe and orbitofrontal and cingulate cortices, it is of special interest in OCD research. Therefore, this
study aimed to investigate P300 in OCD and its associations with brain activity during decision-making: P300, evoked by
an auditory oddball paradigm, was analysed in 19 OCD patients and 19 healthy controls regarding peak latency, amplitude
and source density power in parietal cortex areas by sLORETA. Afterwards, using a fMRI paradigm, Blood–oxygen-level-
dependent (BOLD) contrast imaging was conducted during a delay-discounting paradigm. We hypothesised differences
between groups regarding P300 characteristics and associations with frontal activity during delay-discounting. The P300
did not differ between groups, however, the P300 latency over the P4 electrode correlated negatively with the NEO-FFI
score openness to experience in patients with OCD. In healthy controls, P300 source density power correlated with activity
in frontal regions when processing rewards, a finding which was absent in OCD patients. To conclude, associations of P300
with frontal brain activation during delay-discounting were found, suggesting a contribution of attentional or context updating
processes. Since this association was absent in patients with OCD, the findings could be interpreted as being indeed related
to dysfunctions of fronto-striatal neural networks in patients with OCD.
Keywords OCD · Event-related potentials · P300 · Delay discounting · Neuroimaging · fMRI
Introduction
Obsessive–compulsive disorder (OCD) is a psychiatric con-
dition that involves neurobiological dysfunctions of fronto-
striatal neural networks. Neuroimaging methods have con-
tributed to a better understanding of the pathogenesis of this
disorder, however, findings are not consistent across all stud-
ies. Although efficacious treatments have been developed
and established, patients in clinical settings often show inad-
equate responses to treatment attempts. Several studies indi-
cate a neurobiological basis of OCD, resulting in two main
* Georg Juckel
[email protected]
1 Department of Psychiatry, LWL-University Hospital, Ruhr
University Bochum, Alexandrinenstr. 1, 44791 Bochum,
Germany
2 Department of Psychiatry and Psychotherapy, University
Hospital Lübeck (UKSH), Ratzeburger Allee 160,
23538 Lübeck, Germany
hypotheses: neuroanatomical and serotonergic. Studies using
neurochemical and neuroimaging methods have shown that
various neurotransmitters are implicated in the pathophysiol-
ogy of this disorder, including serotonin [1], dopamine [2]
and glutamate [3]. To date, the highest impact is attributed
to the neurochemical model of OCD that postulates a central
serotonergic dysfunction, mainly based on the efficacy of
selective serotonin reuptake inhibitor (SSRI) treatment in
OCD. However, the underlying therapeutic mechanism of
SSRIs in OCD remains unclear because there are discrep-
ant findings across studies of structural and functional brain
changes before and after SSRI treatment in patients with
OCD [4].
In addition, it has been suggested that OCD is caused
by abnormal activity in the cortico-striato-thalamo-cor-
tical (CSTC) circuits, including the orbitofrontal cortex
(OFC), the striatum within basal ganglia and the thala-
mus [5, 6], which is summarised as the neuroanatomical
hypothesis. It was postulated that OCD symptoms may be
related to increased activity in the OFC, as a consequence
Vol.:(0123456789)1 3
328
European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
of diminished inhibitory effects of the striatum (especially
the globus pallidus internus) on the thalamus. Furthermore,
this hypothesis suggests that OCD could be associated with
dysfunctional cognitive and metacognitive processing.
In order to investigate the proposed OFC hyperactivity in
OCD patients, the P300 component of auditory event-related
potentials (ERPs) could be a suitable tool, as it is proposed
that P300 is generated in the medial temporal lobe, OFC and
cingulate cortex [7]. Furthermore, the appearance of P300
during oddball paradigms is suggested to reflect cognitive
and attentional processes. In detail, the P300 occurs with a
latency of approximately 300–500 ms after the occurrence
of rare or task-related stimuli or after a target stimulus (com-
pared to non-target stimuli) and was measured over frontal-
to-temporal and parietal electrodes.
A tremendous range of literature revealed inconsistent
cognitive neuropsychological findings e.g. attentional defi-
cits in OCD, which were found using various behavioural
tests. The investigation of biological markers, such as the
P300 component, also contributed to the understanding of
cognitive alterations in OCD. Unfortunately, inconsistent
P300 abnormalities were reported for patients with OCD
with several previous studies reporting shortened latencies
[8–12] and increased amplitudes [8, 13–15], whereas other
studies showed decreased P300 amplitudes in these patients
[16, 17]. Thus, additional future research is necessary to
clarify the P300 alterations and in OCD, which was one aim
of the present study.
Further studies aimed to investigate ERPs, especially
the P300, and their changes during decision-making tasks
[18–22]. Here, the P300 was found to be linked to risky deci-
sion making, with larger P300 amplitudes associated with
riskier behaviours. Besides these findings, only a few stud-
ies exist that were interested in specific delay-discounting
effects on P300, such as the effect of intertemporal choices
[23–26].
Research regarding decision making in difficult tasks,
such as the Iowa Gambling Task [27] and the Game of Dice
Task [28], also showed abnormal performances in OCD
patients [29–32], but did not clarify which neural processes
were altered. Delayed reward discounting is a behavioural
economic index of impulsivity and numerous studies have
examined delayed reward discounting in substance use dis-
order [33, 34]. However, few empirical data is available on
delayed reward discounting in patients with OCD [35]. In
a series of functional magnetic resonance imaging (fMRI)
studies, scientists reported activities primarily within the
OFC during delay-discounting tasks [36], therefore, this task
may be a suitable tool for assessing the activity state within
the OFC. However, it remains unclear whether neurotrans-
mitters, especially serotonin, are involved in the abnormali-
ties of the CSTC circuit in OCD. In a recently published
study, using the same dataset as the present investigation, the
results indicated that activation of dorsolateral and medial
prefrontal cortex (PFC) as well as ventral striatum activa-
tion differed between OCD patients and healthy volunteers
during the delay-discounting paradigm (immediate reward
vs. control) [37]. Based on previous literature and theoretical
considerations, we propose that P300, as a marker of cogni-
tive and attentional processes, would be increased in OCD,
due to altered attention and accelerated cognitive and motor
processes. Higher P300 processing would be observable
as lower amplitudes and longer latencies [38]. Moreover,
it would be of interest to examine whether general cogni-
tive processing would be associated specifically with OFC
activity in patients with OCD. The OFC activity would be
of special interest in OCD since it is suggested to be a region
which is functionally altered in OCD, during a task that is
known to elicit deviating behaviour in OCD patients com-
pared und unaffected individuals.
To our knowledge, the approach of combining oddball
P300 measures with BOLD contrasts of delay-discounting
has not been investigated previously and is the secondary
aim of the present investigation. We combined the previ-
ously performed fMRI analysis during the delayed discount-
ing paradigm with EEG, cortical and source analysis con-
cerning the P300 component, whereby these measurements
were conducted consecutively. For the fMRI analysis, func-
tional BOLD signal was extracted from selected anatomi-
cally defined regions of interest in the OFC, next to whole
brain fMRI analysis [37].
We hypothesised that a significant association of P300
during EEG recording would be found with activation of the
reward processing system during the fMRI-delay-discount-
ing task. Here, we proposed that a higher cognitive demand
during the P300 paradigm would be related to increased
OFC activity. In addition, we expected to detect differences
in P300 amplitude and latency between healthy participants
and patients with OCD. More detailed, we hypothesised to
find longer latencies and decreased amplitudes in patients
with OCD, since we suggested increased cognitive impair-
ment in these patients.
Method
Subjects
N i n et e e n p a t i e n t s ( e i g h t fe m a l e s ; m e a n a ge
33.37 ± 11.73 years) with unequivocal diagnosis of OCD
were recruited. Diagnosis was based on the diagnostic crite-
ria of the 4th edition of the Diagnostic and Statistical Man-
ual of Mental Disorders (DSM-IV) [39] and 10th revision of
the International Statistical Classification of Diseases and
Related Health Disorders (ICD-10: F42.X) [40]. Exclusion
criteria included organic disorders according to the ICD-10
1 3 European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
329
(F0X) or recent concomitant neurological or other medical
disorders and the presence of severe alcohol or substance
abuse. No patient met the criteria for Tourette syndrome
or any psychotic disorder. Table 1 shows the sociodemo-
graphic and clinical data of the nineteen patients included
in the study. Seventeen patients were medicated at the time
of assessment: Thirteen were taking SSRIs (fluoxetine,
40–60 mg/day; sertraline, 50–150 mg/day; escitalopram,
Table 1 Sociodemographic and clinical characteristics of patients
with obsessive–compulsive disorder (OCD) and healthy controls
OCD (n = 19)
Controls (n = 19)
Gender
Female
Male
Age (years)
Marital status
Married
Cohabitating
Single
Education
Upper grade
Middle grade
Lower grade
Occupational status
Employed
Unemployed
Student
Retired, unable to work
Duration of illness (years)
Age of onset (years)
HAM-D
BDI
Y-BOCS, obsessions
Y-BOCS, compulsions
Y-BOCS, total
MOCI
STAI I
STAI II
CGI
MWST-IQ
NEO-FFI, total
BIS-11, total
PSP
8(42.1%)
11(57.9%)
33.37 ± 11.73
8 (42.1%)
11 (57.9%)
31.63 ± 10.79
3 (15.8%)
10 (52.6%)
6 (31.6%)
15 (78.9%)
4 (21.1%)
0
8 (42.1%)
3 (15.8%)
6 (31.6%)
2 (10.2%)
14.27 ± 12.39
19.21 ± 6.71
12.42 ± 6.13
14.68 ± 10.12
10.74 ± 2.53
10.53 ± 3.73
21.79 ± 6.59
14.84 ± 5.93
42.89 ± 13.72
50.26 ± 11.75
4.58 ± 0.69
109.63 ± 12.08
2.77 ± 0.55
59.00 ± 8.72
67.16 ± 14.08
4 (21.1%)
8 (42.1%)
7 (36.8%)
16 (84.2%)
3 (15.8%)
0
13 (68.4%)
0
6 (31.6%)
0
1.42 ± 2.01*
3.89 ± 2.96*
30.21 ± 5.06*
30.58 ± 7.95*
1.00 ± 0*
119.58 ± 13.22*
2.69 ± 0.69
56.37 ± 7.43
100*
Values are numbers and percentages or means and standard devia-
tions (SD); *p < 0.05
HAM-D Hamilton Depression Scale, BDI Beck Depression Inventory,
Y-BOCS Yale–Brown Obsessive Compulsive Scale, MOCI Maudsley
Obsessive–Compulsive Inventory, STAI Stait–Trait Anxiety Inven-
tory, CGI Clinical Global Impression scale, MWST-IQ Mehrfach-
Wortschatztest, NEO-FFI NEO Five-Factor Inventory, BIS-11 Barratt
Impulsiveness Scale, PSP Personal and Social Performance scale
10 mg/day; citalopram, 20–60 mg/day), one received clo-
mipramine (200 mg/day) and three received a serotonin–nor-
epinephrine reuptake inhibitor (SNRI: venlafaxine, 300 mg/
day, n = 2; or duloxetine, 90 mg/day, n = 1). None of the
patients were engaged in cognitive-behavioural therapy dur-
ing the study period.
Nineteen healthy volunteers (eight females; mean age
31.63 ± 10.79 years) without any neurological or psychi-
atric disorder in their personal or family history served as
a control group, matched for age, gender, education level
and handedness (18 right-handed). The volunteers under-
went the Mini International Neuropsychiatric Interview for
DSM-IV and ICD-10 disorders (MINI-PLUS) [41, 42] and
psychometric tests for obsessive–compulsive, depressive and
anxiety symptoms.
All participants underwent the same study design with
fMRI, P300-based electroencephalography (EEG) and
psychometric assessments within a few hours on a single
day. All participants started with the EEG recording and
questionnaires in the morning and the fMRI recording was
done in the afternoon. For one control participant, the fMRI
recording was done the next morning, still within 24 h.
Clinical assessment
The severity of OCD symptoms was assessed by the
Yale–Brown Obsessive Compulsive Scale (Y-BOCS) [43,
44] and the Maudsley Obsessive–Compulsive Inventory
(MOCI) [45]. To validate the presence of OCD symptoms,
we used the Y-BOCS symptom checklist.
The severity of depressive symptoms was assessed using
the Hamilton Depression Rating Scale (HAM-D) [46] and
self-ratings were assessed by the Beck Depression Inven-
tory (BDI) [47]. Anxiety symptoms were measured using
the State-Trait Anxiety Inventory (STAI I and II) [48, 49].
The overall severity of the psychiatric disorder was quan-
tified using the Clinical Global Impression (CGI) score
(NIMH) [50]. Psychosocial functioning was measured by
the Personal and Social Performance scale (PSP) [51] and
impulsivity was assessed by the Barratt Impulsiveness Scale
(BIS-11) [52, 53]. The NEO Five-Factor Inventory (NEO-
FFI) [54] was used to assess personality characteristics such
as extraversion, neuroticism and conscientiousness. Par-
ticipants’ verbal intelligence was estimated with the Mehr-
fachwahl–Wortschatztest (MWT) [55].
P300
During the oddball paradigm, two different kinds of stimuli
(80% non-target, 400 sinus tones, 500 Hz; 20% target stim-
uli, 100 sinus tones, 1000 Hz) were presented in pseudoran-
domized order (80 dB SPL, 40 ms duration, 10 ms rise and
fall time, interstimulus interval 1.5 s) via headphones (Sony
1 3 330
European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
Stereo Headphones MDR-1A, Sony® Corporation) and Pres-
entation® software (Neurobehavioral Systems, Inc., Version
14.9, Berkeley, CA: www. neuro bs. com) to the participants.
All participants were instructed to press a response button
with their dominant hand whenever they heard the target
stimulus.
EEG recording and data analysis
Subjects sat in a comfortable armchair in an electrically
shielded and sound-attenuated room. Auditory-evoked
potentials were recorded with 32 non-polarizable Ag–AgCl
electrodes referred to as FCz, placed according to the inter-
national 10/20 system. Impedances were kept at 5 kΏ or
below. EEG was filtered using a bandpass of 0.16–70 Hz
and data were collected at a sampling rate of 250 Hz using a
BrainAmp MR amplifier and BrainVision recorder software
(Version 1.20.001: Brain Products GmbH, Gilching, Ger-
many). Data analysis was performed using the BrainVision
Analyzer 2.0 (Version 2.01.3931: Brain Products GmbH,
Gilching, Germany). The recorded data were re-referenced
to the mastoid electrodes and filtered using bandpass and
notch filters (0.5–20 Hz and 50 Hz). For artifact rejection,
all trials were excluded if the voltage exceeded ± 70 µV in
any channel. The epochs (− 200 to 1000 ms) were averaged
separately for the target and non-target stimuli and corrected
to the baseline (− 200 ms). Only subjects with at least 40
trials free of artefacts for both stimuli were included.
The P300 amplitudes and peak latencies were analysed
(P300 defined as the most positive peak within 250–500 ms
after stimuli onset for the P3, P4 and Pz electrodes because
P300 is suggested to be maximal over parietal electrodes
[56]. This was also true for the present study. As shown in
Fig. 1, the maximal amplitude was recorded over parietal
electrodes, independent of group.
sLORETA analysis
For the analysis of source P300 data, sLORETA Software
[57] was used. Therefore, the re-referencing was conducted
to the average of all electrodes and the average of segments
from target tones were exported. First, we compared the cur-
rent density power, measured as µA/mm2, between groups.
Therefore, a voxel-by-voxel t-test was performed on log-
transformed data for the timeframe from 240 to 580 ms after
target tone. As previously done, a non-parametric randomi-
sation approach was applied [58] for correction for multiple
comparisons. In addition, a ROI analysis was performed to
investigate the electric neuronal activity as current source
density power in the parietal cortices comprising all voxels
of the Brodmann areas 5, 7, 39 and 40 (see Fig. 2). Here,
Brodmann areas belonging to the posterior parietal cortex
were selected, due to the involvement of this regions in
higher-order functions [59]. Since we are interested in cog-
nitive processing, as represented by P300, we chose the pos-
terior parietal cortex and excluded anterior parietal cortical
regions, which are also involved in somatosensory processes.
In this study, the BA 5-ROI covered a region extended in
Talairach space from x: 0–40 and 0 to − 40, y: − 35 to − 50,
z: 50–70 and included all voxels. The BA 7-ROI covered the
region from x: 0 to − 40 and 40, y: − 50 to − 80, z: 30–70,
also including all voxels. Similarly, BA 39 extended from x:
− 35 to − 60 and 35–60, y: − 55 to − 80, z: 10–40 and BA
40 BA 39 from x: − 25 to − 65 and 25–65, y: − 20 to − 60, z:
15–60. The ROI analysis was done with the “ROI-Extractor”
tool which averages the CSD values in the specified vox-
els. The brain model of LORETA is based on the Montreal
Neurological Institute average MRI brain map (MNI 152),
while the solution space is limited to the cortical grey mat-
ter, comprising 6239 voxels of 5-mm3 resolution. The mean
source density power at each ROI within the time frame of
240−580 ms after target tone onset was computed for every
Fig. 1 Topographic maps of
brain activity after onset of the
target tones from 0 to 500 ms
in healthy controls (left) and
patients with OCD (middle),
measured by EEG. The right
topographic maps show the
difference between patients with
OCD and healthy controls
1 3 European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
331
Fig. 2 Comparison of brain activity for the P300 between patients with OCD and healthy controls by sLORETA. Here, the ROIs, namely BA39,
BA4, BA 5 and BA7 are marked
participant. Finally, we calculated the average of all ROIs
for each participant.
Behavioural practice session of delay‑discounting
We used a slightly modified version of an established
decision-making paradigm described previously by Peters
and Büchel [36]. Before scanning, all subjects completed
an identical practice version of the task. The results of the
pretest were used to adequately compute offers for the fMRI
sessions and estimate the individual discounting rate k. The
participants were ask to choose between a fixed immediate
reward of €20 and higher but delayed rewards in 2, 7, 14, 28
or 40 days. The delayed rewards were computed individually
for each participant to ensure that the delayed offer was
chosen in approximately 50% of all trials. The amount of
money at which the participants switched from accepting the
immediate fixed reward to the delayed reward, also called
the indifference amount, was calculated and converted into
proportions of the fixed reward. Based on the hyperbolic
function, these data were used to obtain the best-fitting dis-
counting parameter k.
fMRI
During fMRI, each trial began with a short cue symbol
(500 ms) followed by presentation of the reward options
(immediate vs. delayed) for 2000 ms. After a jittered
1 3 332
European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
anticipation period of 2000–3000 ms, participants had to
choose the preferred reward option using an MR-compatible
response box. After a short feedback period of 2000 ms, a
jittered intertrial interval (3000–5000 ms) was presented.
Each delay condition consisted of 14 trials, resulting in 70
trials per run. During 10 control condition trials, the partici-
pants were asked to choose either the left or right side of the
screen without getting a reward. The experiment consisted
of two runs of approximately 18 min each. Functional data
were collected using a 3-T whole-body MRI system (Philips
Achieva 3.0 T TX) equipped with a 32-channel Philips
SENSE head coil. A total of 32 T2*-weighted echo-planar
images per volume with blood-oxygen-level-dependent
(BOLD) contrast were obtained using a sensitivity-encoded
single-shot echo-planar imaging protocol (SENSE-sshEPI).
For further details of fMRI procedures, see our previous
publications [37, 60]. The functional data were preprocessed
and statistically analysed using SPM8 (Wellcome Depart-
ment of Cognitive Neuroscience, University College Lon-
don, UK: http:// www. fil. ion. ucl. ac. uk) and MATLAB 7.11
(Mathworks Inc., Natick, MA, USA). In addition to the
whole brain analyses described elsewhere [37, 60], activ-
ity in anatomically defined regions of interest based on our
previous work were analysed. These regions, namely the left
and right OFC, respectively (inferior frontal gyrus, orbital
part; superior frontal gyrus, medial orbital part (SFG/MO);
middle frontal gyrus, orbital part; superior frontal gyrus,
orbital part; gyrus rectus) were generated using both AAL
and WFU PickAtlas software. More in detail, percent signal
changes (based on the beta values for each event) derived
from the above-mentioned regions were extracted using the
standard routines implemented in MarsBar [61].
Statistical analysis
Statistical analyses of the data were performed using IBM
SPSS Statistics for Windows, Version 25.0 (IBM Corp.,
Armonk, NY, USA). The analyses of P300, questionnaire
data and neuroimaging results were performed with non-
parametric Mann–Whitney U tests and Spearman correlation
coefficients due to violations of normal distribution. Statis-
tical significance was defined as p < 0.05. For correlations
of questionnaires and P300 data, Bonferroni correction due
to multiple testing was applied, whereby related variables,
e.g. P3 amplitude and P4 amplitude, were considered as one
factor. The p-value threshold was shifted accordingly (for
eight questionnaires: BDI, MOCI, STAI, CGI, MWST-IQ,
NEO-FFI, BIS-11, PSP and three P300 variables: latency,
amplitude and source density power: p = 0.05/11 = 0.0045).
In the patients group, additional correlations were calculated
for Y-BOCS scores. For correlations between fMRI data,
based on ROI-analysis, and P300 (source P300 data), the
significance level was set to p < 0.025 (since OFC regions
are related and considered as one factor; correction for test-
ing of left and right hemisphere was applied). For correla-
tions between functional BOLD responses and P300, the
significance threshold was adjusted for six different, unre-
lated regions and P300 (p = 0.05/7 = 0.007). The correla-
tions with fMRI data were performed for the three different
contrasts separately, i.e. [∆ immediate reward − control],
[∆ delayed reward—control] and [∆ immediate reward—
delayed reward].
Results
Sociodemographic and clinical findings
Patients with OCD reported significantly more severe psy-
chopathological symptoms with higher scores in depression,
anxiety and obsessive–compulsive symptom questionnaires
compared to the control group (Table 1). Regarding per-
sonality characteristics, patients showed lower neuroticism
(OCD: M = 1.31, SD = 0.69; control: M = 2.27, SD = 0.62;
U = 48.0, Z = − 3.87, p < 0.001) and higher extraversion
(OCD: M = 2.65, SD = 0.49; control: M = 2.05, SD = 0.60;
U = 70.0, Z = − 3.23, p = 0.001) and openness to expe-
rience (OCD: M = 2.72, SD = 0.61; control: M = 2.32,
SD = 0.48; U = 82.5, Z = − 2.86, p = 0.003). No differ-
ences between groups were observed for agreeableness and
conscientiousness.
Although no differences between groups were observed
for the BIS-11 total score, distinct differences emerged for
the BIS-11 subscales, with OCD patients reaching lower
scores in attentional impulsiveness (OCD: M = 12.58,
SD = 3.08; control: M = 17.63, SD = 4.0; U = 42.5,
Z = − 4.05, p < 0.001) and higher scores in motor impul-
siveness compared to the control group (OCD: M = 21.47,
SD = 2.59; control: M = 19.32, SD = 3.13; U = 104.0,
Z = − 2.25, p = 0.025).
EEG: P300 findings
The waveforms evoked by the target tones are shown in
Fig. 3 for the parietal electrodes of interest (P3, P4 and Pz)
and for additional central (C3, C4 and Cz) and frontal (F3,
F4 and Fz) electrodes. Here, the parietal maximum of the
P300 component is again observable. P300 amplitude and
latency did not differ significantly between OCD patients
and controls at P3, P4 and Pz. In OCD patients, amplitudes
reached 8.5 µV (SD = 4.6), 7.0 µV (SD = 3.3) and 7.0 µV
(SD = 3.5 µV) and latencies were 375.6 ms (SD = 53.0),
366.3 ms (SD = 50.8) and 373.7 ms (SD = 47.0) for Pz,
P3 and P4, respectively. In healthy controls, amplitudes
reached 8.5 µV (SD = 3.6), 6.9 µV (SD = 3.1) and 6.8 µV
(SD = 2.9). There was a visual tendency towards shorter
1 3 European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
333
Fig. 3 Grand-average waveforms showing the ERP components
evoked by the target tones during the oddball paradigm. The wave-
forms of electrodes F3, F4 (first line), C3, C4 (second line), P3, P4
(third line), Fz, Cz (forth line) and Pz and the legend (sixth line) are
presented. Healthy controls (blue) and patients with OCD (brown) are
indicated by separate lines
latencies within the control group (357.5 ms, SD = 28.5;
357.0 ms, SD = 23.5; 364.0 ms, SD = 36.9) for Pz, P3 and
P4, respectively (see Fig. 3), compared to patients with
OCD. This tendency is also visible in Figs. 1, which shows
the parietal maximum in controls in the time window of
252–376 ms, and for patients with OCD, the most posi-
tive activity is observable in the last timeframe from 376
to 500 ms.
Similar to the cortical P300 results, no differences
between groups were found for source P300 results as cal-
culated by sLORETA (maximum t = 2.419, p < 0.05; all
p’s > 0.05; see Fig. 4). Accordingly, no differences were
found for the ROI analyses (Fig. 2).
Correlations between P300 (EEG) and clinical
outcome
In the group of healthy controls, questionnaire scores cor-
related with P300 characteristics, as measured by EEG.
However, after correction for multiple testing (for nine ques-
tionnaires: BDI, MOCI, STAI, CGI, MWST-IQ, NEO-FFI,
BIS-11, PSP and three P300 variables: latency, amplitude
1 3 334
European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
Fig. 4 T-test comparison of
current source density power
by sLORETA between patients
with OCD and healthy controls.
The marked differences did not
reach statistical significance.
In a, parietal brain regions are
shown and in b frontal regions
are visible
signal change derived from anatomically based ROIs, and
the P300 during EEG recording, we calculated the Spearman
correlation coefficient for the fMRI signal for [∆ immediate
reward—control] and P300 characteristics (source density
power). Within the OCD group, no significant correlation
was found. In contrast, we were able to detect significant
positive correlations between activations, i.e. the signal
change for the contrast, in the left middle frontal gyrus
(orbital part) and P300 source density power (r = 0.535,
p = 0.018; see Fig. 6) in the healthy subgroup. For the con-
trast [∆ delayed reward—control], significant correlations
were again only observable in the control group between the
signal change in the left middle frontal gyrus (orbital part)
and the left superior frontal gyrus (orbital part) and the P300
source density power (r = 0.544, p = 0.016). For the contrast
[∆ immediate reward—delayed reward], no significant asso-
ciation was found.
fMRI BOLD responses and correlations with P300
For brain activations during the fMRI task in both groups,
see Table 2 and for details see [37]. In brief, a main effect of
task was observable in the bilateral inferior frontal gyrus, the
bilateral supramarginal gyrus, the left middle frontal gyrus,
the left middle occipital cortex and the angular gyrus. A
group effect was observed for the left ventral striatum/puta-
men and the right dorsolateral prefrontal cortex. Correla-
tions between these functional BOLD measures (FOI), based
Fig. 5 Correlation between the NEO-FFI score openness to experi-
ence score and the P300 latency over the P4 electrode in patients with
OCD
and source density power: p = 0.05/11 = 0.0045), no correla-
tion remained significant. In the patients group, a significant
correlation between the NEO-FFI openness to experience
score and the P4 P300 latency survived Bonferroni correc-
tion (r = − 0.697, p = 0.001; see Fig. 5; correction for all
variables mentioned above plus Y-BOCS).
Correlations between P300 (EEG) and ROI‑activation
(BOLD)
Regarding a possible relationship between reward-related
neuronal activity during fMRI acquisition, extracted as
1 3 European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
335
Fig. 6 Correlations between
activation in the left middle
frontal gyrus (orbital part) for
the difference [∆ immediate
reward—control] and the P300
power over left parietal brain
areas for healthy controls and
patients with OCD
Table 2 Activations in healthy
subjects and patients with
obsessive–compulsive disorder
(OCD)
Hemi-
sphere
Region
Extent k
Z value
Statistical valuea
F-contrast [main effect of task] collapsed over groups
− 38, 6, 28
− 58, − 34, 34
60, − 38, 34
− 34, 28, 38
56, 12, 8
− 30, − 76, 22
28, − 54, 42
T-contrast [Interaction group × task], i.e. “immediate reward: accepted” vs. “delayed reward: accepted” in
Inferior frontal gyrus, opercular part
Supramarginal gyrus
Supramarginal gyrus
Middle frontal gyrus/dlPFC
Inferior frontal gyrus, opercular partb
Middle occipital cortexc
Angular gyrusc
16.89
17.93
18.91
16.51
9.06
12.64
11.13
4.94
5.09
5.23
4.88
3.51
4.24
3.95
24
69
81
31
16
50
12
L
L
R
L
R
L
R
heathy vs. OCD patients
L
− 22, 16, − 2
R dlPFC (BA8)b
16, 20, 56
Putamen/ventral striatumb
12
52
3.56
3.68
3.67
3.8
Initial threshold p[FWE] < 0.05 for an extent k > 10 voxels or F > 10.0 for k > 10
a t or F value. bp[FWE] < 0.05 after small volume correction with 5 mm radius. cp[FWE] < 0.05 on cluster
level. BA Brodmann area, dlPFC dorsolateral prefrontal cortex
on the contrasts, and P300 source density power did not
survive correction for multiple testing.
Discussion
The present study investigated P300 ERPs and their asso-
ciations with fMRI activation in a delay-discounting task in
OCD patients and healthy controls. The two matched groups
differed regarding psychopathology, personality characteris-
tics and impulsivity but did not differ in P300 amplitudes or
latencies or P300 source density power in parietal regions.
Thus, our hypothesis that the groups will differ regarding
P300 characteristics was not confirmed. Regarding personal-
ity characteristics, patients showed lower neuroticism, but
higher extraversion and openness to experience. In previous
studies, higher neuroticism and lower extraversion has been
reported frequently for patients with OCD [62, 63]. Here,
the findings also seem inconsistent, whereas it has been pro-
posed by another study that facets of openness may impact
on the particular expression and severity of obsessive–com-
pulsive symptoms [64]. In our study, the factor openness
was negatively correlated with P300 latency over P4 in the
patients group (see Fig. 5). Thus, higher openness is related
to smaller peak latencies, i.e. lower controlled processing.
This could, very speculatively, interpreted as lower inhibi-
tion in individuals scoring high in openness to experience.
Even if patients with OCD did not differ from healthy
controls with regard to P300 latencies and amplitudes, a
tendency towards prolonged P300 latency was observed for
OCD patients. Previous studies reported prolonged laten-
cies and larger P300 amplitudes in OCD [8–14]. However,
it should be noted that the existing literature on P300 EEG
abnormalities in studies of patients with OCD is rather dis-
crepant. Sanz et al. [17] found lower P300 amplitudes in
combination with prolonged P300 latencies in a sample of
drug-free adult OCD patients compared to healthy controls.
In addition, a trend towards increased P300 amplitude was
observed in patients after treatment with clomipramine,
whereas, no modification in P300 latency was shown.
Dayan-Riva and colleagues [65] utilised pictures showing
neutral and angry facial expressions instead of auditory
stimuli. They reported higher P3 amplitudes in patients with
OCD compared to unaffected controls for neutral stimuli
only, with no differences regarding angry facial expressions
[65]. In this study, no differences were found between groups
for latencies, suggesting that the different findings observed
in OCD patients compared to healthy controls may depend
1 3 336
European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
crucially on the tasks used. In addition, it is known that P300
latencies have a much lower reliability than P300 ampli-
tudes, whereas, perhaps data on P300 onset latency could
have been mixed up with data on P300 peak latency. Most
of these studies reported a shortened P300 latency whereas
others detected prolonged latencies [9, 17]. In previous stud-
ies, shorter latencies in OCD patients were found only for
P3b [8, 13, 14]. Thus, recent research brought several argu-
ments for altered P300 amplitudes and more sparse support
for latency differences in OCD. Regarding source analysis of
P300, less research is existing, whereas, one study reported
higher P300-related activity in patients with OCD in the left
orbitofrontal cortex, left prefrontal, parietal and temporal
areas compared to controls [13]. Thus, there are hints that
altered P300 could play a role in OCD, whereas, the results
may depend on the tasks used, the sample sizes investigated
and medication of samples. Furthermore, the data analy-
sis may have varied across studies, e.g. with regard to peak
latency vs. onset latency analysis or the investigation of P3a
and P3b subcomponents.
P300 and delay‑discounting in OCD
Previous researchers have revealed that P300 reflects the
updating of cognitive models in order to make an appropri-
ate response in the sense of an evaluation process for making
a decision [22, 66]. In our fMRI study part, as previously
reported for the present dataset, it has been shown that acti-
vations of dorsolateral PFC and ventral striatum activations
differed between OCD patients and control participants dur-
ing a delay-discounting paradigm (see [37]). Thus, it was
known that P300 (context updating) during EEG recording
and delay-discounting behaviour and processing were altered
in OCD. Therefore, the question was whether P300, meas-
ured by EEG, is related to brain activations, measured by
fMRI, during decision making, which was the secondary
subject of the present study. In healthy participants, source
density power of P300 over parietal brain areas correlated
positively with activations in the left middle and superior
frontal gyri (orbital parts) for the [∆reward—control] con-
trasts during the fMRI task. No such correlations were found
in the patient’s group. The correlations in healthy controls
are consistent with previous results, showing larger P300
amplitudes in contexts causing higher risk tendencies [21].
Furthermore, Bellebaum et al. [67] reported that P300 was
larger for positive outcomes and showed an effect of poten-
tial reward magnitude that was independent of valence.
Thus, findings regarding the relationship between P300 and
decision-making suggested that P300 was modulated by
reward magnitude.
This association was absent in patients with OCD, as no
correlations of brain activation during the fMRI-task and
P300 power density were found. There are several potential
reasons for these findings. First, as we found lower scores
for attentional impulsiveness in patients with OCD com-
pared to healthy controls, a general reduced attention
could attenuate the association of P300 with brain activa-
tion during the delay-discounting paradigm. Second, it has
been suggested that patients with OCD exhibit prolonged
deliberation during decision-making, implicating emotional
valence or risk due to altered processing in relevant brain
regions, including frontal and limbic regions [68]. Third,
previous studies reported impaired adaption of the decision
strategy during a decision-making task, suggesting lower
flexibility in OCD [69, 70]. It can be speculated that the
reduced flexibility could be related to reduced attention. In
summary, previous research indicated decreased flexibility,
and therefore, decreased capacity in OCD to focus attention
in a goal-directed manner. In addition, deficits may occur
due to delayed attention to relevant cues in OCD ([71]; for
review, see [72]).
In fact, this interpretation is speculative and not based on
our results. Based on our data, one can propose that these
negative findings in the patient group could be caused by
altered cognitive controlled processing in these patients,
whereby controlled processing is not directly related to
reward processing in the OFC, a region which is proposed
to be hyperactive due to diminished inhibitory effects of
the striatum in OCD. Altered activations of the dorsolat-
eral PFC and the ventral striatum has been shown for the
present group of patients, wherefore the results suggest
that the OCD group showed indeed altered processing in
cortico-striato-thalamo-cortical (CSTC) circuits during the
fMRI-task. Therefore, the missing link between parietal cog-
nitive processing, measured by EEG, and OFC activation
during reward processing in the fMRI scanner in patients
might reflect deviating CSTC circuit processing compared
to processes observed in healthy individuals. Another pos-
sible reason for the missing association between general
cognitive processing (EEG), and reward processing, meas-
ured by fMRI, in patients with OCD could be a diminishing
effect of the psychopharmacological medication the patients
received. In the present study, most of the patients received
antidepressant medication. However, it has been shown that
psychopharmacological medication affect P300 and OFC
activity [73, 74]. Therefore, future research also might inves-
tigate the effect of psychopharmacological medication in
cognitive processing. Finally, the sample sizes were small
in the present study, wherefore significant results, also for
P300 analyses between groups, would possibly appear in
larger samples.
Conclusion
In the present study, a negative correlation between the
factor openness with P300 latency over P4 was observed
1 3 European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
337
exclusively in the patients group. We found distinct associa-
tions in healthy controls showing correlations of brain acti-
vation, as measured by fMRI during reward processing with
P300 power, which were absent in the group of patients with
OCD. Since cognitive processing, as indicated by P300, did
not differ between the groups, the missing association in the
group of patients with OCD could be interpreted as altered
CSTC circuit activity, which would disrupt the association
with general cognitive processing observed in unaffected
individuals.
Limitations
Some limitations of this current study should be noted. First,
as mentioned above, our sample consists of patients receiv-
ing SSRI medication, which may have affected the results.
Second, the small sample size does not enable a meaningful
investigation of the specific OCD subgroup characteristics
or maybe even group differences at all. Furthermore, P300 as
well as fMRI BOLD contrasts during the delay-discounting
task are both indirect measurements of brain activity. Finally,
both measurements were recorded in sequence within a few
hours, but not simultaneously, possibly producing a bias.
Furthermore, the proportion of trait and state properties of
P300 characteristics and brain activity during the delay-dis-
counting task remains difficult to determine exactly.
Funding Open Access funding enabled and organized by Projekt
DEAL. This study was supported by the FORUM of Medical Depart-
ment of Psychiatry of the Ruhr University Bochum (K038-09).
Declarations
Conflict of interest The authors declare that there is no conflict of in-
terest.
Ethical approval All subjects gave written informed consent after the
study was fully explained to them. In accordance with the Helsinki
Declaration of 1975, the study was approved by the local university
ethics committee of the Ruhr University Bochum, Germany.
Open Access This article is licensed under a Creative Commons Attri-
bution 4.0 International License, which permits use, sharing, adapta-
tion, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source,
provide a link to the Creative Commons licence, and indicate if changes
were made. The images or other third party material in this article are
included in the article’s Creative Commons licence, unless indicated
otherwise in a credit line to the material. If material is not included in
the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a
copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
References
1. Goodman WK, Grice DE, Lapidus KAB, Coffey BJ (2014) Obses-
sive-compulsive disorder. Psychiatr Clin N Am 37(3):257–267.
https:// doi. org/ 10. 1016/j. psc. 2014. 06. 004
2. Klanker M, Feenstra M, Denys D (2013) Dopaminergic control of
cognitive flexibility in humans and animals. Front Neurosci 7:201.
https:// doi. org/ 10. 3389/ fnins. 2013. 00201#
3. Pittenger C, Bloch MH, Williams K (2011) Glutamate abnormali-
ties in obsessive compulsive disorder: neurobiology, pathophysiol-
ogy, and treatment. Pharmacol Ther 132(3):314–332. https:// doi.
org/ 10. 1016/j. pharm thera. 2011. 09. 006
4. Nakao T, Okada K, Kanba S (2014) Neurobiological model of
obsessive-compulsive disorder: evidence from recent neuropsy-
chological and neuroimaging findings. Psychiatry Clin Neurosci
68(8):587–605. https:// doi. org/ 10. 1111/ pcn. 12195
5. Burguière E, Monteiro P, Mallet L, Feng G, Graybiel AM (2015)
Striatal circuits, habits, and implications for obsessive-compul-
sive disorder. Curr Opin Neurobiol 30:59–65. https:// doi. org/ 10.
1016/j. conb. 2014. 08. 008
6. Menzies L, Chamberlain SR, Laird AR, Thelen SM, Sahakian
BJ, Bullmore ET (2008) Integrating evidence from neuroim-
aging and neuropsychological studies of obsessive-compulsive
disorder: the orbitofronto-striatal model revisited. Neurosci
Biobehav Rev 32(3):525–549. https:// doi. org/ 10. 1016/j. neubi
orev. 2007. 09. 005
7. Huang WJ, Chen WW, Zhang X (2015) The neurophysiology
of P 300–an integrated review. Eur Rev Med Pharmacol Sci
19(8):1480–1488
8. Mavrogiorgou P, Juckel G, Frodl T, Gallinat J, Hauke W, Zaudig
M, Dammann G et al (2002) P300 subcomponents in obsessive-
compulsive disorder. J Psychiatr Res 36(6):399–406. https:// doi.
org/ 10. 1016/ s0022- 3956(02) 00055-9
9. Miyata A, Matsunaga H, Kiriike N, Iwasaki Y, Takei Y, Yamag-
ami S (1998) Event-related potentials in patients with obsessive-
compulsive disorder. Psychiatry Clin Neurosci 52(5):513–518.
https:// doi. org/ 10. 1046/j. 1440- 1819. 1998. 00427.x
10. Morault PM, Bourgeois M, Laville J, Bensch C, Paty J (1997)
Psychophysiological and clinical value of event-related potentials
in obsessive-compulsive disorder. Biol Psychiatry 42(1):46–56.
https:// doi. org/ 10. 1016/ S0006- 3223(96) 00228-4
11. Towey J, Bruder G, Hollander E, Friedman D, Erhan H, Liebowitz
M, Sutton S (1990) Endogenous event-related potentials in obses-
sive-compulsive disorder. Biol Psychiatry 28(2):92–98. https://
doi. org/ 10. 1016/ 0006- 3223(90) 90626-d
12. Towey J, Bruder G, Tenke C, Leite P, DeCaria C, Friedman D,
Hollander E (1993) Event-related potential and clinical correlates
of neurodysfunction in obsessive-compulsive disorder. Psychia-
try Res 49(2):167–181. https:// doi. org/ 10. 1016/ 0165- 1781(93)
90103-n
13. Andreou C, Leicht G, Popescu V, Pogarell O, Mavrogiorgou P,
Rujescu D, Giegling I et al (2013) P300 in obsessive-compulsive
disorder: source localization and the effects of treatment. J Psy-
chiatr Res 47(12):1975–1983. https:// doi. org/ 10. 1016/j. jpsyc hires.
2013. 09. 003
14. Gohle D, Juckel G, Mavrogiorgou P, Pogarell O, Mulert C,
Rujescu D, Giegling I et al (2008) Electrophysiological evidence
for cortical abnormalities in obsessive-compulsive disorder—a
replication study using auditory event-related P300 subcompo-
nents. J Psychiatr Res 42(4):297–303. https:// doi. org/ 10. 1016/j.
jpsyc hires. 2007. 01. 003
15. Towey JP, Tenke CE, Bruder GE, Leite P, Friedman D, Liebow-
itz M, Hollander E (1994) Brain event-related potential corre-
lates of overfocused attention in obsessive-compulsive disorder.
1 3 338
European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
Psychophysiology 31(6):535–543. https:// doi. org/ 10. 1111/j. 1469-
8986. 1994. tb023 46.x
16. Malloy P, Rasmussen S, Braden W, Haier RJ (1989) Topographic
evoked potential mapping in obsessive-compulsive disorder: evi-
dence of frontal lobe dysfunction. Psychiatry Res 28(1):63–71.
https:// doi. org/ 10. 1016/ 0165- 1781(89) 90198-4
17. Sanz M, Molina V, Martin-Loeches M, Calcedo A, Rubia FJ
(2001) Auditory P300 event related potential and serotonin
reuptake inhibitor treatment in obsessive-compulsive disorder
patients. Psychiatry Res 101(1):75–81. https:// doi. org/ 10. 1016/
s0165- 1781(00) 00250-x
18. Ma Q, Wang X, Shu L, Dai S (2008) P300 and categorization in
brand extension. Neurosci Lett 431(1):57–61. https:// doi. org/ 10.
1016/j. neulet. 2007. 11. 022
19. Polezzi D, Daum I, Rubaltelli E, Lotto L, Civai C, Sartori G,
Rumiati R (2008) Mentalizing in economic decision-making.
Behav Brain Res 190(2):218–223. https:// doi. org/ 10. 1016/j. bbr.
2008. 03. 003
20. Polezzi D, Sartori G, Rumiati R, Vidotto G, Daum I (2010) Brain
correlates of risky decision-making. Neuroimage 49(2):1886–
1894. https:// doi. org/ 10. 1016/j. neuro image. 2009. 08. 068
21. Rigoni D, Polezzi D, Rumiati R, Guarino R, Sartori G (2010)
When people matter more than money: an ERPs study. Brain Res
Bull 81(4–5):445–452. https:// doi. org/ 10. 1016/j. brain resbu ll.
2009. 12. 003
22. Yeung N, Sanfey AG (2004) Independent coding of reward mag-
nitude and valence in the human brain. J Neurosci 24(28):6258–
6264. https:// doi. org/ 10. 1523/ JNEUR OSCI. 4537- 03. 2004
23. He JM, Huang XT, Yuan H, Chen YG (2012) Neural activity in
relation to temporal distance: differences in past and future tem-
poral discounting. Conscious Cogn 21(4):1662–1672. https:// doi.
org/ 10. 1016/j. concog. 2012. 08. 006
24. Li JZ, Gui DY, Feng CL, Wang WZ, Du BQ, Gan T, Luo YJ
(2012) Victims’ time discounting 2.5 years after the Wenchuan
earthquake: an ERP study. PLoS ONE 7(7):e40316. https:// doi.
org/ 10. 1371/ journ al. pone. 00403 16
25. Wang J, Chen J, Lei Y, Li P (2014) P300, not feedback error-
related negativity, manifests the waiting cost of receiving reward
information. NeuroReport 25(13):1044–1048. https:// doi. org/ 10.
1097/ WNR. 00000 00000 000226
26. Wu H, Gui D, Lin W, Gu R, Zhu X, Liu X (2016) The procras-
tinators want it now: behavioral and event-related potential evi-
dence of the procrastination of intertemporal choices. Brain Cogn
107:16–23. https:// doi. org/ 10. 1016/j. bandc. 2016. 06. 005
27. Brevers D, Bechara A, Cleeremans A, Noël X (2013) Iowa Gam-
bling Task (IGT): twenty years after—gambling disorder and IGT.
Front Psychol 4:665. https:// doi. org/ 10. 3389/ fpsyg. 2013. 00665
28. Brand M, Labudda K, Markowitsch HJ (2006) Neuropsychologi-
cal correlates of decision-making in ambiguous and risky situa-
tions. Neural Netw 19(8):1266–1276. https:// doi. org/ 10. 1016/j.
neunet. 2006. 03. 001
29. Cavedini P, Zorzi C, Piccinni M, Cavallini MC, Bellodi L (2010)
Executive dysfunctions in obsessive-compulsive patients and
unaffected relatives: searching for a new intermediate phenotype.
Biol Psychiatry 67(12):1178–1184. https:// doi. org/ 10. 1016/j. biops
ych. 2010. 02. 012
30. Starcke K, Tuschen-Caffier B, Markowitsch HJ, Brand M (2009)
Skin conductance responses during decisions in ambiguous and
risky situations in obsessive-compulsive disorder. Cogn Neu-
ropsychiatry 14(3):199–216. https:// doi. org/ 10. 1080/ 13546 80090
29968 31
31. Starcke K, Tuschen-Caffier B, Markowitsch HJ, Brand M (2010)
Dissociation of decisions in ambiguous and risky situations in
obsessive-compulsive disorder. Psychiatry Res 175(1–2):114–
120. https:// doi. org/ 10. 1016/j. psych res. 2008. 10. 022
32. Zhang L, Dong Y, Ji Y, Zhu C, Yu F, Ma H, Chen X, Wang K
(2015) Dissociation of decision making under ambiguity and deci-
sion making under risk: a neurocognitive endophenotype candi-
date for obsessive-compulsive disorder. Prog Neuropsychophar-
macol Biol Psychiatry 57:60–68. https:// doi. org/ 10. 1016/j. pnpbp.
2014. 09. 005
33. MacKillop J, Amlung MT, Few LR, Ray LA, Sweet LH, Munafò
MR (2011) Delayed reward discounting and addictive behavior: a
meta-analysis. Psychopharmacology 216(3):305–321. https:// doi.
org/ 10. 1007/ s00213- 011- 2229-0
34. Stanger C, Elton A, Ryan SR, James GA, Budney AJ, Kilts CD
(2013) Neuroeconomics and adolescent substance abuse: indi-
vidual differences in neural networks and delay discounting. J Am
Acad Child Adolesc Psychiatry 52(7):747-755.e6. https:// doi. org/
10. 1016/j. jaac. 2013. 04. 013
35. Pinto A, Steinglass JE, Greene AL, Weber EU, Simpson HB
(2014) Capacity to delay reward differentiates obsessive-com-
pulsive disorder and obsessive-compulsive personality disorder.
Biol Psychiatry 75(8):653–659. https:// doi. org/ 10. 1016/j. biops
ych. 2013. 09. 007
36. Peters J, Büchel C (2010) Episodic future thinking reduces reward
delay discounting through an enhancement of prefrontal-medi-
otemporal interactions. Neuron 66(1):138–148. https:// doi. org/ 10.
1016/j. neuron. 2010. 03. 026
37. Mavrogiorgou P, Enzi B, Steinmann S, Mulert C, Juckel G (2018)
Relationship between neuroanatomical and serotonergic hypoth-
eses of obsessive-compulsive disorder: a combined functional
magnetic resonance imaging-evoked potential study. J Clin Psy-
chiatry. https:// doi. org/ 10. 4088/ JCP. 17m11 811
38. Polich J (2007) Updating P300: an integrative theory of P3a and
P3b. Clinical neurophysiology. Clin Neurophysiol 118(10):2128–
2148. https:// doi. org/ 10. 1016/j. clinph. 2007. 04. 019
39. American Psychiatric Association (2000) Diagnostic and statisti-
cal manual of mental disorders: DSM-IV-TR, 4th edn. American
Psychiatric Association, Washington
40. Dilling H (Ed) (1999) Taschenführer zur ICD-10 klassifikation
psychischer störungen: mit glossar und diagnostischen kriterien
ICD-10, DCR-10 (1. Aufl) Huber, Bern, Göttingen, Toronto,
Seattle
41. Ackenheil M, Stotz G, Dietz-Bauer R (1997) MINI PLUS: Version
4.5. München, Germany
42. Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J,
Weiller E, Hergueta T et al (1998) The Mini-International Neu-
ropsychiatric Interview (MINI): the development and validation
of a structured diagnostic psychiatric interview for DSM-IV and
ICD-10. J Clin Psychiatry 59(Suppl 20):22–33 (Quiz:34–57)
43. Goodman WK, Price LH, Rasmussen SA, Mazure C, Fleischmann
RL, Hill CL, Heninger GR, Charney DS (1989) The Yale-brown
obsessive compulsive scale. I. Development, use, and reliability.
Arch Gen Psychiatry 46(11):1006–1011. https:// doi. org/ 10. 1001/
archp syc. 1989. 01810 11004 8007
44. Goodman WK, Price LH, Rasmussen SA, Mazure C, Delgado
P, Heninger GR, Charney DS (1989) The Yale-brown obsessive
compulsive scale. II. Validity. Arch Gen Psychiatry 46(11):1012–
1016. https:// doi. org/ 10. 1001/ archp syc. 1989. 01810 11005 4008
45. Hodgson RJ, Rachman S (1977) Obsessional-compulsive com-
plaints. Behav Res Ther 15(5):389–395. https:// doi. org/ 10. 1016/
0005- 7967(77) 90042-0
46. Hamilton M (1967) Development of a rating scale for primary
depressive illness. Br J Soc Clin Psychol 6(4):278–296. https://
doi. org/ 10. 1111/j. 2044- 8260. 1967. tb005 30.x
47. Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J (1961)
An inventory for measuring depression. Arch Gen Psychiatry
4:561–571. https:// doi. org/ 10. 1001/ archp syc. 1961. 01710 12003
1004
1 3 European Archives of Psychiatry and Clinical Neuroscience (2022) 272:327–339
339
48. Spielberger CD (1970) Manual for the state-trait anxiety inven-
tory: self-evaluation questionnaire. Consulting Psychologists
Press, Palo Alto
49. Laux L, Spielberger CD (1981) Das State-Trait-Angstinventar:
STAI; theoretische Grundlagen und Handanweisung. Beltz-Test.
Beltz, Weinheim
50. Guy W, Bonato RR (1970) Manual for the ECDEU assessment
battery. Department of Health, Education, and Welfare, Bethesda
51. Morosini PL, Magliano L, Brambilla L, Ugolini S, Pioli R (2000)
Development, reliability and acceptability of a new version of
the DSM-IV Social and Occupational Functioning Assessment
Scale (SOFAS) to assess routine social functioning. Acta Psychi-
atr Scand 101(4):323–329
52. Barratt ES (1959) Anxiety and impulsiveness related to psycho-
motor efficiency. Percept Mot Skills 9(3):191–198. https:// doi. org/
10. 2466/ PMS. 1959.9. 3. 191
53. Preuss UW, Rujescu D, Giegling I, Koller G, Bottlender M, Engel
RR, Möller HJ, Soyka M (2003) Evaluation der deutschen Ver-
sion der Barratt Impulsiveness Scale (BIS 5). Fortschr Neurol
Psychiatr 71(10):527–534. https:// doi. org/ 10. 1055/s- 2003- 42872
54. Borkenau P, Ostendorf F (1993) NEO-Fünf-Faktoren-Inventar
(NEO-FFI) nach Costa und McCrae: Handanweisung. Hogrefe,
Göttingen
55. Lehrl S, Triebig G, Fischer B (1995) Multiple choice vocabulary
test MWT as a valid and short test to estimate premorbid intel-
ligence. Acta Neurol Scand 91(5):335–345. https:// doi. org/ 10.
1111/j. 1600- 0404. 1995. tb070 18.x
56. Polich J (2007) Updating P300: an integrative theory of P3a and
P3b. Clin Neurophysiol 118(10):2128–2148. https:// doi. org/ 10.
1016/j. clinph. 2007. 04. 019
57. Pascual-Marqui RD (2002) Standardized low-resolution brain
electromagnetic tomography (sLORETA): technical details. Meth-
ods Find Exp Clin Pharmacol 24(Suppl D):5–12
58. Nichols TE, Holmes AP (2002) Nonparametric permutation tests
for functional neuroimaging: a primer with examples. Hum Brain
Mapp 15:R691–R695. https:// doi. org/ 10. 1002/ hbm. 1058
59. Whitlock JR (2017) Posterior parietal cortex. Curr Biol
27(14):R691–R695. https:// doi. org/ 10. 1016/j. cub. 2017. 06. 007
60. Mavrogiorgou P, Enzi B, Klimm AK, Köhler E, Roser P, Norra C,
Juckel G (2017) Serotonergic modulation of orbitofrontal activity
and its relevance for decision making and impulsivity. Hum Brain
Mapp 38(3):1507–1517. https:// doi. org/ 10. 1002/ hbm. 23468
61. Brett M, Anton JL, Valabrgue R, Poline JB (2002) Region of
interest analysis using an spm toolbox [abstract]. Presented at the
8th international conference on functional mapping of the human
brain, June 2–6. Sendai, Japan
62. Rector NA, Hood K, Richter MA, Bagby RM (2002) Obsessive-
compulsive disorder and the five-factor model of personality:
assessing the distinction and overlap with major depressive dis-
order. Behav Res Ther 40:1205–1219. https:// doi. org/ 10. 1016/
S0005- 7967(02) 00024-4
63. Samuels J, Nestadt G, Bienvenu OJ, Costa PT Jr, Riddle MA,
Liang KY, Hoehn-Saric R, Grados MA, Cullen BA (2000) Person-
ality disorders and normal personality dimensions in obsessive-
compulsive disorder. Br J Psychiatry 177:457–462. https:// doi.
org/ 10. 1192/ bjp. 177.5. 457
64. Rector NA, Richter MA, Bagby RM (2005) The impact of per-
sonality on symptom expression in obsessive-compulsive disorder.
J Nerv Ment Dis 193:231–236. https:// doi. org/ 10. 1097/ 01. nmd.
00001 58378. 14019. 89
65. Dayan-Riva A, Berger A, Anholt GE (2019) Early cognitive
processes in OCD: an ERP study. J Affect Disord 246:429–436.
https:// doi. org/ 10. 1016/j. jad. 2018. 12. 109
66. Linden DEJ (2005) The p300: where in the brain is it produced
and what does it tell us? Neuroscientist 11(6):563–576. https://
doi. org/ 10. 1177/ 10738 58405 280524
67. Bellebaum C, Polezzi D, Daum I (2010) It is less than you
expected: the feedback-related negativity reflects violations of
reward magnitude expectations. Neuropsychologia 48(11):3343–
3350. https:// doi. org/ 10. 1016/j. neuro psych ologia. 2010. 07. 023
68. Sachdev PS, Malhi GS (2005) Obsessive-compulsive behaviour:
a disorder of decision-making. Aust N Z J Psychiatry 39(9):757–
763. https:// doi. org/ 10. 1080/j. 1440- 1614. 2005. 01680.x
69. Gruner P, Anticevic A, Lee D, Pittenger C (2016) Arbitration
between action strategies in obsessive-compulsive disorder. Neu-
roscientist 22(2):188–198. https:// doi. org/ 10. 1177/ 10738 58414
568317
70. Pushkarskaya H, Tolin D, Ruderman L, Kirshenbaum A, Kelly
JM, Pittenger C, Levy I (2015) Decision-making under uncer-
tainty in obsessive-compulsive disorder. J Psychiatr Res 69:166–
173. https:// doi. org/ 10. 1016/j. jpsyc hires. 2015. 08. 011
71. Okasha A, Rafaat M, Mahallawy N, El Nahas G, El Dawla AS,
Sayed M, El Kholi S (2000) Cognitive dysfunction in obsessive-
compulsive disorder. Acta Psychiatr Scand 101(4):281–285
72. Benzina N, Mallet L, Burguière E, N’Diaye K, Pelissolo A
(2016) Cognitive dysfunction in obsessive-compulsive dis-
order. Curr Psychiatry Rep 18(9):80. https:// doi. org/ 10. 1007/
s11920- 016- 0720-3
73. Kenemans J, Kähkönen S (2011) How human electrophysiology
informs psychopharmacology: from bottom-up driven processing
to top-down control. Neuropsychopharmacol 36:26–51. https://
doi. org/ 10. 1038/ npp. 2010. 157
74. Fettes P, Schulze L, Downar J (2017) Cortico-striatal-thalamic
loop circuits of the orbitofrontal cortex: promising therapeutic
targets in psychiatric illness. Front Syst Neurosci 11:25. https://
doi. org/ 10. 3389/ fnsys. 2017. 00025
1 3
| null |
10.1038_s42003-023-04981-1.pdf
|
Data availability
Plasmid sequences are available through the PLSDB database (https://ccb-microbe.cs.
uni-saarland.de/plsdb/) while plasmid metadata files are hosted in a Github repository
(https://github.com/LBHarrison/Lociq/). Source data for Figs. 3, 6 and 8 are provided in
Supplementary Data 3, 7 and 8, respectively.
Code availability
The Lociq program is available through the Github repository https://github.com/
LBHarrison/Lociq/38.
|
Data availability Plasmid sequences are available through the PLSDB database ( https://ccb-microbe.cs. uni-saarland.de/plsdb/ ) while plasmid metadata files are hosted in a Github repository ( https://github.com/LBHarrison/Lociq/ ). Source data for Figs. 3, 6 and 8 are provided in Supplementary Data 3, 7 and 8, respectively. Code availability The Lociq program is available through the Github repository https://github.com/ LBHarrison/Lociq/ 38 .
|
ARTICLE
https://doi.org/10.1038/s42003-023-04981-1
OPEN
Lociq provides a loci-seeking approach for
enhanced plasmid subtyping and structural
characterization
1✉
Lucas Harrison
, Shaohua Zhao1, Cong Li1, Patrick F. McDermott1, Gregory H. Tyson1 & Errol Strain1
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
Antimicrobial resistance (AMR) monitoring for public health is relying more on whole gen-
ome sequencing to characterize and compare resistant strains. This requires new approaches
to describe and track AMR that take full advantage of the detailed data provided by genomic
technologies. The plasmid-mediated transfer of AMR genes is a primary concern for AMR
monitoring because plasmid rearrangement events can integrate new AMR genes into the
plasmid backbone or promote hybridization of multiple plasmids. To better monitor plasmid
evolution and dissemination, we developed the Lociq subtyping method to classify plasmids
by variations in the sequence and arrangement of core plasmid genetic elements. Subtyping
with Lociq provides an alpha-numeric nomenclature that can be used to denominate plasmid
population diversity and characterize the relevant features of individual plasmids. Here we
demonstrate how Lociq generates typing schema to track and characterize the origin, evo-
lution and epidemiology of multidrug resistant plasmids.
1 Center for Veterinary Medicine, U.S. Food and Drug Administration, Laurel, MD, USA.
✉
email: [email protected]
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
1
ARTICLE
COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-04981-1
Plasmid-mediated antimicrobial resistance (AMR) allows
bacteria to resist exposure to every major class of antibiotics.
Transferrable plasmids can disseminate AMR genes between
bacterial genera and facilitate the spread of antimicrobial-resistant
pathogens among host species1. The National Antimicrobial
Resistance Monitoring System (NARMS) recognizes plasmid-
mediated AMR as a key threat to human, companion animal and
food animal health. However, plasmids that encode for AMR genes
are prone to genetic recombination events2. This capacity for
genetic remodeling contributes to the great sequence diversity seen
among plasmids and confounds current efforts to track and char-
acterize these clinically relevant molecules3–8.
that
The most common plasmid typing method categorizes plas-
mids by a single conserved region on the plasmid replicon9. This
method, plasmid incompatibility group typing (Inc typing), uses a
PCR based replicon typing approach and emerged from research
on the effects of plasmid replicon pairs and plasmid replication
efficiency10. Plasmid combinations
in decreased
replication efficiency when concurrently occupying the same cell
are classified within the same incompatibility group. This criter-
ion was well-suited for analysis with contemporaneous molecular
or in silico methods because it only required identification of a
single target11. However, this reliance on a single genetic target
does not address the great sequence diversity present among
plasmids within a single Inc group and often does not detect
hybrid plasmids12. This shortfall of using a single target for
plasmid typing is apparent when the plasmid contains multiple
replicon sequences13,14.
result
A complementary method to plasmid Inc typing known as
MOB typing categorizes plasmids by the sequence of
their
relaxase protein15,16. The relaxase protein is an essential com-
ponent in mobilizable plasmids that binds to the plasmid origin
of transfer, introduces a single stranded nick and facilitates the
transfer of the single plasmid strand to the bacterial plasmid
secretion system17. Relaxase proteins have been phylogenetically
grouped into six MOB families and plasmids are assigned to a
MOB group based on the relaxase protein sequence16. Unfortu-
nately, MOB typing methods are limited in their ability to cate-
gorize non-mobilizable plasmids and like Inc typing are based on
only a single target.
One promising typing approach classifies plasmids by their
average nucleotide identity18. This approach has a notable
advantage over other typing schema because it uses the entire
plasmid sequence to identify plasmid taxonomic units (PTUs)
instead of using a single target. The PTU method identifies
conserved taxonomic units using a sequence-length dependent
comparison between plasmids. One of the main advantages of
this method is that it classifies plasmids independent from any
predicted phenotypic trait or function. This sequence-based
approach has shown strong associations between PTU group and
bacterial host specificity19. However, this approach does have
limitations. First, because this method makes length-based com-
parisons between plasmids it is possible to miss regions of
sequence similarity in smaller plasmids when they fall below the
method’s cutoff threshold. Second, the naming schema of the
PTU system is independent of other typing systems, complicating
comparisons to historical plasmid data. Finally, similar to other
average nucleotide identity clustering methods, this method does
not take into account variations in the plasmid structure resulting
from recombination events. These limitations hinder the ability to
make detailed comparisons between plasmids using the PTU
designation alone.
Plasmid multilocus sequence typing (PMLST) addresses some of
the challenges of Inc, MOB and average nucleotide identity typing
methods. Schema that contain more than one target for plasmid
typing are able to account for a greater degree of sequence diversity
within a plasmid type3. Unlike the MOB methods, PMLST is able
to categorize non-mobilizable plasmids as well. PMLST methods
are compatible with existing plasmid typing nomenclature and the
typing loci are defined sequences that can be used in downstream
analysis. The IncA/C3, IncF4, IncHI6, IncH25, IncI17 and IncN8
PMLST schema contain 2-6 typing loci each and have contributed
greatly to the understanding of plasmid sequence diversity. The
IncA/C PMLST schema is used to differentiate the plasmids of the
IncC plasmid type. The IncC plasmids are commonly associated
with the carriage of clinically-relevant antimicrobial resistance
genes and contribute to the spread of the multi-drug resistant
phenotype20. Core genome plasmid multilocus sequence typing
(cgPMLST) expands on PMLST methods further by identifying the
genes essential
for plasmid maintenance and using them as
sequence typing targets3. This method has been applied to IncA/C
plasmids to increase the number of typing loci to 28. However,
while more targets are used for PMLST- based plasmid classifica-
tion, they only represent a small percentage of the entire plasmid
sequence and provide little information on structural differences
between plasmids.
One factor that has hindered the progress of sequence-based
plasmid typing systems is the difficulty of assembling plasmids
from short read sequencing data21. However, as long-read
sequencing technologies become more accessible, more closed
plasmid assemblies are available to researchers. Closed plasmid
assemblies offer two main advantages over gapped, or draft,
plasmid assemblies. First, closed plasmid assemblies account for
every nucleotide on the plasmid molecule. This provides a full
accounting of all the coding and intergenic regions on the plas-
mid. The second advantage closed assemblies provide is the
ability to determine which sequences are missing from a plasmid.
For comparison, draft assemblies do not contain the entire
plasmid sequence and cannot be used to determine if a given
sequence is missing. Finally, closed assemblies can be used to
identify the relative position of any genetic element on the plas-
mid. This attribute is useful in epidemiological operations such as
antimicrobial resistance monitoring where the proximity of an
AMR gene to an insertion sequence or transposon can help assess
the risk of gene transfer.
These three attributes of closed plasmid assemblies are ideal
factors for plasmid typing. First, the ability to account for every
nucleotide on the plasmid increases the likelihood of identifying
common sequences shared among different plasmids. Second,
the ability to equate absence of sequence in an assembly to
absence of sequence in the cognate plasmid allows for plasmid
classification methods based on the presence or absence of
genetic elements. Finally, analyzing the relative position of each
genetic element on the plasmid can account for differences in
the plasmid structure resulting from plasmid recombination
and insertion events.
Here we present a plasmid subtyping method that uses closed
plasmid assemblies to identify the conserved sequences and pat-
terns of loci found among plasmids of a given plasmid type. In this
paper, we propose to subtype plasmids of the IncC plasmid type as
a demonstration of the Lociq method. We chose the IncC plasmids,
not only because of their role in the transmission of AMR genes,
but also because we can compare results of the Lociq method to the
PMLST and cgPMLST profiles of this well-characterized plasmid
type. By identifying these conserved genetic elements and patterns,
we aim to develop a scalable approach to plasmid classification that
allows the user to first identify large families of plasmids and then
apply additional typing criteria to differentiate between individual
plasmids. The purpose of this paper is to introduce the plasmid
subtyping method, demonstrate its ability to subtype IncC plas-
mids, compare it to existing plasmid typing methods, and show
how the results of the subtyping method can be used to facilitate
2
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-04981-1
ARTICLE
Fig. 1 Lociq workflow for the identification of plasmid typing Loci. Overview of the typing loci identification process of the Lociq method.
research in plasmid biology, which has the potential to enhance
pathogen surveillance for public health.
Results
We demonstrated the utility of the Lociq plasmid typing method
by performing an analysis of closed plasmid assemblies and
generating subtyping definitions for the IncC plasmids. Identifi-
cation of the typing loci was performed by using the Roary and
piggy programs to define the pangenome of 459 closed plasmid
sequences22,23. Prevalence thresholds were used to determine
which pangenomic loci were indicative of and exclusive to a given
plasmid type. Finally, the candidate typing loci were validated
against an external database (Fig. 1). We then compared the Lociq
typing method results to Inc, MOB and PTU typing methods, as
well as PMLST and cgMLST subtyping methods. Finally, we
demonstrated how the Lociq method organizes the results to
facilitate downstream analyses.
Plasmid subtyping method. The full dataset of Salmonella and E.
isolates contained 459 closed plasmid assemblies and 46
coli
plasmid Inc types. These 46 plasmid types were represented by
398 plasmids and the remaining 61 plasmids did not belong to
any plasmid Inc group. The combined pangenome for all 459
plasmids contained 6726 unique coding and intergenic regions, as
generated by the Roary & piggy programs. These 6726 genetic
elements were the library of plasmid loci found among our
plasmids. The pangenome was analyzed as a binary presence/
absence matrix in R where plasmids were grouped by the simi-
larity of their loci profiles accounting for both the coding and
intergenic regions. This grouping was performed first by com-
puting a distance matrix of the binary matrix data, then clustering
with the hclust function using complete linkage. The resulting
presence absence matrix was used for downstream subtyping of
the Inc group plasmid typing schema (Fig. 2).
Next, we identified the IncC cluster on the presence-absence
matrix and selected the loci indicative of and exclusive to IncC
plasmids. Identification of the IncC plasmids revealed that the
loci composition of IncC plasmids is not uniform and only a
subset of loci is shared among the IncC plasmids (Fig. 2). Next,
we identified the loci indicative of and selective for IncC plasmids
by comparing the prevalence of each 6726 loci among IncC
plasmids to their prevalence in non-IncC plasmids. Seventy-five
loci were present in >90% of the IncC plasmids and fewer
than 10% of the non-IncC plasmids (Supplementary Fig. 1).
This initial set of IncC typing loci contained 59 coding and 16
intergenic regions.
Following the initial identification of typing loci, we evaluated
the loci against plasmids in an external
the prevalence of
database. The purpose of this analysis was to reduce the bias in
loci selection that may be introduced if the initial dataset was not
representative of the broader plasmid population. For example, in
this demonstration, all plasmids were harvested from Salmonella
and E. coli strains that were isolated from retail meats or food
animal cecal samples. The plasmid data set did not contain
plasmids harvested from other genera of bacteria and none of the
bacteria were isolated from clinical or environmental sources. To
address this, we evaluated the prevalence of the 75 loci among the
34,513 plasmids of the PLSDB v.2021_06_23_v2 database24. We
compared the prevalence of typing loci between IncC and non-
IncC plasmids in the database. Seventy-two of the seventy-five
IncC typing loci met the two criteria of being present in > 90% of
the plasmids that contained at least one typing locus and being
present in < 1% of plasmids without a typing locus in the PLSDB
database. The resulting complement of 72 IncC typing loci
accounted for 40,091 bp and contained 58 coding regions and 14
intergenic regions. Further, a 90% prevalence of loci threshold
was sufficient to identify all 534 IncC plasmids in the PLSDB
database.
In the next stage of this plasmid subtyping demonstration, we
identified the patterns of contiguous plasmid loci that were
conserved among the IncC plasmids in the PLSDB database.
These conserved contiguous regions were identified as fragments
of the plasmid backbone. The fragment analysis that allowed for
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
3
ARTICLE
COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-04981-1
Fig. 2 Pangenome of 459 closed plasmid sequences. Presence Absence matrix of loci in the plasmid dataset. Plasmids are arranged along the x-axis while
the loci of coding and intergenic regions are organized along the y-axis. Dark red color indicates the presence of a locus in a given plasmid. Plasmid Inc
group assignments of the most common Inc types are visually located above each plasmid as well as in tabular format in Supplementary Data 1. Multiple
Inc group markers in a single plasmid are stacked vertically. The larger cluster of IncC plasmids is delineated by solid lines while the two IncC subclusters
are subdivided by a dotted line.
no greater than 500 bp between neighboring loci revealed that the
IncC plasmids contained 8 conserved plasmid fragments. These
fragments contained 2–31 typing loci (Fig. 3) and the loci
sequences on the fragments ranged in size from 236 bp to
13,836 bp (Supplementary Data 2). The mean correlation
coefficient for the 31 loci on the largest plasmid fragment was
0.989 (Supplementary Data 3). Fragment 6 had the lowest mean
correlation coefficient among its 4 loci with an R-value of 0.916.
The 417 bp PlasmidFinder IncC marker was contained within a
1066 bp locus found on plasmid fragment 1. This fragment
contained 8 loci with a mean R-value among its loci of 0.942.
In the final stage of our demonstration of IncC plasmid analysis
with the Lociq method, we used the sequence and position of the
typing loci to characterize all plasmids from the external database
that contained at least 1 IncC typing locus (Supplementary Fig. 2).
Plasmid characterization was performed by assigning a numeric
identifier to each unique pattern of sequence type, fragment type
and loci type (Fig. 4). The plasmid sequence type was defined by the
complement of plasmid alleles in the plasmid, regardless of their
position. The plasmid fragment type was determined by how the
plasmid fragments were ordered along the plasmid, relative to a
type was
semi-conservative starting locus. The plasmid loci
determined by rearranging the plasmid fragments in ascending
order of their numeric identifier and recovering the arrangement of
loci
from the re-ordered plasmid fragments. This temporary
rearrangement of plasmid fragments for loci typing allows the loci
type to be independent of the fragment type.
the analysis
In addition to the 534 IncC plasmids in the database of 34,513
identified 31 IncC hybrid plasmids
plasmids,
that contained at least 1 of the IncC typing loci. The 534 IncC
plasmids were then subdivided into unique patterns of 52 fragment
types, 260 loci types and 388 sequence types. There were 397
unique combinations of fragment type, loci type and sequence type
represented among the 534 IncC plasmids (Supplementary Data 4).
Further, the addition of the interfragment distance values to the
subtyping criteria increased the number of unique combinations to
515. As a result, the 534 IncC plasmids could be divided into 515
unique combinations of fragment type, loci type, sequence type and
interfragment distances.
The Lociq plasmid subtyping method includes features for
analysis of the results. First, the results can be evaluated in a web-
browser using an R-shiny application25. This application allows the
user to compare plasmids by generating a graphical map (Fig. 5) of
each plasmid in the database (Supplementary Fig. 3), a report of
plasmid features (Supplementary Fig. 4) and a searchable table of
AMR genes that are present in the plasmid (Supplementary Fig. 5)
4
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-04981-1
ARTICLE
Fig. 3 Correlation matrix and fragment assignments of IncC Plasmid Loci. A correlogram illustrating the strength of correlation between any two loci
occurring on the same contiguous region of an IncC plasmid (n = 565 plasmid sequences that contain IncC typing loci). The conserved fragments can be
seen as dark blue circles in eight distinct triangular shapes along the diagonal. Correlation coefficients corresponding to insignificant loci interactions
(p ≥ 0.05) are represented as blank cells. Source data for the correlogram may be found in Supplementary Data 3.
and the full plasmid database (Supplementary Fig. 6). Second, this
subtyping method generates a tabular typing summary of all the
plasmids that were evaluated (Table 1). This summary includes
the plasmid ID, plasmid typing category, fragment type,
loci
type, plasmid sequence type, fragment sequence types and the
interfragment distances (Supplementary Data 4). Third,
the
method produces sequence (Supplementary Data 5) and pattern
(Supplementary Data 6) definitions for downstream analysis.
Finally, this subtyping method includes a script that allows the user
to characterize their own plasmid sequences using the database of
results generated by the Lociq method. The database will also
update the plasmid typing reference database if the user’s plasmid
sequences contain variants in sequence or structure that were not
previously accounted for.
Comparison to existing methods. Our subtyping method clas-
sifies plasmids by variations in loci sequence and relative position
on the plasmid. We compared the total number of subtyping
groups, the size of each group and the Simpson diversity index
across four plasmid typing methods and the Lociq typing method
to evaluate their discriminatory power (Fig. 6). The first two
methods we evaluated were the MOB type and the PTU typing
methods. While neither of these classification methods were
designed for IncC plasmid subtyping, both are valuable alter-
natives to the Inc typing system. PTU classification of the plas-
mids was able to assign 479 of the 534 IncC plasmids to a PTU
group. The largest group contained 475 plasmids and the results
that were generated had a Simpson’s index of diversity of 0.199.
MOB typing of the 534 IncC plasmids revealed 15 MOB types,
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
5
ARTICLE
COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-04981-1
Fig. 4 Metrics for plasmid typing. Endpoints evaluated in the Lociq plasmid typing method (a). Boxes represent plasmid loci while the numbered clusters
of loci correspond to plasmid fragments. Examples of how the endpoints can be used to differentiate between two example plasmids A and B (b) using the
sequence of plasmid loci to determine plasmid sequence type, order of the plasmid loci to determine loci type, order of the plasmid fragments to determine
fragment type or the distances between the plasmid fragments as a metric for interfragment distances.
Fig. 5 Lociq characterization of IncC plasmids. A graphical representation of nine IncC plasmids generated by the Lociq companion application. The
numbered black bars represent plasmid fragments, red bars represent AMR genes and yellow bars represent stress-tolerance genes. Strand orientation is
in relation to the plasmid indexing locus and forward orientation is represented by gene presence above the sequence line.
6
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-04981-1
ARTICLE
Table 1 Sample Results from IncC Plasmid Subtyping.
Plasmid
Fragment Pattern
Loci Pattern
Sequence Type
Interfragment Distance (bp)
AP022381.1
AP022385.1
AP024844.1
CP063757.1
NZ_CP045517.1
NZ_CP048384.1
NZ_MH995506.1
NZ_LT985224.1
NZ_CP065463.1
25
25
25
27
24
30
45
30
30
3
3
2
134
205
77
255
151
79
376
376
381
147
126
226
319
75
239
2942, 14165, 870, 115695, 17808, 7594
2942, 14165, 870, 84976, 17808, 7594
2942, 1030, 870, 103967, 17766, 3933, 3133
1226, 1030, 870, 40258, 2955, 1100, 13770
1226, 1030, 870, 15107, 637, 688, 829, 25736
1226, 1030, 870, 130822, 829, 688, 637, 11581
637, 11153, 1142, 1226, 1030, 870, 80076, 829
1226, 1030, 11687, 870, 70622, 829, 688, 637, 27441
6378, 870, 71929, 829, 688, 637, 25439
Fig. 6 Comparison of plasmid typing methods. Classification comparison of 6 plasmid typing methods. A stacked barplot comparing the size of each
subgroup that was identified from the initial dataset of 534 IncC plasmids across 6 typing methods. Subgroups for each method are arranged in decreasing
size from left to right and source data are available in Supplementary Data 7.
the largest of which contained 425 plasmids. MOB typing of this
dataset generated a Simpson’s index of diversity of 0.359. The
next two methods we evaluated were specifically designed to
subtype IncC plasmids and showed greater ability to differentiate
between plasmids. The first of these methods was the 5 loci IncA/
C PMLST schema which produced 28 groups. The largest group
classified by this method contained 363 plasmids and the diver-
sity index for this method was 0.492. The final comparator typing
method was the 28 loci IncA/C cgPMLST schema. There were
180 unique combinations of IncA/C cgPMLST alleles represented
in the dataset and the most common combination was identified
in 87 plasmids. Typing with the IncA/C cgPMLST loci showed
the greatest discriminatory power of all the comparator methods
with a Simpson’s diversity index of 0.954.
Next, we evaluated the typing schema generated in our plasmid
subtyping method (Fig. 6). Structural characterization of the
plasmids by the order of their fragments grouped the 534 IncC
plasmids into 53 groups, the largest of which contained 386
plasmids. Fragment typing had slightly greater discriminatory
power than MOB typing, as indicated by a Simpson’s diversity
index of 0.475. Structural characterization of plasmids by the
order of their loci classified the plasmids into 260 groups. The
largest group contained 171 plasmids and the Simpson’s
diversity index for this schema was 0.896. This value was
slightly less than the diversity index of the IncA/C cgMLST
method. The plasmid classification schema that grouped
plasmids by the plasmid loci sequence type that were generated
in our method grouped the plasmids into 388 groups. The
largest group that was produced with this schema contained 15
plasmids. This schema had the second highest Simpson’s
diversity index of 0.996. The final schema that we evaluated
combined all the structural and sequence features that were
generated in the analysis. For this aggregate schema, plasmids
were evaluated by their fragment type, loci type, sequence type
and the distances between their fragments. This separated the
plasmids into 515 groups, and the largest group contained 4
plasmids. This schema had the greatest discriminatory power
with a Simpson’s diversity index >0.999.
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
7
ARTICLE
COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-04981-1
Fig. 7 Alignment of IncC plasmids with insertion sequence custom annotations. Alignment of two IncC plasmids with Insertion Sequence (IS) element
annotations and sequence alignments added. Black bars represent IncC plasmid fragments, red bars represent AMR genes, yellow bars represent stress
tolerance genes and purple bars represent IS elements. Light red alignments depict conserved regions in the same orientation in both plasmids while dark
blue alignments show sequence inversions. Notice the two prominent sequence inversions in AP024125.1 are immediately flanked by IS6 elements.
of plasmids NC_012690.1 and AP024125.1 illustrates the proxi-
mity of IS elements to AMR and heavy metal resistance genes
(Fig. 7). In addition, two inverted sequence regions are flanked by
IS6 elements in AP024125.1 relative to NC_012690.1.
Downstream analysis of AMR positions in a dataset. As a sec-
ond downstream analysis, we can leverage the AMR gene location
data to identify trends in gene position among the plasmid dataset.
We analyzed the location of blaCMY-2 among our IncC plasmids. Of
the 117 plasmids that encoded for blaCMY-2, 93 plasmids bore the
gene downstream of IncC fragment 8 and upstream of IncC frag-
ment 6. All but 2 of the blaCMY-2 genes in this subset were located in
a range that peaked at 28 kb upstream of IncC fragment 6 (Fig. 8).
The blaCMY-2 genes were located in two ranges downstream of
IncC fragment 8. One range peaked at 30 kb downstream of frag-
ment 8 and the other at 80 kb. Two blaCMY-2 genes were found
outside of these ranges: 1 was identified 242,666 bp downstream of
fragment 8 in plasmid NZ_CP028804.1 and the other 190,156 bp
downstream of fragment 8 in plasmid NZ_CP019001.1. The shift
in the location of blaCMY-2 in both cases was associated with
a potential insertion event upstream of the gene. Upstream of
blaCMY-2 in NZ_CP028804.1 is a region that contains genes asso-
ciated with resistance to silver, copper and arsenic as well as
heat shock tolerance as well as the genetic markers for the plasmid
replicons
IncFIC(FII)_1(AP001918)
and IncFII_1(AY458016). Similarly, upstream of blaCMY-2 in
NZ_CP028804.1 is a region encoding for the iucA, iucB, iucC, iucD,
iutA virulence genes and the genetic markers for plasmid replicons
IncFIB(K)_1(JN233704) and IncFII(K)_1(CP000648). The pre-
sence of multiple plasmid replicons combined with the relative
position of blaCMY-2 from the IncC plasmid fragments indicates
these two plasmids are the result of a recombination event between
an IncC plasmid and a plasmid of the IncF family of plasmid
groups. The Lociq typing method records the gene position data for
all AMR and user-defined accessory genes and as a result, this gene
location analysis can be performed for any gene represented in the
plasmid dataset.
IncFIA_1(AP001918),
Lociq typing of draft assemblies. Draft plasmid assemblies can
be analyzed by using the allele definitions of the Lociq results. The
Lociq program cannot analyze draft assemblies for structural
variations of loci or fragment order, but it can perform plasmid
Fig. 8 Location of blaCMY-2 relative to IncC plasmid fragments. A density
plot of the position of blaCMY-2 upstream of IncC plasmid fragment 6 and
downstream of IncC plasmid fragment 8. The two small peaks near
190,000 bp and 240,000 bp correspond to plasmids NZ_CP019001.1 and
NZ_CP028804.1, respectively. Source data are available in Supplementary
Data 8.
Downstream analyses of hybrid plasmids and custom annota-
tions. Typing plasmids using the Lociq method allows us to
standardize the language surrounding plasmid feature diversity.
Here, we present four demonstrations of how the Lociq results
can be applied to downstream analyses. First, because the plasmid
typing process indexes the plasmids to a common starting point,
the data are organized to facilitate the incorporation of custom
feature annotations. This can aid the identification of hybrid
plasmids containing elements from multiple plasmid types, such
as the hybrid plasmid NZ_CP028197.1 that contains both IncC
and IncHI2A elements (Supplementary Fig. 7). The Lociq method
can also be used to identify custom features such as IS elements to
indicate potential sites of plasmid recombination. A comparison
8
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-04981-1
ARTICLE
MLST to identify which plasmids in the Lociq results most closely
match the draft assembly. To demonstrate this plasmid typing
function, we queried the NCBI database for draft assemblies
containing IncC plasmid sequence and filtered the results for
assemblies generated from short reads. From this, we selected the
whole genome shotgun sequencing record for Klebsiella pneu-
moniae K184 (JAANYS000000000) that contained 1,477 contigs.
A BLAST query of the IncC typing loci against the 6.7 Mb draft
assembly identified 62 IncC typing loci in the sequence. Thirty
five of the 62 loci were partial matches that either occurred at the
end of a contig (Supplementary Data 9) or aligned to com-
(Supplementary
plementary ends of
Data 10). Of the remaining 27 loci, 22 matched known alleles in
the Lociq results (Supplementary Data 11). Our analysis revealed
that this grouping of 22 alleles was conserved among 86 plasmids
in our dataset of 534 IncC plasmids. This subset of 86 IncC
plasmids represented the closest matches to the plasmid identified
in the whole genome shotgun sequencing assembly based on our
typing method.
two separate
contigs
Analysis of plasmids in a clinical setting. The final demon-
stration shows how subtyping with the Lociq method can aid in
tracking the evolution of a plasmid in a clinical setting. To do this,
we used the Lociq method to visualize the results of a study in a
major hospital in Taiwan that tracked the transmission of blaOXA-
48 from a plasmid to a K. pneumoniae chromosome over a three-
year period26. During this time, an accessory IncC plasmid that
was retained among the K. pneumoniae strains had lost ~20 kb
of sequence containing 9 AMR genes. The study closed the
sequences of 4 IncC plasmids that were recovered from isolates in
the blood of a patient suffering from bacteremia, urine of two
patients suffering from urinary tract infections and pus from a
patient suffering from pneumonia. Analysis of the 4 IncC plas-
mids revealed that all belonged to the IncC Lociq sequence type
74 (IncC Lociq ST74) and the loci and fragment patterns were
identical among all four plasmids (Fig. 9). However, the inter-
fragment distances and arrangement of AMR genes among the
plasmids differed, indicating that each of the plasmids that was
recovered represented a different stage in the evolution of the
plasmid at the hospital. The primary study indicated that the first
stage of plasmid evolution was observed between the plasmids
NZ_CP040034.1 and NZ_CP040029.1 that were isolated in the
first year of
the sample period. These plasmids showed an
inversion of a ~ 20 kb resistance cassette containing erm(42)-
blaTEM-31-rmtb1-tet(G)-floR2-sul1-qacEdelta1-aadA2-dfrA12 that
was located between IncC fragments 4 & 2. The next step in
plasmid evolution indicated in the primary study was observed in
plasmids recovered later in the sampling period. These plasmids
showed a reduction in size due to the loss of an overlapping
resistance cassette containing aac(3)-IId-dfrA12-aadA2-qacE-
delta1-sul1-floR2-tet(G)-rmtb1-blaTEM-31 but leaving erm(42) in
the plasmid. The proposed final step was the loss of blaCTX-M-14
that was embedded between two sections of IncC fragment 3.
This quick analysis revealed that even though the plasmids were
identical in sequence type, loci pattern and fragment pattern, the
difference in interfragment distance showed that the plasmids
were not identical. Further, the fragments of the Lociq typing
method provided common reference point among the plasmids to
identify where each plasmid restructuring event had taken place.
Next, we compared the four IncC Lociq ST74 plasmids recovered
from K. pneumoniae isolates in a Taiwanese hospital to the only five
IncC Lociq ST75 plasmids in our results. These two plasmid
sequence types differ by a single allele that encodes for an
uncharacterized protein. Even though the four ST74 plasmids were
all recovered from a single location and single species, the five ST75
plasmids were recovered from multiple species and multiple sites.
The smaller two IncC Lociq ST74 plasmids shared the same loci
and fragment pattern with the IncC Lociq ST75 plasmids
NZ_LT985224.1 and NZ_MF150121.1, however the ST75 plasmids
were recovered from E. coli in France and K. pneumoniae in Brazil,
respectively (Supplementary Fig. 8). Alignment of the plasmids
revealed 98% coverage and > 99% identity between NZ_CP040024.1
and NZ_LT985224.1 and 97% coverage and > 99% identity between
NZ_MF150121.1 and NZ_CP040039.1. The third IncC Lociq
ST74 plasmid NZ_CP040029.1 shared the same plasmid structure
the IncC Lociq ST75 plasmids
and AMR composition of
NZ_MF150118.1 and NZ_CP028996.1, but the IncC Lociq ST75
plasmids were recovered from P. mirabilis in Brazil and K.
pneumoniae in USA (Supplementary Fig. 9). Both NZ_MF150118.1
and NZ_CP028996.1 aligned to the ST74 NZ_CP040029.1 with
100% coverage and >99% identity. Finally, the fourth IncC Lociq
Fig. 9 Alignment of IncC sequence type 74 plasmids. Visual comparison of all IncC Lociq ST74 plasmids in our results. All plasmids were recovered from
a single hospital in Taiwan between 2013 and 2015 and illustrate how a plasmid that is established in a single location can change over time. Shaded
regions indicate differences between plasmids. The numbered black bars represent plasmid fragments, red bars represent AMR genes and yellow bars
represent stress-tolerance genes. Strand orientation is in relation to the plasmid indexing locus and forward orientation is represented by gene presence
above the sequence line.
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
9
ARTICLE
COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-04981-1
ST74 plasmid NZ_CP040034.1 shared the same inverted sequence
upstream of IncC plasmid fragment 2 that was observed in the IncC
Lociq ST75 plasmids NZ_CP023724.1 and NZ_AP018672.1. These
last two ST75 plasmids were recovered from a hypervirulent K.
pneumoniae clinical
isolate in Taiwan and a K. pneumoniae
environmental isolate in Japan (Supplementary Fig. 10)27. Both
ST75 plasmids shared 96% coverage and 99% identity with the ST74
plasmid, and the decreased coverage was affected by the partial loss
of a resistance cassette between IncC plasmid fragment 4 and 2.
This final application of the Lociq method demonstrates how
one plasmid type that was recovered solely from K. pneumoniae
isolates from a major hospital in Taiwan was genetically similar to
plasmids isolated from K. pneumoniae, E. coli and P. mirabilis in
4 different continents. This demonstration indicates that with the
appropriate supporting epidemiological data, the results of the
Lociq method can be used to support efforts to track the spread of
clinically relevant plasmids.
Discussion
We have demonstrated how the Lociq method uses closed plas-
mid assemblies to identify core genetic elements and structural
patterns conserved among IncC plasmids. This method can be
applied to a single dataset to identify typing metrics for any other
plasmid group that share a core set of loci. Further, because this
method can characterize plasmids that do not contain the full
complement of typing loci, it is ideal for characterizing plasmids
that contain elements from multiple plasmid types. This feature
also allows for increased characterization of plasmids from draft
assemblies where not all of the plasmid typing loci are repre-
sented in the assembled sequence. The Lociq method provides a
common language to describe plasmid diversity using the end-
points of fragment pattern, loci pattern, plasmid sequence type
and interfragment distances. These properties make the Lociq
method a powerful tool to track and study the evolution and
routes of transmission of any plasmid of interest.
The Lociq method generates multiple typing schema, each with
a different discriminatory power. Typing schema with a low
discriminatory power, such as the Lociq fragment type, are suited
to identify larger groups of similar plasmids. The schema that
accounts for all metrics of the Lociq method had the greatest
discriminatory power of all the evaluated methods and is best
suited to differentiate between similar plasmids. The comparator
method whose metrics generated greatest discriminatory power
was
the IncA/C
cgPMLST schema was only designed to differentiate between
IncA/C plasmids while the Lociq method can theoretically be
applied to characterize any plasmid type that shares a common
set of core loci. Further, the cgPMLST was developed through
resource intensive transposon disruption assays while the bioin-
formatic Lociq subtyping method can be run on a desktop
computer3.
the IncA/C cgPMLST method. However,
The Lociq method adds two features that are not common in
other typing methods. First, this method identifies conserved
intergenic regions and codifies them as typing loci. This has the
dual benefit of not only increasing the pool of plasmid loci, but
also facilitating the construction of larger contiguous regions of
neighboring typing loci. The second feature the Lociq method
adds is an analysis of variations in the plasmid structure.
Structural analysis of the arrangement of elements is relevant to
plasmid typing because it can identify common recombination
events in plasmids such as deletions, insertions, duplications or
rearrangement events. The structural analysis also accounts for
differences in the length of sequence between the plasmid
fragments. Variations in interfragment distances can notify
researchers not only that a recombination event occurred, but
also the region of the plasmid where the recombination event
took place.
While the Lociq method increases the discriminatory power of
plasmid subtyping through the addition of structural compar-
isons, the method does have limitations. First, the user needs to
have access to a library of high-quality closed plasmid assemblies
to construct their initial dataset. Second, the plasmid dataset
should contain sufficient genetic diversity to represent the plas-
mid type of interest. The Lociq method also requires the user to
input threshold values for loci selection and interfragment dis-
tance limits. The method provides graphics to help inform the
user of plasmid loci distribution, but no equation to determine
the optimum cutoff value for prevalence within a plasmid type is
supplied with the method. Finally, even though the Lociq results
demonstrated greater discriminatory power than other typing
schema, increased discriminatory power is not always ideal when
the objective is to identify similar members of a group. For-
tunately, the Lociq method generates multiple outputs that allow
the user to select the testing metric that is appropriate for their
the IncC-specific typing
purposes. Due to these limitations,
definitions that we obtained from our sample set of foodborne
pathogens are not intended to classify the full diversity of the
extant IncC population. Rather, developing the plasmid typing
definitions
the diversity of plasmid
sequence and structure will require collaboration with a number
of partners that represent a diverse set of isolation locations,
biological compartments, host organisms and isolation dates.
that accurately reflect
In addition to the applications demonstrated earlier, this typing
method has promising implications for plasmid research. First,
the Lociq method can be used to characterize plasmids that are
currently untyped. The initial stage of the Lociq program orga-
nizes plasmids independent of plasmid type through hierarchical
clustering of loci presence/absence data. Plasmids belonging to
clusters without a known plasmid type can be characterized by
the Lociq method using the typing loci unique to that cluster.
Second, the Lociq method can facilitate analyses between plasmid
sequence and plasmid metadata. These comparisons may be
made either by evaluating the sequence composition of the
plasmid typing alleles, or by evaluating alleles present in sub-
clusters of a plasmid type as was seen in the clustering of the IncC
plasmids (Fig. 1). Finally, the library of typing loci may help to
reconcile draft plasmid assemblies by providing a template for
contig extension and gap closure when partial matches of plasmid
typing loci map to the end of draft assembly contigs.
The Lociq method combines structural and sequence variants
to increase the discriminatory power of existing plasmid typing
methods. By reducing plasmids to their component parts, the
Lociq method standardizes comparison metrics among plasmid
types and allows for enhanced investigations between plasmid loci
and plasmid metadata such as AMR gene composition, isolation
source or plasmid lineage. The results of the Lociq method will
not only benefit basic plasmid biology research they will also aid
public health monitoring programs such as NARMS to track the
spread of plasmid lineages and better identify the origin of multi-
drug resistant plasmids.
Methods
Sequences and core annotations. The initial dataset of long read sequences from
175 Salmonella and E. coli retail meat and cecal sample NARMS isolates were
generated using PacBio Sequel platform with sequencing kit v3.0 (Pacific Bios-
ciences, Menlo Park, CA). Sequencing libraries were prepared with the PacBio
SMRTbell template prep kit v1.0 and the resulting reads were assembled into closed
contigs using the PacBio Hierarchical Genome Assembly Process 4.0 and Circlator
v1.5.528,29. Plasmid Inc type was determined using PlasmidFinder definitions
(accessed 4-27-2022) and closed plasmid assemblies were annotated with PROKKA
v1.14.530,31. The reference database of plasmid sequences evaluated was the PLSDB
database v. 2021_06_23_v224.
10
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-04981-1
ARTICLE
Lociq method. The scripts for operation of the Lociq method are available for
download at http://www.github.com/LBHarrison/Lociq/. Required input for the
method includes annotation files of closed plasmid assemblies, access to reference
database and plasmid type metadata. Additionally, the program requires user
defined thresholds for prevalence and distance to account for variability in diversity
among different plasmid groups and comparator schema.
Identification of plasmid typing Loci. The pangenome and intergenic regions of
the closed plasmid dataset were obtained using Roary and piggy, respectively23,22.
Data for the coding and intergenic pangenomes were merged and passed to R for
clustering as binary data with complete linkage32. Plasmid typing loci among the Inc
groups was determined in a two-stage process (Fig. 1). First, putative plasmid typing
loci were identified by selecting the loci with a user-defined threshold of high pre-
valence in the plasmid group of interest and a user-defined threshold of low pre-
valence in the other plasmid groups. Second, loci were queried against an external
plasmid database as a validation step using an 80% identity threshold. Loci that met
or exceeded user-defined prevalence thresholds for membership within a plasmid
group were identified as the plasmid typing loci for the current plasmid group.
Identification of conserved plasmid fragments. Sequence coordinates of typing
loci were obtained through a BLAST query of the loci against an external plasmid
database. Clusters of loci separated by less than a user-defined threshold value
defined the contiguous sequence regions of a plasmid. These data were used to
generate a contingency table displaying an all vs all tally of loci occurring in the
same contiguous sequence region. The contingency table was analyzed as a cor-
relation matrix evaluating the Pearson’s correlation coefficient (R) for all loci
interactions using the R Hmisc v4.7-0 package33. Loci clusters with a mean cor-
relation coefficient ≥ 0.9 represent conserved contiguous sequence regions in the
plasmid dataset and are referred to as plasmid fragment. Loci clusters with a mean
R-value < 0.9 were subjected to increasingly stringent clustering parameters until
the resulting plasmid fragments had a mean R-value≥ 0.9.
Plasmid subtyping. Plasmid sequences were indexed to begin at the typing locus
present in the greatest number of plasmids and the sequences were analyzed with
AMRFinder plus to identify AMR genes and stress tolerance genes34. Plasmids were
then subtyped using the metrics of: sequence type, organization of loci, organization
of plasmid fragments and the distances between the plasmid fragments (Fig. 4).
Unique numeric identifiers of the typing metrics are generated as part of the sum-
mary file output from the Lociq program (Supplementary Data 4). Plasmid sequence
type was determined by the allelic composition of plasmid typing loci. Loci position
data were extracted from the BLAST results and a unique numeric identifier was
assigned to each unique organization of loci among the plasmid typing fragments. A
similar process was applied to the order of plasmid fragments to determine the
plasmid fragment type. Finally, the distances between each plasmid fragment were
recorded to identify each plasmid’s set of interfragment distance values.
Comparator typing methods. Comparator typing schema were used to evaluate the
discriminatory power of the Lociq method. PTU designation and MOB type were
determined using COPLA (updated 6-30-2021 using the RS84 definitions)19. IncC
plasmids were further characterized by the IncA/C PMLST and IncA/C cgPMLST
allelic profiles as recorded in PubMLST (Accessed 8-18-2022)3,35. Discriminatory
power of the typing schema was determined by Simpson’s diversity index.
Downstream analyses. Downstream analyses were performed to demonstrate four
additional applications of Lociq method. In the first demonstration of custom
annotations, insertion sequence (IS) elements were identified in the dataset using
ISEscan v1.7.23 and the results merged with the Lociq annotation file36. Sequence
alignments were generated with NCBI BLAST and visualizations were generated using
the R genoPlotR package37. In the second demonstration that identified trends in the
position of AMR genes in the dataset, the distance of an AMR gene of interest to its
nearest plasmid fragments were visualized on a density plot in the base R package.
Third, the Lociq method was used to improve characterization of plasmid draft
assemblies. This was done by performing a BLAST query of typing loci against the draft
plasmid assembly to identify loci present in the sequence. The results were filtered by
requiring an identity > 70% and coverage >90%. The draft assembly loci sequences
were compared to the reference plasmid loci sequences to determine which specific
plasmid typing alleles were present in the draft assembly. Alleles present in the draft
plasmid assembly were used to construct an m x n presence/absence matrix where m
was equal the number of plasmids that were analyzed with the Lociq method + the
plasmid draft assembly and n was equal to number of unique alleles in the plasmid
draft assembly. The presence/absence matrix was used to create a distance matrix of
plasmids using the dist function in R with the method parameter set to binary. The row
corresponding to the plasmid draft assembly was extracted and the distance values
were evaluated to identify the least dissimilar plasmids from the Lociq results.
two plasmid typing loci occurring on the same region of the plasmid. Source data
and numeric results are available in Supplementary Data 3. The sample size for
these tests accounted for the 72 plasmid typing loci that were identified in the IncC
plasmid demonstration dataset.
Reporting summary. Further information on research design is available in the Nature
Portfolio Reporting Summary linked to this article.
Data availability
Plasmid sequences are available through the PLSDB database (https://ccb-microbe.cs.
uni-saarland.de/plsdb/) while plasmid metadata files are hosted in a Github repository
(https://github.com/LBHarrison/Lociq/). Source data for Figs. 3, 6 and 8 are provided in
Supplementary Data 3, 7 and 8, respectively.
Code availability
The Lociq program is available through the Github repository https://github.com/
LBHarrison/Lociq/38.
Received: 5 January 2023; Accepted: 25 May 2023;
References
1. Rozwandowicz, M. et al. Plasmids carrying antimicrobial resistance genes in
Enterobacteriaceae. J. Antimicrob. Chemother. 73, 1121–1137 (2018).
2. Carattoli, A. Plasmid-mediated antimicrobial resistance in Salmonella
enterica. Curr. Issues Mol. Biol. 5, 113–122 (2003).
3. Hancock, S. J. et al. Identification of IncA/C plasmid replication and
maintenance genes and development of a plasmid multilocus sequence typing
scheme. Antimicrob. agents Chemother. 61, e01740–01716 (2017).
4. Villa, L., García-Fernández, A., Fortini, D. & Carattoli, A. Replicon sequence
typing of IncF plasmids carrying virulence and resistance determinants. J.
Antimicrob. Chemother. 65, 2518–2529 (2010).
5. García-Fernández, A. & Carattoli, A. Plasmid double locus sequence typing for
IncHI2 plasmids, a subtyping scheme for the characterization of IncHI2
plasmids carrying extended-spectrum β-lactamase and quinolone resistance
genes. J. Antimicrob. Chemother. 65, 1155–1161 (2010).
Phan, M.-D. et al. Variation in Salmonella enterica serovar typhi IncHI1
plasmids during the global spread of resistant typhoid fever. Antimicrob.
Agents Chemother. 53, 716–727 (2009).
6.
7. García-Fernández, A. et al. Multilocus sequence typing of IncI1 plasmids carrying
extended-spectrum β-lactamases in Escherichia coli and Salmonella of human
and animal origin. J. Antimicrob. Chemother. 61, 1229–1233 (2008).
8. García-Fernández, A. et al. Multilocus sequence typing of IncN plasmids. J.
Antimicrob. Chemother. 66, 1987–1991 (2011).
9. Novick, R. P. Plasmid incompatibility. Microbiol. Rev. 51, 381–395 (1987).
10. Datta, N. & Hedges, R. Compatibility groups among fi− R factors. Nature 234,
222–223 (1971).
11. Carattoli, A. et al. Identification of plasmids by PCR-based replicon typing. J.
Microbiol. methods 63, 219–228 (2005).
12. Williams, L. E., Wireman, J., Hilliard, V. C. & Summers, A. O. Large plasmids
of Escherichia coli and Salmonella encode highly diverse arrays of accessory
genes on common replicon families. Plasmid 69, 36–48 (2013).
13. Du, P. et al. Novel IS26-mediated hybrid plasmid harbouring tet (X4) in
Escherichia coli. J. Glob. Antimicrob. resist. 21, 162–168 (2020).
14. Sun, R.-Y. et al. Global clonal spread of mcr-3-carrying MDR ST34 Salmonella
enterica serotype Typhimurium and monophasic 1, 4,[5], 12: i:− variants
from clinical isolates. J. Antimicrob. Chemother. 75, 1756–1765 (2020).
15. Robertson, J. & Nash, J. H. MOB-suite: Software tools for clustering, reconstruction
and typing of plasmids from draft assemblies. Microb. genomics 4, e000206 (2018).
16. Alvarado, A., Garcillán-Barcia, M. P. & de la Cruz, F. A degenerate primer
MOB typing (DPMT) method to classify gamma-proteobacterial plasmids in
clinical and environmental settings. PLoS One 7, e40438 (2012).
17. Chen, I., Christie, P. J. & Dubnau, D. The ins and outs of DNA transfer in
bacteria. Science 310, 1456–1460 (2005).
18. Redondo-Salvo, S. et al. Pathways for horizontal gene transfer in bacteria
revealed by a global map of their plasmids. Nat. Commun. 11, 1–13 (2020).
19. Redondo-Salvo, S. et al. COPLA, a taxonomic classifier of plasmids. BMC
Bioinforma. 22, 1–9 (2021).
Statistics and reproducibility. Statistical tests were performed in R with the
Hmsic package32,33. Specifically, the Pearson’s correlation coefficient and corre-
sponding probabilities were calculated to evaluate the pairwise likelihood of any
20. Qi, Q., Kamruzzaman, M. & Iredell, J. R. The higBA-type toxin-antitoxin
system in IncC plasmids is a mobilizable ciprofloxacin-inducible system.
Msphere 6, e00424–00421 (2021).
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
11
ARTICLE
COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-04981-1
21. Arredondo-Alonso, S., Willems, R. J., Van Schaik, W. & Schürch, A. C. On the
(im) possibility of reconstructing plasmids from whole-genome short-read
sequencing data. Microb. genomics 3, e000128 (2017).
22. Page, A. J. et al. Roary: Rapid large-scale prokaryote pan genome analysis.
Bioinformatics 31, 3691–3693 (2015).
23. Thorpe, H. A., Bayliss, S. C., Sheppard, S. K. & Feil, E. J. Piggy: A rapid, large-
scale pan-genome analysis tool for intergenic regions in bacteria. Gigascience
7, giy015 (2018).
24. Galata, V., Fehlmann, T., Backes, C. & Keller, A. PLSDB: a resource of
complete bacterial plasmids. Nucleic acids Res. 47, D195–D202 (2019).
25. Chang, W., Cheng, J., Allaire, J. J., Xie, Y. & McPherson, J. (2020).
26. Lu, M.-C. et al. Transmission and evolution of OXA-48-producing Klebsiella
pneumoniae ST11 in a single hospital in Taiwan. J. Antimicrob. Chemother.
75, 318–326 (2020).
27. Huang, Y.-H. et al. Emergence of an XDR and carbapenemase-producing
hypervirulent Klebsiella pneumoniae strain in Taiwan. J. Antimicrob.
Chemother. 73, 2039–2046 (2018).
28. Hunt, M. et al. Circlator: Automated circularization of genome assemblies
using long sequencing reads. Genome Biol. 16, 1–10 (2015).
29. Chin, C.-S. et al. Nonhybrid, finished microbial genome assemblies from long-
read SMRT sequencing data. Nat. Methods 10, 563–569 (2013).
30. Seemann, T. Prokka: rapid prokaryotic genome annotation. Bioinformatics 30,
2068–2069 (2014).
31. Carattoli, A. et al. PlasmidFinder and pMLST: in silico detection and typing of
plasmids. Antimicrob. Agents Chemother. (2014).
32. Team, R. C. R.: A language and environment for statistical computing. R.
Found. Stat. Comput., Vienna, Austria. https://www.R-project.org/ (2021).
33. Harrell, F. E. & Dupont, C. Hmisc: Harrell miscellaneous. R package version
4.1-1. R Found. Stat. Comput. https://CRAN.R-project.org/package=Hmisc
(accessed 16 Feb. 2018) (2018).
34. Feldgarden, M. et al. Validating the AMRFinder tool and resistance gene
35.
database by using antimicrobial resistance genotype-phenotype correlations in a
collection of isolates. Antimicrob. Agents Chemother. 63, e00483–00419 (2019).
Jolley, K. A., Bray, J. E. & Maiden, M. C. Open-access bacterial population
genomics: BIGSdb software, the PubMLST. org website and their applications.
Wellcome Open Res. 3, 124 (2018).
36. Xie, Z. & Tang, H. ISEScan: automated identification of insertion
sequence elements in prokaryotic genomes. Bioinformatics 33, 3340–3347 (2017).
37. Guy, L., Roat Kultima, J. & Andersson, S. G. genoPlotR: Comparative gene
and genome visualization in R. Bioinformatics 26, 2334–2335 (2010).
38. Harrison, L. LBHarrison./Lociq: Lociq v1.0.0 (v1.0.0) v. 05/16/2023
(Zenodo, 2023). https://doi.org/10.5281/zenodo.7942565
may not reflect the official policy of the FDA, the Department of Health and Human
Services, or the U.S. Government. Reference to any commercial materials, equipment,
or process does not in any way constitute approval, endorsement, or recommendation
by the FDA.
Author contributions
C.L., E.S., G.H.T., L.H., P.F.M., and S.Z. contributed to data analysis & interpretation as
well as manuscript revision. G.H.T., L.H. and S.Z. were responsible for conception and
design of the project while SZ was responsible for project oversight. L.H. was responsible
for the figure and software creation as well as the draft manuscript. C.L. was responsible
for data acquisition. Funding was obtained by P.F.M. and S.Z.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information The online version contains supplementary material
available at https://doi.org/10.1038/s42003-023-04981-1.
Correspondence and requests for materials should be addressed to Lucas Harrison.
Peer review information Communications Biology thanks Michael Feldgarden, Joseph
Nesme and the other, anonymous, reviewer(s) for their contribution to the peer review of
this work. Primary Handling Editor: George Inglis. A peer review file is available.
Reprints and permission information is available at http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
material in this article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not included in the
article’s Creative Commons license and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.
Acknowledgements
This work was supported through the U.S. FDA National Antimicrobial Resistance
Monitoring System. The views expressed in this paper are those of the authors and
This is a U.S. Government work and not under copyright protection in the US; foreign
copyright protection may apply 2023
12
COMMUNICATIONS BIOLOGY |
(2023) 6:595 | https://doi.org/10.1038/s42003-023-04981-1 | www.nature.com/commsbio
| null |
10.1038_s41588-023-01451-6.pdf
|
Data availability
Public and private data can be accessed through their respective por-
tals. Private data will require prior authorization. Data can be cleaned
and normalized using any standard or well-established procedure
for variant analysis or the procedures described in this paper, includ-
ing referenced papers or procedures. The integrated, curated and
de-duplicated data (to the best of our ability) are available in Sup-
plementary Table 1. No additional data or intermediate results will be
available upon request given the high manual burden to verify access
to a variety of private portals, repositories and patients.
Code availability
Variants were processed using well-established procedures described
in the referenced papers. Datasets from diverse sources were inte-
grated using a combination of code (to automate certain steps) and
manual curation. Thus, the standalone code is not sufficient to regen-
erate the integrated dataset. Therefore, this code and intermediate
results from dataset integration and curation is not available upon
request. The code used for analysis and to generate figures is avail-
able under Creative Commons license through Zenodo at https://doi.
org/10.5281/zenodo.8008632. Analyses were executed in Python (v3.7),
R (v4.1.1), GraphPad Prism (v92.2), matplotlib(v3.3.1), circos (v0.69-9)
and seaborn (v0.11.1).
PyMOL v2.4.0 was used to visualize structures. The Consurf online
server was used for conservation analysis. Geneious Prime v2021.2.2
was used for multiple sequence alignmentss. The PolyPhen2 online
server using the HumVar model was used to predict the severity/patho-
genicity of the compiled NDD mutations. Unless otherwise noted,
mutational counts, bar plots, pie charts, and Venn diagrams throughout
were made using a combination of Python (v3.7), R (v4.1.1), GraphPad
Prism (v92.2), matplotlib(v3.3.1) and seaborn (v0.11.1). The lollipop
portion of the 2D schematics were created using the St. Jude PeCan
Protein Paint software. Missense substitutions were visualized as a
Sankey diagram using Google Charts. The Circos plot was made using
the Circos software (v0.69-9). Missense substitutions were visualized
as a Sankey diagram using Google Charts. The code used to process
and visualize the data are available under the MIT license at Zenodo at
|
Data availability Public and private data can be accessed through their respective portals. Private data will require prior authorization. Data can be cleaned and normalized using any standard or well-established procedure for variant analysis or the procedures described in this paper, including referenced papers or procedures. The integrated, curated and de-duplicated data (to the best of our ability) are available in Supplementary Table 1 . No additional data or intermediate results will be available upon request given the high manual burden to verify access to a variety of private portals, repositories and patients. Code availability Variants were processed using well-established procedures described in the referenced papers. Datasets from diverse sources were integrated using a combination of code (to automate certain steps) and manual curation. Thus, the standalone code is not sufficient to regenerate the integrated dataset. Therefore, this code and intermediate results from dataset integration and curation is not available upon request. The code used for analysis and to generate figures is available under Creative Commons license through Zenodo at https://doi. org/10.5281/zenodo.8008632 . Analyses were executed in Python (v3.7), R (v4.1.1), GraphPad Prism (v92.2), matplotlib(v3.3.1), circos (v0.69-9) and seaborn (v0.11.1). PyMOL v2.4.0 was used to visualize structures. The Consurf online server was used for conservation analysis. Geneious Prime v2021.2.2 was used for multiple sequence alignmentss. The PolyPhen2 online server using the HumVar model was used to predict the severity/pathogenicity of the compiled NDD mutations. Unless otherwise noted, mutational counts, bar plots, pie charts, and Venn diagrams throughout were made using a combination of Python (v3.7), R (v4.1.1), GraphPad Prism (v92.2), matplotlib(v3.3.1) and seaborn (v0.11.1). The lollipop portion of the 2D schematics were created using the St. Jude PeCan Protein Paint software. Missense substitutions were visualized as a Sankey diagram using Google Charts. The Circos plot was made using the Circos software (v0. 69-9). Missense substitutions were visualized as a Sankey diagram using Google Charts. The code used to process and visualize the data are available under the MIT license at Zenodo at https://doi.org/10.5281/zenodo.8008632 .
|
Landscape of mSWI/SNF chromatin
remodeling complex perturbations
in neurodevelopmental disorders
https://doi.org/10.1038/s41588-023-01451-6
Received: 4 October 2022
Accepted: 20 June 2023
Published online: 27 July 2023
Check for updates
Alfredo M. Valencia1,2,3,10,11,12, Akshay Sankar1,3,12, Pleuntje J. van der Sluijs
F. Kyle Satterstrom
Samantha A. Schrier Vergano
5,6, Jack Fu6, Michael E. Talkowski
7,8, Gijs W. E. Santen
4 & Cigall Kadoch
5,6,
4,
1,3,9
DNA sequencing-based studies of neurodevelopmental disorders
(NDDs) have identified a wide range of genetic determinants. However, a
comprehensive analysis of these data, in aggregate, has not to date been
performed. Here, we find that genes encoding the mammalian
SWI/SNF (mSWI/SNF or BAF) family of ATP-dependent chromatin
remodeling protein complexes harbor the greatest number of de novo
missense and protein-truncating variants among nuclear protein
complexes. Non-truncating NDD-associated protein variants predominantly
disrupt the cBAF subcomplex and cluster in four key structural regions
associated with high disease severity, including mSWI/SNF-nucleosome
interfaces, the ATPase-core ARID-armadillo repeat (ARM) module insertion
site, the Arp module and DNA-binding domains. Although over 70% of the
residues perturbed in NDDs overlap with those mutated in cancer, ~60% of
amino acid changes are NDD-specific. These findings provide a foundation
to functionally group variants and link complex aberrancies to phenotypic
severity, serving as a resource for the chromatin, clinical genetics and
neurodevelopment communities.
Sequencing studies have revealed extensive involvement of chroma-
tin regulatory processes in a range of human diseases, with frequent
mutations in the genes encoding proteins that govern chromatin archi-
tecture1–4. Four families of multi-subunit ATP-dependent chromatin
remodeling complexes (SWI/SNF, ISWI, CHD and INO80) modulate
chromatin topology and gene expression by mobilizing their nucleo-
some substrates5. Recent advances in cryo-electron microscopy
(cryo-EM), cross-linking mass spectrometry and homology modeling
have begun to uncover the three-dimensional (3D) structure and modes
of nucleosome substrate engagement of these large heterogeneous
entities, informing mechanistic studies6.
Mutations in the genes encoding mammalian SWI/SNF (mSWI/SNF)
chromatin remodeling complex are found in over 20% of cases in can-
cer, which has stimulated a range of basic and translational efforts over
the past several years7–9. A wealth of mutational data of neurodevel-
opmental disorders (NDDs), such as intellectual disability and autism
1Dana-Farber Cancer Institute and Harvard Medical School, Boston, MA, USA. 2Chemical Biology Program, Harvard University, Cambridge, MA, USA.
3Broad Institute of MIT and Harvard, Cambridge, MA, USA. 4Department of Clinical Genetics, Leiden University Medical Center, Leiden, the Netherlands.
5Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard, Cambridge, MA, USA. 6Massachusetts General Hospital, Boston, MA, USA.
7Children’s Hospital of the King’s Daughters, Norfolk, Virginia, USA. 8Department of Pediatrics, Eastern Virginia Medical School, Norfolk, Virginia, USA.
9Howard Hughes Medical Institute, Chevy Chase, MD, USA. 10Present address: Department of Psychiatry and Behavioral Sciences, Stanford University,
Stanford, CA, USA. 11Present address: Stanford Brain Organogenesis, Wu Tsai Neurosciences Institute, Stanford University, Stanford, CA, USA.
12These authors contributed equally: Alfredo M. Valencia, Akshay Sankar.
e-mail: [email protected]
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1400
nature geneticsAnalysis spectrum disorders, has also recently emphasized a high mutational
burden of chromatin regulatory genes in NDD, presenting an oppor-
tunity to dissect the molecular underpinnings and to inform poten-
tial strategies to remedy the comorbid issues associated with these
disorders2,10–14.
Most cancer-associated mSWI/SNF mutations result in subunit
deletions or gene silencing, which has presented the field with opportu-
nities to understand the impact of full subunit losses and the impact on
complex disassembly15–18. NDD-associated mSWI/SNF genetic variants
present particularly unique opportunities for functional dissection, in
that 1) mutations are often missense, affecting single amino acids and
clustering i n defined domains within subunits; 2) mutations are p r e d -
o m i na ntly h e t e r o z yg ous, u n d er s c oring t h e h i g h d e g ree o f d o s age sen-
sitivity; and 3) mutations are often found as the sole genetic cause of
these disorders. Furthermore, for trios in which parents’ genetic infor-
mation is available, mSWI/SNF gene variants are predominantly de novo
(absent in parents), indicating their causative role19–21. Together, these
features enable functional assignment and prioritization for specific
subunit domains and even individual protein residues. Identifying
and mechanistically defining these variants will be critical for the
assignment of specific chromatin remodeling complex functions and,
ultimately, informing therapeutic approaches for a range of human
diseases driven by mSWI/SNF complex disruption.
Here, we sought to comprehensively catalog and integrate mSWI/
SNF complex sequence variants across a diverse collection of datasets,
including the Simon’s Foundation Research Initiative (SFARI) (Simons
Foundation Powering Autism Research for Knowledge (SPARK), Simons
Searchlight Collection–Autism Sequencing Consortium (SSC-ASC)),
the Deciphering Developmental Disorders project (DDD), the DECI-
PHER database22, ClinVar23, the Leiden Open Variation Database
(LOVD)24, de novo sequence variants from the literature (as performed
in McRae et al. (https://github.com/jeremymcrae/dnm_cohorts)3,25–39),
NDD-associated mSWI/SNF sequence variants from the litera-
ture3,19–21,35,40–81 and 85 previously unreported NDD-associated mSWI/SNF
cases, including 72 novel variants, focused on protein coding mutations
stemming from single-nucleotide variants (SNVs) and small insertions/
deletions (indels) (Supplementary Table 1). These analyses encompass
2,539 total cases of which the majority (67.1%, n = 1,703) result in mis-
sense and in-frame indels that collectively reveal 1,204 unique variants.
Results
Chromatin remodelers carry a high mutational burden in NDDs
Single amino acid mutations and protein-truncating variants (PTVs)
in chromatin regulatory genes are pathogenic for a variety of NDDs,
including syndromic and non-syndromic intellectual disabilities and
autism spectrum disorders3, but their relative prevalence remains
undefined. We collated and analyzed all SNVs and small indels reported
in DECIPHER (DatabasE of genomiC varIation and Phenotype in Humans
using Ensembl Resources)22 (https://www.deciphergenomics.org/),
a repository of clinical and genetic information on individuals with
developmental disorders. Remarkably, we found that epigenetic and
chromatin-related genes (EpiFactor gene list, Supplementary Table 2)82
were more frequently mutated than synapse-related genes (SynGO
gene list, Supplementary Table 2)83, which are known to be highly
implicated in NDDs (Extended Data Fig. 1a). By examining the top 50
Gene Ontology molecular functions (GOMFs) of genes in the Develop-
ment Disorder Genotype–Phenotype Database (DDG2P), we found that
top-ranked disrupted processes were enriched for transcription- and
chromatin-related processes, with transcription and chromatin binding
terms ranking highest among them (Fig. 1a and Extended Data Fig. 1b,c).
Performing this analysis with variants identified from the SFARI Autism
Spectrum Disorder (ASD) SPARK, SSC-ASC and developmental disorder
(DD) DDD study datasets (ASD + DD) revealed similar results, includ-
ing transcription-, synapse- and chromatin-related GOMFs (that is, 1:
transcriptional coregulator activity, 2: voltage-gated channel activity,
3: voltage-gated cation channel activity and 4: chromatin DNA binding)
(Extended Data Fig. 1a,c–e). We then analyzed de novo missense and
PTV frequencies from ASD + DD datasets by protein complex associa-
tions and by chromatin regulatory activity, which revealed the great-
est number of variants occurred in SWI/SNF chromatin remodeling
complex genes (protein complex, n = 404 sequence variants, rank 1),
followed by SET1 methyltransferase family (protein family, n = 346,
rank 2), lysine acetyltransferases (protein family, n = 300, rank 3) and
CHD chromatin remodeling complex genes (protein complex, n = 232,
rank 4) (Fig. 1b and Supplementary Table 2). This result was consistent
using DECIPHER data (Extended Data Fig. 1f) and chromatin-related
protein complexes from EpiFactor using ASD + DD data (Extended Data
Fig. 1g). Of note, several histone modifying complexes, including the
histone–lysine N-methyltransferase (KMT2 or MLL) family of com-
plexes, the histone acetyltransferase MOZ/MORF complexes and
Polycomb repressive deubiquitinase (PR-DUB) complexes had a greater
average of mutations when normalized by gene set size, owing to
lower numbers of defined components relative to mSWI/SNF com-
plexes (average ~6 components versus ~19 components for mSWI/
SNF) (Extended Data Fig. 1h and Supplementary Table 2). Neverthe-
less, when normalized by protein length (or gene exon length), cBAF
complexes maintained the highest average number of de novo muta-
tions and PTVs compared to all EpiFactor complexes (Extended Data
Fig. 1i). Interestingly, separating ASD and DD datasets revealed cBAF was
the most frequently mutated gene set in DD but ranked fourth in ASD,
potentially suggesting a subtle distinction between ASD-associated
variants from SFARI compared to a mixture of ASD and other NDDs
reported in the DDD database (Extended Data Fig. 1j).
Expanding our analysis to include copy-number variants in addi-
tion to SNVs/indels using DECIPHER, we found that genes encoding all
members of mammalian chromatin remodeling complexes (across all
families) are implicated in approximately one in ten of all DECIPHER
cases (9.34%, 5,196/55,645) (Fig. 1c,d and Extended Data Fig. 1k). The
29 genes encoding the mSWI/SNF complex are affected in the greatest
percentage (4.10%, 2,281/55,645), the majority of which are classified
as ‘pathogenic’ or ‘likely pathogenic’ (67.9%, 1,548/2,281), 39.2% of
which were confirmed de novo and 34.4 % of unknown inheritance
(Extended Data Fig. 1l). Many mSWI/SNF genes are also implicated in
ASD, as characterized by SFARI database (Fig. 1d)84. Notably, genes such
as ARID1B, SMARCA4 and SMARCA2 were among the top mSWI/SNF
genes with most de novo missense and PTVs across all ASD + DD cases,
with ARID1B having the most variants, followed by ANKRD11, KMT2A,
and SCN2A (Extended Data Fig. 1m–n). When including CNV losses and
sequence variants from DECIPHER, the top mSWI/SNF genes implicated
were SMARCB1 and SMARCA2, mutations in which cause the most severe
phenotypes of mSWI/SNF-related NDDs, CSS and Nicolaides-Baraitser
syndrome (NCBRS), respectively85 (Fig. 1c). Nevertheless, multiple
genes may be disrupted in a given CNV, making genotype-phenotype
correlations more challenging to directly assess. As compared to
cancer, wherein mutations in mSWI/SNF genes are present in 20.3% of
all cases sequenced86 (COSMIC: the Catalog of Somatic Mutations in
Cancer), specific mSWI/SNF subunits were more frequently mutated
in NDD relative to other mSWI/SNF genes. These included ARID1B,
the paralog of which, ARID1A, is among one of the most frequently
mutated genes in cancer, SMARCA4, and SMARCA2 (Extended Data
Fig. 1o). Notably, genes encoding PBAF and ncBAF components such
as PBRM1, ARID2, BICRAL (GLTSCR1L) and others were found to be more
frequently mutated in cancer than in NDD (Extended Data Fig. 1p). As
the most frequently mutated chromatin remodeler in NDDs and cancer,
the remainder of this Analysis is centered on the mSWI/SNF family of
chromatin remodeling complexes.
mSWI/SNF NDD variants accumulate in functional domains
To comprehensively examine the full constellation of mSWI/SNF
sequence variants in NDD, we combined mSWI/SNF gene mutations
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1401
Analysishttps://doi.org/10.1038/s41588-023-01451-6 a
i
I
s
V
N
S
R
E
H
P
C
E
D
n
g
n
e
b
-
n
o
n
f
o
r
e
b
m
u
n
n
a
e
M
b
s
n
o
i
t
a
t
u
m
o
v
o
n
e
d
f
o
r
e
b
m
u
N
400
350
300
250
200
150
100
50
0
Distribution of non-benign DECIPHER SNVs
across top 50 GOMFs in DDG2P genes
ECM structural constituent conferring tensile strength (8.74)
RNA Pol-II specific DNA binding TF binding (8.45)
Chromatin DNA
binding (11.12)
Transcription coregulator
activity (12.69)
Protein N terminus binding (8.30)
DNA binding TF binding (8.22)
Protein C terminus binding (8.07)
ATP hydrolysis activity (7.85)
Gated channel activity (7.70)
Transcription regulator activity (7.61)
Sequence-specific DNA binding (7.03)
Chromatin binding (10.35)
Beta catenin binding (10.00)
Voltage-gated
channel activity (9.09)
TF binding (8.49)
Voltage-gated cation
channel activity (8.32)
Cation channel activity (7.42)
Metal ion transmembrane
transporter activity (7.33)
Scaffold protein binding (6.67)
Protein domain specific binding (6.62)
ATP-dependent activity (6.62)
Passive transmembrane transporter activity (6.49)
Cytoskeletal protein binding (6.31)
Hydrolase activity acting on acid anhydrides (6.25)
Protein containing complex binding (5.83)
12
10
8
6
4
2
0
50
40
30
20
10
0
Rank
Chromatin
Structure
Voltage/transmembrane
Protein binding
ATP/catalytic/transferase/hydrolase/ligase/tRNA
NDD (ASD + DD) de novo variants
(missense + PTV) across chromatin regulators
i
e
s
a
n
k
e
n
o
t
s
i
H
x
e
l
p
m
o
C
n
i
s
e
h
o
C
C
R
C
0
8
O
N
I
n
o
i
t
a
l
y
h
t
e
m
A
N
D
D
R
u
N
/
2
D
B
M
D
R
u
N
/
3
D
B
M
C
R
C
F
N
S
/
I
W
S
m
)
r
e
h
t
o
(
C
R
C
D
H
C
y
l
i
e
s
a
r
e
f
s
n
a
r
t
l
y
t
e
c
a
e
n
i
s
y
L
m
a
f
e
s
a
r
e
f
s
n
a
r
t
l
y
h
t
e
m
1
T
E
S
e
s
a
l
y
h
t
e
m
e
d
e
n
i
s
y
L
y
l
i
m
a
f
e
s
a
r
e
f
s
n
a
r
t
l
y
h
t
e
m
2
T
E
S
C
R
C
I
W
S
I
y
l
i
m
a
f
e
s
a
r
e
f
s
n
a
r
t
l
y
h
t
e
m
9
3
V
U
S
e
s
a
t
o
h
p
s
o
h
p
e
n
o
t
s
i
H
)
I
0
6
P
T
/
0
0
4
P
(
C
R
C
0
8
O
N
I
s
e
i
l
i
m
a
f
e
s
a
r
e
f
s
n
a
r
t
l
y
h
t
e
m
r
e
h
t
O
1
C
R
P
2
C
R
P
C
A
D
H
e
s
a
g
i
l
2
P
C
e
M
I
I
s
s
a
l
C
b
U
e
n
i
s
y
L
x
e
l
p
m
o
c
n
i
s
n
e
d
n
o
C
C
A
D
H
I
I
I
s
s
a
l
C
C
A
D
H
I
s
s
a
l
C
i
i
e
s
a
l
y
s
o
b
o
r
e
n
i
s
y
L
e
s
a
n
m
e
d
e
n
n
g
r
A
i
i
e
s
a
l
y
h
t
e
m
e
d
e
n
n
g
r
A
i
i
n
o
i
t
a
l
y
h
t
e
m
e
d
A
N
D
i
e
s
a
n
i
t
i
u
q
b
u
e
d
e
n
i
s
y
L
e
s
a
r
e
f
s
n
a
r
t
l
y
h
t
e
m
e
n
n
g
r
A
i
i
)
X
X
A
D
/
X
R
T
A
(
C
R
C
e
k
i
l
-
F
N
S
/
I
W
S
Subfamily
c
Distribution of DECIPHER mutations (SNV + indels + CNV loss)
across CRCs
separated by complex and gene
% of all
DECIPHER
variants:
mSWI/SNF
(4.10%)
CHD
(2.04%)
ISWI
(1.94%)
INO80
(3.61%)
All CRC
Genes
(9.34%)
Total variants (SNV/
indel/CNV loss/gain):
SNV/indel/CNV loss:
SNV/indel:
CNV loss:
n = 2,281
n = 1,415
n = 255
n = 1,160
n = 1,134
n = 664
n = 176
n = 488
n = 1,077
n = 580
n = 26
n = 554
n = 2,009
n = 1,125
n = 97
n = 1,028
n = 5,196
n = 3,193
n = 572
n = 2,621
SMARCA2/4
SMARCB1
ARID1B/A
BCL7B/A/C
BRD9
SMARCD3/2/1
PHF10
ACTB
ACTL6B/A
SS18/L1
ARID2
DPF1/2/3
SMARCC2/1
BICRA/L
BRD7
PBRM1
SMARCE1
CHD3/4/5
CECR2
CHD8/7/9/6
BAZ1A/B
CHD2/1
RBBP4/7
MTA1/3/2
SMARCA1/5
RBBP7/4
BPTF
GATAD2A/B
BAZ2A/B
HDAC2/1
CHRAC15/17
MBD3/2
RSF1
INO80E
NFRKB
ACTB
YY1
INO80B/C/D
SRCAP
MORF4L1/2
EPC1/2
EP400
TRRAP
ZNHIT1
ING3
ACTR5/6/8
MRGBP
H2AFZ
RUVBL1/2
UCHL5
BRD8
DMAP1
MEAF6
YEATS4
INO80
VPS72
MCRS1
ACTL6A
KAT5
TFPT
t
n
u
o
c
l
a
t
o
T
461
250
200
150
100
50
0
s
s
o
s
s
o
l
l
V
N
C
s
l
e
d
n
i
/
V
N
S
V
N
C
/
s
l
e
d
n
i
/
s
V
N
S
s
s
o
s
s
o
l
l
V
N
C
s
l
e
d
n
i
/
V
N
S
V
N
C
/
s
l
e
d
n
i
/
s
V
N
S
s
s
o
s
s
o
l
l
V
N
C
s
l
e
d
n
i
/
V
N
S
V
N
C
/
s
l
e
d
n
i
/
s
V
N
S
s
s
o
s
s
o
l
l
V
N
C
s
l
e
d
n
i
/
V
N
S
V
N
C
/
s
l
e
d
n
i
/
s
V
N
S
d
mSWI/SNF complexes
CHD complexes
ISWI complexes
INO80 complexes
cBAF complex
Nucleosome
acidic patch
DPF
SMARCB1
Coffin-Siris syndrome
severe ID with hydrocephaly
Kleefstra-like syndrome
DPF1/2*/3
Coffin-Siris syndrome
ARID1A/B
ASD
Coffin-Siris syndrome
generalized non-syndromic ID
Polybromo BAF (PBAF)
subunits
SS18/L1
SMARCA2/4
ATPase subunits
ASD
Coffin-Siris syndrome
Nicolaides baraitser syndrome
Blepharophimosis-impaired intellectual
development syndrome
Syndromic-ID
ACTB
ASD
Baraitser-Winter syndrome
non-syndromic ID
BCL7A/B/C
ACTL6A/B
ASD
syndromic ID
epileptic encephalopathy
SMARCD1/2/3
Syndromic ID
SMARCC1/C2
ASD
Syndromic ID
Coffin-Siris syndrome
SMARCE1
Coffin-Siris syndrome
non-canonical BAF
(ncBAF) subunits
ASD
Pilarowski-Bjornsson
syndrome
CHD1/2
CHD1/2
ATPase subunits
ASD
epileptic
encephalopathy
CERF
CECR2
SMARCA1/5*
ATPase subunits
Lowe syndrome
syndromic ID
SMARCA5/1
(SNF2H/L)
CHD
6/7/8/9
CHD6/7/8/9
ATPase subunits
CHARGE syndrome
Hypogonadotropic hypogonadism
ASD
Snijder Blok-Campeau
syndrome
GAND
syndrome
CHD3/4/5
MTA1/2/3
RBBP4/7
GATAD2A/B
MBD2/3
Sifrim-Hitz-Weiss
syndrome
MTA1/2/3
RBBP4/7
NoRC
BAZ2A/B*
Additional
ISWI subunit(s)
WICH
BAZ1B
(WSTF)
CHRAC/ACF
BAZ1A
(ACF)
CHRAC1
(CHRAC15)
POLE3
(CHRAC17)
RSF
RSF1
Williams-Beuren
syndrome
NURF
RBBP4
BPTF
RBBP7
NDD with dysmorphic
facies and distal
limb anomalies
INO80 complex
INO80B
INO80E
INO80
ACTB
ACTL6A
Gabriele-
de Vries
syndrome
ACTR5/8
INO80C
YY1
MCRS1
RUVBL1/2
INO80D
UCHL5
(UCH37)
TFPT
(AMIDA)
NFRKB
(INO80G)
Johanson-Blizzar syndrome
Jacobsen syndrome
PHF10
ARID2
ASD
CSS-like
generalized non-syndromic ID
PBRM1
SFARI gene
ASD risk score
S-Syndromic
1-High confidence
2-Strong candidate
3-Suggestive evidence
BRD7
BICRA*/L
CSS-like SWI/SNF-related
intellectual disability disorder
BRD9
SWI/SNF-like
DAXX/ATRX complex
ATRX
X-linked Alpha-thalassemia
intellectual disability syndrome
DAXX
HDAC1/2
HDAC1/2
NuRD complex
ASD
Sifrim-Hitz-Weiss syndrome (SIHIWES)
Snijder Blok-Campeau syndrome
GAND syndrome
TIP60/P400 complex
SRCAP complex
Floating-Harbor
syndrome
SRCAP
ACTB
ACTL6A
(ARP4)
EP400
ACTB
ACTL6A
(ARP4)
ING3
(MEAF4)
ZNHIT1
BRD8
EPC1/2
ACTR6
DMAP1
RUVBL1/2
YEATS4
(GAS41)
MEAF6
(MRG15)
YL1
(VPS72)
MRGBP
MORF4L1/2
(MRG15/MRGX)
RUVBL1/2
YEATS4
(GAS41)
YL1
(VPS72)
NDD with dysmorphic facies,
sleep disturbence,
and brain anomalies
KAT5
(TIP60)
BRD8
DMAP1
ASD
Developmental delay with
or without dysmoriphic facies
TRRAP
Fig. 1 | Genes encoding chromatin regulatory complexes represent the
most frequently mutated gene classes in human NDDs. a, Scatterplot of the
average numbers of non-benign SNVs in DECIPHER corresponding to the top
50 GOMF gene sets enriched in DDG2P developmental disorder-associated
genes, ranked by the mutational burden of each gene set. b, Bar graph depicting
the total number of NDD-associated missense and protein truncating variants
(PTVs) for a curated list of chromatin regulatory and epigenetic gene sets,
ranked by mutational burden of each gene set in autism spectrum disorders and
developmental disorders (ASD + DD) from the Simons Foundation Research
Initiative (SFARI) datasets (SPARK: Simons Foundation Powering Autism
Research + SSC-ASC: Simons Searchlight Collection–Autism Sequencing
Consortium, and DDD: Deciphering developmental disorders studies). The
mSWI/SNF chromatin remodeling complex gene set is emphasized in red.
c, Heatmaps depicting the mutational frequency for genes encoding members of
the SWI/SNF, CHD, ISWI, and INO80 complex families in DECIPHER. Total number
of variants (including copy-number variant (CNV) gain, copy number variant
loss (CNV loss), single nucleotide variant (SNV) and indel mutational frequencies
are indicated. Percentage of total DECIPHER sequence variants are indicated for
each chromatin remodeling complex family (top). d, Cartoon representations of
the four classes of chromatin remodelers (SWI/SNF, CHD, ISWI and INO80) and
respective subcomplex or related complex associations, colored by CNV loss/
SNV/indel variation frequency from panel c. Interchangeable subunit paralogs
are colored by their combined mutational frequency. Autism spectrum disorder
(ASD) risk score (SFARI) and developmental disorder associations curated from
literature and OMIM (Online Mendelian Inheritance in Man, a catalog of human
genes and genetic disorders; https://www.omim.org/) are indicated. Asterisk (*)
indicates paralog implicated in NDD. Where possible, cartoons were based on 3D
structural data available from human and yeast structures; ovals are used in in
lieu of structural cartoons for components lacking structural data.
from the DECIPHER, ClinVar, LOVD, SFARI SPARK and SSC–ASC datasets
and merged these with mutations reported in published literature as well
as n = 85 novel, previously unreported cases (Supplementary Table 1).
After removing duplicates, variants with a mutant allele frequency of
>0.5% in the general population as assessed by gnomAD87, and filtering
for missense, inframeshift (herein defined as non-frameshift inducing
insertions/deletions), frameshift and nonsense variants, we identi-
fied 2539 variants in mSWI/SNF genes, 61.5% of which were missense
(Fig. 2a). Variants resulted predominantly in missense or inframeshift
(67.1%) (Fig. 2a,b), with the exception of ARID1B and ARID2, for which
the majority of variants were nonsense or frameshift (Fig. 2b). The
greatest number of missense variants stemmed from G > A and C > T
base pair conversions, resulting in a variety of amino acid changes
(Extended Data Fig. 2a–e). The most frequently altered residues were
Arginine (R), Proline (P), Alanine (A), and Glycine (G), together making
up 47% (815/1703) of all missense and inframeshift affected residues
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1402
Analysishttps://doi.org/10.1038/s41588-023-01451-6
in the dataset (Extended Data Fig. 2b–e, Supplementary Table 1).
Furthermore, the most common missense amino acid substitution
was Arginine to Histidine (Arg>His; R > H), indicating reductions in
both the relative size and pKa of the amino acid side chain (Arg pKa
12.48 – His pKa 6.0) (Extended Data Fig. 2e).
A high percentage of missense and indel mSWI/SNF mutations
localized to highly conserved regions (53.1% high, 24.7% moderate
conservation) (Fig. 2c). Mutations in subunits such as ACTB, ACTL6A/B,
DPF2, and SMARCB1 entirely or nearly entirely occurred in intra-domain
structured regions, whereas variants in BCL7A/B, PHF10, and ARID1A/B
subunits were skewed toward interdomain disordered regions
(Fig. 2d, Extended Data Fig. 2f and Supplementary Table 3). Intrigu-
ingly, mutations in SMARCA2 clustered in the ATPase/helicase domain,
whereas mutations in SMARCA4 were more dispersed throughout the
protein, including the structurally unresolved N terminus (Fig. 2e).
Interestingly, whereas mutations within the SMARCA2 helicase
cause NCBRS, SMARCA2 mutations outside of this domain are impli-
cated in a distinct disorder, blepharophimosis-impaired intellectual
disability syndrome88. Among mSWI/SNF paralogs, frameshift muta-
tions were more enriched in ARID1B, whereas missense mutations in
specific regions were enriched in ARID1A, clustering namely in the
ARID DNA-binding domain, the structurally unresolved N terminus and
the C-terminal armadillo repeat domain (ARM or core binding region)
(Fig. 2e). A possibility underlying this difference is that ARID1A haplo-
insufficient mutations lead to a more severe phenotype, as suggested
by the frequent occurrence of mosaic variants69 and further substanti-
ated during the review process by an analysis of fetal cases89.
Genotype-phenotype clinical studies have suggested that ARID1B
truncating mutations are generally linked to the mildest cases of
CSS-related intellectual disability, including some individuals without
intellectual disability90, whereas single amino acid mutations of the
SMARCB1 protein are correlated with the most severe cognitive impair-
ment and growth delay in CSS21,69,85. SMARCA2-ATPase mutations result
in severe intellectual disability cases of NCBRS, but SMARCE1-HMG
and DPF2-PHD mutations are correlated to moderate-severe and mild
intellectual disability phenotypes, respectively72,74,91. We examined
non-truncating variants through predicted phenotypic severity
score analysis (PolyPhen HumVar92), which highlighted domains such
as the SMARCB1-CTD, ARID2-ARID and SMARCA2-Helicase-C and
SMARCA2-post-Helicase-C as those predicted to result in most severe
disease phenotypes, in agreement with published phenotypic data
(Fig. 2f and Supplementary Table 3). This analysis also highlighted the
SMARCC1-post-SWIRM interdomain with a particularly high PolyPhen
score and average number of mutations; this region lacks 3D structural
definition, implicating an alternative contribution to mSWI/SNF func-
tion (Fig. 2f). Collectively, these results highlight convergent clinical
outcomes stemming from mSWI/SNF gene disruption, with variation in
severity observed across distinct proteins and even domains of mSWI/
SNF complex components.
Mapping NDD missense/inframeshift variants on 3D
SWI/SNF-nucleosome models
We next integrated these sequence variant data with recently solved
structures of mSWI/SNF cBAF complexes93,94, which allowed for map-
ping of 238 unique positions comprising 44.08% (655/1,486) of the theo-
retically mappable cBAF-specific NDD missense and in-frame indels
on the recombinant cBAF cryo-EM structure, and 51.55% (766/1,486)
on the endogenous structure for all cBAF paralogs (Fig. 3, Extended
Data Fig. 3a, b and Supplementary Table 3)95,96. These results highlight
the need for further structural efforts as well as studies to define the
roles and interactions of non-structured, disordered regions. Mapping
subcomplex-specific positions onto the recently solved PBAF complex
bound to a nucleosome97 resolved 20 additional PBAF-specific subunit
mutations across ARID2, PBRM1 and BRD7 (Extended Data Fig. 3c). For
ARID1B, SMARCA2 and ACTL6B, paralog subunits that are not part of
the solved protein complex, we mapped mutant residues on to the
respective paralogs following paralog alignment (Fig. 3 and Extended
Data Fig. 3a).
This structural analysis reveals that BAF complex compromises in
NDD cluster primarily in four distinct regions on mSWI/SNF complexes:
the catalytic ATPase module, the mSWI/SNF core, the Arp module, and
the SMARCB1 BAF-nucleosome contact point (Fig. 4a–d). As demon-
strated initially through our previous work98 and later resolved in 3D
structural efforts, CSS-associated mutations in SMARCB1 localize to the
SMARCB1-CTD, the key and only interface connecting the mSWI/SNF
core module to the nucleosome acidic patch (Fig. 4a and Extended
Data Fig. 4a). Second, mutations in the SMARCA4 ATPase subunit
are primarily situated in the ATP-coordinating and DNA-binding resi-
dues near the nucleosome, with additional mutations accumulating
within the region of SMARCA4 interfacing within the mSWI/SNF core
(Fig. 4b,c and Extended Data Fig. 4b). We also identified a cluster of
variants are found throughout the ACTB subunit of the Arp module,
whose mutation is associated with severe cases of Baraitser-Winter
cerebrofrontofacial syndrome75,95 (Fig. 4d).
Intriguingly, whereas mutations to positively charged residues
within the SMARCB1-CTD disrupt binding to the nucleosome and result
in severe intellectual disability93,94,98, we report two novel variants in
the SMARCB1-CTD, D369E and R376K, in which a positive or negative
charge is maintained, and which are phenotypically associated with less
severe disease (Fig. 4a, red, and Supplementary Table 1), underscoring
that defining chemical properties of distinct mutations, even within a
given subunit domain, may inform intellectual disability severity and
phenotypic outcomes.
We next mapped cBAF NDD-mutant residues by amino acid char-
acteristics (that is, charged, polar, nonpolar, etc). This map highlighted
that many NDD-associated ACTB residues are nonpolar, the mutation
of which is predicted to disrupt hydrophobic core as further suggested
by Missense3D96,99 (Extended Data Fig. 4c and Supplementary Table 3;
http://missense3d.bc.ic.ac.uk/). Within the context of mSWI/SNF (ACTB
Fig. 2 | Analysis of NDD-associated SNV and indel mutations in mSWI/SNF
complex components. a, Pie chart reflecting the distribution of n = 2,539 mSWI/
SNF NDD-associated SNV and in-frame indel mutations from an integrated dataset
containing data from SPARK, SSC-ASC, DDD, DECIPHER, ClinVar, LOVD, literature
curation and novel variants reported in this study. b, Bar chart summarizing total
NDD-associated missense/in-frame deletions and insertions (red) and nonsense/
frameshift-inducing mutations (blue) across all mSWI/SNF genes. c, Scatterplot
of the negative-normalized ConSurf conservation score versus the mutational
recurrence at each mSWI/SNF complex subunit residue for NDD missense and
in-frame variants in the integrated dataset. Highly conserved and highly mutated
positions are labeled. d, Stacked bar chart summarizing proportion of NDD-
associated missense and in-frame insertion/deletion variants falling within (intra,
blue) or outside (inter, orange) of mSWI/SNF subunit domains in the integrated
dataset. Domains annotated by PFAM, UniProtKB, manual curation, and
structurally resolved domains (see also Supplementary Table 3). e, Lollipop plots
of NDD mutations in the integrated dataset across protein domain schematics
of ARID1A/B, ARID2, SMARCA2/4, SMARCB1, SMARCC1, SMARCE1, and DPF2
subunits generated with Protein Paint. Missense (blue), nonsense (orange),
frameshift (red), in-frame deletions (gray) and insertions (brown) are shown.
Kernel density estimates (relative frequency distribution) of gnomAD missense
mutations (purple line) are overlaid. Domain annotations informed by PFAM,
UniProtKB, manual curation, or by structurally resolved domains are indicated.
ConSurf conservation scores are shown in a cyan-white-magenta heatmap in
increasing conservation order, and structural coverages of the nucleosome core
particle (NCP)-bound human cBAF (light orange, PDB: 6LTJ), endogenous human
cBAF-NCP bound (red, PDBDEV00000056), and by both structures (brown).
Structural coverage for the NCP-bound PBAF complex is also shown for ARID2
(light green, PDB:7VDV). f, PolyPhen HumVar predicted phenotypic severity
score and missense mutational recurrence of mSWI/SNF gene mutations from the
integrated dataset in intra (blue) and inter (orange) domains.
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1403
Analysishttps://doi.org/10.1038/s41588-023-01451-6 is also a member of INO80 and TIP60 complexes; Supplementary
Table 2), ACTB mutations are predicted to alter buried hydrophobic
cavities, as well as interaction with the ACTL6A Arp module binding
partner, and even the HSA helix of the SMARCA4 ATPase (Extended
Data Fig. 4d). Intriguingly, some of the most recurrent ACTL6A and
ACTL6B mutations of the Arp module, R377W and G343R, are located
in close proximity to one another when mapped onto ACTL6A subu-
nit on the cBAF structure (Fig. 4d). Although not interfacing other
246
230
208
145
103
98
91
81
77
76
a
Mutational distribution of
SNVs + short sequence variants
142
(5.6%)
487
(19.2%)
349
(13.7%)
1561
(61.5%)
Total = 2,539
Missense
Inframeshift
Frameshift
Nonsense
b
ARID1B
SMARCA4
SMARCA2
ARID1A
ACTB
ARID2
SMARCC2
BICRA
ACTL6B
SMARCB1
PBRM1
SMARCC1
SMARCD2
DPF2
SMARCD1
DPF3
BRD9
ACTL6A
SMARCE1
SMARCD3
BCL7C
SS18L1
DPF1
BICRAL
BCL7B
BCL7A
SS18
PHF10
BRD7
44
36
34
33
27
23
23
20
20
18
15
12
8
8
4
3
3
2
Missense/inframeshift
Nonsense/frameshift
851
c
e
r
o
c
s
n
o
i
t
a
v
r
e
s
n
o
c
f
r
u
s
n
o
C
)
d
e
v
r
e
s
n
o
c
e
r
o
m
s
i
r
e
h
g
h
i
;
e
v
i
t
a
g
e
n
(
1
0
−1
−2
−3
0
200
400
600
800 1,000
Total variants
Proportion of missense and in-frame indel
NDD mutations within domains
ARID2 (E98K)
SMARCB1 (K364del)
SMARCC1 (M582I/V)
ACTB
(R196C/H/L/S)
ACTB (R183G/W)
PBRM1 (R365C)
ACTL6B (G343R)
d
s
n
o
i
t
a
t
u
m
D
D
N
f
o
n
o
i
t
r
o
p
o
r
P
1.0
0.8
0.6
0.4
0.2
0
0
5
.
2
0
5
.
5
.
7
.
0
0
1
5
.
2
1
.
0
5
1
5
.
7
1
.
0
0
2
Number of NDD missense and
in-frame indel mutations
B
T
C
A
A
6
L
T
C
A
B
6
L
T
C
A
2
F
P
D
1
B
C
R
A
M
S
1
E
C
R
A
M
S
I
2
D
R
A
2
A
C
R
A
M
S
C
7
L
C
B
2
C
C
R
A
M
S
4
A
C
R
A
M
S
3
F
P
D
1
C
C
R
A
M
S
8
1
S
S
1
F
P
D
9
D
R
B
0
1
F
H
P
1
L
8
1
S
S
A
1
D
R
A
I
3
D
C
R
A
M
S
1
M
R
B
P
B
1
D
R
A
I
2
D
C
R
A
M
S
1
D
C
R
A
M
S
A
R
C
B
I
B
7
L
C
B
7
D
R
B
A
7
L
C
B
L
A
R
C
B
I
Subunit
Inter domain
Intra domain
e
gnomAD
Domains
Conserved
1
2
3 4 5 6 7 8 9
Coverage
SMARCA4
ARID1A
QLQ
HSA
BRK
Helicase-N
Heli-C
SnAC
BD
ARID
ARM (CBRB)
0
200
400
600
800
1,000
1,200
1,400
1,600
0
500
1,000
1,500
2,000
cBAF recombinant
cBAF endogenous
Both structures
SMARCA2
ARID1B
gnomAD
Domains
Conserved
1
2
3 4 5 6 7 8 9
Coverage
QLQ
HSA
BRK
Helicase-N
Heli-C
SnAC
BD
ARID
ARM (CBRB)
0
200
400
600
800
1,000
1,200
1,400
0
250
500
750
1,000
1,250
1,500
1,750
2,000
SMARCB1
ARID2
gnomAD
Domains
Conserved
1
2
3 4 5 6 7 8 9
Coverage
WH
RPT1
RPT2
CC
ARID
ARM (CBR)
RFX_DBD
ZnF
0
50
100
150
200
250
300
350
0
250
PBAF structure coverage
500
750
1,000
1,250
1,500
1,750
SMARCE1
SMARCC1
gnomAD
Domains
Conserved
1
2
3 4 5 6 7 8 9
Coverage
HMG
CC
SWIRM
SANT SWIRM-Assoc
CC
0
50
100
150
200
250
300
350
400
0
200
400
600
800
1,000
DPF2
gnomAD
Domains
Conserved
1
2
3 4 5 6 7 8 9
Coverage
Requiem
PHDs
0
50
100
150
200
250
300
350
f
e
r
o
c
s
y
t
i
r
e
v
e
s
r
a
V
m
u
H
n
e
h
P
y
l
o
P
1.0
0.8
0.6
0.4
0.2
0
ARID1A-ARID
ARID1B-ARID
SMARCA4-HeliN.
SMARCA2-HeliC.
ARID2-ARID
SMARCA2-post-HeliC.
SMARCB1-CTD
SMARCC1-
post-SWIRM
DPF2-PHDs
SMARCA4-Cterm
SMARCE1-HMG
Inter domain
Intra domain
ACTB-ACTB
SMARCC2-pre-CC
PBRM1-post-BD2
SMARCC2-N-term
0
0.5
1.0
1.5
2.0
Average number of NDD missense mutations per residue
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1404
Analysishttps://doi.org/10.1038/s41588-023-01451-63S11Afs*913A44_A45dup3A88_G92del6G125SG125V4A167dup5A247dup3Q372Sfs*195P392H2D1050YD1050E6L1831V3K2045Rfs*522G11RI28M7R37HR121GE122DR153GT163AE184KI195Mfs*13D196NR201QE216KA238TI253_L254delinsMV262IH269R2G271EA294TK296NL301P2R341WL352Tfs*92T357RE362_I365delinsV3K363N2K363E21K364del2R366CR366GD367GD369VD369E5R374QR374WR374Yfs*48R376K3R377HP69Q2Y73CY73S3Y76CK92NW104S3R105QE120KY126DK256R2R313CN337YH346N2E359delG370V8R359QR359W2I366MI366F2A450SA450D3R451CR451P3P616L4P674L2K744NK744E4R885C2R885H3M886V3V902M5R979Q2R1043WR1043QR1043L3A1186VA1186T3R1203HR1203CMISSENSE, n = 235NONSENSE, n = 3PROTEINDEL, n = 6FRAMESHIFT, n = 212R885*7R1089*R1089Q8Q1183Pfs*14Q1183Sfs*15Q1183*6E1743Afs*9E1743*11F1785Lfs*528K1844Sfs*178R1977*4R525H2R525C2G752A2G752R2S783L2S783W4E852K3E852QE852DH854LH854YH854R3P883L2P883QP883A4R937H2R937LR937C5H939Y3G1098DG1098R5R1105HR1105SR1105PR1105CR1105GG1132SG1132RG1132D3R1159G3R1159QR1159L4R1162H2R1162CR1162S6A1201VA1201EMISSENSE, n = 219PROTEINDEL, n = 3FRAMESHIFT, n = 4PROTEININS, n = 2NONSENSE, n = 2MISSENSE, n = 51FRAMESHIFT, n = 3PROTEINDEL, n = 23MISSENSE, n = 18PROTEINDEL, n = 22R60QR78CL122V2R179HG189VS200C2I201LD274A2C276FQ288HC298RC298GC298S3A315VW322RC330WC345GD346ND346G2D349N2R350HM354Tfs*24C356GC356Pfs*52T358Nfs*6P365RW369RC374Wfs*4S390YMISSENSE, n = 30FRAMESHIFT, n = 42R53Efs*5R53Sfs*1210E98K2N387*2Y423Afs*392V846Lfs*3Y939Cfs*11Y939Sfs*162G1139Sfs*202Q1440*2H1481Ifs*4V81MK103RR135QK179*I187TW279GT415Kfs*29Y454CC462RP472LR499HH526PQ531Vfs*33V535Sfs*29Q538HD561Ifs*23V576G13M582IM582VQ597PL702FA717VQ742RS822Vfs*3K891Rfs*6R912*Q956*M958T2H971RQ972Sfs*19Q1005*FRAMESHIFT, n = 32PROTEININS, n = 22PROTEINDEL, n = 14MISSENSE, n = 115NONSENSE, n = 25MISSENSE, n = 223PROTEININS, n = 24FRAMESHIFT, n = 345PROTEINDEL, n = 13NONSENSE, n = 246MISSENSE, n = 36PROTEINDEL, n = 1FRAMESHIFT, n = 43NONSENSE, n = 23MISSENSE, n = 33NONSENSE, n = 4FRAMESHIFT, n = 7
SMARCA4 (ATPase)
T1170A
S1176G
A1186V/T
A1190V
V1199M
V1201M
R1203H/C
R1308W
H1309R
I1347N
D1351N
G784R
S813L
A826S
R841I
R842Q/W
T859M
K865E
R874H
G883S
R885C/H
M886V
C891Y
T894M
V902M
R906C
P913L
L921F
C936S
R978Q
R979Q
ACTL6A
I18T
I70T
R76C
M92T/I
H105R
M122L
P126L
T175I
I340V
R377W
R389W
SMARCA4
N477D
R359Q/W
I366M/F
R370C
L430P
N435D
E449K
A450S/D
R451C/P
T453I
E457D
K458_E465del
SMARCA4 (ATPase)
R1043Q/W/L
H1010R
M1011T
S1079L
L1092R
L1108P
M1109V
M1112T
A1117V
R1119C
G1120S
T1156A
R1157Q/G
D1169G
SMARCB1
G11R
R37H
G271E
R341W
T357R
K363N/E
K364del
R366C/G
D369V/E
R374Q/W
R377H
I28M
E184K
D196N
R201Q
E216K
A238T
V262I
H269R
A294T
K296N
L301P
E362_I365delinsV
D367G
R376K
ARID1A
T1649R
P1651L
R1656K
D1697G
G1711E
L1713P
L1731S
G1737S
T1743M
N1820D
L1831V
R1833H
I1988T
F1993_V1994dup
N1997D
G2012R
L2061S
N2066Y
E2078K
V2084G
H2090R
A2137D
R2164W
L2171dup
N2173I
G2177R
M2231T
R2233Q
R2236C
P2265L
N2194_L2195delinsGN
Red text: indicates recurrence of 3 or greater
Bold text: indicates newly reported case
SMARCD1
Q145P
P293T
D330E
I428V
R446G
V470del
A487G
F495L
SMARCE1
K256R
DPF2
R60Q
R78C
BAF NDD mutations
Primary subunit (in structure)
Paralog mapped
In both primary and paralog
SMARCC2
R443Q
A445T
L609P
L610P
L613P
C635R
A868T
E893G
M896V
E900G
E921K
ACTB
D2Y
D3F
A7T/V
D11del
N12H/D
A22T
D24N
A26V
R28G
F31L
H87D
F90del
N92S
V103L
A108D
P109S
E117V/D/K
M119T
T120I
I136V
L140V
S145C
T149I
P164S/H
L171F
H173Y
L176P
D179E
R183W/G
D184Y
T194A
R196H/C/L/S
G197S
A204G
R206Q/W
V209M
R210H/C
K215N
T229M
P243L
G245S
P258L
A260E
G268R/C
Y294D
N296K
V298A
G302A
M305T
K315del
P322R
I330L
R335P
S338F
I341V
S348L
W356R
E364K
H371R
Fig. 3 | Mapping of 238 unique NDD-associated variant positions onto the
structure of the human cBAF complex. NDD-associated variants, including
14 novel variants, mapped on to the 3D structure of the human cBAF complex
(PDB:6LTJ). Residues shown in red spheres represent NDD-associated variants
in the subunit indicated, residues in blue represent those mapped from the
paralog subunit, and residues in purple represent NDD-variants mapped in both
the primary subunit present on the cBAF structure and paralog mapped subunit.
Recurrent variants (n ≥ 3) are emphasized in red text. Caution is needed when
evaluating these variants in a clinical context since not all variants are confirmed
as causal.
mSWI/SNF subunits, these residues are oriented toward the DNA exit,
and we hypothesize that the ACTL6A-R377 residue may stably bind the
DNA backbone adjacent to the nucleosome, which would be disrupted
upon mutation to a nonpolar residue such as tryptophan (R377W).
Conversely, the addition of a positive charge in ACTL6B from side
chain-absent glycine (G) to arginine (R) upon mutation may impart
affinity to the nucleosomal DNA.
We predicted that SMARCB1 mutations in the RPT2 domain may
disrupt the RPT domain cavity (Extended Data Fig. 4e). Further, the
recurrent SMARCB1-R37H mutation in the winged-helix DNA-binding
domain, which causes severe intellectual disability and Kleefstra-like
syndrome, also demonstrated hydrogen bonding with the carbonyl
backbone of ARID1A-L2073 and Y2076 that is likely disrupted upon
mutation (Extended Data Fig. 4e). Intriguingly, the SMARCB1-WH
domain is isolated from the SMARCB1 C-terminus on the recombinant
cBAF structure but is predicted to be repositioned closer to the nucleo-
some binding lobe in the PBAF structure97, suggesting potentially
distinct roles and functional impacts of the SMARCB1-R37H muta-
tion in cBAF compared to PBAF, perhaps independent of remodeling
activity as the SMARCB1-R37H mutation does not impact cBAF nucleo-
some remodeling activity in vitro98.
Yeast SWI/SNF ATPases offer NDD variant functional insights
Given the high frequency of mutations within the catalytic ATPase
subunits of mSWI/SNF chromatin remodeling complexes, SMARCA2
and SMARCA4, we mapped conserved mutant residues onto the
nucleosome-bound yeast SWI/SNF and SNF2 structures100,101 (Extended
Data Fig. 4f). Interestingly, the current human cBAF structures do not
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1405
Analysishttps://doi.org/10.1038/s41588-023-01451-6 Nucleosome engagement:
SMARCB1-CTD
b
Nucleosome engagement/ATP coordination:
SMARCA4/2- ATPase (helicase)
Nucleosome acidic patch
NDD-associated mutated residues
SMARCB1
C-terminal α-helix
R366C/G
K363N/E
R374Q/W
K364del
T357R
R377H
D369V/E
R341W
G271E
SMARCB1
NDD-associated
mutated residues:
SMARCA4
SMARCA2
ATPase
ATP binding
pocket
R1308W
R1203H
A1186V/T
R1157Q/G
A1117V
R1043W/L/Q
K364del
R37H
R374Q/W
K363E/N
R366C/G
R377H
G11R
G271E
R341W
T357R
D369E/V
0
3
6
R341W
D369E/V
R376K
0
9
12
15
Recurrence
18
21
SMARCB1
1
Recurrence
2
R979Q
A826S
R885C/H
M886V
T859M
R842Q/W
R906C
V902M
a
c
ATPase-ARM core module insertion:
SMARCA4/2-ARID1A/B interface
ARID1A
NDD-associated
mutated residues
SMARCA4/ARID1A
SMARCA2/ARID1B
ARID1A
L1831V
D1697G
6
7
d
Arp module:
ACTL6A/B and ACTB
ACTB
NDD-associated
mutated residues
ACTB
3
4
5
Recurrence
ARID1B
SMARCA4
R451C/P
A450S/D
P287S
L1054_Y1055delinsH
K1072_W1073delinsR
M1273I
R1593Q
N2194_L2195delinsGN
R2233Q
0
0
K1032N
M1686I
V2013_L2015dup
C2032R
ARID1A
Recurrence
ARID1B
Recurrence
1
1
ACTL6A
NDD-associated
mutated residues
ACTL6A
ACTL6B
0
1
2
3
Recurrence
4
5
6
0
1
2
G125S/V
L1831V
A247dup
P392H
A167dup
A44_A45dup
A88_G92del
D1050E/Y
P877L
Q1333_Q1334dup
A1136T/V
L1011F
A345_A349dup
R774C
P407L/S
R376W
A166_A167dup,A166_A167del
D1697G
G6D,G6_A8dup
D1772N
G144_A147del,G144_A147dup
Y550C/H
G1303R
M1366T
R2199C/H
Q1315H/R
R1901Q
P1836H/R
E1687G
P1560A
P1401A/L
A1540S/V
S320_G328del,
Y792C/H
S320_G327del,S320_G324dup
G778C/S
A274dup
A45_A47dup
G327A
P724S
I2018N/T
V2013_L2015dup
G1098R
H1955P/R
M1952T
P1881S
R1832C/H
N1659I/S
E1592K
A1474V
G1351S
R1313H
A346_A350del
M1067L
D1061N
K1032N
C878Y
Q876E
A744V
A460dup,A460_A464del
A459T,A459_A460del
G377V
A341T
G333_G335dup
V2106I
R1105C/G/H/P/S
E852D/K/Q
R937C/H/L
R1159G/L/Q
R1162C/H/S
A1201E/V
R525C/H
P883A/L/Q
H939Y
G752A/R
S783L/W
G1098D/R
E929V
A1188E/P/T
G1132D/R/S
H854L/R/Y
R855G/Q
R505Q
L766V
Q1241E
Q1165H/K
Q1155P
K951N/R
L946F/S
G881R/V
D851G/H
N787K
Q296P,Q296_A301del
T756I
Q676L
D534N/Y
L529V
G513D/V
N486K
H484N
N453S
R1245G
M886V
M98V
R841I
R979Q
G1031D
R1043Q
T1156A
T1170A
V1201M
D1298N
I330L
A7T/V
N12H/D
R28G
G268R/C
T120I
S145C
P164S/H
D179E
R183G/W
D184Y
R196H/C/S/L
G197S
R206Q/W
V209M
R210H/C
E117D/V/K
ACTL6A
R377W
T175I
H105
M92I/T
[ACTL6B]
G346 [G343R]
F150 [F147del]
E80 [D77G]
L209 [L206P]
G352 [G349S]
R133 [R130Q]
M140 [M137I]
A232 [P232A/T]
I338 [I335T]
V424 [V421M]
SMARCA2
SMARCA4
R359Q/W
R885C/H
R979Q
R451C/P
P674L
R1043L/W/Q
A1186T/V
R1203C/H
V902M
Q1606R
R1411Q
M886V
K744E/N
P616L
A450D/S
I366F/M
R1308W
G951R
E1023K
A1117V
L1126F
G1146S
R1157G/Q
P1277L
D1298E/N
L1609P
R1329H
R1405W
A1419T
D1435N
T1459I
R1608Q
T859M
P1624L
V1626M
R906C
K546del
R842Q/W
V318I
A136V
S141L
P195H
P221T
G237S
G241_P244del
P277S
P305L
S323L
A826S
A340T
R370C
R521G/W
R549C/L
G630D
A677T
A703V
R726C
S1631C
0 1
2 3 4 5 6 7 8 9
Recurrence
SMARCA4
0
1
Recurrence
2
0 1
2 3 4 5 6 7 8 9
Recurrence
ACTB
R196C/H/L/S
R183G/W
R206Q/W
A7T/V
P70A/H/L
H73D/Q/Y
V209M
R210C/H
E117D/K/V
S145C
G268C/R
G197S
D184Y
D179E
P164H/S
S60N/R
T120I
I75M/T
G74S
L65F/V
R28G
N12D/H
I330L
H105R
R377W
T175I
M92I/T
E227Q
G343R
F147del
D77G
L206P
G349S
R130Q
M137I
P232A/T
R298Q
I335T
V421M,V421_C425del
0
2
4
6
8
10
Recurrence
12
14
16
ACTL6A
0
1
2
Recurrence
3
4
ACTL6B
0
2
4
6
8
10
12
14
Recurrence
Fig. 4 | NDD-associated mutations cluster within key structural hubs
of mSWI/SNF complexes. a, Zoomed-in view of the SMARCB1 C-terminal
alpha-helix domain (PDB:6LTJ) with the nucleosome acidic patch interaction
site highlighted in yellow (left). NDD-associated mutations in SMARCB1 are
emphasized in red. All NDD-associated SMARCB1–C terminal alpha-helix
mutations ranked by frequency (right). Novel SMARCB1 variant cases reported
in this study shown in red bar chart. b, Zoomed-in view of the SMARCA4 ATPase
subunit within the cBAF complex (PDB:6LTJ) at its interface with the nucleosome
(left). Mutations in SMARCA4 are indicated in red; mutations in SMARCA2 are
indicated in blue, shared mapped in purple. ATP binding pocket is highlighted
in yellow. NDD-associated missense and inframeshift variants in SMARCA4
and SMARCA2, ranked by frequency, filtered for recurrence of n ≥ 2 by position
(right). Novel SMARCA4 cases reported in this study shown in red bar chart.
c, NDD-associated mutations in ARID1A and ARID1B, ranked by frequency,
filtered for recurrence of n ≥ 2 by position (left). Zoomed-in view of the
SMARCA4-ARID1A interface within the core module of the cBAF complex (right).
SMARCA4 is shown in tan and ARID1A in light purple, with mutations in SMARCA4
and ARID1A shown in red and those in their respective paralogs SMARCA2 and
ARID1B shown in blue. Novel ARID1A/B variant cases reported in this study shown
in red bar chart. d, Left, zoomed-in view of the ACTB (tan) and ACTL6A (light
purple) subunits within the Arp module of the cBAF complex, with mutations
indicated in red and blue for ACTL6A paralog subunit, ACTL6B. NDD-associated
mutations in ACTL6A, ACTL6B and Actin, ranked by frequency, filtered for
recurrence of n ≥ 2 by position (right). Recurrent ACTL6B variants donated in
brackets mapped onto ACTL6A indicated.
resolve the brace helices, and we highlight residues that are buried
in the brace helices (SMARCA4 978-979) (Extended Data Fig. 4f).
Cancer- and NDD-associated mutations (R973W and R1243W) in the
brace helices of SMARCA4 were recently found to diminish nucleo-
some remodeling activity of PBAF complexes in vitro97. Given their
proximity to this region and the ATP pocket of SMARCA4, we posit
that additional variants in the brace helices and the nearby R978Q and
R979Q variants would have similar deficits in nucleosome remodeling
in human cells (Extended Data Fig. 4g). To assess the potential impact
that NDD-associated mutations might have on ATP engagement, given
that structures are static, we mapped conserved SMARCA2/4 mutant
residues onto the open state, ADP bound (similar to apo structure) and
onto the closed, ADP-BeFx-bound yeast SNF2 nucleosome bound struc-
tures102, which allows mapping of ~85% of all SMARCA2/4-ATPase posi-
tions (Extended Data Fig. 4h). Furthermore, this mapping highlighted
NDD-associated nucleosome binding residues such as N1050 and
K1057 (corresponding NDD variants: SMARCA2-N1007K and K1044E),
which were previously shown to dramatically diminish nucleosome
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1406
Analysishttps://doi.org/10.1038/s41588-023-01451-6 a
Unique cancer and NDD
missense + inframe variants
Cancer overlap of NDD variants
(missense + inframe indel)
b
Shared sorted by cancer
Shared sorted by NDD
NDD unique
Cancer
and NDD
41.6%
overlap
19,364
502
702
NDD
Cancer
(cBioPortal PanCan/GENIE
+COSMICv94)
c
ACLT6B-G343R
Gene
Mutation NDD
Cancer
combined*
cBioPortal
-PanCan
cBioPortal
-GENIE
COSMIC
Domain
Gene
Mutation NDD
Cancer
combined*
cBioPortal-
PanCan
cBioPortal-
GENIE
COSMIC
Domain
Gene
Mutation
NDD
Domain
CTD
CBR
CTD
CTD
Nterm
Nterm
Nterm
SMARCB1
K364del
PBRM1
R365C
SMARCC1 M582I
SMARCA4
R359Q
ACTB
BICRA
R196H
R1475H
SMARCB1
R37H
ARID1A R1
ARID1A
L1831V
SMARCA4 R6
ARID1A
ARID1A R1
ACTB
G125S
R196C
21
13
13
8
8
8
7
6
6
6
74
3
1
24
4
1
5
26
7
5
16
3
0
2
2
1
0
3
0
2
50
0
0
22
0
0
5
20
6
0
8
0
1
0
2
0
0
3
1
3
CTD
ACTL6B
G343R
PBRM1 R3
ARID2
SMARCC1 R3
ACTB
E98K
R183W
SMARCA4 R2
SMARCC2
E788G
ACTB
BRD9
E61D
BICRA R2
SMARCA2 H939Y
Winged_Helix
ACTB
R206Q
ARID1AR3
SMARCD2 M178L
ARID1AR1
SMARCC2 D805N
ACTB
ACTL6B
F147del
14
10
9
8
5
5
5
4
4
4
Actin
ARID
ACTB
SMARCC2 R5
BRD9 R1
Nterm
ACTB
SMARCD2 R1
SMARCC2 R5
Actin
Local NDD enrichment
Local cancer enrichment
Enriched in NDD
Enriched in cancer
1
0
1
0
1
0
–1
Nonrecurrent
in cancer
25.2% (304)
Recurrent
in cancer
16.4% (198)
NDD-only
58.3%
(702)
Total = 1,204
SMARCB1
R377H
ARID2
R314C
SMARCB1
K364del
SMARCB1
R374Q
SMARCA4
P913L
SMARCA4
S813L
SMARCA4
R885H
ARID1A
A45del
SMARCA4
P1277L
ARID1A
A167dup
3
1
21
5
1
1
2
1
2
4
180
99
74
61
57
33
31
31
30
29
43
22
16
17
12
5
4
7
4
2
103
52
50
30
33
22
23
17
22
25
34
25
8
14
12
6
4
7
4
2
d
NDD-cancer normalized
enrichment (PDB:6LTJ)
S
E
N
r
e
c
n
a
c
-
D
D
N
Enriched
in NDD
1
0
Enriched
in cancer
–1
SMARCB1-K364del
DPF3-V37M
ARID1B-D1772N
SMARCD1-I428V
SMARCD2-I444T
SMARCD3-I396T
ACTB-R183W
SMARCD2-M178L
SMARCA4-R359Q
SMARCD1-Q145P
e
ARID1A-ARID
DNA binding domain
NDD mutations
DNA binding residues
DPF2-PHD domain
(histone tail binding)
NDD mutations
Zinc ion
Q1066
G1062
F1103
Y1027
W1023
R1020
E1017
K1021
K1072
R1074
L1092
L1054
R1053
L1049
D1050
L1011
C298
W322
D349
D274
C276
A315
Q288
SMARCE1-HMG
DNA binding domain
NDD mutations
K92
SMARCB1-WH
DNA binding domain
NDD mutations
DNA binding residues
R105
W104
Y76
Y73 P69
E120
SMARCA4
SMARCB1-WH
R37
I28
Y126
ARID1A
G11
C330
C356
R350
C345
P365
W369
D346
ARID1B-ARID
DPF2-PHDs
SMARCE1-HMG
D1061N
M1067L
G1098R
W1059R
L1085P
R1089Q
G1099D
0
1
Recurrence
2
ARID1A-ARID
Variable
Average
Conserved
D1050E/Y
R1020T
K1021E
W1023R
Y1027H
L1049R
R1053H
L1054_Y1055delinsH
G1062V
Q1066H
K1072_W1073delinsR
R1074W
L1092F
F1103S
C298G/R/S
C276F
D346G/N
R350H
D274A
Q288H
A315V
W322R
C330W
C345G
D349N
C356G
P365R
W369R
Variable
Average
Conserved
Y73C/S
R105Q
P69Q
Y76C
K92N
W104S
E120K
Y126D
0
1
2
3
Recurrence
Conserved
Average
Variable
N/A
Variable
Average
Conserved
0
1
2
3
Recurrence
SMARCB1-WH
R37H
G11R
I28M
0 1 2 3 4 5 6 7
Recurrence
0
1
2
3
Recurrence
Fig. 5 | Comparison of NDD- and cancer-associated mutations in mSWI/SNF
complex components. a, Venn diagram overlapping unique cancer and NDD
missense and in-frame variants (left). Pie chart reflecting breakdown between
NDD- and cancer-associated mSWI/SNF missense and in-frame mutations (right).
The breakdown of recurrent and non-recurrent cancer variants is shown. b, Top
ten recurrent missense and in-frame indel mutations specific to NDD and those
shared between NDD and cancer, sorted by frequency in each disease type.
Inter- and intradomains are indicated. c, Heatmap representation of mutation
differences between NDD and cancer (NDD - Cancer normalized enrichment
scores (Methods)) reflected on the 3D structure of the human cBAF complex
(PDB:6LTJ). Red regions represent those enriched in NDD, blue represent those
enriched in cancer (−1, maximally enriched in cancer; 1, maximally enriched
in NDD). Labels for NDD hotspots are shown. d, Circos plot reflecting regions
of top-mutated mSWI/SNF subunits and the local enrichment of missense and
in-frame indel mutations in NDD (green), Cancer (red) or NDD-Cancer difference
(represented as NDD-Cancer NES): NDD (orange) or cancer (purple); interactions
between subunits, determined by cross-linking mass-spectrometry (CX-MS)
performed on endogenous cBAF complexes are shown (NCP-bound endogenous
cBAF, from Mashtalir et al.93). Scaled local recurrence, and NDD-Cancer NES
were calculated similarly to panel c with one exception, where all secondary
paralog mutations were preserved instead of remapping to paralogs. Enrichment
scores were bounded from 0 to 1 for local recurrence and −1 to 1 for differential
enrichment of mutations. Domains are represented as darker bands in the first
inner ring of the Circos plot. e, NDD-associated mutant residues emphasized as
red spheres on the structures of the ARID1A-ARID domain (PDB:1RYU), the DPF2-
PHD domain (PDB:5B79), the SMARCE1-HMG DNA-binding domain (PDB:7CYU)
and the SMARCB1-winged-helix DNA-binding domain (PDB:6LTJ). NDD-
associated missense and inframeshift variants, ranked by frequency, are shown
as bar charts. ConSurf conservation scores are mapped onto each domain
structure with cyan-white-magenta color scale in increasing conservation order.
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1407
Analysishttps://doi.org/10.1038/s41588-023-01451-6ARID1AARID1BACTBACTL6AACTL6BSMARCA2SMARCA4SMARCB1SMARCE1
Canonical BAF (cBAF)
NDD mutations methods of disruption
Perturbations to
catalytic activity &
helicase DNA binding
NDD phenotype: severe
Perturbations to
nucleosome engagement
NDD phenotype: severe
ARP module
disruptions
NDD phenotype: mild-moderate
SMARCA2/4
SMARCB1
ACTB, ACTL6A/B
Nucleosome
acidic patch
SMARCB1
C-term
DPF
ATPase
module
ARP
module
Disruption of
histone tail binding
NDD phenotype: mild-moderate
DPF2
PHD
domain
ARID
ARM
BAF
core
module
Mutations in
DNA binding domains
NDD phenotype: mild-severe
DPF2-PHD
DPF1/3-PHD?
PHF10-PHD?
PBRM1-Bromo?
BRD7/9-Bromo?
NDD phenotype
Severe
Moderate
Mild
ARID1A/B-ARID domain
SMARCB1-WH domain
SMARCE1-HMG domain
Fig. 6 | Summary of widely disrupted mSWI/SNF complex hubs in NDDs.
NDD-associated mSWI/SNF mutations occur across several subunits of the
mSWI/SNF family of chromatin remodeling complexes and cluster in key
structural hubs. Missense and in-frame deletions accumulate within the catalytic
ATPase, nucleosome interacting, histone-binding or DNA-binding domains,
as well as the ARP module, underscoring their convergence in producing
neurodevelopmental aberrations. Interpretation of NDD-associated variants
in the context of this framework enables mechanistic dissection of mSWI/SNF
activities and provides functional links relevant to clinical phenotypes.
remodeling activity without disrupting ATPase consumption102. Muta-
tion of additional nucleosome DNA-binding residues including K878,
R1164 and R1142 (corresponding NDD variants: A4-K865E, A4-R1157Q/G,
A2-R1105H/G/C/P/S) may have similar biochemical outcomes (Extended
Data Fig. 4i). However, NDD-mutant residues in the ATP binding pocket
are expected to disrupt the fundamental ATPase activity of SNF2. For
example, mutation of either G797 or G795 (corresponding NDD vari-
ants A2-754A, A2-G752A and A4-G784R) residues, which provide space
for ATP to bind to the ATP pocket, may reduce mSWI/SNF nucleosome
remodeling activity (Extended Data Fig. 4i). Further work is required
to define how mutations might impact the dynamic activity of these
complexes as well as fully characterizing the structural domains not
yet resolved in SMARCA2/4.
Comparing cancer and NDD mutations reveals disruption hubs
Previous studies have examined the distribution of cancer-associated
single-residue mutations on the cBAF complex structure93,94,97. For our
analysis, we examined the overlap of unique missense and inframeshift
mutations identified in the context of NDD with those in human cancer
(cBioPortal-PanCancer103,104, AACR Project GENIE105 and COSMIC86)
(Extended Data Fig. 5a). We found that the majority (58.3%) of unique
mutations found in NDD were specific to NDD (Fig. 5a, Supplemen-
tary Table 4). Further, among the 41.6% of shared cancer mutations,
16.4% were found to be recurrent among the three cancer datasets
analyzed (Fig. 5a). Shared recurrent mutations in both NDD and can-
cer included those localized to the C-terminal domain of SMARCB1,
the SMARCA4 N terminus, as well as within PBRM1 and ACTB sub-
units (Fig. 5b and Supplementary Table 4). By examining mutational
positions rather than unique mutations, we found that over two
thirds (69.3%) of NDD-mutant positions are also altered in cancer, with
similar breakdown of the shared mutational recurrence (Extended Data
Fig. 5b,c). Given the difficulty of de-duplicating cancer variants across
the three cancer databases used in this study (cBioPortal PanCan/GENIE
and COSMIC datasets), we used the cumulative recurrence across the
three datasets for comparison to NDD recurrence (Fig. 5b, Extended
Data Fig. 5c,d and Supplementary Table 4).
A minor positive correlation was observed between the recur-
rence of shared cancer (cBioPortal-PanCan) and NDD sequence vari-
ants (Extended Data Fig. 5e). Although normalization of both NDD
and cancer mutational frequencies can mask regions highly mutated
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1408
Analysishttps://doi.org/10.1038/s41588-023-01451-6 in both disease settings, mutational enrichment analyses revealed
several unique mutational hot spots specific to human NDD (Fig. 5c).
Mutations in Arp module subunits, ACTB and ACTL6A/B, were nearly
selectively enriched in NDDs, whereas mutations in the helicase domain
of SMARCA4 were more enriched in cancer (Fig. 5b,c). Mutations over-
lapping with those in cancer localize to the SMARCA4 ATP binding
pocket and nucleosomal DNA-binding residues, the SMARCB1-CTD,
and the SMARCA4-BAF core module entry point (Fig. 5b and Extended
Data Fig. 5f). Finally, we used cross-linking mass spectrometry (CX-MS)
datasets from previous studies performed on endogenous cBAF
complexes16, which further demonstrated region-specific enrich-
ment of NDD-versus cancer-associated mutations throughout cBAF
subunits (Fig. 5d).
Mutations in structurally and functionally elusive domains
To date, 3D structural studies have resolved only ~44% of the total
cBAF complex (by molecular weight), owing to the presence of
low-complexity or disordered regions within many subunits (with
to-date unassigned functions). Further, and given that such regions are
often spaced between structured domains, several structured domains,
many solved in isolation, have not been solved in the context of full 3D
cBAF or PBAF complexes. We thus mapped all NDD non-truncating
variants to the highly mutated ARID1A-ARID DNA-binding domain, the
SMARCE1-HMG domain, the DPF2-PHD domains and the SMARCB1-WH
domain to previously resolved high-resolution apo structures106–109
(Fig. 5e and Extended Data Fig. 5g–j). Intriguingly, the majority of
ARID1A-ARID domain and SMARCB1 WH domain non-truncating vari-
ants do not overlap with the DNA-binding residues, and we therefore
predict that they disrupt intradomain structural integrity (Fig. 5e and
Extended Data Fig. 5g–h)108. As has been demonstrated previously,
mutations in the DPF2-PHD domains disrupt zinc-binding residues
which are important for PHD domain structural formation, resulting
in decreased affinity to modified histone substrates (Fig. 5e and
Extended Data Fig. 5i)109. NDD-associated mutations in the SMARCE1-
HMG domain accumulate on the DNA-binding interface of the
structure (Fig. 5e and Extended Data Fig. 5j)107 and hence are predicted
to inhibit DNA binding.
Discussion
Here, we demonstrate that mSWI/SNF complex genes are the most
frequently disrupted chromatin regulatory entity in NDD, with
perturbation of several key structural ‘hubs’ within this multicompo-
nent complex displaying a phenotypic convergence that yields NDD
features associated in the literature with the greatest level of NDD
severity (Fig. 1d and Fig. 6). Our study serves as a powerful founda-
tion upon which to pursue integrated efforts between the chromatin
biology and neurobiology communities to functionally characterize
and prioritize these frequent disruptions.
It should be noted that because the products of mSWI/SNF
complex genes are assembled into a highly heterogeneous group of
complexes, the total extent of mutational burden of this complex
reported here may not be completely recognized, even with genes such
as ARID1B ranking among the most highly mutated in NDDs (Extended
Data Fig. 1n)110–112. Disruption of both structured and unstructured
domains presented here may impart altered mSWI/SNF complex
localization and activity on the genome via a range of mechanisms
requiring extensive further investigation. Additionally, further exami-
nation of zygosity and how missense variants within the same protein
differentially impact protein activity may reveal distinct functions.
For example, both dominant and recessive single amino acid variants
affected ACTL6B have been identified113. Although the ACTL6B G393R
recessive variant has been shown to reduce ACTL6B protein expression,
behaving as a loss-of-function mutation114, the dominant G343R vari-
ant is predicted to impart dominant-negative effects that disrupting
mSWI/SNF activity42.
In this study, we curated a list of chromatin regulatory genes in
combination with the EpiFactor database to investigate the preva-
lence of chromatin-related process disruptions in NDD. However, addi-
tional work is needed to define a maximally complete set of chromatin
regulators, regulatory complexes and their subunit membership.
Further, functional studies must be performed to define mecha-
nisms by which variants alter activity or other functions, especially
given that 3D structures are based on a range of complex states and
conformations, which may vary in biologic relevance. Importantly,
although we have obtained information on recurrence of sequence
variants for which distinct cases were clear, potential duplicates were
omitted in processing in cases for which we could not verify distinct
cases between literature and databases used, meaning that recurrence
of some variants may be artificially reduced. Further, cross-referencing
of additional private databases such as FoundationCORE may be
useful in follow-up analyses115. To prevent inclusion of false positives,
we omitted NDD-associated mSWI/SNF sequence variants which
are also present in gnomAD with a minor allele frequency of
>0.5%, predicted to be benign. Although the overwhelming major-
ity (96%) of DECIPHER variants reported to date are heterozygous
(Extended Data Fig. 1l), zygosity data were not included in this
study, and this remains a limitation. By centering the majority of
our analysis on de novo variants, we expect these to be patho-
genic; however, future studies must be performed to assess the full
scope of the molecular and pathophysiological consequences of
these mutations.
Online content
Any methods, additional references, Nature Portfolio reporting sum-
maries, source data, extended data, supplementary information,
acknowledgements, peer review information; details of author contri-
butions and competing interests; and statements of data and code avail-
ability are available at https://doi.org/10.1038/s41588-023-01451-6.
References
1. Bailey, M. H. et al. Comprehensive characterization of cancer
driver genes and mutations. Cell 173, 371–385 (2018).
2. Gabriele, M., Tobon, A. L., D’Agostino, G. & Testa, G. The
chromatin basis of neurodevelopmental disorders: Rethinking
dysfunction along the molecular and temporal axes. Prog.
Neuropsychopharmacol. Biol. Psychiatry 84, 306–327 (2018).
3. Rubeis, S. D. et al. Synaptic, transcriptional and chromatin genes
disrupted in autism. Nature 515, 209–215 (2014).
4. Satterstrom, F. K. et al. Large-scale exome sequencing study
implicates both developmental and functional changes in the
neurobiology of autism. Cell 180, 568–584 (2020).
5. Clapier, C. R., Iwasa, J., Cairns, B. R. & Peterson, C. L.
Mechanisms of action and regulation of ATP-dependent
chromatin-remodelling complexes. Nat. Rev. Mol. Cell Biol. 18,
407–422 (2017).
6. Sundaramoorthy, R. & Owen-Hughes, T. Chromatin remodelling
comes into focus. F1000Res. 9, Faculty Rev-1011 (2020).
7. Valencia, A. M. & Kadoch, C. Chromatin regulatory mechanisms
and therapeutic opportunities in cancer. Nat. Cell Biol. 21, 152–161
(2019).
8. Kadoch, C. & Crabtree, G. R. Mammalian SWI/SNF chromatin
remodeling complexes and cancer: Mechanistic insights gained
from human genomics. Sci. Adv. 1, e1500447 (2015).
9. Kadoch, C. et al. Proteomic and bioinformatic analysis of
mammalian SWI/SNF complexes identifies extensive roles in
human malignancy. Nat. Genet. 45, 592–601 (2013).
10. Maulik, P. K., Mascarenhas, M. N., Mathers, C. D., Dua, T. &
Saxena, S. Prevalence of intellectual disability: A meta-analysis
of population-based studies. Res. Dev. Disabil. 32, 419–436
(2011).
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1409
Analysishttps://doi.org/10.1038/s41588-023-01451-6 11. Robertson, J., Hatton, C., Emerson, E. & Baines, S. Prevalence of
epilepsy among people with intellectual disabilities: A systematic
review. Seizure 29, 46–62 (2015).
33. Kaplanis, J. et al. Evidence for 28 genetic disorders discovered by
combining healthcare and research data. Nature 586, 757–762
(2020).
12. Kleefstra, T., Schenck, A., Kramer, J. M. & van Bokhoven, H. The
34. McRae, J. F. et al. Prevalence and architecture of de novo
genetics of cognitive epigenetics. Neuropharmacology 80, 83–94
(2014).
mutations in developmental disorders. Nature 542, 433–438
(2017).
13. Ronan, J. L., Wu, W. & Crabtree, G. R. From neural development to
cognition: Unexpected roles for chromatin. Nat. Rev. Genet. 14,
347–359 (2013).
35. Lelieveld, S. H. et al. Meta-analysis of 2,104 trios provides support
for 10 new genes for intellectual disability. Nat. Neurosci. 19,
1194–1196 (2016).
14. Valencia, A. M. & Pașca, S. P. Chromatin dynamics in human
brain development and disease. Trends Cell Biol. 32, 98–101
(2021).
36. Ligt, Jde et al. Diagnostic exome sequencing in persons with
severe intellectual disability. N. Engl. J. Med. 367, 1921–1929
(2012).
15. Kelso, T. W. R. et al. Chromatin accessibility underlies synthetic
lethality of SWI/SNF subunits in ARID1A-mutant cancers. Elife 6,
e30506 (2017).
37. Sanders, S. J. et al. De novo mutations revealed by whole-exome
sequencing are strongly associated with autism. Nature 485,
237–241 (2012).
16. Mashtalir, N. et al. Modular organization and assembly of SWI/SNF
family chromatin remodeling complexes. Cell 175, 1272–1288.e20
(2018).
38. Rauch, A. et al. Range of genetic mutations associated with
severe non-syndromic sporadic intellectual disability: an exome
sequencing study. Lancet 380, 1674–1682 (2012).
17. Nakayama, R. T. et al. SMARCB1 is required for widespread
39. Sanders, S. J. et al. Insights into autism spectrum disorder
BAF complex-mediated activation of enhancers and bivalent
promoters. Nat. Genet. 49, 1613–1623 (2017).
genomic architecture and biology from 71 risk loci. Neuron 87,
1215–1233 (2015).
18. Pan, J. et al. The ATPase module of mammalian SWI/SNF
family complexes mediates subcomplex identity and catalytic
activity-independent genomic targeting. Nat. Genet. 51, 618–626
(2019).
19. Santen, G. W. E. et al. Mutations in SWI/SNF chromatin
remodeling complex gene ARID1B cause Coffin-Siris syndrome.
Nat. Genet. 44, 379–380 (2012).
40. Alvarez-Mora, M. I. et al. Comprehensive molecular testing
in patients with high functioning autism spectrum disorder.
Mutat. Res Fundam. Mol. Mech. Mutagen 784, 46–52 (2016).
41. Aref-Eshghi, E. et al. BAFopathies’ DNA methylation epi-signatures
demonstrate diagnostic utility and functional continuum of
Coffin–Siris and Nicolaides–Baraitser syndromes. Nat. Commun.
9, 4885 (2018).
20. Tsurusaki, Y. et al. Mutations affecting components of the
42. Bell, S. et al. Mutations in ACTL6B Cause Neurodevelopmental
SWI/SNF complex cause Coffin-Siris syndrome. Nat. Genet. 44,
376–378 (2012).
Deficits and Epilepsy and Lead to Loss of Dendrites in Human
Neurons. Am. J. Hum. Genet. 104, 815–834 (2019).
21. Wieczorek, D. et al. A comprehensive molecular study on
43. Bowling, K. M. et al. Genomic diagnosis for children with
Coffin–Siris and Nicolaides–Baraitser syndromes identifies
a broad molecular and clinical spectrum converging on
altered chromatin remodeling. Hum. Mol. Genet. 22, 5121–5135
(2013).
22. Firth, H. V. et al. DECIPHER: Database of chromosomal imbalance
and phenotype in humans using Ensembl resources. Am. J. Hum.
Genet. 84, 524–533 (2009).
23. Landrum, M. J. et al. ClinVar: improving access to variant
interpretations and supporting evidence. Nucleic Acids Res. 46,
D1062–D1067 (2018).
24. Fokkema, I. F. A. C. et al. LOVD v.2.0: the next generation in gene
variant databases. Hum. Mutat. 32, 557–563 (2011).
25. An, J.-Y. et al. Genome-wide de novo risk score implicates
promoter variation in autism spectrum disorder. Science 362,
eaat6576 (2018).
intellectual disability and/or developmental delay. Genome Med.
9, 43 (2017).
44. Yuen, R. K. C. et al. Whole genome sequencing resource
identifies 18 new candidate genes for autism spectrum disorder.
Nat. Neurosci. 20, 602–611 (2017).
45. Campeau, P. M. & Hennekam, R. C., group, T. D. syndrome
collaborative. DOORS syndrome: Phenotype, genotype and
comparison with Coffin‐Siris syndrome. Am. J. Med. Genet. Part C.
Semin. Med. Genet. 166, 327–332 (2014).
46. Chérot, E. et al. Using medical exome sequencing to identify the
causes of neurodevelopmental disorders: Experience of 2 clinical
units and 216 patients. Clin. Genet. 93, 567–576 (2018).
47. D’Gama, A. M. et al. Targeted DNA Sequencing from Autism
Spectrum Disorder Brains Implicates Multiple Genetic
Mechanisms. Neuron 88, 910–917 (2015).
26. Appenzeller, S. et al. De novo mutations in synaptic transmission
genes including DNM1 cause epileptic encephalopathies. Am. J.
Hum. Genet. 95, 360–370 (2014).
27. Gilissen, C. et al. Genome sequencing identifies major causes of
48. Diets, I. J. et al. A recurrent de novo missense pathogenic variant
in SMARCB1 causes severe intellectual disability and choroid
plexus hyperplasia with resultant hydrocephalus. Genet. Med. 21,
572–579 (2019).
severe intellectual disability. Nature 511, 344–347 (2014).
28. Halldorsson, B. V. et al. Characterizing mutagenic effects of
recombination through a sequence-level genetic map. Science
363, eaau1043 (2019).
29. Homsy, J. et al. De novo mutations in congenital heart disease
with neurodevelopmental and other congenital anomalies.
Science 350, 1262–1266 (2015).
30. Iossifov, I. et al. De novo gene disruptions in children on the
31.
autistic spectrum. Neuron 74, 285–299 (2012).
Iossifov, I. et al. The contribution of de novo coding mutations to
autism spectrum disorder. Nature 515, 216–221 (2014).
49. Doan, R. N. et al. Mutations in Human Accelerated Regions Disrupt
Cognition and Social Behavior. Cell 167, 341–354.e12 (2016).
50. Farwell, K. D. et al. Enhanced utility of family-centered diagnostic
exome sequencing with inheritance model-based analysis:
results from 500 unselected families with undiagnosed genetic
conditions. Genet. Med. 17, 578–586 (2015).
51. Fichera, M. et al. Mutations in ACTL6B, coding for a subunit of
the neuron-specific chromatin remodeling complex nBAF, cause
early onset severe developmental and epileptic encephalopathy
with brain hypomyelination and cerebellar atrophy. Hum. Genet.
138, 187–198 (2019).
32. Jin, S. C. et al. Contribution of rare inherited and de novo variants
in 2,871 congenital heart disease probands. Nat. Genet 49,
1593–1601 (2017).
52. Geisheker, M. R. et al. Hotspots of missense mutation identify
novel neurodevelopmental disorder genes and functional
domains. Nat. Neurosci. 20, 1043–1051 (2017).
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1410
Analysishttps://doi.org/10.1038/s41588-023-01451-6 53. Guo, H. et al. Genome sequencing identifies multiple deleterious
variants in autism patients with more severe phenotypes.
Genet. Med. 21, 1611–1620 (2019).
54. Karaca, E. et al. Genes that affect brain structure and function
identified by rare variant analyses of Mendelian neurologic
disease. Neuron 88, 499–513 (2015).
74. Vasileiou, G. et al. Mutations in the BAF-complex subunit DPF2 are
associated with Coffin-Siris syndrome. Am. J. Hum. Genet. 102,
468–479 (2018).
75. Verloes, A. et al. Baraitser–Winter cerebrofrontofacial syndrome:
Delineation of the spectrum in 42 cases. Eur. J. Hum. Genet. 23,
292–301 (2015).
55. Kleefstra, T. et al. Disruption of an EHMT1-associated
76. Vissers, L. E. L. M. et al. A clinical utility study of exome
chromatin-modification module causes intellectual disability. Am.
J. Hum. Genet. 91, 73–82 (2012).
sequencing versus conventional genetic testing in pediatric
neurology. Genet. Med. 19, 1055–1063 (2017).
56. Koga, M. et al. Involvement of SMARCA2/BRM in the SWI/SNF
chromatin-remodeling complex in schizophrenia. Hum. Mol.
Genet 18, 2483–2494 (2009).
57. Krumm, N. et al. Excess of rare, inherited truncating mutations in
autism. Nat. Genet. 47, 582–588 (2015).
58. Lecoquierre, F. et al. Variant recurrence in neurodevelopmental
disorders: the use of publicly available genomic data identifies
clinically relevant pathogenic missense variants. Genet. Med. 21,
2504–2511 (2019).
59. Li, J. et al. Targeted sequencing and functional analysis reveal
brain-size-related genes and their networks in autism spectrum
disorders. Mol. Psychiatr. 22, 1282–1290 (2017).
60. Consortium, A. S. et al. Rates, distribution and implications of
postzygotic mosaic mutations in autism spectrum disorder. Nat.
Neurosci. 20, 1217–1224 (2017).
61. Machol, K. et al. Expanding the spectrum of BAF-related
disorders: De novo variants in SMARCC2 cause a syndrome with
intellectual disability and developmental delay. Am. J. Hum.
Genet. 104, 164–178 (2019).
62. Mannino, E. A., Miyawaki, H., Santen, G. & Vergano, S. A. S.
First data from a parent‐reported registry of 81 individuals
with Coffin–Siris syndrome: Natural history and management
recommendations. Am. J. Med. Genet. A 176, 2250–2258 (2018).
63. Marom, R. et al. Heterozygous variants in ACTL6A, encoding a
component of the BAF complex, are associated with intellectual
disability. Hum. Mutat. 38, 1365–1371 (2017).
64. Mignot, C. et al. ARID1B mutations are the major genetic cause of
corpus callosum anomalies in patients with intellectual disability.
Brain 139, e64 (2016).
65. Monies, D. et al. Lessons learned from large-scale, first-tier clinical
exome sequencing in a highly consanguineous population. Am. J.
Hum. Genet. 104, 1182–1201 (2019).
66. Nixon, K. C. J. et al. A syndromic neurodevelopmental disorder
caused by mutations in SMARCD1, a core SWI/SNF subunit
needed for context-dependent neuronal gene regulation in flies.
Am. J. Hum. Genet. 104, 596–610 (2019).
67. Pascolini, G., Agolini, E., Novelli, A., Majore, S. & Grammatico, P.
The p.Arg377Trp variant in ACTL6A underlines a recognizable
BAF‐opathy phenotype. Clin. Genet 97, 672–674 (2020).
68. Sandestig, A. et al. Could dissimilar phenotypic effects of ACTB
missense mutations reflect the actin conformational change
two novel mutations and literature review. Mol. Syndromol. 9,
259–265 (2019).
69. Santen, G. W. E. et al. Coffin–Siris syndrome and the BAF complex:
Genotype–phenotype study in 63 patients. Hum. Mutat. 34,
1519–1528 (2013).
70. Sekiguchi, F. et al. Genetic abnormalities in a large cohort of
Coffin–Siris syndrome patients. J. Hum. Genet. 64, 1173–1186 (2019).
71. Tsurusaki, Y. et al. Coffin–Siris syndrome is a SWI/SNF complex
77. Wolff, D. et al. In-frame deletion and missense mutations of the
C-terminal helicase domain of SMARCA2 in three patients with
Nicolaides-Baraitser syndrome. Mol. Syndromol. 2, 237–244
(2012).
78. Wu, H. et al. Phenotype‐to‐genotype approach reveals head‐
circumference‐associated genes in an autism spectrum disorder
cohort. Clin. Genet 97, 338–346 (2020).
79. Xiong, J. et al. Neurological diseases with autism spectrum
disorder: Role of ASD risk genes. Front. Neurosci. 13, 349
(2019).
80. Yu, Y. et al. De novo mutations in ARID1B associated with both
syndromic and non-syndromic short stature. Bmc Genomics 16,
701 (2015).
81. Zhao, J. J. et al. Exome sequencing reveals NAA15 and PUF60 as
candidate genes associated with intellectual disability. Am. J.
Med. Genet. 177, 10–20 (2018).
82. Medvedeva, Y. A. et al. EpiFactors: A comprehensive database
of human epigenetic factors and complexes. Database (Oxford)
2015, bav067 (2015).
83. Koopmans, F. et al. SynGO: An evidence-based, expert-curated
knowledge base for the synapse. Neuron 103, 217–234 (2019).
84. Abrahams, B. S. et al. SFARI Gene 2.0: A community-driven
knowledgebase for the autism spectrum disorders (ASDs).
Mol. Autism 4, 36 (2013).
85. Bögershausen, N. & Wollnik, B. Mutational landscapes and
phenotypic spectrum of SWI/SNF-related intellectual disability
disorders. Front Mol. Neurosci. 11, 252 (2018).
86. Tate, J. G. et al. COSMIC: The catalogue of somatic mutations In
cancer. Nucleic Acids Res. 47, D941–D947 (2019).
87. Karczewski, K. J. et al. The mutational constraint spectrum
quantified from variation in 141,456 humans. Nature 581, 434–443
(2020).
88. Cappuccio, G. et al. De novo SMARCA2 variants clustered
outside the helicase domain cause a new recognizable syndrome
with intellectual disability and blepharophimosis distinct from
Nicolaides–Baraitser syndrome. Genet. Med. 22, 1838–1850
(2020).
89. Sluijs, P. Jvander et al. Discovering a new part of the phenotypic
spectrum of Coffin-Siris syndrome in a fetal cohort. Genet. Med.
24, 1753–1760 (2022).
90. van der Sluijs, P. J. et al. A case series of familial ARID1B variants
illustrating variable expression and suggestions to update the
ACMG criteria. Genes (Basel) 12, 1275 (2021).
91. Milone, R., Gnazzo, M., Stefanutti, E., Serafin, D. & Novelli,
A. A new missense mutation in DPF2 gene related to Coffin
Siris syndrome 7: Description of a mild phenotype expanding
DPF2-related clinical spectrum and differential diagnosis among
similar syndromes epigenetically determined. Brain Dev. 42,
192–198 (2019).
disorder. Clin. Genet. 85, 548–554 (2014).
92. Adzhubei, I. A. et al. A method and server for predicting damaging
72. Zarate, Y. A. et al. SMARCE1, a rare cause of Coffin–Siris syndrome:
Clinical description of three additional cases. Am. J. Med Genet. A
170, 1967–1973 (2016).
73. Houdt, J. K. J. V. et al. Heterozygous missense mutations in
SMARCA2 cause Nicolaides-Baraitser syndrome. Nat. Genet. 44,
445–449 (2012).
missense mutations. Nat. Methods 7, 248–249 (2010).
93. Mashtalir, N. et al. A structural model of the endogenous human
BAF complex informs disease mechanisms. Cell 183, 802–817.e24
(2020).
94. He, S. et al. Structure of nucleosome-bound human BAF complex.
Science 367, 875–881 (2020).
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1411
Analysishttps://doi.org/10.1038/s41588-023-01451-6 95. Donato, N. D. et al. Severe forms of Baraitser–Winter syndrome
are caused by ACTB mutations rather than ACTG1 mutations.
Eur. J. Hum. Genet. 22, 179–183 (2014).
96. Ittisoponpisan, S. et al. Can predicted protein 3D structures
provide reliable insights into whether missense variants are
disease associated? J. Mol. Biol. 431, 2197–2212 (2019).
97. Yuan, J., Chen, K., Zhang, W. & Chen, Z. Structure of human
chromatin-remodelling PBAF complex bound to a nucleosome.
Nature 605, 166–171 (2022).
98. Valencia, A. M. et al. Recurrent SMARCB1 mutations reveal a
nucleosome acidic patch interaction site that potentiates
mSWI/SNF complex chromatin remodeling. Cell 179, 1342–1356.
e23 (2019).
109. Xiong, X. et al. Selective recognition of histone crotonylation
by double PHD fingers of MOZ and DPF2. Nat. Chem. Biol. 12,
1111–1118 (2016).
110. Hoyer, J. et al. Haploinsufficiency of ARID1B, a member of the
SWI/SNF-A chromatin-remodeling complex, is a frequent
cause of intellectual disability. Am. J. Hum. Genet. 90, 565–572
(2012).
111. Sluijs, P. Jvander et al. The ARID1B spectrum in 143 patients: From
nonsyndromic intellectual disability to Coffin–Siris syndrome.
Genet. Med. 21, 1295–1307 (2019).
112. Wright, C. F. et al. Genetic diagnosis of developmental disorders
in the DDD study: a scalable analysis of genome-wide research
data. Lancet Lond. Engl. 385, 1305–1314 (2015).
99. Khanna, T., Hanna, G., Sternberg, M. J. E. & David, A. Missense3D-DB
web catalogue: An atom-based analysis and repository of 4M human
protein-coding genetic variants. Hum. Genet. 140, 805–812 (2021).
113. Rowland, M. E., Jajarmi, J. M., Osborne, T. S. M. & Ciernia, A. V.
Insights Into the emerging role of Baf53b in autism spectrum
disorder. Front. Mol. Neurosci. 15, 805158 (2022).
100. Han, Y., Reyes, A. A., Malik, S. & He, Y. Cryo-EM structure of
114. Wenderski, W. et al. Loss of the neural-specific BAF subunit
SWI/SNF chromatin remodeling complex with nucleosome.
Nature 579, 452–455 (2020).
101. Liu, X., Li, M., Xia, X., Li, X. & Chen, Z. Mechanism of chromatin
remodelling revealed by the Snf2-nucleosome structure. Nature
544, 440–445 (2017).
102. Li, M. et al. Mechanism of DNA translocation underlying
chromatin remodelling by Snf2. Nature 567, 409–413 (2019).
103. Gao, J. et al. Integrative analysis of complex cancer genomics and
clinical profiles using the cBioPortal. Sci. Signal 6, pl1 (2013).
104. Cerami, E. et al. The cBio Cancer Genomics Portal: An open
platform for exploring multidimensional cancer genomics data.
Cancer Discov. 2, 401–404 (2012).
105. The AACR Project GENIE Consortium. AACR Project GENIE:
Powering precision medicine through an international
consortium. Cancer Discov. 7, 818–831 (2017).
106. Allen, M. D., Freund, S. M. V., Zinzalla, G. & Bycroft, M. The
SWI/SNF subunit INI1 contains an N-terminal winged helix DNA
binding domain thatis a target for mutations in schwannomatosis.
Struct. Lond. Engl. 1993, 1344–1349 (2015).
107. Heo, Y. et al. Crystal structure of the HMG domain of human
BAF57 and its interaction with four-way junction DNA. Biochem.
Biophys. Res. Commun. 533, 919–924 (2020).
108. Kim, S., Zhang, Z., Upchurch, S., Isern, N. & Chen, Y. Structure and
DNA-binding sites of the SWI1 AT-rich interaction domain (ARID)
suggest determinants for sequence-specific DNA recognition.
J. Biol. Chem. 279, 16670–16676 (2004).
ACTL6B relieves repression of early response genes and causes
recessive autism. Proc. Natl Acad. Sci. USA 117, 10055–10066
(2020).
115. Hartmaier, R. J. et al. High-throughput genomic profiling of adult
solid tumors reveals novel insights into cancer pathogenesis.
Cancer Res. 77, 2464–2475 (2017).
Publisher’s note Springer Nature remains neutral with regard
to jurisdictional claims in published maps and institutional
affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format,
as long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate
if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons license and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this license, visit http://creativecommons.
org/licenses/by/4.0/.
© The Author(s) 2023
Nature Genetics | Volume 55 | August 2023 | 1400–1412
1412
Analysishttps://doi.org/10.1038/s41588-023-01451-6 Methods
Novel variant collection
Novel NDD-related mSWI/SNF gene variants reported in this study were
identified through physician referrals and the Coffin-Siris syndrome
registry. Variants from Leiden University Medical Center were identi-
fied in a diagnostic setting, and genetic data were retrieved from the
generated reports or shared with us by the treating physician with
consent from the patient or parents. The institutional review board
of Leiden University Medical Center provided approval waivers for
using de-identified data and publishing aggregated data (G18.098
and G21.129) without obtaining specific informed consent. Individuals
identified through Eastern Virginia Medical School were recruited to
the Coffin-Siris syndrome registry through clinicians, social media and
patient foundations. Individuals completed an online consent form fol-
lowed by a registry survey with phenotypic questions. The Coffin-Siris
Syndrome Registry has been approved by the Eastern Virginia Medical
School institutional review board (15-03-EX-0058). Novel variants
reported in this study have been deposited in LOVD (https://www.
lovd.nl/)24. Variants identified through this method that were present
in previously published literature or deposited in an online repository
were excluded for analysis in this study to prevent reporting potential
duplicates (Curating mSWI/SNF gene NDD-associated variants section).
Given that our paper centers on the mutational rather than phenotypic
outcomes of NDD-related mSWI/SNF variants, future clinical papers
will further explore the phenotypes associated with novel variants
published in this manuscript. During the review process, some novel
variants included in this study were published with detailed clinical
information89.
Mutational datasets
Open-access mutations publicly available on the DECIPHER database
(https://www.deciphergenomics.org/; accessed June 22, 2022) (ref. 22)
were used for broader chromatin gene analysis (Fig. 1c,d and Extended
Data Fig. 1k,l). The queried chromatin remodeling complex gene list
(SWI/SNF, CHD, INO80 and ISWI) was manually curated from a literature
review detailed below (Supplementary Table 2).
Chromatin regulatory gene sets (Supplementary Table 2)
Chromatin remodeling complex gene lists were curated from a variety
of sources, including HGNC gene groups SWI/SNF and INO80 (https://
www.genenames.org/data/genegroup/#!/), as well as a literature review
of all chromatin remodeling complexes116,117, mSWI/SNF16, ISWI118, CHD119
and INO80 (refs. 120–124). The histone modifier gene list was gathered
from HISTome2 (refs. 125,126) (http://www.actrec.gov.in/histome2/).
Polycomb repressive complex genes and DNA methylation regulatory
genes were informed by the literature127,128. Additional chromatin regu-
latory complexes were obtained from EpiFactor82 (https://epifactors.
autosome.org/protein_complexes). The full set of cBAF, PBAF and
ncBAF genes were included in the EpiFactor complexes if absent.
Curating mSWI/SNF gene NDD-associated variants
The set of rare inherited and de novo variants included data from three
cohorts of individuals with autism spectrum disorders or other devel-
opmental disorders: the Simons SSC/ASC, SPARK and DDD cohorts.
Details about merging and de-duplicating the data are described in
Fu et al.129. Briefly, duplicated samples were identified and excluded
by IBD and other metadata, and the filtered samples were merged to
provide a single unified set of de-duplicated de novo variants in autism
spectrum disorders and other developmental disorders. The recur-
rence of NDD de novo variants across BAF genes and several gene sets
of interest, including a curated set of chromatin remodelers, epigenetic
modifiers and synaptic genes were visualized with scatter plots and bar
charts using matplotlib130. The set of de novo variants and non-benign
SNVs in DECIPHER were used for all summary calculations in Fig. 1 and
Extended Data Fig. 1 and for comparisons between the BAF genes,
Nature Genetics
chromatin regulatory genes, epigenetic modifier genes and synaptic
genes. The queried chromatin regulatory gene list was based on EpiFactor
(https://epifactors.autosome.ru/genes; accessed 2 September 2021)
(ref. 82 updated to include all mSWI/SNF genes (Supplementary Table 2).
The queried synaptic gene list was based on the SynGO gene list
(https://www.syngoportal.org/; accessed 2 September 2021) (ref. 83).
The development disorder DECIPHER gene list was based on DDG2P
genes in DECIPHER (accessed 13 June 2022).
A comprehensive list of SNV and short in-frame indels (inframeshift
variants) was compiled from an extensive literature review, the com-
bined set of rare inherited and de novo variants from the Simons
SSC/ASC, SPARK, and DDD cohorts (the ‘combined cohort study’),
the DECIPHER database of SNVs (https://www.deciphergenomics.
org/), the merged set of de novo mutations from the DNM effort by
McRae et al.34 NDD-associated ClinVar mutations (accessed 5/15/2021),
NDD-associated variants from LOVD (LOVD v3.0 accessed June 2022)
and 85 previously unreported cases published in this study collected
through the laboratories of S.A.S.V. (Eastern Virginia Medical School)
and G.W.E.S. (Leiden University Medical Center).
First, the combined set of rare inherited and de novo variants was
split into a set of rare inherited variants and a set of de novo variants.
All rare inherited PTVs, in-frame indel variants and de novo variants
were included in the integrated dataset. Guided by the analysis in
Fu et al.129, where missense variants with MPC scores (missense bad-
ness, PolyPhen-2 and constraint) of 1 or more were observed to confer
moderate to strong levels of risk in developing autism and missense
rare inherited variants with MPC scores ≥1 were included in the inte-
grated dataset. All other rare inherited variants from the combined
cohort study were excluded. Then, samples were cross-referenced
between the combined cohort study, DECIPHER database, and the
DNM cohort of de novo mutations and identical variants from the
same samples (using available sample IDs or aliases) were removed
to de-deduplicate the data between these three cohorts/databases.
Separately, a list of de novo variants in BAF genes across several other
studies in the literature not covered previously by the cohorts used in
DECIPHER and the combined cohort study (SSC/ASC, SPARK and DDD)
were manually curated and de-duplicated to form the compiled set of
mutations from the literature. Additionally, NDD-associated mutations
from the LOVD database were compiled and filtered to include all PTV
and in-frame indels and de novo/likely de novo missense variants. All
benign/likely benign variants were excluded. The filtered set of LOVD
variants and the manually curated variants from the literature were
merged and de-duplicated based on sample IDs or aliases (if avail-
able) and study ID / reference (if sample IDs were not available). For
shared variants between LOVD and the literature, where it was not
clear whether these variants were duplicates, only shared variants
from the manually curated literature dataset were kept, effectively
de-duplicating the data. Minimal overlap was assumed between the
de-duplicated set of LOVD/literature variants and the de-duplicated
set of SSC + ASC/SPARK/DDD/DECIPHER/DNM variants. These two
sets were merged, followed by a round of manual curation to double
check that as many duplicates or potential duplicates were removed
during dataset integration. The set of 85 novel cases identified by
S.A.S.V. and G.W.E.S. were added to this merged dataset. In parallel,
a curated set of ClinVar variants from samples with NDD-associated
clinical features and unknown/likely pathogenic/pathogenic clinical
significance was generated. Benign and likely benign ClinVar variants
were excluded. Additionally, ClinVar variants submitted by GeneDx
were excluded due to substantial overlap with the comprehensive
analysis of de novo mutations in NDD by Kaplanis et al. included in the
DNM database of de novo mutations. Samples were de-duplicated
between ClinVar and the LOVD/literature dataset using SCV codes wher-
ever available. Finally, this de-duplicated ClinVar dataset was used to
adjust the counts of the previously merged dataset of NDD-associated
BAF mutations from the combined cohort study (SSC/ASC, SPARK
Analysishttps://doi.org/10.1038/s41588-023-01451-6 and DDD), DECIPHER SNVs, DNM, LOVD and the literature. It was dif-
ficult (and sometimes impossible) to track, match and assign each fil-
tered NDD-associated ClinVar SCV (submitted record for each variant)
with the list of available sample IDs or aliases in the previously merged
dataset. Thus, the total counts for each variant were adjusted to the
total counts found in ClinVar (based on the number of submissions
for each variant using SCV IDs) to eliminate the possibility of double
counting if the ClinVar total count for a variant was more than the total
count from the previously merged dataset. This procedure assumes
submissions to ClinVar overlap entirely with the previously merged
dataset, so it is possible the new merged dataset containing ClinVar
variants might undercount some NDD-associated BAF variants. This
integrated dataset was compared to gnomAD v3.1.2 to remove potential
SNPs and other variants that occur frequently in a collection of healthy
individuals. A more stringent MAF threshold of ≥0.5% MAF was used
to exclude potentially common variants in the integrated dataset. This
final integrated dataset was manually checked once more to exclude
potential duplicates and likely benign variants before freezing for all
downstream analyses. A total of 2,539 NDD-associated BAF variants
are included in this dataset, including 85 novel cases and 72 previously
unreported variants.
To standardize the data, all variants were remapped to the UniProt
canonical BAF protein isoforms (see Supplementary Table 3), and
duplicates that could not be confirmed unique cases were removed.
Unless otherwise noted, remapping of all variants (both NDD variants
and cancer variants) to different isoforms was performed using
the Ensembl Variant Effect Predictor (VEP) online web server131.
gnomAD variants of the general population were derived from the
gnomAD v3 dataset (accessed 11 January 2021).
Cancer dataset cleaning and compilation
PanCancer datasets from TCGA and cBioPortal103,104 were cleaned and
compiled for all downstream analyses related to NDD versus cancer
comparisons.
The TCGA MC3 PanCancer dataset was used for NDD versus
cancer comparisons in Extended Data Fig. 1. Briefly, known SNPs were
removed and BAF gene mutations were remapped to the canonical
UniProt transcripts (Supplementary Table 3). Missense, nonsense
and frameshift mutations were included, and all other mutations
were excluded. This filtered set of mutations merged with the com-
bined cohort study of NDD-associated mutations from the combined
SSC/ASC, SPARK and DDD cohorts. Total cancer missense, frameshift
and nonsense mutational recurrence was log normalized, compared to
total de novo NDD-associated missense and PTV mutational recurrence
for each gene, and visualized as a scatterplot using matplotlib130, with
BAF genes indicated in red. The total proportion of NDD and Cancer
missense and PTV mutations across the BAF genes were visualized as
a grouped bar chart using matplotlib130.
Mutations across BAF genes from the curated set of nonredun-
dant studies in cBioPortal, the AACR Project GENIE (accessed through
cBioPortal) and COSMIC were compiled and filtered for NDD versus
cancer comparative analyses across the BAF genes. Briefly, the BAF
mutations were remapped to the UniProt canonical BAF protein
isoforms (Supplementary Table 3) using the Ensembl VEP online
web server131. Missense, frameshift, nonsense and in-frame indels
were included, and all other mutations were excluded. Additionally,
duplicate mutations in patients with multiple samples were excluded.
This filtered set of mutations from cBioPortal103,104 was used for down-
stream BAF cancer versus NDD comparative analyses.
NDD gene set enrichment analysis
A custom Perl132 script was used to determine the enrichment of GOMF
gene sets enriched in DDG2P genes, a list of genes known to be associ-
ated with developmental disorders. All BAF genes were added back to
DDG2P gene list if absent. Specifically, GOMF gene sets were overlapped
Nature Genetics
with DDG2P using gene symbols and a hypergeometric distribution test
(for example, statistical overrepresentation test) was used to evaluate
the significance (P value) of enrichment of each GOMF. Additionally,
the total and mean number of de novo missense and PTVs in ASD + DD
using the combined cohort study was calculated for the overlapping
genes (using gene symbols) between each GOMF gene set and DDG2P
genes. The enrichment of GOMFs in DDG2P genes were visualized as
scatterplots and ranked by significance (P value) and total de novo
missense and PTV mutational recurrence for the overlapping genes
(using gene symbols) with the top 10 GOMFs labeled. Additionally,
the top 50 most enriched GOMFs by statistical significance (P value)
were ranked by the mean number of de novo missense and PTVs in the
overlapping genes (using gene symbols) in the combined cohort study
and the mean number of non-benign DECIPHER SNVs in the overlap-
ping genes (using gene symbols) and visualized as scatter plots with
the top 25 GOMFs indicated.
Further, the top 50 most enriched GOMFs by significance (P value)
were categorized into five major groups and colored accordingly in the
scatter plots. Additionally, the total number of non-benign DECIPHER
SNVs for the overlapping genes (using gene symbols) in these five major
groups and chromatin remodeling complexes from the curated list of
chromatin regulators were visualized as a bar chart (GOMF chromatin
gene sets and chromatin regulatory complexes were merged into one
group).
The GOMFs gene sets were obtained from MSigDB v7.5.1 (GOMF
v7.5.1; https://www.gsea-msigdb.org/gsea/msigdb/). The ARID2,
BCL7A/C and BICRAL BAF genes were added to the chromatin
binding GOMF gene set.
Benign and likely benign SNVs in DECIPHER were excluded to cre-
ate the set of non-benign DECIPHER SNVs. The development disorder
DECIPHER gene list was based on DDG2P genes on DECIPHER (accessed
on 15 May 22).
NDD recurrence in chromatin regulatory complexes,
epigenetic modifiers and synaptic genes
Queried chromatin remodeling gene lists (Supplementary Table 2)
were used for all downstream analysis in Fig. 1/Extended Data Fig. 1.
The total number of de novo missense and PTVs in the combined
cohorts ASD + DD study (SSC/ASC, SPARK, and DDD) across a curated
list of chromatin regulators and EpiFactor complexes were visualized as
bar charts. The total number of de novo missense and PTVs in DD (DDD)
and ASD (SSC/ASC and SPARK) across EpiFactor complexes were visual-
ized separately as bar charts. The total number of de novo missense and
PTVs in ASD + DD for every gene was visualized as a scatter plot with BAF
genes indicated in red. The mean number of de novo missense and PTVs
in ASD + DD (SSC/ASC, SPARK, and DDD) across EpiFactor complexes
were visualized as a bar chart. Protein lengths were obtained from the
top reviewed UniProtKB accession for each gene. The total de novo
missense and PTVs in ASD + DD (SSC/ASC, SPARK, and DDD) for each
EpiFactor complex was divided by the total protein length of each
EpiFactor complex to obtain protein length-normalized NDD de novo
mutational recurrence (that is average number of de novo missense
and PTVs per residue in each EpiFactor complex). The protein
length-normalized de novo mutational recurrence for EpiFactor
complexes were visualized as a bar chart.
Benign and likely benign SNVs in DECIPHER were excluded to
create the set of non-benign DECIPHER SNVs. The mean number
of non-benign DECIPHER SNVs and de novo missense and PTVs in
ASD + DD across all EpiFactor complex genes, mSWI/SNF genes and
SynGO synaptic genes were visualized as bar charts. The total number
of non-benign DECIPHER SNVs across a curated list of chromatin regula-
tors were visualized as a bar chart.
All bar charts were created using matplotlib130, and mSWI/SNF and
cBAF, PBAF and ncBAF gene sets are indicated in red. Ensembl gene IDs
(ENSG IDs) were used to overlap genes, merge datasets, and calculate
Analysishttps://doi.org/10.1038/s41588-023-01451-6 the total or mean number of de novo missense and PTVs in ASD + DD
and non-benign DECIPHER SNVs for gene sets in the list of curated chro-
matin regulators and EpiFactor complexes (Supplementary Table 2).
Structure figures
The mapping of unique SNV and short in-frame insertion/deletion
mutations was visualized using PyMol (v2.4.0) (ref. 133). The struc-
tural models used for this study were the following: Recombinant
cBAF structure bound to nucleosome (PDB: 6LTJ), Endogenous cBAF
structure bound to nucleosome (PDBDEV: PDBDEV_00000056), PBAF
complex bound to nucleosome (7VDV), SNF2h (5X0Y), yeast SWI/SNF
(6UXW), ARID1A-ARID (1RYU), DPF2-PHD (5B79), SMARCE1-HMG (7CYU)
and SMARCB1-WH (6LTJ). Domain annotations were obtained from
the PFAM and the literature, and manually curated (Supplementary
Table 3).
PolyPhen mutational analysis
The PolyPhen HumVar92 model was used to predict the severity of each
missense mutation in the list of compiled NDD mutations. The number
of NDD missense mutations for each intradomain (within-domain) and
interdomain (between-domain) region was divided by the lengths of
these regions to calculate the average number of NDD missense muta-
tions per residue for each interdomain or intradomain region. The
PolyPhen HumVar predicted severity scores for each residue in each
interdomain and intradomain were summed and divided by the length
of each region to calculate the average PolyPhen HumVar predicted
severity score for each inter-domain and intra-domain region. The aver-
age predicted PolyPhen HumVar predicted severity score and average
number of NDD missense mutations were visualized as a scatter plot
with interdomain and intradomain status indicated by color. All BAF
genes were used for this analysis.
Conservation analysis
Conservation analysis was performed for the recombinant cBAF struc-
ture (PDB:6LTJ; SMARCA4, chain I and SMARCB1, chain M), and the
ARID1A-ARID (1RYU), DPF2-PHD (PDB: 5B79), and SMARCE-HMG (PDB:
7CYU) domains using the ConSurf Server (https://consurf.tau.ac.il/)134.
Briefly, Protein Data Bank (PDB) IDs were selected and run through
ConSurf analysis using standard parameters (HMMR search algorithm,
UNIREF-90 protein database, automatic homolog selection and MAFFT
multiple sequence alignment method). Once completed, amended PDB
files color coded by conservation were downloaded and instructions
to ‘create high resolution figures’ were followed as instructed by the
ConSurf server.
Pairwise alignment
Multiple sequence alignments of the SMARCA4-ATPase, SMARCB1-CTD,
ARID1A-ARID, SMARCB1-WH, DPF2-PHD and SMARCE1-HMG domains
with their respective homologous proteins were performed using
Geneious Prime (v2021.2.2) using standard parameters.
General
Unless otherwise noted, mutational counts, bar plots, heatmaps and
pie charts throughout were made using a combination of R (v4.1.1),
GraphPad Prism (v9.2.0) and matplotlib (v3.3.1), and seaborn.
ConSurf mutational analysis
Full-length FASTA sequences of the UniProt canonical transcript for
all mSWI/SNF genes were uploaded to the ConSurf server with default
parameters to obtain predicted conservation scores. The number of
missense and in-frame indel NDD mutations by gene and position
and the predicted ConSurf conservation score (negative-transformed
so that higher scores indicate more conserved residues) were
visualized as a scatter plot. All mSWI/SNF genes were used for this
analysis.
NDD domain mutation analysis
The proportion of NDD mutations from the compiled list (missense,
in-frame indels, frameshift and nonsense mutations) were summed for
each gene, domain and inter-domain regions (Supplementary Table 3).
The proportion of NDD mutations within domains (intradomain) and
between domains (interdomain) were visualized as a stacked bar plot.
Domains were defined by PFAM, UniProtKB, manual curation and
resolved structures.
NDD disorder analysis
The proportion of NDD mutations from the compiled list (missense,
in-frame indels, frameshift and nonsense mutations) falling within
disordered (defined by MobiDB-lite; Supplementary Table 3) and
structured regions were visualized as a stacked bar chart for individual
BAF genes and BAF genes as a whole c ol le ct ion.
2D schematics
The distribution of gnomAD (v3) missense SNPs were visualized as a
kernel density estimate plot using the seaborn kdeplot with default
parameters. The gnomAD (v3) missense mutations for SMARCA2,
SMARCA4, ARID1A, ARID1B, SMARCB1, SMARCE1 and DPF2 were used
to compute the missense recurrence by position across the length of
each protein, which was used as input into the kernel density estimate
analysis. The NDD compiled list of mutations (missense, in-frame
indels, frameshift and nonsense mutations) for the aforementioned
genes were visualized using the St. Jude PeCan Protein Paint software
with default settings (https://proteinpaint.stjude.org/). Special care
was taken to map the mutations on the canonical UniProt isoform
(Supplementary Table 3). Domains using the annotations compiled
from PFAM, InterPro and the literature, and manually curated based on
the AlphaFold EMBL-EBI structural predictions. ConSurf conservation
scores were visualized as horizontal bars using the ConSurf provided
‘COLOR’ column with an aggregation of scores (1, 2 or 3, cyan; 4, 5 or 6,
white; 7, 8 or 9, violet). The coverage of the two available recombinant
(PDB:6LTJ) and endogenous nucleosome-bound cBAF structures were
visualized as horizontal bars (recombinant coverage in orange, endo-
genous coverage in red and dual coverage in brown).
Missense DNA and protein changes
The frequencies of DNA point substitutions (all SNVs) and protein
amino acid substitutions (top 20) in the compiled NDD mutation data-
set (missense only) were visualized as bar plots. Additionally, the amino
acid substitutions for the missense subset of mutations in the com-
piled NDD mutation dataset was visualized as Sankey Diagram using
Google Charts. Additionally, these amino acid substitutions were aggre-
gated into functional changes (negative, positive, polar, nonpolar and
miscellaneous) and visualized as proportions in stacked bar charts.
Mappability of NDD mutations
The proportion of NDD mutations in the compiled NDD mutation
dataset (missense, in-frame indels, frameshift and nonsense mutations)
mappable across the endogenous and recombinant (PDB:6LTJ) were
visualized as a group bar plot (Supplementary Table 3).
NDD versus cancer overlap analysis
The recurrence of every unique gene-mutation combination for mis-
sense and in-frame indel mutations from the NDD compiled dataset
and the cBioPortal (accessed June 2022) cancer dataset was computed
and visualized as a pie chart or tables.
NDD versus cancer NESs and comparative analyses
The missense and in-frame indel mutations from the compiled NDD
mutation dataset and the cBioPortal cancer dataset were used to com-
pute the NDD and cancer mutation recurrence by position across
each BAF gene. This recurrence was scaled between 0 and 1 using the
Nature Genetics
Analysishttps://doi.org/10.1038/s41588-023-01451-6 MinMaxScaler preprocessing function in scikit-learn. The rescaled
mutation recurrence for cancer was subtracted from the rescaled
mutation recurrence for NDD to compute the NDD-Cancer normalized
enrichment scores (NESs). Specifically, cancer NESs were calculated
using a four-step process. First, paralogs were pairwise aligned to the
primary paralog, and mutations on conserved residues were remapped
from the secondary to the primary paralogs. Second, the mutational
recurrence by residue position of NDD- and cancer-associated mis-
sense and in-frame indel mutations were calculated across all mSWI/
SNF subunits and averaged over a window size of 21 aa centered at each
residue (10 amino acids on each side). Third, these smoothed averages
were scaled to a range between 0 (no recurrence) and 1 (highest recur-
rence) to generate the local recurrence of NDD- and cancer-associated
missense and in-frame indel mutations. Fourth, the local recurrence
maps across all mSWI/SNF for NDD- and cancer-associated muta-
tions were subtracted (NDD-cancer) to form the NDD-cancer NES on
a range bounded by −1 (maximally enriched in cancer) and 1 (maxi-
mally enriched in NDD). NDD- and cancer-associated missense and
in-frame mutations were derived as described in (Fig. 5a). These local
and NESs were visualized across the specific paralogs in the recombi-
nant cBAF structure (PDB ID 6LTJ) as various colored heatmaps (local
NDD recurrence scaled in green, local cancer recurrence scaled in red,
NDD-Cancer NESs in blue-white-red: blue = enriched in cancer, red =
enriched in NDD) and across specific paralogs indicated in the Circos
plot as a purple-orange histogram (purple, enriched in cancer; orange,
enriched in NDD). The local enrichment scores for NDD (green) and
cancer (red) were visualized as histograms in the outer bands of the
Circos plot. Previously published nucleosome-bound cBAF cross-
linking mass spectrometry data were combined and visualized as inner
links on the Circos plot, where link thickness is proportional to the
frequency of cross-links (the maximum frequency of cross-links is
capped at 10 units). The Circos plot was made using the Circos software135.
Rolling averages of cancer and NDD mutational recurrence
(missense and in-frame indels only) were calculated for BAF genes
and visualized as a scatter plot with a regression line using the
seaborn136 regplot function.
NDD functional mutation analysis
Specific NDD residues predicted (by structural analysis) to disrupt
buried residues (altering cavities), buried charged residues and
hydrogen-bonds, BAF subunit or BAF module interaction, and BAF
domain interaction were visualized in PyMol, with the disruptive NDD
mutations indicated in red and putative interacting/proximal residues
in blue or purple. Additionally, Missense 3D webserver with recombi-
nant NCP-bound cBAF complex as input was used to assign functional
consequences of some of these disruptive NDD mutations.
NDD human versus yeast analysis
Select NDD residues in the integrated dataset were mapped to the
recombinant NCP-bound cBAF complex (PDB: 6LTJ), yeast Swi/Snf
(PDB:6UXW) and Snf2-nuclesome structures (PDB:5X0Y, 5X0X) were
used to show that seemingly exposed residues on the cBAF structure
are in fact buried by the brace helices in SMARCA2/A4 and that certain
side-chain orientations in cBAF structure have different orientations in
the yeast structures. SMARCA2/4 variant residues were mapped onto
additional yeast Snf2-nucleosome structures (PDB:5Z3O, 5Z3U) to
explore the open (ADP-bound) and closed (ADP-BeFx-bound) ATPase
states and emphasize ATP and DNA interacting residues of the ATPase
domain.
Statistics and reproducibility
A hypergeometric test was used to determine the enrichment of genes
of interest in a given gene set representing a specific biological process,
molecular function, pathway or meaningful biological collection of
genes. This analysis is more thoroughly described under NDD Gene Set
Nature Genetics
Enrichment Analysis. OLS regression analysis was carried out using the
default parameters in the seaborn regplot function.
No statistical method was used to predetermine sample size.
Samples sizes for the hypergeometric test were determined using
the standard procedure for GO, enrichment, or overrepresentation
analysis.
Known duplicate samples or potentially duplicate samples
from manual curation were excluded from analysis. Criteria for
exclusion are thoroughly described under Curating mSWI/SNF
gene NDD-associated variants. No other data were excluded from
the analyses from variants collected from the aforementioned
public or private databases. The experiments were not randomized.
The investigators were not blinded to allocation during experiments
and outcome assessment.
Reporting summary
Further information on research design is available in the Nature
Portfolio Reporting Summary linked to this article.
Data availability
Public and private data can be accessed through their respective por-
tals. Private data will require prior authorization. Data can be cleaned
and normalized using any standard or well-established procedure
for variant analysis or the procedures described in this paper, includ-
ing referenced papers or procedures. The integrated, curated and
de-duplicated data (to the best of our ability) are available in Sup-
plementary Table 1. No additional data or intermediate results will be
available upon request given the high manual burden to verify access
to a variety of private portals, repositories and patients.
Code availability
Variants were processed using well-established procedures described
in the referenced papers. Datasets from diverse sources were inte-
grated using a combination of code (to automate certain steps) and
manual curation. Thus, the standalone code is not sufficient to regen-
erate the integrated dataset. Therefore, this code and intermediate
results from dataset integration and curation is not available upon
request. The code used for analysis and to generate figures is avail-
able under Creative Commons license through Zenodo at https://doi.
org/10.5281/zenodo.8008632. Analyses were executed in Python (v3.7),
R (v4.1.1), GraphPad Prism (v92.2), matplotlib(v3.3.1), circos (v0.69-9)
and seaborn (v0.11.1).
PyMOL v2.4.0 was used to visualize structures. The Consurf online
server was used for conservation analysis. Geneious Prime v2021.2.2
was used for multiple sequence alignmentss. The PolyPhen2 online
server using the HumVar model was used to predict the severity/patho-
genicity of the compiled NDD mutations. Unless otherwise noted,
mutational counts, bar plots, pie charts, and Venn diagrams throughout
were made using a combination of Python (v3.7), R (v4.1.1), GraphPad
Prism (v92.2), matplotlib(v3.3.1) and seaborn (v0.11.1). The lollipop
portion of the 2D schematics were created using the St. Jude PeCan
Protein Paint software. Missense substitutions were visualized as a
Sankey diagram using Google Charts. The Circos plot was made using
the Circos software (v0.69-9). Missense substitutions were visualized
as a Sankey diagram using Google Charts. The code used to process
and visualize the data are available under the MIT license at Zenodo at
https://doi.org/10.5281/zenodo.8008632.
References
116. Hargreaves, D. C. & Crabtree, G. R. ATP-dependent chromatin
remodeling: genetics, genomics and mechanisms. Cell Res. 21,
396–420 (2011).
117. Sokpor, G., Castro-Hernandez, R., Rosenbusch, J., Staiger, J. F. &
Tuoc, T. ATP-dependent chromatin remodeling during cortical
neurogenesis. Front. Neurosci. 12, 226 (2018).
Analysishttps://doi.org/10.1038/s41588-023-01451-6 118. Li, Y. et al. The emerging role of ISWI chromatin remodeling
complexes in cancer. J. Exp. Clin. Cancer Res. 40, 346 (2021).
119. Torrado, M. et al. Refinement of the subunit interaction network
within the nucleosome remodelling and deacetylase (NuRD)
complex. FEBS J. 284, 4216–4232 (2017).
120. Sardiu, M. E. et al. Conserved abundance and topological
features in chromatin-remodeling protein interaction networks.
EMBO Rep. 16, 116–126 (2015).
121. Giaimo, B. D., Ferrante, F., Herchenröther, A., Hake, S. B. &
Borggrefe, T. The histone variant H2A.Z in gene regulation.
Epigenetetics Chromatin. 12, 37 (2019).
122. Fröb, F. & Wegner, M. The role of chromatin remodeling
complexes in Schwann cell development. Glia 68, 1596–1603
(2020).
123. Willhoft, O. & Wigley, D. B. INO80 and SWR1 complexes: Rhe
non-identical twins of chromatin remodelling. Curr. Opin. Struc.
Biol. 61, 50–58 (2020).
124. Conaway, R. C. & Conaway, J. W. The INO80 chromatin
remodeling complex in transcription, replication and repair.
Trends Biochem. Sci. 34, 71–77 (2009).
125. Shah, S. G. et al. HISTome2: a database of histone proteins,
modifiers for multiple organisms and epidrugs. Epigenetetics
Chromatin 13, 31 (2020).
126. Khare, S. P. et al. HIstome—A relational knowledgebase of human
histone proteins and histone modifying enzymes. Nucleic Acids
Res. 40, D337–D342 (2012).
127. Croce, L. D. & Helin, K. Transcriptional regulation by Polycomb
group proteins. Nat. Struct. Mol. Biol. 20, 1147–1155 (2013).
128. Greenberg, M. V. C. & Bourc’his, D. The diverse roles of DNA
methylation in mammalian development and disease. Nat. Rev.
Mol. Cell Biol. 20, 590–607 (2019).
129. Fu, J. M. et al. Rare coding variation provides insight into the
genetic architecture and phenotypic context of autism.
Nat. Genet. 54, 1320–1331 (2022).
130. Hunter, J. D. Matplotlib: A 2D graphics environment. Comput. Sci.
Eng. 9, 90–95 (2007).
131. McLaren, W. et al. The Ensembl Variant Effect Predictor. Genome
Biol. 17, 122 (2016).
132. Wall, L., Christiansen, T., & Orwant, J. Programming perl (O’Reilly
Media, 2000).
133. Schrödinger, L., & DeLano, W. PyMOL. http://www.pymol.org/
pymol (2020).
134. Ashkenazy, H. et al. ConSurf 2016: an improved methodology
to estimate and visualize evolutionary conservation in
macromolecules. Nucleic Acids Res. 44, W344–W350 (2016).
135. Krzywinski, M. et al. Circos: An information aesthetic for
comparative genomics. Genome Res. 19, 1639–1645 (2009).
136. Waskom, M. seaborn: statistical data visualization. J. Open Source
Softw. 6, 3021 (2021).
Acknowledgements
We are grateful to all members of the Kadoch laboratory and our
collaborators in the Santen and Vergano research groups for helpful
discussions. This analysis includes data generated through the
Coffin-Siris Syndrome Registry (S.A.S.V., Children’s Hospital of the
King’s and Daughters and Eastern Virginia Medical School) under IRB
approval number EVMS #15-03-0058, the ARID1B registry (G.W.E.S.,
Leiden University Medical Center, http://www.arid1bgene.com/) and
the sharing of de-identified patient variants identified from individuals
through Leiden University Medical Center was approved through
the Institutional Review Board of Leiden University Medical Center
(approval waivers no: G18.098 and G21.129). This study also uses
data generated by the DECIPHER community. A full list of centers
contributing to DECIPHER is available from https://deciphergenomics.
org/about/stats and via email from [email protected].
Funding for the DECIPHER project was provided by the Wellcome
Sanger Trust. Those who carried out the original analysis and
collection of data in the DECIPHER database bear no responsibility
for the further analysis or interpretation of the data. This study makes
use of DDD study. The DDD study presents independent research
commissioned by the Health Innovation Challenge Fund (grant
number HICF-1009-003), a parallel funding partnership between
Wellcome and the Department of Health, and the Wellcome Sanger
Institute (grant number WT098051). The views expressed in this
publication are those of the author(s) and not necessarily those of
Wellcome or the Department of Health. We would like to acknowledge
the American Association for Cancer Research and its financial and
material support in the development of the AACR Project GENIE
registry, as well as members of the consortium for their commitment
to data sharing. Interpretations are the responsibility of study authors.
The study has UK Research Ethics Committee approval (10/H0305/83,
granted by the Cambridge South REC, and GEN/284/12 granted
by the Republic of Ireland REC). The research team acknowledges
the support of the National Institute for Health Research, through
the Comprehensive Clinical Research Network. This work was
supported in part by the HHMI Gilliam Fellowship (A.M.V.) and the Ford
Foundation Predoctoral Fellowship (A.M.V.).
Author contributions
A.M.V. and C.K. conceived of and directed the study. A.S. performed
all computational and statistical analyses. F.K.S., J.F. and M.T. analyzed
and curated the SFARI and DDD datasets used in this analysis. P.J.v.d.S.
curated and presented newly reported NDD-associated mutations.
S.A.S.V. and G.W.E.S. curated and contributed novel human genetic
sequencing data and edited the manuscript. C.K. and A.M.V. wrote
the manuscript and all authors critically reviewed and edited the
manuscript.
Competing interests
C.K. is the scientific founder, scientific advisor to the Board of
Directors, scientific advisory board member, shareholder and
consultant for Foghorn Therapeutics. C.K. is also a member of the
scientific advisory board and is a shareholder of Nested Therapeutics,
Nereid Therapeutics and Accent Therapeutics, serves on the scientific
advisory board for Fibrogen and serves as a consultant for Google
Ventures and Cell Signaling Technologies. C.K. and A.M.V. hold
patents in the field of mSWI/SNF complex targeting therapeutics.
S.A.S.V. is a member of the scientific advisory board at Ambry
Genetics, for which no compensation is received. The other authors
declare no competing interests.
Additional information
Extended data is available for this paper at
https://doi.org/10.1038/s41588-023-01451-6.
Supplementary information The online version contains supplementary
material available at https://doi.org/10.1038/s41588-023-01451-6.
Correspondence and requests for materials should be addressed to
Cigall Kadoch.
Peer review information Nature Genetics thanks the anonymous
reviewer(s) for their contribution to the peer review of this work.
Reprints and permissions information is available at
www.nature.com/reprints.
Nature Genetics
Analysishttps://doi.org/10.1038/s41588-023-01451-6 Extended Data Fig. 1 | See next page for caption.
Nature Genetics
Analysishttps://doi.org/10.1038/s41588-023-01451-6 Extended Data Fig. 1 | SWI/SNF complex genes are among the most frequently
mutated genes in human NDD. a, Bar charts depicting mean number of non-
benign SNVs in DECIPHER and ASD+DD across gene sets indicated. b, Bar graph
summarizing the number of non-benign DECIPHER SNVs across top 5 categories
from Fig. 1a. c–e, Rank plots depicting GOMF gene sets in (c) DDG2P, (d) ranked
by total number of ASD+DD de novo missense variants, (e) ranked by mutation
frequency, top 50 GOMFs. f–j, Bar charts showing distribution of variants across
sets indicated in each title. mSWI/SNF or cBAF, PBAF, and ncBAF are highlighted
in red. k, Heatmap summarizing DECIPHER database mutational frequency for
each chromatin remodeling complex separated by variant type (all variants,
copy number variants (CNV), and SNVs/indels). l, Pie charts showing inheritance,
pathogenicity, and zygosity breakdown of all mSWI/SNF complex variants from
DECIPHER. m, Heatmaps depicting the mutational frequency of chromatin
remodeling genes in SWI/SNF, CHD, ISWI, and INO80 complex family classes in
the ASD+DD dataset. Total number of SNV and indel variants per protein complex
family indicated. n, Scatterplot of the total number of de novo missense and PTVs
in ASD+DD for all genes ranked by the mutational burden of each gene. mSWI/
SNF genes are shown in red. o, Scatterplot of the log normalized total number
of cancer missense, frameshift, and nonsense mutations in the TCGA MC3
PanCancer dataset versus the total number of NDD de novo missense and PTVs
in ASD+DD datasets. mSWI/SNF genes shown in red. p, Grouped bar graph of
the proportion of NDD (blue) and cancer (orange) missense and PTV mutations
across all mSWI/SNF genes sorted by decreasing NDD mutational proportion.
Nature Genetics
Analysishttps://doi.org/10.1038/s41588-023-01451-6 Extended Data Fig. 2 | Characteristics of NDD-associated single-residue
amino acid perturbations in mSWI/SNF components. a, Distribution
of single-nucleotide variants (SNVs) found in NDD-associated missense
mutations of mSWI/SNF family genes (Supplementary Table 1) in the integrated
dataset (n=2539). b, Horizontal bar graphs of the top 20 amino acid missense
substitutions in the integrated dataset (Supplementary Table 1). c, Bar chart
characterizing amino acid chemical property changes upon missense mutation
for NDD-associated variants in the integrated dataset. d, Stacked bar graphs
of the distribution of amino acid substitution chemical property changes in
NDD-associated missense mutations in the integrated dataset. e, Sankey diagram
of the distribution of NDD-associated missense substitutions in the integrated
dataset. Ribbon thickness represents frequency of substitutions in the integrated
dataset. f, Stacked bar chart summarizing percentage of NDD-associated
missense and in-frame indel mutations in the integrated dataset falling within
intrinsically disordered (defined by MobiDB-lite) or structured regions for (left)
each mSWI/SNF subunit and (right) all mSWI/SNF subunits combined.
Nature Genetics
Analysishttps://doi.org/10.1038/s41588-023-01451-6 Extended Data Fig. 3 | NDD-associated missense variants mapped on cBAF
and PBAF 3D structures. a, NDD-associated missense and inframe indel
variants mapped on to the 3D structure of the endogenous human cBAF complex
(PDBDEV_00000056). Red spheres represent NDD-associated variants in the
subunit indicated, blue spheres represent those mapped from the paralog
subunit, and residues in purple represent NDD-variants mapped in both primary
subunit present on cBAF structure and paralog subunit. Variants that map
exclusively on endogenous complex are indicated. Recurrent variants (n>3)
are emphasized in red. b, Bar chart indicating proportion of NDD-associated
missense and in-frame indel mutations in the integrated dataset mappable to
current mSWI/SNF complex structures separated by subunits. c, NDD-associated
missense and inframe indel variants mapped on to the 3D structure of the
PBAF complex (PDB 7VDV). Red and blue spheres represent NDD-associated
variants in the subunit indicated. Blue spheres and annotations emphasize PBAF
subcomplex specific variants mapped.
Nature Genetics
Analysishttps://doi.org/10.1038/s41588-023-01451-6 Extended Data Fig. 4 | See next page for caption.
Nature Genetics
Analysishttps://doi.org/10.1038/s41588-023-01451-6 Extended Data Fig. 4 | Structural dissection of mSWI/SNF subunit mutations
across the ARP, Core, and ATPase modules. a, b, (a) SMARCB1-C terminal alpha
helix and (b) SMARCA4-ATPase domain (top) ConSurf conservation mapping and
(bottom) multiple sequence alignment using D. melanogaster, C. elegans, and
S. cerevisiae homologs. c, NDD-associated missense and in-frame indel variants
mapped onto the 3D structure of the cBAF complex (PDB:6LTJ) color coded by
residue chemical characteristics: red: positive charge, blue: negative charge,
green: polar, orange: nonpolar. Nonpolar residues of the ACTB (Arp module) and
Table of nonpolar mutations predicted to structurally disrupt ACTB are shown.
d, ACTB NDD mutations may alter internal hydrophobic cavities, interfaces with
ACTL6A/B, and interfaces with SMARCA2/A4-HSA. Mutant residues shown in red
and putative proximal/interacting residues shown in blue/purple. e, SMARCB1-
RPT and WH domain NDD mutations predicted to disrupt internal cavity
integrity, and hydrogen bonding to interacting ARID1A main chain carbonyls,
respectively. Top, selected NDD-associated SMARCB1 missense mutations are
labeled, and major domains of SMARCB1 are colored, including RPT1 (blue),
RPT2 (orange), and CTD (red). Bottom, mutant residue shown in red and
putative proximal/interacting residues shown in blue. f, Mapping of conserved
SMARCA2/4 NDD mutant residues (red) on the yeast Snf2 ATPase domain (5X0Y
and 6UXW) compared to the recombinant cBAF SMARCA4 ATPase (6LTJ). Brace
helices (indicated in yeast structures) are not resolved in human cBAF structure,
but demonstrate that certain residues, emphasized in yellow, are buried by the
SMARCA2/4 brace helices, rather than exposed. g, Mapping of SMARCA2/4 brace
helix NDD variants onto the closed state of the SMARCA4 ATPase domain using
the PBAF structure (7VDV). NDD variants clustered in brace helices are predicted
to disrupt nucleosome remodeling activity as has been shown with R1243 and
R973 NDD and cancer-associated mutations indicated in panel97. h, Mapping of
SMARCA2/4 NDD mutant residues on the Snf2 ATPase open (gray) and closed
(pale cyan) states (PDB IDs: 5Z3O, 5Z3U). NDD residues colored blue in open state
and red in closed state. i, SMARCA2/4 NDD mutant residues (left) within 5Å of the
ADP-BeFx and (right) interacting with nucleosomal DNA mapped onto the closed
yeast Snf2 ATPase structure (5Z3U).
Nature Genetics
Analysishttps://doi.org/10.1038/s41588-023-01451-6 g
i
Consensus
Identity
Mean PI
Sequence
Logo
H. sapiens (ARID1A)
M. musculus (Arid1a)
D. melanogaster (Osa)
Consensus
Sequence
Logo
Mean PI
Identity
H. sapiens (DPF2)
M. musculus (Dpf2)
D. melanogaster (Dpf2)
ARID1A/B ARID DNA-Binding Domain
L1054
G1062Q1066
K1072
R1053
R1074
Y1027
L1092
W1023
K1021
R1020
G1017
L1011
D1050
L1049
h
SMARCB1 Winged-Helix DNA-Binding Domain
F1103
G11
I28
R37
DPF2 PHD zinc-finger domain
j
D274
C276
Q288
C298
A315
W322
C330
D346
D349-R350
W369
C345
C356 P365
Consensus
Sequence
Logo
Mean PI
Identity
M. musculus (Smarcb1)
H. sapiens (SMARCB1)
D. melanogaster (Snr1)
D. rerio (smarcb1a)
Consensus
Identity
Mean PI
Sequence
Logo
H. sapiens (SMARCE1)
M. musculus (Smarce1)
D. rerio (smarce1)
D. melanogaster (Bap111)
SMARCE1 HMG DNA-Binding Domain
P69
Y73
Y76
K92
W104-R105
E120
Y126
Extended Data Fig. 5 | See next page for caption.
Nature Genetics
Analysishttps://doi.org/10.1038/s41588-023-01451-6247257267277287297307317327337347357367377260270280290300310320330340350360370380390247257267277287297307317327337347357367377247257267277287297307317327337347357367377247257267277287297307317327337347357367377260269279289299309319329339349359369379389 Extended Data Fig. 5 | Perturbed subunit positions shared between cancer
and NDD highlight ATPase, nucleosome binding regions, and Arp module.
a, Venn diagram overlapping unique cancer missense and inframeshift variants
identified from cBioPortal_PanCan, cBioPortal_GENIE and COSMICv94
cancer genetics datasets. b, Venn diagram overlapping unique cancer and
NDD (Supplementary Table 1) missense and inframe variants by amino acid
position regardless of mutation consequence. NDD mutations derived from
Supplementary Table 1, cancer mutations derived by combining cBioPortal_
PanCan, cBioPortal_GENIE and COSMICv94 datasets. c, Top ten most recurrent
mutant residue amino acid positions shared between Cancer and NDD sorted
by frequency in each disease type. Highest recurrence of NDD mutations also
included. NDD- and cancer-associated mutations were derived as described
in (b). d, Bar plot showing the total number of unique missense/indel mSWI/
SNF mutations across the following cancer datasets: cBioPortal_PanCan,
cBioPortal_GENIE, COSMICv94. e, Correlation of missense and inframeshift
mutations shared between cancer (cBioPortal_PanCan only) and NDD across
recombinant cBAF structure. Briefly, NDD- and cancer-associated missense
and in-frame indel mutations were remapped onto the primary paralogs of the
recombinant cBAF (PDB ID: 6LTJ) structure. A rolling average with a window size
of 11aa centered on each residue (5aa on each side) of mutation recurrence by
residue position for NDD and cancer was used for the scatterplot and correlation
calculation. NDD- and cancer-associated mutations were derived from
Supplementary Table 1 (NDD) and cBioPortal_PanCan datasets. The translucent
bands around the regression line represent the 95% confidence interval
estimated using a bootstrap for 100 iterations. f, Heatmap representation of
scaled local enrichment of NDD- and cancer-associated missense and in-frame
indel mutational burden of (left, in green) NDD and (right, in red) cancer reflected
on the 3D structure of the human cBAF complex (PDB: 6LTJ). Local enrichment
scores were computed as described in (Fig. 5e). NDD- and cancer-associated
mutations were derived as described in (Fig. 5e). g–j, Multiple sequence
alignment of (g) ARID1A-ARID domain, (h) SMARCB1-WH domain, (i) DPF2-
PHD domain, and ( j) SMARCE1-HMG domain, with variety of related homologs
(including M. musculis, D. rerio, D. melanogaster, C. elegans, and S. cerevisiae,
where possible).
Nature Genetics
Analysishttps://doi.org/10.1038/s41588-023-01451-6
| null |
10.1371_journal.pone.0240269.pdf
| null |
Genomics data are now available at the NCBI repository: https://www . Results
|
RESEARCH ARTICLE
Using association rule mining to jointly detect
clinical features and differentially expressed
genes related to chronic inflammatory
diseases
Rosana VeronezeID
RochaID
Fernando J. Von Zuben1, Raquel Mantuaneli Scarel-Caminaga2
3, Cla´ udia V. Maurer-Morelli3, Silvana Regina Perez Orrico4,5, Joni A. CirelliID
1*, Saˆ mia Cruz Tfaile Corbi2, Ba´ rbara Roque da Silva2, Cristiane de S.
4,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Veroneze R, Cruz Tfaile Corbi S, Roque da
Silva B, de S. Rocha C, V. Maurer-Morelli C, Perez
Orrico SR, et al. (2020) Using association rule
mining to jointly detect clinical features and
differentially expressed genes related to chronic
inflammatory diseases. PLoS ONE 15(10):
e0240269. https://doi.org/10.1371/journal.
pone.0240269
Editor: Paolo Magni, Università degli Studi di
Milano, ITALY
Received: June 25, 2020
Accepted: September 23, 2020
Published: October 2, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0240269
Copyright: © 2020 Veroneze et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Genomics data are
now available at the NCBI repository: https://www.
1 Department of Computer Engineering and Industrial Automation, School of Electrical and Computer
Engineering, University of Campinas (UNICAMP), Campinas, SP, Brazil, 2 Department of Morphology,
Genetics, Orthodontics and Pediatric Dentistry, School of Dentistry at Araraquara, São Paulo State
University (UNESP), Araraquara, SP, Brazil, 3 Department of Medical Genetics and Genomic Medicine,
University of Campinas (UNICAMP), Campinas, SP, Brazil, 4 Department of Diagnosis and Surgery, School
of Dentistry at Araraquara, São Paulo State University (UNESP), Araraquara, SP, Brazil, 5 Advanced
Research Center in Medicine, Union of the Colleges of the Great Lakes (UNILAGO), São Jose´ do Rio Preto,
SP, Brazil
* [email protected], [email protected]
Abstract
Objective
It is increasingly common to find patients affected by a combination of type 2 diabetes melli-
tus (T2DM), dyslipidemia (DLP) and periodontitis (PD), which are chronic inflammatory dis-
eases. More studies able to capture unknown relationships among these diseases will
contribute to raise biological and clinical evidence. The aim of this study was to apply associ-
ation rule mining (ARM) to discover whether there are consistent patterns of clinical features
(CFs) and differentially expressed genes (DEGs) relevant to these diseases. We intend to
reinforce the evidence of the T2DM-DLP-PD-interplay and demonstrate the ARM ability to
provide new insights into multivariate pattern discovery.
Methods
We utilized 29 clinical glycemic, lipid and periodontal parameters from 143 patients divided
into five groups based upon diabetic, dyslipidemic and periodontal conditions (including a
healthy-control group). At least 5 patients from each group were selected to assess the tran-
scriptome by microarray. ARM was utilized to assess relevant association rules considering:
(i) only CFs; and (ii) CFs+DEGs, such that the identified DEGs, specific to each group of
patients, were submitted to gene expression validation by quantitative polymerase chain
reaction (qPCR).
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
1 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Results
We obtained 78 CF-rules and 161 CF+DEG-rules. Based on their clinical significance, Peri-
odontists and Geneticist experts selected 11 CF-rules, and 5 CF+DEG-rules. From the five
DEGs prospected by the rules, four of them were validated by qPCR as significantly different
from the control group; and two of them validated the previous microarray findings.
Conclusions
ARM was a powerful data analysis technique to identify multivariate patterns involving clini-
cal and molecular profiles of patients affected by specific pathological panels. ARM proved
to be an effective mining approach to analyze gene expression with the advantage of includ-
ing patient’s CFs. A combination of CFs and DEGs might be employed in modeling the
patient’s chance to develop complex diseases, such as those studied here.
ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=
GSE156993.
Funding: RV and FJVZ are supported by São Paulo
Research Foundation (FAPESP - http://www.
fapesp.br/) Grant 2017/21174-8, Coordination of
Superior Level Staff Improvement (CAPES - https://
www.capes.gov.br/) and Brazilian National Council
for Scientific and Technological Development
(CNPq - http://www.cnpq.br/) Grant 307228/2018-
5. RMSC is supported by FAPESP Grants 2007/
08362-8, 2009/16233-9, 2010/10882-2, 2014/
16148-0 and 2016/25418-6, CAPES and CNPq
Grant 304570/2017-6. The funders had no role in
study design, data collection and analysis, decision
to publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Introduction
As a metabolic disorder, diabetes mellitus (DM) is caused either by a deficiency of insulin’s
mechanism of action, by an insulin secretion deficit, or by both [1]. As recently reported by
Jeong et al. [2], the prevalence of DM has increased exponentially in recent decades, being
expected to affect 693 million patients within 25 years. Of all adults newly diagnosed with DM,
more than 90% are affected by type 2 diabetes mellitus (T2DM) [3]. According to Jeong et al.
[2], in 2017 the estimated total global healthcare expenditure considering DM was USD 850
billion, with a relevant proportion of these costs arising from the treatment of various compli-
cations associated with the progression of DM. Over a period of years most T2DM patients
progress to three major groups of complications: microvascular, macrovascular, and miscella-
neous [4]. Regarding miscellaneous T2DM complications, Jeong et al. [2] recently reported
that dyslipidemia had the highest relative incidence risk of comorbidities that evolved after a
diagnosis of T2DM in Koreans. In 2010, the third cause of premature deaths (before the age of
70 years) in Brazilian subjects was regarded as diabetes, with high fasting plasma glucose and
high body mass index (BMI) being some of the major risk factors related to diabetes mortality
(53,353 individuals, or 12%) [5].
Dyslipidemia (DLP) is a metabolic dysfunction that results from an increased level of lipo-
proteins in the blood [6, 7]. Some studies have revealed that DLP could be one factor associ-
ated with DM-induced immune cell alterations [7–9]. It is believed that pro-inflammatory
cytokines produce an insulin resistance syndrome similar to that observed in DM [7, 9]. Find-
ings concerning chronically elevated levels of inflammatory markers suggest that poor glyce-
mic control of T2DM patients could increase risk for cardiovascular disease and infectious
diseases, including periodontitis [8, 10].
Periodontitis (PD) is a common chronic inflammatory disease characterized by destruction
of the periodontium, which is the supporting structures of the teeth, such as gingiva, periodon-
tal ligament and alveolar bone [11]. PD is a microbially induced oral disease, in which the bac-
terial biofilm is formed on the surfaces of teeth providing a chronic microbial stimulus that
elicits a local inflammatory response in the gingival tissues [12]. PD is also considered an
inflammatory disorder influenced by factors such as genetics [13], immune system reactions,
smoking [14] and the occurrence of systemic diseases, including DM [15]. Periodontal infec-
tion and DM have a two-way relationship [16] and PD can be recognized as the sixth largest
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
2 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
complication associated with DM [17]. In response to bacterial products after periodontium
infection, there are local and systemic elevations of pro-inflammatory cytokines [18], which
may induce alterations in the metabolism of lipids, contributing to DLP in these patients [7,
9]. Some studies indicate an association between elevation in blood lipoproteins and alter-
ations in the periodontal condition [6, 19–21].
Currently, the interplay of T2DM, DLP and PD has been increasingly affecting patients
worldwide. Those are chronic inflammatory diseases, including systemic T2DM and DLP,
while PD is localized at the periodontium of the patient. Growing evidence indicates a biologi-
cal connection among T2DM-DLP-PD, demonstrated by the finding that these patients pres-
ent a hyperinflammatory state promoted by systemically increased levels of pro-inflammatory
molecules, as reviewed by Soory et al. [22]. Moreover, all of them are considered chronic and
complex diseases, since they are caused by a combination of genetic, environmental and life-
style factors [23]. Therefore, more studies focused on detecting unknown relationships in data-
sets of diseased patients will contribute to a better understanding of the interplay of T2DM,
DLP and PD.
Association rule mining (ARM) has been widely used to discover hidden relationships
established by multiple attributes that characterize a complex process under investigation. It
has several applications in the medical domain (for instance, see [24–26]) promoting highly
interpretable explanations without requiring data mining expertise [27]. In addition to
interpretability, another reason that makes ARM a widely used data mining technique is that
the obtained rules are capable of summarizing the joint impact of several factors [27, 28].
Thus, ARM is a powerful technique to assess the supposed interplay of T2DM, DLP and PD.
The ARM was previously used to assess the T2DM survival risk [29], and to determine the
T2DM comorbidities in large amounts of clinical data [30]. Ramezankhani et al. [31] showed
that ARM is a useful approach to determine the most frequent subsets of attributes in people
who will develop diabetes. However, this is the first study using ARM to simultaneously iden-
tify the potential clinical patterns and genetic markers of this group of diseases, thus revealing
clinical features and differentially expressed genes capable of properly characterizing these
chronic inflammatory diseases.
The outline of this paper is as follows. Section Materials and Methods presents the literature
review and our proposed methodology. Section Results and Discussion presents the experi-
mental results and an analytical explanation of their implications, followed by concluding
remarks in Section Conclusion.
Materials and methods
Datasets
Studied population. This research was approved by the Ethics in Human Research Com-
mittee of School of Dentistry at Araraquara (UNESP; Protocol number 50/06). Patients who
voluntarily sought dental treatment at the School of Dentistry at Araraquara (UNESP), Brazil,
were informed about the aims and methods of the study, providing their written consent to
participate; therefore, the whole study was conducted according to the ethical principles of the
Declaration of Helsinki.
The patients were characterized by the following criteria: age from 35 to 60 years, presence
of at least 15 natural teeth and similar socioeconomic level. Pre-selected patients, according to
their medical history, had their glycemic and lipid profiles investigated by biochemical blood
analysis, and were submitted to full periodontal examination. Then, 143 patients were divided
into five groups based upon diabetic, dyslipidemic and periodontal conditions:
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
3 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
1. Group 1: poorly controlled T2DM with DLP and PD. Number of subjects = 28.
2. Group 2: well-controlled T2DM with DLP and PD. Number of subjects = 29.
3. Group 3: DLP and PD. Number of subjects = 29.
4. Group 4: systemically healthy individuals with PD. Number of subjects = 29.
5. Group 5: systemically and periodontally healthy individuals (control group). Number of
subjects = 28.
No patient in those five groups presented: history of antibiotic therapy in the previous 3
months and/or nonsteroidal anti-inflammatory drug therapy in the previous 6 months, preg-
nancy or use of contraceptives or any other hormone, current or former smoking addiction,
history of anemia, periodontal treatment or surgery in the preceding 6 months, use of hypoli-
pidemic drugs such as statins or fibrates, and history of diseases that interfere with lipid metab-
olism, such as hypothyroidism and hypopituitarism.
Additionally, patients enrolled in this study were previously investigated regarding malo-
naldehyde (MDA) quantification and some inflammatory cytokine levels [32], micronuclei
frequency (DNA damage evaluation) [33] and lipid peroxidation [32]. In these previous stud-
ies, power analysis based on a pilot study determined that at least 20 patients in each group
would be sufficient to assess differences in those molecules with 90% power and 95% confi-
dence interval.
Biochemical, physical and periodontal evaluations. Clinical criteria to include each
patient in the studied group are presented in what follows. Subjects were submitted to physical
and anthropometric examination for evaluating obesity such as abdominal circumference
(cm), height (m), weight (kg), waist (cm), hip (cm) and body mass index [33].
After a 12-hour overnight fast, each subject was referred to a clinical analysis laboratory
that collected a blood sample for evaluating: glycated haemoglobin (HbA1c) by enzymatic
immunoturbidimetry, fasting plasma glucose (mg/dL) by the modified Bondar & Mead
method, high-sensitivity C-reactive protein by the nephelometric method and insulin levels
by the chemiluminescence method (U/L). The homeostasis model assessment (HOMA) was
evaluated to calculate insulin resistance (IR). The diagnosis of T2DM was made by an endocri-
nologist who monitored the glycemic levels of each patient by evaluation of HbA1c; being
patients considered poorly controlled (HbA1c �8.0%) or well-controlled (HbA1c �7.0%).
Normoglycemic (nondiabetic) individuals presented fasting glucose levels <100 mg/dL and
HbA1c <5.7% [34–36].
The lipid profile [triglycerides (TG), total cholesterol (TC), and high density lipoprotein
(HDL)] was performed by enzymatic methods. Low density lipoprotein (LDL) was determined
by the Friedewald formula. Individuals with transitory DLP were not included here by consid-
ering the highest cutoff values: TC �240 mg/dL, LDL �160 mg/dL, HDL <40 mg/dL, and
TGs �200 mg/dL, according to the 2018 AHA / ACC / AACVPR / AAPA / ABC / ACPM /
ADA / AGS / APhA / ASPC / NLA / PCNA Guideline on the Management of Blood Choles-
terol [37]. It was also considered in this analysis the non-HDL-cholesterol (N-HDL-C), given
by N-HDL-C = TC—HDL, being the abnormal cutoff value �130 mg/dL, which is considered
to be a good predictor of cardiovascular disease (CVD) risk [38].
Diagnosis of periodontitis in at least 4 non-adjacent teeth, including local signs of inflam-
mation, loss of the connective tissue attachment of gingiva to teeth (clinical attachment loss,
CAL �4mm), and tissue destruction (presence of deep periodontal pockets �6mm) was
adopted according to the American Academy of Periodontology [39]. Each subject underwent
a periodontal clinical examination performed at 6 sites per tooth. The presence of deep
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
4 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
periodontal pockets �6mm with CAL �5mm and bleeding on probing in at least 8 sites dis-
tributed in different quadrants of the dentition were the criteria of severe periodontitis [40].
Regarding the mutagenesis analysis, the description of the peripheral blood sampling, cell
culture and cytokinesis-block micronucleus (CBMN) assay can be found in Corbi et al. [33].
Table 1 summarizes the clinical features collected from the 143 investigated subjects. The
clinical feature dataset is available in S1 File.
Isolation of peripheral blood mononuclear cells, RNA extraction and microarray analy-
sis. Patients with greater glycemic, lipid and periodontal homogeneity parameters had their
transcriptome investigated (30 subjects in total) from peripheral blood mononuclear cells
(PBMCs), divided into: Group 1 (number of subjects = 5), Group 2 (number of subjects = 7),
Group 3 (number of subjects = 6), Group 4 (number of subjects = 6) and Group 5 (number of
subjects = 6). PBMCs were isolated, and total RNA was extracted using TRizol (Invitrogen,
Rockville, MD, USA) and purified by an RNeasy Protection Mini Kit (Qiagen, Hilden, Ger-
many) according to the manufacturer’s instructions. RNA was quantified by a NanoVue Spec-
trophotometer (GE Healthcare Life Sciences, Oslo, Norway), and its integrity was assessed by
agarose gel electrophoresis (1%). Only RNA samples in the λ(260/280) and λ(260/230) reasons
between 1.8 and 2.2 were used for microarray and quantitative real-time PCR analyses. Micro-
array data were generated from 500 nanograms of RNA as the initial input of each sample in
the GeneChip IVT Labeling Kit and hybridized to the U133 Plus 2.0 (Affymetrix Inc., Santa
Clara, CA, USA) arrays, which comprise 54,675 human transcripts. The U133 Plus 2.0 arrays
were scanned twice using the GeneChip Scanner 3000 7G (Affymetrix Inc., Santa Clara, CA,
USA). The Robust Multichip Average (RMA) strategy was used to preprocess raw .CEL files
[41, 42]. This strategy performs background correction through a normal-exponential convo-
lution model, quantile normalizes the probe intensities and summarizes them into probeset-
level quantities using an additive model fit through the median-polish strategy [43]. The gene
expression dataset is available in S2 File.
Association rule mining
Let An×m be a binary data matrix with the row index set X = {1, 2, . . ., n} and the column index
set Y = {1, 2, . . ., m}. Each row represents a transaction, and each column represents an item.
Each element aij 2 A holds the binary relationship between transaction i and item j. Let (X, Y)
denote the entire matrix A and (I, J) denote a submatrix of A with I � X and J � Y.
Definition 1 A subset J = {j1, . . ., js} � Y is called an itemset.
For a subset J � Y, we define J# = {x 2 X|axj = 1, 8j 2 J} as the set of transactions common to
all the items in J. The support of an itemset J is given by σ(J) = |J#|.
The problem of mining all frequent itemsets can be described as follows: determine all sub-
sets J � Y such that σ(J)�minSup, where minSup is a user-defined parameter.
To reduce the computational cost of the frequent itemset (pattern) mining problem, some
algorithms mine only the maximal frequent itemsets, i.e., those frequent itemsets from which
all supersets are infrequent and all subsets are frequent. The problem of this approach is that it
leads to loss of information since the supports of the subsets of the maximal frequent itemsets
are not available. An option to reduce the computational cost of the frequent pattern mining
problem without loss of information is to mine only the closed frequent itemsets. A frequent
itemset J is called closed if there exists no superset H aˆSˇƒ J with H# = J#. Remarkably, the set of
closed frequent itemsets uniquely determines the exact frequency of all frequent itemsets, and
it can be orders of magnitude smaller than the set of all frequent itemsets [44]. Therefore, this
approach drastically reduces the number of rules that have to be presented to the user, without
any information loss [45].
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
5 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Table 1. Description of the clinical features of the 143 subjects enrolled in this study (%ts stands for % of tooth sites).
Characteristic
Demographic
# Attribute
1
Sex
2 Age
Cardiovascular and obesity risk
3
Body Mass Index
4 Waist / Hip Ratio
5 Abdominal Circumference
Type 2 Diabetes Mellitus
6
Fasting Plasma Glucose
Alias
Sex
Age
BMI
WHR
AC
FPG
Unit Domain
1. Female
2. Male
1. �50
2. >50
yr.
m/kg2
1. Underweight: <18.5
2. Normal weight: [18.5, 25)
3. Overweight: [25, 30)
4. Obesity class I: [30, 35)
5. Obesity class II e III: �35
cm/cm (see Table 2)
cm
(see Table 3)
mg/dL
1. Normoglycemic: <100
2. Prediabetes or high-risk: [100, 126)
3. Established diabetes: �126
7
Insulin
INS
U/L
1. Normal: �25
2. Altered: >25
8 Glycated Haemoglobin
HbA1c
%
1. Normoglycemic: <5.7
9 HOMA-IR
HOMA-IR
2. Prediabetes or high-risk: [5.7, 6.5)
3. Decompensation (transitory): [6.5, 8)
4. Decompensation (defined): �8
1. Normal: �2.15
2. Altered: >2.15
Dyslipidemia
10 Total Cholesterol
TC
mg/dL
1.<150 (Optimal)
11 HDL cholesterol
HDL
mg/dL
1. <40 (Low)
2. [40, 60]
3. >60
12
LDL cholesterol
LDL
mg/dL
1. <100 (Optimal)
2. [150, 200)
3. [200, 240)
4. �240
2. [100, 130)
3. [130, 160)
4. [160, 190)
5. �190
13 Triglycerides
TG
mg/dL
1. <150 (Optimal)
2. [150, 200)
3. �200
14 Non-HDL-Cholesterol
N-HDL-C
mg/dL
1. <130 (Optimal)
2. [130, 160)
3. [160, 190)
4. [190, 220)
5. �220
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
6 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Table 1. (Continued)
Characteristic
Periodontal
# Attribute
15 Visible Plaque
Alias
VP
Unit Domain
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
16 Gingival Index bleeding
GI
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
17 Bleeding on probing
BOP
%ts
1. Low: <30
18 Total Number of Teeth
TNT
2. Medium: [30, 50]
3. High: >50
1. low number teeth: �20
2. high number teeth: >20
19
Interproximal periodontal pocket depth (PPDi) �3mm PPDi3mm
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
20 PPDi = 4—5mm
PPDi4-5mm
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
21 PPDi � 6mm
PPDi6mm
%ts
1. Low: <30
2. Medium and High: �30
22
Interproximal clinical attachment loss (CALi) �2mm
CALi2mm
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
23 CALi = 3-4mm
CALi3-4mm
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
24 CALi � 5mm
CALi5mm
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
25
Suppuration
SUPP
%ts
1. Abscence: <1
Mutagenesis
26 Nuclear Division Index
NDI
2. Moderate: [1, 16)
3. Severe: �16
1. Low: <1.87
2. Moderate: [1.87, 2.08)
3. High: �2.08
27
Frequency of Binucleated cells with Micronuclei
MNCF
%
1. Low: <3.05
28 Micronucleus Frequency
MNF
%
1. Low: <3.5
29
Frequency of Nucleoplasmic Bridges
FNB
%
1. Low: <1.21
2. Moderate: [3.5, 6.1)
3. High: �6.1
2. Moderate: [3.05, 7.2)
3. High: �7.2
2. Moderate: [1.21, 2.7)
3. High: �2.7
https://doi.org/10.1371/journal.pone.0240269.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
7 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Definition 2 An association rule (AR) is an expression of the form J ) H, where J and H are
itemsets, H \ J = ;. J is called antecedent (or head) and H is called consequent (or tail) of the
rule.
The support of an association rule J ) H is the number of transactions that contain the
itemset J[H: σ(J ) H) = σ(J[H). The confidence of an association rule J ) H measures its pre-
dictive accuracy and is given by conf(J ) H) = σ(J ) H)/σ(J). A rule is considered a strong rule
if conf(J ) H)�minConf, where minConf is a user-defined parameter. The completeness (or
recall) is given by comp(J ) H) = σ(J ) H)/σ(H). Remark that confidence and completeness
are not symmetric measures because by definition they are conditional on the antecedent and
consequent, respectively. The metric lift measures the degree of surprise of a rule and is given
by lift(J ) H) = σ(J ) H)/(σ(J) × σ(H)).
A user can be interested in a more specific set of association rules, where the consequents of
the rules describe a target attribute. These rules are known as class association rules (CARs).
Definition 3 A class association rule (CAR) is an expression of the form J ) c, where J is an
itemset and c is a class label (a target item).
In this work, each item is given by an attribute-value pair. Thus, for instance, FPG = 3 is an
item; {AC = 3, FPG = 3, HbA1c = 4} is an itemset; and {AC = 3, FPG = 3, HbA1c = 4} )
{GI = 3, BOP = 3} is an association rule.
Given that the result to be presented to the user is more parsimonious, we will focus on
closed frequent itemsets here. The patterns will be mined using the RIn-Close_CVCP algo-
rithm [46, 47], which is a fast algorithm and avoids the necessity of the itemization step [47].
Its implementation is available at https://github.com/rveroneze/rinclose.
Association rule mining from the clinical features alone. T2DM, DLP, and PD have
their own specific characteristics (features or attributes) generally taken as decision variables
to perform a diagnosis. However, given the increasing incidence of patients affected by differ-
ent interplays of T2DM-DLP-PD, we originally used ARM to assess whether there are joint
attributes present in patients with these comorbidities that might indicate the biological inter-
relationship among them.
Fig 1 shows a flowchart that summarizes the process of association rule mining from the
dataset containing solely clinical features. From the clinical features collected from the investi-
gated patients (presented in Table 1), we selected the most clinically relevant to diagnose
T2DM, DLP and PD diseases isolated. We did not use the mutagenesis attributes because they
are not applied in a clinical routine for disease diagnosis. The following 17 clinical features
were selected for this analysis: BMI, WHR, AC, FPG, HbA1c, HOMA-IR, TC, HDL, LDL, TG,
N-HDL-C, GI, BOP, PPDi6mm, CALi34mm, CALi5mm and SUPP. Thus, the dataset to be
analyzed has 143 subjects and 17 attributes. BMI, WHR and AC attributes represent character-
istics that confer cardiovascular and obesity risk, according to the World Health Organization
[19, 48]. The N-HDL-C attribute is considered a good predictor of CVD risk [38]. The glyce-
mic parameters: FPG, HbA1c and HOMA-IR (Homeostasis Model Assessment to calculate
the insulin resistance) are considered essential for the diagnosis of T2DM and its metabolic
control [35, 36]. TC, HDL, LDL and TG are important lipid parameters to diagnose DLP [37].
Regarding periodontitis, the American Academy of Periodontology (AAP) utilizes the clinical
periodontal parameters: GI, BOP, PPDi6mm, CALi3-4mm, CALi5mm and SUPP [39, 40].
The parameters used in ARM were: minSup = 14 and minConf = 70%. A rule was consid-
ered interesting whenever at least one of the following attributes is present: PPDi6mm = 2; GI,
BOP, CALi34mm, CALi5mm, SUPP 2{2, 3}. We followed those clinical periodontal parame-
ters, as recommended by the AAP, because they indicate periodontal disease activity. Those
selected attributes are considered relevant to identify individuals undoubtedly affected by
moderate or severe periodontitis, allowing us to check if there is an evident association
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
8 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Fig 1. Flowchart that summarizes the process of association rule mining from the dataset containing solely clinical features.
https://doi.org/10.1371/journal.pone.0240269.g001
between both systemic diseases (T2DM and DLP) and PD. In this way, we corroborate the
existence of a T2DM-DLP-PD biological interrelationship.
In addition, we performed an analysis focusing on the cardiovascular and obesity risk attri-
butes to determine whether they are associated with periodontal disease. Therefore, we per-
formed an analysis with only the cardiovascular and obesity risk attributes in the antecedent
part of the rule (BMI, WHR, AC, FPG, N-HDL-C), and the same attributes in the consequent
part of the rule. We also performed an analysis comprising only T2DM patients presenting
diabetic dyslipidemia, which are the 10 patients from Groups 1 and 2 having TG �204 mg/dL
and HDL <38 mg/dL [49, 50].
The results of these analysis will be presented and discussed in Section Results and
Discussion.
Association rule mining from the clinical features and gene expression datasets in con-
junction. The transcriptome of the patients studied here obtained from PBMCs by microar-
ray was analyzed utilizing bioinformatics and statistical tools, as described in topic Isolation of
peripheral blood mononuclear cells, RNA extraction and microarray analysis. Those analyses,
developed as regularly, produced a list of differentially expressed genes (DEGs). However, in
that kind of analysis the gene expression profile obtained by the probesets did not consider the
patient’s clinical features (CFs). In conventional bioinformatics and statistical tools, adequate
clinical diagnosis of each group of patients is used to determine whether a DEG is related to a
specific pathological condition. Here, we used ARM to identify the joint interplay of CFs and
DEGs, having the advantage of taking together CFs and genetic markers to identify each com-
bination of T2DM-DLP-PD complex diseases. This approach might contribute to better iden-
tifying new targets for the diagnosis of each combination of those complex diseases, as well as
for modeling the patient’s chance to develop them.
Fig 2 shows a flowchart that summarizes the process of class association rule mining from
the dataset containing both CFs and DEGs. First, we performed the preprocessing of the
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
9 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Fig 2. Flowchart that summarizes the process of class association rule mining from the dataset containing both clinical features (CFs) and
differentially expressed genes (DEGs).
https://doi.org/10.1371/journal.pone.0240269.g002
original gene expression dataset (GED), which has the gene expression profile of 54,675 genes
obtained from the transcriptome of the 30 subjects, in the following three steps:
1. Gene selection: we filtered out genes with small profile variance, in specific we filtered
out gene expression profiles with variation less than 0.1 when considering the difference
between its maximum and minimum values. It was done because gene profiling experi-
ments typically include genes that exhibit little variation in their profile and these genes are
usually uninteresting. Thus, these genes are commonly removed from the analysis. With
this filter, 50.441 genes were removed, leaving 4.234 genes for the subsequent analysis.
2. Normalization: we used zero-mean normalization to adjust the values measured on different
scales to a common scale. Let g be the gene expression profile of a gene g for the 30 subjects
of our study. The normalized gene expression profile ^g is given by ^g ¼ ðg (cid:0) avgðgÞÞ=stdðgÞ,
where avg(g) and std(g) are, respectively, the sample average and the sample standard devia-
tion of g.
3. Discretization: if a normalized gene expression value was above 1.0, it was considered
over-expressed (and it is represented by the value 1 in our results); if a normalized gene
expression value was below -1.0, it was considered under-expressed (and it is represented
by the value -1 in our results); otherwise the gene expression value was considered uninter-
esting and was ignored.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
10 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
We performed the mining of CARs in this preprocessed GED with the following parame-
ters: minSup = 3 and minConf = 90%. The group of each individual (Groups 1 to 5) is the tar-
get attribute. The result, containing 118 CARs, was used for a new phase of gene selection as
described in what follows. The 118 CARs have a coverage of 1081 genes (this means that
1081 genes are presented in these rules). Of these 1081 genes, 17 genes are present in con-
flicting rules, exhibiting the same value for the control group (Group 5) and for the other
groups (Groups 1 to 4). Therefore, these 17 genes were discarded. Thus, 1081 − 17 = 1064
genes were selected for the new phase of analysis, together with the 29 CFs listed in Table 1.
In this new phase of analysis, we performed the mining of CARs with the same parameters,
i.e., minSup = 3 and minConf = 90%. The results will be presented and discussed in Section
Results and Discussion.
Reverse transcription-quantitative polymerase chain reaction (RT-qPCR)
Real-Time Analysis
To biologically validate the genes selected from the CARs considering the CFs+DEGs, we con-
ducted RT-qPCR analyses in all 143 patients (including the 30 patients who were analyzed
by microarray) distributed into the 5 groups, according to the subitem Studied population.
Reverse transcription reactions were performed utilizing the High Capacity Kit (Thermo
Fisher Scientific). Complementary DNA (cDNA) was used to perform qPCR reactions for the
selected DEGs, which are represented as probe sets in Table 7. To investigate the expression of
the probe (or gene) identified by the rule selected for each group of patients, the TaqMan1
gene expression assay specific for each of these “target” genes was utilized. Each target gene is
normalized by a gene considered an endogenous control of the qPCR reactions, in this case,
we utilized the GAPDH -Glyceraldehyde-3-Phosphate Dehydrogenase gene (Hs02758991_g1),
due to its housekeeping expression pattern.
All reactions were performed in duplicate utilizing the 7500 Real-Time PCR-System
(Thermo Fisher Scientific, Foster City, CA, USA). To calculate gene expression, Expression
Suite Software was used (Thermo Fisher Scientific, Foster City, CA, USA), which employs the
comparative Cycle Threshold (ΔCt) method for multivariate data analysis. Statistical analysis
to find differences in the gene expression by the values of 2−ΔCt between the groups was per-
formed by the Mann-Whitney test, utilizing GraphPad Prism software, version 5.0, and con-
sidering a significance level of 0.05 [51].
Results and discussion
Association rules for the dataset of clinical features (CFs)
It was obtained 78 rules comprising the CF dataset, which are presented in S1 Table. The peri-
odontists and geneticist experts analyzed those rules to select examples of rules of high clinical
relevance to demonstrate the T2DM-DLP-PD interrelationship. To select the rules, the follow-
ing requirements were established in decreasing order of relevance:
1. In the antecedent part of the rule, the joint presence of attributes with altered values in
these characteristics of Tables 1, 2 and 3: cardiovascular and obesity risk; T2DM; and DLP;
2. The highest confidence value.
The rules of Table 4 present, in general, WHR = 4 and AC = 3, which represent very high
cardiovascular and obesity risk for all ages of both male and female (see Tables 2 and 3);
FPG = 3, HbA1c = 4 and HOMA-IR = 2 represent the worst glycemic parameters, evidencing
that those patients have established T2DM with defined metabolic decompensation and
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
11 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Table 2. Waist / hip ratio domain.
1. Low
2. Moderate
3. High
4. Very High
1. Low
2. Moderate
3. High
4. Very High
Age � 39
<0.72
[0.72, 0.79)
[0.79, 0.84]
> 0.84
Age � 39
<0.84
[0.84, 0.92)
[0.92, 0.96]
> 0.96
https://doi.org/10.1371/journal.pone.0240269.t002
Female
39 < Age � 49
<0.73
[0.73, 0.80)
[0.80, 0.87]
> 0.87
Male
39 < Age � 49
<0.88
[0.88, 0.96)
[0.96, 1]
> 1
Age > 49
<0.74
[0.74, 0.82)
[0.82, 0.88]
> 0.88
Age > 49
<0.90
[0.90, 0.97)
[0.97, 1.02]
> 1.02
Table 3. Table caption Nulla mi mi, venenatis sed ipsum varius, volutpat euismod diam.
1. Low risk
2. High risk
3. Very high risk
Female
< 80
[80, 88)
� 88
Male
< 94
[94, 102)
� 102
https://doi.org/10.1371/journal.pone.0240269.t003
insulin resistance; the patients are also dyslipidemic as demonstrated by the highest levels of
total cholesterol (TC = 4) and triglycerides (TG = 3). The consequent part of those rules is
BOP = 3, which means that more than 50% of tooth sites bleed during the periodontal exam,
demonstrating wide and active inflammation of the periodontal tissues including the gingiva.
There are 4 rules showing as consequent SUPP = 2, meaning that those patients have a moder-
ate suppuration, since it affects 1% to 16% of tooth sites, indicating the presence of an estab-
lished periodontitis. The seventh and eighth rules of Table 4 show TC = 4 and N-HDL-C = 5,
meaning that individuals with the highest levels of TC and N-HDL-C have 78% of confidence
of presenting BOP = 3 or SUPP = 2, demonstrating wide and active inflammation of the peri-
odontal tissues and an established periodontitis.
Table 4. Association rules for the clinical feature dataset.
Rule
WHR = 4, FPG = 3, HbA1c = 4, TG = 3 ) BOP = 3
AC = 3, FPG = 3, HbA1c = 4, HOMA-IR = 2, TG = 3 ) BOP = 3
AC = 3, FPG = 3, HOMA-IR = 2, TG = 3 ) BOP = 3
WHR = 4, AC = 3, FPG = 3, HOMA-IR = 2, TG = 3 ) BOP = 3
WHR = 4, FPG = 3, TG = 3 ) BOP = 3
WHR = 4, FPG = 3, HOMA-IR = 2, TG = 3 ) BOP = 3
AC = 3, HOMA-IR = 2, TC = 4, TG = 3 ) SUPP = 2
TC = 4, N-HDL-C = 5 ) BOP = 3
TC = 4, N-HDL-C = 5 ) SUPP = 2
AC = 3, HOMA-IR = 2, TC = 4 ) SUPP = 2
WHR = 4, AC = 3, HOMA-IR = 2, TC = 4 ) SUPP = 2
https://doi.org/10.1371/journal.pone.0240269.t004
σrule
14
15
22
19
21
20
15
18
18
23
18
σhead
14
15
26
23
26
25
19
23
23
31
25
σtail
74
74
74
74
74
74
67
74
67
67
67
%Conf.
100.00
100.00
84.62
82.61
80.77
80.00
78.95
78.26
78.26
74.19
72.00
Lift
1.93
1.93
1.64
1.60
1.56
1.55
1.68
1.51
1.67
1.58
1.54
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
12 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Table 5. Association rules for the clinical feature dataset—Cardiovascular risk.
Rule
WHR = 4, AC = 3, FPG = 3 ) BOP = 3
FPG = 3 ) BOP = 3
AC = 3, FPG = 3 ) BOP = 3
WHR = 4, FPG = 3 ) BOP = 3
N-HDL-C = 5 ) BOP = 3
N-HDL-C = 5 ) SUPP = 2
BMI = 3, WHR = 4, AC = 3 ) SUPP = 2
BMI = 3, WHR = 4 ) SUPP = 2
https://doi.org/10.1371/journal.pone.0240269.t005
σrule
24
35
28
26
18
18
18
20
σhead
28
41
33
31
23
23
24
28
σtail
74
74
74
74
74
67
67
67
%Conf.
85.71
85.37
84.85
83.87
78.26
78.26
75.00
71.43
Lift
1.66
1.65
1.64
1.62
1.51
1.67
1.60
1.52
There was interest in verifying the association of cardiovascular and obesity parameters
with the presence of periodontitis. In that analysis we also included the N-HDL-C attribute,
which predicts CVD risk even better than LDL [52]. The rules obtained by focusing on only
those 11 attributes are presented in Table 5. We highlighted the rules: BMI = 3, WHR = 4,
AC = 3 ) SUPP = 2 and N-HDL-C = 5 ) BOP = 3, as supporting the evidence of an associa-
tion between cardiovascular risk factors and periodontitis. The obtained rules support the
clear association between N-HDL-C and parameters of periodontitis. The N-HDL-C was the
best predictor among all cholesterol measures, both for coronary artery disease events and for
strokes [53]. More recently, this was confirmed, since the highest N-HDL-C concentrations in
blood (�220 mg/dL, which is equivalent to �5.7 mmol/L) were associated with the highest
long-term risk of atherosclerotic cardiovascular disease [54]. Here we observed exactly this
highest level of N-HDL-C in the rules of Table 5. Interestingly, there are good reasons for
the usefulness of N-HDL-C in monitoring patients, since unlike LDL, N-HDL-C does not
require the triglyceride concentration to be 4.5 mmol/L (400 mg/dL), and has an additional
advantage of not requiring patients to fast before blood sampling. Therefore, it is certainly a
better measure than calculated LDL for patients with increased plasma triglyceride concentra-
tions [38, 53].
In general, these rules demonstrate the interplay between cardiovascular and obesity risk,
T2DM, DLP and PD, which is in line with some studies as reviewed by Soory [22] and Khu-
maedi et al. [8]. These diseases manifest persistent elevation of systemic inflammatory media-
tors, characterizing chronic inflammation [8]. It is known to be one of the atherosclerosis non-
traditional risk factors and has a role in every phase of atherogenesis [8]. Atherogenic dyslipi-
demia is expressive among T2DM individuals, for example, in 10 − 15% of the European popu-
lation [49, 50]. Therefore, we performed an analysis comprising only our 10 T2DM patients
presenting diabetic dyslipidemia [49, 50]. The rules found for this pathologic condition are
presented in Table 6. We highlighted the rule: FPG = 3, HOMA-IR = 2, TC = 2, HDL = 1,
TG = 3 ) BOP = 3, as it demonstrated that diabetic dyslipidemia was associated with
more than 50% of tooth sites bleeding, one of the main significant signals of periodontium
Table 6. Association rules for the clinical feature dataset—Diabetic dyslipidemia.
Rule
AC = 3, FPG = 3, HOMA-IR = 2, HDL = 1, TG = 3 ) GI = 3
AC = 3, FPG = 3, HOMA-IR = 2, TC = 2, HDL = 1, TG = 3 ) GI = 3, BOP = 3
FPG = 3, HOMA-IR = 2, TC = 2, HDL = 1, TG = 3 ) BOP = 3
AC = 3, FPG = 3, HOMA-IR = 2, HDL = 1, TG = 3 ) GI = 3, PPDi6mm = 1
https://doi.org/10.1371/journal.pone.0240269.t006
σrule
6
5
6
5
σhead
6
5
6
6
σtail
6
5
6
5
%Conf.
100.00
100.00
100.00
83.33
Lift
1.67
2.00
1.67
1.67
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
13 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
inflammation. Periodontitis is the most common cause of chronic inflammation in diabetic
patients. Both periodontitis and diabetes have detrimental effects on each other in terms of
alveolar bone destruction and poor metabolic control, by continuous inflammatory mediator
activation [8].
Association rules for the datasets of clinical features and differentially
expressed genes in conjunction
Remark that we used ARM to obtain rules with joint patterns of CFs and DEGs, having the
advantage of taking together the clinical characteristics and the genetic markers to identify
each T2DM-DLP-PD combination of complex diseases. Also different from the rules consider-
ing only CFs (Table 4), the CF+DEG-rules were obtained for identifying specifically a group of
patients. Therefore, both CFs and DEGs were considered in the antecedent part of the rules,
and the consequent part of the rules is given by the number representing the groups (Groups 1
to 5). It was obtained 161 CF+DEG-rules, which are presented in S2 Table.
Because of the importance of biologically validating the CF+DEG-rules, Periodontists and
Geneticist experts selected only one discriminant rule for each of the five groups, as presented
in Table 7. The Periodontists and Geneticist experts make the decision of the CF+DEG-rules’s
choice following these criteria in decreasing order of relevance:
1. The joint presence of attributes showing values as altered as possible (according to the
reference values presented in Tables 1, 2 and 3) referring to the cardiovascular and obe-
sity risk, T2DM, DLP, PD, and also, at lower relevance, mutagenesis and demographic
characteristics;
2. The presence of one probe representing an over-expressed gene, such as ‘229026_at = 1’;
3. The highest confidence value;
4. The highest completeness value.
All the selected rules in Table 7 have 100% of confidence, which means that all subjects who
give support to a rule are from the same group.
Specifically to Group 1 of patients (poorly controlled T2DM with DLP and PD), the selected
rule means that 80% of the patients of Group 1 have high abdominal circumference (AC = 3),
meaning high CHD risk; altered glycemic parameters (FPG = 3, HbA1c = 4, HOMA-IR = 2),
evidencing that those patients have established T2DM with defined metabolic decompensation
Table 7. Association rules for the clinical feature and gene expression datasets in conjunction.
Rule
AC = 3, FPG = 3, INS = 1, HbA1c = 4, HOMA-IR = 2, HDL = 2, TG = 3, VP = 3, BOP = 3,
PPDi6mm = 1, CALi2mm = 1, SUPP = 2, 223130_s_at = -1, 229026_at = 1 ) 1
HOMA-IR = 2, TC = 4, TG = 3, N-HDL-C = 5, 208485_x_at = 1, 212386_at = -1 ) 2
FPG = 1, HDL = 2, PPDi3mm = 3, PPDi6mm = 1, MNCF = 2, 223422_s_at = 1,
224902_at = 1 ) 3
BMI = 2, FPG = 1, INS = 1, HbA1c = 1, HOMA-IR = 1, HDL = 2, TG = 1, TNT = 2,
PPDi6mm = 1, CALi2mm = 1, CALi3-4mm = 3, N-HDL-C = 1, 1560999_a_at = 1,
228766_at = -1, 244413_at = 1 ) 4
Age = 1, FPG = 1, INS = 1, HbA1c = 1, TG = 1, VP = 1, GI = 1, BOP = 1, TNT = 2,
PPDi3mm = 3, PPDi4-5mm = 1, PPDi6mm = 1, CALi5mm = 1, SUPP = 1, NDI = 2,
MNCF = 1, MNF = 1, FNB = 1, 236395_at = 1 ) 5
https://doi.org/10.1371/journal.pone.0240269.t007
%Comp. %Conf. Lift
80.00
100.00
6.00
71.00
67.00
100.00
100.00
4.29
5.00
67.00
100.00
5.00
67.00
100.00
5.00
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
14 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
and insulin resistance; high triglyceride level (TG = 3); established severe periodontitis as
denoted by VP = 3 (more than 50% of tooth sites showing poor oral hygiene), BOP = 3 (more
than 50% of tooth sites bleeding), PPDi6mm = 1 (up to 30% of tooth sites with deep periodon-
tal pockets), and SUPP = 2 (suppuration at maximum of 16% of tooth sites). Though the fol-
lowing attributes did not contribute to the identification of Group 1, they also did not disturb
it: INS = 1, HDL = 2 and CALi2mm = 1.
The rule selected for Group 2 (well-controlled T2DM with DLP and PD) means that 71%
of the patients of Group 2 have insulin resistance demonstrated by HOMA-IR = 2; and the
highest levels of total cholesterol (TC = 4), triglycerides (TG = 3) and non-HDL-cholesterol
(N-HDL-C = 5). Surprisingly, considering the first criterion for selecting these 5 rules, for
identifying Group 2 of patients, a few rules were obtained. Because of this, in the selected rule
there were no attributes regarding the cardiovascular and obesity risk and PD. Moreover, it
should be taken into account that the rules obtained for Group 2 of patients should reflect
the clinical criteria defined to select the patients. For example, in comparison with Group 1,
Group 2 of patients differs only by the better metabolic control of T2DM.
The rule selected for Group 3 (DLP and PD) means that 67% of the patients have normal
fasting plasma glucose (FPG = 1) which is expected since they are not affected by T2DM; they
present altered HDL levels (HDL = 2), and they are affected by PD, since up to 30% of tooth
sites present very deep periodontal pockets (PPDi6mm = 1). Moreover, in this rule the moder-
ate frequency of binucleated cells with micronuclei (MNCF = 2) means that the circulating
blood of the patients is affected by a moderate level of mutagenesis, probably as a consequence
of the altered lipid metabolism of the patients. Indeed, a previous study of our research group
enrolling the same patients showed significantly higher mRNA levels of leptin in dyslipidemic
individuals (Groups 1, 2 and 3). Moreover, those leptin mRNA levels were significantly corre-
lated with periodontal parameters such as BOP, suppuration and mainly CALi � 5 mm [55].
Regarding Group 4 (systemically healthy individuals with PD), the selected rule means that
67% of the patients of this group are not obese, diabetic or dyslipidemic, as expected by the
underlined clinical criteria for selecting them. Those patients are only affected by generalized
periodontitis with pronounced alveolar bone loss, since they present more than 50% of tooth
sites with 3 to 4 mm of clinical attachment loss (CALi34mm = 3), and up to 30% of tooth sites
with very deep periodontal pockets (PPDi6mm = 1).
The rule selected for Group 5 (systemically and periodontally healthy individuals, or control
group) means that 67% of the patients of this group are not characterized by obesity, T2DM or
DLP, as expected by the underlined clinical criteria for selecting them. In addition, they did
not present active PD because it was not present in the rule any domain of bleeding or inflam-
mation, and the presence of the shallow periodontal pockets (PPDi3mm = 3) in at least 50% of
tooth sites is not an indicator of periodontal disease. Conversely, the occurrence of up to 30%
of tooth sites with PPDi45mm, PPDi6mm = 1, and clinical attachment loss (CALi5mm = 1)
suggests that those patients were previously affected by localized PD. Moreover, although the
rule includes the mutagenic parameters, their values are not altered.
To proceed to the biological validation of DEGs, we chose to validate by RT-qPCR (see Sub-
section Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) Real-Time
Analysis) one highly expressed gene in each of the five rules. Certainly, more rules with more
probes/DEGs could be selected for validation, but we had limitations in the volume of the bio-
logical sample of the patients (RNA obtained from PBMCs).
For Group 1, we selected the probe 229026_at = 1, whose gene is CDC42SE2 (Cell Division
Cycle 42 Small Effector 2), detected by the TaqMan assay Hs00184113_m1. Although there is
another gene in the rule of Group 1 (23130_s_at), this gene was down-regulated, and therefore
did not meet the criteria of choice. The CDC42SE2 gene has diverse biological functions, such
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
15 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Fig 3. Validation results by RT-qPCR of the genes considering the different Group (G) comparisons. All mRNA
levels of the investigated genes were normalized to the GAPDH endogenous control gene. (A) CDC42SE2 gene
expression, �p � 0.0001; (B) CFLAR gene expression, no statistical difference among the groups; (C) PDPR gene
expression, �p � 0.0002; (D) Validation of the CLECL1 gene expression, �p � 0.0064; (E) Validation of the MEF2C
gene expression, �p � 0.0425. Data represent the mean ± SEM 2−ΔCt of all patients in that group (Mann–Whitney U
test; α = 5%).
https://doi.org/10.1371/journal.pone.0240269.g003
as the organization of the actin cytoskeleton by acting downstream of CDC42SE2, inducing
actin filament assembly, and it may play a role in early contractile events in phagocytosis in
macrophages. Accordingly, the CDC42SE2 gene alters CDC42-induced cell shape changes. In
activated T-cells, the CDC42SE2 gene may play a role in CDC42-mediated F-actin accumula-
tion at the immunological synapse [56]. The CDC42 (Cell Division Cycle 42) gene encodes a
small GTPase protein belonging to the Rho-subfamily, which regulates signaling pathways that
control diverse cellular functions including cell morphology, migration, endocytosis and cell
cycle progression [56].
In Fig 3(A), it can be observed that the CDC42SE2 gene was down-regulated in the decom-
pensated T2DM, dyslipidemic and PD patients (Group 1) (p-value � 0.0001) in comparison
to the healthy patients (Group 5). Actually, this finding obtained by qPCR is contrary to the
expected by the rule based on the microarray data (denoted by the positive 1 value of the
‘229026_at’). Therefore, the qPCR method showed discordant gene expression levels from
those detected by the microarray. Actually, it is not uncommon to find discrepant results of
gene expression between qPCR and microarray, either because the gene expression between
the diseased and control groups did not reach statistical difference or because conflicting
results were found between the qPCR and microarray methods [51]. The discordant
CDC42SE2 gene expression between qPCR and microarray (not validation) means more a lim-
itation of the method for identification of gene expression levels than a limitation of CAR min-
ing. In addition, considering that Group 2 of patients only differs from Group 1 in patients’
metabolic control, we also investigated the CDC42SE2 gene expression in the well-controlled
T2DM-DLP-PD (Group 2) patients, and we observed significantly lower levels in Group 1 but
no significant difference in Group 2 in comparison to the control Group 5. Therefore, when
we performed the CDC42SE2 gene expression comparison involving Groups 1, 2 and 5, we
observed the lowest expression in the worst metabolic condition of patients (Group 1), while
the patients with adequate metabolic control (Group 2) had similar CDC42SE2 expression
when compared with the healthy patients of Group 5.
For Group 2, the selected probe is 208485_x_at = 1, which is the CFLAR (CASP8 and FADD
Like Apoptosis Regulator) gene, detected by the TaqMan assay Hs01117851_m1. The protein
encoded by the CFLAR gene is a regulator of apoptosis which may function as a crucial link
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
16 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
between cell survival and cell death pathways. Additionally, this protein acts as an inhibitor of
TNF receptor superfamily member 6 (TNFRSF6) mediated apoptosis [56]. Considering the
rule, an over-expression of the CFLAR gene was expected in Group 2 compared to Group 5.
However, there was a similarly high expression of the CFLAR gene in both Groups 2 and 5 (see
Fig 3(B)). We also performed the analysis of the CFLAR gene expression for Groups 1, 2 and 5,
observing no significant difference among them, although a lower gene expression can be
found in the patients with the worst metabolic condition (Group 1).
For Group 3, the rule has 2 highly expressed genes/probes, and we selected the 224902_at
probe for further analysis, which is the PDPR (Pyruvate Dehydrogenase Phosphatase Regulatory
Subunit) gene, detected by the TaqMan assay Hs01663324_m1, because it takes part in a more
interesting metabolic pathway. This gene acts on the pyruvate dehydrogenase complex by cata-
lyzing the oxidative decarboxylation of pyruvate and linking glycolysis to the tricarboxylic acid
cycle and to the synthesis of fatty acids [56]. The observed significant down-regulation of the
PDPR gene in Group 3 (DLP-PD) in comparison with the healthy Group 5 (p-value � 0.0002)
by qPCR was discordant from those detected by the microarray, as shown in Fig 3(C).
Regarding Group 4 (patients affected by only PD), the rule also has 2 highly expressed
genes/probes: the IL12RB2 gene (1560999_a_at), and the CLECL1 gene (244413_at), which
was chosen to validate the gene expression by using the TaqMan assay Hs00416849_m1. The
CLECL1 (C-Type Lectin Like 1) gene acts as a co-stimulating molecule of T cells and plays a
role in the interaction of dendritic cells with T cells and the cells of the adaptive immune
response [56]. In the comparison between Group 4 and Group 5, there was a highly statistically
significant (p-value � 0.0064) expression of the CLECL1 gene in Group 4, validating the DEG
detected by microarray, as shown in Fig 3(D).
For Group 5 (healthy patients), the only highly expressed gene is the MEF2C (Myocyte
Enhancer Factor 2C) gene (identified by the 236395_at probe), and detected by the TaqMan
assay Hs00231149_m1. The MEF2C gene is involved in several normal pathways of muscular,
vascular, neural, megakaryocyte and platelet development, bone marrow B lymphopoiesis, B
cell survival and proliferation in response to BCR stimulation, efficient responses of IgG1 anti-
bodies to T cell dependent antigens and normal induction of B cells from the germinal center
[56]. The MEF2C gene expression by qPCR validated the DEG detected by microarray, as sig-
nificantly highly expressed in Group 5 when compared with Group 1 (p-value � 0.0425) (see
Fig 3(E)). It is interesting to compare PBMC gene expression between patients with the most
opposite healthy conditions, such as Groups 1, 2 and 5, in which the worst metabolic condition
(Group 1) showed the lowest level of MEF2C gene expression.
To our knowledge, this is the first initiative to investigate the expression of CDC42SE2 and
CLECL1 genes in the context of T2DM, DLP and PD, demonstrating the innovative character
of this study. Regarding CFLAR gene expression, only one study was reported in the literature
investigating the relationship between body composition and BMI in children and DNA meth-
ylation. CFLAR gene expression was positively regulated in PBMCs of obese children [57].
Similarly, only one study investigated the PDPR gene with the genetic risk for DM, but the
authors focused on type 1 DM, not allowing direct comparison with the T2DM results [58].
Two previous studies reported changes in the function of the MEF2C gene: Yuasa et al. [59]
found MEF2C transcriptional repression in patients with T2DM, and Davegårdh et al. [60]
verified a down-regulation of MEF2C related to obesity. Such results are in agreement with the
findings of our study, with MEF2C being more highly expressed in patients in Group 5 (sys-
temically and periodontally healthy individuals) than in Groups 1 and 2 (individuals with met-
abolic and periodontal involvement).
Although we originally utilized the ARM to investigate CFs and DEGs relevant in the con-
text of T2DM, DLP and PD, it is important to attest that:
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
17 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
1. We just considered the periodontitis parameters as the consequent part of the rules because
the literature demands more evidences regarding the association between systemic diseases
like T2DM and DLP, with PD;
2. Regarding the CF+DEG rules, more rules could be selected for each patient group, permit-
ting biological validation of up- or down-regulated probesets/genes, but we had limitations
in the volume of biological samples of the patients (RNA obtained from PBMCs) necessary
for the RT-qPCR technique.
Conclusion
We demonstrated that ARM is a powerful data analysis technique to identify consistent pat-
terns between the clinical and molecular profiles of patients affected by specific pathological
panels. In addition, ARM was able to evidence relevant associations among important parame-
ters of the periodontal, glycemic, lipid, cardiovascular and obesity risk conditions of the
patients. Considering the qPCR validation results of the DEGs prospected by the CARs of each
group of patients, four of the five genes revealed significant differences in comparison to the
control group; two of them CLECL1 and MEF2C genes validated the previous microarray find-
ings. These last genes were referred to groups without systemic metabolic impairment (Group
4 and Group 5). Further studies will investigate other DEGs and other rules. Additionally, as
an alternative to other commonly used techniques, ARM can be applied as a highly-interpret-
able mining approach to analyze the gene expression signal, with the advantage of including
the patient’s clinical features. Moreover, the combination of CFs and DEGs can be utilized to
further estimate the patient’s chance of developing complex diseases, such as those studied
here.
Supporting information
S1 File. Clinical feature dataset.
(CSV)
S2 File. Gene expression dataset.
(TXT)
S1 Table. Association rules mined from the clinical feature dataset.
(XLS)
S2 Table. Class association rules mined from clinical feature and gene expression datasets
in conjunction.
(XLS)
Author Contributions
Conceptualization: Rosana Veroneze, Fernando J. Von Zuben, Raquel Mantuaneli Scarel-
Caminaga.
Data curation: Rosana Veroneze, Saˆmia Cruz Tfaile Corbi, Cristiane de S. Rocha, Cla´udia V.
Maurer-Morelli, Silvana Regina Perez Orrico, Joni A. Cirelli, Raquel Mantuaneli Scarel-
Caminaga.
Formal analysis: Rosana Veroneze, Saˆmia Cruz Tfaile Corbi, Ba´rbara Roque da Silva, Cris-
tiane de S. Rocha, Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico, Joni A. Cirelli.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
18 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Funding acquisition: Rosana Veroneze, Silvana Regina Perez Orrico, Fernando J. Von Zuben,
Raquel Mantuaneli Scarel-Caminaga.
Investigation: Rosana Veroneze, Saˆmia Cruz Tfaile Corbi, Ba´rbara Roque da Silva, Cristiane
de S. Rocha.
Methodology: Rosana Veroneze, Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico,
Fernando J. Von Zuben, Raquel Mantuaneli Scarel-Caminaga.
Project administration: Rosana Veroneze, Silvana Regina Perez Orrico, Fernando J. Von
Zuben, Raquel Mantuaneli Scarel-Caminaga.
Resources: Rosana Veroneze, Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico, Fer-
nando J. Von Zuben, Raquel Mantuaneli Scarel-Caminaga.
Software: Rosana Veroneze, Fernando J. Von Zuben.
Supervision: Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico, Joni A. Cirelli, Fer-
nando J. Von Zuben, Raquel Mantuaneli Scarel-Caminaga.
Validation: Saˆmia Cruz Tfaile Corbi, Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico,
Joni A. Cirelli, Raquel Mantuaneli Scarel-Caminaga.
Visualization: Rosana Veroneze, Ba´rbara Roque da Silva, Silvana Regina Perez Orrico, Joni A.
Cirelli, Raquel Mantuaneli Scarel-Caminaga.
Writing – original draft: Rosana Veroneze, Ba´rbara Roque da Silva, Fernando J. Von Zuben,
Raquel Mantuaneli Scarel-Caminaga.
Writing – review & editing: Rosana Veroneze, Saˆmia Cruz Tfaile Corbi, Ba´rbara Roque da
Silva, Cristiane de S. Rocha, Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico, Joni
A. Cirelli, Fernando J. Von Zuben, Raquel Mantuaneli Scarel-Caminaga.
References
1. American Diabetes Association and others. Diagnosis and classification of diabetes mellitus. Diabetes
care. 2014; 37(Supplement 1):S81–S90. https://doi.org/10.2337/dc14-S081 PMID: 24357215
2.
Jeong E, Park N, Kim Y, Jeon JY, Chung WY, Yoon D. Temporal trajectories of accompanying comor-
bidities in patients with type 2 diabetes: a Korean nationwide observational study. Scientific Reports.
2020; 10(1):1–10. https://doi.org/10.1038/s41598-020-62482-1 PMID: 32218498
3. Holman N, Young B, Gadsby R. Current prevalence of Type 1 and Type 2 diabetes in adults and chil-
dren in the UK. Diabetic Medicine. 2015; 32(9):1119–1120. https://doi.org/10.1111/dme.12791 PMID:
25962518
4. Papatheodorou K, Papanas N, Banach M, Papazoglou D, Edmonds M. Complications of diabetes
2016. Journal of diabetes research. 2016; 2016.
5.
de Oliveira Otto MC, Afshin A, Micha R, Khatibzadeh S, Fahimi S, Singh G, et al. The impact of dietary
and metabolic risk factors on cardiovascular diseases and type 2 diabetes mortality in Brazil. PLoS
One. 2016; 11(3):e0151503. https://doi.org/10.1371/journal.pone.0151503 PMID: 26990765
6. Almeida Abdo J, Cirano FR, Casati MZ, Ribeiro FV, Giampaoli V, Viana Casarin RC, et al. Influence of
dyslipidemia and diabetes mellitus on chronic periodontal disease. Journal of periodontology. 2013; 84
(10):1401–1408. https://doi.org/10.1902/jop.2012.120366 PMID: 23136946
7. Nassar PO, Walker CS, Salvador CS, Felipetti FA, Orrico SRP, Nassar CA. Lipid profile of people with
diabetes mellitus type 2 and periodontal disease. Diabetes research and clinical practice. 2012; 96
(1):35–39. https://doi.org/10.1016/j.diabres.2011.11.017 PMID: 22154377
8. Khumaedi AI, Purnamasari D, Wijaya IP, Soeroso Y. The relationship of diabetes, periodontitis and car-
diovascular disease. Diabetes & Metabolic Syndrome: Clinical Research & Reviews. 2019; 13(2):1675–
1678. https://doi.org/10.1016/j.dsx.2019.03.023
9.
Zhou X, Zhang W, Liu X, Zhang W, Li Y. Interrelationship between diabetes and periodontitis: role of
hyperlipidemia. Archives of Oral Biology. 2015; 60(4):667–674. https://doi.org/10.1016/j.archoralbio.
2014.11.008 PMID: 25443979
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
19 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
10. Prattichizzo F, De Nigris V, Spiga R, Mancuso E, La Sala L, Antonicelli R, et al. Inflammageing and
metaflammation: the yin and yang of type 2 diabetes. Ageing Research Reviews. 2018; 41:1–17.
https://doi.org/10.1016/j.arr.2017.10.003 PMID: 29081381
11. Preshaw P, Alba A, Herrera D, Jepsen S, Konstantinidis A, Makrilakis K, et al. Periodontitis and diabe-
tes: a two-way relationship. Diabetologia. 2012; 55(1):21–31. https://doi.org/10.1007/s00125-011-
2342-y PMID: 22057194
12.
Lalla E, Papapanou PN. Diabetes mellitus and periodontitis: a tale of two common interrelated diseases.
Nature Reviews Endocrinology. 2011; 7(12):738. https://doi.org/10.1038/nrendo.2011.106 PMID:
21709707
13. Michalowicz BS, Diehl SR, Gunsolley JC, Sparks BS, Brooks CN, Koertge TE, et al. Evidence of a sub-
stantial genetic basis for risk of adult periodontitis. Journal of periodontology. 2000; 71(11):1699–1707.
https://doi.org/10.1902/jop.2000.71.11.1699 PMID: 11128917
14.
Leite FR, Nascimento GG, Scheutz F, Lopez R. Effect of smoking on periodontitis: a systematic review
and meta-regression. American Journal of Preventive Medicine. 2018; 54(6):831–841. https://doi.org/
10.1016/j.amepre.2018.02.014 PMID: 29656920
15. Mealey BL, Oates TW. Diabetes mellitus and periodontal diseases. Journal of periodontology. 2006; 77
(8):1289–1303. https://doi.org/10.1902/jop.2006.050459 PMID: 16881798
16.
17.
Lalla E, Lamster IB. Assessment and management of patients with diabetes mellitus in the dental office.
Dental Clinics. 2012; 56(4):819–829. PMID: 23017553
Lo¨ e H. Periodontal disease: the sixth complication of diabetes mellitus. Diabetes care. 1993; 16
(1):329–334. https://doi.org/10.2337/diacare.16.1.329 PMID: 8422804
18. Carrizales-Sepu´lveda EF, Ordaz-Farı´as A, Vera-Pineda R, Flores-Ramı´rez R. Periodontal disease,
systemic inflammation and the risk of cardiovascular disease. Heart, Lung and Circulation. 2018; 27
(11):1327–1334. https://doi.org/10.1016/j.hlc.2018.05.102 PMID: 29903685
19. Arboleda S, Vargas M, Losada S, Pinto A. Review of obesity and periodontitis: an epidemiological view.
20.
British dental journal. 2019; 227(3):235–239. https://doi.org/10.1038/s41415-019-0611-1
Fentoğlu O¨ , Kırzıoğlu F, O¨ zdem M, Koc¸ak H, Su¨ tc¸u¨ R, Sert T. Proinflammatory cytokine levels in hyper-
lipidemic patients with periodontitis after periodontal treatment. Oral diseases. 2012; 18(3):299–306.
https://doi.org/10.1111/j.1601-0825.2011.01880.x PMID: 22151458
21. Nepomuceno R, Pigossi SC, Finoti LS, Orrico SR, Cirelli JA, Barros SP, et al. Serum lipid levels in
patients with periodontal disease: A meta-analysis and meta-regression. Journal of Clinical Periodontol-
ogy. 2017; 44(12):1192–1207. https://doi.org/10.1111/jcpe.12792 PMID: 28782128
22. Soory M. Inflammatory mechanisms and redox status in periodontal and cardiometabolic diseases:
effects of adjunctive nutritional antioxidants and statins. Infectious Disorders-Drug Targets (Formerly
Current Drug Targets-Infectious Disorders). 2012; 12(4):301–315. PMID: 22697128
23. McAllister K, Mechanic LE, Amos C, Aschard H, Blair IA, Chatterjee N, et al. Current challenges and
new opportunities for gene-environment interaction studies of complex diseases. American journal of
epidemiology. 2017; 186(7):753–761. https://doi.org/10.1093/aje/kwx227 PMID: 28978193
24. Buxton EK, Vohra S, Guo Y, Fogleman A, Patel R. Pediatric population health analysis of southern and
central Illinois region: A cross sectional retrospective study using association rule mining and multiple
logistic regression. Computer methods and programs in biomedicine. 2019; 178:145–153. https://doi.
org/10.1016/j.cmpb.2019.06.020 PMID: 31416543
25.
Ivančević V, Tusˇek I, Tusˇek J, Knezˇević M, Elheshk S, Luković I. Using association rule mining to iden-
tify risk factors for early childhood caries. Computer Methods and programs in Biomedicine. 2015; 122
(2):175–181. https://doi.org/10.1016/j.cmpb.2015.07.008 PMID: 26271408
26. Kalgotra P, Sharda R. BIARAM: A process for analyzing correlated brain regions using association rule
mining. Computer methods and programs in biomedicine. 2018; 162:99–108. https://doi.org/10.1016/j.
cmpb.2018.05.001 PMID: 29903499
27.
28.
Toti G, Vilalta R, Lindner P, Lefer B, Macias C, Price D. Analysis of correlation between pediatric
asthma exacerbation and exposure to pollutant mixtures with association rule mining. Artificial intelli-
gence in medicine. 2016; 74:44–52. PMID: 27964802
Lin Y, Qian X, Krischer J, Vehik K, Lee HS, Huang S. A rule-based prognostic model for type 1 diabetes
by identifying and synthesizing baseline profile patterns. PloS one. 2014; 9(6):e91095. https://doi.org/
10.1371/journal.pone.0091095 PMID: 24926781
29. Simon GJ, Schrom J, Castro MR, Li PW, Caraballo PJ. Survival association rule mining towards type 2
diabetes risk assessment. In: AMIA annual symposium proceedings. vol. 2013. American Medical Infor-
matics Association; 2013. p. 1293.
30. Kim HS, Shin AM, Kim MK, Kim YN. Comorbidity study on type 2 diabetes mellitus using data mining.
The Korean journal of internal medicine. 2012; 27(2):197. https://doi.org/10.3904/kjim.2012.27.2.197
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
20 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
31. Ramezankhani A, Pournik O, Shahrabi J, Azizi F, Hadaegh F. An application of association rule mining
to extract risk pattern for type 2 diabetes using tehran lipid and glucose study database. International
journal of endocrinology and metabolism. 2015; 13(2). https://doi.org/10.5812/ijem.25389 PMID:
25926855
32. Bastos AS, Graves DT, Loureiro APdM, Rossa Ju´ nior C, Abdalla DSP, Faulin TdES, et al. Lipid peroxi-
dation is associated with the severity of periodontal disease and local inflammatory markers in patients
with type 2 diabetes. The Journal of Clinical Endocrinology & Metabolism. 2012; 97(8):E1353–E1362.
https://doi.org/10.1210/jc.2011-3397 PMID: 22564665
33. Corbi SC, Bastos AS, Orrico SR, Secolin R, Dos Santos RA, Takahashi CS, et al. Elevated micronu-
cleus frequency in patients with type 2 diabetes, dyslipidemia and periodontitis. Mutagenesis. 2014; 29
(6):433–439. https://doi.org/10.1093/mutage/geu043 PMID: 25239120
34.
de Souza Bastos A, Graves DT, de Melo Loureiro AP, Ju´ nior CR, Corbi SCT, Frizzera F, et al. Diabetes
and increased lipid peroxidation are associated with systemic inflammation even in well-controlled
patients. Journal of Diabetes and its Complications. 2016; 30(8):1593–1599. https://doi.org/10.1016/j.
jdiacomp.2016.07.011 PMID: 27497685
35. Association AD, et al. 2. Classification and diagnosis of diabetes: standards of medical care in diabetes
—2019. Diabetes Care. 2019; 42(Supplement 1):S13–S28. https://doi.org/10.2337/dc19-S002
36. Association AD, et al. 6. Glycemic Targets: Standards of Medical Care in Diabetes-2019. Diabetes
Care. 2019; 42(Suppl 1):S61. PMID: 30559232
37. Grundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS, et al. 2018 AHA/ACC/
AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA guideline on the management of
blood cholesterol: executive summary: a report of the American College of Cardiology/American Heart
Association Task Force on Clinical Practice Guidelines. Journal of the American College of Cardiology.
2019; 73(24):3168–3209. https://doi.org/10.1016/j.jacc.2018.11.002 PMID: 30423391
38. Piepoli MF, Hoes AW, Agewall S, Albus C, Brotons C, Catapano AL, et al. 2016 European Guidelines
on cardiovascular disease prevention in clinical practice: The Sixth Joint Task Force of the European
Society of Cardiology and Other Societies on Cardiovascular Disease Prevention in Clinical Practice
(constituted by representatives of 10 societies and by invited experts) Developed with the special contri-
bution of the European Association for Cardiovascular Prevention & Rehabilitation (EACPR). European
heart journal. 2016; 37(29):2315–2381. https://doi.org/10.1093/eurheartj/ehw106 PMID: 27222591
39. Periodontology AA. International workshop for a classification of periodontal diseases and conditions.
Ann Periodontol. 1999; 4(1).
40. Koromantzos PA, Makrilakis K, Dereka X, Katsilambros N, Vrotsos IA, Madianos PN. A randomized,
controlled trial on the effect of non-surgical periodontal therapy in patients with type 2 diabetes. Part I:
effect on periodontal status and glycaemic control. Journal of clinical periodontology. 2011; 38(2):142–
147. https://doi.org/10.1111/j.1600-051X.2010.01652.x PMID: 21114680
41. Gautier L, Cope L, Bolstad BM, Irizarry RA. affy—analysis of Affymetrix GeneChip data at the probe
level. Bioinformatics. 2004; 20(3):307–315. https://doi.org/10.1093/bioinformatics/btg405
42. Breitling R, Armengaud P, Amtmann A, Herzyk P. Rank products: a simple, yet powerful, new method
to detect differentially regulated genes in replicated microarray experiments. FEBS letters. 2004; 573(1-
3):83–92. https://doi.org/10.1016/j.febslet.2004.07.055 PMID: 15327980
43. Corbi SC, de Vasconcellos JF, Bastos AS, Bussaneli DG, da Silva BR, Santos RA, et al. circulating lym-
phocytes and monocytes transcriptomic analysis of patients with type 2 diabetes mellitus, dyslipidemia
and periodontitis. Scientific Reports. 2020; 10(1):1–14. https://doi.org/10.1038/s41598-020-65042-9
PMID: 32424199
44.
45.
Zaki MJ, Hsiao CJ. CHARM: An Efficient Algorithm for Closed Itemset Mining. In: Proceedings of the
2002 SIAM International Conference on Data Mining. vol. 2; 2002. p. 457–473.
Lakhal L, Stumme G. Efficient mining of association rules based on formal concept analysis. In: Formal
concept analysis: Foundations and Applications. Springer; 2005. p. 180–195.
46. Veroneze R, Banerjee A, Von Zuben FJ. Enumerating all maximal biclusters in numerical datasets.
Information Sciences. 2017; 379:288–309. https://doi.org/10.1016/j.ins.2016.10.029
47. Veroneze R, Von Zuben FJ. New advances in enumerative biclustering algorithms with online partition-
ing. arXiv preprint arXiv:200304726. 2020;.
48. Organization WH. Obesity: preventing and managing the global epidemic. World Health Organization;
2000.
49.
Fruchart JC, Santos RD, Aguilar-Salinas C, Aikawa M, Al Rasadi K, Amarenco P, et al. The selective
peroxisome proliferator-activated receptor alpha modulator (SPPARMα) paradigm: conceptual frame-
work and therapeutic potential. Cardiovascular diabetology. 2019; 18(1):71. https://doi.org/10.1186/
s12933-019-0864-7 PMID: 31164165
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
21 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
50. Halcox JP, Banegas JR, Roy C, Dallongeville J, De Backer G, Guallar E, et al. Prevalence and treat-
ment of atherogenic dyslipidemia in the primary prevention of cardiovascular disease in Europe: EUR-
IKA, a cross-sectional observational study. BMC cardiovascular disorders. 2017; 17(1):160. https://doi.
org/10.1186/s12872-017-0591-5 PMID: 28623902
51. Corbi SCT, Bastos AS, Nepomuceno R, Cirelli T, Santos RAd, Takahashi CS, et al. Expression profile
of genes potentially associated with adequate glycemic control in patients with type 2 diabetes mellitus.
Journal of diabetes research. 2017; 2017. https://doi.org/10.1155/2017/2180819 PMID: 28812028
52. Robinson JG, Wang S, Smith BJ, Jacobson TA. Meta-analysis of the relationship between non–high-
density lipoprotein cholesterol reduction and coronary heart disease risk. Journal of the American Col-
lege of Cardiology. 2009; 53(4):316–322. https://doi.org/10.1016/j.jacc.2008.10.024
53. Virani SS. Non-HDL cholesterol as a metric of good quality of care: opportunities and challenges. Texas
Heart Institute Journal. 2011; 38(2):160. PMID: 21494527
54. Brunner FJ, Waldeyer C, Ojeda F, Salomaa V, Kee F, Sans S, et al. Application of non-HDL cholesterol
for population-based cardiovascular risk stratification: results from the Multinational Cardiovascular
Risk Consortium. The Lancet. 2019; 394(10215):2173–2183. https://doi.org/10.1016/S0140-6736(19)
32519-X PMID: 31810609
55. Nepomuceno R, Vallerini BdF, da Silva RL, Corbi SC, Bastos AdS, dos Santos RA, et al. Systemic
expression of genes related to inflammation and lipid metabolism in patients with dyslipidemia, type 2
diabetes mellitus and chronic periodontitis. Diabetes & Metabolic Syndrome: Clinical Research &
Reviews. 2019; 13(4):2715–2722. https://doi.org/10.1016/j.dsx.2019.07.003
56. Weizmann Institute of Science. GeneCards: The Human Data Base; 2020. https://www.genecards.org.
57. Rzehak P, Covic M, Saffery R, Reischl E, Wahl S, Grote V, et al. DNA-methylation and body composi-
tion in preschool children: epigenome-wide-analysis in the European Childhood Obesity Project
(CHOP)-Study. Scientific reports. 2017; 7(1):1–13. https://doi.org/10.1038/s41598-017-13099-4 PMID:
29084944
58. Wallace C, Rotival M, Cooper JD, Rice CM, Yang JH, McNeill M, et al. Statistical colocalization of mono-
cyte gene expression and genetic risk variants for type 1 diabetes. Human molecular genetics. 2012; 21
(12):2815–2824. https://doi.org/10.1093/hmg/dds098 PMID: 22403184
59. Yuasa K, Aoki N, Hijikata T. JAZF1 promotes proliferation of C2C12 cells, but retards their myogenic dif-
ferentiation through transcriptional repression of MEF2C and MRF4—Implications for the role of Jazf1
variants in oncogenesis and type 2 diabetes. Experimental cell research. 2015; 336(2):287–297. https://
doi.org/10.1016/j.yexcr.2015.06.009 PMID: 26101156
60. Davegårdh C, Broholm C, Perfilyev A, Henriksen T, Garcı´a-Calzo´ n S, Peijs L, et al. Abnormal epige-
netic changes during differentiation of human skeletal muscle stem cells from obese subjects. BMC
medicine. 2017; 15(1):39. https://doi.org/10.1186/s12916-017-0792-x PMID: 28222718
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
22 / 22
PLOS ONE
|
Could not heal snippet
|
10.1186_s12889-023-15632-9.pdf
|
Data availability
All datasets used for supporting the conclusions of this paper are available
from the corresponding author on request.
|
Data availability All datasets used for supporting the conclusions of this paper are available from the corresponding author on request.
|
Joseph et al. BMC Public Health (2023) 23:748
https://doi.org/10.1186/s12889-023-15632-9
BMC Public Health
Who are the vulnerable, and how do we reach
them? Perspectives of health system actors
and community leaders in Kerala, India
Jaison Joseph1*, Hari Sankar1, Gloria Benny1 and Devaki Nambiar1,2,3
Abstract
Background Among the core principles of the 2030 agenda of Sustainable Development Goals (SDGs) is the call
to Leave no One behind (LNOB), a principle that gained resonance as the world contended with the COVID-19
pandemic. The south Indian state of Kerala received acclaim globally for its efforts in managing COVID-19 pandemic.
Less attention has been paid, however, to how inclusive this management was, as well as if and how those “left
behind” in testing, care, treatment, and vaccination efforts were identified and catered to. Filling this gap was the aim
of our study.
Methods We conducted In-depth interviews with 80 participants from four districts of Kerala from July to October
2021. Participants included elected local self-government members, medical and public health staff, as well as
community leaders. Following written informed consent procedures, each interviewee was asked questions
about whom they considered the most “vulnerable” in their areas. They were also asked if there were any special
programmes/schemes to support the access of “vulnerable” groups to general and COVID related health services, as
well as other needs. Recordings were transliterated into English and analysed thematically by a team of researchers
using ATLAS.ti 9.1 software.
Results The age range of participants was between 35 and 60 years. Vulnerability was described differentially by
geography and economic context; for e.g., fisherfolk were identified in coastal areas while migrant labourers were
considered as vulnerable in semi-urban areas. In the context of COVID-19, some participants reflected that everyone
was vulnerable. In most cases, vulnerable groups were already beneficiaries of various government schemes
within and beyond the health sector. During COVID, the government prioritized access to COVID-19 testing and
vaccination among marginalized population groups like palliative care patients, the elderly, migrant labourers, as
well as Scheduled Caste and Scheduled Tribes communities. Livelihood support like food kits, community kitchen,
and patient transportation were provided by the LSGs to support these groups. This involved coordination between
health and other departments, which may be formalised, streamlined and optimised in the future.
Conclusion Health system actors and local self-government members were aware of vulnerable populations
prioritized under various schemes but did not describe vulnerable groups beyond this. Emphasis was placed on
the broad range of services made available to these “left behind” groups through interdepartmental and multi-
*Correspondence:
Jaison Joseph
[email protected]
Full list of author information is available at the end of the article
© The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use,
sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and
the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included
in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The
Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available
in this article, unless otherwise stated in a credit line to the data.
RESEARCHOpen Access Page 2 of 11
stakeholder collaboration. Further study (currently underway) may offer insights into how these communities –
identified as vulnerable – perceive themselves, and whether/how they receive, and experience schemes designed for
them. At the program level, inclusive and innovative identification and recruitment mechanisms need to be devised
to identify populations who are currently left behind but may still be invisible to system actors and leaders.
Keywords Vulnerable Population, Health Equity, Sex Differences, Universal Health Coverage, Primary Health Care,
Health Systems, Primary Care Cost, Primary Care Utilization
Introduction
The core aim of the 2030 agenda of Sustainable Develop-
ment Goals (SDGs) is to bring in transformation through
Sustainable Development which requires nations to
Leave no One behind (LNOB) [1]. Populations left
behind are defined as being “at greater risk of poor health
status and healthcare access, who experience significant
disparities in life expectancy, access to and use of health-
care services, morbidity and mortality” [2]. These popu-
lations sometimes experience multiple morbidities which
results in complex health care needs which are further
exacerbated by intersecting deleterious social and eco-
nomic conditions [2]
Globally, each nation has the prerogative to define
“left behind” groups or communities based on the social,
economic, cultural and political factors, which in turn
may vary across geographies subnationally [3]. In India,
groups face vulnerability or marginalization on the basis
of age, disability, socio-economic status, which in turn
restricts the access of these communities to health and
healthcare [4]. Groups that are officially considered vul-
nerable in India according to the country’s main think
tank, the NITI Aayog, include persons who are clas-
sified as those in Scheduled Castes (SCs), Scheduled
Tribes (STs), Other Backward Classes (OBCs), Economi-
cally Backward Classes (EBCs), Religious Minorities,
Nomadic, Semi-Nomadic and De-Notified Tribes (NT,
SNT & DNTs), people who work in sanitation, known
in Hindi as Safai karmacharis (SKs), Senior Citizens/
the elderly, Transgendered persons, Persons engaging in
Substance Abuse, as well as those who are destitute and
involved with begging[4–6]These population subgroups
are prioritised for various government welfare schemes.
Across the country, participation of under-represented
groups in planning an decision-making is instituted
through affirmative action: SC, ST and Other Back-
ward Classes (OBCs) are provided reservations in public
service.
In the health domain, Below Poverty Line (BPL) house-
holds are covered under Ayushman Bharat Pradhan
Mantri Jan Arogya Yojana (AB-PMJAY) providing insur-
ance coverage in the amount of 500,000 INR (~ 6,050
USD) per family for secondary and tertiary care hospi-
talization expenditure through empanelled health care
providers [7, 8]. In the Southern Indian state of Kerala,
Ayushman Bharat benefits are extended to a broader
beneficiary group, comprising Mahatma Gandhi National
Rural Employment Guarantee Act (MGNREGA) house-
holds, households of unorganized workers and additional
population subgroups recognised as facing disadvantage
by the state.
Kerala has the lowest level of multidimensional poverty
according to the NITI Aayog, which suggests that the
population of “vulnerable” may be relatively lower in this
setting [9]. Overall, this bears out: the state’s develop-
ment pattern also indicates relatively low inequalities in
health and education outcomes [10]. The state nonethe-
less takes seriously the process of identifying and cater-
ing to “vulnerable” population groups. It has a range of
programmes for people recognised as having Scheduled
Caste (SC) and Scheduled Tribe (ST) status, women,
children, elderly and persons living with disabilities
[11]. We identified no less than around 35 schemes and
population-specific programs introduced by the state in
the past half decade to support groups facing disadvan-
tage: these include earmarked funds, subsidy schemes,
as well as reservations in education and employment
[3, 12]. Health programs have also been put in place by
non-health departments and agencies. For example, the
Scheduled Tribes Development Department implements
many programs to address the general healthcare needs
of tribal populations, which include allopathic health
care institutions, medical reimbursement through hos-
pitals, a tribal relief fund for emergency expenditure,
assistance for sickle-cell anaemia patients, assistance to
traditional tribal healers and mobile medical units [13].
One of the objectives of the Health and Family Welfare
Department’s recently launched Aardram mission was
to improve access of marginalized/vulnerable popula-
tions to comprehensive health services [14]. The state
is also implementing free health insurance scheme
called “Awaz” for interstate migrant workers, covering
Rs.15,000/- (~ 181.82 USD) for medical treatment per
year and an amount Rs.200,000/- Lakhs (~ 2424 USD) for
accident deaths [15]
Although the state has several welfare measures and
schemes to improve healthcare access for vulnerable
groups, challenges remain. For one, impoverishment due
to health is a major barrier that disproportionately affects
those already facing marginalisation: such groups cannot
rely on the public sector for services and end up impov-
erished due to health expenditures in the private sector
Joseph et al. BMC Public Health (2023) 23:748 Page 3 of 11
[16]. In fact, high Out-of-Pocket-Expenditure (OOPE)
and rising health care cost for hospitalization have
resulted in reducing health seeking [17]. Vulnerabilities
therefore, are changing almost continuously. This makes
the task of identifying vulnerable groups difficult – given
the dynamic, complex, historically, and contextually con-
tingent nature of vulnerability [18]. And yet, both global
and national goals call for identification, responses and
monitoring of outcomes in these population groups [1,
19].
As part of a larger health systems study, we placed
emphasis on how vulnerability is defined in the state,
and how vulnerabilities are addressed through schemes
and equity-oriented reforms introduced in the state. It is
important to understand the perspective of primary care
health system actors on vulnerability and who are vul-
nerable, as they are at the forefront of delivering essen-
tial health care services and identification and catering
to the needs of vulnerable population. Such an exercise
has been carried out, for example in other regions with
the support of the World Health Organization, [20]. as
well as in other projects focused on equity integration
in health programming and planning [21–23]. Barring a
rare example published in 2015 [24], we were not able to
identify such initiatives or studies in the Indian context,
particularly ones that viewed “vulnerability” and efforts
at inclusion from an implementer’s perspective. Seek-
ing to fill this gap, we undertook a qualitative analysis
of perspectives from Kerala’s health system actors, local
self-government representatives and community leaders
involved with Primary Healthcare Reforms (PHCR) in
Kerala about their definitions and understandings of who
is vulnerable in the state, what is being done to address
their vulnerabilities, both within and outside of the con-
text of COVID-19.
Methods
This study is the qualitative component of a larger health
system research study in Kerala; our detailed methodol-
ogy is reported elsewhere[25]. In summary, Kerala’s 14
districts were grouped into four categories using princi-
pal components analysis, using indicators from the fourth
round of the National Family Health Survey (NFHS)
(2015–16) [26]. One district was randomly selected from
each group, within which catchment areas served by two
randomly selected primary health facilities (one recently
upgraded by Aardram and one slated for later upgrada-
tion) were also randomly selected.
In-depth interviews (IDIs) were carried out in the
four selected districts between July and October 2021.
Participants for this study were staff from two primary
healthcare facilities per district and elected representa-
tives from their corresponding Local Self Governments
(LSGs). We adopted purposive criterion sampling
technique for the selection and recruitment of study
participants. For the identification and selection of par-
ticipants we employed a two-pronged strategy. As an ini-
tial step we line-listed the potential health system actors
(HSAs) and community leaders who could be part of
this study. From each facility we enrolled HSAs includ-
ing medical and public health staff, community leaders
and Local Self Government representatives to obtain a
comprehensive HSAs perception of vulnerable popula-
tion their area. Medical and public health staff included,
Medical Officer (MO), Staff Nurse/Nursing Officer,
Health Inspector (HI), Junior Health Inspector (JHI),
Public Health Nurse (PHNs), Junior Public Health Nurse
(JPHNs), Palliative Care Nurse and Accredited Social
Health Activists (ASHAs). Community members eligible
for recruitment included Panchayat Presidents and Vice
Presidents, Health Standing Committee member and
Ward Members. We identified additional community
leaders from these areas through the HSAs, LSG mem-
bers and non-governmental organizations to capture the
perspective of the community. On an average we enrolled
10 HSA per facility, a total of 83 HSAs were contacted for
this study and three of them could not participate due to
their busy schedule.
The Institutional Ethics Committee of the George Insti-
tute for Global Health (Project Number 05/2019) issued
ethical approval for this study. In each facility area, in-
depth interviews for this study were carried out by three
researchers trained in qualitative research methods (HS,
JJ & GB). The research team comprised of two male
research fellows and a female research assistant and was
supervised by a senior health systems researcher (DN).
Administrative approval was taken from the Depart-
ment of Health and Family Welfare, Government of Ker-
ala. The team met the District Medical Officers (DMO)
of four districts, shared the departmental permissions,
outlined the study objectives, and shared findings of an
earlier primary survey carried out in the same catch-
ment areas. After the permissions were issued from the
DMOs, the team of three researchers (HS, GB, JJ) took
appointments with Medical Officers and briefed them
about the study and sought their permission for conduct-
ing IDIs with the staff under their institutions. Further,
each of the HSAs were met in person and appointments
for interviews were sought based on their convenience.
As per their convenience IDIs were carried out in-person
or through online platforms (i.e. Zoom). For carrying out
the IDIs with LSG representatives, the team met with the
panchayat presidents of the respective LSGs and briefed
on the purpose of study and sought their permission
to carry out the IDIs with other identified LSG mem-
bers. Community leaders were contacted over phone, to
brief them on the purpose of the study and as per their
Joseph et al. BMC Public Health (2023) 23:748 convenience the researcher met them in person to carry
out the interviews.
All the participants were handed over with a hard copy
of the topic guides and Participant Information Sheet
(PIS) in English and Malayalam before the in-person
interviews. Each participant’s signed informed consent
was taken for participating in the study and for record-
ing interviews. For those interviews conducted over
online platforms, a soft copy of the topic guide, PIS and
consent form were shared in advance with the partici-
pants. Before commencing the interview, the participants
shared the dully signed consent form with the research-
ers. Malayalam was the medium of conversation and
each of the IDIs lasted between 20 and 60 min. To obtain
context and perspectives of HSAs in various capacities
and geographies pertaining to each of the study sites
across four districts the interviews with all the pre-set list
of participants were completed even though achieving
early data saturation was reached with some of the study
topics.
Three participants could not participate in the inter-
view due to their busy schedules and after multiple failed
attempts to schedule, we decided to remove them from
the study. All IDIs were recorded; interview record-
ings and field notes were stored and secured in a pass-
word protected database after the completion of each
interview and were accessible only to the research team
members. Recordings were transliterated into English by
a third-party agency empanelled by The George Institute
for Global Health, India, which signed confidentiality
agreements prior to accessing data. All the transliterated
transcripts were reviewed by a three-member research
team to ensure quality.
Table 1 Participant characteristics
Category
Local Self
Government
Representatives
Health System
Actors
Designation
Panchayat President
Panchayat Vice-President
Health Standing Com-
mittee Member
Ward Member
Community Leader
Medical Officer
Health Inspector (HI)
Public Health Nurse
(PHN)
Junior Health Inspector
(JHI)
Junior Public Health
Nurse (JPHN)
Nursing Officer
Palliative Nurse
Community Health
Worker
Total Participants
Female
3
0
3
0
1
5
1
4
0
11
3
1
16
48
Male
4
1
5
Total
7
1
8
1
6
3
5
7
0
0
0
0
1
7
8
6
4
7
11
3
1
16
32
80
Page 4 of 11
Transliterated transcripts were thematically analysed
using ATLAS.ti 9 software by a four-member research
team (DN, HS, JJ, GB). An inductive approach was used:
the thematic structure and code book were finalized
after multiple discussions among the four-member team.
Finally, the coded manuscripts from the team members
were merged using ATLAS.ti 9 software. Codes of inter-
est for this analysis were indexed and themes consoli-
dated based on further discussions and core questions of
interest (i.e., who is left behind? How are they reached?
and impact of COVID-19 among those left behind). A
narrative was then constructed around these questions
and compiled by the lead author with inputs, edits, and
review by other authors.
Results
Participant characteristics
Data for a total of 80 participants was included in the
study, of which more than half (60%) were women (see
Table 1). From this group of participants, we received
information on who they considered was being left
behind from health programming in Kerala, as well as
what was being done to support them and/or address
their needs (in general, and in the COVID context).
Who is left behind?
Participants in all districts would often first identify
Scheduled Caste and Scheduled Tribe communities
as vulnerable; these are nationally established catego-
ries defined as facing vulnerability. Apart from this, we
observed geographical variation across districts in who
was described as vulnerable population by stakeholders
(see Table 2). Migrant labourers were identified as vul-
nerable in the semi-urban areas, while fisherfolk in the
coastal areas (inland and seafaring).
It was found that most of the places where the vulner-
able population were identified, faced challenges related
to living and working conditions - social determinants of
health like sanitation, nutrition, crowding/housing were
raised. According to a Medical Officer,
…there is the SC/ST community- they have colo-
nies1here… they have drinking water issues, food
issues, improper waste management, and crowded
places. It is a dengue hotspot and communicable
diseases (hotspot). Also, COVID is a big issue there,
1 While system actors often mentioned colonies of SC and ST communi-
ties, in subsequent fieldwork, SC communities in particular felt offended by
the label of “colony” used to describe their places of residence. This could be
seen as being akin to what Wacquant has called “territorial stigma,” which
automatically assigns ignominy to a geographic category.(27) Although
Wacquant’s theorization referred to the urban context in Chicago and Paris
alone, we saw resonance of the concept for urban and rural residents of “col-
onies.” The concept of the “colony,” of course, has other problematic histories
and legacies.
Joseph et al. BMC Public Health (2023) 23:748 Page 5 of 11
This view was held by another JPHN as well who took
the view that
There are no marginalised communities in my area.
All the people here are from similar backgrounds
since it is a coastal area. I do not know if they have
any issues. Most of the people over there depend
on their daily income and even when they must
undergo quarantine, the authorities have delivered
them essential commodities and resolved the prob-
lems that came up. So, there were no issues, all such
troubles were taken care of.
Programs to support those left behind
We found that schemes and programmes targeting vul-
nerable populations were being implemented across the
state in most cases. The possible exception we found was
the case of fisherfolk and farmers, who were defined as
vulnerable, but were not described as being covered
by many government health schemes. Recently imple-
mented primary health care reforms had reportedly
improved access to healthcare for vulnerable groups in
some areas. In many cases this involved interdepartmen-
tal coordination. A Panchayat president took the follow-
ing view:
Our Family Health Centre works from 7 AM till 8
PM even now. The service of a gynaecology specialist
is provided twice a week. Then, we have an eye spe-
cialist. We have been getting the services of a phys-
iotherapy specialist. People from the rural areas,
including the Adivasi community, were able to ben-
efit from these changes. The Tribal Department has
been conducting camps in the places where Adivasis
[tribal persons] live
According to a Health Inspector, there was empha-
sis placed on going to where communities were to offer
them care/support and the role of labour department and
private employers in health service delivery:
We have a lot of migrants around here. The labour
office is holding special camps for them. Their
employers also sometimes book slots in bulk and get
the workers vaccinated. As far as we are concerned,
we go to their companies and conduct tests and pro-
vide other services there.
We also found that joint programs implemented by
LSGs and the Department of Social Justice, such as the
Table 2 Vulnerable Population Identified by Participants across
Districts
Thiruvananthapuram Kollam Alappuzha
X
X
X
People
from
Sched-
uled
Tribe
People
from
Sched-
uled
Caste
Pal-
liative
Care
pa-
tients
Fisher-
folk
Farm-
ers
Mi-
grants
X
X
X
X
X
(inland)
X
(seafaring)
X
X
Kasara-
god
X
X
X
X
X
because if it affects one person, the spread will be too
much…because even the children run around and
enter all the houses.
We also found that climate change (subsequent floods in
the state) and COVID-19 pandemic had affected popu-
lation subgroups and added to their vulnerability. Farm
workers were affected by the consequent floods in the
state and fisherfolk were affected by the COVID-19 pan-
demic. One Community Leader noted this:
…Especially when there were floods, farm work-
ers were there…. the one who is mostly engaged with
paddy fields. Last financial year was a time when
the yield was maximum but there was a technical
difficulty in harvesting it. During such a situation,
the farmers had to face a lot of trouble.People turn
out to be marginalised when they cannot har-
vest their crop. The situation is similar in the case
of fisheries as well. Due to COVID, they could not
go fishing for several days. Even if they went, there
was a situation that people turned COVID posi-
tive because there were about 40 people in a fishing
boat...
On the other hand, a few people we spoke to also men-
tioned that nobody was vulnerable, because the needs of
all were catered to, as per need. A Junior Public Health
Nurse said: “I don’t think such a marginalised community
exists anymore in this era. We all are equal. I do not think
any community is being sidelined nowadays.”
Joseph et al. BMC Public Health (2023) 23:748 Page 6 of 11
Kudumbasree2-self help program for women, as well as
programs focussing on the elderly population, migrants,
destitute and palliative care patients were intended to
increase access to healthcare and to improve quality
of life for groups facing these forms of disadvantage. A
Health Standing Committee Member added:
…for palliative patients, we provide support from
Panchayat and the FHC. Other than this, we have
a scheme called Ashraya for the destitute. We pro-
vide them with kits through Kudumbasree. We have
another scheme called Santhwanam. Under this,
through Kudumbasree we conduct an event once a
year. Ashraya scheme falls under the ambit of this
one. Ashraya is for people with no means of support.
According to a Community Health worker, the Panchayat
placed emphasis on palliation and also on the health and
welfare of guest or migrant workers:
Yes, Panchayat provides it. Even medicines and
hospital-related services are arranged by the Pan-
chayat. Similarly, the Panchayat has appointed a
nurse for palliative care. We visit their homes along
with the palliative nurse and provide all possible
services to them. If any guest workers come here, we
treat them like our own people, and both the Pan-
chayat and the FHC provide them with all kinds of
assistance.
This was corroborated by a Panchayat President in
another district as well:
We have proper facilities for ensuring the health of
people including migrant labourers. …. Grama Pan-
chayat has facilitated the treatment for numerous
cancer patients in the area as well as for those with
other related diseases. The area has around 250 pal-
liative patients. We have implemented various pro-
grams for helping all such patients.
There was seen to be, therefore, responsibility taken by
local leaders for vulnerable groups and the idea that these
were “our own people,” whose needs related to health and
beyond, were given due attention.
COVID Outreach for vulnerable populations
Many study participants felt that during the COVID-19
pandemic and consequent lockdowns, vulnerable popu-
lations were prioritised. Various health service design
2 Kudumbashree is the poverty eradication and women empowerment pro-
gramme implemented by the State Poverty Eradication Mission (SPEM) of
the Government of Kerala.[28]. More information is available at: https://
www.kudumbashree.org.
changes were described as being introduced to ensure
the delivery of essential health care and related services
under the stewardship of LSGs. A Junior Public Health
Nurse described them as follows:
We used to provide food to these side-lined people
from the community kitchen, and provide medicines
from our Tele-OP [out-patient services], when the
first wave of COVID started. When COVID started
and there were strict lockdowns, from the side of the
health department, every day there was one or two
vehicles that were arranged from the side of LSGD
and in that vehicle, our staff would take details from
each area of the positive cases, and create a calcu-
lation on how many of them need medicine, and
how many homes we need to put a sticker etc, and
both these vehicles would cover two different areas
without overlapping and delivered, medicine kit is,
NCD medicines and Tele OP medicines everywhere
promptly.
Another Panchayat President noted the greater risks of
exposure in certain populations and how they were pri-
oritised commensurably, saying that “we have distributed
kits in every ward. Due to COVID and lockdown, people
were not able to go outside so we distributed kits to every-
one. We especially distributed masks and sanitisers in the
S[cheduled] T[ribe] colonies and other marginalised colo-
nies. Because they were residing in a densely populated
area and there is a high chance of spreading, we provided
the kits.”
A Nursing Officer also noted the role played by pan-
chayat leaders in mobilising support during lockdowns,
“when migrants could not go back to their homes, vol-
unteers intervened and helped them. Whatever needed,
from food to shelter was provided from the side of the
Panchayat.”
Vulnerable populations were prioritized for receiv-
ing COVID-19 vaccinations. There were efforts from the
health systems and LSGs to deliver vaccines at the door-
steps of these population. A community health worker
described how separate, priority vaccination drives were
held for fisherfolk, SC and ST groups. She said simply:
“They were given more preference.” A Medical Officer
noted that in their area, SC, ST, persons living with dis-
abilities and migrants were the first to achieve complete
vaccination. This was echoed by a frontline worker in
another district who noted that
Bedridden patients were given vaccination doses
at their houses. Palliative patients were given
the vaccination at their places. We have also vac-
cinated people above 80 years of age after visiting
their houses. We visited the houses of those who
Joseph et al. BMC Public Health (2023) 23:748 Page 7 of 11
cannot come and got them inoculated. We also con-
duct health camps in colonies. A class on vaccina-
tion programs was also given for them and all these
were organised by the PHC.
Discussion
Our study sought to identify who was defined as vul-
nerable by health system and LSG actors in the state of
Kerala and what schemes and arrangements were in
place to address their health issues. In the current study,
we observed that a number of groups identified at the
national level as vulnerable were also identified by our
study participants, alongside other population groups
that were uniquely identified in Kerala. This is consis-
tent with the findings of Kerala State Poverty Eradication
Plan presented to NITI Aayog, which reported that SC
populations were concentrated in colonies (including in
urban areas), ST populations continued to be sequestered
in remote and rural locations, consistent with nationally
identified groups in need [29]. However, this report also
indicated the need to support coastal populations like
fisherfolk who for economic reasons were also confined
to particular, hard to reach geographies [29]. Decen-
tralized planning in Kerala has helped keep the issue of
inclusion and marginalisation on the agenda of decision-
makers and implementers, even as newer groups facing
vulnerability were being identified, like migrant workers
[11]. Migrant workers also faced confinement in their
work settings, while palliative care patients were confined
due to their health situation. This distance – physical or
social – was a defining feature of vulnerability from the
perspective of these supply side actors. This kind of a dis-
tance based vulnerability has been found in a national
studies from Uttar Pradesh, Madhya Pradesh, Bihar
Assam and Jharkhand during pre and post COVID-19
periods [30], although the view of health system actors
or decision-makers on this was not specifically indicated
in the literature. Other studies in LMICs have identi-
fied vulnerability on the basis of racial, ethnic and gen-
der minoritization, economically disadvantage, having
chronic health issues, as well as those at extremes of age
[1, 31, 32]
It was also observed that it was not merely in the con-
text of health, but the larger social determinants that vul-
nerable populations were “hard to reach.” The residential
areas of the marginalized population were underdevel-
oped: providing quality health service delivery remained
challenging without addressing the social determinants
of health. This is consistent with the findings of the 6th
Kerala Administrative Reforms Commission report
(2020) which noted lack of land, improper housing, inad-
equate infrastructure, poor quality of education, lack of
sanitation services and unsafe drinking water among the
marginalized population [33]. This report also gave spe-
cial emphasis on the condition of SC and other “back-
ward” communities who continue to live and work in
highly dangerous and pathogenic conditions [33]. It has
been deemed vital to address social determinants among
the marginalized to improve their health status as they
are important factor in management and prevention of
communicable and non-communicable diseases alike
[34]. Studies conducted in LMICs have reported lower
access to safe drinking water, sanitation, and hygiene
(WASH), conditions which are fundamental to living and
working, are both reflective of vulnerability and are what
drive disparities in health burdens, health seeking, and
health outcomes [35–37]
We found that natural disasters (floods) and COVID-19
pandemic added to the vulnerabilities faced by farmers
and fisherfolk, suggesting that vulnerability is not a static
phenomenon. A study conducted by a panel of experts
in Kerala immediately after the 2018 floods reported
that the vulnerable population who were the victims of
floods lagged behind their peer groups in levels of human
development, in part because they faced differential and
layered exposures and vulnerabilities compared to other
groups [38]. Another study by the Palliative Care Con-
sortium on the effect of 2018 floods on elderly living
alone found serious after effects of the disaster especially
among the elderly women, also the palliative care ser-
vices and medications were disrupted [39]. COVID-19
lockdowns imposed by the Government during the first
wave (2020) affected the coastal community in the state
in accessing healthcare and in resourcing the essential
commodities. Along with it the declaration of some of
the overcrowded coastal regions as containment zones,
with restriction of movement leading to reduced work-
ing hours and income further increased their vulner-
ability [40]. A study conducted by Kattungi et al. (2020)
assessing the impact of COVID-19 on the livelihood
of fishermen in Puducherry found loss of employment
among many fishermen which has resulted in increas-
ing inequities and poverty [41]. Aura CM et al. (2020),
in their study which assesses the consequences of flood-
ing and COVID-19 Pandemic among inland fisherfolk in
Kenya in East Africa, found that natural calamities and
pandemic affected the livelihood of fisherfolk, reduced
fishing time and trips, decline in consumables such as
boat fuel resulting low fish catches etc [42, 43]. COVID-
19 has negatively affected small scale farmers in LMICs
which resulted in low production, low income and higher
food insecurity which has increased their vulnerability
[44, 45]
There has been a fairly high degree of multisectoral
action and coordination to reaching the “vulnerable” in
Kerala. We found a fascinating convergence in the views
of those who identified vulnerable groups and those
Joseph et al. BMC Public Health (2023) 23:748 Page 8 of 11
who did not. Both noted that schemes existed and that
vulnerable groups (or everyone!) were taken care of the
state through schemes implemented by government
departments. This includes multisectoral action led by
the State government in prevention and control of Non-
communicable Diseases (NCDs) [46, 47], convergence
to support awareness of and enrolment in the Depart-
ment of Labour’s health insurance scheme (supported
greatly by LSG leaders and Kudumbasree mission work-
ers under Department of Social Justice), [48]. as well as
other schemes introduced by the Kerala Social Security
Mission [49–51]
The state’s response in handling the COVID-19 pan-
demic was another example of multi-sectoral coordina-
tion backed by decentralized governance, along with
whole of society approaches where community action
complemented the work of health system actors [52,
53]. During COVID-19, a community kitchen initia-
tive was introduced through LSGs with the support of
Kudumbasree, which provided free meals to labourers,
people who were under quarantine, the destitute and
other needy marginalized population [54]. Grassroots
agencies were also involved with delivering free food
kits universally, which required a special focus on vul-
nerable population typically excluded from social secu-
rity benefit programmes like transgender persons [53].
In a scoping review by Hasan et al. (2021) about the
response of LMICs in management of COVID-19 found
that decentralized governance coupled with stewardship
and multisectoral collaboration facilitated the delivery of
integrated health service delivery[55] ,which was found
through our study in Kerala.
Another interesting feature in Kerala was seen dur-
ing COVID-19 in the context of vaccination. Initially
COVID-19 vaccination in Kerala followed global norms
by prioritising health workers followed by frontline work-
ers [56], then national norms prioritising citizens above
the age of 60 years and citizens aged between 45 and 59
with specified comorbidities [57]. However, by April 2021
Kerala created state specific norms by way of 32 prior-
ity categories in the age group of 18–45 which included
other frontline workers, seafarers, field staff, teachers,
students and more [58]. This demonstrates the possibility
of defining and redefining those in need in the context of
a crisis. It is less clear, however, if such prioritization of
populations in need could be done on an ongoing basis,
helping the state to identify those who may face unique
disadvantages and may need to be reached by program-
ming beyond the existing ambit. This is a clear area for
further research.
Beyond this, there are other areas warranting further
research: greater attention to how multi-sectoral policy
processes for the “vulnerable” take place, in what con-
texts, could offer lessons for their replication in other
contexts, and also for their enhancement in Kerala.
Moreover, it is unclear, at present, how intersections of
vulnerability may be addressed in current programming,
for e.g. SC or ST populations receiving palliative care,
women involved with the fishing industry. Whether or
not such programs are catering to these intersectional
needs would be a critical area for future policymaking.
Finally, there is a very little understanding of those fac-
ing vulnerability as being more than “target populations”
or “beneficiaries” of services. Other research on UHC
has shown that just producing interventions and consid-
ering communities passive recipients can easily alienate
and exclude them from health reform processes[59]. Fur-
ther study is needed – across all these and more groups
facing vulnerability – on how they perceive themselves,
and how they receive, and experience schemes designed
for them, and in the absence of such schemes, how they
manage their health and related needs. This would have
to be given more attention in research and policymaking
and is a limitation in the framing of our study as well.
Limitations
This analysis is based on the perceptions of government
health system actors. It therefore does not include the
perceptions of the general population as well as those
who constitute “those left behind.” Research is currently
underway to understand the care seeking experiences of
these, “demand side” actors and is a crucial part of our
understanding of vulnerability.
Conclusion
Our analysis sought to understand supply side perspec-
tives in the health sector on who is left behind in the
southern Indian state of Kerala. Health system actors and
local self-government members were aware of vulnerable
population prioritized under various schemes but did not
describe vulnerable groups beyond this. Emphasis was
placed on the broad range of services available to these
“left behind” groups. Further study (currently underway)
may offer insights into how these communities – identi-
fied as vulnerable – perceive themselves, and how they
receive, and experience schemes designed for them.
Innovative sampling and recruitment mechanisms need
to be devised to identify populations who are currently
left behind but may also be invisible to system actors and
leaders.
While the Kerala government has shown initiative
in carrying out a mapping of poorest households in the
state, there are other critical forms of vulnerability that
affect residents in the state; continuous monitoring of
“who is being left behind,“ in partnership with academic
and civil society institutions, could help enhance such
initiatives.
Joseph et al. BMC Public Health (2023) 23:748 List of abbreviations
SDGs
LNOB
SC
STs
OBCs
EBCs
SKs
BPL
AB-PMJAY
MGNREGA
OOPE
PHCR
IDIs
HSAs
FHC
LSG
MO
HI
JHI
PHN
JPHN
ASHAs
PIS
Sustainable Development Goals
Leave No One Behind
Schedule Caste
Schedule Tribes
Other Backward Castes
Economically Backward Castes
Safai Karmacharis
Below Poverty Line
Ayushman Bharat Pradhan Mantri Jan Arogya Yojana
Mahatma Gandhi National Rural Employment Guarantee Act
Out-of-Pocket Expenditure
Primary Health Care Reform
In-depth Interviews
Health System Actors
Family Health Centre
Local Self-Government
Medical Officer
Health Inspector
Junior Health Inspector
Public Health Nurse
Junior Public Health Nurse
Accredited Social Health Activists
Participation Information Sheet
Acknowledgements
We are grateful to Mr. Santosh Sharma, Research Fellow, The George Institute
for Global Health, India, for his key reflections and critical inputs.
Author contributions
Conceptualization: JJ Methodology: JJ, HS, DN Formal analysis and
investigation: JJ, GB Writing - original draft preparation: JJ, HS, GB Writing -
review and editing: JJ, HS, GB, DN Funding acquisition: DN Supervision: DN.
Funding
We wish to indicate that this work was supported by the Wellcome Trust/DBT
India Alliance Fellowship(https://www.indiaalliance.org) Grant number IA/
CPHI/16/1/502653) awarded to Dr. Devaki Nambiar. The funder had no role in
study design, data collection and analysis, decision to publish, or preparation
of the manuscript. The funder provided support in the form of salaries and
research materials and field work support for authors DN, HS, GB and JJ but
did not have any additional role in the study design, data collection and
analysis, decision to publish, or preparation of the manuscript. The specific
roles of these authors are articulated in the ‘author contributions’ section.
Data availability
All datasets used for supporting the conclusions of this paper are available
from the corresponding author on request.
Declarations
Ethics approval
of the study was received from the institutional ethics committee of George
Institute for Global Health (Project Number 05/2019). All participants gave
written informed consent before taking part in the study including Illiterate
participants in the survey who were read out and explained the consent
form in the local language. Thereafter, they were able to sign their names. The
ethics committee that approved the study also approved this procedure of
obtaining written informed consent from these participants. All methods were
carried out in accordance with relevant guidelines and regulations.
Consent to publish
Not applicable.
Competing interests
The authors declare no competing interests.
Author details
1The George Institute for Global Health, New Delhi, India
2Faculty of Medicine, University of New South Wales, Sydney, Australia
Page 9 of 11
3Prasanna School of Public Health, Manipal Academy of Higher Education,
Manipal, India
Received: 6 September 2022 / Accepted: 7 April 2023
References
1. United Nations Sustainable Development Group. Leave No One Behind
[Internet]. 2022 [cited 2022 Jun 9]. Available from: https://unsdg.
un.org/2030-agenda/universal-values/leave-no-one-behind
3.
2. No authors listed. Vulnerable Populations: Who Are They? The American Jour-
nal of Managed Care [Internet]. 2006 Nov 1 [cited 2022 May 30]; Available
from: https://www.ajmc.com/view/nov06-2390ps348-s352
Balan PP, George S, Kunhikannan TP, Marginalisation. and Deprivation Studies
in Multiple Vulnerabilities [Internet]. Thrissur, Kerala: Kerala Institute of Local
Administration; 2016 [cited 2022 Jun 1]. Available from: http://dspace.kila.
ac.in:8080/jspui/bitstream/123456789/276/1/Marginalisation%20and%20
Deprivation%20Studies%20in%20Multiple%20Vulnerabilities.pdf
Agarwal M. Vulnerable Groups in India: Status, Schemes, Constitution of India
[Internet]. 2022 [cited 2022 May 30]. Available from: http://www.legalser-
vicesindia.com/article/1079/Vulnerable-Groups-in-IndiaStatus,-Schemes,-
Constitution-of-India.html
Aayog N. Role of Social Justice & Empowerment Division [Internet]. NITI
Aayog; 2019 [cited 2022 Jul 9]. Available from: https://www.niti.gov.in/sites/
default/files/2019-08/Role-of-SJE-Division-in-NITI.pdf
4.
5.
6. George AA, Mechanisms. Laws, Institutions and Bodies for Vulnerable Sec-
7.
tions [Internet]. Clear IAS. 2020 [cited 2022 Jul 9]. Available from: https://www.
clearias.com/mechanisms-laws-institutions-bodies-vulnerable-sections/
Karunya Arogya Suraksha Padahathi, State Health Agency Kerala. Schedules
for Service Contract of AB-PMJAY-KASP [Internet]. Department of Health and
Family Welfare, Government of Kerala; 2020 [cited 2022 Jun 9]. Available from:
https://arogyakeralam.gov.in/wp-content/uploads/2020/03/Schedules.pdf
8. Government of India. DPE-GM-15/0001/2016-GM-FTS-5921, Brochure on
Reservation for SCs/STs and OBCs [Internet]. Ministry of Heavy Industries &
Public Enterprises; 2016 [cited 2022 Jun 9]. Available from: https://dpe.gov.in/
sites/default/files/Reservation_Brochure-2.pdf
Board SP. Govt. of Kerala. Economic Review 2021 [Internet]. Thiruvanantha-
puram; 2021 [cited 2022 Jun 9]. Available from: https://spb.kerala.gov.in/sites/
default/files/2022-03/ECNO_%20ENG_21_%20Vol_1.pdf
9.
10. Ghosh M. Regional Disparities in Education, Health and Human Develop-
ment in India. Indian Journal of Human Development [Internet]. 2011 Jan
[cited 2022 Sep 6];5. Available from: https://www.researchgate.net/publica-
tion/302964390_Regional_Disparities_in_Education_Health_and_Human_
Development_in_India
11. Marginalisation. and Deprivation Studies in Multiple Vulnerabilities.pdf [Inter-
net]. [cited 2022 Jun 1]. Available from: http://dspace.kila.ac.in:8080/jspui/
bitstream/123456789/276/1/Marginalisation%20and%20Deprivation%20
Studies%20in%20Multiple%20Vulnerabilities.pdf
12. Chatterjee CB, Sheoran G. Centre for Enquiry into Health & Allied Themes
(Mumbai I. vulnerable groups in India. Mumbai: Centre for Enquiry into
Health and Allied Themes; 2007.
13. Scheduled Tribes Development Department., Govt. of Kerala. Healthcare
Schemes [Internet]. 2022 [cited 2022 Jun 9]. Available from: https://www.
stdd.kerala.gov.in/healthcare-schemes
14. Aarogyakeralam National Health Mission. Aardram [Internet]. 2018 [cited
2022 Jun 9]. Available from: https://arogyakeralam.gov.in/2020/04/01/
aardram/
15. Staff Reporter. 5.13 lakh migrant workers registered under Awaz scheme. The
Hindu [Internet]. 2022 Mar [cited 2022 Jun 10]; Available from: https://www.
thehindu.com/news/national/kerala/513-lakh-migrant-workers-registered-
under-awaz-scheme/article65223711.ece
16. Haddad S, Baris E, Narayana D. Safeguarding the Health Sector in Times of
Macroeconomic Instability: Policy Lessons for Low- and Middle-Income
Countries [Internet]. African World Press & International Development
Research Centre; 2008 [cited 2022 May 30]. Available from: https://www.idrc.
ca/en/book/safeguarding-health-sector-times-macroeconomic-instability-
policy-lessons-low-and-middle
Joseph et al. BMC Public Health (2023) 23:748 Page 10 of 11
17. Mukherjee S, Haddad S. Social class related inequalities in household health
expenditure and economic burden: evidence from Kerala, South India.
International journal for equity in health. 2011 Jan7;10:1.
18. Gallardo M. Identifying vulnerability to poverty: a critical survey. J Economic
Surveys. 2018;32(4):1074–105.
19. Ministry of Health and Family Welfare, Government of India. National Health
Policy India [Internet]. New Delhi; 2017 [cited 2022 Aug 4]. Available from:
https://www.nhp.gov.in/nhpfiles/national_health_policy_2017.pdf
20. Solar O, Valentine N, Castedo A, Brandt GS, Sathyandran J, Ahmed Z et al.
Action on the social determinants for advancing health equity in the time
of COVID-19: perspectives of actors engaged in a WHO Special Initiative.
International Journal for Equity in Health [Internet]. 2023 Jan 24 [cited 2023
Feb 8];21(3):193. Available from: https://doi.org/10.1186/s12939-022-01798-y
21. Brixi H, Mu Y, Targa B, Hipgrave D. Engaging sub-national governments in
addressing health equities: challenges and opportunities in China’s health
system reform. Health Policy and Planning [Internet]. 2013 Dec 1 [cited 2023
Feb 8];28(8):809–24. Available from: https://doi.org/10.1093/heapol/czs120
22. Chanchien Parajón L, Hinshaw J, Sanchez V, Minkler M, Wallerstein N.
Practicing Hope: Enhancing Empowerment in Primary Health Care through
Community-based Participatory Research. American Journal of Community
Psychology [Internet]. 2021 [cited 2023 Feb 8];67(3–4):297–311. Available
from: https://onlinelibrary.wiley.com/doi/abs/https://doi.org/10.1002/
ajcp.12526
23. Chin MH, King PT, Jones RG, Jones B, Ameratunga SN, Muramatsu N, et al.
Lessons for achieving health equity comparing Aotearoa/New Zealand and
the United States. Health Policy. 2018 Aug;122(8):837–53.
24. Nambiar D, Muralidharan A, Garg S, Daruwalla N, Ganesan P. Analysing
implementer narratives on addressing health inequity through convergent
action on the social determinants of health in India. Int J Equity Health
[Internet]. 2015 Dec [cited 2021 Aug 2];14(1):133. Available from: http://www.
equityhealthj.com/content/14/1/133
25. Nambiar D, Sankar H, Negi J, Nair A, Sadanandan R. Field-testing of primary
health-care indicators, India. Bull World Health Organ. 2020 Nov;98(1):747–53.
26. Ministry of Health and Family Welfare, Govt. of India. National Family Health
Survey (NFHS-4) [Internet]. IIPS Mumbai. ; 2015 [cited 2023 Feb 9]. Available
from: http://rchiips.org/nfhs/factsheet_nfhs-4.shtml
Middle-Income Countries during the COVID-19 Pandemic through a Gen-
dered Lens. Int J Environ Res Public Health. 2022 Sep 20;19(19):11845.
38. Thummarukudy M, Peter B. Leaving no one behind: Lessons from the Kerala
disasters - India [Internet]. Perumbavoor, Kerala: Centre for Migration and
Inclusive Development; 2019 Oct [cited 2022 Jun 10]. Available from: https://
reliefweb.int/report/india/leaving-no-one-behind-lessons-kerala-disasters
39. Kerala State Disaster Management Authority, India P, Indian Associa-
tion of Palliative Care. Addressing the Needs of Elderly and People Under
Palliative Care in Disaster Management [Internet]. Trivandrum; 2019 Jun
[cited 2022 Jun 10]. Available from: https://sdma.kerala.gov.in/wp-content/
uploads/2020/11/Report-Palliative-Care-Consultation-cum-Workshop.pdf
40. Sumitha TS, Thelly AS, Medona B, Lijimol AS, Rose MJ, Rajagopal MR.
Response to COVID-19 Crisis with Facilitated Community Partnership among
a Vulnerable Population in Kerala, India - A Short Report. Indian J Palliat Care.
2022;28(1):115–9.
41. Sekhar KV, Sutha DAI, Devi DRU, IMPACT OF COVID-19 ON THE LIVELIHOODS
OF FISHERMEN COMMUNITY IN YANAM., PUDUCHERRY: AN ANALYSIS. Euro-
pean Journal of Molecular & Clinical Medicine. 2020 Nov 30;7(8):869–80.
42. Aura CM, Nyamweya CS, Odoli CO, Owiti H, Njiru JM, Otuo PW, et al.
Consequences of calamities and their management: the case of COVID-19
pandemic and flooding on inland capture fisheries in Kenya. J Great Lakes
Res. 2020 Dec;46(6):1767–75.
43. Love DC, Allison EH, Asche F, Belton B, Cottrell RS, Froehlich HE, et al. Emerg-
ing COVID-19 impacts, responses, and lessons for building resilience in the
seafood system. Glob Food Sect. 2021 Mar;28:100494.
44. Löhr K, Mugabe P, Turetta APD, Steinke J, Lozano C, Bonatti M, et al. Assessing
impacts of COVID-19 and their responses among smallholder farmers in
Brazil, Madagascar and Tanzania. Outlook Agric. 2022 Dec;51(4):460–9.
45. Hammond J, Milner D, van Wijk M. How COVID controls hit farm-
ers in 7 low-income countries, most in Africa [Internet]. 2022 [cited
2023 Mar 30]. Available from: https://www.preventionweb.net/news/
how-covid-controls-hit-farmers-7-low-income-countries-most-africa
46. Sen S, Gautam P, Kerala Wins UN, Award For. “Outstanding Contribution”
Towards Control of Non-Communicable Diseases. The Logical Indian Crew
[Internet]. 2020 Sep [cited 2022 Jul 29]; Available from: https://thelogicalin-
dian.com/uplifting/kerala-un-award-23980
27. Wacquant L, Slater T, Pereira VB. Territorial Stigmatization in Action. Environ
47. Arora M, Chauhan K, John S, Mukhopadhyay A. Multi-Sectoral Action for
Plan A [Internet]. 2014 Jun 1 [cited 2023 Mar 31];46(6):1270–80. Available
from: https://doi.org/10.1068/a4606ge
28. State Poverty Eradication Mission, Government of Kerala. Kudumbashree |
What is Kudumbashree [Internet]. 2023 [cited 2023 Mar 31]. Available from:
https://www.kudumbashree.org/pages/171
29. Kerala State Planning Board. Kerala State Poverty Eradication Plan [Internet].
2018 [cited 2022 Jun 10]. Available from: https://www.niti.gov.in/writeread-
data/files/Kerala%20presentation.pdf
30. Bhattacharya M, Banerjee P. COVID-19: Indices of economic and health
vulnerability for the Indian states. Social Sciences & Humanities Open. 2021
Jan 1;4(1):100157.
31. 5 Vulnerable Populations in Healthcare [Internet]. AJMC. 2018
[cited 2023 Mar 30]. Available from: https://www.ajmc.com/
view/5-vulnerable-populations-in-healthcare
32. Peters DH, Garg A, Bloom G, Walker DG, Brieger WR, Hafizur Rahman M.
Poverty and Access to Health Care in developing countries. Ann N Y Acad Sci.
2008;1136(1):161–71.
33. Administrative Reforms Commission, Government of Kerala. Welfare to Rights
II, Implementation of Select Legislations, Review Phase II [Internet]. Kerala;
2020 Oct [cited 2022 Jun 10]. Report No.: 6. Available from: https://arc.kerala.
gov.in/sites/default/files/inline-files/Welfare%20to%20Rights%202%20-%20
File%20Copy%20-%2001-01-2020.pdf
34. Marmot M. Social determinants of health inequalities. The Lancet. 2005
Mar;19(9464):1099–104.
35. Hutton G, Chase C. Water Supply, Sanitation, and Hygiene. In: Mock CN,
Nugent R, Kobusingye O, Smith KR, editors. Injury Prevention and Environ-
mental Health [Internet]. 3rd ed. Washington (DC): The International Bank for
Reconstruction and Development / The World Bank; 2017 [cited 2023 Mar
30]. Available from: http://www.ncbi.nlm.nih.gov/books/NBK525207/
36. Prüss-Ustün A, Wolf J, Bartram J, Clasen T, Cumming O, Freeman MC, et al.
Burden of disease from inadequate water, sanitation and hygiene for selected
adverse health outcomes: an updated analysis with a focus on low- and
middle-income countries. Int J Hyg Environ Health. 2019 Jun;222(5):765–77.
37. Sahoo KC, Dubey S, Dash GC, Sahoo RK, Sahay MR, Negi S et al. A System-
atic Review of Water, Sanitation, and Hygiene for Urban Poor in Low- and
48.
Addressing Social Determinants of Noncommunicable Diseases and Main-
streaming Health Promotion in National Health Programmes in India. Indian
Journal of Community Medicine. 2011 Dec 1;36(5):43.
Joy J. The Impact of RSBY-CHIS on utilisation of Healthcare Services in Kerala
[Internet]. Cochin: SCMS Cochin School of Business; 2019 [cited 2022 Jul 29].
Available from: https://spb.kerala.gov.in/sites/default/files/inline-files/RSBY-
CHIS.pdf
49. Kerala Social Security Mission., Govt. of Kerala. Schemes [Internet]. 2022 [cited
2022 Jul 9]. Available from: https://socialsecuritymission.gov.in/scheme_info.
php?id=2
50. Kerala Social Security Mission., Govt. of Kerala. Samashwasam [Internet].
2020 [cited 2022 Jul 9]. Available from: https://socialsecuritymission.gov.in/
scheme_info.php?id=MTk=
51. Madhavan M, Ravindran RM, Shinu KS. Functioning of Arogyakiranam
programme in Kerala: a qualitative study. J Family Med Prim Care. 2021
Nov;10(11):4117–23.
52. Rahim AA, Chacko TV. Replicating the Kerala State’s successful COVID-19
Containment Model: insights on what worked. Indian J Community Med.
2020;45(3):261–5.
53. Solomon N, K A G, Krishnan A, Cicily Joseph A, Dhanuraj D. Kerala Model of
Respone to COVID-19 [Internet]. Kochi: Centre for Public Policy Research
(CPPR); 2020 [cited 2022 Jun 10]. Available from: https://www.cppr.in/wp-
content/uploads/2020/10/KERALA-MODEL-OF-RESPONSE-TO-COVID-19.pdf
54. World Health Organization. Responding to COVID-19 - Learn-
ings from Kerala [Internet]. 2020 [cited 2022 Jun 10]. Available
from: https://www.who.int/india/news/feature-stories/detail/
responding-to-covid-19---learnings-from-kerala
55. Hasan MZ, Neill R, Das P, Venugopal V, Arora D, Bishai D et al. Integrated
health service delivery during COVID-19: a scoping review of published
evidence from low-income and lower-middle-income countries. BMJ Global
Health. 2021 Jun 1;6(6):e005667.
IANS. Kerala vaccination drive: Focus now on frontline workers in 2nd
phase. Business Standard India [Internet]. 2021 Feb 11 [cited 2022 Jun 10];
Available from: https://www.business-standard.com/article/current-affairs/
56.
Joseph et al. BMC Public Health (2023) 23:748 Page 11 of 11
kerala-vaccination-drive-focus-now-on-frontline-workers-in-2nd-
phase-121021100593_1.html
57. Health and Family Welfare Department. Guidelines for COVID-19 Vaccination
for the priority groups [Internet]. Government of Kerala; 2021 [cited 2022 Jun
10]. Available from: https://dhs.kerala.gov.in/wp-content/uploads/2021/03/
Guideline-for-COVID19-Vaccination-for-the-priority-groups_compressed.pdf
58. Health and Family Welfare Department. Government of, Kerala GO. (Rt) No.
1114/2021/H&FWD, Health & Family Welfare Department: Prioritization for
Vaccination in the age group of 18–45 years, Modified Orders issued [Inter-
net]. 2021 [cited 2022 Jun 10]. Available from: https://arogyakeralam.gov.in/
wp-content/uploads/2020/03/Prioritization-of-vaccinnation-in-18-45.pdf
59. George MS, Davey R, Mohanty I, Upton P. “Everything is provided free, but
they are still hesitant to access healthcare services”: why does the indigenous
community in Attapadi, Kerala continue to experience poor access to
healthcare? International Journal for Equity in Health [Internet]. 2020 Jun
26 [cited 2022 Jul 22];19(1):105. Available from: https://doi.org/10.1186/
s12939-020-01216-1
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
Joseph et al. BMC Public Health (2023) 23:748
| null |
10.1016_j.yjsbx.2023.100085.pdf
|
Data availability
Micrographs are available at EMPIAR-11397. Reconstructions are
available on EMDB with the following accession codes: for Glacios 0-50
nm, EMD-29566; 50-100 nm, EMD-29567; 100-150 nm, EMD-29568;
150-200 nm, EMD-29569; 200-500 nm, EMD-29570. For Arctica in
counting mode 0-50 nm, EMD-29573; 50-100 nm, EMD-29574; 100-150
nm, EMD-29575; 150-200 nm, EMD-29576; 200-500 nm, EMD-29577.
For Arctica in super resolution mode 0-50 nm, EMD-29589; 50-100
nm, EMD-29591; 100-150 nm, EMD-29592; 150-200 nm, EMD-29593;
200-500 nm, EMD-29594. For Krios unfiltered 0-50 nm, EMD-29554;
50-100 nm EMD-29555; 100-150 nm, EMD-29556; 150-200 nm, EMD-
29557; 200-500 nm, EMD-29558. For Krios filtered 0-50 nm, EMD-
29536; 50-100 nm, EMD-29535; 100-150 nm, EMD-29559; 150-200
nm, EMD-29513; 200-500 nm, EMD-29393.
|
Data availability Micrographs are available at EMPIAR-11397. Reconstructions are available on EMDB with the following accession codes: for Glacios 0-50 nm, EMD-29566; 50-100 nm, EMD-29567; 100-150 nm, EMD-29568; 150-200 nm, EMD-29569; 200-500 nm, EMD-29570. For Arctica in counting mode 0-50 nm, EMD-29573; 50-100 nm, EMD-29574; 100-150 nm, EMD-29575; 150-200 nm, EMD-29576; 200-500 nm, EMD-29577. For Arctica in super resolution mode 0-50 nm, EMD-29589; 50-100 nm, EMD-29591; 100-150 nm, EMD-29592; 150-200 nm, EMD-29593; 200-500 nm, EMD-29594. For Krios unfiltered 0-50 nm, EMD-29554; 50-100 nm EMD-29555; 100-150 nm, EMD-29556; 150-200 nm, EMD-29557; 200-500 nm, EMD-29558. For Krios filtered 0-50 nm, EMD-29536; 50-100 nm, EMD-29535; 100-150 nm, EMD-29559; 150-200 nm, EMD-29513; 200-500 nm, EMD-29393.
|
Contents lists available at ScienceDirect
Journal of Structural Biology: X
journal homepage: www.sciencedirect.com/journal/journal-of-structural-biology-x
Measuring the effects of ice thickness on resolution in single particle
cryo-EM
Kasahun Neselu a, Bing Wang b, William J. Rice b, c, Clinton S. Potter a, Bridget Carragher a,*,
Eugene Y.D. Chua a, *
a Simons Electron Microscopy Center, New York Structural Biology Center, New York, NY, USA
b Cryo-Electron Microscopy Core, New York University Grossman School of Medicine, New York, NY, USA
c Department of Cell Biology, New York University Grossman School of Medicine, New York, NY, USA
A R T I C L E I N F O
A B S T R A C T
Keywords:
Cryo-EM
Ice thickness
Single particle analysis
Energy filter
High tension
Resolution
Ice thickness is a critical parameter in single particle cryo-EM – too thin ice can break during imaging or exclude
the sample of interest, while ice that is too thick contributes to more inelastic scattering that precludes obtaining
high resolution reconstructions. Here we present the practical effects of ice thickness on resolution, and the
influence of energy filters, accelerating voltage, or detector mode. We collected apoferritin data with a wide
range of ice thicknesses on three microscopes with different instrumentation and settings. We show that on a 300
kV microscope, using a 20 eV energy filter slit has a greater effect on improving resolution in thicker ice; that
operating at 300 kV instead of 200 kV accelerating voltage provides significant resolution improvements at an ice
thickness above 150 nm; and that on a 200 kV microscope using a detector operating in super resolution mode
enables good reconstructions for up to 200 nm ice thickness, while collecting in counting instead of linear mode
leads to improvements in resolution for ice of 50–150 nm thickness. Our findings can serve as a guide for users
seeking to optimize data collection or sample preparation routines for both single particle and in situ cryo-EM.
We note that most in situ data collection is done on samples in a range of ice thickness above 150 nm so these
results may be especially relevant to that community.
Introduction
The goal of sample preparation for single particle cryo-electron mi-
croscopy (cryo-EM) is to capture the sample in optimal conditions on a
cryo-EM grid. “Optimal conditions” means the biological sample is
embedded in vitreous ice suspended over holes in the grid foil, has
enough well-distributed particles in different orientations, and that the
sample is found in ice that is as thin as possible, typically 10–100 nm
(Noble et al., 2018). While the thinnest possible ice might be expected to
yield the highest resolution reconstructions, there is usually a “Goldi-
locks” zone for ice thickness for each sample (Olek et al., 2022). If the ice
is too thin, the sample can be excluded from the holes, adopt a preferred
orientation, or break during imaging. On the other hand if the ice is too
thick, increased inelastic scattering from the additional ice may nega-
tively affect reconstruction resolutions (Wu et al., 2016). In most cases,
the thinnest possible ice that yields good particles is desirable for data
collection. This ideal ice thickness depends on the sample, and can range
from 15 nm for apoferritin (12 nm in diameter) (Brown & Hanssen,
2022) to 750 nm for the Giant Mimivirus (500 nm in diameter) (Xiao
et al., 2005). Quite often, however, ice much thicker than the diameter
of the particle is required to avoid particles adopting a preferred
orientation (e.g. Huntington et al., 2022).
Although ice thickness is an important parameter both for the sample
integrity and optimal data collection, it is not currently possible to finely
control ice thicknesses during cryo-EM sample preparation. With
commonly-used plunge freezers, or even with modern automated sam-
ple preparation devices such as the chameleon (Darrow et al., 2019,
2021), ice thicknesses often vary both within a grid square and across
the grid. Some areas of a grid may have good particle distribution and
ideal ice thickness while others may have too thin ice which excludes
particles, or too thick ice that has reduced contrast.
Problems of variations in ice thickness on a grid can be solved in
several ways. First, by setting automated data collection parameters to
only collect on the desired ice thicknesses (Brown & Hanssen, 2022;
Cheng et al., 2021; Rheinberger et al., 2021). Collecting good quality
data by skipping over targets with too thin or too thick ice is important
* Corresponding authors.
E-mail addresses: [email protected] (B. Carragher), [email protected] (E.Y.D. Chua).
https://doi.org/10.1016/j.yjsbx.2023.100085
Received 29 November 2022; Received in revised form 10 January 2023; Accepted 23 January 2023
JournalofStructuralBiology:X7(2023)100085Availableonline24January20232590-1524/©2023TheAuthors.PublishedbyElsevierInc.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/). K. Neselu et al.
for optimizing data collection and storage efficiency, and for achieving
highest resolution reconstructions. Second, post-specimen energy filters
can be used (Schr¨oder et al., 1990; Yonekura et al., 2006), which remove
inelastically scattered electrons to reduce background noise, especially
in regions with thicker ice. Using an energy filter should increase the
upper range of ice thicknesses useful for achieving a desired resolution.
Third, increasing the accelerating voltage of a microscope reduces the
inelastic mean free path of scattering (Dickerson et al., 2022; Henderson,
1995; Martynowycz et al., 2021; Peet et al., 2019). This means that
given the same sample thickness, electrons that have higher energy are
less likely to undergo inelastic scattering than those with lower energy,
and so will contribute less noise in those micrographs.
While the theoretical effects of ice thickness on single particle
analysis and available strategies to optimize data collection are known,
the practical effects of ice thickness on single particle analysis recon-
struction resolutions have to our knowledge not been experimentally
quantified. To this end we collected large apoferritin datasets over a
wide range of ice thickness (15–500 nm) using a variety of instrumen-
tation. This included both 200 kV and 300 kV microscopes (Glacios,
Arctica, and Krios); direct electron detectors operating in integrating
(Glacios with Falcon3), counting (Arctica with K3 and Krios with K3),
and super resolution mode (Arctica with K3); and with a 20 eV energy
filter slit inserted or retracted (Krios with K3). The data were sorted into
groups based on ice thickness and each batch was independently pro-
cessed to measure the impact of ice thickness and imaging technique on
reconstruction resolution. We show that using a 20 eV energy filter slit
has a greater effect in thicker ice; that operating at 300 kV instead of
200 kV accelerating voltage provides significant resolution improve-
ments at an ice thickness above 150 nm; that collecting data in super
resolution mode provides the most improvement in 150–200 nm
thickness; and finally that using a detector operating in counting instead
of linear mode has the greatest positive effect in < 150 nm ice thickness.
Our findings can serve as a guide for users seeking to optimize data
collection or sample preparation routines for both single particle and in
situ cryo-EM. We also note that most in situ data collection is done on
samples in a range of ice thickness above 150 nm so these results may be
especially relevant to that community.
Methods
Sample preparation
Mouse apoferritin in a pET24a vector (Danev et al., 2019) was
expressed in BL21(DE3) pLys cells. Cells were lysed, and apoferritin
precipitated with 60% ammonium sulfate. After resuspension in 30 mM
HEPES pH 7.5, 1 M NaCl, and 1 mM DTT, apoferritin was injected onto a
HiTrap Q column and eluted with a 0–0.5 M NaCl gradient over 4 col-
umn volumes. The elution peak was pooled and concentrated for puri-
fication on a Superdex 200 16/60 column in 30 mM HEPES pH 7.5, 150
mM NaCl, and 1 mM DTT.
UltrAuFoil R1.2/1.3 300 mesh grids (Quantifoil) were plasma
cleaned using a Solarus II (Gatan) with Ar:O₂ (26.3:8.7) at 15 W for 10 s.
3 μL mouse 8 mg/ml apoferritin was applied onto the plasma cleaned
grids. After a 30 s incubation at 100% relative humidity and 22
C the
grids were blotted for 4–5 s then plunge frozen into liquid ethane using a
Vitrobot Mark IV (Thermo Fisher Scientific).
◦
Data collection
Cryo-EM data was collected on three different microscopes. (1) A
Titan Krios (Thermo Fisher Scientific) microscope operating at 300 kV
and equipped with a BioQuantum energy filter (Gatan) and K3 camera
(Gatan) in counting mode. Krios data was collected either with a 20 eV
energy filter slit, or with the slit open, on the same grid during the same
data collection session. (2) A Talos Arctica microscope operating at 200
kV and equipped with a K3 detector operating in counting or super
resolution mode. Data was collected on a different apoferritin grid. (3) A
Glacios microscope operating at 200 kV and equipped with a Falcon3
camera (ThermoFisher Scientific) operating in integrating mode. Data
was collected on a third apoferritin grid. Data collection parameters are
found in Table 1. Leginon (Cheng et al., 2021; Suloway et al., 2005) was
used for automated data collection for all sessions. Ice thickness on the
Arctica and Glacios was measured by using aperture limited scattering
(ALS) method, and on the Krios by using the zero loss peak (ZLP) method
(Rice et al., 2018). During data collection, the incoming images were
motion corrected and dose weighted with motioncor2 (Zheng et al.,
2017) in Appion (Lander et al., 2010).
Image processing
Frame-aligned and dose-weighted images were sorted into 5
different ice thickness groups (0–50 nm, 50–100 nm, 100–150 nm,
150–200 nm, and 200–500 nm) using a Python script. The micrographs
were then imported into different workspaces and processed using cry-
oSPARC (Punjani et al., 2017). After importing the micrographs from
each ice thickness group, the CTF was estimated. Next, the micrographs
were manually curated to exclude bad micrographs, using the same
exclusion criteria for all ice thickness groups. 200 micrographs were
then randomly selected for further image processing. Particles were
manually picked from some of these micrographs to generate good
picking templates. Next, template picking was done on all 200 micro-
graphs. The picks were then inspected, and obvious bad picks were
excluded. The good picks were then extracted in a 256-pixel box and
connected to a 2D class averaging job. The resulting 2D classifications
were evaluated and only good class averages with good signal to noise
ratio were kept. From the set of good particles, 2 to 4 mutually exclusive
sets of 14,000 particles were created for further processing, depending
on the number of particles available. Homogeneous refinement with
defocus and CTF refinement was done on each set of particles, and the
best and average reconstruction statistics are reported here. For the
Glacios dataset, there was overfitting in the 3D reconstructions for ice
thicknesses above 100 nm resulting in an overestimation of the resolu-
tion. To overcome this, the same soft mask around the apoferritin den-
sity was applied to all reconstructions from all Glacios ice thickness
groups.
Analysis
Once a 3D reconstruction was obtained, the density was evaluated
using UCSF Chimera (Pettersen et al., 2004). Reconstructions from the
different microscopes and ice thickness groups were compared against
one another to evaluate which ice thickness and microscope setup gave
the best results. Linear regressions were done in Microsoft Excel using
the midpoint of each ice thickness group as the value on the x-axis. Map-
to-map Fourier shell correlations (FSCs) were calculated on the EMDB
FSC server https://www.ebi.ac.uk/emdb/validation/fsc/.
Table 1
Cryo-EM data collection parameters.
Dataset
“Krios (Filtered and
Unfiltered)”
“Arctica (Counting and
Super Resolution)”
“Glacios”
Titan Krios
300
Talos Arctica
200
Microscope
Accelerating
voltage (kV)
Energy filter slit
width (eV)
Pixel size (Å/pix)
Exposure time
(ms)
Frame time (ms)
Number of frames
Total dose (e/Å2)
Session name
20
1.083
2000
40
50
51.22
22may20b
N/A
1.096
2400
Glacios
200
N/A
1.204
2000
50
48
50.34
22sep21a, 22sep22a
40
50
50.53
22feb15b
JournalofStructuralBiology:X7(2023)1000852 K. Neselu et al.
Results
The thinner the ice, the better the resolution
To study the effects of ice thickness on resolution, we collected
apoferritin data with a wide range of ice thicknesses (15–500 nm) on the
Krios, Arctica, and Glacios microscopes. We observed the expected trend
that as ice thickness increases, resolution decreases (Fig. 1 & Supple-
mentary Fig. 1). With data collected on differently configured micro-
scopes, we can quantify the contributions from the energy filter,
accelerating voltage, and detector mode, to reconstruction resolutions at
varying ice thicknesses. It is important to bear in mind that the numbers
presented here are for a very specific data collection scenario, and do not
represent the performance limit of these microscope setups. What a
reconstruction can achieve practically will also depend on the number of
particles, sample size, and homogeneity.
Here, we report both the best (Fig. 1 and Tables 2a and 2b) and
average (Supplementary Fig. 1 and Supplementary Table 1) recon-
struction statistics from mutually exclusive sets of 14,000 particles
processed with the same data processing parameters and settings, so as
to have a holistic view on our processing, and to report on the variability
we encountered in the process.
The energy filter reduces the rate of resolution decay
Comparing the Krios datasets with and without the 20 eV energy
filter slit shows that the main advantage of using the slit is to reduce the
rate at which the resolution decays with increasing ice thickness. Fitting
linear regressions into the 0–150 nm range of the resolution plot
Table 2a
Accompaniment table to Fig. 1A. Table of highest apoferritin reconstruction
resolutions obtained from micrographs of various ice thicknesses, and with
microscopes of different configurations (see Table 1 for microscope configura-
tion details).
0–50
nm
3.40
2.76
50–100
nm
100–150
nm
150–200
nm
200–500
nm
4.64
2.91
10.18
3.19
9.63
6.83
2.61
2.78
3.04
4.11
2.41
2.53
2.84
3.21
2.36
2.46
2.58
2.76
2.92
9.84
8.13
8.82
6.67
Glacios
Arctica
(Counting)
Arctica (Super
Resolution)
Krios
(Unfiltered)
Krios (Filtered)
(Table 2a) reveals that both data have very similar intercepts (2.27 Å for
unfiltered, and 2.30 Å for filtered), but the slope of the unfiltered data, at
(cid:0) 1, is ~ 2-fold higher than that of filtered data at 0.0022 Å
0.0043 Å nm
(cid:0) 1 (Table 3a). This indicates that for apoferritin at the thinnest
nm
possible ice, the energy filter has minimal effect; however, with every
nm of increasing ice thickness, the resolution of these reconstructions
from data collected without an energy filter suffer 2-fold more than with
an energy filter, up to 150 nm. Similarly, the rate of B-factor decay
(Table 3b) on unfiltered data is 1.6-fold worse than that of filtered data,
up to 150 nm ice thickness. Above 150 nm thickness, however, the
resolution of unfiltered reconstructions starts to decay more rapidly,
reaching a best of only 6.67 Å in 200–500 nm ice thickness, compared to
2.92 Å for energy filtered data. Practically speaking, most single particle
data collected at the Simons Electron Microscopy Center is in ice
Fig. 1. (A) Plot of the best apoferritin resolutions obtained from micrographs of various ice thicknesses, and with microscopes of different configurations (see Table 1
for microscope configuration details). (B) Guinier plot B-factors from the best reconstructions versus ice thickness group. The data from each ice thickness group are
plotted on the midpoint ice thickness value on the x-axis, i.e. 25, 75, 125, 175, and 350 nm. The numbers giving rise to these plots can be found in Table 2a and 2b.
(C) Fig. 1(A) with a rescaled y-axis from 2.3 to 4 Å. (D) Fig. 1(B) with a rescaled y-axis from 50 to 300 Å2.
JournalofStructuralBiology:X7(2023)1000853 K. Neselu et al.
Table 2b
Accompaniment table to Fig. 1B. Table of Guinier plot B-factors from Table 1(A)
for each ice thickness group. DNE = Did not estimate; that is, the 3D refinement
job did not return a B-factor.
0–50
nm
220.1
117
50–100
nm
377.2
126.6
100–150
nm
150–200
nm
200–500
nm
DNE
129.6
DNE
756.8
DNE
1470.7
106
113.8
118.5
217.4
1287.1
87.5
91
108.5
128.4
766.4
85.5
90.7
98.3
104.8
116.7
Glacios
Arctica
(Counting)
Arctica (Super
Resolution)
Krios
(Unfiltered)
Krios (Filtered)
Table 3a
Linear regression fits into resolution vs ice thickness plots. Fits were done into
the linear portions of the graph to allow for the best comparisons between plots.
DNF = did not fit.
Dataset
Fit range
Linear regression
Glacios
Arctica (Counting)
Arctica (Super Resolution)
Krios (Unfiltered)
Krios (Filtered)
DNF
0–150 nm
0–150 nm
0–150 nm
0–150 nm
DNF
y = 0.0043x + 2.6308
y = 0.0043x + 2.4875
y = 0.0043x + 2.2708
y = 0.0022x + 2.3017
Table 3b
Linear regression fits into Guinier plot B-factor vs ice thickness plots.
Dataset
Fit range
Linear regression
Glacios
Arctica (Counting)
Arctica (Super Resolution)
Krios (Unfiltered)
Krios (Filtered)
DNF
0–150 nm
0–150 nm
0–150 nm
0–150 nm
DNF
y = 0.126x + 114.95
y = 0.125x + 103.39
y = 0.21x + 79.917
y = 0.128x + 81.9
R2
DNF
0.9704
0.9856
0.9389
0.9973
R2
DNF
0.9162
0.9799
0.871
0.9884
thickness < 100 nm, for which the improvement in resolution by
inserting the energy filter slit is small. This is expected since this ice
thickness range is well below the inelastic mean free path of 350 ~ 440
nm at 300 kV (Yonekura et al., 2006). Since the 20 eV slit provided the
greatest resolution improvement in the thickest 200–500 nm ice thick-
ness group (Fig. 1A and Table 1A), this may be of particular interest for
in situ data collection from FIB-milled lamella where thickness is more
likely to be in the range 150–250 nm.
Increasing high tension from 200 to 300 kV has the greatest effect in
thicker ice
Next, we compared the Arctica counting data with the unfiltered
Krios data. Since both microscopes were operated with a K3 detector in
counting mode, we could concentrate on the effects of 200 vs 300 kV
accelerating voltages. In ice of 0–150 nm, 200 kV data performed
slightly worse than 300 kV data: linear regression fits reached intercepts
of 2.63 Å (for 200 kV) vs 2.27 Å (for 300 kV), although the rates of
(cid:0) 1 (Table 3a). The
resolution decay were the same, at 0.0043 Å nm
biggest differences were observed at > 150 nm ice thickness, where the
200 kV Arctica counting data achieved only 6.83–8.13 Å re-
constructions, compared to 3.21–6.67 Å for 300 kV Krios counting data
(Table 2a). The data shows that increasing the accelerating voltage from
200 to 300 kV provides the greatest improvement at the 150–200 nm
thickness range. The corresponding ~ 6-fold increase in B-factors
(128.4 Å2 for 300 kV vs 756.8 Å2 for 200 kV) indicates that for this ice
thickness, a much larger amount of 200 kV data would need to be
collected to compensate for the loss of information due to inelastic
scattering.
Super resolution > counting > integrating mode
In integrating mode, a direct electron detector integrates the total
charge imparted by an electron, distributed by the microscope’s point
spread function, across several pixels. Operating in counting mode al-
lows for the localization of single electron events on the camera with
pixel accuracy, reducing Landau and readout noise, and improving the
DQE of a detector compared to integrating mode (Gatan, 2022; Li et al.,
2013). A further improvement in DQE can be gained by collecting data
in super resolution mode which makes use of high-speed detector elec-
tronics to determine the sub-pixel location of each electron event,
digitally increasing the number of pixels by 4x (Booth, 2012; Li et al.,
2013).
The poorer performance of 200 kV Arctica counting data compared
to 300 kV Krios counting data in 150–200 nm ice can be somewhat
rescued by collecting data in super resolution mode. This improved the
reconstruction resolution from 6.83 to 4.11 Å, and the B-factors from
756.8 to 217.4 Å2 (Table 2a and 2b) which are more comparable to 300
kV Krios counting data.
By comparing Arctica data with Glacios data, we could compare the
performance of a K3 detector operating in counting mode with a Falcon3
in integrating mode respectively. This is not an ideal comparison of
counting vs integrating collection modes, since the Falcon3 and K3 have
slightly different DQEs (Booth, 2019; Morado, 2020). Nevertheless, we
include this data in the interest of completeness. We observed the most
significant improvements from using counting mode below 150 nm ice
thickness. Above 150 nm ice thickness, counting and integrating modes
achieved similar resolutions and B-factors, suggesting that the noise
from increased inelastic scattering and the subsequent reduction in
image contrast dominates the gain in signal-to-noise from counting.
We observed that our Glacios data performs poorly at all ice thick-
nesses above 50 nm. While the data may appear to indicate that the
resolution remains stable in 100–500 nm ice, in contrast to the other
datasets where the resolutions and B-factors continued to worsen in the
same ice thickness range (Fig. 1 and Supplementary Fig. 1), we believe
this is just an artifact of generally poor reconstructions. Visual exami-
nation of the maps for reconstructions above 100 nm thickness revealed
no real structural features that might be expected for a map 9 ~ 10 Å in
resolution, and instead showed that the ~ 9.5 Å reported resolutions
were due to misalignments to noise (Fig. 2). Map-to-map FSCs of the
maps from thicker ice calculated against the map from 0 to 50 nm ice
show that the Glacios maps from > 100 nm ice thickness have, at best, a
20 Å correlation to the map from 0 to 50 nm thickness (Supplementary
Fig. 2). We conclude that Glacios data collected in integrating mode in
ice thicker than 100 nm produces unreliable reconstructions that are, for
apoferritin, significantly worse than 7 Å.
As ice thickness increases, we might expect a smooth decrease in
reconstruction resolution. Instead, across all our datasets, we observed
that resolution would decrease up to ~ 4 Å, after which there is a “jump”
to ~ 7 Å without an intermediate 5–6 Å reconstruction (Fig. 1A and
Table 1A). We hypothesize that this is because at better than ~ 4 Å there
are side chain densities that reconstruction programs can align to;
however, for apoferritin, which contains only alpha helices and no beta
sheets or any other significant structural features, the next feature that
can be aligned are alpha helices at ~ 7 Å, which results in the observed
“jump” in resolution.
Discussion
Thicker ice produces more inelastic scattering events, which de-
creases single-to-noise ratios and worsens reconstruction alignment ac-
curacy, resulting in poorer reconstruction resolutions. Here, we observe
that using a 20 eV energy filter slit slows down the rate of resolution
decay with increasing ice thickness by ~ 2-fold on a 300 kV microscope.
Using 300 kV accelerating voltage provides the greatest benefit over
200 kV at > 150 nm ice thickness, improving our apoferritin
JournalofStructuralBiology:X7(2023)1000854 K. Neselu et al.
Fig. 2. Apoferritin reconstructions from Glacios data collected in integrating mode, by ice thickness group.
reconstruction from 6.83 to 3.21 Å in 150–200 nm ice. Using super
resolution mode provides the most improvement in < 200 nm ice, and
collecting data in counting instead of integrating mode improves re-
constructions most noticeably in ice thinner than 150 nm. Combining
these effects, we obtained the highest resolution reconstructions across
all ice thickness groups from energy filtered 300 kV Krios data, followed
by unfiltered Krios, 200 kV Arctica with a K3 in super resolution then
counting mode, and lastly with a 200 kV Glacios with a Falcon3 in
integrating mode. For 200 kV instruments, the best imaging setup of
using a K3 in super resolution mode enabled high resolution re-
constructions < 200 nm ice. In situations where thick (> 200 nm) ice
cannot be avoided, for example with a large virus, large macromolecular
complex, or in situ sample, it is most critical to use a microscope with
high kV and an energy filter to obtain the highest resolution data.
In thin ice (0–50 nm), the best reconstructions from our comparable
200 and 300 kV data (Arctica with K3 in counting mode and Krios
unfiltered) perform similarly, at 2.76 and 2.41 Å respectively. A visual
examination of the maps showed little difference between the two
(Supplementary Fig. 4). The advantages of using a lower accelerating
voltage for single particle cryo-EM experiments have recently been more
thoroughly described (Naydenova et al., 2019; Peet et al., 2019), and the
data show that 200 keV electrons are better for single particle cryo-EM
than 300 keV when specimen thickness is not considered (Peet et al.,
2019). Specifically, the ratio of elastic scattering at 200 keV to 300 keV
is 1.24, whereas the ratio of inelastic scattering at 200 keV to 300 keV is
1.14. For specimens thinner than ~ 100 nm, electron energies lower
than 300 keV were shown to contain more useful information for single
particle cryo-EM (Peet et al., 2019). This improvement is likely some-
what offset by the detective quantum efficiency (DQE) of existing
counting direct detectors being slightly worse at 200 keV than at 300
keV.
During processing, we observed that in thicker ice, reconstructions
from mutually exclusive sets of 14,000 good particles randomly selected
from each ice thickness group could achieve very different resolutions
(Supplementary Fig. 1A). In the 200–500 nm ice thickness group, while
the best reconstruction we obtained with filtered Krios data was 2.92 Å,
across 4 independent reconstructions from mutually exclusive sets of
14,000 particles from the same dataset, we obtained reconstructions
that ranged up to 10.74 Å, with an average of 7.39 Å (Supplementary
Fig. 1A and Supplementary Table 1). This does not appear to be because
some groups of 14,000 particles were from thinner ice than others.
Analysis of the per-particle distribution of ice thicknesses for each of the
four Krios filtered reconstructions from the 200–500 nm ice thickness
group showed that particles that gave rise to the 2.92 Å reconstruction
did not have significantly better ice thicknesses than the other 7–10 Å
reconstructions (Supplementary Fig. 3). One hypothesis to explain this is
that variability can arise during the random initialization process of 3D
reconstruction: if a subset of higher quality particles happens to be
selected to initialize the reconstruction, this could lead to better align-
ment and resolution for that reconstruction. However, since these high-
quality particles are not ubiquitous in the thick ice data, obtaining these
reconstructions can be hit-or-miss. Another interesting observation was
that turning off both defocus and CTF optimization during 3D recon-
struction in thick ice could sometimes give higher resolution re-
constructions than if we turned on both options. For example, the best
unfiltered Krios 200–500 nm reconstruction achieved 6.67 Å with
defocus and CTF refinement, but 3.65 Å with both options deactivated.
We think this could be because in thick ice the particles have very little
high-resolution signal, so the defocus and CTF optimizing algorithms are
fitting to noise, and turning them off can potentially yield a better
reconstruction.
There are several additional considerations for improving on the
existing imaging setups tested in this work. Firstly, considering that the
inelastic mean free path is shorter for slower electrons, there will be
more inelastic scattering in a 200 kV microscope, which means that
installing a post-specimen energy filter will make a bigger impact on
JournalofStructuralBiology:X7(2023)1000855 K. Neselu et al.
such a setup than on a 300 kV microscope. Secondly, on a microscope
with a post-specimen energy filter, when collecting data in thick ice, it
would also be beneficial to reduce the width of the energy filter slit to ~
10 eV to optimally eliminate inelastically scattered electrons and
improve the reconstruction (Nakane et al., 2020; ThermoFisher Scien-
tific, 2022). Thirdly, reducing inelastic scattering from ice by energy
filtration will have the most benefit for small particles, since they have
the lowest signal-to-noise ratios, and energy filtration also improves
amplitude contrast allowing for better alignments during reconstruction
(Danev et al., 2021). Since apoferritin has a molecular weight of ~ 450
kDa, this same experiment should be done with a smaller protein of <
200 kDa to better evaluate the benefits for small macromolecules.
We observed that a Falcon3 detector in integrating mode (Glacios
dataset) was only useful in the thinnest ice, < 50 nm. While using
integrating mode on a Falcon3 reduces exposure times from 60 s to 1–2
s, making it much faster for a quick survey of the grid, the data from
integrating mode is not likely to provide useful reconstructions except in
very thin ice. For more challenging samples that prefer thicker ice or
when working with suboptimal grids with thicker ice (as is commonly
the case on a screening microscope), reconstructions obtained from the
Glacios in integrating mode may not be an accurate reflection of what
can be obtained from a better imaging setup. Here at the Simons Elec-
tron Microscopy Center, preliminary data are commonly collected on
our Glacios in integrating mode before a full data collection on a Krios
instrument. The data in this paper provide a useful benchmark for how
reconstructions from a Glacios dataset can be extrapolated to re-
constructions from a Krios dataset, given an ice thickness range. The
case can be made here for either collecting data in counting mode on the
Falcon3 on our Glacios microscope, or else for upgrading the camera,
say to a K3, for faster speeds and better reconstructions.
Preparing samples in the thinnest ice possible remains the best global
solution to obtaining high resolution. Where thick ice is necessary, for
example with large macromolecules or in situ samples, using the best
available imaging setup is essential for reaching high resolution with the
greatest possible speed.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Data availability
Micrographs are available at EMPIAR-11397. Reconstructions are
available on EMDB with the following accession codes: for Glacios 0-50
nm, EMD-29566; 50-100 nm, EMD-29567; 100-150 nm, EMD-29568;
150-200 nm, EMD-29569; 200-500 nm, EMD-29570. For Arctica in
counting mode 0-50 nm, EMD-29573; 50-100 nm, EMD-29574; 100-150
nm, EMD-29575; 150-200 nm, EMD-29576; 200-500 nm, EMD-29577.
For Arctica in super resolution mode 0-50 nm, EMD-29589; 50-100
nm, EMD-29591; 100-150 nm, EMD-29592; 150-200 nm, EMD-29593;
200-500 nm, EMD-29594. For Krios unfiltered 0-50 nm, EMD-29554;
50-100 nm EMD-29555; 100-150 nm, EMD-29556; 150-200 nm, EMD-
29557; 200-500 nm, EMD-29558. For Krios filtered 0-50 nm, EMD-
29536; 50-100 nm, EMD-29535; 100-150 nm, EMD-29559; 150-200
nm, EMD-29513; 200-500 nm, EMD-29393.
Acknowledgements
We thank Dr. Masahide Kikkawa (University of Tokyo) for the apo-
ferritin plasmid, and Dr. Brian Kloss (NYSBC) for expressing and pur-
ifying the apoferritin sample.
This work was supported by the Simons Electron Microscopy Center
and National Resource for Automated Molecular Microscopy located at
the New York Structural Biology Center, supported by grants from the
Simons Foundation (SF349247) and the NIH National Institute of Gen-
eral Medical Sciences (GM103310). All Arctica data in this work was
collected at the NYU Langone Health Cryo-EM core facility (RRID:
SCR_019202).
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.yjsbx.2023.100085.
References
Booth, C., 2012. K2: a super-resolution electron counting direct detection camera for
Cryo-EM. Microsc. Microanal. 18 (S2), 78–79. https://doi.org/10.1017/
S1431927612002243.
Booth, C. (2019). Detection Technologies for Cryo-Electron Microscopy. https://cryoem.slac.
stanford.edu/s2c2/sites/s2c2.cryoem.slac.stanford.edu/files/Cameras%20for%20Cr
yo-EM%20Stanford2019-Booth.pdf.
Brown, H.G., Hanssen, E., 2022. MeasureIce: accessible on-the-fly measurement of ice
thickness in cryo-electron microscopy. Commun. Biol. 5 (1), 817. https://doi.org/
10.1038/s42003-022-03698-x.
Cheng, A., Negro, C., Bruhn, J.F., Rice, W.J., Dallakyan, S., Eng, E.T., Waterman, D.G.,
Potter, C.S., Carragher, B., 2021. Leginon: new features and applications. Protein Sci.
30 (1), 136–150. https://doi.org/10.1002/pro.3967.
Danev, R., Yanagisawa, H., Kikkawa, M., 2019. Cryo-electron microscopy methodology:
current aspects and future directions. Trends Biochem. Sci 44 (10), 837–848.
https://doi.org/10.1016/j.tibs.2019.04.008.
Danev, R., Belousoff, M., Liang, Y.-L., Zhang, X., Eisenstein, F., Wootten, D., Sexton, P.M.,
2021. Routine sub-2.5 Å cryo-EM structure determination of GPCRs. Nat. Commun.
12 (1), 4333. https://doi.org/10.1038/s41467-021-24650-3.
Darrow, M.C., Moore, J.P., Walker, R.J., Doering, K., King, R.S., 2019. Chameleon: next
generation sample preparation for CryoEM based on Spotiton. Microsc. Microanal.
25 (S2), 994–995. https://doi.org/10.1017/S1431927619005701.
Darrow, M.C., Booth, T., Moore, J.P., Doering, K., Thaw, P., King, R.S., 2021. Enabling a
Paradigm Shift in CryoEM Sample Preparation with chameleon. Microsc. Microanal.
27 (S1), 524–525. https://doi.org/10.1017/S1431927621002336.
Dickerson, J.L., Lu, P.-H., Hristov, D., Dunin-Borkowski, R.E., Russo, C.J., 2022. Imaging
biological macromolecules in thick specimens: The role of inelastic scattering in
cryoEM. Ultramicroscopy 237, 113510. https://doi.org/10.1016/j.
ultramic.2022.113510.
Gatan. (2022). Improving DQE with Counting and Super-Resolution. https://www.gatan.
com/improving-dqe-counting-and-super-resolution.
Henderson, R., 1995. The potential and limitations of neutrons, electrons and X-rays for
atomic resolution microscopy of unstained biological molecules. Q. Rev. Biophys. 28
(2), 171–193. https://doi.org/10.1017/S003358350000305X.
Huntington, B., Zhao, L., Bron, P., Shahul Hameed, U.F., Arold, S.T., Qureshi, B.M., 2022.
Thicker Ice Improves the Integrity and Angular Distribution of CDC48A Hexamers on
Cryo-EM Grids. Front. Mol. Biosci. 9, 890390 https://doi.org/10.3389/
fmolb.2022.890390.
Lander, G. C., Stagg, S. M., Voss, N. R., Cheng, A., Fellmann, D., Yoshioka, C., Irving, C.,
Mulder, A., Lau, P.-W., Potter, C. S., & Carragher, B. (2010). Appion: An integrated,
database-driven pipeline to facilitate EM image processing. 16.
Li, X., Mooney, P., Zheng, S., Booth, C.R., Braunfeld, M.B., Gubbens, S., Agard, D.A.,
Cheng, Y., 2013. Electron counting and beam-induced motion correction enable
near-atomic-resolution single-particle cryo-EM. Nat. Methods 10 (6), 584–590.
https://doi.org/10.1038/nmeth.2472.
Martynowycz, M. W., Clabbers, M. T. B., Unge, J., Hattne, J., & Gonen, T. (2021).
Benchmarking the ideal sample thickness in cryo-EM. Proceedings of the National
Academy of Sciences, 118(49), e2108884118. https://doi.org/10.1073/
pnas.2108884118.
Morado, D. (2020). Some timing tests of the Gatan K3. https://forum.scilifelab.se/t/some-t
iming-tests-of-the-gatan-k3/139.
Nakane, T., Kotecha, A., Sente, A., McMullan, G., Masiulis, S., Brown, P.M.G.E.,
Grigoras, I.T., Malinauskaite, L., Malinauskas, T., Miehling, J., Ucha´nski, T., Yu, L.,
Karia, D., Pechnikova, E.V., de Jong, E., Keizer, J., Bischoff, M., McCormack, J.,
Tiemeijer, P., Hardwick, S.W., Chirgadze, D.Y., Murshudov, G., Aricescu, A.R.,
Scheres, S.H.W., 2020. Single-particle cryo-EM at atomic resolution. Nature 587
(7832), 152–156.
Naydenova, K., McMullan, G., Peet, M.J., Lee, Y., Edwards, P.C., Chen, S., Leahy, E.,
Scotcher, S., Henderson, R., Russo, C.J., 2019. CryoEM at 100 keV: A demonstration
and prospects. IUCrJ 6 (6), 1086–1098. https://doi.org/10.1107/
S2052252519012612.
Noble, A.J., Dandey, V.P., Wei, H., Brasch, J., Chase, J., Acharya, P., Tan, Y.Z., Zhang, Z.,
Kim, L.Y., Scapin, G., Rapp, M., Eng, E.T., Rice, W.J., Cheng, A., Negro, C.J.,
Shapiro, L., Kwong, P.D., Jeruzalmi, D., des Georges, A., Carragher, B., 2018.
Routine single particle CryoEM sample and grid characterization by tomography.
Elife 7, e34257.
Peet, M.J., Henderson, R., Russo, C.J., 2019. The energy dependence of contrast and
damage in electron cryomicroscopy of biological molecules. Ultramicroscopy 203,
125–131. https://doi.org/10.1016/j.ultramic.2019.02.007.
Pettersen, E.F., Goddard, T.D., Huang, C.C., Couch, G.S., Greenblatt, D.M., Meng, E.C.,
Ferrin, T.E., 2004. UCSF Chimera—A visualization system for exploratory research
JournalofStructuralBiology:X7(2023)1000856 K. Neselu et al.
and analysis. J. Comput. Chem. 25 (13), 1605–1612. https://doi.org/10.1002/
jcc.20084.
ThermoFisher Scientific. (2022). Selectris Applications Notes. https://cryoem.jhmi.
edu/wp-content/uploads/2022/07/Selectris_apps_note_v2_1.pdf.
Punjani, A., Rubinstein, J.L., Fleet, D.J., Brubaker, M.A., 2017. cryoSPARC: algorithms
for rapid unsupervised cryo-EM structure determination. Nat. Methods 14 (3),
290–296.
Wu, S., Armache, J.-P., Cheng, Y., 2016. Single-particle cryo-EM data acquisition by
using direct electron detection camera. Microscopy 65 (1), 35–41. https://doi.org/
10.1093/jmicro/dfv355.
Rheinberger, J., Oostergetel, G., Resch, G.P., Paulino, C., 2021. Optimized cryo-EM data-
acquisition workflow by sample-thickness determination. Acta Crystallogr. Sect. D
Struct. Biol. 77 (5), 565–571. https://doi.org/10.1107/S205979832100334X.
Rice, W.J., Cheng, A., Noble, A.J., Eng, E.T., Kim, L.Y., Carragher, B., Potter, C.S., 2018.
Routine determination of ice thickness for cryo-EM grids. J. Struct. Biol. 204 (1),
38–44. https://doi.org/10.1016/j.jsb.2018.06.007.
Schr¨oder, R.R., Hofmann, W., M´en´etret, J.-F., 1990. Zero-loss energy filtering as
improved imaging mode in cryoelectronmicroscopy of frozen-hydrated specimens.
J. Struct. Biol. 105 (1–3), 28–34. https://doi.org/10.1016/1047-8477(90)90095-T.
Suloway, C., Pulokas, J., Fellmann, D., Cheng, A., Guerra, F., Quispe, J., Stagg, S.,
Potter, C.S., Carragher, B., 2005. Automated molecular microscopy: The new
Leginon system. J. Struct. Biol. 151 (1), 41–60. https://doi.org/10.1016/j.
jsb.2005.03.010.
Xiao, C., Chipman, P.R., Battisti, A.J., Bowman, V.D., Renesto, P., Raoult, D.,
Rossmann, M.G., 2005. Cryo-electron microscopy of the giant mimivirus. J. Mol.
Biol. 353 (3), 493–496. https://doi.org/10.1016/j.jmb.2005.08.060.
Yonekura, K., Braunfeld, M.B., Maki-Yonekura, S., Agard, D.A., 2006. Electron energy
filtering significantly improves amplitude contrast of frozen-hydrated protein at
300kV. J. Struct. Biol. 156 (3), 524–536. https://doi.org/10.1016/j.
jsb.2006.07.016.
Zheng, S.Q., Palovcak, E., Armache, J.-P., Verba, K.A., Cheng, Y., Agard, D.A., 2017.
MotionCor2: Anisotropic correction of beam-induced motion for improved cryo-
electron microscopy. Nat. Methods 14 (4), 331–332. https://doi.org/10.1038/
nmeth.4193.
JournalofStructuralBiology:X7(2023)1000857
| null |
10.1016_S2213-2600(20)30559-2.pdf
| null |
Data sharing Access to all data and samples collected by ISARIC4C are controlled by an Independent Data and Materials Access Committee composed of representatives of research funders, academia, clinical medicine, public health, and industry. The application process for access to the data is available on the ISARIC4C website .
|
PDF file not found
|
|
10.1371_journal.pgph.0002891.pdf
|
Data Availability Statement: Data are included in
supporting materials.
|
Data are included in supporting materials.
|
RESEARCH ARTICLE
HIV epidemiologic trends among occupational
groups in Rakai, Uganda: A population-based
longitudinal study, 1999–2016
1, Joseph Kagaayi2,3, Joseph Ssekasanvu1,2, Robert Ssekubugu2,
Victor O. PopoolaID
Grace Kigozi2, Anthony Ndyanabo2, Fred Nalugoda2, Larry W. Chang1,2,4, Tom Lutalo2,
Aaron A. R. Tobian1,5, Donna Kabatesi6, Stella Alamo6, Lisa A. MillsID
6, Godfrey Kigozi2,
Maria J. Wawer1,2, John SantelliID
David Serwadda2,3, Justin Lessler1,9,10, M. Kate Grabowski1,2,5*
7, Ronald H. Gray1,2, Steven J. Reynolds4,8,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Popoola VO, Kagaayi J, Ssekasanvu J,
Ssekubugu R, Kigozi G, Ndyanabo A, et al. (2024)
HIV epidemiologic trends among occupational
groups in Rakai, Uganda: A population-based
longitudinal study, 1999–2016. PLOS Glob Public
Health 4(2): e0002891. https://doi.org/10.1371/
journal.pgph.0002891
Editor: Siyan Yi, National University of Singapore,
SINGAPORE
Received: August 8, 2023
Accepted: January 12, 2024
Published: February 20, 2024
Copyright: This is an open access article, free of all
copyright, and may be freely reproduced,
distributed, transmitted, modified, built upon, or
otherwise used by anyone for any lawful purpose.
The work is made available under the Creative
Commons CC0 public domain dedication.
Data Availability Statement: Data are included in
supporting materials.
Funding: This study was supported by the National
Institute of Allergy and Infectious Diseases (grants
R01AI110324, U01AI100031, and U01AI075115 to
RHG, R01AI143333 to LWC, R01AI155080 and
K01AI125086-01 to MKG), the National Institute of
Mental Health (grants R01MH107275 to LWC and
R01MH105313 to CK), the Eunice Kennedy Shriver
National Institute of Child Health and Human
1 Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland,
United States of America, 2 Rakai Health Sciences Program, Entebbe, Uganda, 3 Makerere University
School of Public Health, Kampala, Uganda, 4 Department of Medicine, Division of Infectious Diseases, Johns
Hopkins School of Medicine, Baltimore, Maryland, United States of America, 5 Department of Pathology,
Johns Hopkins School of Medicine, Baltimore, Maryland, United States of America, 6 Division of Global HIV
and TB, Centers for Disease Control and Prevention Uganda, Kampala, Uganda, 7 Department of Population
and Family Health and Pediatrics, Columbia University, New York, New York, United States of America,
8 Laboratory of Immunoregulation, Division of Intramural Research, National Institute for Allergy and
Infectious Diseases, National Institutes of Health, Bethesda, Maryland, United States of America,
9 Department of Epidemiology, UNC Gillings School of Global Public Health, Chapel Hill, North Carolina,
United States of America, 10 Carolina Population Center, Chapel Hill, North Carolina, United States of
America
* [email protected]
Abstract
Certain occupations have been associated with heightened risk of HIV acquisition and
spread in sub-Saharan Africa, including female bar and restaurant work and male transpor-
tation work. However, data on changes in population prevalence of HIV infection and HIV
incidence within occupations following mass scale-up of African HIV treatment and preven-
tion programs is very limited. We evaluated prospective data collected between 1999 and
2016 from the Rakai Community Cohort Study, a longitudinal population-based study of 15-
to 49-year-old persons in Uganda. Adjusted prevalence risk ratios for overall, treated, and
untreated, prevalent HIV infection, and incidence rate ratios for HIV incidence with 95% con-
fidence intervals were estimated using Poisson regression to assess changes in HIV out-
comes by occupation. Analyses were stratified by gender. There were 33,866 participants,
including 19,113 (56%) women. Overall, HIV seroprevalence declined in most occupational
subgroups among men, but increased or remained mostly stable among women. In con-
trast, prevalence of untreated HIV substantially declined between 1999 and 2016 in most
occupations, irrespective of gender, including by 70% among men (12.3 to 4.2%; adjPRR =
0.30; 95%CI:0.23–0.41) and by 78% among women (14.7 to 4.0%; adjPRR = 0.22; 95%
CI:0.18–0.27) working in agriculture, the most common self-reported primary occupation.
Exceptions included men working in transportation. HIV incidence similarly declined in most
occupations, but there were no reductions in incidence among female bar and restaurant
workers, women working in local crafts, or men working in transportation. In summary,
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
1 / 18
PLOS GLOBAL PUBLIC HEALTH Development (grants R01HD070769 and
R01HD050180 to MJW), the Division of Intramural
Research of the National Institute for Allergy and
Infectious Diseases (to SJR), the Johns Hopkins
University Center for AIDS Research (grant
P30AI094189 to MKG), and the President’s
Emergency Plan for AIDS Relief through the
Centers for Disease Control and Prevention (grant
NU2GGH000817 to DS). The findings and
conclusions in this article are those of the authors
and do not necessarily represent the official
position of the funding agencies. The funders had
no role in study design, data collection and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: We have read the journal’s
policy and the authors of this manuscript have the
following competing interests: Drs. Wawer and
Gray are paid consultants to the Rakai Health
Sciences Program and serve on its Board of
Directors. These arrangements have been reviewed
and approved by Johns Hopkins University in
accordance with its conflict of interest policies.
HIV epidemiologic trends among occupational groups in Rakai, Uganda
untreated HIV infection and HIV incidence have declined within most occupational groups in
Uganda. However, women working in bars/restaurants and local crafts and men working in
transportation continue to have a relatively high burden of untreated HIV and HIV incidence,
and as such, should be considered priority populations for HIV programming.
Introduction
The scale-up of combination HIV treatment and prevention interventions (CHI) in sub-Saha-
ran Africa has led to significant declines in HIV incidence [1–4]. However, rates of new HIV
infection remain significantly above elimination thresholds in most countries [5,6]. Demo-
graphic heterogeneities in population-level risk of HIV acquisition and onward transmission
likely drive continued virus spread, but they remain poorly characterized. A detailed under-
standing of such heterogeneities may facilitate targeted control efforts leading to further
declines in HIV incidence and, ultimately, disease elimination.
Decades-old data established a person’s occupation as a salient risk factor for HIV acquisi-
tion in Africa. Occupations historically associated with increased HIV risk have included min-
ing, bar work, truck driving, sex work, fishing, trading, and construction [3,4,7–10]. For
example, a study of HIV risk in Uganda, conducted in 1992, prior to the availability of antire-
troviral therapy (ART), found that bar and restaurant work, trading, and truck and taxi driving
were associated with three times higher odds of HIV acquisition compared to agricultural
work [4]. In southern Africa, truck driving, factory work, and mining have been strongly
linked to higher HIV burden [10–12]. While historical studies have provided useful insights
into HIV risk by occupation, there are very limited data comparatively assessing key HIV out-
comes within occupational subgroups since the widespread rollout of HIV interventions in
sub-Saharan Africa. Given that an individual’s occupation can be readily assessed in program-
matic settings, understanding whether HIV burden currently varies by occupation may facili-
tate efficient targeting of interventions.
Here, we assessed the extent to which occupation-specific population prevalence of HIV
and HIV incidence have changed since the implementation of combination HIV interventions
(CHIs) including ART, using data from the Rakai Community Cohort Study (RCCS), a popu-
lation-based HIV surveillance cohort in southern Uganda. We have previously measured
trends in HIV prevalence and incidence in the RCCS and shown a 42% reduction in HIV inci-
dence with ART rollout beginning in 2004 and VMMC scale-up beginning in 2007.13 How-
ever, it remains unclear whether or not untreated HIV prevalence and incidence declines have
occurred uniformly across occupational subgroups in this population. We hypothesized that
while the burdens of HIV, untreated HIV, and HIV incidence have declined within all occupa-
tions, heterogeneities in HIV outcomes by occupation persist.
Methods
Study population and procedures
The Rakai Community Cohort Study (RCCS) is conducted by the Rakai Health Sciences Pro-
gram and is an open, population-based census and cohort study including consenting individ-
uals aged 15–49 years across 40 communities in southern Uganda [13]. Individuals are
followed at ~18-month intervals. Briefly, the RCCS conducts a household census to enumerate
all individuals who are residents in the household, irrespective of presence or absence in the
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
2 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
home at time of census, based on sex, age, and how long they have been resident in the com-
munity. The census is followed by a survey of residents aged 15 to 49 years. All RCCS partici-
pants provide written informed consent prior to interviews. Participant interviews provide
self-reported data on socio-demographic characteristics, sexual behaviors, male circumcision
status, and ART use. Two attempts are made to contact individuals who are censused and eligi-
ble but who do not participate in the surveys.
To determine individual participant HIV serostatus in RCCS, venous blood samples are
obtained for HIV testing. Prior to October 2011, HIV testing used enzyme immunoassays
(EIAs) with confirmation via western blot. Subsequently, a field-validated, parallel three-test,
rapid HIV testing algorithm was introduced with demonstrated high sensitivity (>99.5%) and
specificity (>99.5%). All rapid test positives in RCCS are confirmed by two EIAs, with western
blot or PCR for discordant EIA results [14,15].
In this study, we included data from 12 consecutive RCCS survey rounds conducted between
April 6, 1999, and September 2, 2016, collected from 30 continuously surveyed communities.
The 12 surveys are herein denoted as Surveys 1 through 12: start and completion dates for each
survey are included in S1 Table. Participation rates among census-eligible persons present in
the community at the time of survey ranged from 74% to 98% (59%-66%, including those
absent from the community) across survey rounds [16]. There were generally lower levels of
participation in earlier survey rounds due to higher refusal rates. During the study period, par-
ticipant retention (i.e., follow-up between consecutive survey rounds) decreased from 73% to
55% [16,17]. Loss to follow-up was due mostly to out-migration to non-eligible study communi-
ties. When considering only participants who were resident in the community at time of survey
(e.g. excluding non-eligible migrants), retention decreased over the analysis from 93% to 80%.
For this study, RCCS data were accessed from December 15, 2018 through December 15,
2022. This study was approved by the Research and Ethics Committee of the Uganda Virus
Research Institute and the Johns Hopkins School of Medicine Institutional Review Board. This
study was also approved for the inclusion of children as ’research not involving greater than
minimal risk’ with the permission of at least one parent.
Measurement and classification of participant occupation
Occupational data were collected as self-reported primary occupations at the time of RCCS
interviews. Participants were asked, “What kind of work do you do, or what kind of activities
keep you busy during an average day, whether you get money for them or not.” There were 23
occupational subgroups that participants could select from on the questionnaire, including
“other.” Individuals who listed “other” were asked to provide occupational details as a free-text
response. Free-text responses were reviewed and re-assigned into pre-existing categories, or
new categories were created as needed. There were 36 self-reported primary occupations,
which were subsequently aggregated into 15 primary occupational subgroups (S2 Table). Of
these larger subgroups, eight among men (agriculture, trading, student, construction, civil ser-
vice, causal labor, mechanic, transportation) and nine among women (agriculture, trading,
student, bar/restaurant work, civil service, hairdressing, local crafts, tailoring/laundry, house-
keeping) contained a median number of � 50 observations per survey across all surveys
(S3A and S3B Table). These eight occupational subgroups among men and nine among
women were the primary exposure units for all subsequent occupational analyses.
Primary and secondary outcomes
Our primary study outcomes were (1) prevalent HIV infection, (2) prevalent untreated HIV
infection, and (3) incident HIV infection. We defined prevalent HIV infection as any HIV
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
3 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
infection in an individual (whether treated or untreated) and untreated HIV as HIV infection
in an individual with HIV who did not self-report ART use at time of survey. We have previ-
ously shown that self-reported ART use has high specificity (99%) and moderate sensitivity
(77%) in this population, and that this does not substantially vary by self-reported occupation
[18]. We note that the prevalence of untreated HIV infection in the overall population (includ-
ing seronegative individuals and persons living with HIV) was measured as a surrogate mea-
sure for population prevalence of viremia, which previous studies have shown is predictive of
HIV incidence [16,19]. Incident HIV infection was defined as a first HIV seropositive test
result in a person with a prior seronegative test result irrespective of HIV treatment status at
first positive visit. The unit of analysis for HIV incidence was person-years of follow-up
between surveys among persons who were initially HIV-seronegative and who contributed
two consecutive survey visits or two visits with no more than one missing intervening survey.
Incident infections were assumed to have occurred at the mid-point of the visit interval. Our
secondary outcome was self-reported ART use among persons with HIV.
Scale-up and measurement of combination HIV intervention coverage in
Rakai
During the analysis period, ART rollout in Uganda, including Rakai, was phased as follows: in
2004, ART was offered to persons with a CD4-T-cell count of <250 cells/mm3; in 2011, the
CD4 T-cell criterion was raised to <350; and in 2013, it was further increased to <500 and
ART was also offered to all individuals with HIV, regardless of CD4 T-cell count, if they were
pregnant, in a serodiscordant relationship, or self-identified as a sex worker or fisherfolk. The
prevalence of self-reported ART use had risen to 69% among all persons with HIV by 2016. In
addition to ART, the Rakai Health Sciences Program has provided free VMMC since 2007 to
adolescents and men aged 13 years or older [16]. The prevalence of male circumcision
increased from 15% in 1999 to 59% by 2016 [16]. Impacts of universal HIV test and treat and
pre-exposure prophylaxis were not assessed in this study as implementation of these programs
occurred after the analysis period in 2017 and 2018, respectively.
To assess changes in HIV incidence by occupation over calendar time, we divided the study
period into pre-CHI (surveys 1–5; 1999–2004), early-CHI scale up (surveys 6–9; 2005–2011),
and mature-CHI (surveys 10–12; 2011–2016) periods. Period-specific baselines were estab-
lished as the first survey during each period, while the study baseline for individual partici-
pants was defined as their first survey during the entire study period.
Statistical analysis
Demographic characteristics of participants at period-specific baselines were summarized
using descriptive statistics, including median and interquartile ranges for continuous variables
and frequencies and percentages for categorical variables. The prevalence of each primary
occupation was estimated as the number of participants self-reporting that occupation,
expressed as a proportion of all participants surveyed, and was stratified by sex. Self-reported
ART use among participants with HIV was assessed during the early and mature-CHI periods
and at the final study visit. Overall and untreated HIV prevalence were assessed at each of the
12 study visits and HIV incidence was estimated during the eleven inter-survey intervals over
the 17-year analysis period. To evaluate changes in prevalence of untreated HIV infection and
HIV incidence within occupational subgroups, we constructed log-binomial regression mod-
els to estimate prevalence risk ratios (PRR) and Poisson regression models to estimate inci-
dence rate ratios (IRR). Because our primary objective was to describe patterns of HIV
infection within occupational subgroups as opposed to causal inference, PRRs and IRRs were
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
4 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
only adjusted for age and marital status to ensure demographic comparability across popula-
tions. We calculated IRRs for HIV infection, comparing incidence rates during the pre-, early-
, and late-CHI periods. All statistical analyses were performed in Stata version 15 and the R sta-
tistical software (Version 3.6).
Results
Characteristics of study participants
Overall, 33,866 individuals (including 19,113 (56%) women) participated, contributing to a
total of 102,759 person visits. Of these participants, 17,840 women and 14,244 men who were
HIV-seronegative at their first study visit contributed 57,912 and 49,403 person-years to the
incidence cohort, respectively. S4 Table shows characteristics of the study population by sex at
the first (baseline) study visit within the CHI periods. Among women, during the pre-CHI
baseline visit, median age was 25 years (IQR: 20–34), 59% (2056/3474) were married, and the
prevalence of untreated HIV was 16%. Median age at the late-CHI baseline visit for women
was somewhat older at 28 years (IQR: 22–34), 60% (2265/3758) were married, and prevalence
of untreated HIV was 9.1%. Among men, during the first pre-CHI baseline visit, median age
was 26 years (IQR: 20–33), 56% (1418/2518) were married, 15% (374/2518) were circumcised,
and the prevalence of untreated HIV was 8.1%. In comparison, median age at the late-CHI
baseline visit for men was 27 years (IQR: 20–36), 52% (1524/2944) were married, 46% (1359/
2944) were circumcised, and the prevalence of untreated HIV was 6.4%.
Population prevalence of occupations over calendar time
Fig 1 shows the proportion of participants in each occupational subgroup over calendar time
stratified by gender (see also S5A and S5B Table). At the initial visit (1999–2000), the majority
Fig 1. Prevalence of primary occupational subgroups by gender in the Rakai Community Cohort Study, 1999–2016.
https://doi.org/10.1371/journal.pgph.0002891.g001
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
5 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
of women (61%) reported agriculture as their primary occupation. While agriculture remained
the most commonly reported female occupation at the final visit (2015–16), its prevalence sig-
nificantly declined to 40% (PRR = 0.66; 95%CI: 0.62–0.69) (Fig 1). Declines in agricultural
work among women were accompanied by an increase in the average age within the occupa-
tion (S1 Fig) and were predominately offset by the proportion of women who reported work-
ing in trading (9.4% in 1999 vs.16% in 2016, PRR = 1.7; 95%CI: 1.49–1.91) and being a student
(7.3% vs. 14%, PRR = 1.97; 95%CI: 1.72–2.27). Notably, no women or men reported sex work
as a primary occupation, and very few people reported being unemployed (n<7 at all study vis-
its; S6A and S6B Table).
Men similarly reported agriculture and trading as their most common primary occupations
(Fig 1). Between the first (1999–2000) and last (2015–2016) study visit, there was a decrease in
the proportion of male participants reporting agriculture (39% vs. 29%, PRR = 0.74; 95%CI:
0.68–0.80), while a greater proportion reported being a student (13% vs. 22%, PRR = 1.74; 95%
CI: 1.53–1.96), mechanic (2.5% vs. 5.6%, PRR = 2.29, 95%CI: 1.74–3.01), or working in trans-
portation (1.9% vs. 4.7%, PRR = 2.42, 95%CI: 1.78–3.28).
Trends in the prevalence of HIV, ART use, and untreated HIV within
occupations
The prevalence of HIV remained unchanged in most occupational groups among women
(Table 1), but increased among women working in agriculture (adjPRR = 1.19; 95%CI: 1.04–
1.35) and decreased among hairdressers (adjPRR = 0.27; 95%CI: 0.18–0.41) and housekeepers
(adjPRR = 0.68; 95%CI: 0.47–0.98). Among men, HIV prevalence decreased or trended down-
wards in most occupational groups but non-significantly trended upwards among men work-
ing in transportation (8.2% vs. 15.1%; adjPRR = 1.71; 95% CI: 0.64–4.58) and men working in
casual labor (10.6% vs. 16.7%; adjPRR = 1.26; 95% CI: 0.58–2.73).
The proportion of male and female participants with HIV self-reporting ART use increased
over time among all occupational subgroups (Table 2A and 2B). During the late-CHI period
and at the final study visit, levels of ART use were highest among women working in agricul-
ture and lowest among female students. ART use was statistically significantly lower among
female traders (adjPRR = 0.91; 95%CI: 0.83–0.98) and bar and restaurant workers
(adjPRR = 0.87; 95%CI: 0.78–0.97) compared to women working in agriculture during the late
CHI-period. Among men, ART use was highest among those working in civil service over the
entire analysis period. During the late CHI period, ART use was statistically significantly lower
among men working in trading (adjPRR = 0.91; 95%CI: 0.83–0.98) and male students
(adjPRR = 0.59; 95%CI: 0.41–0.84) compared to men working in agriculture.
Figs 2 and 3 show the prevalence of untreated HIV within occupational subgroups among
men and women at each of the 12 survey visits, respectively. Significant declines in the preva-
lence of untreated HIV were observed in nearly all occupational subgroups, irrespective of
gender, with scale-up of ART use. Relative changes in untreated HIV prevalence between the
first and final study visits are shown in Table 3 for each occupational subgroup by gender. The
prevalence of untreated HIV significantly decreased within most occupations. For example,
among women working in agriculture, prevalence of untreated HIV decreased from 14.7% to
4.0% (adjPRR = 0.22; 95%CI: 0.18–0.27), and among men, prevalence of untreated HIV
decreased from 12.3% to 4.2% (adjPRR = 0.30, 95%CI: 0.23–0.41). Women working in bars
and restaurants had among the highest HIV burdens across all occupational subgroups (Fig 3).
The prevalence of untreated HIV significantly declined among female bar and restaurant
workers from a high of 34.7% in 1999–2000 to 12.0% by 2015–2016 (adjPRR = 0.38; 95%CI:
0.25–0.58) (Table 3). However, these women had a 41.6% overall HIV seroprevalence at the
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
6 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
Table 1. Changes in prevalence of HIV infection between RCCS survey visit 1 (1999–2000) and RCCS survey visit 12 (2015–2016) by primary occupational sub-
group and gender of study participants.
Occupational
subgroup
Agriculture
Women
N = 10,121
Unadjusted
PRR
(95% CI)
Visit 12
(2015–2016),
HIV
prevalence, %
(n/T)
n = 6647
18.0 (481/
2669)
1.23 (1.08–
1.40)
Visit 1
(1999–
2000),
HIV
prevalence,
% (n/T)
n = 3474
14.7 (313/
2128)
Construction
-
-
-
Trading
21.5 (70/
325)
19.2 (201/
1048)
0.89 (0.70–
1.13)
Casual labor
-
-
-
Civil service
11.2 (19/
170)
10.8 (57/529)
Student
2.4 (6/254)
3.1 (30/959)
Mechanic
Transportation
-
-
-
-
Bar/Restaurant
worker
34.7 (50/
144)
41.6 (111/267)
Local crafts
19.8 (20/
101)
24.8 (34/137)
Hairdressing
46.3 (19/41)
13.7 (41/300)
Tailoring/
laundry
4.0 (2/50)
13.1 (16/122)
Housekeeping
16.7 (38/
227)
12.7 (69/545)
Other
occupations
17.6 (6/34)
26.8 (19/71)
0.96 (0.59–
1.57)
1.32 (0.56–
3.15)
-
-
1.20 (0.92–
1.56)
1.25 (0.77–
2.05)
0.30 (0.19–
0.46)
3.28 (0.78–
13.79)
0.76 (0.53–
1.09)
1.52 (0.66–
3.46)
All occupations
15.6 (543/
3474)
15.9 (1059/
6647)
1.02 (0.93–
1.12)
Men
N = 7,876
adjPRR*
(95% CI)
adjPRR
p-value
Visit 1 (1999–
2000),
HIV
prevalence, %
(n/T)
n = 2,518
Visit 12 (2015–
2016),
untreated HIV
prevalence, % (n/
T)
n = 5,358
Unadjusted
PRR (95% CI)
adjPRR**
(95% CI)
adjPRR
p-value
1.19
(1.04–
1.35)
-
0.82
(0.64–
1.05)
-
0.92
(0.54–
1.58)
0.91
(0.37–
2.26)
-
-
1.22
(0.93–
1.61)
0.98
(0.59–
1.65)
0.27
(0.18–
0.41)
2.49
(0.58–
10.62)
0.68
(0.47–
0.98)
1.48
(0.63–
3.50)
0.95
(0.86–
1.04)
0.010
12.3 (120/975)
11.9 (183/1538)
-
12.6 (36/285)
10.8 (54/500)
0.112
12.7 (51/401)
11.0 (83/756)
-
10.6 (7/66)
16.7 (22/132)
0.772
10.4 (23/221)
6.1 (26/429)
0.837
0.6 (2/319)
0.5 (6/1178)
-
-
0.158
0.945
<0.001
0.218
0.040
9.7 (6/62)
4.6 (14/302)
8.2 (4/49)
15.1 (38/252)
-
-
-
-
-
-
-
-
-
-
0.97 (0.78–
1.20)
0.83 (0.67–
1.02)
0.081
0.86 (0.58–
1.27)
0.86 (0.62–
1.20)
1.57 (0.71–
3.50)
0.58 (0.34–
1.00)
0.64 (0.43–
0.94
0.59 (0.41–
0.84)
1.26 (0.58–
2.73)
0.50 (0.30–
0.83)
0.025
0.004
0.558
0.008
0.81 (0.17–
4.01)
0.59 (0.12–
2.99)
0.524
0.48 (0.19–
1.20)
1.85 (0.69–
4.95)
0.38 (0.15–
0.99)
1.71 (0.64–
4.58)
0.049
0.286
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
0.369
11.4 (16/140)
14.0 (38/271)
1.23 (0.71–
2.12)
1.15 (0.66–
1.99)
0.619
0.243
10.5 (265/
2518)
8.7 (464/5358)
0.82 (0.71–
0.95)
0.72 (0.62–
0.83)
<0.001
PRR = prevalence risk ratios; adjPRR = adjusted prevalence risk; *Models adjusted for age and marital status of study participants.
https://doi.org/10.1371/journal.pgph.0002891.t001
final study visit in 2016 and still maintained a three-fold higher burden of untreated HIV com-
pared to women working in agriculture at the final versus initial visits (12.0% versus 4.0%).
Women working in local crafts and in trading also continued to have a high prevalence of
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
7 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
Table 2. a. Prevalence of self-reported ART use among women with HIV during the early and late-CHI periods and at the final study visit (Visit 12). b. Prevalence
of self-reported ART use among men with HIV during the early and late-CHI periods and at the final study visit (Visit 12).
Early–CHI (2004–2011)
N = 3,352
Late–CHI (2011–2016)
N = 2,695
Visit 12
N = 1,059
% self-reporting ART
(n/T)
PRR
(95% CI)
adjPRR
(95% CI)
% self-reporting ART
(n/T)
PRR
(95% CI)
adjPRR
(95% CI)
% self-reporting ART
(n/T)
PRR
(95% CI)
adjPRR
(95% CI)
Agriculture
24.9 (432/1734)
Trading
25.5 (149/584)
Casual labor
Civil service
-
19.5 (42/215)
Student
Bar/restaurant
worker
Local crafts
11.8
(2/17)
22.7 (65/287)
12.8
(12/94)
Hairdressing
21.0 (22/105)
Tailoring/laundry
23.7
(9/38)
Housekeeping
12.4 (28/226)
Other occupations
23.1
(12/52)
Ref
1.02
(0.87–
1.20)
-
0.78*
(0.59–
1.04)
0.47
(0.13–
1.74)
0.91
(0.72–
1.14)
0.51**
(0.30–
0.88)
0.84
(0.58–
1.23)
0.95
(0.53–
1.69)
0.50***
(0.35–
0.71)
0.93
(0.56–
1.53)
Ref
1.04
(0.89–
1.22)
-
0.89
(0.68–
1.17)
1.31
(0.35–
4.92)
0.95
(0.76–
1.20)
0.57**
(0.34–
0.95)
1.13
(0.76–
1.66)
1.07
(0.66–
1.75)
0.73*
(0.52–
1.04)
0.82
(0.50–
1.33)
64.6 (811/1256)
56.9 (302/531
-
58.2 (85/146)
38.0
(19/50)
55.9 (160/286)
50.0
(32/64)
54.3
(51/94)
60.0
(18/30)
52.5 (94/179)
66.1
(39/59)
Ref
0.88***
(0.81–
0.96)
-
0.90
(0.78–
1.04)
0.59***
(0.41–
0.84)
0.87**
(0.78–
0.97)
0.77**
(0.60–
0.99)
0.84*
(0.70–
1.02)
0.93
(0.69–
1.25)
0.81***
(0.70–
0.94)
1.02
(0.85–
1.24)
Ref
0.91**
(0.83–
0.98)
-
0.93
(0.81–
1.07)
0.83
(0.58–
1.20)
0.89**
(0.80–
0.99)
0.82
(0.65–
1.05)
0.95
(0.79–
1.15)
0.99
(0.75–
1.31)
0.96
(0.83–
1.11)
1.04
(0.85–
1.27)
78.0 (375/481)
68.2 (137/201)
-
70.2
(40/57)
53.3
(16/30)
71.2 (79/111)
58.8
(20/34)
65.9
(27/41)
75.0
(12/16)
63.8
(44/69)
79.0
(15/19)
Ref
0.87**
(0.79–
0.97)
-
0.90
(0.76–
1.07)
0.68**
(0.49–
0.96)
0.91
(0.80–
1.04)
0.76*
(0.57–
1.00)
0.85
(0.67–
1.06)
0.96
(0.72–
1.28)
0.82**
(0.68–
0.98)
1.01
(0.80–
1.28)
Ref
0.90**
(0.81–
0.99)
-
0.93
(0.78–
1.11)
0.86
(0.60–
1.22)
0.93
(0.82–
1.06)
0.79
(0.60–
1.05)
0.92
(0.74–
1.15)
1.01
(0.75–
1.35)
0.90
(0.75–
1.08)
1.05
(0.82–
1.36)
Early–CHI (2004–2011)
N = 1,702
Late–CHI (2011–2016)
N = 1,260
Visit 12
N = 464
% self-reporting ART
(n/T)
PRR
(95% CI)
adjPRR
(95% CI)
% self-reporting ART
(n/T)
PRR
(95% CI)
adjPRR
(95% CI)
% self-reporting ART
(n/T)
PRR
(95% CI)
adjPRR
(95% CI)
Agriculture
Construction
21.2 (140/661)
7.8 (14/180)
Trading
14.9 (47/316)
Casual labor
21.7 (15/69)
Civil service
24.5 (34/139)
Student
33.3 (4/12)
Mechanic
18.6 (11/59)
Ref
0.37***
(0.22–
0.62)
0.70**
(0.52–
0.95)
1.03
(0.64–
1.64)
1.16
(0.83–
1.60)
1.57
(0.70–
3.55)
0.88
(0.51–
1.53)
Ref
0.48***
(0.29–
0.80)
0.76*
(0.57–
1.01)
1.26
(0.79–
2.02)
0.96
(0.69–
1.34)
11.83***
(4.72–
29.68)
0.87
(0.54–
1.40)
54.4 (262/482)
41.0 (64/156)
42.0 (94/224)
36.9 (24/65)
60.6 (43/71)
35.7 (5/14)
43.3 (13/30)
Ref
0.76***
(0.62–
0.93)
0.77***
(0.65–
0.92)
0.68**
(0.49–
0.94)
1.11
(0.91–
1.37)
0.66
(0.32–
1.33)
0.80
(0.53–
1.21)
Ref
0.86
(0.70–
1.06)
0.80***
(0.67–
0.94)
0.70**
(0.51–
0.96)
1.03
(0.84–
1.25)
1.20
(0.56–
2.55)
0.80
(0.52–
1.23)
64.5 (118/183)
55.6 (30/54)
57.8 (48/83)
59.1 (13/22)
84.6 (22/26)
50.0 (3/6)
50.0 (7/14)
Ref
0.86
(0.66–
1.12)
0.90
(0.73–
1.11)
0.92
(0.64–
1.32)
1.31***
(1.08–
1.60)
0.78
(0.35–
1.74)
0.78
(0.45–
1.32)
Ref
0.91
(0.69–
1.19)
0.91
(0.74–
1.12)
0.91
(0.64–
1.29)
1.21*
(1.00–
1.48)
1.21
(0.49–
2.96)
0.82
(0.50–
1.34)
(Continued )
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
8 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
Table 2. (Continued)
Early–CHI (2004–2011)
N = 3,352
Late–CHI (2011–2016)
N = 2,695
Visit 12
N = 1,059
% self-reporting ART
(n/T)
PRR
(95% CI)
adjPRR
(95% CI)
% self-reporting ART
(n/T)
Transportation
15.4 (18/117)
Other occupations
16.1 (24/149)
0.73
(0.46–
1.14)
0.76
(0.51–
1.13)
1.01
(0.67–
1.52)
0.83
(0.57–
1.21)
42.9 (48/112)
50.9 (54/106)
https://doi.org/10.1371/journal.pgph.0002891.t002
PRR
(95% CI)
0.79**
(0.63–
0.99)
0.94
(0.76–
1.15)
adjPRR
(95% CI)
% self-reporting ART
(n/T)
PRR
(95% CI)
adjPRR
(95% CI)
0.94
(0.76–
1.17)
1.01
(0.83–
1.24)
52.6 (20/38)
60.5 (23/38)
0.82
(0.59–
1.13)
0.94
(0.71–
1.24)
0.94
(0.70–
1.28)
0.99
(0.76–
1.29)
untreated HIV compared to women in agriculture at the final visit (Table 3). Men working in
transportation did not have significantly higher HIV prevalence than other male occupations
at the initial visit (Table 3). However, we observed no declines in untreated HIV in this popula-
tion over the analysis period, and by the final visit, they had the highest prevalence of untreated
HIV among all male occupations at 7.1%.
Changes in HIV incidence within occupations before and during scale-up
of CHI programs
Table 4 shows HIV incidence by occupation, gender, and calendar time. In the early CHI
period, HIV incidence rates ranged from 0.4 to 2.3 per 100 person-years between occupational
subgroups among women, and from 0.1 to 1.8 per 100 person-years among men. Between the
early and late CHI periods, HIV incidence declined or trended downwards among most occu-
pational subgroups. For example, among those working in agriculture, HIV incidence declined
by 67% among men (adjIRR = 0.33; 95%CI: 0.21–0.54) and 38% among women (adjIRR = 0.62;
95%CI: 0.45–0.86). HIV incidence trends in most other occupations showed a decline, but
Fig 2. Trends in HIV prevalence (overall and untreated) among men by primary occupational subgroup in the Rakai Community Cohort Study (RCCS),
1999–2016; Untreated prevalence and 95% confidence intervals are shown as solid lines; overall HIV prevalence is shown as dashed lines with 95%
confidence bands in gray. Data are plotted at the calendar midpoint of each study visit.
https://doi.org/10.1371/journal.pgph.0002891.g002
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
9 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
Fig 3. Trends in HIV prevalence (overall and untreated) among women by primary occupational subgroup in the Rakai Community Cohort Study
(RCCS), 1999–2016; Untreated prevalence and 95% confidence intervals are shown as solid lines; overall HIV prevalence is shown as dashed lines with
95% confidence bands in gray. Data are plotted at the calendar midpoint of each study visit.
https://doi.org/10.1371/journal.pgph.0002891.g003
Table 3. Changes in prevalence of untreated HIV infection between RCCS survey visit 1 (1999–2000) and RCCS survey visit 12 (2015–2016) by primary occupa-
tional subgroup and gender of study participants.
Men
N = 7,876
Unadjusted PRR
(95% CI)
adjPRR**
(95% CI)
adjPRR
p-value
Occupational
subgroup
Women
N = 10,121
Visit 1
(1999–2000),
untreated
HIV
prevalence,
% (n/T)
n = 3474
14.7 (313/
2128)
Agriculture
Unadjusted PRR
(95% CI)
adjPRR*
(95% CI)
adjPRR
p-value
Visit 12 (2015–
2016),
untreated HIV
prevalence, %
(n/T)
n = 6647
Visit 1 (1999–
2000),
untreated HIV
prevalence, %
(n/T)
n = 2,518
Visit 12 (2015–
2016),
untreated HIV
prevalence, %
(n/T)
n = 5,358
4.0 (106/2669)
0.27 (0.22–0.33)
0.22 (0.18–0.27) <0.001
12.3 (120/975)
4.2 (65/1538)
0.34 (0.26–0.46)
0.30 (0.23–0.41) <0.001
Construction
-
-
-
-
-
12.6 (36/285)
4.8 (24/500)
0.38 (0.23–0.62
0.27 (0.17–0.43 <0.001
Trading
21.5 (70/325)
6.1 (64/1048)
0.28 (0.21–0.39)
0.09 (0.06–0.12) <0.001
12.7 (51/401)
4.6 (35/756)
0.36 (0.24–0.55)
0.27 (0.17–0.43) <0.001
Casual labor
-
-
-
-
-
10.6 (7/66)
6.8 (9/132)
0.64 (0.25–1.65)
0.54 (0.22–1.35)
0.185
Civil service
11.2 (19/170)
3.2 (17/529)
0.29 (0.15–0.54)
0.07 (0.04–0.15) <0.001
10.4 (23/221)
0.9 (4/429)
0.09 (0.03–0.26)
0.09 (0.03–0.25) <0.001
Student
Mechanic
Transportation
Bar/Restaurant
worker
2.4 (6/254)
1.5 (14/959)
0.62 (0.24–1.59)
0.11 (0.04–0.30) <0.001
0.6 (2/319)
0.3 (3/1178)
0.41 (0.07–2.42)
0.23 (0.04–1.19)
0.079
-
-
-
-
-
-
-
-
-
-
9.7 (6/62)
8.2 (4/49)
2.3 (7/302)
0.24 (0.08–0.69)
0.22 (0.07–0.67)
0.007
7.1 (18/252)
0.88 (0.31–2.47)
0.90 (0.30–2.73)
0.858
34.7 (50/144)
12.0 (32/267)
0.35 (0.23–0.51)
0.21 (0.14–0.31) <0.001
Local crafts
19.8 (20/101)
10.2 (14/137)
0.52 (0.27–0.97)
0.29 (0.14–0.58) <0.001
Hairdressing
46.3 (19/41)
4.7 (14/300)
0.10 (0.05–0.19)
0.01 (0.01–0.02) <0.001
Tailoring/
laundry
4.0 (2/50)
3.3 (4/122)
0.82 (0.16–4.33)
0.25 (0.05–1.35)
0.107
Housekeeping
16.7 (38/227)
4.6 (25/545)
0.27 (0.17–0.44)
0.11 (0.07–0.17) <0.001
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Other
occupations
All occupations
17.6 (6/34)
5.6 (4/71)
0.32 (0.10–1.06)
0.28 (0.08–1.02)
0.053
11.4 (16/140)
5.5 (15/271)
0.48 (0.25–0.95)
0.48 (0.25–0.96)
0.037
15.6 (543/
3474)
4.4 (294/6647)
0.28 (0.25–0.32)
0.27 (0.24–0.31) <0.001
10.5 (265/2518)
3.4 (180/5358)
0.32 (0.27–0.38)
0.29 (0.24–0.35) <0.001
PRR = prevalence risk ratios; adjPRR = adjusted prevalence risk; *Models adjusted for age and marital status of study participants.
https://doi.org/10.1371/journal.pgph.0002891.t003
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
10 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
Table 4. Incidence of HIV infection by primary occupational subgroup, sex, and CHI (combination HIV intervention) calendar period.
Occupation
Incidence rate per 100 py (n/py)
Women (N = 17,840)
IRR (95% CI)
adjIRR (95%CI)
Pre–CHI
(1999–2004)
Early–CHI
(2004–2011)
Late–CHI
(2011–2016)
Early–CHI vs. Pre-
CHI (ref)
Agriculture
1.1 (97/8490)
0.9 (129/
13785)
0.7 (62/9515)
0.82 (0.63–1.07)
Late–CHI vs. Pre-
CHI (ref)
0.57* (0.41–0.79)
Early–CHI vs. Pre-
CHI (ref)
0.87 (0.67–1.13)
Late–CHI vs. Pre-
CHI (ref)
0.62* (0.45–0.86)
Bar/restaurant
worker
1.1 (4/365)
2.1 (17/813)
2.0 (12/605)
1.91 (0.64–5.69)
1.81 (0.58–5.66)
2.13 (0.72–6.36)
2.79 (0.85–9.19)
Trading
1.4 (17/1222) 1.4 (49/3474) 0.7 (23/3117)
1.01 (0.58–1.77)
Hairdressing
2.3 (3/133)
1.6 (10/624)
0.8 (6/774)
0.71 (0.19–2.62)
Civil service
1.1 (9/851)
0.6 (14/2399)
0.3 (5/1849)
0.55 (0.24–1.28)
Student
0.3 (2/623)
0.4 (6/1433)
0.7 (14/2059)
1.30 (0.26–6.48)
Housekeeping
1.2 (6/496)
1.1 (15/1388) 0.9 (12/1278)
0.89 (0.35–2.32)
Local crafts
1.3 (5/387)
2.3 (12/518)
1.6 (5/308)
1.79 (0.63–5.13)
Tailoring/laundry
1.7 (2/121)
1.4 (5/355)
1.1 (3/261)
0.86 (0.16–4.47)
Other occupations
1.1 (1/93)
0.0 (0/169)
1.2 (5/406)
-
All occupations
1.1 (146/
12781)
1.0 (257/
24958)
0.7 (147/
20173)
0.90 (0.74–1.11)
0.53* (0.28–1.00)
0.34 (0.08–1.39)
0.26* (0.09–0.77)
2.12 (0.48–9.35)
0.78 (0.29–2.08)
1.25 (0.36–4.38)
0.70 (0.12–4.23)
1.14 (0.13–9.96)
0.64* (0.51–0.80)
1.19 (0.69–2.06)
0.73 (0.20–2.68)
0.76 (0.31–1.86)
1.23 (0.25–6.11)
1.0 (0.37–2.67)
1.88 (0.65–5.44)
0.94 (0.17–5.10)
-
0.93 (0.76–1.14)
0.72 (0.38–1.37)
0.36 (0.09–1.43)
0.49 (0.13–1.78)
1.93 (0.44–8.36)
0.89 (0.30–2.59)
1.36 (0.38–4.91)
0.85 (0.15–4.84)
1.41 (0.16–12.48)
0.66* (0.53–0.83)
Men (N = 14,244)
IRR (95% CI)
adjIRR (95%CI)
Incidence rate per 100 py (n/py)
Pre–CHI
(1999–2004)
Early–CHI
(2004–2011)
Late–CHI
(2011–2016)
Agriculture
1.4 (53/3894) 0.8 (62/8046) 0.4 (25/5834)
Construction
1.6 (18/1128) 1.2 (28/2322)
0.5 (9/1690)
Early–CHI vs. Pre-
CHI (ref)
0.57* (0.39–0.82)
0.76 (0.42–1.37)
Trading
0.8 (13/1541) 0.7 (26/3613) 0.5 (15/2746)
0.85 (0.44–1.67)
Casual labor
1.2 (3/259)
1.6 (7/431)
0.6 (2/322)
1.40 (0.36–5.49)
Civil service
0.7 (7/981)
0.6 (12/2121)
0.4 (6/1673)
0.79 (0.31–2.02)
Student
Mechanic
0.1 (1/955)
0.05 (1/1905)
0.1 (3/2840)
0.50 (0.03–8.03)
0.0 (0/228)
1.0 (8/780)
0.4 (4/901)
-
Late–CHI vs Pre-CHI
(ref)
0.32* (0.20–0.51)
0.33* (0.15–0.75)
0.65 (0.31–1.37)
0.54 (0.09–3.25)
0.50 (0.17–1.50)
1.01 (0.11–9.71)
-
Early–CHI vs. Pre-
CHI (ref)
0.58* (0.40–0.84)
0.79 (0.42–1.47)
0.89 (0.46–1.70)
1.67 (0.42–6.66)
0.78 (0.31–2.00)
0.51 (0.03–8.23)
-
Late–CHI vs Pre-CHI
(ref)
0.33* (0.21–0.54)
0.35* (0.15–0.83)
0.69 (0.33–1.45)
0.68 (0.11–4.30)
0.49 (0.16–1.51)
0.44 (0.04–4.86)
-
Transportation
1.4 (4/287)
1.8 (21/1181)
1.2 (12/964)
1.28 (0.43–3.75)
0.89 (0.29–2.80)
1.33 (0.45–3.91)
1.10 (0.35–3.50)
Other occupations
1.8 (10/546)
1.9 (21/1099) 0.9 (10/1115)
All occupations
1.1 (109/
9821)
0.9 (186/
21498)
0.5 (86/
18085)
1.04 (0.49–2.23)
0.78* (0.62–0.99)
0.49 (0.20–1.19)
0.43* (0.32–0.57)
1.06 (0.50–2.26)
0.79* (0.62–1.0)
0.50 (0.21–1.23)
0.44* (0.33–0.59)
py = person years; IRR = incidence rate ratio; adjIRR = adjusted incidence rate ratio for age and marital status; IRR not presented for other occupations (women, Early-
CHI) and mechanic (men) because there were no cases in the numerator and denominator respectively; CHI = combination HIV intervention
*p<0.05.
https://doi.org/10.1371/journal.pgph.0002891.t004
were not statistically significant. While HIV incidence did not decline among students, inci-
dence in this population was low overall. HIV incidence rates also did not decline among men
working in transportation, and women working in bars and restaurants or local crafts. S7
Table shows the adjusted relative risk of HIV acquisition by occupation during the late CHI
period. Compared to women working in agriculture, female bar and restaurant workers had a
three-fold higher rate of HIV incidence (adjIRR = 2.88; 95%CI: 1.51–5.49). Men working in
transportation also had significantly higher HIV incidence compared to agricultural workers
(adjIRR = 2.75; 95% CI: 1.37–5.50). Regardless of sex, students had a significantly lower risk of
HIV acquisition compared to persons working in agriculture (men: adjIRR = 0.19; 95% CI:
0.05–0.73; women: adjIRR = 0.36; 95% CI: 0.18–0.72).
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
11 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
Discussion
In this population-based study, overall prevalence of HIV (treated and untreated) mostly
declined among men, but remained stable or increased in most occupational subgroups
among women. We also observed declining prevalence of untreated HIV and HIV incidence
among most occupational subgroups with the scale up of HIV treatment and prevention pro-
grams in Uganda. Among men and women working in agriculture, the most common self-
reported primary occupation, prevalence of untreated HIV and HIV incidence declined by
more than two-thirds. However, this downward trend was not always the case for other occu-
pations. While women working in bars and restaurants made up a small proportion of the
overall population, they had among the highest burdens of untreated HIV prior to HIV inter-
vention scale-up, with no declines in HIV incidence over the analysis period. We also found
no significant reduction in HIV incidence among male transportation workers. Moreover,
both female bar and restaurant workers and male transportation workers had the highest prev-
alence of untreated HIV at the final study visit. HIV incidence rates among women reporting
student and crafting as primary occupations also showed no decrease following CHI scale-up,
although students had a very low HIV burden overall. Taken together, these results suggest
that members of traditionally high-risk occupations continue to experience elevated rates of
HIV incidence and remain sub-optimally served by HIV programs.
Other studies have reported high HIV prevalence among female bar workers in sub-Saha-
ran Africa [20,21]. In this study, HIV prevalence among female bar and restaurant workers
exceeded 40% with rising prevalence in recent years. While the prevalence of untreated HIV
significantly declined in this population, it was three times higher than among women working
in agriculture at the final study visit. The high burden of HIV among these women has been
linked to female sex work, alcohol use, and mobility [22–24]. In a systematic review of socio-
demographic characteristics and risk factors for HIV among female bar workers, high rural-
to-urban mobility, transactional sex, and inconsistent condom use were common and associ-
ated with financial needs and social marginalization [22]. Our results underscore that female
bar and restaurant workers should be a priority population for African HIV treatment and
prevention programs. While key population-based programs in Africa include female sex
workers, and female bar workers are often engaged in sex work, not all women working in
bars and restaurants at high risk of HIV classify themselves as sex workers [22]. Multi-level,
social influence, and structural HIV prevention interventions targeting alcohol-serving estab-
lishments, including enhanced sexually transmitted infection clinic services, portable health
services, and peer education, have been shown to be effective in settings outside Africa, for
reducing HIV risk and increasing treatment uptake [25,26].
Prior research has shown that men working in transportation are highly mobile and often
engage in transactional sex [27–29]. We found that the prevalence of untreated HIV did not sig-
nificantly decline in this occupational sub-group with the increasing availability of treatment and
prevention. Prior research has linked male transportation workers, including truck drivers, to
higher risk of HIV transmission [27], and has shown that men working in this occupation fre-
quently engage with sex workers and women working in bars and restaurants [28,30]. Supplies of
free condoms, roadside clinics, and free HIV testing services at truck stops are some HIV preven-
tion interventions that have been targeted to male transportation workers [10,30]; however, levels
of awareness and uptake of such services in this population have been low [10,31].
Adolescent girls and young women aged 15 to 24 years have a disproportionately high risk
of HIV acquisition in Africa [32–35], but HIV risk was significantly lower among young peo-
ple who list their occupation as “student” and who have higher education attainment, regard-
less of sex [36–39]. During the study period, HIV prevalence declined in female students by
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
12 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
nearly 90%. Incidence of HIV remained stable for both male and female students, but com-
pared to those in agriculture, students of both sexes had lower HIV incidence during the late-
CHI period. Research from South Africa has shown that students tend to have smaller sexual
networks and are less likely to report high-risk sexual behaviors compared to those not in
school [37]. Lower HIV incidence and prevalence among female students have also been
attributed to avoiding the consequences of unprotected sex and increased self-efficacy for
negotiating safer sex with their partners [40]. Interventions that increase school enrollment of
adolescent girls and young women may decrease sexual initiation, high-risk sexual behavior,
and HIV risk [32].
Since the onset of the COVID-19 pandemic in Uganda during the spring of 2020, schools
remained fully or partially closed until 2022. A review of adolescent sexual and reproductive
health during the COVID-19 pandemic found an increase in teenage pregnancies and gender-
based violence [41]. Given the strong protective effects of schooling on HIV acquisition,
understanding the extent to which school closures impact HIV and other reproductive health
outcomes, such as unplanned pregnancy, is an urgent public health priority.
Earlier studies have established migration and mobility as a key risk factor for HIV acquisi-
tion and transmission [23,42,43]. Overall, we found that the occupations which tend to have
high mobility also had higher prevalence of untreated HIV and HIV incidence despite scale-
up of HIV interventions. Both female bar and restaurant work and male transportation work
are associated with increased mobility as well as high-risk sexual behaviors, including concur-
rent sexual partnerships and inconsistent condom use [28,29,44]. Specialized service-delivery
tailored to mobile populations, such as client-managed groups, adherence clubs, community
drug distribution points, and multi-month prescriptions may reduce HIV burden in these
populations [45–47].
The shifting distribution of the occupational makeup in our study population away from
agriculture likely reflects the increasing urbanization happening across the African continent
[48]. Little data exists on the impact of urbanization on HIV transmission; however, in sub-
Saharan Africa, HIV prevalence and incidence have been reported to be higher in urban than
in rural centers [49,50]. This has been attributed to factors such as relative affluence in urban
centers, increased social interaction, and higher-risk behaviors such as transactional sex and
concurrent sexual partnerships [51–53]. More research is needed to elucidate the impact of
increasing urbanization on HIV transmission within African populations.
Our study has important limitations. First, both occupation and ART use were self-reported
and may be subject to bias. However, we have previously shown that self-reported ART use
has high specificity and moderate sensitivity in this same study population, and does not sub-
stantially vary by self-reported occupation [18]. Second, female sex work in Uganda is crimi-
nalized and was likely substantially underreported in our survey [4]. Third, PEPFAR-
supported key population HIV prevention programs began in this region in 2017, after the
time of the analysis, and so their impact cannot be assessed. Given previously reported links
between female sex work and bar work [21], our findings support PEPFAR’s ongoing focus on
targeted HIV prevention and treatment to female sex workers. However, many bar workers do
not identify as sex workers (none in this study), suggesting that they and other population sub-
groups may merit additional programmatic consideration. Neither bar and restaurant workers
nor male transportation workers are presently considered priority populations for HIV pro-
gramming in Uganda. Fourth, while the longitudinal nature of this study is a strength, analysis
of incident HIV infections were limited by a small number of events in some occupational sub-
groups, which may have obscured significant trends. Additionally, non-differential non-
response and loss to follow-up may have biased our results but in earlier studies from this
same population, sensitivity analyses showed little to no impact of selection bias on incidence
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
13 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
estimates [16]. Lastly, because participants become aware of the risk of contracting HIV, their
HIV status, and available treatments and prevention through their participation in the study,
they may be more likely to take up and adhere to preventative measures or treatment, and so
our results may not be generalizable to other populations. However, we expect the Hawthorne
effect to be limited in this open cohort with substantial in- and out-migration.
In summary, prevalence of untreated HIV infection and HIV incidence declined in most
occupational subgroups following the mass scale-up of HIV prevention and treatment inter-
ventions in rural southern Uganda. However, HIV burden remained relatively high in some
occupations, including the traditionally high-risk occupations of female bar and restaurant
work and male transportation work. HIV programs that meet the unique needs of these high-
risk populations, which tend to be more mobile with higher levels of HIV-associated risk
behaviors, may help achieve HIV epidemic control.
Supporting information
S1 Checklist. Inclusivity in global research.
(DOCX)
S2 Checklist. STROBE Statement—checklist of items that should be included in reports of
observational studies.
(DOCX)
S1 Fig. Boxplots of age in years at each study visit, among RCCS agricultural workers.
(TIF)
S1 Table. Rakai Community Cohort Study (RCCS) survey start and end dates.
(DOCX)
S2 Table. Recategorization of 36 self-reported primary occupations into occupational sub-
groups.
(DOCX)
S3 Table. A. Number of male observations at each study visit by primary occupational sub-
group. B. Number of female observations in each primary occupational subgroup at each
study visit.
(DOCX)
S4 Table. Characteristics of the study population at the baseline visit within each CHI cal-
endar period by gender.
(DOCX)
S5 Table. A. Prevalence (%) of major occupations among women by visit. B. Prevalence (%)
of major occupations among men by visit.
(DOCX)
S6 Table. A. Self-reported primary occupations by male RCCS study participants at each
study visit. B. Self-reported primary occupations by female RCCS study participants at each
study visit.
(DOCX)
S7 Table. Adjusted incidence rate ratios of HIV infection comparing all occupations vs.
agriculture during the late-CHI period.
(DOCX)
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
14 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
Acknowledgments
We thank the RCCS participants and many staff and investigators who have made this study
possible over the years. Additionally, we thank the personnel at the Office of Cyberinfrastruc-
ture and Computational Biology at the National Institute of Allergy and Infectious Diseases
for data management support.
Author Contributions
Conceptualization: Victor O. Popoola, M. Kate Grabowski.
Data curation: Victor O. Popoola, Joseph Kagaayi, Joseph Ssekasanvu, Robert Ssekubugu,
Grace Kigozi, Anthony Ndyanabo, Fred Nalugoda, Larry W. Chang, Tom Lutalo, Aaron A.
R. Tobian, Godfrey Kigozi, Ronald H. Gray, Steven J. Reynolds, David Serwadda.
Formal analysis: Victor O. Popoola, Justin Lessler.
Funding acquisition: Joseph Kagaayi, Maria J. Wawer, Ronald H. Gray.
Investigation: Victor O. Popoola.
Methodology: Victor O. Popoola, Justin Lessler, M. Kate Grabowski.
Project administration: Joseph Kagaayi, Robert Ssekubugu, Grace Kigozi, Fred Nalugoda,
Larry W. Chang, Donna Kabatesi, Stella Alamo, Lisa A. Mills, Maria J. Wawer, John San-
telli, Ronald H. Gray, David Serwadda, M. Kate Grabowski.
Software: Victor O. Popoola.
Supervision: M. Kate Grabowski.
Validation: Joseph Ssekasanvu.
Writing – original draft: Victor O. Popoola, M. Kate Grabowski.
Writing – review & editing: Victor O. Popoola, Joseph Kagaayi, Joseph Ssekasanvu, Robert
Ssekubugu, Grace Kigozi, Anthony Ndyanabo, Fred Nalugoda, Larry W. Chang, Tom
Lutalo, Aaron A. R. Tobian, Donna Kabatesi, Stella Alamo, Lisa A. Mills, Godfrey Kigozi,
Maria J. Wawer, John Santelli, Ronald H. Gray, Steven J. Reynolds, David Serwadda, Justin
Lessler, M. Kate Grabowski.
References
1.
Joshi K, Lessler J, Olawore O, Loevinsohn G, Bushey S, Tobian AAR, et al. Declining HIV incidence in
sub-Saharan Africa: a systematic review and meta-analysis of empiric data. medRxiv.
2020:2020.12.08.20246066. https://doi.org/10.1002/jia2.25818.
2. UNAIDS. Report–AIDS 2020. UNAIDS; 2020 https://aids2020.unaids.org/report/.
3. Diallo BL, Alary M, Rashed S, Barry A. [HIV prevalence, associated risk factors and evolution among
truck drivers from 2001 to 2007 in Guinea]. Med Trop (Mars). 2011; 71(2):142–6. https://pubmed.ncbi.
nlm.nih.gov/21695870/.
4. Kagaayi J, Gray RH, Whalen C, Fu P, Neuhauser D, McGrath JW, et al. Indices to measure risk of HIV
acquisition in Rakai, Uganda. PloS one. 2014; 9(4):e92015. https://doi.org/10.1371/journal.pone.
0092015 PMID: 24704778
5. Granich RM, Gilks CF, Dye C, De Cock KM, Williams BG. Universal voluntary HIV testing with immedi-
ate antiretroviral therapy as a strategy for elimination of HIV transmission: a mathematical model. Lan-
cet. 2009; 373(9657):48–57. https://doi.org/10.1016/S0140-6736(08)61697-9 PMID: 19038438
6. Hontelez JA, Lurie MN, Barnighausen T, Bakker R, Baltussen R, Tanser F, et al. Elimination of HIV in
South Africa through expanded access to antiretroviral therapy: a model comparison study. PLoS Med.
2013; 10(10):e1001534. https://doi.org/10.1371/journal.pmed.1001534 PMID: 24167449
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
15 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
7. Bowen P, Govender R, Edwards P, Lake A. HIV infection in the South African construction industry.
Psychol Health Med. 2018; 23(5):612–8. https://doi.org/10.1080/13548506.2017.1380836 PMID:
28931303
8. Mbugua GG, Muthami LN, Mutura CW, Oogo SA, Waiyaki PG, Lindan CP, et al. Epidemiology of HIV
infection among long distance truck drivers in Kenya. East Afr Med J. 1995; 72(8):515–8. https://
europepmc.org/article/med/7588147. PMID: 7588147
9. Mafigiri R, Matovu JK, Makumbi FE, Ndyanabo A, Nabukalu D, Sakor M, et al. HIV prevalence and
uptake of HIV/AIDS services among youths (15–24 Years) in fishing and neighboring communities of
Kasensero, Rakai District, South Western Uganda. BMC public health. 2017; 17(1):251. https://doi.org/
10.1186/s12889-017-4166-2 PMID: 28288604
10. Delany-Moretlwe S, Bello B, Kinross P, Oliff M, Chersich M, Kleinschmidt I, et al. HIV prevalence and
risk in long-distance truck drivers in South Africa: a national cross-sectional survey. International journal
of STD & AIDS. 2014; 25(6):428–38. https://doi.org/10.1177/0956462413512803.
11. Mabathoana RS, Wyk CV, Adefuye AO. Factors influencing HIV risk-taking behaviours amongst textile
factory workers living with HIV in Lesotho. Pan Afr Med J. 2019; 33:166. https://doi.org/10.11604/pamj.
2019.33.166.18961 PMID: 31565127
12. Baltazar CS, Horth R, Inguane C, Sathane I, Cesar F, Ricardo H, et al. HIV prevalence and risk behav-
iors among Mozambicans working in South African mines. AIDS and behavior. 2015; 19 Suppl 1:S59–
67. https://doi.org/10.1007/s10461-014-0941-6 PMID: 25398418
13. Wawer MJ, Sewankambo NK, Serwadda D, Quinn TC, Paxton LA, Kiwanuka N, et al. Control of sexu-
ally transmitted diseases for AIDS prevention in Uganda: a randomised community trial. Rakai Project
Study Group. Lancet. 1999; 353(9152):525–35. https://doi.org/10.1016/s0140-6736(98)06439-3 PMID:
10028980
14. Kagulire SC, Opendi P, Stamper PD, Nakavuma JL, Mills LA, Makumbi F, et al. Field evaluation of five
rapid diagnostic tests for screening of HIV-1 infections in rural Rakai, Uganda. International journal of
STD & AIDS. 2011; 22(6):308–9. https://doi.org/10.1258/ijsa.2009.009352 PMID: 21680664
15. Gray RH, Makumbi F, Serwadda D, Lutalo T, Nalugoda F, Opendi P, et al. Limitations of rapid HIV-1
tests during screening for trials in Uganda: diagnostic test accuracy study. BMJ. 2007; 335:188. https://
doi.org/10.1136/bmj.39210.582801.BE PMID: 17545184
16. Grabowski MK, Serwadda DM, Gray RH, Nakigozi G, Kigozi G, Kagaayi J, et al. HIV Prevention Efforts
and Incidence of HIV in Uganda. N Engl J Med. 2017; 377(22):2154–66. https://doi.org/10.1056/
NEJMoa1702150 PMID: 29171817
17. Kagaayi J, Chang LW, Ssempijja V, Grabowski MK, Ssekubugu R, Nakigozi G, et al. Impact of combi-
nation HIV interventions on HIV incidence in hyperendemic fishing communities in Uganda: a prospec-
tive cohort study. Lancet HIV. 2019; 6(10):e680–e7. https://doi.org/10.1016/S2352-3018(19)30190-0
PMID: 31533894
18. Grabowski MK, Reynolds SJ, Kagaayi J, Gray RH, Clarke W, Chang LW, et al. The validity of self-
reported antiretroviral use in persons living with HIV: a population-based study. AIDS. 2018; 32(3):363–
9. https://doi.org/10.1097/QAD.0000000000001706 PMID: 29194115
19.
Larmarange J, Bachanas P, Skalland T, Balzer LB, Iwuji C, Floyd S, et al. Population-level viremia pre-
dicts HIV incidence at the community level across the Universal Testing and Treatment Trials in eastern
and southern Africa. PLOS Glob Public Health. 2023; 3(7):e0002157. https://doi.org/10.1371/journal.
pgph.0002157 PMID: 37450436
20. Nagot N, Ouangre A, Ouedraogo A, Cartoux M, Huygens P, Defer MC, et al. Spectrum of commercial
sex activity in Burkina Faso: classification model and risk of exposure to HIV. J Acquir Immune Defic
Syndr. 2002; 29(5):517–21. https://doi.org/10.1097/00126334-200204150-00013 PMID: 11981369
21. Riedner G, Rusizoka M, Hoffmann O, Nichombe F, Lyamuya E, Mmbando D, et al. Baseline survey of
sexually transmitted infections in a cohort of female bar workers in Mbeya Region, Tanzania. Sex
Transm Infect. 2003; 79(5):382–7. https://doi.org/10.1136/sti.79.5.382 PMID: 14573833
22. Dambach P, Mahenge B, Mashasi I, Muya A, Barnhart DA, Barnighausen TW, et al. Socio-demographic
characteristics and risk factors for HIV transmission in female bar workers in sub-Saharan Africa: a sys-
tematic literature review. BMC public health. 2020; 20(1):697. https://doi.org/10.1186/s12889-020-
08838-8 PMID: 32414352
23. Olawore O, Tobian AAR, Kagaayi J, Bazaale JM, Nantume B, Kigozi G, et al. Migration and risk of HIV
acquisition in Rakai, Uganda: a population-based cohort study. Lancet HIV. 2018; 5(4):e181–e9.
https://doi.org/10.1016/S2352-3018(18)30009-2 PMID: 29490875
24. Barnhart DA, Harling G, Muya A, Ortblad KF, Mashasi I, Dambach P, et al. Structural, interpersonal,
psychosocial, and behavioral risk factors for HIV acquisition among female bar workers in Dar es
Salaam, Tanzania. AIDS Care. 2019; 31(9):1096–105. https://doi.org/10.1080/09540121.2019.
1612018.
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
16 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
25. Kalichman SC. Social and structural HIV prevention in alcohol-serving establishments: review of inter-
national interventions across populations. Alcohol Res Health. 2010; 33(3):184–94. https://pubmed.
ncbi.nlm.nih.gov/23584060/. PMID: 23584060
26. Pitpitan EV, Kalichman SC. Reducing HIV Risks in the Places Where People Drink: Prevention Inter-
ventions in Alcohol Venues. AIDS and behavior. 2016; 20 Suppl 1:S119–33. https://doi.org/10.1007/
s10461-015-1116-9 PMID: 26099244
27. Botao C, Horth RZ, Frank H, Cummings B, Inguane C, Sathane I, et al. Prevalence of HIV and Associ-
ated Risk Factors Among Long Distance Truck Drivers in Inchope, Mozambique, 2012. AIDS and
behavior. 2016; 20(4):811–20. https://doi.org/10.1007/s10461-015-1194-8 PMID: 26395193
28. Costenbader EC, Lancaster K, Bufumbo L, Akol A, Guest G. On the road again: concurrency and con-
dom use among Uganda truck drivers. African journal of AIDS research: AJAR. 2015; 14(2):117–25.
https://doi.org/10.2989/16085906.2015.1040810.
29. Morris CN, Ferguson AG. Estimation of the sexual transmission of HIV in Kenya and Uganda on the
trans-Africa highway: the continuing role for prevention in high risk groups. Sex Transm Infect. 2006; 82
(5):368–71. https://doi.org/10.1136/sti.2006.020933 PMID: 16854995
30. Makhakhe NF, Lane T, McIntyre J, Struthers H. Sexual transactions between long distance truck drivers
and female sex workers in South Africa. Glob Health Action. 2017; 10(1):1346164. https://doi.org/10.
1080/16549716.2017.1346164 PMID: 28764585
31. Kelvin EA, George G, Mwai E, Nyaga E, Mantell JE, Romo ML, et al. Offering self-administered oral
HIV testing to truck drivers in Kenya to increase testing: a randomized controlled trial. AIDS Care. 2018;
30(1):47–55. https://doi.org/10.1080/09540121.2017.1360997 PMID: 28826229
32. Birdthistle I, Tanton C, Tomita A, de Graaf K, Schaffnit SB, Tanser F, et al. Recent levels and trends in
HIV incidence rates among adolescent girls and young women in ten high-prevalence African countries:
a systematic review and meta-analysis. Lancet Glob Health. 2019; 7(11):e1521–e40. https://doi.org/10.
1016/S2214-109X(19)30410-3 PMID: 31607465
33. Blaizot S, Kim AA, Zeh C, Riche B, Maman D, De Cock KM, et al. Estimating HIV Incidence Using a
Cross-Sectional Survey: Comparison of Three Approaches in a Hyperendemic Setting, Ndhiwa Sub-
county, Kenya, 2012. AIDS Res Hum Retroviruses. 2017; 33(5):472–81. https://doi.org/10.1089/AID.
2016.0123 PMID: 27824254
34. Kharsany AB, Buthelezi TJ, Frohlich JA, Yende-Zuma N, Samsunder N, Mahlase G, et al. HIV infection
in high school students in rural South Africa: role of transmissions among students. AIDS Res Hum Ret-
roviruses. 2014; 30(10):956–65. https://doi.org/10.1089/AID.2014.0110 PMID: 25077861
35. Simbayi LC ZK, Zungu N, Moyo S, Marinda E, Jooste S, Mabaso M, Ramlagan S, North A, van Zyl J,
Mohlabane N, Dietrich C, Naidoo I, SABSSM V Team. South African National HIV Prevalence, Inci-
dence, Behavior and Communication Survey, 2017. HSRC Press. 2019.
36. Barnighausen T, Hosegood V, Timaeus IM, Newell ML. The socioeconomic determinants of HIV inci-
dence: evidence from a longitudinal, population-based study in rural South Africa. AIDS. 2007; 21 Suppl
7:S29–38. https://doi.org/10.1097/01.aids.0000300533.59483.95 PMID: 18040162
37. Hargreaves JR, Morison LA, Kim JC, Bonell CP, Porter JD, Watts C, et al. The association between
school attendance, HIV infection and sexual behaviour among young people in rural South Africa. J Epi-
demiol Community Health. 2008; 62(2):113–9. https://doi.org/10.1136/jech.2006.053827 PMID:
18192598
38. Pettifor AE, Levandowski BA, MacPhail C, Padian NS, Cohen MS, Rees HV. Keep them in school: the
importance of education as a protective factor against HIV infection among young South African
women. Int J Epidemiol. 2008; 37(6):1266–73. https://doi.org/10.1093/ije/dyn131 PMID: 18614609
39. Santelli JS, Edelstein ZR, Mathur S, Wei Y, Zhang W, Orr MG, et al. Behavioral, biological, and demo-
graphic risk and protective factors for new HIV infections among youth in Rakai, Uganda. J Acquir
Immune Defic Syndr. 2013; 63(3):393–400. https://doi.org/10.1097/QAI.0b013e3182926795 PMID:
23535293
40.
Jukes M, Simmons S, Bundy D. Education and vulnerability: the role of schools in protecting young
women and girls from HIV in southern Africa. AIDS. 2008; 22 Suppl 4:S41–56. https://doi.org/10.1097/
01.aids.0000341776.71253.04 PMID: 19033754
41. Groenewald C, Isaacs N, Isaacs D. Adolescent Sexual and Reproductive Health During the COVID-19
Pandemic: A Mini Review. Front Reprod Health. 2022; 4:794477. https://doi.org/10.3389/frph.2022.
794477 PMID: 36303613
42. McGrath N, Eaton JW, Newell ML, Hosegood V. Migration, sexual behaviour, and HIV risk: a general
population cohort in rural South Africa. Lancet HIV. 2015; 2(6):e252–9. https://doi.org/10.1016/S2352-
3018(15)00045-4 PMID: 26280016
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
17 / 18
PLOS GLOBAL PUBLIC HEALTH HIV epidemiologic trends among occupational groups in Rakai, Uganda
43. Phillips TK, Clouse K, Zerbe A, Orrell C, Abrams EJ, Myer L. Linkage to care, mobility and retention of
HIV-positive postpartum women in antiretroviral therapy services in South Africa. Journal of the Interna-
tional AIDS Society. 2018; 21 Suppl 4:e25114. https://doi.org/10.1002/jia2.25114 PMID: 30027583
44. Harling G, Muya A, Ortblad KF, Mashasi I, Dambach P, Ulenga N, et al. HIV risk and pre-exposure pro-
phylaxis interest among female bar workers in Dar es Salaam: cross-sectional survey. BMJ open. 2019;
9(3):e023272. https://doi.org/10.1136/bmjopen-2018-023272 PMID: 30898799
45. Grimsrud A, Barnabas RV, Ehrenkranz P, Ford N. Evidence for scale up: the differentiated care
research agenda. Journal of the International AIDS Society. 2017; 20(Suppl 4):22024. https://doi.org/
10.7448/IAS.20.5.22024 PMID: 28770588
46. Okoboi S, Ding E, Persuad S, Wangisi J, Birungi J, Shurgold S, et al. Community-based ART distribu-
tion system can effectively facilitate long-term program retention and low-rates of death and virologic
failure in rural Uganda. AIDS Res Ther. 2015; 12:37. https://doi.org/10.1186/s12981-015-0077-4 PMID:
26566390
47. Alamo ST, Wagner GJ, Ouma J, Sunday P, Marie L, Colebunders R, et al. Strategies for optimizing
clinic efficiency in a community-based antiretroviral treatment programme in Uganda. AIDS and behav-
ior. 2013; 17(1):274–83. https://doi.org/10.1007/s10461-012-0199-9 PMID: 22610422
48. Angel S, Parent J, Civco DL, Blei A, Potere D. The dimensions of global urban expansion: Estimates
and projections for all countries, 2000–2050. Progress in Planning. 2011; 75(2):53–107. https://doi.org/
10.1016/j.progress.2011.04.001.
49. Dyson T. HIV/AIDS and Urbanization. Population and Development Review. 2004; 29(3):427–42.
https://doi.org/10.1111/j.1728-4457.2003.00427.x.
50. Maulide Cane R, Melesse DY, Kayeyi N, Manu A, Wado YD, Barros A, et al. HIV trends and disparities
by gender and urban-rural residence among adolescents in sub-Saharan Africa. Reprod Health. 2021;
18(Suppl 1):120. https://doi.org/10.1186/s12978-021-01118-7 PMID: 34134720
51. Shelton JD, Cassell MM, Adetunji J. Is poverty or wealth at the root of HIV? Lancet. 2005; 366
(9491):1057–8. https://doi.org/10.1016/S0140-6736(05)67401-6 PMID: 16182881
52. Hargreaves JR. Socioeconomic status and risk of HIV infection in an urban population in Kenya. Trop
Med Int Health. 2002; 7(9):793–802. https://doi.org/10.1046/j.1365-3156.2002.00943.x PMID:
12225512
53. Awusabo-Asare K, Annim SK. Wealth status and risky sexual behaviour in Ghana and Kenya. Appl
Health Econ Health Policy. 2008; 6(1):27–39. https://doi.org/10.2165/00148365-200806010-00003
PMID: 18774868
PLOS Global Public Health | https://doi.org/10.1371/journal.pgph.0002891 February 20, 2024
18 / 18
PLOS GLOBAL PUBLIC HEALTH
| null |
10.1038_s41467-023-38529-y.pdf
|
Data availability
Spike recording data in NWB format are available for download at
https://dandiarchive.org/dandiset/000060/draft. Source data are
provided with this paper.
|
Data availability Spike recording data in NWB format are available for download at https://dandiarchive.org/dandiset/000060/draft . Source data are provided with this paper. Code availability The Julia code for training spiking neural network is available at https:// github.com/SpikingNetwork/distributedActivity 75 .
|
Article
https://doi.org/10.1038/s41467-023-38529-y
Distributing task-related neural activity
across a cortical network through
task-independent connections
Received: 18 July 2022
Accepted: 5 May 2023
Check for updates
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
Christopher M. Kim 1,2
Karel Svoboda 2,5 & Ran Darshan 2
, Arseny Finkelstein 3,4, Carson C. Chow 1,
Task-related neural activity is widespread across populations of neurons
during goal-directed behaviors. However, little is known about the synaptic
reorganization and circuit mechanisms that lead to broad activity changes.
Here we trained a subset of neurons in a spiking network with strong synaptic
interactions to reproduce the activity of neurons in the motor cortex during a
decision-making task. Task-related activity, resembling the neural data,
emerged across the network, even in the untrained neurons. Analysis of
trained networks showed that strong untrained synapses, which were inde-
pendent of the task and determined the dynamical state of the network,
mediated the spread of task-related activity. Optogenetic perturbations sug-
gest that the motor cortex is strongly-coupled, supporting the applicability of
the mechanism to cortical networks. Our results reveal a cortical mechanism
that facilitates distributed representations of task-variables by spreading the
activity from a subset of plastic neurons to the entire network through task-
independent strong synapses.
Large-scale measurements of neural activity show that learning can
rapidly change the activity of many neurons, resulting in widespread
changes in task-related neural activity1–5. For instance, a goal-directed
behavior involving motor planning leads to widespread changes across
the motor cortex1.
To gain insights into the circuit mechanism behind the observed
widespread activity, it is critical to understand how interconnected
neural circuits modulate their synaptic connections to produce the
observed changes in task-related neural activity. Tracking synaptic
modifications during learning6–9 and manipulating them to demon-
strate a causal link with behavioral outputs10–14, show that synaptic
plasticity underlies learned behaviors and changes in neural
activity15,16. However, it is highly challenging to conduct multi-scale
experiments that monitor and manipulate learning-specific synaptic
changes at cellular resolution across a wide region of cortical
networks, while measuring the resulting neural activity17. Thus, it
remains unclear what aspects of the synaptic connections are modified
to produce the widespread changes in task-related neural activity.
Here we investigated if the task-related activity, learned locally by
modifying synaptic inputs to a dedicated subset of neurons, can
spread across the network through pre-existing, task-independent,
synaptic pathways. Although distributed neural activity may result
from broad changes in synaptic connections across a neural network,
we hypothesized that recruiting only a small subset of neurons is
sufficient to generate the distributed task-related neural activity. To
test this hypothesis, we used recurrent neural networks that provide a
powerful data-driven approach for investigating how synaptic mod-
ifications can support the observed task-related neural activity18–22.
In typical implementations of network training, the synaptic
inputs to all the neurons in the network are considered to be plastic, in
1Laboratory of Biological Modeling, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD, USA. 2Janelia
Research Campus, Howard Hughes Medical Institute, Ashburn, VA, USA. 3Department of Physiology and Pharmacology, Sackler Faculty of Medicine, Tel Aviv
University, Tel Aviv, Israel. 4Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel. 5Allen Institute for Neural Dynamics, Seattle, WA, USA.
e-mail: [email protected]; [email protected]
Nature Communications |
(2023) 14:2851
1
Article
https://doi.org/10.1038/s41467-023-38529-y
that the activity of every neuron is fit to the activity of experimentally
recorded neurons, thereby constraining the entire network activity to
the neural data19–21. In this study, we instead trained the synaptic inputs
to only a subset of neurons in a biologically plausible network to
reproduce the activity of recorded neurons. The network consisted of
excitatory and inhibitory populations of spiking neurons with sparse
and strong connections23–25. Such pre-existing, task-independent,
connections made the network settle into a cortical-like dynamical
regime, where excitation and inhibition balanced each other23,25,26,
resulting in temporally irregular spikes and heterogeneous spike
rates27.
We applied our modeling framework to study the spread of task-
related activity in the anterior lateral motor cortex (ALM) of mice
performing a memory-guided decision-making task21. Similarly to
neurons in the primate motor cortex28–30, the activity of many neurons
in ALM ramps slowly during motor preparation and is selective to
future actions21,31,32. These task-related activity patterns are widely
distributed across the ALM and are highly heterogeneous across
neurons.
When a small number of synapses was trained to reproduce the
ALM activity in a subset of neurons, we found that, surprisingly, the
emerging activity in the untrained model neurons closely matched
the responses of ALM neurons held out from training. In other words,
the task-related ALM activity, learned by modifying synaptic inputs to
only a subset of neurons, spread to other untrained neurons in the
network without further training and produced activity that resembled
the actual responses of ALM neurons. Analysis of the trained networks
revealed that the pre-existing strong synapses between the neurons
mediated the propagation of the task-related activity. The trained
activity failed to spread in networks of neurons that were not strongly
coupled. Optogenetic perturbation experiments of ALM activity pro-
vided additional evidence that the ALM network is strongly coupled,
supporting the applicability of the proposed mechanism for spreading
the trained activity to cortical networks.
Our work provides a general circuit mechanism for spreading
activity in cortical networks. It suggests that task-related activity
observed in cortical regions during behavior can emerge from sparse
synaptic reorganization to a subset of neurons and then propagate to
the rest of the network through the strong, task-independent
synapses.
Results
Training strongly coupled spiking networks with sparse
synapses
Our spiking network was based on a cortical circuit model that pro-
vided mechanistic explanation of the canonical features of cortical
activity, such as temporally irregular spike trains, large trial-to-trial
variability and a wide range of firing rates across neurons23–25,27,33. It
consisted of excitatory (E) and inhibitory (I) neurons sparsely and
randomly connected by strong synapses (Fig. 1A, solid arrows). This
initial EI network structure, due to its random connectivity, was inde-
pendent of the task to be learned. In addition, the strong excitatory
and inhibitory synaptic inputs were dynamically balanced to maintain
a stable network state, known as the balanced regime. As in the cortex,
neurons, driven by fast fluctuating synaptic inputs, emitted spikes
stochastically in this network state.
We developed a training scheme to train these spiking networks,
while keeping them in the balanced regime (see Methods and below).
We used this scheme to train sparse synapses to a subset of neurons in
the EI spiking network, referred to as Subset Training, to generate
target activity patterns in the subset of neurons, while keeping the
synaptic inputs to rest of the neurons untrained (Fig. 1A, left). After
training the synaptic inputs to the selected subset of neurons, we
analyzed if the learning-related changes in activity spread throughout
the network (Fig. 1A, right).
p
To model the effects of learning in the subset of neurons, we
introduced a relatively small number of plastic synapses (Fig. 1A,
magenta arrows) to an existing EI network (Fig. 1A, solid arrows), with
no overlap between the plastic and existing EI synapses. The plastic
synapses were connected to the selected subset of neurons from
randomly chosen excitatory and inhibitory neurons in the network and
also from a pool of external neurons emitting stochastic spikes mod-
eled by the Poisson process (see Methods for details). These plastic
synapses were sparser than the static, task-independent (random) EI
connections, in part, motivated by the synaptic connectivity found in
the cortex that is sparse but functionally biased34. For instance, with
≈30 plastic
K = 1000 static synapses, there were of the order of
synapses per neuron (Fig. S1). Superimposing plastic synapses to the
existing EI network resulted in synaptic input to the subset that con-
sisted of two components: 1) a component that entered through the
strongly coupled random EI network connectivity that were not
trained, but made the network operate in the balanced regime (Fig. 1B,
ubal), and 2) a plastic component that entered through the plastic
synapses that were optimized by the learning process (Fig. 1B, uplas).
During network training, a synaptic learning rule based on recursive
least squares19,35–37 optimized the strength of plastic synapses to neu-
rons in the subset, so that total input to each trained neuron (Fig. 1B,
ubal + uplas) followed the neuron’s target activity pattern (Fig. 1B, cyan
sine wave). We note that the plastic connections to trained neurons
were allowed to flip their signs after training (see Fig. S8A for the
distribution of plastic weights and Methods); the untrained neurons,
on the other hand, only received synaptic inputs through the initial EI
network connections.
ffiffiffiffi
K
This arrangement of plastic synapses, which connected only to
the selected subset of neurons, allowed us to examine the role of the
pre-existing recurrent connections of the EI network in spreading the
trained activity to the untrained neurons, which were not targeted by
the plastic synapses. In addition, due to their sparsity, the plastic
synaptic inputs were substantially weaker than the strong excitatory
and inhibitory synaptic inputs of the existing EI network (Fig. 1B). This
allowed the network to stay in the balanced regime after training and
supported robust network training, independent of the density of
synaptic connections (Fig. S1; see Methods for full description of the
training and details on the sparse plastic synapses).
In the trained network, the total synaptic input to each trained
neuron was able to successfully follow the target patterns (Fig. 1B, left;
Fig. 1C). The statistics of spiking activity of the trained network were
similar to those of untrained, strongly coupled EI networks, thus
consistent with the spiking activity of cortical neurons. Specifically,
due to the highly fluctuating balanced input, the spike trains of each
neuron were irregular and exhibited large trial-to-trial variability
(Fig. 1D, Fano factor = 1.4)23,24,38. The firing rate distribution was also
highly skewed and was well approximated by a log-normal distribution
(trained model: Fig. 1E, neural data: Fig. S2D)27.
We demonstrate in following sections that the temporally irre-
gular and heterogeneous spiking activity is not just cosmetics. Instead,
the strongly coupled excitatory-inhibitory connections responsible for
generating noisy spikes have consequences on how task-related neural
activity is represented in the cortical network.
Spread of trained neural activity to untrained neurons
We applied the Subset Training method to reproduce the firing rate
patterns of cortical neurons recorded from the anterior lateral motor
cortex (ALM) during a memory-guided decision-making task21. Mice
learned to respond to an optogenetic stimulation of neurons in the
vibrissal somatosensory cortex (vS1) by licking right when stimulated
and licking left otherwise (Fig. 2A). For training networks and analysis
of trained networks, we used the electrophysiological recordings in
ALM of the spiking activity of putative pyramidal neurons (excitatory;
Npyr = 1824) and putative fast-spiking neurons (inhibitory; Nfs = 306)
Nature Communications |
(2023) 14:2851
2
Article
https://doi.org/10.1038/s41467-023-38529-y
Fig. 1 | Training a subset of neurons in a strongly coupled spiking neural net-
work with sparse plastic synapses. A Schematic of the Subset Training method
(left). The network consisted of excitatory (green) and inhibitory (orange) neurons.
Selected neurons (dark magenta, left) were trained to generate task-related activity
patterns, modeled here as 1Hz sine waves with random phases (cyan curves), by
modifying plastic synapses (dashed arrows, magenta) to the selected neurons. The
static synapses (solid arrows, excitatory: green, inhibitory: orange) remained
unchanged throughout training. External stimulus (blue pulse) triggered the neu-
rons to generate the trained activity patterns. After training (right), task-related
activity could potentially spread to the rest of the untrained neurons (light
magenta, right). B The total synaptic input (left panel, in arbitrary units (a.u.)) to a
trained neuron followed the target pattern (cyan) when triggered by an external
stimulus (blue region). The total input is the sum of the balanced input, denoted as
ubal, from static synapses (black; middle panel) and the plastic input, denoted as
uplas, from plastic synapses (magenta; right panel). The balanced and plastic inputs
can be further divided into excitatory (green) and inhibitory (orange) inputs. The
spike-threshold of the neuron is at 1 (red dotted line). Note the scale difference
between the balanced and plastic inputs. C Additional examples of the total
synaptic inputs (same as the left panel in (B)) to trained neurons (bottom) following
the target patterns (cyan); the 200ms moving average is shown in gray. Spike trains
emitted by the neurons across 30 trials are shown on the top. D Fano factor of spike
counts across 30 trials. E The log of firing rates of trained neurons. All neurons in
the network were trained in this demonstration of the Subset Training method.
Source data are provided as a Source Data file.
when the mice responded correctly to lick-right and lick-left
conditions.
We asked what aspects of the network connectivity should
change to reproduce the activity of ALM neurons in a strongly cou-
pled spiking neural network. In previous studies, in which networks
were trained to generate specific patterns of neural activity, all the
units in the network were trained to reproduce the target activity
patterns18–21. Here, we trained only a subset of neurons, embedded in
the strongly coupled EI network, to reproduce the spiking activity of
ALM neurons. By analyzing the activity of neurons in the trained
network, we found that synaptic reorganization to a subset of neu-
rons was sufficient to generate the observed ALM activity throughout
the entire network, including the untrained neurons. Importantly,
the spread of target activity patterns from the subset of trained
neurons to the rest of neurons was not observed in a network that
was not strongly coupled (Figs. S9, S10). This suggests that the
spread of trained activity to untrained neurons is a characteristic of
strongly-coupled networks, but not a general outcome of recurrent
networks.
The network connectivity of initial balanced network was set up,
such that the excitatory and inhibitory population rates were con-
sistent with the population rates of ALM pyramidal and fast-spiking
neurons, respectively. In addition, the firing rates of model and ALM
neurons were both log-normally distributed27, which allowed us to
easily pair each ALM neuron with a model neuron to be trained based
on the proximity of their mean firing rates (Fig. S2D, Methods). Our
Nature Communications |
(2023) 14:2851
3
Article
https://doi.org/10.1038/s41467-023-38529-y
Fig. 2 | Reproducing ALM activity in a subset of neurons and the spread of
trained activity to the entire network. A Schematic of the experimental setup.
Mice learned to lick right when the optogenetic simulation was delivered to
vibrissal somatosensory (vS1) neurons and to lick left when there was no stimu-
lation. The spiking activity of ALM neurons was recorded during the task. B Trial-
averaged firing rates and raster plots of the spike trains across multiple trials (lick-
right: blue, lick-left: red). Trained excitatory model neurons (top) and pyramidal
ALM neurons used for training the excitatory model neurons (bottom). C A subset
of excitatory neurons in the spiking neural network learned to reproduce the
PSTHs of pyramidal ALM neurons. The rest of the neurons in the network were
not trained. After training, the activity of untrained inhibitory model neurons was
compared to the activity of fast-spiking ALM neurons. D Trial-averaged firing
rates and raster plots of the spike trains across multiple trials (lick-right: blue, lick-
left: red). Untrained inhibitory model neurons (top) and fast-spiking ALM neurons
(bottom) that best resembled the PSTHs of the inhibitory model neurons (see also
Fig. 3 and Fig. S4). E, F The principal components (PCs) of the PSTHs of excita-
tory/pyramidal and inhibitory/fast-spiking neurons (model/data) and the cumu-
lative variances explained by the PCs. Source data are provided as a Source
Data file. Adapted from ref. 76.
modeling approach assumed that the firing rate dynamics generating
the noisy spike trains of ALM neurons change smoothly in time.
Hence, for the training targets, we used smoothed trial-averaged
peri-stimulus time histogram (PSTH) of pyramidal neurons recorded
in ALM during the delay period (Fig. 2B, bottom; for details see
Methods).
Following this training scheme, we trained a subset of excitatory
neurons in the model network to reproduce the target activity patterns
(73% of the excitatory or 36% of all the neurons, Fig. 2C). Each trained
neuron received recurrent plastic synapses from randomly selected
excitatory and inhibitory neurons in the network and feedforward
plastic synapses from external neurons, which accounted for the
potential inputs from outside the ALM. By modifying the plastic
synapses, the trained neurons reproduced two PSTHs, corresponding
to lick-right and lick-left conditions, when evoked by two different
stimuli. The rest of the excitatory, as well as all of the inhibitory neu-
rons in the network, were not trained (Fig. 2C).
After training, the firing rate of trained excitatory neurons suc-
cessfully reproduced the PSTHs of pyramidal neurons (Fig. 2B, top;
Fig. S2A,B), even though the plastic synaptic inputs were substantially
weaker than the excitatory and inhibitory inputs from the static
synapses (Fig. S2C, right). To estimate the smooth PSTHs in model
neurons, we simulated the trained network over multiple trials and
used the trial-averaged firing rates of the model neurons (the
smoothness of which depended on the number of trial averages). We
estimated the correlations between single neuron PSTHs in the model
and in the data (Fig. S2C, left), as well as the similarity in their popu-
lation activity (Fig. 2E, left) to asses the success of the training. For the
latter, we performed Principal Component Analysis (PCA) on the
PSTHs of neurons, which is a dimensionality reduction technique used
for identifying a set of activity patterns that captures a large fraction of
variance in the population activity. We found that the projection of the
PSTH of a pyramidal ALM neuron onto the first PC was a good indicator
for how well a trained excitatory neuron could fit the pyramidal ALM
Nature Communications |
(2023) 14:2851
4
Article
https://doi.org/10.1038/s41467-023-38529-y
neurons (Fig. S2C). The principal components (PCs) of the PSTHs of
the trained excitatory neurons closely matched the PCs of the pyr-
amidal neurons. Moreover, the first six PCs explained close to 80% of
the trained neurons’ activity, thus the recurrent network displayed
low-dimensional dynamics as in the pyramidal neurons in ALM (Fig. 2E,
right)39.
Next, we examined the activity of the untrained model neurons
(64% of the neurons). Similarly to the trained excitatory neurons, their
activity tended to ramp before go-cue and was choice-selective
(Fig. 2D, top). The PCs of their PSTHs were almost identical to the
trained excitatory neurons (Fig. 2F, right; Fig. S3E). This finding showed
that cortical-like activity generated within the subset of excitatory
neurons spread to the rest of the network without additional synaptic
reorganization to the untrained neurons.
Finally, we found that the PCs of the PSTHs of the fast-spiking
ALM neurons, whose activity was not learned by the network, were
almost identical to the PCs of the untrained inhibitory model neu-
rons (Fig. 2F, right). A good agreement between the untrained model
neurons and the held-out neural data supported the hypothesis that
cortical-like activity learned within a subset of neurons can spread
and produce cortical-like activity in the entire network. This could
explain why the activity of putative fast-spiking neurons in ALM is
heterogeneous, yet is very similar to the activity of putative pyr-
amidal neurons21.
Similarity between untrained model neurons and ALM neurons
To further investigate the similarities between the activities of the
untrained inhibitory neurons in the trained network and the fast-
spiking ALM neurons, we compared their PSTHs at the single neuron
and population levels.
At the single neuron level, we identified an untrained inhibitory
neuron that best matched each fast-spiking ALM neuron, based on the
mean-squared-error of the PSTHs of all possible pairings between the
ALM neuron and the population of inhibitory model neurons. Fig-
ure 3A shows the PSTHs of several matched pairs and their correlations
for the lick-right and lick-left trials (see Fig. S4 for all the matched
pairs). Evaluating the correlations of all the matched pairs showed that
they were significantly higher than the spurious correlations obtained
by matching the fast-spiking ALM neurons to inhibitory neurons in an
untrained balanced network (Fig. 3B, left; two sample Kolmogorov-
Smirnov tests; p-value < 0.0001).
To elucidate which aspects of the fast-spiking ALM activity were
captured by the untrained inhibitory neurons in the trained network,
we examined if certain activity patterns of the fast-spiking ALM neu-
rons were indicative of the goodness-of-fit to the model neurons. We
found that the projection of the PSTHs of the fast-spiking ALM neurons
onto their first PC, a ramping mode that captured over 70% of the
variance in the ALM activity (see PC1 in Fig. 2F), was a good indicator
for how well the untrained model neurons could fit the fast-spiking
ALM neurons (Fig. 3B, right). This analysis suggested that the ramping
mode was the dominant component of the trained activity that was
transferred to the untrained inhibitory neurons and shared with the
fast-spiking ALM neurons.
We systematically examined the activity patterns shared by the
populations of untrained inhibitory neurons and fast-spiking ALM
neurons, by analyzing the shared-variance between the two population
activities. The shared-variance analysis identified population vectors
along which two population activities co-varied maximally and yielded
population-averaged activity along those directions (shared compo-
nents) and the proportion of variance explained by the shared com-
ponents (shared variance; see40 and Methods for details). The shared
components (SCs) were similar to the PCs of the untrained inhibitory
subnetwork and fast-spiking ALM activities (compare the SCs in Fig. 3C
to the PCs in Fig. 2F), and the first four components captured most of
the shared variance (Fig. 3C). In particular, consistent with the single
neuron analysis shown in Fig. 3B, the first shared component was a
ramping mode (SC1 in Fig. 3C).
In addition to the spiking activity patterns, we asked if functional
properties, such as choice selectivity, were transferred to the
untrained neurons in the trained network. It has been shown that
pyramidal ALM neurons in mice display selectivity to the animal’s
choice21, 39,41 (Fig. S5; absolute value of selectivity index: 0.43 ± 0.35,
mean ± SD; see Methods). As expected, the excitatory model neurons,
trained to reproduce the activity of pyramidal ALM neurons, also dis-
played choice selectivity (absolute selectivity: 0.33 ± 0.28). Interest-
ingly, we also found that the fast-spiking ALM neurons in the neural
data were choice selective (Fig. 3D,E; see also Supplementary Fig. 2 in21;
absolute selectivity: 0.40 ± 0.39). These observations led us to exam-
ine if the untrained inhibitory neurons in the trained network exhibited
choice selectivity, as in the fast-spiking ALM neurons.
To this end, we analyzed the difference of the PSTHs to two trial
types (lick-right versus lick-left) in all the untrained inhibitory neu-
rons and found that they displayed choice selectivity (Fig. 3D;
absolute selectivity: 0.22 ± 0.19, compared with 0.031 ± 0.036 of the
null model of Fig. S10). Moreover, the distribution of the choice
selectivity of fast-spiking ALM neurons and untrained inhibitory
neurons were in good agreement (Fig. 3E), although the selectivity of
the inhibitory model neurons were slightly weaker than the fast-
spiking ALM neurons, potentially due to the weaker selectivity of
trained excitatory model neurons with respect to selectivity of pyr-
amidal neurons, caused by imperfect training. This finding shows
that not only the trained neural activity can propagate throughout
the network, but the choice selectivity emerged in a subset of neu-
rons can spread to the untrained parts of the network as well. In
particular, this suggests an alternative mechanism for how selectivity
may emerge in inhibitory neurons. In contrast to previous models
that required specific connectivity between excitatory-inhibitory
neurons for selective responses to emerge42,43, our model suggests
that choice selectivity in inhibitory neurons can arise in strongly
coupled networks even when the connections to the inhibitory
neurons are non-specific.
Training inhibitory neurons improves the spread of activity
So far, we showed that the cortical-like activity originating from the
excitatory neurons can spread to the untrained inhibitory neurons.
Next, we asked how the spreading of trained activity may depend on
the type of neurons being trained. To address this question, we con-
sidered two training scenarios where either the excitatory or the
inhibitory subnetwork (but not both) was trained to generate the tar-
get activity patterns (Fig. 4A, right).
The number of fast-spiking ALM neurons recorded from the mice
(Nfs = 306) was, however, too small to train the inhibitory neurons in
large-scale spiking neural networks. We thus developed a method to
generate synthetic neural activity that had similar low-dimensional
dynamics as the ALM neurons. Briefly, we first performed principal
component analysis on the PSTHs of ALM neurons to obtain the PCs
(Fig. 2E,F) and the empirical distribution of each PC’s loading on the
neurons. To construct a synthetic target activity for neuron i, we
sampled (1) a baseline rate ri from the firing rate distribution and (2)
each PC’s loading on the neuron from the empirical distribution,
conditioned on the rate ri (see Fig. S6 and Methods for details).
Applying this method to the lick-left and lick-right trial types and to
pyramidal and fast-spiking neuronal types, we were able to generate an
unlimited number of cortical-like PSTHs needed for training large-
scale networks consisting of, e.g., N = 30, 000 neurons. In particular,
these synthetic neural activities had statistically identical
low-
dimensional dynamics as the ALM neurons (Fig. S6E).
Using the synthetic neural activity as the target activity patterns,
we performed the two training scenarios where we trained a subset of
neurons in the excitatory or the inhibitory subnetwork to reproduce
Nature Communications |
(2023) 14:2851
5
Article
https://doi.org/10.1038/s41467-023-38529-y
Fig. 3 | Untrained inhibitory neurons in the trained network display similar
task-related activity to the fast-spiking neurons recorded in the ALM.
A Examples of the PSTHs of untrained inhibitory neurons (lick-left: red, lick-right:
blue) that best fit the PSTHs of fast-spiking ALM neurons (black). Correlations of the
matched pairs are shown in each panel. B Correlations between the PSTHs of all the
matched inhibitory model neurons and the fast-spiking ALM neurons for the lick-
right and lick-left trial types (left). The null network shows the correlation between
the PSTHs of the fast-spiking ALM neurons and the best-fit neurons in the initial
balanced network, i.e., without training. The PSTHs of the neurons in the trained
and the null networks were both obtained by averaging the spike trains from 400
trials starting at random initial conditions. The p-values of the two-sample Kol-
mogorov-Smirnov tests between the trained and null networks for both trial types
are shown (p-value = 10−51 (Right), 10−18 (Left)). PC1 (right) represents the projection
of the PSTH of a fast-spiking ALM neuron onto the first PC, i.e., the ramping mode
(see Fig. 2F). R-squared value of the linear regression is shown. C Shared compo-
nents (SC) and the cumulative shared variance explained by them for the lick-right
(top) and lick-left trial types (bottom). The null network shows the shared variance
between the fast-spiking ALM neurons and the initial balanced network. D Choice
selectivity of all the untrained inhibitory neurons in the trained network (left) and
the fast-spiking ALM neurons (middle). Choice selectivity was defined at each time
point as the difference of the PSTHs to the lick-right and lick-left trial types, and
then normalized by the average firing rate of each neuron. E Distribution of choice
selectivity of untrained inhibitory neurons in the trained network (orange) and fast-
spiking ALM neurons (black). Choice selectivity of a neuron shown here was
obtained by averaging the choice selectivity over the 2 second time window shown
in (D). Note that there are more right selective fast-spiking ALM neurons than
expected by the model. This might result from asymmetries in the task design. The
mouse is instructed to lick right by optogenetic activation of sensory neurons,
while it learned to lick left in the absence of such activation. In addition, most of the
data were acquired from left ALM, which previous studies also showed that this
leads to a bias for right selective neurons (e.g.21). We did not model these effects.
Source data are provided as a Source Data file.
the synthetic neural activity. Following training, we compared the
spiking activities of the untrained neurons in the subnetworks that
were not trained.
We first observed that the PCs of synthetic neural activity was
transferred to the untrained neurons when a sufficient number of
neurons were trained (Fig. 4D, right). Such transfer of PCs was similar
to what we found in the untrained inhibitory neurons when the exci-
tatory neurons were trained on the activity of ALM pyramidal neurons
(Fig. 2E,F). Based on the transfer of PCs and the low dimensionality of
ALM activity, we used the variance explained by the first six PCs of the
PSTHs of the untrained neurons to quantify the transferred cortical-
like activity. In the trained neurons, the first six PCs explained 80% of
the activity, regardless of the trained neuronal type (E or I) or the
fraction of trained neurons (Fig. S7A). On the other hand, the cortical-
like activity transferred to the untrained neurons gradually increased
with the fraction of trained neurons. Moreover, the transferred activity
was stronger by 20% when the inhibitory subnetwork was trained,
compared to when the excitatory subnetwork was trained (Fig. 4A,
left). Using the first six PCs of the ALM fast-spiking neurons (i.e., data
PCs), instead of the PCs of untrained model neurons (i.e., model PCs),
Nature Communications |
(2023) 14:2851
6
Article
https://doi.org/10.1038/s41467-023-38529-y
Fig. 4 | Trained activity originating from the inhibitory subnetwork spreads
more effectively than the trained activity from the excitatory subnetwork.
A Schematic of two training scenarios (right). A subset of neurons in the excitatory
subnetwork (top) or the inhibitory subnetwork (bottom) was trained to reproduce
synthetic neural activity. The fraction equals 1 (left) if all the neurons in the trained
subnetwork are trained. The transferred activity was defined as the variance
explained by the first six PCs of the PSTHs of all the neurons in the untrained
subnetwork. B The strength of PCs of the PSTHs of pyramidal and fast-spiking ALM
neurons (left). The absolute value of the loading of each PC on all the neurons in the
population was averaged to obtain the average strength of each PC, denoted as k.
Examples of centered PSTHs (right), i.e., mean rate subtracted, showing that the
strength of the ith PC, denoted as ki, was stronger in the fast-spiking ALM neurons.
C The modulation of the trained synthetic inhibitory rate was adjusted by scaling
the centered PSTH by a multiplicative factor, referred to as the relative strength of
modulation. For instance, it equals 2 if the centered PSTHs are doubled. As in (A),
the transferred activity was defined as the variance explained by the first six PCs of
the PSTHs of all the untrained excitatory neurons. The fraction of trained inhibitory
neurons in the inhibitory subnetwork was 0.4. D The fidelity of transferred PCs
(left) was defined by the correlation between the PCs of the trained and transferred
activity. A subset of neurons in the excitatory subnetwork was trained, and the
activity of the untrained inhibitory subnetwork was analyzed to obtain the trans-
ferred PCs. Examples of transferred PCs in the untrained inhibitory neurons (right)
are shown, as the fraction of trained neurons is varied.
to quantify the cortical-like activity in the untrained neurons yielded
similar results.
To understand what allowed the activity patterns of inhibitory
neurons to spread better to the untrained neurons, we examined the
differences in the spiking activities of the pyramidal and fast-spiking
ALM neurons. The mean firing rate of each neuron was subtracted
from its PSTH to remove the differences in the baseline firing rates of
the pyramidal and fast-spiking ALM neurons (Fig. 4B, right). The
principal component analysis of the centered PSTHs revealed that the
strength of every PC was stronger in the fast-spiking neurons than in
the pyramidal neurons, when the loadings on each PC were averaged
over the population of neurons (Fig. 4B, left). This analysis showed that
the modulation of firing rate around the mean rate was larger in the
fast-spiking neurons, raising the possibility that stronger rate mod-
ulation leads to stronger activity transfer.
To test if stronger modulations in the trained activity patterns
would increase the transferred activity to the untrained neurons, we
adjusted the modulation strength of the synthetic inhibitory activity
and trained a fixed subset of inhibitory neurons (40% of the inhibitory
neurons) to generate target activity patterns with different levels of
rate modulations. We found that, in the untrained excitatory neurons,
the variance explained by the cortical-like activity increased
monotonically with the modulation strength of the trained inhibitory
neurons. (Fig. 4C). These results suggested that the stronger rate
modulations in the fast-spiking ALM neurons enabled the trained
inhibitory neurons in the model to spread their activity patterns to the
untrained neurons more effectively. It also suggested that inhibitory
neurons, whose baseline spiking rates are typically higher than the
excitatory neurons in cortex (e.g., mean firing rates of ALM pyramidal
and fast-spiking neurons were ~ 4Hz and ~ 11Hz, respectively, in our
data), can support stronger rate modulations and potentially play a
more significant role in spreading the trained activity patterns.
The finding that activity patterns with strong rate modulation
spread better was also observed across the PCs. The leading PC modes
of the ALM spiking activity showed stronger modulation than the other
PC modes, as expected, since the leading PC modes capture more
variance (Fig. 4B). To quantify how well the trained PCs transferred to
the untrained neurons, we computed the correlation between the PC
modes of the trained and transferred activity (Fig. 4D, left). The leading
PC modes (PC1 to PC3) transferred with high fidelity even when only
20% of the neurons were trained. On the other hand, the transfer of
PC4 to PC6 improved gradually when the fraction of trained neurons
increased. This result suggested that the leading PC modes, due to
their strong modulations, can spread more robustly to the rest of the
Nature Communications |
(2023) 14:2851
7
Article
https://doi.org/10.1038/s41467-023-38529-y
neurons, promoting low-dimensional neural dynamics across a
strongly coupled network.
Taken together, our results demonstrate that trained activity
patterns with stronger rate modulations, which can emerge from the
fast-spiking ALM neurons or leading PC modes, have greater influence
on the untrained neurons in the network.
A network mechanism for distributing trained neural activity
In a recurrent neural network, in which neurons are highly inter-
connected, it may seem obvious that task-related activity can spread
from one part of the network to another part through the recurrent
connections. However, this intuition becomes less clear when the
activity of a neuron is determined by integrating a large number of
heterogeneous presynaptic activities through synapses that are not
optimized for the task, as considered in our network model and is the
case in the cortical network. Indeed, a close examination of networks
with a large number of connections reveals that whether the task-
related activity can spread depends on the operating regime of the
network, as well as on the coherence level of the learned task-related
activity.
When the activity of the trained neurons is coherent, for example,
if most neurons would increase their firing rates before the go-cue in
the delayed-response task, then indeed activity could spread to the
untrained neurons, which will also ramp-up before the go-cue. This will
result in a coherent task-related activity, in which both trained and
untrained neurons are increasing their rate before the go cue. How-
ever, the activity of neurons in ALM during the delayed-response task
are highly heterogeneous and are far from being coherent (examples
in Fig. S2B and also Fig. 5H below). In fact, the average firing rate of the
neurons barely varies during the delay period31. Thus, if the synaptic
connections to an untrained neuron randomly sample and sum het-
erogeneous activity patterns of pre-synaptic neurons, one could
expect that the post-synaptic input to the untrained neuron will be
averaged-out. Then, the untrained neuron will not display any task-
related activity patterns.
To directly demonstrate that ALM activity patterns do not spread
if the network does not operate in the balanced regime, we con-
structed a network whose synaptic weights merely averaged the
spiking activities of presynaptic neurons. Unlike the balanced network
that internally generated highly fluctuating synaptic currents, we
injected external noise to neurons in this network to mimic the sto-
chastic spiking of cortical neurons (see Methods and Fig. S10 for
details). We found that the trained excitatory neurons successfully
reproduced the spiking activity of ALM pyramidal neurons and showed
choice selectivity. In contrast, the untrained inhibitory neurons did not
exhibit any temporally structured activity patterns, and, when mat-
ched with the activity patterns of ALM fast-spiking neurons, the overall
correlation of the best matched pairs was indistinguishable from a null
model of an untrained balanced network. Moreover, the untrained
inhibitory neurons did not exhibit choice selectivity (Fig. S10). These
findings demonstrate that the spread of heterogeneous ALM activity to
untrained neurons is not a general property of recurrent neural net-
works (see also Fig. S9).
In this section, we give an intuitive explanation that hetero-
geneous activity does spread if the network is strongly coupled and
operates in the balanced regime (Fig. 5). In this regime, presynaptic
activity patterns can be preserved in the post-synaptic inputs to
untrained neurons due to the strong synapses, and then manifested in
the untrained neuron’s spiking rate due to the dynamic cancellation of
the large, unmodulated components of the excitatory and inhibitory
inputs. A detailed explanation, together with a mathematical analysis,
is given in the Methods.
To explain the network mechanism underlying the spread of
trained activity patterns to the untrained neurons in the balanced
regime, we considered a training setup where all the excitatory
neurons were trained, while the inhibitory neurons were not. We chose
the training targets to be 2Hz sine functions with random phases.
After training, synaptic inputs to the trained excitatory neurons
followed the target activity patterns (Fig. 5A, top). As a result, the first
two PCs of the trained activities were 2Hz sine and cosine functions
and were the dominant PCs of the trained activities (Fig. 5A, bottom).
To study how the trained activity spread in the network, we next
examined the synaptic inputs to an untrained inhibitory neuron. All
untrained neurons received only static synapses, but no plastic
synapses, from randomly selected trained and untrained presynaptic
neurons. Due to the large number of static synapses and their strong
weights, the mean excitatory (Fig. 5C, uE
t ) and inhibitory (Fig. 5D, uI
t)
inputs to the untrained neuron were much larger, in absolute value,
than the spike-threshold. In addition, the excitatory (Fig. 5C, δutrained
)
and inhibitory (Fig. 5D, δuuntrained
) temporal modulations around their
respective mean inputs developed sizable patterns, which were, how-
ever, significantly smaller than the mean inputs. Since the network
operated in the balanced regime, the large mean excitatory and inhi-
bitory inputs dynamically canceled each other, resulting in the net
mean input to the untrained neuron being around the spike-threshold
(Fig. 5E, ut). Consequently, the spiking pattern of the untrained neuron
was determined by the temporal modulations around the net mean
input (i.e., δutrained
).
and δuuntrained
t
t
t
t
t
We further examined the synaptic modulations driven by the
trained excitatory and untrained inhibitory presynaptic neurons.
Analysis of synaptic modulation driven by the trained excitatory neu-
rons (Fig. 5C, δutrained
) showed that it was dominated by the same PCs
the excitatory neurons were trained to generate (Fig. 5A). This trained
synaptic modulation then led the total input to the untrained neuron
to be modulated as well (Fig. 5E, ut). As a result, the untrained neurons
produced modulated activity (Fig. 5B), which then provided modu-
lated inputs to other neurons in the network (Fig. 5D, δuuntrained
). The
modulated synaptic drive that the untrained neurons received (Fig. 5E)
and provided to other neurons (Fig. 5D, right) both showed strong
temporal modulations compatible with the PCs acquired from
training.
t
One of the predictions of this spreading mechanism is that each
PC loading of the synaptic inputs to untrained neurons follows a
Gaussian distribution. This results from the task-independent synapses
that randomly sample the presynaptic activity patterns (and their PC
loadings) and summing them to generate the synaptic inputs (and
their PC loadings) to the untrained neurons. Then,
if the task-
independent synapses have strong weights, the sum of the randomly
sampled PC loadings (i.e., the PC loading of the synaptic inputs to
untrained neurons) converges to a Gaussian distribution with a finite
variance (see Methods for details). Indeed, this was the case for the
loadings of the first two PCs in the network trained on sine functions
(Fig. 5G). Then we analyzed the loadings of the dominant PC mode in
the ALM data, which were the slopes of the ramping activity of the
synaptic inputs. Since the synaptic inputs to ALM neurons were not
available, we estimated them by finding inputs to the transfer function
of the model neuron that yielded the observed firing rates of ALM
neurons. We found that the statistics of these loadings were also well-
fitted by a Gaussian distribution, supporting the biological plausibility
of the proposed mechanism (Fig. 5H).
The same network mechanism also provides an explanation for
how functional properties, such as choice selectivity, can spread from
neurons trained to be choice-selective to other neurons that are not
trained (Fig. 3E). It stems from the fact that the differences in the
activity of the lick-left and lick-right trials in the trained neurons spread
through the random static synapses and are realized into two different
responses in the untrained neurons, thus producing choice selectivity
in them (see Methods for details). In addition, our mathematical ana-
lysis of the network mechanism is consistent with the findings that, due
to their strong temporal modulations, inhibitory activity patterns
Nature Communications |
(2023) 14:2851
8
Article
https://doi.org/10.1038/s41467-023-38529-y
Fig. 5 | Network mechanism for distributing trained neural activity to
untrained neurons through strong, non-specific connections. A Excitatory
neurons were trained to generate 2Hz sinusoidal synaptic activity patterns with
random phases. Examples (top) of trained synaptic inputs (black) to the excitatory
neurons and their moving averages over 200ms window (magenta). The absolute
value of the loading of each PC on trained synaptic activities (bottom) was averaged
over all the excitatory neurons to obtain the average strength of the PCs. The first
two PCs, which are the Fourier modes of 2Hz sine waves, are highlighted (magenta)
and shown in the inset (PC1, PC2). fi’s in the circles (right) represent the spiking
activity of trained neurons. Arrows (green) to an untrained neuron represent ran-
dom, static, excitatory synapses with the synaptic weight JE. B Inhibitory neurons in
the network were not trained. Examples of untrained synaptic inputs to inhibitory
neurons (left). ri’s in the circles (right) represent the spiking activity of untrained
neurons. Arrows (orange) to an untrained neuron represent random, static, inhi-
bitory synapses with the synaptic weight JI. C Aggregate synaptic input (in arbitrary
units) from trained excitatory neurons to an untrained inhibitory neuron (uE
t ) and
its temporal modulation (δutrained
) around the mean activity. The PCs (right) show
the average strength of each PC in δutrained
. The PCs corresponding to the trained
t
activity in panel (A) are highlighted (magenta). D Same as in (C) but for the
t
t and uI
aggregate synaptic input from untrained inhibitory neurons in the network to the
same untrained inhibitory neuron shown in (C). E The total synaptic input (ut or the
sum of uE
t , black) to the untrained inhibitory neuron with the moving average
(magenta). More examples are shown in panel (B). The PCs (bottom) show the
strength of each PC in ut, averaged over all the untrained inhibitory neurons. The
PCs corresponding to the trained activity in panel (A) are highlighted (magenta)
and shown in the inset (PC1, PC2). F Schematic of synaptic inputs shown in panels
(A) to (E). Total synaptic input to trained excitatory neurons (A: black arrow) is the
sum of inputs from excitatory and inhibitory neurons (gray arrows). Total synaptic
input to untrained inhibitory neurons (B,E: black arrow) is the sum of inputs from
excitatory (C: gray arrow) and inhibitory neurons (D: gray arrow). G Distributions
(magenta) of PC1 (k1) and PC2 (k2) loadings on the total synaptic input to the
untrained inhibitory neurons (i.e., ut in panels (B) and (E)), overlaid with the
Gaussian fits (black). The PCs are shown in panel (E), bottom. H Distribution (blue)
of PC1 (k1) loadings on the estimated synaptic inputs to ALM pyramidal neurons for
the lick-right trial type, overlaid with the Gaussian fit (black). The transfer function
of the model neuron was used to estimate the synaptic input that yielded ALM
neuron’s firing rate (see Methods). PC1ramp
Fig. 2E. Source data are provided as a Source Data file.
was a ramping mode, similarly to PC1 in
t
spread more effectively than the excitatory activity patterns (Fig. 4A),
and leading PC modes transfer with better fidelity than the other PC
modes (Fig. 4D; see Methods).
The results of our analysis show that trained activity can spread in
the network to untrained neurons as long as the untrained static
synapses are strong, and the network operates in the balanced regime.
As this regime is not sensitive to the number of presynaptic inputs per
neuron, or the network size, this circuit mechanism for distributing
activity in neural networks is robust to variations in these parameters.
Moreover, the slopes of ramping activity in the ALM neurons displayed
Nature Communications |
(2023) 14:2851
9
Article
https://doi.org/10.1038/s41467-023-38529-y
statistics that agreed with the model prediction, providing an evidence
for the biological plausibility of the proposed mechanism.
Perturbation responses suggest that the ALM network is
balanced
We showed that when a subset of neurons was trained to reproduce
the ALM activity, the task-related activity spread to the untrained
neurons, which then also generated spiking activity resembling the
ALM neural data (Figs. 2, 3). Such spreading of activity from trained to
untrained neurons is a general mechanism at play in strongly coupled
spiking networks (Figs. 4, 5). These findings raised the possibility that
the ALM network operated in the same dynamical regime as the
strongly coupled network model when the observed ALM activity was
generated. To test this prediction, we investigated if optogenetic
perturbations to the ALM activity displayed the characteristics of the
balanced regime.
Specifically, we considered the activity modes of population of
neurons responding to perturbations applied during the delay period
(Fig. 6A,B). In strongly coupled networks consisting of excitatory and
inhibitory populations, the projection of the population activity on the
homogeneous mode (i.e. the average firing rate of the excitatory or
inhibitory populations, Fig. 6A) is expected to recover rather fast from
any perturbation. This is because the network dynamics are highly
stable along the homogeneous mode23. To understand this
Nature Communications |
(2023) 14:2851
10
Article
https://doi.org/10.1038/s41467-023-38529-y
Fig. 6 | Fast and slow responses of the network to perturbations (model and
data). A Schematic of the homogeneous mode, which averages the activity of the
neurons. B Schematic of trial-averaged activity for lick-left (red) and lick-right
(blue) trial types together with the choice mode in the neural activity space. This
mode maximally separates trial-averaged activity with respect to licking directions
(See Methods). C Schematic of a trained network receiving perturbation. D Change
in projection on choice mode (blue) and homogeneous mode (black) against time,
averaged over all 10 sessions. Each session consisted of sampling 50 neurons from
the network. The change in projection was calculated as the trial-averaged activity
for perturbed trials minus unperturbed trials (see Methods), with a 50ms
smoothing. Mean ± SEM. Shaded red: time of applied perturbation for perturbed
trials. Dashed lines: exponential fit. E Projection of neural activity on homogeneous
(left) and choice (right) modes for an example session, normalized by subtracting
the average projection over the first 0.5 second of the delay period. Data are pre-
sented as mean values ± SD over trials (shaded area). Orange: significant differences
between perturbed and unperturbed trials, starting from the perturbation time (see
Methods). Dashed red: recovery time of perturbation, estimated as the first time
the change was not significant following the perturbation (see Methods).
F Recovery time for all sessions. Recovery of the homogeneous mode was sig-
nificantly faster (p-value, by One-sampled paired Student t-test). Error bars pre-
senting mean value ± SEM. G Adapted from Finkelstein, A., Fontolan, L., Economo,
M.N. et al. Attractor dynamics gate cortical information flow during decision-
making. Nat Neurosci 24, 843-850 (2021). https://doi.org/10.1038/s41593-021-
00840-6. Schematic of optogenetic perturbation in the mouse cortex. H–J Same as
(D–F), but for putative excitatory neurons in ALM. Here each session corresponds
to simultaneous recordings of ALM neurons on different days. Optogenetic per-
turbation in the data was applied to somatosensory cortex21, whereas in the net-
work model the stimulus that triggered the lick-left response was used to perturb
the lick-right trials. Source data are provided as a Source Data file.
phenomenon, one should consider changes to the average firing rate
of the excitatory population in the network. This will result in a strong
change (on the order of square root the number of inputs per neuron)
to the excitatory drive to each of the neurons, which unless immedi-
ately suppressed by a strong inhibitory current, will destabilize the
network. Therefore, to maintain the stability of network dynamics, in
the balanced regime a perturbation to the homogeneous mode is
expected to decay quickly to its pre-perturbed value due to the strong
and fast inhibition (a phenomenon known as ‘fast tracking’23, 44).
Consistent with this known phenomenon, following a perturba-
tion to the activity of neurons in the strongly coupled network in Fig. 2,
the projection on the homogeneous mode quickly returned to the
baseline (Fig. 6D, black). In contrast, the projection of the activity on
the choice mode (Fig. 6B), a mode that maximally separates trial-
averaged activity with respect to licking directions (see21, 41 and Meth-
ods), returned to the baseline after the perturbation with a significantly
longer recovery time (Fig. 6D, blue; Fig. 6F, paired Student t-test,
p-value = 0.016). The slow recovery suggested that a dynamic attrac-
tor, which formed around the target trajectory due to training, was
able to retract the perturbed activity at a slow timescale along the
coding mode21,45. Importantly, the network was trained only on the
unperturbed ALM activity. Therefore, the fast and slow responses to
perturbations were not dynamical properties acquired directly from
the perturbed ALM activity, but instead they emerged from the
strongly coupled network, when it was trained just on the unperturbed
ALM activity.
To test the model prediction regarding the fast recovery of the
homogeneous mode, we conducted the same analysis on single ses-
sions of simultaneously recorded ALM neurons (Fig. 6G-J). We found
that the response time of the homogeneous mode in ALM was sig-
nificantly faster than that of the choice mode (Fig. 6H-J, paired Student
t-test, p-value = 0.025). Thus, the fast recovery of the homogeneous
mode of ALM network, relative to the slow recovery of the choice
mode, to optogenetic perturbations suggested that the ALM network
operated in the same dynamical regime as the strongly coupled
network.
These findings suggest that the ALM network has the potential to
be endowed with a network-level mechanisms for generating wide-
spread task-related activity, with limited synaptic reorganization on
only a subset of neurons during learning.
Discussion
In this study, we presented a potential circuit mechanism for dis-
tributing task-related activity in cortical networks. We have shown that
neural activity learned by a subset of neurons can spread to the
untrained parts of the network through pre-existing random con-
nectivity, without additional training. This spread of activity occurs as
long as the pre-existing random connections are strong and create a
dynamical state, known as the balanced regime. When a subset of
neurons in the spiking network was trained to reproduce the activity of
ALM neurons, the activity of untrained neurons in the network also
displayed surprising similarity to the activity patterns of neurons in
ALM. Single neuron activity patterns of the untrained neurons were
ramping and selective to future choices, as was observed in ALM. Our
work suggests that only a subset of neurons may be actively engaged in
learning and the rest of the neurons are driven by the structured
activity generated from the trained neurons.
Accumulating evidence shows that inhibition in cortex is highly
plastic (e.g. see review by46). We found that the fidelity of spreading the
activity was higher when the inhibitory neurons were trained instead of
the excitatory ones. For example, all of the excitatory neurons needed
to be trained to explain 70% of the variance in the untrained inhibitory
neurons, while training only 60% of the inhibitory neurons was enough
to induce the same 70% variance in the untrained excitatory neurons
(Fig. 4A). We speculate that this is a characteristic of the operating
regime of cortical networks, in which typically the baseline spiking
rates of inhibitory neurons is higher than the excitatory neurons.
Inhibitory neurons can thus support stronger rate modulations
(Fig. 4B), which in turn improves the fidelity of the spread (Fig. 4A,
Fig. 5, Methods). Our results suggest that synaptic plasticity in inhibi-
tory neurons can lead to wider spread of task-related activity in the
motor cortex. Interestingly, this result is consistent with recent theo-
retical and computational studies showing that patterns of neural
activity are primarily determined by inhibitory connectivity47,48.
In recent studies, the authors of43,49 argued that specific con-
nectivity between excitatory and inhibitory neurons is necessary for
choice selectivity to emerge in these two populations, based on
computational models of their data. Our work suggests an alternative
mechanism in which choice selectivity emerges in one population
during training and spreads to the other population, without any
reorganization of specific connections from the trained to the
untrained populations. The network mechanism that spreads the task-
related activity through random connectivity, as in our trained net-
works, is based on the susceptibility of neurons to modulations of
synaptic inputs in strongly coupled networks (Fig. 5). In brief, the
strong static synapses preserve the temporal variations in the pre-
synaptic activity. It thus results in choice-selective inputs that are on
the order of the spike-threshold. The recurrent inhibition then cancels
the strong mean excitatory input, leading to a total excitatory and
inhibitory inputs that are both on the order of the spike-threshold and
choice-selective (see Prediction 4 in Methods). This is a similar
mechanism that explains how, without training or functional structure,
orientation-selective neurons can emerge in the primary visual cortex
with a ‘salt-and-pepper’ organization50,51.
Overall, the good agreement between the activity of untrained
neurons in the model and the neurons in the data that were held-out of
training resulted from the similarities between the activity of fast-
spiking and pyramidal neurons in the data. It will be interesting to look
Nature Communications |
(2023) 14:2851
11
Article
https://doi.org/10.1038/s41467-023-38529-y
at other data sets in which task-related activity is more diverse between
different cell types, and explore possible network mechanisms that can
spread task-related activity which differ between the trained and
untrained populations. In addition, we point out that one important
property of ALM neurons that make them compatible with the balanced
network is that, on average, their ramping slopes are close to zero
(Fig. 5H), consistent with the fact that the mean rate of ALM neurons is
almost constant during the delay period. This kept the overall rate of the
trained subset constant in time, therefore the trained network did not
deviate significantly from the balanced regime. For neural data with
highly fluctuating average population rates, other network models or
additional network mechanisms may need to be considered to account
for strong changes in population rates that could potentially break the
balance in a subset of the neurons.
Our trained network model showed that the proposed circuit
mechanism (i.e., subset training) for distributing task-related activity
to untrained neurons can explain various aspects of neural data.
However, it still remains an open question whether only a subset of
neurons in the real cortical circuit undergoes synaptic reorganization
when learning. Several recent experimental studies show that synaptic
plasticity and induced neural activity in a subset of neurons can
influence learned behavior and broad network responses. It was shown
that labeling recently potentiated spines in a subset of neurons, cre-
ated through motor learning, and disrupting them by optical shrinkage
were sufficient to reduce acquired motor skills13. In addition, optoge-
netic stimulation studies targeting a small number of specific cells
show that learning a new motor task or producing realistic network
response can be induced from a small number of neurons20,52–54.
Although not conclusive, these studies support the biological feasi-
bility of subset training.
In this study, we computationally explored the amount of subset
training by varying (1) the number of trainable neurons and (2) the
number of plastic synapses to the trained neurons. Given that biolo-
gically plausible learning rules for synaptic reorganization are based
on activity of locally connected neurons55–57, subset training could
potentially be implemented to induce global learning effects across
the network by reorganizing synaptic connections of locally connected
sub-networks of neurons. In addition, when learning resources are
limited (e.g., limited number of trainable synapses), increasing the
number of trainable neurons may not necessarily lead to improved
performance. Instead, subset training could generate desired dynam-
ics in the trained subset without extensively modifying the synapses
across the population of neurons (see Fig. S11 for an example).
On the other hand, an alternative training scheme that could be
implemented in the brain is to train a larger number of neurons and
synapses throughout the network. Using this approach, recent studies
considered training spiking networks with dynamically balanced
excitation and inhibition.
In58 the authors had to break the EI balance in order to achieve
individual
non-linear computations. With our training procedure,
neurons can be trained to perform complex tasks, such as generating
the spiking activity of cortical neurons, without leaving the balanced
regime. The work by59,60 trained all the recurrent weights of the
dynamically balanced spiking networks. To maintain strong excitatory-
inhibitory activities after training, they considered weight regulariza-
tions that constrained the trained weights close to strong initial EI
weights. Instead, in our training setup, the strong initial EI connections
were left unchanged throughout training, thus always provided the
strong excitation and inhibition.
Other studies showed that a larger number of synapses across the
entire network can be trained successfully, as long as they are weaker
than the strong pre-existing random connections19,35,61. Specifically,
several recent studies showed that it is possible to train networks to
perform tasks by training weak presynaptic inputs, while constraining
their connectivity to be of low-rank62,63. In such networks the activity of
every neuron in the network is modulated by trained synapses, a setup
that does not allow one to study the role of untrained synapses in
spreading trained activity. This is different from our work, in which we
train only a subset of the neurons and investigate the role of untrained
synapses in spreading the trained activity to untrained neurons.
There is an ongoing debate if the cortex operates in the balanced
regime64. Experimental evidence that are inconsistent with the
balanced regime hypothesis mainly relies on data from sensory cor-
tices. Here we present evidence that the motor cortex operates in the
balanced regime by analyzing the recovery of neurons in the motor
cortex to optogenetic perturbations. The presence of two recovery
time scales, i.e., fast for the homogeneous mode and slow for the
choice mode, is consistent with the prediction of the balance regime
that the homogeneous mode rapidly tracks inputs, a phenomenon
termed ‘fast tracking’23,44. Our analysis is different from the paradoxical
effect observed in excitatory-inhibitory networks, where strong
recurrent excitation must be compensated by strong feedback inhi-
bition to maintain a stable network state65–68.
We note that, due to the sparse plastic synapses in the trained
networks, the plastic input was moderately strong (i.e., on the order of
the spike-threshold). The network could thus implement non-linear
computations at the individual neuron level. It can also support non-
linear computations at the population level, as long as the computa-
tion is held by subpopulations, such that the overall excitatory and
inhibitory population rates are unchanged after training (see Methods
and also69,70). Thus, the only mode that is strictly linear with the inputs
to the network is the homogeneous mode. This is different from recent
works that portrayed that the strict linear input-output relationship of
balanced networks limits their computational power58,64.
More broadly, our work shows that the same theory that accounts
for the irregular nature of spiking activity of single neurons in the
cortex can also explain a seemingly unrelated phenomenon, which is
why task-related activity is so distributed in the cortex. Distributing
task-related activity can be beneficial for several reasons, such as
robustness to loss of neurons or synapses or an increase in coding
capabilities71. Future research directions could focus on the compu-
tational benefits of cortical networks operating in the balanced regime
in the lens of distributing task-related activity.
To conclude, our work shows that while large changes in network
dynamics can be observed during learning, attributing such changes to
synaptic reorganization between neurons must be taken with care. In
strongly coupled networks that operate in the balanced regime, in
which the motor cortex might operate, widespread changes in neu-
ronal activity can be mainly a result of distributing learned activity
from a dedicated subset of neurons to the rest of the network through
task-independent strong synapses.
Methods
Data acquisition was performed using SpikeGL (https://github.com/
(https://www.janelia.org/open-
cculianu/SpikeGL) and Wavesurfer
science/wavesurfer) software (see21).
Data analysis
Principal component analysis of population rate dynamics. To
obtain the PSTHs of neurons in a trained network, we repeated the
simulation of a trained network 400 times starting at random initial
conditions and applied the same external stimulus to trigger the
trained activity patterns. Subsequently, for each neuron, the spikes
emitted over multiple trials were placed in 20ms time bins, which
ranged over the Ttarget long training window, and averaged across trials
to compute the instantaneous spike rates.
Given the PSTHs, r1, . . . ,rM 2 RT , of a population of M neurons,
we subtracted the mean rate of every neuron from its PSTH to remove
differences in the baseline firing rates. In the following, we use the
same notation ri to refer to the mean subtracted PSTH of neuron i.
Nature Communications |
(2023) 14:2851
12
Article
https://doi.org/10.1038/s41467-023-38529-y
We then performed principal component analysis on the popu-
lation rate dynamics R = (r1, …, rM), which is a T × M matrix, to obtain
the principal components v1, . . . ,vT 2 RT and the principal values
λ1, …, λT. This is equivalent to finding the eigenvectors and eigenvalues
of the covariance matrix, RR⊤. The variance explained by the kth prin-
cipal component was λ2
k
P
=
λ2
i .
i
The same procedure was applied to all the principal component
analyses performed in this study.
Shared variance analysis. We identified population vectors along
which the population activities of inhibitory model neurons and fast-
spiking ALM neurons co-varied maximally40. We also quantified the
fraction of variance that can be explained by the projected population-
averaged activities (Fig. 3).
We first computed the correlation Cij = corr(fi, gj), which an
M1 × M2 matrix, between the PSTH’s of inhibitory model neurons f i 2
RT ,1 ≤ i ≤ M1 and fast-spiking ALM neurons gj 2 RT ,1 ≤ j ≤ M2 where
M1 = 2500, M2 = 306 and T = 100. Then the singular-value decomposi-
tion C = UΣV of the correlation matrix was performed, where U is an
M1 × M1 matrix and V is an M2 × M2 matrix, to obtain the left singular
vectors U = ðu1, . . . ,uM1
Þ with uk 2 RM1 and the right singular vectors
V = ðv1, . . . ,vM2
Þ with vk 2 RM2 .
To obtain the population-averaged activity along the singular
Þ 2 RT × M1
vectors, the matrices of population rate, i.e., F = ðf 1, . . . ,f M1
for the inhibitory model neurons and G = ðg1, . . . ,gM2
Þ 2 RT × M2 for the
fast-spiking ALM neurons, were projected to the corresponding kth
singular vectors uk and vk, respectively, to obtain the kth shared com-
k = Gvk 2 RT . The variance explained by
ponents, α
the kth shared component was defined as ∥αk∥2/∑k∥αk∥2 and ∥βk∥2/
∑k∥βk∥2, respectively.
k = Fuk 2 RT and β
Defining the choice and homogeneous modes. Trial-averaged spike
rate of a neuron i, ri(t, k), were calculated for each trial, k, using 1ms bin
size and were filtered with a 200ms boxcar filter.
We then analyzed the population dynamics of N simultaneously
recorded neurons in a session. During each trial, the population
activity of these neurons, r(t, k), drew a trajectory in the N-dimensional
activity space. We identified the choice mode as N × 1 vector of trial-
averaged spike rate differences of N neurons during trials with lick-
right and lick-left outcomes, averaged within a 1sec window at the end
of the delay epoch, before the go cue21:
C =
p
ffiffiffiffi
N
1
k hrRit,k (cid:2) hrLit,k k
(cid:3)
hrRit,k (cid:2) hrLit,k
(cid:4)
ð1Þ
p
ffiffiffiffi
N
with the L2 norm, ∥x∥, and 〈x〉t,k which is averaging over trials and
time. The
term was introduced to ensure that the projection of the
neural activity is independent of the number of recorded neurons and
for consistency with the homogeneous mode below. Projections of the
neural activity along the choice mode were:
PCðt,kÞ = C (cid:3) rðt,kÞ
ð2Þ
with the SD, after subtracting the average projection over the first
0.5 seconds of the delay period.
We used a statistical hypothesis test (Student t-test) to estimate
the decay time back to the non-perturbed trajectories for the projec-
tions on the modes. Specifically, for each time bin we tested the null
hypothesis that the perturbed and unperturbed trials were from
the same distribution and rejected the null hypothesis with a
p-value < 0.05 (orange dots in Fig. 6E,I). We only analyzed sessions in
which the photostimulation resulted in a significant change in at least
10% of the time points during the photostimulation period ([ − 1.6, 1.2]
sec, 13/17 sessions). To calculate the decay time, we then used the last
significant time bin within the time window of [ − 1.2, 0]sec for which
the derivative was smaller than 10ms (dashed red lines in Fig. 6E,I). The
perturbations in 2/13 sessions were biased and were not included in the
analysis, leaving 11 sessions of simultaneously recorded neurons.
To calculate the decay time over all sessions (Fig. 6D) we averaged
the projection in each of the 11 analyzed sessions and calculated the
difference in the projection between the perturbed and unperturbed
trials (Δ projection). We then took the absolute value and averaged
over all sessions (Fig. 6D, mean ± SEM). Finally, we estimated the decay
rate by an exponential fit.
We note that in the experiments these estimates should be thought
of as an upper bound for the real decay timescale due to multiple
reasons. First, different sessions in the data might originate from
recordings in different mice. Second, even within the same mouse there
might be differences in the dynamical state of the network, which will
affect the firing rate and its decay back to baseline. Third, in contrast to
the model, it is hard to control the optogenetic perturbation in the
experiment. Indeed, our ability to verify that we activated exactly
the same group of neurons in vS1 during the perturbation, and with the
same amplitude, is limited (see also the paragraph above).
Spiking neural networks
Network connectivity. The spiking neural network consisted of ran-
domly connected NE excitatory and NI inhibitory neurons. The recur-
rent synapses consisted of static weights J that remained constant
throughout training and plastic weights W that were modified by the
training algorithm. The static synapses connected neuron j in popu-
lation β to neuron i in population α with probability pαβ = Kαβ/Nβ and
, where Kαβ is the average number of static
(cid:2)
Jαβ=
synaptic weight
connections from population β to α:
q
ffiffiffiffiffiffiffiffi
K αβ
PrðJ
αβ
ij
≠0Þ =
K αβ
Nβ
:
ð3Þ
The strength of plastic synapses,
, was of the same
order as the static weights. However, the plastic synapses connected
neurons with a smaller probability:
(cid:2)
W αβ=
q
ffiffiffiffiffiffiffiffi
K αβ
PrðW
αβ
ij
≠0Þ =
Lαβ
Nβ
with Lαβ = c
q
ffiffiffiffiffiffiffiffi
K αβ
ð4Þ
Similarly, the projection over the homogeneous mode was given by
PHðt,kÞ = 1
N 1 (cid:3) rðt,kÞ, with 1 being a vector of ones.
If an individual neuron was not recorded during a particular trial, its
weight in equation (2) was set to zero, and for the analysis we selected
trials with at least 10 simultaneously recorded neurons.
which made the plastic synapses much sparser than the static
synapses70. Here, c is an order 1 parameter that depends on
training setup.
The static and plastic connections were non-overlapping in that
any two neurons in the network can have only one type of synapse.
Response of the modes to perturbations. To assess the impact of vS1
photostimulation during the delay on the homogeneous and choice
modes in the ALM, we computed for each session the single-trial
projections on each of the modes, PC(t, k) and PH(t, k), for correct lick-
right trials both with and without the photostimulation. The trial-
averaged activity was plotted for one example session in Fig. 6E,I along
αβ
ij W
αβ
ij = 0:
J
ð5Þ
Keeping them disjoint allowed us to maintain the initial network
dynamics generated by the static synapses and, subsequently, intro-
duce trained activity to the initial dynamics by modifying the plastic
synapses.
Nature Communications |
(2023) 14:2851
13
Article
https://doi.org/10.1038/s41467-023-38529-y
Table 1 | Default simulation and training parameters
Neuron parameters
simulation time step
membrane time constant
spike threshold
δt
τm
vthr
vreset
voltage reset after spike
Network parameters
number of neurons
number of excitatory neurons
number of inhibitory neurons
connection probability
Synaptic parameters
static synaptic time constant
plastic synaptic time constant
average number of static synapses to a neuron
average number of excitatory static synapses to a neuron
average number of inhibitory static synapses to a neuron
number of plastic synapses to a neuron
excitatory synaptic weight
inhibitory synaptic weight
external input
N
NE
NI
p
τbal
τplas
K
KE
KI
L
JE
JI
X
JEE
E to E static synaptic weight
JIE
JEI
JII
XE
XI
γE
γI
γX
λ
μ
E to I static synaptic weight
I to E static synaptic weight
I to I static synaptic weight
external input to excitatory neurons
external input to inhibitory neurons
relative strength of WEE to WIE
relative strength of WEI to WII
relative strength of XE to XI
Training parameters
penalty for L2-regularization
penalty for ROWSUM-regularization
Niter
Ttarget
number of training iterations
length of target patterns
Any differences from the above parameters are described in Table 2.
Values
0.1 ms
10 ms
1
0
30000
N/2
N/2
0.2
3 ms
150 ms
pN
pNE
pNI
ffiffiffiffiffiffi
KE
ffiffiffiffiffi
p
KI
ffiffiffiffiffi
KI
see Table 2
p
2:0=
(cid:2)2:0=
p
0:08
γEJE
JE
γIJI
JI
γXX
X
0.15
0.75
1.5
0.05
8.0
200
2 sec
The static recurrent synapses were strong in that the coupling
p , while the
strength between two connected neurons scaled as
1ffiffiffiffiffiffi
K αβ
average number of synaptic inputs scaled as Kαβ. This is in contrast to
the weak, 1/Kαβ, coupling we considered in Fig. S9. As a result of this
strong scaling, the excitatory (uE
bal) synaptic
, thus were
bal) and inhibitory (uI
ffiffiffiffiffiffiffiffi
K αβ
inputs to a neuron from static synapses increased as
q
much larger than the spike-threshold for a large Kαβ. However, uE
bal and
uI
bal were dynamically canceled, and the sum (ubal) was balanced to be
around the spike-threshold (ref. 23, Fig. 1B, middle).
p
q
ffiffiffiffiffiffiffiffi
K αβ
plas, uI
In contrast to the static synapses, each trained neuron received
plastic synapses. This made the plastic synapses
only about
much sparser than the sparse static EI connectivity (e.g., with
ffiffiffiffi
≈30 plastic
K = 1000 static synapses, there are of the order of
K
synapses per neuron). Consequently,
the EI plastic inputs
(uE
plas) of the initial network were independent of Kαβ and
substantially weaker than the EI balanced inputs (uE
bal) for a large
Kαβ. After training the plastic synapses, the total synaptic input
(u = ubal + uplas) to each trained neuron was able to follow the target
patterns (Fig. 1B, left; Fig. 1C), while the plastic input (uplas) stayed
around the spike-threshold (Fig. 1B, right). With this scaling of
plastic synapses, training was robust to variations in the number of
bal,uI
Nature Communications |
(2023) 14:2851
synaptic connections, Kαβ. Network trainings were successful even
when Kαβ was increased, such that the excitatory and inhibitory
balanced inputs were tens of orders of magnitude larger than the
plastic inputs (Fig. S1). All network parameters used in the figures
can be found in Table 1.
Network dynamics. We used integrate-and-fire neuron to model the
membrane potential dynamics of the i’th neuron:
τ
m
_v
α
i = (cid:2) vα
i + uα
i + X
α
i
ð6Þ
where a spike is emitted and the membrane potential is reset to vreset
when the membrane potential crosses the spike-threshold vthr.
Here, uα
i is the total synaptic input to neuron i in population α that
can be divided into static and plastic inputs incoming through the
static and plastic synapses, respectively:
i = uα
uα
bal,i + uα
plas,i
:
ð7Þ
α
i is the total external input that can be divided into constant external
X
input, plastic external input, and the stimulus:
X
α
i = X
α
bal,i + X
α
plas,i + X
α
stim,i
:
ð8Þ
q
α
bal,i is a constant input associated with the initial balanced network. It
X
ffiffiffiffiffiffiffiffi
scales with the number of connections, i.e., proportional to
,
K αβ
determines the firing rate of the initial network and stays unchanged23.
α
plas,i is plastic input provided to trained neurons in the recurrent
X
network from external neurons that emit stochastic spikes with pre-
determined rate patterns. The synaptic weights from the external
neurons to the trained neurons were modified by the training
α
stim,i is the pre-determined stimulus, generated indepen-
algorithm. X
dently from the Ornstein-Uhlenbeck process for each neuron, and
injected to all neurons in the network to trigger the learned responses
in the trained neurons (see details in Network training scheme below).
The synaptic activity was modeled by instantaneous jump of the
synaptic input due to presynaptic neuron’s spike, followed by expo-
nential decay. Since the static and plastic synapses did not overlap, we
separated the total synaptic input into static and plastic components
as mentioned above:
τ
bal
_u
α
bal,i
=
(cid:2)uα
bal,i +
τ
plas
_u
α
plas,i
= (cid:2)uα
plas,i +
P
P
P
αβ
ij
J
δðt (cid:2) tj
kÞ
β2fE,Ig
P
j2β
P
β2fE,Ig
j2β
W
αβ
ij
tj
k
<t
P
tj
k
<t
δðt (cid:2) tj
kÞ:
ð9Þ
with τbal synaptic integration time constant of the static inputs and τplas
the synaptic integration time constant of the plastic inputs. Alter-
natively, the synaptic activity can be expressed as a weighted sum of
filtered spike trains because the synaptic variable equations (equation
(9)) are linear in J and W:
uα
bal,i
uα
plas,i
=
=
P
β,j
P
β,j
αβ
ij r
β
bal,j
J
W
αβ
ij r
β
plas,j
where
τ
bal
_r
β
bal,i
= (cid:2)r
β
bal,i +
τ
plas
_r
β
plas,i
= (cid:2)r
β
plas,i +
P
ti
k <t
P
ti
k <t
δðt (cid:2) ti
kÞ
δðt (cid:2) ti
kÞ
describe the dynamics of synaptically filtered spike trains.
ð10Þ
ð11Þ
14
Article
https://doi.org/10.1038/s41467-023-38529-y
Each external neuron emitted spikes stochastically at a pre-
defined rate that changed over time. The rate patterns, followed by the
external neurons, were randomly generated from an Ornstein-
Ulenbeck process with mean rate of 5 Hz. The synaptically filtered
external spikes were weighted by plastic synapses WX and injected to
trained neurons:
X
α
plas,i =
X
j
W X
ij rX
j
where
τ
plas
_rX
plas,i = (cid:2) rX
plas,i +
X
ti
k <t
δðt (cid:2) ti
kÞ
ð12Þ
ð13Þ
Similarly, the external stimulus Xstim,i applied to each neuron i in the
network to trigger the learned response is generated independently
from the Ornstein-Ulenbeck process.
In the following section, we will use the linearity of W, WX in
equations (10) and (12) to derive the training algorithm that modifies
plastic synaptic weights.
Training recurrent neural networks in the balanced regime using
sparse plastic synapses. From a technical point of view, the choice to
train very sparse plastic synapses made the plastic inputs to be on the
order of the spike-threshold (i.e. order one, independent of the num-
ber of connections). This choice of training only a sparse number of
plastic weights enables training the network without affecting the
mean firing rates of the excitatory and inhibitory populations. It thus
allows the network to generate non-linear dynamics in a macroscopic
number of neurons.
To show this, we write the mean input of each excitatory and
inhibitory neuron in the absence of the transient external stimulus,
X
α
stim,i:
huα
i i + X
n
i = huα
α
bal,ii + X
o
α
bal,i
+ huα
plas,ii + X
α
plas,i
ð14Þ
The terms in the curly brackets in the right hand side of the above
equation are:
n
huα
bal,ii + X
o
=
α
bal,i
(
X
j
αE
ij rE
j +
J
X
j
)
αI
ij rI
j + X
α
bal,i
J
ð15Þ
They consist of a large number of uncorrelated random variables, and
thus in the limit of a large number of presynaptic inputs they converge
to a Gaussian distribution with mean:
p
ffiffiffiffi
JαE rE (cid:2) (cid:2)
½(cid:2)
K
JαI rI +
μα =
(cid:2)
X bal,α(cid:4)
ð16Þ
and an order one variance that needs to be calculated self-consistently.
At the balanced regime, the mean input of the excitatory and inhibitory
populations is Oð1Þ as long as:
JEE rE (cid:2) (cid:2)
(cid:2)
JIE rE (cid:2) (cid:2)
(cid:2)
JEI rI +
JII rI +
(cid:2)
X bal,E
(cid:2)
X bal,I
≈0
≈0
ð17Þ
These two linear equations, termed the ‘balanced equations’44, 72 only
(cid:2)
Jαβ, and the external
involves the strength of the static connections,
(cid:2)
X bal,α. The mean firing rates of the excitatory and inhibitory
inputs,
populations are thus linear in the external inputs, and are independent
of the plastic synapses.
We constructed the plastic synapses in a way that each neuron
synapses, with an average strength of
receives only an order of
Oð1=
Þ. This average strength is kept throughout the training thanks
to the ROWSUM regularization (see below). The plastic inputs,
ffiffiffiffi
K
ffiffiffiffi
K
p
p
ffiffiffiffi
K
plas,ii + X
α
huα
plas,i are thus on the order of the threshold. As they are
smaller than the mean excitatory and inhibitory balanced inputs by a
p
factor of 1=
, they do not enter into the balanced equations, and
cannot affect the mean rates of the excitatory and inhibitory popula-
tions. Yet, they can be trained to drive the neurons to generate non-
trivial dynamics, and lead to non-linear dynamics in sub-populations of
neurons, while keeping the population rates linear in the average
external inputs,
(cid:2)
X bal,α
Alternatively, if the plastic synapses were more abundant in the
network, e.g., on the order of the number of static connections, they
could interfere with the the ability of the strong inhibition to balance
the strong excitation for each neuron in the network. Such inter-
ference significantly limits the ability to train the spiking networks.
Training only a sparse number of plastic connections, on the order of
p
, thus allows to train the networks to perform non-linear compu-
ffiffiffiffi
K
tations, while keeping it in the balanced regime.
Network training scheme
Overview. Prior to training the network, neurons were connected by
the recurrent static synapses and emitted spikes asynchronously at
constant rates. This asynchronous state of the initial network has been
investigated extensively in previous studies23,25,72.
Starting from this asynchronous state, the goal of training was to
produce structured spiking rate patterns in a subset of neurons selected
from the network. Specifically, our training scheme modified the
recurrent and external plastic synapses projecting to the selected
neurons, so that they generated target activity patterns when evoked by
a brief external stimulus. To this end, we first selected M neurons to be
trained from a network consisting of N neurons, and then prepared M
target functions f1(t), …, fM(t) defined on a time interval t ∈ [0, Ttarget] to
be learned by the selected neurons. The plastic synapses projecting to
each selected neuron i were then modified by the training algorithm
such that the total synaptic input ui(t) to neuron i followed the target
pattern fi(t) on the time interval t ∈ [0, Ttarget] after the training.
Initialization of plastic synapses. For each trained neuron, we ran-
domly selected L excitatory and L inhibitory presynaptic neurons that
projected plastic synapses to the trained neuron. When the excitatory
subpopulation was trained, the presynaptic excitatory neurons were
sampled from other trained excitatory neurons while the presynaptic
inhibitory neurons were sampled from the entire inhibitory population.
Similarly, when the inhibitory subpopulation was trained, the pre-
synaptic inhibitory neurons were sampled from other trained inhibitory
neurons while the presynaptic excitatory neurons were sampled from
the entire excitatory population. The untrained neurons did not receive
any plastic synapses. Each trained neuron also received inputs from all
the LX external neurons. The plastic weights from the external neurons
to each trained neuron were trained by the learning algorithm.
While our training algorithm requires only an order of square root
of the pre-existing static connections to be plastic, the specific number
of plastic connections may vary with the complexity of the trained
neural activity. Indeed, almost three times more plastic synapses were
needed to train the neurons to reproduce ALM activity, in which six
PCs explained about 80% of the variance, compared to the number of
in which two PCs
plastic synapses needed to train sine waves,
explained the same amount of the variance (Table 2). It is beyond the
scope of this paper to determine exactly how the prefactor of the
square root term depends on the complexity of the neural activity.
However, previous studies suggest that the number of plastic con-
nections might depend on the dimensionality (i.e., decay rate of the
singular values)73 or decorrelation time of the trained neural activity37.
In addition to having sparse plastic synapses, we modeled their
dynamics using slower integration time constant with respect to the
abundant non-plastic synapses. The timescales of the non-plastic
synapses were on the order τbal = 3ms, consistent with timescales of
Nature Communications |
(2023) 14:2851
15
Article
https://doi.org/10.1038/s41467-023-38529-y
Table 2 | The number of total neurons, trained neurons and plastic synapses in the simulated networks
Target functions
Neurons
N
Ntrained
Static synapses to a neuron
# neurons
# trained neurons
p
K = pN
conn prob of static synapses
# static synapses to a neuron
Plastic synaptic weights to a trained neuron
JE, JI
WEE
WIE
WEI
WII
see Table 1
E to E plastic synaptic weight
E to I plastic synaptic weight
I to E plastic synaptic weight
I to I plastic synaptic weight
Number of plastic synapses to a trained neuron
p
ffiffiffiffi
K
p
ffiffiffiffi
K
ffiffiffiffi
p
K
Lrec = c
Lffwd = c
L = Lrec + Lffwd
order of # plastic synapses
# recurrent plastic synapses
# ffwd plastic synapses
# total plastic synapses
Figure 2
Neural PSTH
5 ⋅ 103
1824
0.2
1000
0.66JE
0.66JE
0.33JI
0.33JI
32
264
300
564
Figure 4
Synthetic PSTH
3 ⋅ 104
3 ⋅ 103 to 1.5 ⋅ 104
Figures 1 & 5
Sine function
3 ⋅ 104
3 ⋅ 104 & 1.5 ⋅ 104
0.2
6000
JE
JE
0.5JI
0.5JI
77
440
200
640
0.213
6400
2JE
2JE
JI
JI
80
226
0
226
Sparsity of plastic synapses
L/K
# plastic/# static synapses
0.564
0.106
0.035
synapses consisting of AMPA and GABA receptors. In contrast, the time
scale of the plastic synapse was significantly slower (τplas = 150ms). In a
previous work we showed that the time scale of the plastic synapses
should be faster than or on the order of the decorrelation time scale of
the target PSTHs. However, the slower τplas is, the sparser the plastic
weights can be37. In this sense, it is better to train networks with
synapses that has a slow ‘NMDA component’, adding another com-
putational advantage to synapses consisting of NMDA receptors in
learning processes.
Cost function. Each trained neuron i had its own private cost function
defined by
Ci½wrec
i
,wX
i (cid:4) =
Z
T target
0
1
2
ðf iðtÞ (cid:2) uiðtÞ (cid:2) X iðtÞÞ2dt +
1
2
Reg½wrec
i
,wX
i (cid:4)
ð18Þ
, . . . ,W iiL
i = ðW ii1
where wrec
Þ is a vector of recurrent plastic synapses to
neuron i from other presynaptic neurons in the network indexed by
i1, . . . ,W X
i1, …, iL. Similarly, wX
Þ is vector of plastic synapses to
iLX
neuron i from the external neurons. The regularization of plastic
weights Reg½wi,wX
i (cid:4) consisted of two terms
i = ðW X
Reg½wrec
i
,wX
i (cid:4) = λðk wrec
i k2 + k wX
i k2Þ + μ
X
α2fE,Ig
ðwrec
i
α
(cid:3) 1
i Þ2:
ð19Þ
α
The first term is a ridge regression that evaluates the L2-norm of the
plastic weights. It allowed us to uniquely solve for the plastic weights in
the training algorithm described below, and the hyperparameter λ
controls the learning rate, i.e., the size of synaptic weight updates. The
second term is called ROWSUM regularization where the elements of
α
α
L Þ are defined to be i
i = ði
the vector 1
k = 1 if the presynaptic
neuron ik belongs to population α and 0 otherwise60. The inner pro-
ducts wrec
i are the aggregate plastic weights to neuron i
from the excitatory and inhibitory populations, respectively. Including
the ROWSUM regularization allowed us to keep the aggregate excita-
tory and inhibitory plastic weights fixed throughout the training. When
the plastic input to a trained neuron is initialized to be around
spike-threshold, the ROWSUM regularization makes it possible to keep
i and wrec
1 , . . . ,i
(cid:3) 1E
(cid:3) 1I
α
i
i
Nature Communications |
(2023) 14:2851
the plastic input to be about the same magnitude in the trained net-
work. Although the ROWSUM regularization term could be further
developed, as studied in48, to impose Dale’s law in networks exhibiting
wide firing rate distributions, the trained plastic weights in our network
were allowed to flip signs, hence violate Dale’s law in the plastic
synapses but not in the initial EI network synapses (see Fig. S8A for the
distribution of trained plastic weights).
Training algorithm. We derived a synaptic update rule that modified
the plastic synapses to learn the target activities. The learning rule was
based on recursive least squares algorithm (RLS) that was previously
applied to train the read-outs to perform tasks35, 36 and the individual
neurons to generate target activity patterns19,37,45. The derivation pre-
sented here closely follows previous papers37,60. For notational sim-
plicity, we dropped the index i in wi and other variables, e.g., fi, ui. We
note that the same synaptic update rule was applied to all the trained
neurons.
The gradient of the cost function with respect to the vector of full
plastic weights w = (wrec, wX) was
∇
wC =
1
2
∇
w
"
X
t
ðf t (cid:2) ubal,t (cid:2) uplas,t (cid:2) X bal (cid:2) X plas,tÞ2 + λ kwk2 + μ
ð(cid:2)f t + ubal,t + X bal + r0
α2E,I
twÞrt + λw + μ
X
=
t
X
#
ðw (cid:3) 1αÞ2
X
1α10
αw:
α2E,I
ð20Þ
Here we substituted the expressions uplas,t = wrec ⋅ rplas,t and
plas,t in the first line to evaluate the gradient with respect
X plas,t = wX (cid:3) rX
to w.
In the second line, we used a condensed expression
plas,tÞ to denote the synaptically filtered spike trains from
rt = ðrplas,t, rX
all plastic inputs. The vectors 1α apply only to the recurrent plastic
weights wrec and take zero elements on wX.
To derive the synaptic update rule, we computed the gradient at
two consecutive time points
0 = ∇
C =
wn
Xn
ð(cid:2)f t + ubal,t + X bal + r0
twnÞrt + λwn + μ
t = 1
X
α2E,I
1α10
αwn
ð21Þ
16
Article
and
0 = ∇
wn(cid:2)1
C =
Xn(cid:2)1
ð(cid:2)f t + ubal,t + X bal + r0
twn(cid:2)1Þrt + λwn(cid:2)1 + μ
t = 1
Subtracting equations (21) and (22) yielded
https://doi.org/10.1038/s41467-023-38529-y
X
α2E,I
1α10
αwn(cid:2)1
:
ð22Þ
i 2 RT for neuron i and trial-type
lick-right trial types. Each PSTH rc
c ∈ L, R was a T = Ttarget/Δt = 100 dimensional vector defined on time
points t = [ − 2 + Δt, …, − Δt, 0]sec, where 0 is the onset of go-cue.
Next, we converted the PSTHs to target synaptic activities to be
used for training the synaptic inputs to selected neurons. For each
it where i = 1, …, M, c = L, R and t = − 2 + Δt, …, 0, we
spike rate rc
obtained the mean synaptic input f c
it that needs to be applied to the the
leaky integrate-and-fire neuron to generate the desired spike rate. To
this end, we numerically solved the nonlinear rate equation
wn = wn(cid:2)1 + enPnrn
en = f n (cid:2) ubal,n (cid:2) X bal (cid:2) wn(cid:2)1 (cid:3) rn
ð23Þ
where
Pn =
"
Xn
t = 1
rtr0
t + λI + μ
X
α2E,I
#
(cid:2)1
1α10
α
for n ≥ 1
ð24Þ
with the initial value
"
P0 = λI + μ
#
(cid:2)1
1α10
α
:
X
α2E,I
ð25Þ
To update Pn iteratively, we used the Woodbury matrix identity
ðA + UCVÞ(cid:2)1 = A(cid:2)1 (cid:2) A(cid:2)1UðC(cid:2)1 + VA(cid:2)1UÞ
(cid:2)1
VA(cid:2)1
ð26Þ
where A is invertible and N × N, U is N × T, C is invertible and T × T and V
is T × N matrices. Then Pn can be calculated iteratively
Pn = Pn(cid:2)1 (cid:2)
Pn(cid:2)1rnr0
1 + r0
nPn(cid:2)1
nPn(cid:2)1rn
:
ð27Þ
External stimulus triggering target activity patterns. To trigger the
target activity patterns learned by the trained neurons, a brief external
stimulus (200ms long) was applied to every neuron in the network
immediately before generating the activity patterns. Two different sets
of stimuli were prepared to trigger the lick-left and lick-right popula-
tion responses. One set of stimuli was used during and after training to
trigger the lick-left response and the other set of stimulus was used for
the lick-right response. The stimulus X c
stim,iðtÞ to each neuron i and trial
type c = L, R was generated independently from the Ornstein-Ulenbeck
ffiffiffiffiffi
p
δt
stim,iðtÞ + τ(cid:2)1X c
where
process:
τ = 20ms, σ = 0.2 and ξ(t) was uncorrelated Gaussian distribution with
zero mean and unit variance.
stim,iðt + δtÞ = X c
X c
δt + σ
ξðtÞ
stim,i
Generating sinusoidal activity patterns. For demonstrating the Sub-
set Training method (Fig. 1) and the network mechanism for spreading
trained activity (Fig. 5), neurons in the network were trained to follow
sine functions with random phases. Specifically, neuron i in the network
learned the target pattern f iðtÞ = a sinðωt + ϕ
iÞ + bi on the time interval
t = [0, 1]sec, where the amplitude a = 0.5, the frequency ω = 1rad/sec
(Fig. 1) and 2rad/sec (Fig. 5), the phase ϕi was sampled from a uniform
distribution [0, 2π], and the offset bi was the mean synaptic input to the
neuron in the initial balanced network prior to training.
Generating target neural trajectories. A subset of excitatory neurons
in the network learned to reproduce the PSTHs of pyramidal neurons
recorded from ALM in21. For each pyramidal neuron, the spikes emitted
across multiple experiment trials were placed in Δt = 20ms time bins
that ranged over the Ttarget = 2 second delay period. The PSTHs were
then smoothed by a moving average over a 300ms time window
centered at each time bin. We obtained two sets of PSTHs rL
M and
1 , . . . ,rR
rR
M from M = 1824 pyramidal neurons for the lick-left and
1 , . . . ,rL
it = ϕðf c
rc
it ,σ2Þ
ð28Þ
(cid:2)1
ffiffiffiffi
π
p R V thr (cid:2)m
σ
V reset (cid:2)m
σ
1 , . . . ,f R
M and f R
dwew2 erf cð(cid:2)wÞ(cid:4)
where ϕðm,σÞ = τ(cid:2)1
m ½
is the transfer func-
tion of the leaky integrate-and-fire neuron given mean input, m, and
variance of the input, σ227,74. We obtained the synaptic fluctuation σ
from the synaptic noise in the neurons of the initial network since the
slow plastic inputs did not significantly change the fast noise
fluctuation. This conversion yielded two sets of target synaptic inputs
1 , . . . ,f L
f L
M 2 RT for M excitatory neurons to be trained.
We chose the parameters of the initial network connectivity such
that the mean rate of the excitatory and inhibitory populations in the
network was close to estimated mean rates of the ALM data (mean
excitatory rate was 4.2 Hz and inhibitory rate was 11.0 Hz). To select the
subset of excitatory neurons to be trained, we compared the mean
firing rates of the neurons in the initial network with the firing rates of
pyramidal neurons and identified the excitatory neuron whose firing
rate’s was closest to the pyramidal neuron. This process was repeated
until all the pyramidal neurons were matched to the excitatory neu-
rons uniquely.
Generating target synthetic trajectories. To generate synthetic data
that shared similar statistics and low-dimensional dynamics as the
neural data, we performed PCA on the PSTHs of pyramidal ALM neu-
rons to identify the principal components v1, . . . ,vD 2 RT
that
explained majority of their variance. We found that D = 9 was large
enough to explain over 95% of the variance. The same procedure was
applied to the PSTHs of the fast-spiking ALM neurons to obtain their
principal components.
n
n
P
D
We sought to construct synthetic trajectories rsynth 2 RT that
resembled the PSTHs of the pyramidal and fast-spiking ALM neurons
(Fig. S6). To this end, we expressed the synthetic trajectory rsynth as a
n = 1 csynth
vn. To
weighted sum of the principal components: rsynth =
find the distribution of the coefficients cneural
of the neural data, we
projected the PSTHs of pyramidal neurons onto the PCs and obtained
the empirical distribution of cneural
= rneural (cid:3) vn. Bootstrapping the
synthetic coefficients csynth
from the empirical distribution of cneural
was performed in two steps. First, the mean firing rate of synthetic
target was sampled from the empirical rate distribution to generate
synthetic PSTHs that had rate distribution statistically identical to the
empirical distribution (Fig. S6A,B). Next, since cneural
depended
strongly on the mean firing rate of neurons, csynth
was bootstrapped
n
whose underlying firing rate was close to the
from a subset of cneural
firing rate of synthetic target (Fig. S6C). In this way, the distributions of
the firing rates and PC loadings of the synthetic and neural data were
almost identical (Fig. S6E).
n
n
n
n
n
In addition, we generated the synthetic PSTHs in pairs for the lick
right and lick left trials. First, the PSTHs for the lick right and lick left
conditions were generated independently. Then, we sorted the PSTH’s
of each condition separately and paired them, to ensure the pairs had
similar level of mean firing rates. Subsequently, we added Gaussian
noise with zero mean and standard deviation equal to the difference of
lick right and lick left mean firing rates, to the PSTH’s of the lick left
condition. This allowed us to introduce choice selectivity to the
synthetic PSTHs.
Nature Communications |
(2023) 14:2851
17
Article
https://doi.org/10.1038/s41467-023-38529-y
Table 3 | Network parameters used for simulating a weakly
coupled network in Figure S10
Target functions
Neurons
N, NE, NI
Ntrained
Supp. Figure S10
Same as target functions
of Fig. 2
# total, exc, inh neurons
Same as param of Fig. 2
# trained neurons
Static synapses to a neuron
p
K, KE, KI
conn prob of static synapses
Same as param of Fig. 2
# total, exc, inh static synap-
ses to a neuron
Static synaptic weights to a trained neuron
Jweak
E
weak excitatory synaptic
weight
Jweak
I
weak inhibitory synaptic
weight
Jweak
EE
Jweak
IE
Jweak
EI
Jweak
II
γE
γI
External inputs to neurons
E to E static synaptic weight
E to I static synaptic weight
I to E static synaptic weight
I to I static synaptic weight
relative strength of WEE to WIE
relative strength of WEI to WII
2.0/KE
− 2.0/KI
γ
EJweak
E
Jweak
E
γ
IJweak
I
Jweak
I
Same as param of Fig. 2
Xweak
Xgaussian
Xweak
E
Xweak
I
weak external input
0.35
Gaussian input to neurons
From an untrained
balanced network
external input to excitatory
neurons
external input to inhibitory
neurons
1.5Xweak
0.8Xweak
Plastic synaptic weights to a trained neuron
normally distributed mean inputs received by neurons in the strongly
coupled balanced network. We also injected external white noise to
neurons, which, together with the additional (normally distributed)
inputs, produced log-normal firing rate distribution in the weakly
coupled network. The white noise was inject to mimic the stochastic
spiking activity of neurons in the balanced network and also produced
exponentially expansive nonlinear activation function27. Finally, a
uniform external excitatory X weak
) input was applied
to all excitatory (inhibitory) neurons to adjust the mean excitatory
(inhibitory) firing rates to be close to the mean firing rate of ALM
pyramidal (fast-spiking) neurons.
(inhibitory X weak
E
I
Mathematical analysis of inputs to untrained neurons
In this part of the methods we use mathematical analysis to show how
random inputs from trained neurons can drive the untrained neurons
to follow the trained activity, without further training, if the network
operates in the balanced regime.
To simplify the analysis, we assumed that only the excitatory
population was trained and the inhibitory population was not. In
addition, we assumed that the target functions, fit for neuron i and
t ∈ [0, Ttarget], were slower than the slow plasticity signal and
that training was perfect. In this case, we can approximate the total
synaptic input to a trained excitatory neuron using the fixed point
equation:
i ðtÞ ≈
uE
X
XNβ
β2fE,Ig
j = 1
JEβ
ij
ϕðu
β
j ðtÞÞ +
X
XNβ
β2fE,Ig
j = 1
W Eβ
ij
ϕðu
β
j ðtÞÞ +
p
ffiffiffiffi
K
X E
ð29Þ
p
ffiffiffiffi
K
with
network23. The transfer function, ϕðuα
function27, 74, with σ2
ϕα = ½hϕα
average over the neurons.
X E the strong external input associated with the balanced
; σαÞ, was the Riccardi
I . The population rate was given by
it i(cid:4), with 〈x〉 denoting the average over the time and [x] the
i Þ = Φðuα
(cid:2)
2
J
EE
(cid:2)
2
J
EI
E =
E +
ϕ
ϕ
i
E to E plastic synaptic weight
Same as param of Fig. 2
Similarly, the total synaptic input to an untrained neuron, which
E to I plastic synaptic weight
I to E plastic synaptic weight
I to I plastic synaptic weight
Number of plastic synapses to a trained neuron
p
order of # plastic synapses
lacked plastic connections, followed:
iðtÞ ≈
uI
X
XNβ
β2fE,Ig
j = 1
JIβ
ij
ϕðu
β
j ðtÞÞ +
p
ffiffiffiffi
K
X I
ð30Þ
Same as param of Fig. 2
WEE
WIE
WEI
WII
ffiffiffiffi
K
Lrec
Lffwd
# recurrent plastic synapses
# ffwd plastic synapses
L = Lrec + Lffwd
# total plastic synapses
The synaptic weights of initial connectivity Jweak and external inputs Xweak, Xgaussian are modified
from the network parameters of Figure 2 to set up a weakly coupled initial network. See Fig-
ure S10 for further explanations of the modified parameters.
with σ2
I =
(cid:2)
2
J
IE
ϕ
E +
(cid:2)
2
J
II
ϕ
I.
Our goal was to analyze the synaptic drive from the trained
(excitatory) neurons to untrained (inhibitory) neurons to make specific
predictions about what aspects of the trained inputs allowed them to
spread effectively to the untrained neurons.
The synthetic PSTHs were then converted into target synaptic
inputs following the same procedure applied to the neural PSTHs.
p
Initializing weakly coupled network. Here we describe how the initial
parameters of a weakly coupled network were set up to match the
population activity of ALM neurons (and the strongly coupled
balanced network). All the network parameters for the weakly coupled
network are reported in Table 3 and explained in Fig. S10. First, the
initial connections of the weakly coupled network were scaled by 1/K,
instead of the 1=
scaling as in the balanced network. The 1/K scaling
produced synaptic weights that averaged the spiking activity of pre-
synaptic neurons. However, with such weak coupling, the network did
not produce a log-normal firing rate distribution, which was needed to
pair ALM neurons with model neurons to be trained based on proxi-
mity of their mean firing rates. Therefore, each neuron i received
additional constant input, denoted by X gaussian
, that varied across
neurons. More specifically, the additional inputs were identical to the
ffiffiffiffi
K
i
Statistics of random inputs from the trained neurons to an
untrained neuron. If an excitatory neuron i is successfully trained, its
firing rate closely follows the target activity fit. We used a shorthand
notation ϕα
i ðtÞÞ and expressed the firing rate of the trained
neuron in the form ϕE
it , with the temporal modulation
δϕE
it. We next considered the singular value decomposition of the
it = ϕðuα
it i + δϕE
it = hϕE
temporal modulation:
δϕE
it =
q
ffiffiffiffiffi
λE
n
V nt
Uin
XT
n = 1
ð31Þ
which is N × T matrix, and where U is a N × N matrix of the left singular
vectors and V is T × T matrix of the right singular vectors. Here, we
considered a discretized version of time with T = Ttarget/Δt, such that
the matrices are of finite size. The values
(SVs) and λE
n are the elements of the spectrum of the covariance matrix
of the trained excitatory neurons. For instance, if we choose the target
are the singular values
ffiffiffiffiffi
λE
n
q
Nature Communications |
(2023) 14:2851
18
Article
https://doi.org/10.1038/s41467-023-38529-y
activity to be sinusoidal functions with random phases (Fig. 5A), the
covariance matrix is stationary and the right singular vectors are the
Fourier modes (e.g., V 1t / sinðωtÞ,V 2t / cosðωtÞ).
Untrained (inhibitory) neurons do not receive plastic synapses.
Thus, the aggregate input from the trained neurons to an untrained
it , is a random summation of trained neurons’ activity. It is
neuron, uIE
given by:
uIE
it =
X
j
JIE
ij
ϕE
jt = ½huIE
it i(cid:4) + ΔuIE
i + δuIE
it
ð32Þ
with the average population input ½huIE
1. The second term in
equation (32) is the quenched disorder44 and its variance is given by:
it i(cid:4) =
ffiffiffiffi
p (cid:2)
JIE
K
ϕ
2
½ðΔuIE
i Þ
(cid:4) =(cid:2)J
2
IE ½hϕE
it i
2
(cid:4) = qIE
ð33Þ
The last term in equation (32) is the temporal modulation of the
P
aggregate trained input, δuIE
jt. Using equation (31), it is
given by:
it =
δϕE
JIE
ij
δuIE
i ðtÞ =
(cid:2)
JIE
q
ffiffiffiffiffi
λE
n
V nt
:
~ain
XT
n = 1
ð34Þ
j
K
p
P
ΛIE
the coefficients
where due to the Central Limit Theorem,
~ain = 1ffiffiffi
ij Unj are Gaussian vectors with zero mean and unit var-
iance in the large K limit (see Prediction 1 below). Here, ΛIE
is the
ij
adjacency matrix, indicating which neurons are connected, and we
assumed that the left singular vectors Uin’s are random variables with
zero mean and unit variance. Importantly, we emphasize that it is the
strong coupling (i.e., synaptic weights scale as 1=
) that allows the
coefficients ~ain
’s to have finite variance. This is not the case if synaptic
weights are weak (see Prediction 2 below). In addition, the variance of
the coefficients of temporal modulation is (cid:2)J
λE
n, which shows that the
ffiffiffiffi
K
2
IE
p
, determine the strength of temporal modulation (see Pre-
q
SVs,
ffiffiffiffiffi
λE
n
diction 3 below).
With this, the synaptic input to an untrained neuron from the
trained population can be written in the following form:
ffiffiffiffi
K
p (cid:2)JIE
p
ffiffiffiffiffiffiffi
qIE
ϕ
E +
uIE
it =
i + δuIE
zE
it
ð35Þ
with zE
variance.
i being a Gaussian random variable with zero mean and unit
For example, when the target functions are sinusoidal functions
with random phases (Figs. 1, 5) these temporal modulations are:
δuIE
it =
(cid:2)
JIE
q
XT
n = 1
ffiffiffiffiffi
λE
n
½ani cosðnωtÞ + bni sinðnωtÞ(cid:4)
ð36Þ
where we replaced ~ani in equation (34) with the even and odd coeffi-
cients of the cosine and sine functions, ani, bni, respectively.
Similarly, in the case of the ALM data, the dominant right singular
vector is a ramping mode (Fig. 2E,F), i.e. V1t ∝ t and the temporal
modulations are dominated by:
δuIE
it
≈ (cid:2)JIE
q
ffiffiffiffiffi
λE
1
~a1it
ð37Þ
with ~a1i
∼ N ð0,1Þ.
The recurrent untrained inputs and implications. The synaptic input
to an untrained inhibitory neuron consists of a large, Oð
Þ, and
positive mean drive from the excitatory neurons. The untrained
neurons will thus fire with high rates and regular spiking, unless the
network operates in the balanced regime, in which the recurrent
ffiffiffiffi
K
p
inhibition cancels most of this large excitatory drive23. In this case,
the untrained neuron will be driven by the temporal modulations
originating from the random summation of the activity of trained
neurons, which are of Oð1Þ due to the strong coupling. This input is
spanned by the principal components (or, equivalently, the right
singular vectors) of the trained population according to equa-
tion (34).
A similar analysis on the recurrent inputs from the untrained
inhibitory population, uII
it , needs to be done to infer the statistics of the
temporal fluctuations of the net input, δuI
it , of the
untrained inhibitory neurons. This analysis needs to be done in a self-
consistent way to determine the statistics of δϕI
73. While this analysis
it
is beyond the scope of the current paper, several observations can be
made already by examining the statistics of the inputs from the trained
population.
it, and rates, δϕI
Prediction 1. No matter what the right singular vectors (which
we refer to as the PCs in the main text) are, their coefficients are
expected to be Gaussian. This prediction is shown in Fig. 5G for
artificial target functions of sine functions with random phases, as
well as in Fig. 5H for the coefficients of the dominant ramping mode
in the neural data.
Prediction 2. The spread of activity in the network is possible only
’s in equation (34) is finite. It is a result of the
p scaling of the synapses,
because the variance of ~ait
strong coupling in the network, i.e. the 1ffiffiffi
K
which guarantees, due to the Central Limit Theorem, that the variance
of the aggregate input from the trained neurons converge is finite. This
(cid:2)
Jαβ
K instead of
is in contrast to the case of weak synapses (e.g., scaling of
(cid:2)
Jαβffiffiffi
p ), where the variance of ~ait converges to zero in the large K limit
K
(Fig. S9, no spreading of trained activity in a weakly coupled network).
Prediction 3. The strength of the transfer of the trained activity to
the untrained neurons depends on the variance of the trained popu-
lation through equation (34). As shown in Fig. 4B, in the ALM data the
variance of the temporal modulations of the inhibitory neurons is
larger than those of the excitatory neurons. This result suggests why
the fidelity of the spread improved when the inhibitory population was
trained instead of the excitatory population. It also explains why
leading PC modes of the activity can spread better in the network, as
in equation (34)) are, by definition, lar-
their corresponding SVs (
ger than those of the higher mode PCs.
ffiffiffiffiffi
λE
n
q
Prediction 4. This framework provides additional insights into
how excitatory neurons trained to be choice-selective can impart the
learned selectivity to the untrained inhibitory neurons through
nonspecific, strong synaptic connections (see Fig. 3E). To show this,
one can estimate the statistics of the difference in the input to an
untrained inhibitory neuron from the trained population for the lick-
right and lick-left trials. For instance, if we consider the target func-
tions to be defined by the dominant ramping mode that captures
over 70% of the variance (Fig. 2E), the relevant basis function would
be V1t ∝ t for t ∈ [0, Ttarget], and the selectivity of the trained inputs
(SIE) yields
SIE
i = uIE,right
it
(cid:2) uIE,lef t
it
≈AΔzi + BΔ~a1it
ð38Þ
i
(cid:2) zE,lef t
i
where A and B determine the variance in the baseline inputs and
ramping rates, respectively. From equation (35), the quenched dis-
order yields Gaussian variables Δzi = zE,right
, with a finite var-
iance A2. From equation (37), the temporal modulation yields a
Gaussian variables Δ~a1i = ~aright
, with a finite variance B2. Because
Δzi and Δ~a1i are random variables with finite variances, the trained
inputs develop choice selectivity, which can then elicit choice selec-
tivity in the untrained inhibitory neurons (Fig. 3D). The good agree-
ment of the distribution of choice-selectivity in the untrained neurons
in the model and the putative fast-spiking neurons in the neural data
(Fig. 3D) is consistent with this prediction.
1i (cid:2) ~alef t
1i
Nature Communications |
(2023) 14:2851
19
Article
https://doi.org/10.1038/s41467-023-38529-y
Reporting summary
Further information on research design is available in the Nature
Portfolio Reporting Summary linked to this article.
21. Finkelstein, A. et al. Attractor dynamics gate cortical information
flow during decision-making. Nat. Neurosci. 24, 843–850 (2021).
22. Andalman, A. S. et al. Neuronal dynamics regulating brain and
Data availability
Spike recording data in NWB format are available for download at
https://dandiarchive.org/dandiset/000060/draft. Source data are
provided with this paper.
23. Van Vreeswijk, C. & Sompolinsky, H. Chaos in neuronal networks
with balanced excitatory and inhibitory activity. Science 274,
1724–1726 (1996).
24. Brunel, N. Dynamics of sparsely connected networks of excitatory
behavioral state transitions. Cell 177, 970–985 (2019).
Code availability
The Julia code for training spiking neural network is available at https://
github.com/SpikingNetwork/distributedActivity75.
References
1.
Chen, T.-W., Li, N., Daie, K. & Svoboda, K. A map of anticipatory
activity in mouse motor cortex. Neuron 94, 866–879 (2017).
2. Allen, W. E. et al. Global representations of goal-directed behavior in
distinct cell types of mouse neocortex. Neuron 94, 891–907 (2017).
3. Makino, H. et al. Transformation of cortex-wide emergent proper-
ties during motor learning. Neuron 94, 880–890 (2017).
4. Steinmetz, N. A., Zatka-Haas, P., Carandini, M. & Harris, K. D. Dis-
tributed coding of choice, action and engagement across the
mouse brain. Nature 576, 266–273 (2019).
5. Allen, W. E. et al. Thirst regulates motivated behavior through
modulation of brainwide neural population dynamics. Science 364,
eaav3932 (2019).
6. Roth, R. H. et al. Cortical synaptic ampa receptor plasticity during
motor learning. Neuron 105, 895–908 (2020).
7. Chen, S. X., Kim, A. N., Peters, A. J. & Komiyama, T. Subtype-specific
plasticity of inhibitory circuits in motor cortex during motor learn-
ing. Nat. Neurosci 18, 1109–1115 (2015).
8. Xu, T. et al. Rapid formation and selective stabilization of synapses
for enduring motor memories. Nature 462, 915–919 (2009).
Yang, G., Pan, F. & Gan, W.-B. Stably maintained dendritic spines are
associated with lifelong memories. Nature 462, 920–924 (2009).
9.
and inhibitory spiking neurons. J. Comput. Neurosci. 8,
183–208 (2000).
25. Renart, A. et al. The asynchronous state in cortical circuits. Science
327, 587–590 (2010).
26. Shadlen, M. N. & Newsome, W. T. The variable discharge of cortical
neurons: implications for connectivity, computation, and informa-
tion coding. J. Neurosci. 18, 3870–3896 (1998).
27. Roxin, A., Brunel, N., Hansel, D., Mongillo, G. & van Vreeswijk, C. On
the distribution of firing rates in networks of cortical neurons. J.
Neurosci. 31, 16217–16226 (2011).
28. Tanji, J. & Evarts, E. V. Anticipatory activity of motor cortex neurons
in relation to direction of an intended movement. J. Neurophysiol.
39, 1062–1068 (1976).
29. Fetz, E. E. Are movement parameters recognizably coded in the
activity of single neurons? Behav. Brain Sci. 154 (1992).
30. Churchland, M. M., Cunningham, J. P., Kaufman, M. T., Ryu, S. I. &
Shenoy, K. V. Cortical preparatory activity: representation of
movement or first cog in a dynamical machine? Neuron 68,
387–400 (2010).
Inagaki, H. K., Inagaki, M., Romani, S. & Svoboda, K. Low-
dimensional and monotonic preparatory activity in mouse anterior
lateral motor cortex. J. Neurosci. 38, 4163–4185 (2018).
31.
32. Guo, Z. V. et al. Flow of cortical activity underlying a tactile decision
in mice. Neuron 81, 179–194 (2014).
33. Rosenbaum, R., Smith, M. A., Kohn, A., Rubin, J. E. & Doiron, B. The
spatial structure of correlated neuronal variability. Nat. Neurosci.
20, 107–114 (2017).
10. Nabavi, S. et al. Engineering a memory with ltd and ltp. Nature 511,
34. Ko, H. et al. Functional specificity of local synaptic connections in
348–352 (2014).
neocortical networks. Nature 473, 87–91 (2011).
11. Humeau, Y. & Choquet, D. The next generation of approaches to
investigate the link between synaptic plasticity and learning. Nat.
Neurosci 22, 1536–1543 (2019).
12. Takemoto, K. et al. Optical inactivation of synaptic ampa receptors
erases fear memory. Nat. Biotechnol. 35, 38–47 (2017).
35. Sussillo, D. & Abbott, L. Generating Coherent Patterns of Activity
from Chaotic Neural Networks. Neuron 63, 544–557 (2009).
36. Nicola, W. & Clopath, C. Supervised learning in spiking neural
networks with force training. Nat. Commun. 8, 2208 (2017).
37. Kim, C. M. & Chow, C. C. Learning recurrent dynamics in spiking
13. Hayashi-Takagi, A. et al. Labelling and optical erasure of synaptic
memory traces in the motor cortex. Nature 525, 333–338 (2015).
14. Kakegawa, W. et al. Optogenetic control of synaptic ampa receptor
endocytosis reveals roles of ltd in motor learning. Neuron 99,
985–998 (2018).
15. Markram, H., Lübke, J., Frotscher, M. & Sakmann, B. Regulation of
synaptic efficacy by coincidence of postsynaptic aps and epsps.
Science 275, 213–215 (1997).
16. Bi, G.-q & Poo, M.-m Synaptic modifications in cultured hippo-
networks. eLife 7, e37124 (2018).
38. Litwin-Kumar, A. & Doiron, B. Slow dynamics and high variability in
balanced cortical networks with clustered connections. Nat. Neu-
rosci. 15, 1498–1505 (2012).
Inagaki, H. K., Fontolan, L., Romani, S. & Svoboda, K. Discrete
attractor dynamics underlies persistent activity in the frontal cortex.
Nature 566, 212–217 (2019).
39.
40. Stringer, C. et al. Spontaneous behaviors drive multidimensional,
brainwide activity. Science. 364 (2019).
campal neurons: dependence on spike timing, synaptic strength,
and postsynaptic cell type. J Neurosci 18, 10464–10472 (1998).
17. Tsutsumi, S. & Hayashi-Takagi, A. Optical interrogation of multi-
41. Li, N., Chen, T.-W., Guo, Z. V., Gerfen, C. R. & Svoboda, K. A motor
cortex circuit for motor planning and movement. Nature 519,
51–56 (2015).
scale neuronal plasticity underlying behavioral learning. Curr. Opin.
Neurobiol. 67, 8–15 (2021).
18. Fisher, D., Olasagasti, I., Tank, D. W., Aksay, E. R. & Goldman, M. S. A
modeling framework for deriving the structural and functional
architecture of a short-term memory microcircuit. Neuron 79,
987–1000 (2013).
19. Rajan, K., Harvey, C. D. & Tank, D. W. Recurrent Network Models of
Sequence Generation and Memory. Neuron 90, 128–142 (2016).
20. Daie, K., Svoboda, K. & Druckmann, S. Targeted photostimulation
uncovers circuit motifs supporting short-term memory. Nat. Neu-
rosci. 24, 259–265 (2021).
42. Khan, A. G. et al. Distinct learning-induced changes in stimulus
selectivity and interactions of gabaergic interneuron classes in
visual cortex. Nat. Neurosci. 21, 851–859 (2018).
43. Najafi, F. et al. Excitatory and inhibitory subnetworks are equally
selective during decision-making and emerge simultaneously dur-
ing learning. Neuron 105, 165–179 (2020).
44. van Vreeswijk, C. & Sompolinsky, H. Chaotic balanced state in a
model of cortical circuits. Neural Comput. 10, 1321–1371 (1998).
45. Laje, R. & Buonomano, D. V. Robust timing and motor patterns by
taming chaos in recurrent neural networks. Nat. Neurosci. 16,
925–933 (2013).
Nature Communications |
(2023) 14:2851
20
Article
https://doi.org/10.1038/s41467-023-38529-y
46. Kullmann, D. M., Moreau, A. W., Bakiri, Y. & Nicholson, E. Plasticity of
inhibition. Neuron 75, 951–962 (2012).
47. Mongillo, G., Rumpel, S. & Loewenstein, Y. Inhibitory connectivity
defines the realm of excitatory plasticity. Nat. Neurosci. 21,
1463–1470 (2018).
48. Kim, R. & Sejnowski, T. J. Strong inhibitory signaling underlies stable
temporal dynamics and working memory in spiking neural net-
works. Nat. Neurosci. 24, 129–139 (2021).
49. Hofer, S. B. et al. Differential connectivity and response dynamics of
excitatory and inhibitory neurons in visual cortex. Nat. Neurosci. 14,
1045–1052 (2011).
50. Hansel, D. & van Vreeswijk, C. The mechanism of orientation
selectivity in primary visual cortex without a functional map. J.
Neurosci. 32, 4049–4064 (2012).
51. Pehlevan, C. & Sompolinsky, H. Selectivity and sparseness in ran-
domly connected balanced networks. PloS one 9, e89992 (2014).
52. Marshel, J. H. et al. Cortical layer–specific critical dynamics trig-
gering perception. Science 365, eaaw5202 (2019).
53. Dalgleish, H. W. et al. How many neurons are sufficient for per-
ception of cortical activity? Elife 9, e58889 (2020).
54. Akitake, B. et al. Amplified cortical neural responses as animals
learn to use novel activity patterns. bioRxiv (2022).
55. Caporale, N. & Dan, Y. et al. Spike timing-dependent plasticity: a
hebbian learning rule. Ann. Rev. Neurosci. 31, 25–46 (2008).
56. Abbott, L. F. & Nelson, S. B. Synaptic plasticity: taming the beast.
Nat. Neurosci. 3, 1178–1183 (2000).
57. Bi, G.-q & Poo, M.-m Synaptic modification by correlated activity:
Hebb’s postulate revisited. Ann. Rev. Neurosci. 24, 139–166 (2001).
58. Baker, C., Zhu, V. & Rosenbaum, R. Nonlinear stimulus representa-
tions in neural circuits with approximate excitatory-inhibitory bal-
ance. PLoS Comput. Biol. 16, e1008192 (2020).
Ingrosso, A. & Abbott, L. Training dynamically balanced excitatory-
inhibitory networks. PloS one 14, e0220547 (2019).
59.
60. Kim, C. M. & Chow, C. C. Training Spiking Neural Networks in the
Strong Coupling Regime. Neural Comput. 33, 1199–1233 (2021).
61. Engelhard, B. et al. Neuronal activity and learning in local cortical
71. Averbeck, B. B., Latham, P. E. & Pouget, A. Neural correlations, popu-
lation coding and computation. Nat. Rev. Neurosci. 7, 358–366 (2006).
72. Darshan, R., Van Vreeswijk, C. & Hansel, D. Strength of correlations in
strongly recurrent neuronal networks. Phys. Rev. X 8, 031072 (2018).
73. Darshan, R. & Rivkind, A. Learning to represent continuous variables
in heterogeneous neural networks. Cell Rep. 39, 110612 (2022).
74. Tuckwell, H. C.Introduction to theoretical neurobiology: linear
cable theory and dendritic structure, vol. 1 (Cambridge University
Press, 1988).
75. Kim, C. M., Finkelstein, A., Chow, C. C., Svoboda, K. & Darshan, R.
Distributing task-related neural activity across a cortical network
through task-independent connections https://doi.org/10.5281/
zenodo.7684907 (2023).
76. Finkelstein, A. et al. Attractor dynamics gate cortical information
flow during decision-making. Nat Neurosci 24, 843–850 (2021).
Acknowledgements
We would like to thank Larry Abbott and Sandro Romani for their valuable
feedback. A.F., K.S. and R.D. were supported by the Howard Hughes
Medical Institute. C.M.K. and C.C.C were supported by the Intramural
Research Program at the NIDDK/NIH. C.M.K. would like to thank the Vis-
iting Scientist Program at Janelia Research Campus for their support.
Author contributions
C.M.K and R.D. conceived the research, ran simulations and analyzed the
data. A.F. and K.S. designed the experiments. A.F. collected the
experimental data. C.M.K, R.D., A.F., C.C.C and K.S. wrote the paper.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information The online version contains
supplementary material available at
https://doi.org/10.1038/s41467-023-38529-y.
networks are modulated by the action-perception state. bioRxiv
537613 (2019).
Correspondence and requests for materials should be addressed to
Christopher M. Kim or Ran Darshan.
62. Mastrogiuseppe, F. & Ostojic, S. Linking connectivity, dynamics,
and computations in low-rank recurrent neural networks. Neuron
99, 609–623 (2018).
63. Schuessler, F., Mastrogiuseppe, F., Dubreuil, A., Ostojic, S. & Barak, O.
The interplay between randomness and structure during learning in
rnns. Adv. Neural Inf. Process. Syst. 33, 13352–13362 (2020).
64. Ahmadian, Y. & Miller, K. D. What is the dynamical regime of cere-
bral cortex? Neuron 109, 3373–3391 (2021).
65. Tsodyks, M. V., Skaggs, W. E., Sejnowski, T. J. & McNaughton, B. L.
Paradoxical effects of external modulation of inhibitory inter-
neurons. J. Neurosci. 17, 4382–4388 (1997).
66. Sanzeni, A. et al. Inhibition stabilization is a widespread property of
cortical networks. Elife 9, e54875 (2020).
67. Sadeh, S. & Clopath, C. Patterned perturbation of inhibition can
reveal the dynamical structure of neural processing. Elife 9,
e52757 (2020).
68. Mahrach, A., Chen, G., Li, N., van Vreeswijk, C. & Hansel, D.
Mechanisms underlying the response of mouse cortical networks to
optogenetic manipulation. Elife 9, e49967 (2020).
69. Darshan, R., Wood, W. E., Peters, S., Leblois, A. & Hansel, D. A
canonical neural mechanism for behavioral variability. Nat. Com-
mun.8, 1–13 (2017).
70. Lebovich, L., Darshan, R., Lavi, Y., Hansel, D. & Loewenstein, Y.
Idiosyncratic choice bias naturally emerges from intrinsic
stochasticity in neuronal dynamics. Nat. Human Behav. 3,
1190–1202 (2019).
Peer review information Nature Communications thanks Omri Barak
and the other, anonymous, reviewer(s) for their contribution to the peer
review of this work. A peer review file is available.
Reprints and permissions information is available at
http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to jur-
isdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as
long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate if
changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons license and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.
© The Author(s) 2023
Nature Communications |
(2023) 14:2851
21
| null |
10.3390_biom11010118.pdf
|
Data Availability Statement: All data are contained within the article or supplementary materia
|
Data Availability Statement: All data are contained within the article or supplementary material.
|
Article
Ginsenoside Rg3 Prevents Oncogenic Long Noncoding RNA
ATXN8OS from Inhibiting Tumor-Suppressive
microRNA-424-5p in Breast Cancer Cells
Heejoo Kim †, Hwee Won Ji †, Hyeon Woo Kim , Sung Hwan Yun, Jae Eun Park and Sun Jung Kim *
Department of Life Science, Dongguk University-Seoul, Goyang 10326, Korea; [email protected] (H.K.);
[email protected] (H.W.J.); [email protected] (H.W.K.); [email protected] (S.H.Y.);
[email protected] (J.E.P.)
* Correspondence: [email protected]; Tel.: +82-31-961-5129
† These authors contributed equally to this work.
Abstract: Ginsenoside Rg3 exerts antiproliferation activity on cancer cells by regulating diverse
noncoding RNAs. However, little is known about the role of long noncoding RNAs (lncRNAs) or their
relationship with competitive endogenous RNA (ceRNA) in Rg3-treated cancer cells. Here, a lncRNA
(ATXN8OS) was found to be downregulated via Rg3-mediated promoter hypermethylation in MCF-7
breast cancer cells. SiRNA-induced downregulation of ATXN8OS decreased cell proliferation but
increased apoptosis, suggesting that the noncoding RNA possessed proproliferation activity. An in
silico search for potential ATXN8OS-targeting microRNAs (miRs) identified a promising candidate
(miR-424-5p) based on its high binding score. As expected, miR-424-5p suppressed proliferation
and stimulated apoptosis of the MCF-7 cells. The in silico miR-target-gene prediction identified
200 potential target genes of miR-424-5p, which were subsequently narrowed down to seven that
underwent hypermethylation at their promoter by Rg3. Among them, three genes (EYA1, DACH1,
and CHRM3) were previously known oncogenes and were proven to be oppositely regulated by
ATXN8OS and miR-424-5p. When taken together, Rg3 downregulated ATXN8OS that inhibited the
tumor-suppressive miR-424-5p, leading to the downregulation of the oncogenic target genes.
Keywords: ceRNA; CpG methylation; ginsenoside Rg3; long noncoding RNA; microRNA
1. Introduction
Ginsenoside Rg3 is a steroidal saponin derivative that is abundant in heat-processed
ginseng extract [1]. Rg3 possesses potent anticancer properties and is known to modulate
diverse cellular events such as cell proliferation, immune response, autophagy, metastasis,
and angiogenesis [2]. Rg3 activates proapoptotic proteins such as caspase-3 and Bax but
suppresses antiapoptotic protein Bcl-2 [3]. In the process, NF-κB, which drives cell-cycle
progression, is inhibited by blocking the phosphorylation of Akt and ERK kinases [4].
In MDA-MB-231 breast-cancer cells, Bcl-2 can be suppressed by destabilizing a mutant P53
with Rg3 [4]. In osteosarcoma cell lines, Rg3 inhibits migration and invasion by suppressing
MMPs and the Wnt/β-catenin pathway, which are related to epithelial-mesenchymal
transition (EMT) and angiogenesis [5]. Rg3-treated gastric cancer cells show a remarkably
lower expression of HIF-1α and VEGF under hypoxia [6]. The SNAIL signaling axis is
another key pathway regulated by Rg3 during metastasis, which regulates EGFR and
fibronectin in cancer stem cells [7].
Rg3 can inhibit cancer-cell growth by modulating epigenetic factors of oncogenes
or tumor suppressors. A genome-wide methylation analysis identified over 250 genes
with significant changes in methylation level at specific CpG sites in Rg3-treated MCF-7
breast-cancer cells [8]. These genes were largely associated with cell-morphology-related
pathways. Notably, NOX4 and KDM5A were hyper- and hypo-methylated on their pro-
Citation: Kim, H.; Ji, H.W.; Kim,
H.W.; Yun, S.H.; Park, J.E.; Kim, S.J.
Ginsenoside Rg3 Prevents Oncogenic
Long Noncoding RNA ATXN8OS
from Inhibiting Tumor-Suppressive
microRNA-424-5p in Breast Cancer
Cells. Biomolecules 2021, 11, 118.
https://doi.org/10.3390/biom
11010118
Received: 30 December 2020
Accepted: 14 January 2021
Published: 18 January 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under
the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Biomolecules 2021, 11, 118. https://doi.org/10.3390/biom11010118
https://www.mdpi.com/journal/biomolecules
biomolecules(cid:1)(cid:2)(cid:3)(cid:1)(cid:4)(cid:5)(cid:6)(cid:7)(cid:8)(cid:1)(cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:6)(cid:7) Biomolecules 2021, 11, 118
2 of 13
moter regions, respectively, which led to gene dysregulation and increases in cell apopto-
sis [8]. Other genes such as p53, Bcl-2, and EGF were affected by Rg3-mediated promoter
methylation in the HepG2-hepatocarcinoma cell line [9]. Approximately a dozen (mi-
croRNAs) miRs are known to be regulated by Rg3, many of which are involved in cancer
malignancy, metastasis, or EMT [10,11]. For example, miR-145 comprises the DNMT3A-
miR-145-FSCN1 axis in ovarian cancer, and its downregulation by Rg3 inhibits EMT [12].
Recently, miRs associated with the Warburg effect [13] and autophagy [14] were identified
as Rg3 targets. Rg3 upregulated miR-519a-5p via reducing DNMT3A-mediated DNA
methylation to inhibit an HIF-1α-stimulated Warburg effect in ovarian cancer [13]. MiR-
181b impaired the antiautophagy effect of Rg3-mediated tumor cytotoxicity by modulating
the CREBRF/CREB3 signaling pathways in gallbladder cancer [14].
LncRNAs (i.e., noncoding RNAs larger than 200 nucleotides) are known to regulate a
variety of genes, leading to tumor-development stimulation or suppression [15]; however,
only a few lncRNAs have been identified as Rg3 targets. LncRNA-CASC2 is upregulated by
Rg3, thereby activating PTEN signaling and suppressing drug-resistant pancreatic cancer
cells [16]. Two tumor-related lncRNAs (RFX3-AS1 and STXBP5-AS1) have been identified
in Rg3-treated MCF-7 cells, and their expression is controlled by promoter methylation [17].
Moreover, lncRNA CCAT1 induces Caco-2 colorectal-cancer-cell proliferation but is also
downregulated by Rg3 [18].
A number of epigenetic factors have been found to act in conjunction to regulate the
expression of specific target genes. Moreover, competitive endogenous RNA (ceRNA)
sponges miR by sharing the same target gene recognition sequence [19]. For example,
lncRNA H19 acts as a miR-340-3p sponge to promote epithelial-mesenchymal transition in
breast-cancer cells [20], thereby disrupting the gene-suppression activity of miR. Although
ginsenosides are known to regulate miRs and lncRNAs in cancer cells, few studies have
characterized the role of ceRNA. In this study, a genome-wide methylation-array dataset
was analyzed to identify lncRNAs that were epigenetically regulated by Rg3. Notably,
the lncRNA ATXN8OS was found to be hypermethylated by Rg3 in MCF-7 breast-cancer
cells. The effect of Rg3 on ATXN8OS expression was then examined, and the role of
the lncRNA in cancer-cell growth was elucidated. A miR that interacts with ATXN8OS
was examined to identify sponge-activity relationships between the two RNAs during
miR-mediated gene regulation in the presence of Rg3.
2. Materials and Methods
2.1. Cell Culture
Human mammary-gland-derived cell lines (MCF-10A, MCF-7, and MDA-MB-231)
were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA).
MCF-10A was cultured in MEGM (Lonza, Basel, Switzerland) with 100 ng/mL cholera
toxin. MCF-7 and MDA-MB-231 were cultured in RPMI 1640 medium (Welgene, Seoul,
Korea) supplemented with 10% fetal bovine serum (Capricorn Scientific, Ebsdorfergrund,
Germany). All cells were supplemented with 2% penicillin/streptomycin (Capricorn
Scientific) and cultured at 37 ◦C with 5% CO2 in a humidified incubator.
2.2. Rg3 Treatment and Transfection
5 × 104 cells were seeded in a 60 mm culture dish with 50% confluence and cultured
for 24 h before Rg3 treatment or transfection. The cells were then treated with 20 and
50 µM of Rg3 using a 20 mM Rg3 stock (LKT Labs, St. Paul, MN, USA) in 100% ethanol.
For transfection, siRNA (Bioneer, Daejon, Korea), mimic miR (Bioneer), and inhibitor miR
(Bioneer) were diluted to final concentrations of 20 and 40 nM in Opti-MEM Medium (In-
vitrogen, Carlsbad, CA, USA), mixed with 5 µL of Lipofectamine RNAiMAX (Invitrogen),
and added to the cell culture. For Rg3 and RNA cotreatments, RNA was processed follow-
ing the aforementioned transfection protocol, and, after 24 h, Rg3 was added. The cells
were further cultured for 24 h and then harvested using 0.05% trypsin-EDTA (Gibco BRL,
Carlsbad, CA, USA).
Biomolecules 2021, 11, 118
3 of 13
2.3. Rg3-Quantitative Reverse-Transcription Polymerase Chain Reaction (qRT-PCR)
Chromosomal DNA and total RNA were extracted from the 60 mm culture dishes
using the ZR-Duet DNA/RNA MiniPrep kit (Zymo Research, Irvine, CA, USA) and eluted
to 50 and 20 µL, respectively. MiR cDNA was synthesized from 1 µg of total RNA using a
miScript II RT kit (Qiagen, Valencia, CA, USA) in 20 µL reactions. qRT-PCR was conducted
with 3 µL cDNA per reaction using the miScript SYBR Green PCR kit (Qiagen) and miScript
Primer Assay kit (Qiagen). mRNA cDNA was synthesized from 2 µg of total RNA using
ReverTra Ace qPCR RT Master Mix (Toyobo, Osaka, Japan) in 10 µL reactions. PCR was
then conducted from 1 µL cDNA using SYBR Fast qPCR Kit Master Mix (Kapa Biosystems,
Wilmington, MA, USA). The expression of miR and mRNA samples was normalized to
that of U6 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH), respectively. PCR
was performed with an ABI 7300 instrument (Applied Biosystems, Foster City, CA, USA),
and the expression level was calculated following the 2−∆∆Ct method. Methylation-specific
PCR was performed with bisulfite-treated DNA, and the methylation level was calculated
by the 1/[1+2− (CTu−CTme)] × 100% method, as previously described [21]. PCR primers are
listed in Supplementary Table S1.
2.4. Data Mining
LncRNAs showing a significant methylation change by Rg3 were retrieved after
analyzing the methylation-array data of the NCBI GEO DataSet (GSE99505). LncBase
Predicted v.2 (http://diana.imis.athena-innovation.gr/DianaTools) and StarBase v3.0 (
http://starbase.sysu.edu.cn/index.php) were used to identify miRs that potentially interact
with ATXN8OS. MiR target genes were selected using five miR target-prediction programs:
MicroT (www.microrna.gr/microT-v4), RNA22 (https://cm.jefferson.edu/rna22), Tar-
getScan7 (http://www.targetscan.org/vert_72), miRWalk (http://http://mirwalk.umm.
uni-heidelberg.de), and miRmap (https://mirmap.ezlab.org).
2.5. Cell Proliferation and Apoptosis Assay
The effect of Rg3 and noncoding RNAs on cell growth was analyzed by a dye-based
cell-proliferation assay as previously described [22]. Briefly, 2 × 103 cells were seeded per
well on a 96-well plate and cultured for 24 h. Afterward, the cells were treated with either
Rg3 or noncoding RNA and cultured for up to six additional days. After an appropriate
culture period, the cells were stained with WST-8 using the Cell Counting Kit-8 (CCK-8)
(Enzo Biochem, New York, NY, USA) to measure cell density at OD450 using a spectropho-
tometer. For the apoptosis analysis, 1 × 106 cells were seeded in a 60 mm plate, treated
with Rg3 or transiently transfected with siRNA, and cultured for 24 h. After harvesting,
1 × 105 cells were suspended in a 1x binding buffer provided with the Annexin V-FITC
Apoptosis Detection kit II (BD Bioscience, San Jose, CA, USA), then stained with FITC
Annexin V(BD Bioscience) and PI (Sigma-Aldrich, St. Louis, MO, USA). Fluorescence was
detected with a BD Accuri C6 flow cytometer (BD Bioscience), and the data were analyzed
with the BD Accuri C6 software (BD Bioscience). Cell-cycle analysis was performed using
a flow cytometer as previously described [23]. The cell-proliferation index was calculated
using the following formula: proliferation index = (S+G2+M)/(G0/G1+S+G2+M) × 100
(%), where each letter represents the number of cells at each stage.
2.6. Western Blot Analysis
Proteins were extracted from the harvested cells using ice-cold RIPA lysis buffer
(Thermo Fisher Scientific, Waltham, MA, USA) with a 1% protease-inhibitor cocktail
(Thermo Fisher Scientific). The proteins (15 µg) were then subjected to SDS-PAGE, blotted
on a PVDF membrane (Sigma-Aldrich), and treated with primary antibodies overnight at 4
◦C. The blot was then incubated with HRP-conjugated antirabbit IgG antibodies (1:1000,
GTX213110-01; GeneTex, Irvine, CA, USA) for 2 h. The signals were visualized with the
ECL reagent (Abfrontier, Seoul, Korea), quantified using the Image Lab software (Bio-Rad,
Hercules, CA, USA), and normalized with β-actin. The antibodies used were anti-CHRM3
Biomolecules 2021, 11, 118
4 of 13
(1:1000, GTX111637; GeneTex), anti-DACH1 (1:1000, A303-556A-M; Bethyl, Montgomery,
TX, USA), and anti-β-actin (1:1000, bs-0061R; Bioss, Woburn, MA, USA).
2.7. Statistical Analyses
All experiments were independently conducted in triplicate, and the results were
expressed as the mean ± SD. Statistical analyses were performed using the SPSS 23.0
software (SPSS, Chicago, IL, USA). T-tests, originally created by Two-tailed Student, were
performed to analyze the qRT-PCR, Western blot, and apoptosis assay results. p-value <
0.05 was considered statistically significant.
3. Results
3.1. Rg3 Induces Hypermethylation and Downregulation of ATXN8OS
We previously performed a genome-wide methylation analysis of Rg3-treated MCF-7
breast-cancer cells [8]. In addition to 866,895 CpGs in protein-coding genes, the array
covered 10,733 CpGs in noncoding RNAs. Six lncRNAs exhibited significant methylation
changes in the promoter (i.e., |methylation level change (∆β)| > 1.5 and |methylation fold
change| > 1.4) (Figure 1A). Given that many lncRNAs have been linked to the development
of various cancer types, our study focused on their regulatory mechanisms. ATXN8OS was
selected for further study as it exhibited the highest methylation level change (∆β = 0.189).
Although little is known about its role in cancer development and progression, previous
studies indicate that ATXN8OS has oncogenic properties and therefore stimulates cancer-
cell growth [24].
The induction of hypermethylation at the ATXN8OS promoter by Rg3 was verified via
methylation-specific PCR in MCF-7 cells treated with 20 and 50 µM of Rg3. This experiment
resulted in a similar methylation change (methylation-fold change = 1.4 and ∆β = 1.5) to
that of the array-based analysis. Moreover, according to the qRT-PCR analysis, ATXN8OS
was downregulated by up to 76% in the Rg3-treated MCF-7 cells (Figure 1B). As Rg3 is
known to share a structural similarity with estrogen [25], regulation of ATXN8OS may
be affected by the estrogen-receptor (ER) status. To test this, the effect was examined
in an ER-negative breast-cancer cell line, MDA-MB-231, and in an ER-positive normal
cell line, MCF-10A. The result showed that expression of ATXN8OS was less affected in
MDA-MB-231 than in the other two cell lines (Supplementary Figure S1), possibly implying
an ER dependence on Rg3 for ATXN8OS regulation.
To address how ATXN8OS contributes to cancer-cell growth, its downregulation was
induced using two siRNAs (siATXN8OS#1 and #2) in MCF-7, which targeted different
sites of ATXN8OS (Supplementary Table S1, Supplementary Figure S2), after which cell
proliferation and apoptosis were monitored. It was found that ATXN8OS siRNA sup-
pressed cancer-cell growth by up to 18%, increased apoptosis by up to 5%, and decreased
the cell-proliferation index from 36.7% to 21.5% (Figure 1C–F; Supplementary Figure S3).
These results suggest that ATXN8OS promotes proliferation by stimulating the MCF-7
cancer-cell growth while also suppressing apoptosis.
Biomolecules 2021, 11, 118
5 of 13
Figure 1. ATXN8OS with proproliferation activity in the MCF-7 cells was downregulated by Rg3 via promoter methylation.
(A) ATXN8OS was among the six lncRNAs that exhibited significant changes in methylation level (|∆β| ≥ 0.15 and |fold
change| ≥ 1.4), as demonstrated by the analysis of an Rg3-treated MCF-7-cell methylation array. (B) MCF-7 cells were treated
with 20 and 50 µM of Rg3, and the methylation and expression of ATXN8OS were examined by methylation-specific PCR
and qRT-PCR, respectively. (C) ATXN8OS was downregulated in MCF-7 using siRNA, and its effect on cell proliferation was
examined in the presence of Rg3 using the CCK-8 assay. (D,E) The effect of ATXN8OS on apoptosis (D) and cell cycle (E) was
monitored using flow cytometry. All experiments were performed in triplicate, and the values are presented as the mean ± SE.
siNC, control siRNA (40 µM); siATXN8OS, ATXN8OS-specific siRNA (40 µM). * p < 0.05, ** p < 0.01, *** p < 0.001.
3.2. ATXN8OS Stimulates Cancer-Cell Proliferation via Sponging miR-424-5p
LncRNAs are known to often interact with and regulate miRs and act as ceRNA to
modulate the expression of miR target genes. Therefore, our study sought to identify
potential miRs for ATXN8OS. Three candidates were identified upon screening the LncBase
Biomolecules 2021, 11, 118
6 of 13
and StarBase public databases, which offer potential partner miRs for lncRNAs (Figure 2A).
MiR-424-5p was selected for further analysis as it showed the highest binding score. Rg3
treatment in MCF-7 cells induced the upregulation of the miR (Figure 2B). To see whether
ATXN8OS could regulate miR-424-5p, the expression of the miR was quantified via qRT-
PCR in MCF-7 cells treated with ATXN8OS-specific siRNA (siATXN8OS). Compared to
the scrambled siRNAs, siATXN8OS significantly increased the expression of miR-424-5p
(Figure 2C). The expression of ATXN8OS was then examined after deregulating miR-424-5p
using a mimic or an inhibitor RNA (Supplementary Figure S1). Interestingly, the miR-424-
5p mimic RNA downregulated ATXN8OS, whereas the inhibitor upregulated the lncRNA
(Figure 2D).
Figure 2. ATXN8OS and miR-424-5p sponge each other. (A) Three miRs that could potentially
bind ATXN8OS were screened in silico using two miR-prediction databases (LncBase Predicted v.3
and StarBase). (B) miR-424-5p exhibited the highest binding score and was therefore examined to
characterize its regulation by Rg3. MCF-7 cells were treated with Rg3, and the RNA expression was
quantified by qRT-PCR. (C,D) The association between the ATXN8OS and miR-424-5p expression
was monitored by examining the expression of each RNA after inhibiting ATXN8OS using siRNA (C)
and overexpressing (40 µM) or inhibiting miR-424-5p (20 µM) (D). All experiments were performed
in triplicate, and the values are presented as the mean ± SE. Testing was done using siNC, negative
control siRNA (40 µM); siATXN8OS, ATXN8OS-specific siRNA (40 µM); mimic NC, negative control
mimic for miR-424-5p (40 µM); and inhibitor NC, negative control inhibitor for miR-424-5p (20 µM).
** p < 0.01, *** p < 0.001.
Afterward, the effect of miR-424-5p on MCF-7 cell proliferation and apoptosis in the
presence of Rg3 was examined after deregulating miR-424-5p in combination with Rg3.
As shown in Figure 3A, cell growth was suppressed by 30% using the miR mimic alone,
and further decreased by Rg3 exposure in a dose-dependent manner. The miR mimic
Biomolecules 2021, 11, 118
7 of 13
increased apoptosis by 15% (Figure 3B). In contrast, the miR-424-5p inhibitor reversed the
effect of the mimic RNA by increasing cell growth while decreasing apoptosis of MCF-7
(Figure 3C,D). Therefore, we concluded that Rg3 inhibited the proproliferation effect of the
miR-424-5p inhibitor.
Figure 3. MiR-424-5p inhibited MCF-7 cell proliferation. MiR-424-5p was deregulated in MCF-7 by transiently transfecting the
cells with a mimic (A,B) or an inhibitor (C,D), after which cell proliferation and apoptosis were assessed with the CCK-8 assay
and flow-cytometry analysis. Rg3 was coadministered with the mimic (40 µM) or inhibitor (20 µM) for the proliferation assay.
Testing was done using mimic NC, negative control miR-424-5p mimic (40 µM) and inhibitor NC, negative control inhibitor for
miR-424-5p (20 µM). All experiments were performed in triplicate, and the results are presented as the mean ± SE. Representative
images are shown for flow-cytometry analysis. * p < 0.05, ** p < 0.01, *** p < 0.001.
3.3. MiR-424-5p Target Genes are Regulated by ATXN8OS
Given the regulatory effect of miRs on target genes, we sought to determine whether
ATXN8OS also affects target-gene expression. Potential targets were first identified using
the five target-gene prediction algorithms described in the Materials and Methods, which
rendered 200 candidate genes according to all five prediction tools (Figure 4A). To nar-
row down the number of target genes, the pool was then further filtered by applying
genome-wide methylation-array data, which were obtained from the Rg3-treated MCF-7
Biomolecules 2021, 11, 118
8 of 13
cells (GSE99505). We aimed to identify target genes that were controlled by miR-424-5p and
subject to promoter methylation by Rg3. Through this double-filtering approach, seven
genes were identified, satisfying both the target-gene prediction and the methylation crite-
ria (|∆β| > 1.5) (Figure 4B). Specifically, our study focused on EYA1, CHRM3, and DACH1
because they had a target sequence for miR-424-5p (Figure 4C) and showed hypermethy-
lation in the array data, suggesting that they were downregulated by Rg3. Additionally,
these three genes had previously been reported to possess oncogenic properties in several
cancer types [26,27], except DACH1, which functioned as either a tumor promoter [28]
or suppressor [29] depending on the cancer type. Consistent with the hypermethylation
status, EYA1, CHRM3, and DACH1 were downregulated by 39–95% by Rg3, as determined
by our qRT-PCR assays (Figure 4D). ATXN8OS inhibition resulted in downregulation of
all the target genes (Figure 4E). Moreover, the miR-424-5p mimic downregulated the three
target genes, whereas an inhibitor upregulated them (Figure 4F,G).
Figure 4. Regulation of miR-424-5p target genes by Rg3 and ATXN8OS. Potential miR-424-5p target genes were identified by
analyzing five public databases (miRmap, miRWalk, TargetScan, MicroT, and RNA22) (A), after which they were compared with
the methylation-array data of the Rg3-treated MCF-7 cells (GSE99505) (B). (C) Potential binding sequence of the target genes on
miR-424-5p. The seed sequence is denoted in bold. (D–G) Effect of Rg3, ATXN8OS, and miR-424-5p on miR-424-5p target-gene
expression. Gene expression was examined by qRT-PCR for samples treated with Rg3 (D), ATXN8OS-specific siRNA (40 µM) (E),
miR-424-5p mimic (40 µM) (F), and a miR-424-5p inhibitor (20 µM) (G). Testing was done using siNC, control siRNA (40 µM);
mimic NC, negative control mimic for miR-424-5p (40 µM); and inhibitor NC, negative control inhibitor for miR-424-5p (20 µM).
All experiments were performed in triplicate, and the results are presented as the mean ± SE. ** p < 0.01, *** p < 0.001.
Biomolecules 2021, 11, 118
9 of 13
The protein expression of DACH1 and CHRM3 was then examined by Western blot
analysis. DACH1 and CHRM3 protein-expression exhibited a similar profile to that of
the transcripts. Specifically, protein expression was downregulated by Rg3, siATXN8OS,
and a miR-424-5p mimic RNA but upregulated by the miR-424-5p inhibitor (Figure 5 and
Supplementary Figure S2). The EYA1 protein was barely detected in MCF-7 as in a previous
study [30]. Therefore, further confirmation of the effect of Rg3 and noncoding RNAs at
the protein level was deemed unnecessary. Overall, Rg3 downregulated EYA1, DACH1,
and CHRM3 via the Rg3/ATXN8OS/miR-424-5p axis, whereas ATXN8OS inhibited the
miR to modulate the expression of the target gene (Figure 6).
Figure 5. Effect of Rg3, ATXN8OS, and miR-424-5p on the target genes of miR-424-5p at the protein
level. Western blot analysis of CHRM3 and DACH1 was performed after treating the MCF-7 cells
with Rg3 (A) or deregulating ATXN8OS (40 µM siRNA) and miR-424-5p (40 µM for mimic and
inhibitor) (B,C). Testing was done using siNC, control siRNA (40 µM); mimic NC, negative control
mimic for miR-424-5p (40 µM); and inhibitor NC, negative control inhibitor for miR-424-5p (20 µM).
The band intensity was measured with the Image Lab software and indicated by bar graphs. * p <
0.05, ** p < 0.01, *** p < 0.001.
Biomolecules 2021, 11, 118
10 of 13
Figure 6. Schematic of the Rg3/ATXN8OS/miR-424-5p axis regulation process. ATXN8OS downreg-
ulates the tumor-suppressive miR-424-5p, which in turn activates oncogenic CHRM3 and DACH1,
leading to cancer-cell proliferation. Rg3 blocks the oncogenic activity of ATXN8OS by inducing
promoter hypermethylation.
4. Discussion
Our study aimed to identify lncRNAs that are dysregulated in Rg3-treated cancer cells
to elucidate the mechanisms by which they control cancer-cell proliferation, with a particu-
lar focus on ceRNA-miR interaction. Most studies on ATXN8OS have so far examined the
genetic expansion of CAG repeats. For instance, spinocerebellar ataxia type 8 (SCA8), an
autosomal dominant neurodegenerative disease, is caused by CTA/CTG repeat expansion
in the ATXN8OS gene [31]. In contrast, little is known about the role of ATXN8OS in
tumor development. Recently, Deng et al. found that ATXN8OS stimulated the prolif-
eration and migration of MCF-7 and MDA-MB-231 breast-cancer cells [24]. Specifically,
the authors reported that ATXN8OS sequestered the tumor-suppressive miR-204. However,
the mechanisms by which miR-204 is regulated by Rg3 remain to be determined. Our study
revealed that the oncogenic ATXN8OS is epigenetically regulated by Rg3 via promoter
methylation. A few other lncRNAs also showed methylation level changes: DOCK4-AS1,
LINC00911, and RFX3-AS1 were hypermethylated, whereas STXBP5-AS1 and LINC01477
were hypomethylated. Notably, LINC00911 and RFX3-AS1 are known as oncogenes [17,32],
whereas STXBP5-AS1 is known as a tumor suppressor [33]. These findings suggest that the
tumor-suppressive activity of Rg3 could be attributed in part to its epigenetic regulation of
tumor-related lncRNAs. However, the mechanisms by which ATXN8OS methylation is
controlled by Rg3 remain to be determined. Moreover, although a close association was
identified between gene methylation and expression levels, additional studies are required
to determine whether inducing hypermethylation could drive gene downregulation.
MiR-424-5p has been shown to reduce cell viability by modulating the PTEN/PI3K/AKT
/mTOR pathway in breast-cancer cells [34], the MAPK pathway in ischemic stroke [35],
and the Hippo-signaling pathway in thyroid cancer [36]. MiR-424-5p target genes have been
Biomolecules 2021, 11, 118
11 of 13
identified in various cancer types, including PD-L1 [34], VEGFA [37], and ARK5 [38]. These
target genes generally exert a protumor activity by promoting proliferation, migration, or
angiogenesis in cancer cells. A few lncRNAs have been found to regulate miR-424-5p in
various cancer cells, including LINC00922 in breast cancer [39], CDNK2B-AS1 in hepatocel-
lular carcinoma [40], and XIST in neuroendocrine tumors [41]. In all the aforementioned
cases, regulation of miR-424-5p by the corresponding lncRNA resulted in cell proliferation or
cancer-progression alterations.
Limited cases of ceRNA have been identified in ginsenosides. However, there are
reports of an Rg3-regulated lncRNA H19 that sponges miR-324-5p to enhance PKM2 expres-
sion by directly binding the miR [42]. In another study, Rg1 inhibited high glucose-induced
mesenchymal activation by downregulating lncRNA RP11-982M15.8 but upregulating
miR-2133 to decrease Zeb1 [43]. The current study suggests a novel ceRNA relationship
between the Rg3-regulated ATXN8OS and miR-424-5p, which is supported by the follow-
ing findings: First, the expression of miR-424-5p increased after ATXN8OS was inhibited
and vice versa. The lncRNA-induced miR regulation may increase through binding sites
with special sequences or paring topology, which would trigger miR degradation upon
binding [44]. Second, the two noncoding RNAs had opposite effects on the target-gene
expression and the MCF-7 cell growth. Nonetheless, the mechanical interaction between
the two RNAs should be elucidated to confirm the proposed ceRNA relationship.
Our study had a few noteworthy limitations. Particularly, all of our findings were
based on the analysis of a single lncRNA. Therefore, data on lncRNAs other than ATXN8OS
should be obtained to comprehensively explore how Rg3-regulated lncRNAs affect cancer-
cell survival or proliferation. Additionally, further studies on other lncRNAs identified
herein such as RFX3-AS1, DOCK4-AS1, and STXBP5-AS1 could provide useful insights.
5. Conclusions
ATXN8OS was identified as a lncRNA that can be downregulated via promoter hy-
permethylation by Rg3 in MCF-7 cancer cells. Moreover, ATXN8OS was found to induce
the proliferation of cancer cells and this was suppressed by Rg3. At the molecular level,
ATXN8OS sponged a tumor-suppressive miR-424-5p, thereby activating key oncogenes
such as EYA1, DACH1, and CHRM3, which could be suppressed by Rg3 treatment. There-
fore, our findings suggest that Rg3 suppresses MCF-7 cancer-cell proliferation but increases
apoptosis by modulating the ATXN8OS/miR-424-5p/target-gene axis.
Supplementary Materials: The following are available online at https://www.mdpi.com/2218-273
X/11/1/118/s1. Table S1: PCR primers, siRNA, miR-mimic, and miR-inhibitor used in this study,
Figure S1: Regulation of ATXN8OS and miR-424-5p by Rg3 in mammary gland-derived cell lines,
Figure S2: Induction of deregulation of ATXN8OS and miR-424-5p in MCF-7, Figure S3: Effect of
ATXN8OS on apoptosis, cell growth, and cell cycle, Figure S4: Uncropped Western blots.
Author Contributions: Conceptualization, S.J.K.; methodology, H.K. and H.W.J.; validation, H.W.K.,
S.H.Y., and J.E.P.; data curation, H.K. and H.W.J.; writing—original draft preparation, H.K., H.W.J.,
and S.J.K.; writing—review and editing, H.K. and S.J.K.; funding acquisition, S.J.K. All authors have
read and agreed to the published version of the manuscript.
Funding: This study was supported by the Basic Science Research Program (NRF-2016R1D1A1B01009235)
of the National Research Foundation of Korea funded by the Ministry of Education, Science, and Tech-
nology.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: All data are contained within the article or supplementary material.
Conflicts of Interest: The authors declare no conflict of interest.
Biomolecules 2021, 11, 118
12 of 13
References
1.
2.
3.
4.
Nakhjavani, M.; Hardingham, J.E.; Palethorpe, H.M.; Tomita, Y.; Smith, E.; Price, T.J.; Townsend, A.R. Ginsenoside Rg3: Potential
molecular targets and therapeutic indication in metastatic breast cancer. Medicines 2019, 6, 17. [CrossRef]
Sun, M.; Ye, Y.; Xiao, L.; Duan, X.; Zhang, Y.; Zhang, H. Anticancer effects of ginsenoside Rg3 (Review). Int. J. Mol. Med. 2017, 39,
507–518. [CrossRef] [PubMed]
Kim, B.M.; Kim, D.H.; Park, J.H.; Na, H.K.; Surh, Y.J. Ginsenoside Rg3 Induces Apoptosis of Human Breast Cancer (MDA-MB-231)
Cells. Eur. J. Cancer Prev. 2013, 18, 177–185. [CrossRef] [PubMed]
Kim, B.-M.; Kim, D.-H.; Park, J.-H.; Surh, Y.-J.; Na, H.-K. Ginsenoside Rg3 inhibits constitutive activation of NF-κB signaling
in human breast cancer (MDA-MB-231) cells: ERK and Akt as potential upstream targets. Eur. J. Cancer Prev. 2014, 19, 23–30.
[CrossRef] [PubMed]
6.
5. Mao, X.; Jin, Y.; Feng, T.; Wang, H.; Liu, D.; Zhou, Z.; Yan, Q.; Yang, H.; Yang, J.; Yang, J.; et al. Ginsenoside Rg3 Inhibits the
Growth of Osteosarcoma and Attenuates Metastasis through the Wnt/β-Catenin and EMT Signaling Pathway. Evid. Based
Complement Altern. Med. 2020, 2020, 6065124. [CrossRef] [PubMed]
Li, B.; Qu, G. Inhibition of the hypoxia-induced factor-1α and vascular endothelial growth factor expression through ginsenoside
Rg3 in human gastric cancer cells. J. Cancer Res. Ther. 2019, 15, 1642–1646. [CrossRef]
Phi, L.T.H.; Wijaya, Y.T.; Sari, I.N.; Kim, K.S.; Yang, Y.G.; Lee, M.W.; Kwon, H.Y. 20(R)-Ginsenoside Rg3 Influences Cancer Stem
Cell Properties and the Epithelial-Mesenchymal Transition in Colorectal Cancer via the SNAIL Signaling Axis. Onco Targets Ther.
2019, 12, 10885–10895. [CrossRef]
7.
8. Ham, J.; Lee, S.; Lee, H.; Jeong, D.; Park, S.; Kim, S.J. Genome-Wide Methylation Analysis Identifies NOX4 and KDM5A as Key
Regulators in Inhibiting Breast Cancer Cell Proliferation by Ginsenoside Rg3. Am. J. Chin. Med. 2018, 46, 1333–1355. [CrossRef]
Teng, S.; Wang, Y.; Li, P.; Liu, J.; Wei, A.; Wang, H.; Meng, X.; Pan, D.; Zhang, X. Effects of R type and S type ginsenoside Rg3 on
DNA methylation in human hepatocarcinoma cells. Mol. Med. Rep. 2017, 15, 2029–2038. [CrossRef]
9.
10. Cheng, Z.; Xing, D. Ginsenoside Rg3 inhibits growth and epithelial-mesenchymal transition of human oral squamous carcinoma
cells by down-regulating miR-221. Eur. J. Pharmacol. 2019, 853, 353–363. [CrossRef]
11. Lee, A.; Yun, E.; Chang, W.; Kim, J. Ginsenoside Rg3 protects against iE-DAP–induced endothelial-to-mesenchymal transition by
regulating the miR-139-5p–NF-κB axis. J. Ginseng Res. 2020, 44, 300–307. [CrossRef] [PubMed]
12. Li, J.; Lu, J.; Ye, Z.; Han, X.; Zheng, X.; Hou, H.; Chen, W.; Li, X.; Zhao, L. 20(S)-Rg3 blocked epithelial-mesenchymal transition
through DNMT3A/miR-145/FSCN1 in ovarian cancer. Oncotarget 2017, 8, 53375–53386. [CrossRef] [PubMed]
13. Lu, J.; Chen, H.; He, F.; You, Y.; Feng, Z.; Chen, W.; Li, X.; Zhao, L. Ginsenoside 20(S)-Rg3 upregulates HIF-1α-targeting
miR-519a-5p to inhibit the Warburg effect in ovarian cancer cells. Clin. Exp. Pharmacol. Physiol. 2020, 47, 1455–1463. [CrossRef]
[PubMed]
14. Wu, K.; Huang, J.; Xu, T.; Ye, Z.; Jin, F.; Li, N.; Lv, B. MicroRNA-181b blocks gensenoside Rg3-mediated tumor suppression of
gallbladder carcinoma by promoting autophagy flux via CREBRF/CREB3 pathway. Am. J. Transl. Res. 2019, 11, 5776–5787.
Schmitt, A.M.; Chang, H.Y. Long Noncoding RNAs in Cancer Pathways. Cancer Cell. 2016, 29, 452–463. [CrossRef] [PubMed]
15.
16. Zou, J.; Su, H.; Zou, C.; Liang, X.; Fei, Z. Ginsenoside Rg3 suppresses the growth of gemcitabine-resistant pancreatic cancer cells
by upregulating lncRNA-CASC2 and activating PTEN signaling. J. Biochem. Mol. Toxicol. 2020, 34, e22480. [CrossRef]
17. Ham, J.; Jeong, D.; Park, S.; Kim, H.W.; Kim, H.; Kim, S.J. Ginsenoside Rg3 and Korean Red Ginseng extract epigenetically
regulate the tumor-related long noncoding RNAs RFX3-AS1 and STXBP5-AS1. J. Ginseng Res. 2019, 43, 625–634. [CrossRef]
18. Li, J.; Qi, Y. Ginsenoside Rg3 inhibits cell growth, migration and invasion in Caco-2 cells by downregulation of lncRNA CCAT1.
Exp. Mol. Pathol. 2019, 106, 131–138. [CrossRef]
19. Tay, Y.; Rinn, J.; Pandolfi, P.P. The multilayered complexity of ceRNA crosstalk and competition. Nature 2014, 505, 344–352.
[CrossRef]
20. Yan, L.; Yang, S.; Yue, C.X.; Wei, X.Y.; Peng, W.; Dong, Z.Y.; Xu, H.N.; Chen, S.L.; Wang, W.R.; Chen, C.J.; et al. Long noncoding
RNA H19 acts as a miR-340-3p sponge to promote epithelial-mesenchymal transition by regulating YWHAZ expression in
paclitaxel-resistant breast cancer cells. Environ. Toxicol. 2020, 35, 1015–1028. [CrossRef]
21. Kim, S.J.; Kelly, W.K.; Fu, A.; Haines, K.; Hoffman, A.; Zheng, T.; Zhu, Y. Genome-wide methylation analysis identifies
involvement of TNF-α mediated cancer pathways in prostate cancer. Cancer Lett. 2011, 302, 47–53. [CrossRef] [PubMed]
22. Lee, S.; Lee, H.; Bae, H.; Choi, E.H.; Kim, S.J. Epigenetic silencing of miR-19a-3p by cold atmospheric plasma contributes to
proliferation inhibition of the MCF-7 breast cancer cell. Sci. Rep. 2016, 6, 30005. [CrossRef] [PubMed]
23. Kang, S.; Kim, B.; Kang, H.S.; Jeong, G.; Bae, H.; Lee, H.; Lee, S.; Kim, S.J. SCTR regulates cell cycle-related genes toward
anti-proliferation in normal breast cells while having pro-proliferation activity in breast cancer cells. Int. J. Oncol. 2015, 47,
1923–1931. [CrossRef] [PubMed]
24. Deng, Z.; Cai, H.; Lin, L.; Zhu, L.; Wu, W.; Yang, S.; Cai, J.; Tan, J. lncRNA ATXN8OS promotes breast cancer by sequestering
miR-204. Mol. Med. Rep. 2019, 20, 1057–1064. [CrossRef]
25. Tian, M.; Li, L.N.; Zheng, R.R.; Yang, L.; Wang, Z.T. Advances on hormone-like activity of Panax ginseng and ginsenosides. Chin.
J. Nat. Med. 2020, 18, 526–535. [CrossRef]
26. Cai, S.; Cheng, X.; Liu, Y.; Lin, Z.; Zeng, W.; Yang, C.; Liu, L.; Chukwuebuka, O.A.; Li, W. EYA1 promotes tumor angiogenesis by
activating the PI3K pathway in colorectal cancer. Exp. Cell Res. 2018, 367, 37–46. [CrossRef]
Biomolecules 2021, 11, 118
13 of 13
27. Wang, N.; Yao, M.; Xu, J.; Quan, Y.; Zhang, K.; Yang, R.; Gao, W.Q. Autocrine Activation of CHRM3 Promotes Prostate Cancer
Growth and Castration Resistance via CaM/CaMKK-Mediated Phosphorylation of Akt. Clin. Cancer Res. 2015, 21, 4676–4685.
[CrossRef]
28. Hu, X.; Zhang, L.; Li, Y.; Ma, X.; Dai, W.; Gao, X.; Rao, X.; Fu, G.; Wang, R.; Pan, M.; et al. Organoid modelling identifies that
DACH1 functions as a tumour promoter in colorectal cancer by modulating BMP signalling. EBioMedicine 2020, 56, 102800.
[CrossRef]
29. Xu, H.; Yu, S.; Yuan, X.; Xiong, J.; Kuang, D.; Pestell, R.G.; Wu, K. DACH1 suppresses breast cancer as a negative regulator of
CD44. Sci. Rep. 2017, 7, 4361. [CrossRef]
30. Wu, K.; Li, Z.; Cai, S.; Tian, L.; Chen, K.; Wang, J.; Hu, J.; Sun, Y.; Li, X.; Ertel, A.; et al. EYA1 phosphatase function is essential to
31.
drive breast cancer cell proliferation through cyclin D1. Cancer Res. 2013, 73, 4488–4499. [CrossRef]
Samukawa, M.; Hirano, M.; Saigoh, K.; Kawai, S.; Hamada, Y.; Takahashi, D.; Nakamura, Y.; Kusunoki, S. PSP-Phenotype in
SCA8: Case Report and Systemic Review. Cerebellum 2019, 18, 76–84. [CrossRef] [PubMed]
32. Li, X.X.; Wang, L.J.; Hou, J.; Liu, H.Y.; Wang, R.; Wang, C.; Xie, W.H. Identification of Long Noncoding RNAs as Predictors of
Survival in Triple-Negative Breast Cancer Based on Network Analysis. BioMed Res. Int. 2020, 2020, 8970340. [CrossRef] [PubMed]
33. Cen, D.; Huang, H.; Yang, L.; Guo, K.; Zhang, J. Long noncoding RNA STXBP5-AS1 inhibits cell proliferation, migration,
and invasion through inhibiting the PI3K/AKT signaling pathway in gastric cancer cells. Onco Targets Ther. 2019, 12, 1929–1936.
[CrossRef]
34. Dastmalchi, N.; Hosseinpourfeizi, M.A.; Khojasteh, S.M.B.; Baradaran, B.; Safaralizadeh, R. Tumor suppressive activity of
miR-424-5p in breast cancer cells through targeting PD-L1 and modulating PTEN/PI3K/AKT/mTOR signaling pathway. Life Sci.
2020, 259, 118239. [CrossRef] [PubMed]
35. Xiang, Y.; Zhang, Y.; Xia, Y.; Zhao, H.; Liu, A.; Chen, Y. LncRNA MEG3 targeting miR-424-5p via MAPK signaling pathway
mediates neuronal apoptosis in ischemic stroke. Aging (Albany N.Y.). 2020, 12, 3156–3174. [CrossRef]
36. Liu, X.; Fu, Y.; Zhang, G.; Zhang, D.; Liang, N.; Li, F.; Li, C.; Sui, C.; Jiang, J.; Lu, H.; et al. miR-424-5p Promotes Anoikis Resistance
and Lung Metastasis by Inactivating Hippo Signaling in Thyroid Cancer. Mol. Ther. Oncolytics. 2019, 15, 248–260. [CrossRef]
37. Vimalraj, S.; Saravanan, S.; Raghunandhakumar, S.; Anuradha, D. Melatonin regulates tumor angiogenesis via miR-424-
5p/VEGFA signaling pathway in osteosarcoma. Life Sci. 2020, 256, 118011. [CrossRef]
38. Wu, J.; Yang, B.; Zhang, Y.; Feng, X.; He, B.; Xie, H.; Zhou, L.; Wu, J.; Zheng, S. miR-424-5p represses the metastasis and invasion
of intrahepatic cholangiocarcinoma by targeting ARK5. Int. J. Biol. Sci. 2019, 15, 1591–1599. [CrossRef]
39. Yue, X.; Wang, Z. Long Intergenic Non-Coding RNA LINC00922 Aggravates the Malignant Phenotype of Breast Cancer by
40.
Regulating the microRNA-424-5p/BDNF Axis. Cancer Manag. Res. 2020, 12, 7539–7552. [CrossRef]
Shen, X.; Li, Y.; He, F.; Kong, J. LncRNA CDKN2B-AS1 Promotes Cell Viability, Migration, and Invasion of Hepatocellular
Carcinoma via Sponging miR-424-5p. Cancer Manag. Res. 2020, 12, 6807–6819. [CrossRef]
41. Zhou, K.; Li, S.; Du, G.; Fan, Y.; Wu, P.; Sun, H.; Zhang, T. LncRNA XIST depletion prevents cancer progression in invasive
pituitary neuroendocrine tumor by inhibiting bFGF via upregulation of microRNA-424-5p. Onco Targets Ther. 2019, 12, 7095–7109.
[CrossRef] [PubMed]
42. Zheng, X.; Zhou, Y.; Chen, W.; Chen, L.; Lu, J.; He, F.; Li, X.; Zhao, L. Ginsenoside 20(S)-Rg3 Prevents PKM2-Targeting miR-324-5p
from H19 Sponging to Antagonize the Warburg Effect in Ovarian Cancer Cells. Cell. Physiol. Biochem. 2018, 51, 1340–1353.
[CrossRef] [PubMed]
43. Xue, L.P.; Fu, X.L.; Hu, M.; Zhang, L.W.; Li, Y.D.; Peng, Y.L.; Ding, P. Rg1 inhibits high glucose-induced mesenchymal activation
and fibrosis via regulating miR-2113/RP11-982M15.8/Zeb1 pathway. Biochem. Biophys. Res. Commun. 2018, 501, 827–832.
[CrossRef] [PubMed]
Figliuzzi, M.; Marinari, E.; De Martino, A. MicroRNAs as a selective channel of communication between competing RNAs: A
steady-state theory. Biophys. J. 2013, 104, 1203–1213. [CrossRef] [PubMed]
44.
| null |
10.1088_1402-4896_ad1320.pdf
|
Data availability statement
All data used are available in the manuscript. The data that support the findings of this study are available upon
reasonable request from the authors.
|
Data availability statement All data used are available in the manuscript. The data that support the findings of this study are available upon reasonable request from the authors.
|
Phys. Scr. 99 (2024) 015920
https://doi.org/10.1088/1402-4896/ad1320
PAPER
RECEIVED
31 October 2023
REVISED
26 November 2023
ACCEPTED FOR PUBLICATION
6 December 2023
PUBLISHED
15 December 2023
CO2 capture and storage by metal and non-metal decorated silicon
carbide nanotubes: a DFT study
Yahaya Saadu Itas1,∗
, Razif Razali2, Sultan Alamri3, Hamid Osman3 and Mayeen Uddin Khandaker4,5,∗
1 Department of Physics, Bauchi State University Gadau, PMB 65, Gadau, Bauchi, Nigeria
2 Department of Physics Faculty of Science, Universiti Teknologi - Johor, Malaysia
3 Radiological Sciences Department, College of Applied Medical Sciences, Taif University, 21944 Taif, Saudi Arabia
4 Applied Physics and Radiation Technologies Group, CCDCU, School of Engineering and Technology, Sunway University, Bandar
Sunway 47500, Selangor, Malaysia
5 Faculty of Graduate Studies, Daffodil International University, Daffodil Smart City, Birulia, Savar, Dhaka—1216, Bangladesh
∗ Authors to whom any correspondence should be addressed.
E-mail: [email protected] and [email protected]
Keywords: CO2 capture, physisorption, silicon carbide nanotubes, CO2 adsorption, CO2 storage
Abstract
This study addressed the nano-mechanism of CO2 capture by Al-doped, B-doped and N-doped
single-walled silicon carbide nanotubes (SWSiCNTs) using the prominent density functional theory.
The results showed absolute interactions between CO2 and B- and N- impurity atoms of the
SWSiCNT surface with the highest adsorption energy of −1.85 eV and −1.83 eV respectively. Analysis
of the binding energy of CO2 to Al-doped SWSiCNT revealed that chemisorption between them is
stronger than B-doped and N-doped SWSiCNTs. Results from optical adsorption spectra revealed
that both B-and N-doped systems adsorb CO2 in the visible region of the electromagnetic spectrum
while B-doped SiCNT shows the highest adsorption. This study recommends B- and N-doped SiCNTs
as candidates for CO2 capture and storage with higher efficiency by B-doped SiCNT, while the
performance of the Al-doped system was underscored.
1. Introduction
The level of carbon dioxide (CO2) in the atmosphere is much higher than it was over 800,000 years ago and is
expected to increase further in the coming decades. Furthermore, the only way to stop the daily increase in world
temperature of 2 degrees Celsius is to dramatically cut greenhouse gas emissions like CO2 and methane, etc
Because of this, this issue has prompted policymakers to shift their focus to a growing field called Carbon
Capture, Usage and Storage (CCUS), which ultimately addresses any effort to extract CO2 from the ecosystem
and consequently eliminates environmental impact. The advent of the industrial revolution have witnessed a
global rise in CO2 levels which increased the atmospheric concentration of CO2 to 47% [1]. Due to this, the issue
of environmental climate change has increased significantly. Another factor which increased too much emission
of CO2 is the burning and over utilization of fossil fuels. For example, the Intergovernmental Panel on Climate
Change in its 5th assessment report has attributed 65% of the CO2 emissions due to fossil fuel combustion alone
in 2010. Therefore, it is very necessary to curtail CO2 emissions in order to save the ecosystem from the tragic
impacts of global warming and climate change. In an attempt to combat the growing CO2 levels in our
ecosystem, various methods have been adopted such as limiting the industrial emission of CO2 and decreasing
the use of fossil fuels. One of the highly recommended most efficient technologies to reduce CO2 in the
atmosphere is CO2 capture, utilization and sequestration (CCUS) which is reported to reduce the CO2 emission
by 45% [2]. Other methods include adsorption, absorption and membrane separation technologies. Many of
these researches have led to the successful capture of CO2 from the atmosphere and then stored in deep
geological formations or converted to other forms such as biofuels. CO2 reduction has been widely studied by
photocatalysis and several semiconductors have been reported as very active in CO2 capture [3], although most
of them were found with poor efficiency and low product selectivity. For materials to successfully reduce CO2,
© 2023 IOP Publishing Ltd
Phys. Scr. 99 (2024) 015920
Y S Itas et al
they have to be soluble, porous, thermally stable, capable of absorbing within visible light and should have
semiconducting behavior. In a research on the evaluation of 1,2,4-Triazolium-based ionic liquids for CO2
capture, the reported results revealed more solubility of energy levels dominated by anions than ions [4], which
showed more tendency to accept CO2 than donate. Experimental studies were conducted to investigate CO2
capture by multi-walled carbon nanotubes [5]. The obtained results revealed that MWCNT was able to adsorb
CO2 up to 200 °C. However, this process failed to report the amount of CO2 absorbed under visible light.
Furthermore, to the best of our survey revealed that the majority of the nanotubes such as MgO nanotubes, BN
nanotubes, SiC nanotubes and ZnO nanotubes were not investigated as candidates for CO2 capture. Meanwhile,
the CO2 absorption of two-dimensional and doped forms of these compounds was rarely reported [6, 7].
In the research on the potentials of metal-doped SiCNT for CO2 capture, the Pb, Cu and Ti decorated SiCNT
showed high chemisorption with Pb and Cu and low physisorption with Ti which underscored its performance
for CO2 capture [8]. Additionally, the ideal overpotential for CO2 reduction is 1.8–2 eV and the majority of the
photocatalysts were found to have a band gap below this range [9]. To bridge many of these gaps, this work
investigated the structural, electronic and optical properties of both metal and nonmetal-doped single-walled
silicon carbide nanotubes (SWSiCNTs) as potential candidates for CO2 capture. The SWSiCNT was chosen as a
representative model for this study because previous reports revealed its CO2 capture inaction under metallic
(Cu, Pb and Ti) doping. Additionally, SiC-based materials have been considered as feasible solvents for the
removal of CO2 and HCl. They have been used in various optoelectronic fields such as photocatalysis, solar cells,
composites for wastewater treatments, hydrogen evolution and so on. SiCNTs have advantages over CNTs in
optoelectronic applications due to their high reactivity of the exterior surface facilitating sidewall decoration,
better hydrogen storage performance and extreme sensitivity to some gaseous molecules [10].
2. Research method
Geometry optimizations were performed in such a way that the Si-C bond length 1.78 Å [11] in SWSiCNT is
consistent with previous studies. Doping to the optimized SiC nanotubes was done by replacing one Si atom
with Al, B, or N atoms with a 3.6% impurity concentration. To ensure accurate determination of the electronic
properties of the investigated systems, we have performed all calculations using the generalized gradient
approximation (GGA) method in terms of Perdew–Burke–Ernzerhof (PBE) exchange functional [12]. Quasi-
particle energies were determined using GW approximation implemented in Yambo codes [13]. A sampling of
the reciprocal lattice was done using 1 × 1 × 60 grids obtained via the Monkhorst−Pack method while an energy
cut-off value of 60 Ry was used. CO2 capture properties of all the doped SWSiCNTs were investigated by
calculating the electronic band’s structure, density of states, adsorption energies and optical properties. Analysis
of the nanotube’s response to the incident CO2 adsorption was determined using the imaginary dielectric
constant which gives an account of the CO2 energy adsorbed [14]. We have determined the formation energy of
Si—C using the following relation,
E
f
=
[
E SiC
T
]
-
[
E Si
T
]
-
[
E C
r
]
(
)
?1
[
E SiCT
where
free C atom.
]
is the total energy of SiCNT,
[
E Si
T
]
is the total energy of free Si atom and
[
E Cr
]
is the energy of
3. Results and discussion
3.1. Structural properties
As reported in [15], doping significantly changes the structural stability of electronic systems depending on the
concentration of dopants. To get a good understanding of the efficiency of the doped SiC nanotubes under
different dopants, it is significant to analyze the structural properties as well as the effects of formation energies.
We have used one unit cell of (7, 7) SWSiCNT which contained 14 atoms each of Si and C respectively. The
system was then doped separately with 3.6% of Al, B and N impurities respectively. After doping, it was observed
that the SiC nanotube was not stable when these dopants were used to replace C atoms; however, the nanotubes
became very stable when the Si atom was replaced by Al, B or N dopants. Additionally, Si atoms were displaced
towards the tube axis while C atoms were displaced in the opposite direction consistent with previous reports
[16]. Figure 1 presents the optimized structures of the CO2-adsorbed SWSiCNT systems with different dopants.
The calculated formation energies of the systems can be found in table 1 using different methods including the
tight binding method. The tight binding (TB) approach was used to determine the formation energies of the
systems before (fixed) and after cell relaxation (relaxed). As can be seen, formation energies were brought to the
appropriate lowest energy to ensure zero unwanted electronic interactions. The obtained results agreed with
the calculated GGA values which can be found in table 1. The lowest value of formation energy (4.01 eV) by
2
Phys. Scr. 99 (2024) 015920
Y S Itas et al
Figure 1. Optimized geometric structures of (a) CO2@Al-doped SWSiCNT, (b) CO2@B-doped t SWSiCNT and (c) CO2@N-doped
SWSiCNT materials. Green balls represent Si atoms, red balls represent C atoms, blue ball represent Al atom, yellow balls represent O
atoms and black ball represent B atom.
Table 1. Calculated formation energies of the optimized doped SWSiCNT materials.
Tight binding (TB)
Material
Fixed (eV)
Relaxed (eV)
Al-doped SWSiCNT
B-doped SWSiCNT
N-doped SWSiCNT
4.85
4.88
4.91
4.01
4.21
4.24
GGA (eV)
4.01
4.19
4.23
Exp. (eV)
4.14
4.20
4.24
Al-doped SWSiCNT signified weak stability by the nanotube. Atoms in the Al-doped SiC nanotube can easily
vibrate and enhance high carrier mobility. In terms of B-doped and N-doped SiC nanotubes, higher values of the
formation energies mean that a higher amount of energy is required to break the bonds between atoms of B, Si,
and C elements. Therefore, these nanotubes are more structurally stable than the metal-doped (Al) SWSiCNT.
Moreover, the calculated values of the formation energies of all three investigated systems were in good
agreement with previous experimental data [17]. The calculated values of bond lengths, bond angles, adsorption
energy values and charge transfer values were presented in Table 2 which all agreed well with the experimental
data for CO2 adsorption [18]. Based on the obtained results, the lower value of the adsorption energy (−1.42 eV)
by the Al-doped SWSiCNT indicated weak interactions with CO2 adsorption. The adsorption energy of CO2 on
top of the B site is −1.85 eV which is the largest negative value and this indicates that CO2 preferred to be
adsorbed well on the B-doped SWSiCNT system.
As shown in table 2, SWSiCNT doped with Al atom was found to have the highest value of bond length
(0.117 nm) because the diameter of Al is larger. On the other hand, SWSiCNT doped with B atom (0.114 nm)
has the lowest bond length because the diameter of the B atom is shorter. Therefore atoms in the Al-doped
SWSiCNT vibrate more freely than atoms in the N- and B-doped systems. In this case, B-doped SWSiCNT is
more stable because more energy is needed to set the atoms of these nanotubes into vibration. Typically, smaller
bond lengths mean the electrons are more tightly bound to the atom, and hence require more energy to remove,
leading to an increased bandgap [19]. Literature reports revealed a lattice constant of 3.12 nm for pure
SWSiCNT. As observed from table 2, the incorporation of impurities significantly changed the lattice parameter
of SWSiCNT in accordance with the size of the dopant, which is in good agreement with others [20].
3.2. Electronic transport mechanisms
Calculations of the electronic properties of the investigated nanotubes were done using both GGA and GW
approximation methods [21]. GGA was used to test the widely reported band gap problem of the Kohn–Sham
DFT while GW approximation was used to obtain an accurate description of the electronic systems. Figure 2(a)
presents the diagram of the electronic band structure of the CO2-adsorbed Al-doped SWSiCNT material. The
narrow band gap of 0.2 eV demonstrated a high charge transfer ratio due to CO2 adsorption which undergoes
spontaneous exothermic reaction accompanied by notable change in band structure. Therefore, the probability
of scattering of the adsorbed molecules of CO2 is higher with this nanotube. Regarding B-doped and N-doped
SWSiCNTs presented in figures 2(b) and (c) respectively, the CO2 adsorption does not significantly change the
band structures of the doped systems. Furthermore, reactions were endothermic accompanied by the formation
of band gap values of 2.57 eV and 2.62 eV respectively. It was also observed that the electronic properties of these
nanotubes were retained after removal of CO2 molecule which demonstrated good physisorption properties of
3
4
Table 2. Optimized geometries and adsorption energy data of the CO2 adsorbed Al-, B- and N-doped SWSiCNT.
Site of adsorption
Bond angle (degrees)
Lattice constants (nm)
Bond length of CO2 (nm)
Adsorption energy (eV)
Charge transfer (e)
Al-doped SWSiCNT with CO2 adsorbed on top of Al site
B-doped SWSiCNT with CO2 adsorbed on top of B site
N-doped SWSiCNT with CO2 adsorbed on top of the N site
129.12
125.01
126.11
3.16
3.13
3.15
0.117
0.114
0.115
−1.42
−1.85
−1.83
−0.615
−0.903
−0.901
P
h
y
s
.
S
c
r
.
9
9
(
2
0
2
4
)
0
1
5
9
2
0
Y
S
I
t
a
s
e
t
a
l
Phys. Scr. 99 (2024) 015920
Y S Itas et al
Figure 2. (a), (b) and (c): Electronic bands structures of CO2 adsorbed Al-, B- and N-doped SWSiCNTs and (d), (e) and (f): DOS of
CO2 adsorbed Al-, B- and N-doped SWSiCNT, respectively.
these nanotubes. Furthermore, the obtained band gaps were found to be higher than the calculated overpotential
value for good CO2 adsorption [9]. As widely known, the electronic density of states (DOS) determines the
systems’ magnetic, electronic, thermal, optical and other related properties [22]. The change in the energy gaps
due to molecular interaction with CO2 can be affected by lattice vibrations. In the DOS diagram of figure 2(d),
the presence of occupied states at the Fermi level revealed more lattice vibrations by the Al-doped SWSiCNT.
However, empty states at Fermi levels of B-doped and N-doped systems presented in figures 2(e) and (f) were
attributed to the semiconducting properties of these nanotubes and also zero lattice vibrations. At these regions,
the CO2 molecules were physically adsorbed on the BSi and NSi sites respectively. Therefore, CO2 finds space for
itself on the surface of these nanotubes in a similar way to water molecules sticking to the walls of the container to
form ice blocks. Compared to the Al-doped system, the intensity of DOS in the conduction bands of B- and
N-doped systems were higher indicating that CO2 was well physisorbed in the conduction band. A comparison
of the calculated electronic band gaps of the present systems under study is tabulated in table 5. It can be seen that
all the reported materials that were used for CO2 capture possessed band gap energy values between 1.8–2.8 eV
which is significantly sufficient to overcome the overpotential for successful CO2 capture.
3.3. Effects of bond length variation
To show the most stable bond length range in which the investigated nanotubes interact with photons, we have
studied the effects of different bond lengths of the Al-, B-, and N-doped SiC nanotubes with respect to CO2
adsorption. Since the bond length in the CO2 molecule is shorter than that of the SiC material, the probability of
its coverage at a given bond length determines the nature of the band shift. Table 3 shows the calculated values of
the band gaps of all investigated nanotubes at different bond lengths and results showed that the band gaps
increased with decreasing bond lengths. Other descriptions of the behaviors observed with these nanotubes can
be obtained in the remark column of table 3.
Regarding CO2 adsorbed Al-doped SWSiCNT shown in figure 3(a), it was found the near metallic properties
of the system has been retained by varying bond length of 1.78 Å, 1.76 Å and 1.74 Å. However, different
5
Phys. Scr. 99 (2024) 015920
Y S Itas et al
Figure 3. (a), (b) and (c): Effects of varying bond lengths of CO2 adsorbed Al-, B- and N-doped SWSiCNTs respectively.
Table 3. Effects of varying bond length on band gap.
Material
Al-doped SWSiCNT
B-doped SWSiCNT
N-doped SWSiCNT
Bond
length (Å)
Band
gap (eV)
Remark
1.78
1.76
1.74
1.76
1.74
1.72
1.77
1.75
1.73
0.20
0.22
0.30
2.57
2.59
3.10
2.62
2.64
2.67
The narrow band gap properties of the systems were retained, demonstrating its
inability for good photo interactions in the visible range
Band gap broadens significantly due to a decrease in bond length. This material
cannot be used to adsorb CO2 above the bond length of 1.74 Å
Band gap also increase due to decreasing bond length and can be used to adsorb
photon up to 1.72 Å
behaviors were observed with respect to B- and N-doped SWSiCNT as presented in figures 3(b) and (c). In terms
B-doped system, bond length was found to vary inversely with a band gap from 2.69 eV—3.0 eV. That is band
gap of CO2 adsorbed N-doped SWSiCNT increased with decreasing bond length because interacting electrons in
the nanotube lattice were tightly bound hence requiring more excitation energy to be removed which agreed well
with others. Moreover, a decrease in the band gap results in the overlapping of different orbital states which
promotes CO2 adsorption in the visible region. The photocatalytic efficiency of the B- and N-doped SWSiCNTs
was also enhanced by the band gap broadening of the valence band accompanied by B and N impurity states
within the band gap. In the CO2 adsorbed N-doped SWSiCNT, the band gap does not reduce beyond the bond
length of 1.77 Å, meaning that further increase in bond length above 1.77 Å does not affect the electronic
properties of the N-doped SWSiCNT system. As a comparative analysis with previous findings. Table 4 presents
the results of altering the bond length of different materials, comparison with table 3 revealed a good correlation
with our work. Various literature reports on band gaps of semiconductors used for carbondioxide capture can
be found in table 5. as can be seen, the band gaps obtained for B- and N-doped systems in the present study
agreed well with previous reports.
3.4. Optical spectra analysis
In order to study the CO2 adsorption properties of interacting systems, it is crucial to analyze the material’s
response to the incident electromagnetic radiation based on electron energy transfer and the imaginary dielectric
constant. Nanotubes interact with photons in different spectral ranges within a broad or narrow band. Materials
that store CO2 adsorb in the visible region of the electromagnetic spectrum and store it under ambient
conditions via a physisorption process accompanied by negligible electron energy loss [33]. The amount of CO2
energy loss was considered in directions parallel and perpendicular to the axes of the nanotubes under study.
Figure 4(a) shows the energy loss spectrum of the Al-doped SWSiCNT. As can be seen, the rising peaks started at
6
Phys. Scr. 99 (2024) 015920
Y S Itas et al
Figure 4. (a), (b) and (c): Electron energy loss spectra of CO2 adsorbed Al-, B- and N-doped SWSiCNTs and (d), (e) and (f): Imaginary
dielectric spectra of CO2 adsorbed Al-, B- and N-doped @SWSiCNTs.
Table 4. Previous reports on the effects of bond length variation on the electronic properties of different materials.
Material
Fe-doped TiO2
MoS2
B-doped MoS2
Carbon nano-
tubes (CNTs)
Fe-doped
SWSiCNTs
Description
Results
Reference
This research worked on the mechanisms of pho-
toabsorption by anatase under different con-
centrations of Fe. Studies were performed via DFT
+ U methods
This work analyzes the properties of MoS2 in
terms of strain using density functional theory.
Computational approaches were used to report
band gap engineering of B-doped MoS2 under dif-
ferent concentrations (2.08% and 4.16%) of B
impurity.
Effects of finite length on the electronic properties
of different chirality of CNTs were discussed in
this work.
Hydrogen energy storage potentials of SWSiCNT
were reported using Fe-doping via optical adsorp-
tion spectra analysis.
[23]
[24]
[25]
[26]
[27]
The band gap of Fe-doped TiO2 decreases due
to an increase in band length.
The effects of strain change the band gap as a
result of altering the bond length. Increasing
external strain tends to enlarge the Mo—
S bond.
Good bonding relations were observed between
MOS2 and B impurity accompanied by the
band gap broadening from 0.48 eV to 1.95 eV
due to a decrease in bond length.
The band gap value of CNTs was found to
decrease due to increasing bond length. Small
oscillations were observed due to the bonding
characteristics of HOMO–LUMO.
The band gap of Fe-doped SWSiCNT was
found to be reduced by increasing bond length.
The stability of the Fe-doped SiC nanotube was
found to be normal between bond lengths
1.73 Å to 1.79 Å, indicating its great potential
for H2 adsorption.
7
Phys. Scr. 99 (2024) 015920
Y S Itas et al
Figure 5. CO2 reflection spectra of (a) Al-doped, (b) B-doped and (c) N-doped SWSiCNTs.
Table 5. Calculated band gap values of different materials used to capture CO2.
Material
B-doped BNNTs
MWCNTs
B-doped SWSiCNTs
N-doped SWSiCNT
MgO-based sorbents
ZnO/activated carbon
TiO2/Anatase
Band gap (eV)
2.31
2.42
2.57
2.62
2.61
2.13
2.31
Applications
CO2 storage
CO2 capture
CO2 Capture
CO2 capture
CO2 capture
CO2 capture
CO2 capture
References
[28]
[29]
Current work
Current work
[30]
[31]
[32]
2.4 eV in both parallel and perpendicular directions, indicating energy loss in these regions. The energy loss is
due to many collisions between photogenerated electrons and holes due to high charge transfer ratios. Regarding
the nonmetal-doped SiC nanotubes, figure 4(b) shows the energy loss of the B-doped SWSiCNT during its
interactions with CO2 gas. The presence of flat electron peaks from 0–3.01 eV indicated that the CO2 energies
were not lost in the visible region and therefore CO2 was well physisorbed. Larger losses can be observed far
above the UV radiation, in these areas, CO2 can only be adsorbed via chemisorption processes. Similarly, the
electron loss spectrum of N-doped SWSiCNT presented in figure 4(c) revealed zero loss in the visible spectrum
which also demonstrated that CO2 was well physisorbed by this nanotube. The imaginary part of the dielectric
constant (Ɛ2) describes the energy absorbed by interacting systems. In regard to this work, we have analyzed the
amount of CO2 adsorbed by all three investigated systems based on the imaginary dielectric spectra calculated
using Kramers-Kronig relations, as shown in equation (1) [34].
e
2
= -
¥
w
2
p
P
ò
0
¢ -
e w
(
)
1
2
¢ -
w
w
1
2
w
d
¢
( )
2
Where ω is real and P denotes the Cauchy principal value.
Regarding the Al-doped SiC nanotube shown in figure 4(d), the edge of adsorption in the parallel direction
was observed at 0.8 eV while 0.7 eV was observed perpendicular to the nanotube axis. However, these values
were not found to be above 1.8 eV for materials that adsorb CO2. In terms of B-doped SWSiCNT, as presented in
figure 4(e), the bound state was clearly observed at 2.4 eV which is in the visible range and above the
overpotential values for good CO2 capture. Therefore, B-doped SWSiCNT presents itself as a good material for
CCUS. The higher peak in the parallel direction indicated more CO2 adsorption than in the perpendicular
direction. Similar properties were observed for N-doped SWSiCNT in figure 4(f), except that the two adsorption
peaks were seen extending from visible to UV. Compared to a B-doped system, its performance can be limited
due to the molecular scattering by ultraviolet rays which promotes the formation of ozone molecules. The
nonmetal doped SiCNT adsorbs CO2 due to the introduction of the dopant impurities which excite the system
by the suitable solar irradiation accompanied by generation of electrons and holes. These photogenerated
electrons then separate and move to the surface of the photocatalyst.
8
Phys. Scr. 99 (2024) 015920
Y S Itas et al
Table 6. Summary of calculated optical properties of Al-, B- and N-doped CO2@SWSiCNTs.
Material
Energy loss (eV)
Energy adsorbed (eV)
Energy reflected (eV)
Remarks
Al-doped SWSiCNT
B-doped SWSiCNT
N-doped SWSiCNT
7.21 (UV)
10.21 (UV)
10.32 (UV)
5.10 (UV)
2.61 (Visible)
2.73 (Visible)
2.51 (UV)
4.93 (UV)
5.20 (UV)
Not efficient for CO2 capture
Very efficient for CO2 capture
Efficient CO2 capture
Table 7. Summary of DFT results on CO2 capture by different materials.
Material
Dopant
Result
CaO
TiO2
SiCNT
BNNT
SWSiCNT
None
B
Ti; Cu
Cu
B
CaO
Ca12Al14O13
CO2 was adsorbed on top O site of CaO horizontally.
CO2 adsorption by B-doped TiO2 significantly improved compared to that of CaO.
CO2 adsorption was not favorable with Ti and Cu-doped SiCNT due to high chemisorption
on the SiCNT surface.
CO2 was much physisorbed by Cu-doped BNNT followed by large adsorption energy.
CO2 molecules were well physisorbed in both parallel and perpendicular directions on the
sites of B accompanied by much adsorption energy.
CO2 adsorption was inhibited due to the pre-adsorption by Ca12Al14O13
Reference
[38]
[39]
[8]
[40]
Current work
[38]
3.5. Surface scattering and reflection
In a photochemical process, light if not absorbed by a surface, is mostly reflected or scattered [35]. The intensity
of the reflected light and angle of scattering determines the potential of material to the absorption of photon
which in turn depends on the frequency of light. We have investigated the scattering properties of all the three
systems considered and appropriate results were obtained. Figure 5 represents the spectrum of the Al-doped
(metal-doped) SWSiCNT in both parallel and perpendicular directions. It has been observed that the material
reflects the adsorbed CO2 in the energy range 0–2.5 eV which corresponds to the visible region, indicating that
little or no CO2 was adsorbed by this system. Additionally, the material reflects in all directions due to high
carrier mobility and electron–hole recombination across the narrow band gap of the Al-doped SWSiCNT. The
properties observed from this nanotube were similar to the previous result obtained regarding CO2 storage
potentials of metal-doped SWSiCNT, hence not better for CO2 capture and storage [36].
Results presented in figures 5(b) and (c) revealed that both B- and N-doped SiCNTs demonstrated good
physisorption behaviors, because the nanoparticles of the CO2 are scattered when incident light irradiates the
scattering surface of the SiC nanotube [37]. Due to this, the scattered light has a longer optical path in the visible
range and hence can be more easily absorbed. Moreover, the doped atoms of B and N lead to significant surface
acidity, thus increasing the adsorption by enhancing the number of photons available to take part in the
photocatalytic activity. Regarding B-doped SWSiCNT, flat peaks from 0-.272 eV means no reflection in that
region hence all the incident CO2 photon have been adsorbed. In terms of the N-doped SWSiCNT
photocatalyst, there is a little reflection in the visible range, therefore the amount of CO2 photon adsorbed is less
than the amount that was adsorbed by the B-doped system. For all the systems, higher reflection occurred in the
UV region which accounted for only 3% of the solar irradiation. Compared to the Al-doped system, B- and
N-doped SiC nanotubes reflect the least CO2 in the visible and more in the UV which makes them good
candidates for CO2 capture, while Al-doped reflects more in the visible and less in the UV and hence poor for
CO2 capture. Table 6 presents the summarized calculated optical properties of the investigated metal-doped and
nonmetal-doped SWSiCNTs for CO2 adsorptions. As can be seen, the Al-doped SiCNT was not favorable to
CO2 storage due to much reflecting in the visible and higher adsorption potentials in the UV range. Regarding B-
and N-doped systems, the adsorption of CO2 molecules was narrowed down to the visible range while recording
energy loss and reflection in the UV. Moreover, the B-doped SiC nanotube was found to be very efficient
compared to the N-doped SiC nanotube because it extended most of the light absorbed to the visible (2.61 eV)
region than the N-doped system (2.73 eV). It also demonstrated less energy loss in the visible region than
N-doped system.
The observed calculated optical properties, therefore, present these nanotubes as good candidates for CCUS.
The outlined data in table 7 revealed that SWSiCNT has been successfully tuned to CO2 adsorption which marks
a significant improvement to the Ti, Pb and Cu-doped SWSiCNT. It was also observed from table 7 that all the
doped systems which store CO2 have done so due to the capture of CO2 molecules on the dopant site.
9
Phys. Scr. 99 (2024) 015920
4. Conclusion
Y S Itas et al
CO2 capture and storage potentials of the metal and nonmetal doped SWSiCNT semiconductor were
investigated using the prominent density functional theory. The properties were investigated based on the
analysis of electronic bands, density of states, adsorption energies and optical spectra analysis. The obtained
results revealed poor performance by the Al-doped SWSiCNT in terms of CO2 capture which agreed well with
the previous results reported in the literature. Regarding B- and N-doped systems, the obtained band gap values
of 2.57 eV and 2.68 eV respectively brought these nanotubes as having good overpotential energies for good CO2
adsorption. Optical adsorption spectra analysis revealed negligible energy loss from 0–2.6 eV and higher CO2
adsorption in this range by B- and N-doped SWSiCNTs. Moreover, these nanotubes (B- and N-doped) were
found to reflect more in the UV region which has nothing to do with adsorption in the visible range. Compared
to the Al-doped system, B- and N-doped SiC nanotubes reflect the least CO2 in the visible and more in the UV
which makes them good candidates for CO2 capture; while Al-doped reflects more in the visible and less in the
UV hence poor for CO2 capture. Based on the obtained results, this study suggests that the B-doped SWSiCNT is
a better candidate for CO2 capture, storage and subsequent utilization.
Acknowledgments
The authors acknowledge the deanship of scientific research of Taif University for funding this research.
Data availability statement
All data used are available in the manuscript. The data that support the findings of this study are available upon
reasonable request from the authors.
Author contributions
Conceptualization, Y S I. and R R.; methodology Y S I. and R.R.; software, M U K, S A., H O. and R R.; formal
analysis, Y S I., R R., and S A.; resources, M U K., R R., S A. and H O. ; data curation, Y S I., R R., and M.U.K.;
writing—original draft preparation, Y S I.; writing—review and editing, M U K.; visualization, Y S I., R R., S A.,
H O. and M.U.K.
All authors have read and agreed to the published version of the manuscript.
ORCID iDs
Yahaya Saadu Itas
Mayeen Uddin Khandaker
https://orcid.org/0000-0002-5251-7082
https://orcid.org/0000-0003-3772-294X
References
[1] Humayun A and Anwar M N 2021 Nanomaterials for CO2 Capture, Storage, Conversion and Utilization,’ in Micro and Nano Technologies
(Elsevier) 1–19
[2] Sharifian R, Wagterveld R M, Digdaya I A, Xiang C and Vermaas D A 2021 Electrochemical carbon dioxide capture to close the carbon
cycle Energy Environ. Sci. 14 781–814
[3] Fu J, Jiang K, Qiu X, Yu J and Liu M 2020 Product selectivity of photocatalytic CO2 reduction reactions Mater. Today 32 222–43
[4] Sulafa A S, Wan Z N, Mohammad A B and Kibria G M 2023 Experimental and computational evaluation of 1,2,4-triazolium-based
ionic liquids for carbon dioxide capture Separations 10 192
[5] Ngoy J M, Wagner N, Riboldi L and Bolland O 2014 ‘A CO2 capture technology using multi-walled carbon nanotubes with
polyaspartamide surfactant Energy Procedia 63 2230–48
[6] Iugai I A and Steksova Y P 2020 MgO/carbon nanofibers composite coatings on porous ceramic surface for CO2 capture Surf. Coat.
Technol. 400 126208
[7] Ikuo Y, Satoshi K and Hidehiko K 2018 Reversible CO2 capture by ZnO slurry leading to formation of fine ZnO particles Adv. Powder
Technol. 29 1239–45
[8] Singh R S 2023 CO2 capture by metal-decorated silicon carbide nanotubes Silicn. 2023 02368–9
[9] Wu J, Huang Y, Ye W and Li Y 2017 CO2 reduction: from the electrochemical to photochemical approach Adv. Sci. 4 1700194
[10] Hongyu Z, Xuejuan Z and Mingwen Z 2010 Atomic-scale characterization of silicon diffusion on carbon nanotubes Physica E 43 610–3
[11] Itas Y S, Suleiman A B, Yamusa A S, Razali R and Danmadami A M 2022 Ab’initio studies of the structural and electronic properties for
single-walled armchair MgONT, SiCNTs and ZnONTs for next generations’ optoelectronics Gadau Journal of Pure and Allied Sciences 1
160–5
[12] Itas Y S, Abdussalam B S, Chifu E N, Lawal A and Khandaker M U 2022 The exchange-correlation effects on the electronic bands of
hybrid armchair single-walled carbon boron nitride nanostructure Crystal 12 394
10
Phys. Scr. 99 (2024) 015920
Y S Itas et al
[13] Golze D, Dvorak M and Rinke P 2019 The GW compendium: a practical guide to theoretical photoemission spectroscopy Front. Chem.
7 00377
[14] Yahaya S I, Balarabe A S, Ebene C E, Khandaker M U and Razif R 2023 Effects of SiO2 and CO2 absorptions on the structural, electronic
and optical properties of (6, 6) magnesium oxide nanotube (MgONT) for optoelectronics applications Silicon 15 5341–52
[15] Chunhong Z, Zhang Z and Yan W 2021 Effect of doping on the photoelectric properties of borophene Adv. Condens. Matter Phys. 2021
3718040
[16] Jian Y, Huang Z, Wang Y and Xing J 2020 Effects of doped elements (Si, Cr, W and Nb) on the stability, mechanical properties and
electronic structures of MoAlB phase by the first-principles calculation Materials (Basel) 4221 13
[17] Zhenyi N, Pi X and Yang D 2017 Density functional theory study on the B doping and B/P codoping of Si nanocrystals embedded in Si
O 2 Phys. Rev. B 95 159905
[18] Kwon S, Choi J I, Lee S G and Jang S S 2014 A density functional theory (DFT) study of CO2 adsorption on Mg-rich minerals by
enhanced charge distribution Comput. Mater. Sci. 95 181–6
[19] Itas Y S, Razif R and Khandaker M U 2023 Structural, mechanical, electronic and optical properties of N-decorated single-walled
silicon carbide nanotube photocatalyst for hydrogen evolution via water splitting: a DFT study Sci. Technol. Adv. Mater. 24 2271912
[20] Mahmood U and Muzaffar I 2017 Effect of Fe doping on the crystallinity of CuO nanotubes and the efficiency of the hybrid solar cells
J. Photochem. Photobiol., A 335 112–8
[21] Itas Y S, Balarabe A S, Chifu E N, Lawal A, Razif R and Khandaker M U 2022 Effects of oxygen absorption on the electronic and optical
properties of armchair and zigzag silicon carbide nanotubes (SiCNTs) Phys. Scr. 98 015824
[22] Itas Y S, Balarabe A B, Ebene C E, Lawal A, Razif R and Khandaker M U 2023 DFT studies of structural, electronic and optical properties
of (5, 5) armchair magnesium oxide nanotubes (MgONTs) Physica E 149 115657
[23] Wu H-C and Lin S-W 2012 Effect of Fe Concentration on Fe-Doped Anatase TiO2 from GGA + U Calculations Int. J. Photoenergy 2012
823498
[24] Chan L, Fan B and Li W 2015 Bandgap engineering of monolayer MoS 2 under strain: a DFT Study Journal- Korean Physical Society 66
1789–93
[25] Tayyab M, Akhtar H, Waqar A and Nabi S 2021 A computational study of MoS2 for band gap engineering by substitutional doping of
TMN (T = transition metal (Cu), M = metalloid (B) and N = non-metal (C)) Mater. Res. Express 8 046301
[26] Mahmud Prince T, Kaisar T, Kaisar T and Mominuzzaman S M M S M 2019 Effect of finite length on the band gap of semiconducting
and metallic carbon nanotube International Conference on Electrical, Computer and Communication Engineering (ECCE) 2019
18566214
[27] Itas Y S, Tata S, Razif R and Khandaker M U 2023 Studies of the hydrogen energy storage potentials of Fe-and Al-doped silicon carbide
nanotubes (SiCNTs) by optical adsorption spectra analysis Journal of Energy Storage 72 108534
[28] Choi H, Choon P, Kim Y and Lee S 2011 Ambient carbon dioxide capture by boron-rich boron nitride nanotube J. Am. Chem. Soc. 133
2084–7
[29] Jacob M N, Nicola W, Luca R and Olay B 2014 A CO2 capture technology using multi-walled carbon nanotubes with polyaspartamide
surfactant Energy Procedia 63 2230–48
[30] Felix D and Muller C 2022 Prospects of MgO-based sorbents for CO2 capture applications at high temperatures Current Opinion in
Green and Sustainable Chemistry 36 100645
[31] Kenji T, Kenji N and Suzuki K 2015 CO2 capture in humid gas using ZnO/activated carbon and ZnO reactivity with CO2 Catalysts 115
563–79
[32] Mino L and Spoto G 2014 CO2 capture by TiO2 anatase surfaces: a combined DFT and FTIR Study J. of Phy. Chem. 118 25016–26
[33] Itas Y S, Khandaker M U, Suleiman A B, Ndikilar C E, Lawal A and Razali R 2023 Studies of H2 storage efficiency of metal-doped carbon
nanotubes by optical adsorption spectra analysis Diam. Relat. Mater. 136 109964
[34] Yahaya S I, Abdussalam B S, Chifu E N, Abdullahi L, Razif R and Khandaker M U 2022 Computational studies of the excitonic and
optical properties of armchair SWCNT and SWBNNT for optoelectronics applications Crystals 12 870
[35] Yahaya I S, Suleiman A B, Ebene C E, Khandaker M U and Razif R 2023 First-principle studies of the structural, electronic, and optical
properties of double-walled carbon boron nitride nanostructures heterosystem under various interwall distances Journal of Chemistry
2023 4574604
[36] Osler K, Dheda D, Ngoy J, Wagner N and Daramola M O 2017 ‘Synthesis and evaluation of carbon nanotubes composite adsorbent for
CO2 capture: a comparative study of CO2 adsorption capacity of single-walled and multi-walled carbon nanotubes International
Journal of Coal Science & Technology 4 41–9
[37] Zuoli H and Que W 2014 Surface scattering and reflecting: the effect on light absorption or photocatalytic activity of TiO2 scattering
microspheres Phys. Chem. Chem. Phys. 3332-3339 4
[38] Xianyao Y, Yingjie L, Zhao J and Wang Z 2020 Density functional theory study on CO2 adsorption by CePromoted CaO in the presence
of steam Energy & Fuels 34 6197–208
[39] Kreft S, Duo W, Junge H and Beller M 2020 Recent advances on TiO2-based photocatalytic CO2 reduction Energy Chem. 2 100044
[40] Mahdavifar Z and Abbasi N 2014 The influence of Cu-doping on aluminum nitride, silicon carbide and boron nitride nanotubes’
ability to detect carbon dioxide; DFT study Physica E 56 268–76
11
| null |
10.1038_s41467-023-37865-3.pdf
|
ndings of this study are available
in the RCSB Protein Data Bank (PDB) under accession numbers: 7LV2
[https://doi.org/10.2210/pdb7LV2/pdb],
7LTU [https://doi.org/10.
7LUX [https://doi.org/10.2210/
2210/pdb7LTU/pdb]
and 7LUZ [https://doi.org/10.2210/
pdb7LUX/pdb]
pdb7LUZ/pdb]. The amino acid sequences of the Nucleocapsid
(form 1),
(form 2),
Nature Communications |
(2023) 14:2379
15
Article
https://doi.org/10.1038/s41467-023-37865-3
proteins of SARS-CoV-2 and SARS-CoV analyzed in this study are
available on UniProtKB, accession numbers: P0DTC9, and P59595
respectively. Amino acid sequences of other coronavirus Nucleocapsid
proteins were accessed from the European Nucleotide Archive [ENA;
https://www.ebi.ac.uk/genomes/virus.html]. Raw EM images, light and
fluorescence microscopy images and fiber diffraction source files
generated in this study have been deposited in the Figshare respiratory
[https://figshare.com/projects/Low_Complexity_Domains_of_the_
at
Nucleocapsid_Protein_of_SARS-CoV-2_Form_Amyloid_Fibrils/162391].
Data for all plots presented in this manuscript are provided wit
|
Data availability Atomic coordinates that support the findings of this study are available in the RCSB Protein Data Bank (PDB) under accession numbers: 7LV2 [ https://doi.org/10.2210/pdb7LV2/pdb ], 7LTU [ https://doi.org/10. 2210/pdb7LTU/pdb ] (form 1), 7LUX [ https://doi.org/10.2210/ pdb7LUX/pdb ] (form 2), and 7LUZ [ https://doi.org/10.2210/ pdb7LUZ/pdb ]. The amino acid sequences of the Nucleocapsid
|
Article
https://doi.org/10.1038/s41467-023-37865-3
Low complexity domains of the nucleocapsid
protein of SARS-CoV-2 form amyloid fibrils
Received: 29 July 2022
Accepted: 3 April 2023
Check for updates
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
1,2,3,4, Xinyi Cheng 1,2,3,4, Lukasz Salwinski
1,2,3,4, Christen E. Tai
6,
Einav Tayeb-Fligelman1,2,3,4, Jeannette T. Bowler
Michael R. Sawaya 1,2,3,4,5, Yi Xiao Jiang 1,2,3,4, Gustavo Garcia Jr
1,2,5,
Sarah L. Griner
Liisa Lutter1,2,3,4, Paul M. Seidler1,2,11, Jiahui Lu1,2,3,4, Gregory M. Rosenberg 1,2,3,4,
Ke Hou 1,2,3,4, Romany Abskharon1,2,3,4, Hope Pan 1,2,3,4, Chih-Te Zee3,
David R. Boyer
Genesis Falcon5, Duilio Cascio 5, Lorena Saelices
Robert Damoiseaux6,7,8,9,10, Vaithilingaraja Arumugaswami6,8,9,
Feng Guo 1,2,10 & David S. Eisenberg 1,2,3,4,5,8
1,2, Daniel H. Anderson1,2,3,4, Kevin A. Murray1,2,3,4,
1,2,12,
1,2,3,4, Yan Li
1,2,
The self-assembly of the Nucleocapsid protein (NCAP) of SARS-CoV-2 is crucial
for its function. Computational analysis of the amino acid sequence of NCAP
reveals low-complexity domains (LCDs) akin to LCDs in other proteins known
to self-assemble as phase separation droplets and amyloid fibrils. Previous
reports have described NCAP’s propensity to phase-separate. Here we show
that the central LCD of NCAP is capable of both, phase separation and amyloid
formation. Within this central LCD we identified three adhesive segments and
determined the atomic structure of the fibrils formed by each. Those struc-
tures guided the design of G12, a peptide that interferes with the self-assembly
of NCAP and demonstrates antiviral activity in SARS-CoV-2 infected cells. Our
work, therefore, demonstrates the amyloid form of the central LCD of NCAP
and suggests that amyloidogenic segments of NCAP could be targeted for
drug development.
The Nucleocapsid protein (NCAP) of severe acute respiratory
syndrome coronavirus 2 (SARS-CoV-2) is an RNA-binding protein
that functions in viral replication by packaging the genomic
viral RNA (vRNA) and aiding virion assembly1–9. During its function,
NCAP engages in multivalent RNA–protein and protein–protein
interactions and self-associates via several interfaces10. Increasing
replication efficiency, NCAP forms concentrated protein–RNA
compartments through a process of phase separation (PS)1,2,4–8,10–12.
NCAP PS is enhanced in low salt buffers4,5 and in the presence of
zinc ions2, and these PS droplets may exist in a liquid or solid-
like state1,2,4,8,11. The liquid state of
the droplets is favored
by NCAP phosphorylation and is presumed to enable vRNA pro-
In contrast, non-
cessing in the early stages of
phosphorylated NCAP oligomerizes and forms solid-like droplets,
possibly to facilitate RNA packaging and nucleocapsid assembly in
later stages4,8.
infection4,8.
1Department of Biological Chemistry, UCLA, Los Angeles, CA 90095, USA. 2Molecular Biology Institute, UCLA, Los Angeles, CA 90095, USA. 3Department of
Chemistry and Biochemistry, UCLA, Los Angeles, CA 90095, USA. 4Howard Hughes Medical Institute, Los Angeles, CA 90095, USA. 5UCLA-DOE Institute of
Genomics and Proteomics, UCLA, Los Angeles, CA 90095, USA. 6Department of Molecular and Medical Pharmacology, UCLA, Los Angeles, CA 90095, USA.
7Department of Bioengineering, UCLA, Los Angeles, CA 90095, USA. 8California NanoSystems Institute, UCLA, Los Angeles, CA 90095, USA. 9Eli and Edythe
Broad Center of Regenerative Medicine and Stem Cell Research, UCLA, Los Angeles, CA 90095, USA. 10Jonsson Comprehensive Cancer Center, UCLA, Los
Angeles, CA 90095, USA. 11Present address: Department of Pharmacology and Pharmaceutical Sciences, University of Southern California School of
Pharmacy, Los Angeles, CA 90089-9121, USA. 12Present address: Center for Alzheimer’s and Neurodegenerative Diseases, Department of Biophysics, Peter
O’Donnell Jr. Brain Institute, University of Texas Southwestern Medical Center, Dallas, TX 75390, USA.
e-mail: [email protected]
Nature Communications |
(2023) 14:2379
1
Article
https://doi.org/10.1038/s41467-023-37865-3
The sequence of NCAP encompasses both RNA-binding and low-
complexity domains. Low-complexity domains (LCDs) are protein
segments comprised of a restricted subset of amino acid residues such
as glycine, arginine, lysine, and serine13–15. Long mysterious in function,
LCDs have recently been established to drive PS and form unbranched,
twisted protein fibrils known as amyloid-like fibrils. Such behavior was
observed in LCD-containing human RNA-binding proteins such as FUS,
TDP-43, and hnRNPA2. By PS and amyloid formation, LCDs non-
covalently link their parent proteins, and in some cases RNAs, into
larger assemblies13,16–18. These larger assemblies are associated with the
formation of subcellular bodies known variously as hydrogels13,19,
condensates20, and membrane-less organelles19. In short, the self-
association of several RNA-binding proteins has been shown to be
driven at least in part by amyloid-like fibrils formed by their LCDs and
to be a regulatory element of RNA metabolism in cells19.
Motivating our study is a medical experience that even efficient
vaccines rarely eradicate viral diseases and their legacies of morbidity
and mortality21, so COVID-19 therapies are needed. Along with others10
we hold that NCAP of SARS-CoV-2 is a worthy drug target and that a
better understanding of the structure and mechanism of action of
NCAP may aid in drug development. NCAP is abundant in SARS-CoV-2-
infected cells and its function is crucial for viral replication and
assembly10. NCAP is also evolutionarily conserved in the coronavirus
genus10, which may render it as an effective target not only for COVID-
19 treatments but possibly also for future coronavirus pandemics.
Here we show that NCAP possesses two fibril-forming LCDs, one
central and one C-terminal. The central LCD forms Thioflavin-S (ThS)-
positive PS droplets and amyloid fibrils that exhibit a characteristic
diffraction pattern. At least three adhesive segments in this central
LCD are capable of mediating amyloid typical interactions, and we
elucidated the atomic structure of the fibrils formed by each. Guided
by these structures, we designed a peptide that shifts NCAP to a less
ordered mode of aggregation and investigated the peptide’s effect on
the infection of human cells by SARS-CoV-2.
Results
NCAP contains central and C-terminal LCDs
Using the SEG algorithm22 we analyzed the sequence of NCAP and
identified a 75-residue LCD (residues 175–249) within NCAP’s central
intrinsically disordered region, as well as a second, lysine-rich LCD of
19 residues (residues 361–379) within its C-terminal tail (CTT)
(Fig. 1a, b). SEG is a widely used algorithm that identifies segments in a
sliding window as either high or low complexity by statistically ana-
lyzing the amino acid distribution as a measure of sequence
complexity22. While not all LCDs identified this way are capable of PS
and amyloid formation, LCDs that do phase separate are readily
identified by SEG23,24. In NCAP, those central and C-terminal LCDs,
along with an N-terminal disordered region, flank the structured RNA-
binding and dimerization domains of the protein (Fig. 1a).
NCAP’s LCDs participate in fibril formation
To assess possible amyloid formation of NCAP’s LCDs and to identify
adhesive segments that drive it, we expressed and purified NCAP and
its LCD-containing segments in E. coli. Those segments included resi-
dues comprising NCAP’s central LCD and surrounding residues (con-
struct named LCD, residues 171–263) and a segment that includes the
C-terminal LCD with the C-terminal tail and dimerization domain
(construct DD-Cterm, residues 257–419) (Fig. 1c). Only RNA-free protein
fractions were combined at the last step of protein purification for use
in subsequent experiments. We then verified that our purified full-
length NCAP protein is capable of PS by mixing it with a 211-nucleotide
5′-genomic vRNA segment named hairpin-Site2 (S2hp; Supplementary
Fig. 1, Supplementary Table 1) in the presence and absence of the PS
2 (Supplementary Fig. 2a and Supplementary text).
enhancing ZnCl2
The S2 vRNA sequence was previously suggested to be a strong NCAP
cross-linking site7, and we extended it by including the adjacent hairpin
regions that improve binding to NCAP25.
Using our recombinant protein system we found that NCAP’s
LCDs are capable of binding the amyloid-dye Thioflavin-T (ThT). In a
ThT amyloid-formation kinetic assay performed over ~35 h of mea-
surement (Fig. 1d, e), S2hp vRNA mixtures (in 4:1 protein: vRNA molar
ratio) of the central LCD and the DD-Cterm segments of NCAP produced
amyloid formation curves. Whereas the DD-Cterm + vRNA curve plateau
after ~3 h of incubation, LCD + vRNA plateaus ~10 h after the start of
measurements while producing a significantly higher fluorescence
signal than that of DD-Cterm. The full-length NCAP also exhibited
increased ThT fluorescence over 5 h of measurement when mixed with
S2hp vRNA, followed by a slight decrease in signal, possibly because of
spontaneous disaggregation (Supplementary Fig. 2e). However, nei-
ther NCAP nor the DD-Cterm segment demonstrated a clear lag phase in
their ThT curves. Also, in the absence of S2hp vRNA, we did not detect
an increase in ThT fluorescence in any of the samples within 35 h of
measurements. This suggests that vRNA promotes the formation of
ThT-positive aggregates from those LCD-containing protein con-
structs, at least in the first 1.5 days of incubation.
Visualization of fibrils by electron microscopy (EM) confirmed the
the LCD-containing constructs to adopt fibrillar
propensity of
morphologies (Fig. 1f and Supplementary Fig. 2c). To observe fibrils of
NCAP and its LCD-containing segments by EM we increased protein
concentration and incubated each protein separately for ~1–2 weeks
with and without S2hp vRNA. Of note, under the conditions used for
the kinetic ThT experiment (Fig. 1d, e and Supplementary Fig. 2e) we
did not detect fibrils by EM, suggesting that the ThT experiment is
more sensitive for the detection of amyloid-like aggregates or that ThT
interacts with pre-fibrillar assemblies of the proteins. Other explana-
tions, such as poor adherence of the protein fibrils to the EM grid and
fibril reversibility are also reasonable. Nevertheless, with increased
protein concentration and incubation time, fibrils were detected by EM
both in the presence and absence of the vRNA. Fibrils of the DD-Cterm
segment with vRNA are morphologically different than those grown in
its absence, however, the central LCD segment produces amyloid-
looking fibrils under both conditions. Indeed, concentrated LCD-only
samples exhibit increased ThT fluorescence signal upon 6 and 11 days
of incubation, but with large sample-to-sample variability (Fig. 1g).
vRNA is, therefore, not essential for fibril formation and ThT binding,
but may promote these processes.
The full-length NCAP also forms fibrillar morphologies in samples
containing higher protein-to-vRNA ratio (40:1 protein:vRNA molar
ratio), as well as when incubated with zinc ions in PBS (Supplementary
Fig. 2b), and particularly in a low ionic strength buffer (Supplementary
Fig. 2d) upon 3–6 days of incubation (as indicated in Supplementary
Fig. 2). NCAP and also the DD-Cterm fibrils are much sparser in EM
images compared to the central LCD, and their morphologies differ
from those of the central LCD or canonical amyloid fibrils. Together,
those observations suggest that NCAP and its LCD-containing seg-
ments, particularly the central LCD segment, are capable of forming
aggregates of fibrillar morphologies as well as ThT-positive species.
The central LCD forms amyloid typical fibrils
To examine the amyloid property of fibrils formed by NCAP and its
LCD segment we used X-ray fiber diffraction. The X-ray fiber diffraction
patterns of the central LCD showed a sharp reflection at 4.7 Å spacing
and a diffuse reflection at 10 Å typical of amyloid fibrils. This is true for
fibrils formed by the LCD alone (no RNA), and by LCD with S2hp vRNA
or a non-specific RNA segment (Fig. 2a). This capacity of the central
LCD segment of NCAP to stack into amyloid fibrils associates it with
LCDs of other RNA-binding proteins that are involved in functional
amyloid-formation and amyloid pathologies13,16,18. We were unable,
however, to obtain a clear diffraction pattern from the full-length
NCAP. This may be a result of low fibril concentration, as evident by EM
Nature Communications |
(2023) 14:2379
2
Article
https://doi.org/10.1038/s41467-023-37865-3
Fig. 1 | NCAP’s LCDs form fibrils and ThT-positive species. a NCAP’s domain
organization. Domain definitions: N-terminal tail (NTT, gray), RNA-binding domain
(red); Central low complexity domain (LCD, purple; residues 175–249), Dimeriza-
tion domain (blue); C-terminal tail (CTT, gray). The C-terminal LCD is highlighted in
yellow (residues 361–379). b Amino acid sequence of the central and C-terminal
LCDs highlighted and colored according to the color scheme in (a). Lowercase
letters represent residues of low complexity while capital letters represent non-low-
complexity residues. No more than five interrupting non-low-complexity residues
between strings of 10 or more low-complexity residues were allowed. Steric-zipper
forming sequences that are discussed below are underlined in the sequence of the
central LCD. c Protein segments used in this study are abbreviated as LCD, con-
sisting of the central LCD and surrounding residues, and as DD-Cterm, consisting of
the dimerization domain (DD) and the C-terminal tail, including the C-terminal LCD.
The LCD and DD-Cterm segments are colored according to the color scheme in (a).
d and e ThT fibril formation kinetic assays of the LCD (d) and DD-Cterm (e) segments
incubated with (purple/navy, respectively) and without (pink/light blue, respec-
tively) hairpin-Site2 (S2hp) viral RNA (vRNA). f Fibril formation from concentrated
LCD and DD-Cterm samples observed by negative stain EM after 6 days of incubation
with and without S2hp vRNA. Scale bar = 500 nm. g Endpoint ThT fluorescence
measurements of concentrated LCD-only samples (pink) and buffer-only controls
(white) at days 1, 6, and 11 of incubation. Dots indicate individual data points and
bars represent mean values ± SD. n = 3 samples. Source data for panels d, e, and
g are provided as a Source Data file.
(Supplementary Fig. 2c), and/or from fibril decomposition during
washing steps meant to eliminate salts from the sample.
The central LCD segment of NCAP also readily forms unbranched
fibrils in the presence of short, unstructured vRNA types such as the
Site1 (S1), Site1.5 (S1.5) and S2 segments, as well as with a non-specific
RNA segment of a similar length (Fig. 2b: Supplementary Table 1), and
even with no RNA (Fig. 1f). When the LCD segment is incubated for one
day with either S1 or S2 vRNA segments, the LCD produces heavily-
stained clusters with fibrils protruding from their edges, but these
clusters disperse after 4 days of incubation. Such behavior is not
observed with S1.5 vRNA or the non-specific RNA segment (Fig. 2b).
This may suggest that the LCD fibril growth process may be altered by
the RNA sequence. Overall, the amyloid formation of the central LCD
offers that this region could potentially promote ordered self-
assembly of NCAP under the appropriate conditions.
NCAP’s central LCD forms PS droplets and solid particles
Next, we examined the capacity of the central LCD to form PS droplets
with different S2hp vRNA concentrations and followed the behavior
and character of the droplets over time in the presence of the amyloid
dye Thioflavin-S (ThS) using light and fluorescence microscopy. In
samples of 4:1 and 40:1 LCD: S2hp vRNA molar ratios we visualized PS
droplets that gradually transition into rough, less circular, seemingly
solid particles (Fig. 3a, b). In the 40:1 LCD: S2hp sample, PS droplets
form and begin to fuse within 30 min of incubation, and ThS partitions
into the droplets and produces rather bright fluorescence (Fig. 3a, c).
Nature Communications |
(2023) 14:2379
3
Article
https://doi.org/10.1038/s41467-023-37865-3
Fig. 2 | The central LCD segment of NCAP demonstrates amyloid-like char-
acteristics. a X-ray diffractions of LCD-only fibrils (No RNA), and LCD fibrils grown
with hairpin-Site2 (S2hp) vRNA or non-specific RNA (antisense siDGCR8-1), show
amyloid-characteristic 4.7 and 10 Å reflections labeled by white arrows. b Negative
stain EM micrographs of LCD fibrils grown in the presence of the short vRNA
segments Site1 (S1; 11 nucleotides), Site 1.5 (S1.5; 22 nucleotides) and Site2 (S2; 22
nucleotides), as well as with a non-specific RNA (antisense siDGCR8-1). All RNA
sequences are given in Supplementary Table 1. This figure shows that the central
LCD produces amyloid-typical fibrils in the absence and presence of different RNA
segments and that the RNA sequence may influence the morphology of the LCD
assemblies over time.
Upon 2 h of incubation,
larger asymmetric droplets appear, and
after 6 h, filamentous structures decorate the droplets. Within 4 days
of incubation, the droplets transform into what appear as solid-like
filamentous particles. At a higher S2hp concentration (4:1 LCD: S2hp
molar ratio), small PS droplets appear after ~30 min, but those
droplets show almost no ThS fluorescence (Fig. 3a, c). Additional PS
droplets form after 2 h of incubation and a weak ThS signal is detected.
However, after 6 h incubation, and even more predominantly after
4 days, most droplets convert into brightly fluorescent particles
(Fig. 3a, c).
An analysis of LCD assemblies (droplets and solid-like particles)
from a series of light microscope images taken at different time points
of incubation shows that the mean area of the 40:1 LCD:vRNA assem-
blies somewhat increases upon the transition from liquid droplets to
the fibrous looking particles. The median value of the mean circularity
of the assemblies (weighted by the size of the droplet/particle) drops
by ~60% between the first (day 1) and last (day 4) measurements
(Fig. 3b, left). A similar analysis of the 4:1 LCD: vRNA sample revealed a
greater increase in the mean area of the assemblies upon 4 days of
incubation, and a greater decrease of ~80% in the median value of the
mean circularity (Fig. 3b, right), suggesting a massive transition of
circular liquid droplet into large, amorphous, solid-like particles.
Quantification of the mean ThS fluorescence from images taken at
0.5 h and 4 days of incubation of both samples show a ~4-fold increase
in ThS fluorescence in the 40:1 LCD:S2hp sample, and ~58-fold increase
in fluorescence intensity in the 4:1 LCD:S2hp ratio (Fig. 3c). Here too,
no fibrils could be detected by EM at the concentration and incubation
times used for the PS assay.
In a separate experiment, we also followed the aggregation of the
central LCD segment when incubated alone or with S2hp vRNA (in 4:1
respective ratio) by measuring turbidity (Fig. 3d). We detected ele-
vated turbidity of the LCD + vRNA sample at the beginning of the
measurement, as opposed to the LCD only sample that was not turbid.
This offers that the central LCD immediately aggregates upon mixing
with vRNA. The LCD + vRNA sample shows biphasic behavior, with a
decrease in turbidity between 0 and 5.5 h of incubation, followed by a
renewed increase. This biphasic behavior of the 4:1 LCD:S2hp vRNA
sample may be related to the transition from liquid droplets to solid
particles visualized in this sample between 2 and 6 h of incubation
(Fig. 3a). Overall, our results indicate that the central LCD of NCAP
forms ThS-positive PS droplets that transition from circular liquid
droplets to fibrous or amorphous solid-like particles, and that the RNA
concentration governs the kinetics of this process and the morphology
of the assemblies.
Structures of LCD-derived steric-zipper-forming segments
To interfere with the self-assembly of the LCD segment, and thereby
possibly of NCAP, we seek structural information of specific amyloid-
like LCD sequences. Amyloid fibrils are stabilized by pairs of tightly
mating β-sheets, with zipper-like interfaces termed steric zippers that
can be predicted by a computer algorithm26 [https://services.mbi.ucla.
edu/zipperdb/]. Within the central LCD, we identified (Supplementary
Fig. 3a) and crystallized three such steric zipper-forming segments:
179GSQASS184, 217AALALL222, and 243GQTVTK248. X-ray structures con-
firmed that each segment forms amyloid-like fibrils composed of pairs
of β-sheets stabilized by steric zipper interfaces (Fig. 4, and
Nature Communications |
(2023) 14:2379
4
Article
https://doi.org/10.1038/s41467-023-37865-3
Fig. 3 | The central LCD segment of NCAP forms ThS fluorescent PS droplet that
transition into amorphous and fibrous solid-like particles. a Brightfield (BF) and
Thioflavin-S (ThS) fluorescence (green) microscopy images of 40:1 and 4:1 LCD:
hairpin-Site2 (S2hp) vRNA molar ratio mixtures incubated for ~0.5, 2, 6 h and 4 days.
b Mean area (purple) and mean circularity (blue; normalized to particle size) of
droplets and particles quantified from a series of light microscopy images of 40:1
and 4:1 LCD: S2hp mixtures imaged at day 1–4 of incubation. The experiment was
performed in three biological repeats, each with technical triplicates. Five images
were collected for every technical replicate. Boxplots show the 25th percentile,
median, and 75th percentile of the mean values for triplicate experiments. The
whiskers extend to the most extreme data points. Observations beyond the whisker
length, shown as circles, are values more than 1.5 times the interquartile range
beyond the bottom or top of the box (n = 9 replicates). c Mean ThS fluorescence
signal measured from background-subtracted fluorescence microscopy images
taken from 40:1 and 4:1 LCD: S2hp mixtures at 0.5 h (white) and 4 days (pink) of
incubation. The experiment was performed in three biological repeats, each with
technical triplicates. Five images were collected for every technical replicate. Data
from all repeats were combined for the quantification. The dots are of individual
data points and the bars represent mean values ± SEM (n = 45 images). Statistical
significance was calculated in Prism using an unpaired two-tailed t-test with Welch’s
correction. The p values are indicated with numbers and stars—****p < 0.0001.
Welch’s corrected t = 5.377/ 8.597 and df = 46.59/44.33 for 40:1 and 4:1 LCD: S2hp
samples, respectively. d Time-dependent shift in turbidity of LCD only (pink) and
4:1 LCD: S2hp (purple) solutions evaluated by measuring absorbance at 600 nm.
Source data for panels b–d are provided as a Source Data file.
Supplementary Figs. 4 and 5; Table 1). GSQASS and GQTVTK segments
both form parallel, in-register β-sheets, whereas the AALALL segment
is crystalized in two forms, both with antiparallel β-sheets27. The
weaker zipper interface of the second form incorporates polyethylene
glycol (Supplementary Fig. 4), and we do not consider it further.
Solvation-free energy calculations based on our crystal structures
(Supplementary Table 2) suggest that the AALALL steric-zipper is the
most stable of the three, consistent with its predominance of hydro-
phobic residues. GSQASS and GQTVTK, on the contrary, contain
mostly polar residues (Fig. 4c). The AALALL segment also overlaps with
a region predicted to participate in context-dependent interactions of
NCAP (Supplementary Fig. 3b, residues 216–221), namely interactions
that change between disordered and ordered modes as a function of
cellular environment and protein interactors and are likely to be
Nature Communications |
(2023) 14:2379
5
Article
https://doi.org/10.1038/s41467-023-37865-3
Fig. 4 | Atomic structures of amyloid-like association of NCAP segments
revealed by crystallography. a Quality of the fit of each atomic model to its
corresponding simulated annealing composite omit maps92. The maps are con-
toured at the 1.0 sigma level. All structural features are well defined by the density.
The view is down the fibril axis. Each chain shown here corresponds to one strand in
a β-sheet. Thousands of identical strands stack above and below the plane of the
page making ~100 micron-long β-sheets. The face of each β-sheet of AALALL (PDB
7LTU) [https://doi.org/10.2210/pdb7LTU/pdb] (form 1) is symmetric with its back.
However, GSQASS (PDB 7LV2) [https://doi.org/10.2210/pdb7LV2/pdb] and
GQTVTK (PDB 7LUZ) [https://doi.org/10.2210/pdb7LUZ/pdb] each reveal two dis-
tinct sheet–sheet interfaces: face-to-face and back-to-back. The tighter associated
pair of sheets is shown in this figure. b 18 strands from each of the steric zippers at a
view nearly perpendicular to the fibril axis. GSQASS and GQTVTK are parallel, in-
register sheets, mated with Class 1 zipper symmetry. The AALALL zippers are
antiparallel, in register sheets, mated with Class 7 zipper symmetry. Trifluoroacetic
acid (TFA) appears in the AALALL-form 1 steric zipper, and polyethylene glycol
(PEG) binds form 2 (PDB 7LUX [https://doi.org/10.2210/pdb7LUX/pdb] (form 2);
Supplementary Fig. 4). Carbon atoms in a and b are shown in orange and het-
eroatoms are colored by atom type. Water molecules are shown as red dots. c Steric
zipper structures (same order as in a) viewed down the fibril axis with residues
colored according to the Kyte and Doolittle hydrophobicity scale (bottom right)
shown with semi-transparent surface representation to emphasize the shape
complementarity and tight fit between the β-sheets. Images in c were rendered with
UCSF Chimera90. A stereo view of all structures is given in Supplementary Fig. 5.
Nature Communications |
(2023) 14:2379
6
Article
https://doi.org/10.1038/s41467-023-37865-3
Table 1 | Crystallographic data collection and refinement statistics from SARS-CoV-2 NCAP segments
179GSQASS184
217AALALL222 Form 1
217AALALL222 Form 2
243GQTVTK248
APS 24-ID-E
P212121
1.30 (1.39–1.30)a
APS 24-ID-E
P1
1.12 (1.18–1.12)
APS 24-ID-E
P21212
1.30 (1.36–1.30)
APS 24-ID-E
P21
1.10 (1.17–1.10)
4.77, 13.60, 42.44
9.45, 11.34, 20.27
44.46, 9.54, 10.95
19.57, 4.78, 22.03
90.0, 90.0, 90.0
74.9, 79.1, 67.8
90.0, 90.0, 90.0
90.0, 94.0, 90.0
Segment
Data collection
Beamline
Space group
Resolution (Å)
Unit cell dimensions:
a,b,c (Å)
Unit cell angles: α,β,γ (°)
Measured reflections
Unique reflections
1833 (338)
809 (139)
Overall completeness (%)
93.2 (95.9)
Overall redundancy
2.3 (2.4)
Overall Rmerge
CC1/2
Overall I/δ
Refinement
Rwork/Rfree
0.126 (1.04)
99.7 (56.7)
3.5 (0.7)
0.259/0.253
RMSD bond length (Å)
0.015
RMSD angle (°)
Number of segment atoms
Number of water atoms
Number of other
solvent atoms
Average B-factor of
peptide (Å2)
Average B-factor of
water (Å2)
Average B-factor other
solvent (Å2)
PDB ID code
1.4
40
2
0
12.3
19.9
N/A
5371 (323)
2270 (136)
78.4 (31.1)
2.4 (2.4)
0.084 (0.397)
98.5 (89.2)
5.9 (2.0)
0.158/0.197
0.009
1.3
180b
1
21
12.3
12.8
20.8
4666 (550)
1234 (139)
93.0 (84.8)
3.8 (4.0)
0.105 (0.808)
99.7 (54.4)
5.9 (1.8)
0.217/0.248
0.010
1.6
40
1
14
14.2
26.6
27.3
4677 (344)
1726 (170)
87.1 (50.9)
2.7 (2.0)
0.085 (0.446)
99.5 (84.3)
6.0 (1.4)
0.133/0.177
0.009
1.5
93b
12
0
8.2
24.7
N/A
7LV2 [https://doi.org/10.
2210/pdb7LV2/pdb]
7LTU [https://doi.org/10.2210/
pdb7LTU/pdb] (form 1)
7LUX [https://doi.org/10.2210/
pdb7LUX/pdb] (form 2)
7LUZ [https://doi.org/10.
2210/pdb7LUZ/pdb]
aNumbers in parentheses report statistics in the highest resolution shell.
bCount includes hydrogen atoms.
responsible for the formation of amyloid fibrils within liquid droplets28.
For drug design, we pursued AALALL and GQTVTK as targets but
excluded GSQASS because it resembles LCDs found in the human
proteome29.
A structure-based disruptor of NCAP’s PS exhibits antiviral
activity
To modulate NCAP’s self-assembly we exploited the propensity of
NCAP’s LCD to form steric-zipper structures. Guided by our amyloid-
spine structures we screened an array of peptides, each designed to
interact with a specific steric-zipper forming segment. We have found
such peptides to inhibit the aggregation and prion-like seeding of
other amyloid-forming proteins (e.g. refs. 30–33). To design the steric-
zipper targeting disruptors of NCAP self-assembly we implemented
two approaches: sequence/structure-based design and Rosetta-based
modeling34. Both approaches produce sequences that bind strongly to
our steric zipper structure targets and contain bulky residues that
block the interactions of additional NCAP molecules via this inter-
face (Fig. 5a).
Screening of a panel of our designed peptides in vitro revealed
that a peptide we named G12 disrupts NCAP’s PS. G12 is a D-amino acid
peptide with the sequence d-(rrffmvlm), designed against the AALALL
steric zipper-forming segment (Fig. 5a; Supplementary Table 3).
Increasing concentrations of G12 disrupt the formation of circular
NCAP PS droplets and instead promote the formation of large
network-like aggregates as judged by light microscopy (Fig. 5b, c and
Supplementary Fig. 6).
We then proceeded to test G12’s antiviral activity in HEK293 cells
that express the human ACE2 receptor (HEK293-ACE2 cells). First, we
verified that HEK293-ACE2 cells transfected with FITC-labeled
G12 show that G12 remains soluble and diffuse in the cytoplasm for
at least 24 h (Supplementary Fig. 7). Next, we used quantitative
immunofluorescence labeling to detect the percentage of SARS-CoV-2
infection in cells transfected with increasing concentrations of G12 or a
vehicle only control. The percentage of infected cells in each G12-
treated culture was normalized to the infected vehicle-only control
(Fig. 5d and Supplementary Fig. 8). Cytotoxicity was tested with the
same cells and G12 concentrations using the LDH toxicity assay
(Fig. 5d, red curve). Whereas G12 concentrations lower than 6 μM
slightly increase the relative percent infectivity of treated cells, in the
range of ~6–16 μM, G12 exhibits dose-dependent antiviral activity while
reducing the amount of virus detected in the culture by up to ~50%
without inflicting cytotoxicity (Fig. 5d and Supplementary Fig. 8). Since
G12 is dissolved in DMSO we could not test higher G12 concentrations
in this cell-based assay to obtain a complete dose–response curve,
however fitting a non-linear regression model to our data allowed a
rough IC50 estimation of 7–11 μM (Fig. 5d). We, therefore, suggest that
G12 serves as a proof-of-concept showing that by targeting amyloi-
dogenic segments within the central LCD of NCAP we interfere with
NCAP’s self-assembly and thereby the viral life cycle.
Discussion
The NCAP protein of SARS-CoV-2 belongs to the subclass of fibril-
forming proteins that contains both an RNA-binding domain and LCDs
Nature Communications |
(2023) 14:2379
7
Article
https://doi.org/10.1038/s41467-023-37865-3
Fig. 5 | Design and evaluation of NCAP’s self-assembly disruptor, G12. a The
Rosetta-based design of G12 templated by the AALALL X-ray crystal structure form 1
(Fig. 4; Table 1). Model of the G12 (blue) capping an AALALL fibril (orange). The top
view is down the fibril axis and the side view is tilted from the axis. Additional
AALALL strands are shown adjacent to the bound G12 to illustrate their steric
clashes (magenta). b Differential interference contrast (DIC) images of NCAP +
S2hp mixtures incubated in the absence (0:1) and presence of increasing con-
centrations of G12 revealing the PS disrupting activity of G12. c Mean area (purple)
and mean circularity (blue; normalized to particle size) of droplets and particles
quantified from a series of light microscopy images of NCAP + S2hp mixtures with
increasing concentrations of G12. The experiment was performed in three
biological repeats, each with technical triplicates. Five images were collected for
every technical replicate. A representative plot is presented. In boxplots the central
mark indicates the median, and the bottom and top edges of the box indicate the
25th and 75th percentiles, respectively. The whiskers extend to the most extreme
data points (n = 3 replicates). d Dose–response analysis of HEK293-ACE2 cells
treated with 10 different concentrations of G12 and fitted with a nonlinear regres-
sion model (black line). The 95% confidence interval of the IC50 for G12 was esti-
mated to be between 7 and 11 μM. Cytotoxicity testing of G12 at each concentration
(red line) indicates that G12 is non-toxic. Each dot represents the mean value of
three technical replicates. Source data for panels c and d are provided as a Source
Data file.
(Fig. 1a, b). NCAP undergoes PS1–8 (Supplementary Fig. 2a), and as we
show here, its central LCD forms amyloid-like fibrils (Figs. 1 and 2).
The central LCD of NCAP forms fibrils with the long and
structured S2hp vRNA segment (Fig. 1 and Supplementary Fig. 1;
Supplementary Table 1), with various short, single-stranded RNA
sequences (Fig. 2; Supplementary Table 1), and also with no RNA
(Fig. 1f). This suggests that specific LCD–RNA interactions are not
required for LCD-amyloid formation. Nevertheless, the LCD does bind
to at least S2hp vRNA25, and LCD fibril maturation is influenced by the
RNA sequence and length (Fig. 2b), so LCD–RNA interactions play a
role. The LCD segment is highly positively charged (Fig. 1b), especially
in its non-phosphorylated form. Therefore we expect it to engage
in non-specific polar interactions with the negatively charged RNA,
which in turn may promote the accumulation of LCD molecules,
including through PS formation (Fig. 3), and their amyloid-like
assembly (Fig. 2a).
The amyloid-like characteristics of the central LCD of NCAP are
similar to those of the LCDs of FUS35,36, hnRNPA237, TDP-4338, and other
RNA-binding proteins that are involved in RNA metabolism in eukar-
yotic cells13,17, and under certain circumstances, also in amyloid-
associated pathologies19. This equivalent ability of the LCD of NCAP to
PS and stack into amyloid-like structures in the presence of RNA pro-
poses its potential function in the yet elusive mechanism of NCAP self-
assembly.
Full-length NCAP is capable of only sparse fibril formation in the
presence and absence of S2hp vRNA and with ZnCl2 (Supplementary
Fig. 2). Whereas fibrils formed in the presence of S2hp do not exhibit
amyloid-typical morphology (Supplementary Fig. 2b, c), the NCAP +
S2hp aggregates produce a ThT amyloid formation curve, but it lacks a
lag phase (Supplementary Fig. 2e). Short or absent lag phase in ThT
curves may result from the existence of pre-formed amyloid seeds in
the tested sample39, or from a fast pickup of the ThT signal prior to
Nature Communications |
(2023) 14:2379
8
Article
https://doi.org/10.1038/s41467-023-37865-3
starting the measurements. The latter may be reasonable given that
NCAP rapidly aggregates and becomes turbid in the presence of S2hp
(Supplementary Fig. 2f). In a parallel study, we show that the struc-
tured regions of S2hp are essential for strong binding to NCAP,
whereas S2 and other short, single-stranded RNA segments bind to it
weakly25. Here, we detected fibrils of NCAP with S2hp (Supplementary
Fig. 2b, c), but its LCD segment is also capable of forming fibrils in the
presence of the short, unstructured vRNA segments S1, S1.5, and S2
(Fig. 2b; Supplementary Table 1). We, therefore, speculate that robust
amyloid formation of full-length NCAP requires strong interactions
with specific vRNA sequences and/or co-factors that we are yet to
identify.
The amyloid formation of the central LCD of NCAP is attributed to
at least three adhesive peptide sequences: 179GSQASS184, 217AALALL222,
and 243GQTVTK248 (Fig. 4 and Supplementary Fig. 3a). 179GSQASS184
and 243GQTVTK248, are predominantly polar (Fig. 4c), similar to the
highly polar reversible amyloid fibrils formed by the LCDs of FUS and
hnRNPA240. The segment 179GSQASS184 is part of a conserved serine/
arginine (SR)-rich region (residues 176–206)4 and it includes the two
phosphorylation sites S180 and S18411. Phosphorylation of the SR-rich
region facilitates the transformation of NCAP’s PS droplets from a solid
to a liquid-like state during viral genome processing. The non-
phosphorylated protein, however, is associated with solid PS dro-
plets and nucleocapsid assembly8. Both S180 and S184 face the dry,
tight interface formed between the β-sheets in the structure of
179GSQASS184 (Fig. 4a). Phosphorylation of those residues is indeed
likely to reverse the solid, amyloid-like packing of this segment. Of
note, all results in this paper showing the ordered, solid-like mode of
aggregation were obtained with non-phosphorylated proteins and
peptides.
The second adhesive segment, 217AALALL222, is highly hydro-
phobic and produces the most stable steric-zipper structure (Supple-
mentary Table 2). 217AALALL222 is also predicted to help switch
between disordered and ordered modes of protein aggregation as a
factor of cellular environment and protein interactors (Supplementary
Fig. 3c, residues 216–221)28. Those properties of 217AALALL222 render it
an important target for the disruption of NCAP’s self-assembly. The
243GQTVTK248 segment, however, resembles sequences in LCDs found
in the human proteome29, and is therefore a poor target for drug
design.
The self-assembly of NCAP is crucial for RNA packaging and SARS-
CoV-2 replication10. The amyloid formation of NCAP’s LCD is a form of
NCAP self-assembly, but it is yet unclear whether NCAP forms and
functions as amyloid in the viral life cycle. Nevertheless, PS-mediated
self-assembly of NCAP was shown to occur in NCAP-transfected and
SARS-CoV-2 infected cells4,7,11,41. By targeting the amyloidogenic
segment 217AALALL222 (Fig. 4 and Supplementary Figs. 4 and 5) with
G12, we inhibited the PS formation of NCAP in vitro (Fig. 5b, c). G12 is a
peptide designed to interact and block the 217AALALL222 interface by
exploiting the tendency of this segment to form steric-zipper struc-
tures (Fig. 5a). G12 is, however, incapable of complete disruption of
NCAP self-assembly, perhaps because assembly is guided by several
proteins interfaces10. Evaluation of G12 in SARS-CoV-2-infected cells
revealed dose-dependent antiviral activity in concentrations of
6–16 μM without inflicting cytotoxicity (Fig. 5d). G12 concentrations
lower than 6 μM, however, led to increased viral infection in treated
cells. We speculate that when administered in subeffective con-
centrations, G12 partitions into NCAP droplets and increases NCAP’s
effective concentration which possibly promotes self-assembly and
formation of new virions. When administered in proper concentra-
tions, we anticipate that the antiviral activity of G12 results from its
interference with the self-assembly of NCAP, as designed, leading to
poor RNA packaging and viral particle assembly.
The three steric-zipper-forming segments we identified in this
work are conserved between the NCAPs of SARS-CoV-2 and SARS-CoV.
The only exception is alanine in position 217 in the sequence of SARS-
CoV-2 which is replaced by threonine in the NCAP of SARS-CoV (Sup-
plementary Fig. 9a). A ZipperDB26
[https://services.mbi.ucla.edu/
zipperdb/] calculation on the LCD of the NCAP of SARS-CoV revealed
that this threonine shifts the steric-zipper forming segment to the
hydrophobic ALALLL sequence (with Rosetta free energy score of
−24.700) that is aligned and conserved with residues 218–223 in the
NCAP of SARS-CoV-2. This suggests that the LCD in the NCAP of SARS-
CoV may also form amyloids, and that future SARS coronaviruses
might share this targetable property. A SEG analysis22 performed on
the sequence of the NCAPs of a number of α, β and γ coronaviruses
from various species showed that many of these viruses contain LCDs
that could potentially participate in amyloid formation (Supplemen-
tary Fig. 9b). This suggests that amyloid formation of NCAP LCDs is a
general mechanism of action and a common targetable trait in
coronaviruses.
Despite the high conservation of NCAP10, some mutations have
been identified in strains that emerged since the initial SARS-CoV-2
outbreak in Wuhan, China. To date, no NCAP mutations were detected
179GSQASS184,
within our amyloid steric-zipper spine segments:
217AALALL222, and 243GQTVTK248. Nevertheless, some mutations were
detected within the central LCD, including the prevalent R203K/M,
G204R/M, and T205I substitutions42–44. The R203K/G204R mutants
exhibit higher PS propensity compared to the Wuhan variant41, and the
R204M mutation promotes RNA packaging and viral replication in the
delta variant45. Also interesting are the G214C (Lambda variant) and
G215C (Delta variant) substututions42–44 that are adjacent to the
217AALALL222 steric-zipper segment. The Delta variant spread faster
and caused more infection compared to its predecessors46–49. The
Delta variant also carries a D377Y mutation in the C-terminal LCD
segment of NCAP. It is possible that mutations in NCAP’s LCD enhance
amyloid formation, similarly to mutations in other RNA-binding
proteins35,50–52. This is important to explore since amyloid fibrils are
associated with numerous dementias and movement disorders53,54.
Amyloid cross-talk and hetero-amyloid aggregation,
including
between microbial and human amyloid proteins (e.g. refs. 55–58), is a
well-known phenomenon that is postulated to exacerbate amyloid
pathology59.
The possible connection of amyloid formation of NCAP to neu-
rodegeneration was already recently suggested. NCAP was shown to
interact and accelerate the amyloid formation of the Parkinson’s
disease-related protein, α-synuclein, which may explain the correlation
between Parkinsonism and SARS-CoV-2 infection60. NCAP was also
shown to partition into PS droplets5 and accelerate amyloid
formation61 of FUS, TDP-43, hnRNPA1, and hnRNPA2.
In certain
forms, those proteins are associated with neurodegenerative and
movement disorders19. In SARS-CoV-2 infected cells, NCAP impairs
the disassembly of stress granules into which it partitions, and in
cells expressing an ALS-associated mutant of FUS, NCAP enhances
FUS aggregation into amyloid-containing puncta61. Those observa-
tions, together with the capacity of NCAP’s central LCD to form amy-
loid, call for further investigation of the possible NCAP-amyloid
formation and regulation in SARS-CoV-2-infected cells, and of the
possible involvement of NCAP in amyloid cross-talk and human
neurodegeneration.
Our study of the amyloid formation of NCAP expands an emer-
ging class of known amyloid-forming viral proteins. In the Influenza A
virus, the full-length and N-terminal segment of the PB1-F2 protein
form cytotoxic amyloid fibrils when mixed with liposomes, and the
C-terminal segment forms cytotoxic amyloid oligomers62. A 111-residue
segment from the V protein of Hendra virus, a respiratory virus that
may progress in humans to severe encephalitis, was shown to undergo
a liquid-to-hydrogel transition of its PS droplets and to produce
amyloid-like fibrils63. The RIP-homotypic interaction motif containing
segments of the herpes simplex virus 1 (HSV-1) protein ICP664, the
Nature Communications |
(2023) 14:2379
9
Article
https://doi.org/10.1038/s41467-023-37865-3
murine cytomegalovirus protein M4565 and the varicella-zoster virus
protein ORF2066 are capable of forming heteromeric amyloid com-
plexes with host proteins. Other examples of amyloid-forming peptide
segments include avibirnavirus viral protease that contributes to
from the fiber protein of
protease self-assembly67, peptides
adenovirus68,69, and a nine-residue peptide from the C-terminus of the
SARS-CoV envelope protein70. Recent studies also showed the amy-
loidogenic properties of various segments of the spike protein71, and
other regions in the proteome72 of SARS-CoV-2. None of these pre-
viously studied viral amyloids, however, was associated with NCAPs.
Nevertheless, LCDs and prion-like sequences, such as those that
exist in NCAP5 were identified in over two million eukaryotic viruses73.
Therefore, our finding of the amyloid formation of this viral
RNA-binding protein may foreshadow a much wider field for
investigation.
In summary, this work extends knowledge of amyloidogenic viral
proteins and their LCD segments, associates NCAP with known
amyloid-forming RNA-binding proteins, and may inspire future inves-
tigation of NCAP amyloid formation in SARS-CoV-2 infection. Finally,
we also suggest an approach for the development of SARS-CoV-2
therapeutics via disruption of NCAP self-assembly by targeting and
capping amyloid-driving steric-zipper segments of NCAP.
Methods
Molecular biology reagents
Phusion HF DNA polymerase, Quick Ligase, and restriction enzymes
were purchased from New England BioLabs. Custom DNA oligonu-
cleotides were synthesized by IDT (Coralville, IA). RNA oligonucleo-
tides, S1, S1.5, S2, and the non-specific RNA (siDGCR8-1, antisense
strand) were synthesized by Horizon Discovery Biosciences.
Computational predictions and sequence alignment
Prediction of low-complexity sequences in the NCAP of SARS-CoV-2.
The amino acid sequence of the Nucleocapsid protein of SARS-CoV-2
(NCAP; UniProtKB74 accession number: P0DTC9 [https://covid-19.
uniprot.org/uniprotkb/P0DTC9#Sequence]) was evaluated using
SEG22 with default settings: window length = 12, trigger complexity 2.2,
extension complexity 2.5. LCDs were defined by strings of at least 10
low-complexity residues. Long LCDs, such as the central NCAP-LCD,
were allowed no more than five interrupting non-low-complexity
residues between strings of 10 or more low-complexity residues22.
Prediction of LCDs in the NCAPs of various coronaviruses. A list of
coronavirus Nucleocapsid proteins was downloaded from the Eur-
opean Nucleotide Archive (ENA; [https://www.ebi.ac.uk/genomes/
virus.html]), and protein sequences were retrieved from Uniprot
[https://www.uniprot.org/]. Low complexity residues were identified
using the SEG algorithm22 with default parameters (see above).
Redundant low-complexity region sequences from strains of indivi-
dual viruses were removed. Low-complexity region sequences were
aligned in BioEdit using the ClustalW algorithm with gap penalties set
to 100 in order to avoid the insertion of gaps in the aligned sequences.
Gaps consisting of hyphens in between amino acid stretches in an
individual sequence represent an interrupting, non-low-complexity
segment of at least 20% the length of the longest LCD in the protein
rather than defined gaps in the alignment. Some of these gaps were
manually made larger or smaller to achieve a more accurate alignment.
Supplementary Fig. 9b is the representation of this alignment in
Jalview.
Prediction of steric-zipper forming segments. This was done on the
Nucleocapsid proteins of SARS-CoV-2 (UniProtKB74 accession number:
number:
P0DTC9)
P59595) using the ZipperDB algorithm26 [https://services.mbi.ucla.
edu/zipperdb/].
(UniProtKB74
SARS-CoV
accession
and
Prediction of PS forming regions and context-dependent interac-
tions. Was performed using the Fuzdrop algorithm28 [https://fuzdrop.
bio.unipd.it/predictor] on the Nucleocapsid protein of SARS-CoV-2
(UniProtKB74 accession number: P0DTC9 [https://covid-19.uniprot.
org/uniprotkb/P0DTC9#Sequence]).
Sequence conservation
Sequence conservation analysis was performed on the LCDs of the
NCAPs of SARS-CoV and SARS-CoV-2 (UniprotKB74 accession numbers:
P59595 and P0DTC9, respectively). The sequences were aligned and
colored according to conservation in Jalview.
Construct design
Full-length SARS-CoV-2 Nucleocapsid protein gene and its fragments
were PCR amplified from 2019-nCoV Control Plasmid (IDT Inc., cat. no.
10006625) and spliced with N-terminal 6xHis-SUMO tag75 using spli-
cing by overlap extension (SOE) technique76. 5’ KpnI and 3’ SacI
restriction sites introduced with the flanking primers were used to
ligate the resulting fragments into pET28a vector. When needed, an
additional round of SOE was performed to generate internal Nucleo-
capsid protein deletion mutants. Construct sequences were confirmed
by Sanger sequencing (Laragen, Culver City, CA). Primers used for
cloning are given in Supplementary Table 4, DNA sequences and
alignment of translated amino acid sequences from Sanger sequencing
are given in Supplementary Figs. 10 and 11, respectively.
Protein expression and purification
NCAP segments and full-length protein were expressed as fusions to
6xHis-SUMO (6xHis-SUMO-NCAP). Plasmids were transformed into
Escherichia coli Rosetta2 (DE3) strain (MilliporeSigma cat. no 71-397-4)
and small-scale cultures were grown at 37 °C overnight in LB with
35 μg/mL kanamycin and 25 μg/mL chloramphenicol. TB with 35 μg/μL
kanamycin was inoculated with overnight starter culture at a 1:100
ratio and large-scale cultures were grown at 37 °C with 225 rpm shaking
until the OD600 reached ~0.6. Protein expression was induced with
1 mM IPTG and cultures were further incubated with shaking at 28 °C
overnight, then harvested at 5000×g at 4 °C for 15 min. Bacterial pellets
were either used right away or stored at −20 °C. Pellets from 2–4 L of
culture were re-suspended in ~200 mL chilled Buffer A (20 mM Tris pH
8.0, 1 M NaCl) supplemented with Halt Protease Inhibitor Cocktail
(ThermoScientific cat. no. 87785) and sonicated on the ice at 80%
amplitude for a total sonication time of 15 min, with pauses at regular
intervals so the sample does not exceed 15 °C. Cell debris was removed
via centrifugation at 24,000×g at 4 °C for 30- 60 min, filtered twice
through 0.45 μm high particulate syringe filters (MilliporeSigma cat.
no. SLCRM25NS), and imidazole added to 5 mM. Filtered clarified
lysate was loaded onto HisTrap HP columns (GE Healthcare) and
proteins were eluted over a step-gradient with Buffer B (20 mM Tris pH
8.0, 1 M NaCl, 500 mM imidazole), with extensive low-imidazole
(<20%) washes to improve purity. NCAP proteins were generally
eluted in 20–50% Buffer B. Fractions were analyzed by SDS–PAGE,
pooled and dialyzed against 20 mM Tris pH 8.0, 250 mM NaCl at 4 °C
overnight. Following dialysis, the sample was concentrated using
Amicon Ultra-Centrifugal filters (MilliporeSigma) and urea was added
up to 1 M final concentration if protein precipitation was observed.
Ulp1 protease (homemade) was added at a 1:100–1:200 w/w ratio to
purified proteins, along with 1 mM DTT, and the sample was incubated
at 30 °C with 195 rpm shaking for 1–2 h. After cleavage, NaCl was added
to 1 M final concentration to reduce aggregation, and the sample was
incubated with HisPur Ni-NTA resin (Thermo Scientific cat. no.
PI88222) equilibrated in Buffer A at 25 °C with 140 rpm shaking for
30 min. Cleaved NCAP proteins were eluted from the resin via gravity
flow chromatography, then the resin was washed twice with Buffer A,
twice with Buffer A + 5 mM imidazole, and finally with Buffer B. The
flow-through and appropriate washes were concentrated and flash-
Nature Communications |
(2023) 14:2379
10
Article
https://doi.org/10.1038/s41467-023-37865-3
frozen for storage or further purified by gel filtration. Directly prior to
gel filtration, the sample was centrifuged at 21,000×g for 30 min at 4 °C
to remove large aggregates. Soluble protein was injected on a HiLoad
Sephadex 16/600 S200 (for proteins larger than ~25 kDa) or S75 (for
proteins smaller than ~25 kDa) (GE Healthcare) equilibrated in SEC
buffer (20 mM Tris pH 8.0, 300 mM NaCl) and run at a flow rate of
1 mL/min. Elution fractions were assessed by SDS–PAGE for purity, and
confirmed to have low RNA contamination as assessed by 260/280 nm
absorbance ratio. Pooled fractions were concentrated and 0.2-μm fil-
tered. Protein concentration was measured by A280 absorbance using
a NanoDrop One (ThermoScientific) and calculated by the sequence-
specific extinction coefficient, and aliquots were flash-frozen and
stored at −80 °C. Of note, the first N-terminal residue in all purified
proteins (residue #1) is a threonine remaining from cleavage of the
6xHis-SUMO tag during protein purification.
for
templates
the design of peptide inhibitors
Rosetta-based peptide inhibitor design
Crystal structures of LCD segments GSQASS and AALALL (form 1) were
used as
in
Rosetta3 software34. 5 layers of the steric zipper structure were gen-
erated. A 6-residue peptide chain was placed at the top or bottom of
the fibril-like structure. Rosetta Design was used to sample all amino
acids and their rotamers on the sidechains of the fixed peptide back-
bone. The lowest energy conformations of the sidechains were
determined by minimizing an energy function containing terms for
Lennard–Jones potential, orientation-dependent hydrogen bond
potential, solvation energy, amino acid-dependent reference energies,
and statistical torsional potential dependent on the backbone and
sidechain dihedral angles. Buried surface area and shape com-
plementarity were scored by AREAIMOL77 and Sc78, respectively, from
the CCP4 suite of crystallographic programs79. Solvation-free energy
estimates were calculated using software available here: [https://doi.
org/10.5281/zenodo/6321286]. Design candidates were selected based
on their calculated binding energy to the top or bottom of the fibril-
like structure, shape complementarity, and propensity for self-
aggregation. The binding energy for an additional strand of the
native sequence (i.e., AALALL) was computed for comparison with
peptide inhibitor designs. The structural model of each candidate
peptide was manually inspected in PyMOL80. Many computational
designs produced sequences with high hydrophobic content, thus two
arginine residues were added onto the N-terminal end to increase
peptide solubility. Candidate G12 was the most effective inhibitor in
preliminary screens and therefore was chosen for further evaluation.
Peptide synthesis and purification
The NCAP steric zipper segments 179GSQASS184 and 243GQTVTK248
were synthesized by LifeTein. The inhibitor candidate G12 was syn-
thesized by LifeTein and GenScript. All peptides were synthesized at
over 98% purity. The NCAP segment 217AALALL222 was synthesized and
purified in-house as H-AALALL-OH. Peptide synthesis was carried out
at a 0.1 mmol scale. A 2-chlorotrityl chloride resin (Advanced Chem-
tech) was selected as the solid support with a nominal loading of
1.0 mmol/g. Each loading of the first amino acid was executed by
adding 0.1 mmol of Fmoc-Leu-OH (Advanced Chemtech FL2350/
32771) and 0.4 mmol of diisopropylethylamine (DIPEA), dissolved in
10 mL of dichloromethane (DCM), to 0.5 g of resin. This mixture was
gently agitated by bubbling with air. After 30 min, the supernatant was
drained, and the resin was rinsed twice with 15 mL aliquots of the
capping solution, consisting of 17:2:1 DCM/MeOH/DIPEA. With the first
amino acid loaded, the elongation of each polypeptide was completed
in a CEM Liberty BlueTM Microwave Peptide Synthesizer. A 1.0 M solu-
tion of N,N’-diisopropylcarbodiimide (DIC) in DMF was used as the
primary activator, and a 1.0 M solution of ethyl cyanohydrox-
yiminoacetate (oxyma) in DMF, buffered by 0.1 M of DIPEA was used as
a coupling additive. The Fmoc-L-Ala-OH used was also purchased from
Advanced Chemtech (FA2100/32786). The microwave synthesizer
utilizes 0.2 M solutions of each amino acid. For the deprotection of N-
termini, Fmoc protecting groups, a 9% w/v solution of piperazine in 9:1
N-Methyl-2-Pyrrolidone to EtOH buffered with 0.1 M of oxyma was
used. For 0.1 mmol deprotection reactions, 4 mL of the above depro-
tection solution was added to the resin. The mixture was then heated
to 90 °C for 2 min while bubbled with nitrogen gas. The solution was
drained, and the resin was washed 4 times with 4 mL aliquots of DMF.
For 0.1 mmol couplings, 2.5 mL of 0.2 M amino acid solution
(0.5 mmol) was added to the resin along with 1 mL of the DIC solution
(1.0 mmol) and 0.5 mL of oxyma solution (0.5 mmol). This mixture was
agitated by bubbling for 2 min at 25 °C, then heated to 50 °C followed
by 8 min of bubbling. After the last deprotection, the resin was washed
with methanol, diethyl ether, dried over the vacuum, and introduced
to a cleavage cocktail consisting of 20 mL of trifluoroacetic acid (TFA),
0.50 mL of water, 0.50 mL of triisopropylsilane (TIS). After 2 h of vig-
orous stirring, the mixture was filtered, and the filtrate was con-
centrated in vacuo. The residue was triturated with cold diethyl ether,
and precipitated, the crude peptide was collected by filtration. The
crude peptide was then purified by RP-HPLC, using an Interchim pur-
iFlash® 4125 Preparative Liquid Chromatography System equipped
with a Luna (Phenomenex, C18(2), 5 μm, 100 Å, 30 × 100 mm) column.
For purification, two buffer systems were utilized. Initial purifications
and salt exchanges were executed with a 13 mM aqueous solution of
trifluoroacetic acid (TFA; [A]) and a 2:3 water to acetonitrile solution,
buffered by 13 mM of TFA ([B]). For the better resolution of diaster-
eomers and other impurities, ultrapure water, buffered by 14 mM of
HClO4, and a 2:3 water to acetonitrile solution, buffered by 5.6 mM of
HClO4, were selected as mobile phases A and B, respectively. The
purity of the purified fractions was analyzed by RP-HPLC, using an
Agilent 1100 Liquid Chromatography System equipped with a Kinetex
(Phenomenex, C18, 5 μm, 100 Å, 4.6 × 250 mm) column. Ultrapure
water with 0.1% TFA, and a 1:9 water to acetonitrile solution with
0.095% TFA were selected as mobile phases [A] and [B], respectively.
The flow rate was set at 1.0 mL/min and the gradient used is detailed in
Supplementary Table 5. The UV absorption at 214 nm was monitored.
The resulting chromatogram is shown in Supplementary Fig. 12.
RNA in vitro transcription and purification
The nucleic acid sequence corresponding to S2hp (Supplementary
Table 1) was cloned from a gBlock (IDT) of the first 1000 nucleotides of
the 5′-end of the SARS-CoV-2 genome into pUC19 vectors using the
restriction sites EcoRI and KpnI. Forward primer P2627 (5′-TAAT
ACGACTCACTATAGGCTGTGTGGCTGTCACTCG-3′) containing the T7
promoter sequence was added at a low concentration of 0.5 nM in
(5′-GCGAATTCTAATACGACT
addition to forward primer P1471
CACTATAGG-3′) containing the EcoRI restriction sequence and T7
promoter sequence at the normal concentration of 500 nM. Reverse
primer P2644 (5′-CGGGGTACCTCGTTGAAACCAGGGACAAG-3′) con-
taining the KpnI restriction sequence was added at 500 nM. The clone
was sequence-confirmed and the miniprep was used as a template for
PCR. The forward primer for PCR containing the T7 promoter
sequence was biotinylated on the 5′ end for removal of PCR template
after transcription. The PCR product was purified by HiTrap column.
The running buffer solutions (0.2-μm filtered) contained 2 M NaCl,
10 mM HEPES pH 7.0 (buffer A), and 10 mM NaCl, 10 mM HEPES pH 7.0
(buffer B). The purified PCR products were concentrated using Amicon
Ultra centrifugal filter units (Millipore) and buffer-exchanged against
10 mM Na/HEPES pH 7.0. Transcription reactions ranging from 5 to
100 mL were set up. The transcription reaction was incubated at 37 °C
with gentle shaking for one hour. After transcription, streptavidin
beads (ThermoFisher) were added to the transcription and set on a
rotator at room temperature for an additional 15 min. The transcrip-
tion reaction was centrifuged at 500×g for 10 min at 4 °C. The super-
natant was decanted and the pellet containing any PCR template
Nature Communications |
(2023) 14:2379
11
Article
https://doi.org/10.1038/s41467-023-37865-3
remaining was discarded. The transcription reaction was then purified
by 5 mL HiTrap Q HP column in several rounds, loading ~5 mL into the
column each round. The purified RNA was concentrated using Amicon
Ultra-15 (Millipore) and the buffer was exchanged into 10 mM HEPES
pH 7.0. The purity of the RNA was confirmed using denaturing poly-
acrylamide gels. The concentration was calculated by measuring
OD260 and a conversion factor of 40 μg/mL/OD260.
PS assays
All solutions were prepared using DNase/Rnase-free water (ultrapure
water) and were filtered twice using a 0.22-μm syringe filter. Prepara-
tions were done under sterile conditions and using sterile filter pipette
tips to prevent RNA degradation.
PS of NCAP with ThS staining (Supplementary Fig. 2a). Experiments
were carried out in 96-well black/clear glass-bottom plates (Cellvis
glass-bottom plates cat. no. P96-1.5H-N). S2hp RNA, stored at −20 °C
was thawed, then annealed by heating at 95 °C for 3 min and trans-
ferring quickly on the ice. The RNA was diluted by its original buffer of
10 mM HEPES pH 7.0 to 750 μM and 75 μM working solutions. 1 mM
ZnCl2 was prepared in ultrapure water and filtered twice with a 0.22-
μm syringe filter. Fresh Thioflavin S (ThS) solution was prepared from
powder (MP Biomedicals) in ultrapure water at 0.002% w/v and fil-
tered. Purified NCAP stock solution was centrifuged at 15,000×g for
15 min at 4 °C to remove large aggregates. NCAP, S2hp vRNA, and
ZnCl2 were mixed in PBS at final concentrations of 30 μM NCAP with 0
or 0.75 μM S2hp vRNA, and 0 or 20 μM ZnCl2 as indicated in the figure.
ThS was diluted into the wells to a final concentration of 0.0002% w/v.
Blank solutions containing everything but NCAP were prepared as
controls. After dispensing the samples the plates were immediately
covered with optical film (Corning Sealing Tape Universal Optical) and
incubated in a plate reader (BMG LABTECH FLUOstar Omega) at 37 °C
with 700 rpm shaking. The plates were imaged at indicated time points
of incubation. All samples were imaged with ZEISS Axio Observer D1
fluorescence microscope with ZEN 2 software, equipped with a 100x
oil objective lens, using the 1,4-Diphenylbutadiene fluorescence
channel with a DAPI filter for ThS, as well as a DIC filter. Images were
processed and rendered with FIJI (imageJ)81.
PS of the LCD segment with ThS staining (Fig. 3). S2hp and ThS
solutions were prepared as above. Purified LCD protein solution was
centrifuged at 15,000×g for 15 min at 4 °C to remove large aggregates.
The Protein, RNA, and ThS were then mixed in wells of 96-well black/
clear glass-bottom plate (Cellvis glass-bottom plates cat. no. P96-1.5H-
N) at 40:1 and 4:1 LCD: S2hp vRNA molar ratios in triplicates. ThS was
added to 0.0002% w/v final concentration. This experiment was
repeated with both 30 and 10 μM final LCD concentrations showing
similar results. Respective protein and RNA blank solutions were pre-
pared as controls. The plate was immediately covered with an optical
film (Corning Sealing Tape Universal Optical) and incubated at 37 °C
with 700 rpm shaking in a plate reader (BMG LABTECH FLUOstar
Omega). Images were obtained at indicated time points and processed
as above.
PS of NCAP with G12 (Fig. 5). Directly prior to assay setup, purified
NCAP protein was centrifuged at 15,000×g for 15 min at 4 °C to remove
large aggregates and the supernatant was used for the experiment.
S2hp RNA was briefly annealed by heating at 95 °C for 3 min and
transferring quickly on the ice. G12 stock solutions were prepared in
DMSO in 1 mM concentration from lyophilized peptide powder and
serially diluted in PBS buffer complemented with 10 % DMSO and was
added to wells of 384-well black/clear glass-bottom plate containing
10 μM NCAP protein and 0.25 μM S2hp RNA (40:1 molar ratio) in PBS
buffer. NCAP: G12 (or buffer control) molar ratios are indicated in the
figure. The final DMSO concentration in all wells was 1%. The plate was
covered with optical film (Corning Sealing Tape Universal Optical) and
incubated for ~4 h at room temperature without shaking prior to
imaging. Images were acquired using an Axio Observer D1 microscope
(Zeiss) with ZEN 2 software, equipped with a ×100 oil objective lens
using a DIC filter. Images were processed and rendered with FIJI
(imageJ)81. Mean area and mean circularity (weighted by particle size)
of particles and droplets were calculated using MATLAB as described
in the Brightfield Image Segmentation and Shape Analysis section.
PS with FITC labeled G12 (Supplementary Fig. 6). FITC-labeled
G12 stock solution (made in DMSO) was added to non-labeled stocks at
a 1:9 labeled:non-labeled ratio. The mixture was then added to a 96-
well plate with glass bottom at a final concentration of 10 μM NCAP,
0.25 μM S2hp RNA (40:1 molar ratio), and 0 or 10 μM G12 in 20 mM Tris
pH 8, 50 mM NaCl, and 20 μM ZnCl2. The final DMSO concentration in
all wells was 0.5 %. The plate was covered with optical film (Corning
Sealing Tape Universal Optical) and incubated at 37 °C without shaking
for 24 h prior to imaging with ZEISS Axio Observer D1 fluorescence
microscope with ZEN 2 software, equipped with a ×100 oil objective
lens, using the FITC fluorescence channel with a GFP filter and a DIC
filter. Images were processed and rendered with FIJI (imageJ)81.
Measurements of ThS fluorescence in LCD PS droplets (Fig. 3c)
The PS experiment of the LCD segment with ThS staining was per-
formed as described above. To evaluate the change in ThS fluores-
cence upon incubation of the PS droplets we combined for each
experimental condition and time point 5 fluorescence images per well
from triplicate wells and 3 biological repeats (n = 45 images per con-
dition per time point). Background fluorescence was subtracted indi-
vidually from each image using FIJI after measuring the mean gray
value and STD of a region containing no features of interest and cal-
culating it with Eq. (1):
Background fluorescence signal = 3 × STD + mean gray value
ð1Þ
Then the mean fluorescence (gray value) of the entire background
subtracted image was measured and averaged across all images from
the same condition and time point. The plot was rendered in Prism
software and error bars represent standard error of the mean. Two-
tailed t-test with Welch’s correction was performed in Prism to evalu-
ate statistical significance of the change in ThS fluorescence between
time points of each condition. Mean area and mean circularity
(weighted by particle size) of particles and droplets were calculated
using MATLAB as described below in the Brightfield Image Segmen-
tation and Shape Analysis section.
Image segmentation and Shape analysis (Figs. 3 and 5)
Brightfield microscopy images were imported into MATLAB 9.13.0
(R2022b) where all subsequent processing and image analysis were
carried out. Image segmentation was carried out by initial Gaussian
filtering of each image to achieve local smoothing of the image data.
Each image was filtered using a Gaussian kernel with a standard
deviation of 5 pixels (px). The Laplacian of the Gaussian-filtered image
was then found to highlight areas of rapid change in intensity to
facilitate edge detection. Edge detection was performed on each
image by finding points of maximum local gradients, using the Sobel
approximation to derivatives that are implemented using the MATLAB
Image Processing Toolbox. The detected edges on the resulting binary
image were then dilated and holes, defined by the connectivity of
edges and corners, filled. Regions with an area <100 px2 were removed
to reduce segmentation errors. For small regions, defined by an area
less than 10,000 px2, refined segmentation was then carried out in
which each region was extracted from the unprocessed image data
using a padded square extraction box with a side length of 1.5 times the
maximal length of the region on the xy-plane. Image segmentation was
Nature Communications |
(2023) 14:2379
12
Article
https://doi.org/10.1038/s41467-023-37865-3
carried out on each extracted small region individually as described
above, with the difference of using the Canny algorithm, implemented
using the MATLAB Image Processing Toolbox, for edge detection82.
Each detected region of the segmented image then represented an
area of interest for which shape analysis was carried out. For each
region, its area was found from the total number of pixels and the
circularity of the area was calculated as shown in Eq. (2).
circularity = 4π × area × perimeter
(cid:2)2
ð2Þ
For the representation of LCD assemblies with S2hp vRNA (Fig. 3),
calculated circularity measures were then weighted by the area of each
corresponding region in the analysis of the sample means and stan-
dard errors of the means. Statistical analysis of the area and weighted
circularity of segmented regions from a total of 45 images combined
from 5 individual images collected from each technical triplicate of 3
biological repeats, was finally performed and visualized as boxplots
showing the 25th percentile, median, and 75th percentile of the mean
values for triplicate experiments. The whiskers of the plots extend to
the most extreme data points. Observations beyond the whisker length
(shown as circles in the figure) are values more than 1.5 times the
interquartile range beyond the bottom or top of the box. For the
representation of NCAP particles with G12 (Fig. 5), calculated circu-
larity measures were weighted by the area of each corresponding
region in the analysis of the sample means. Mean area and mean
weighted circularity was calculated across regions of 15 images per
experimental condition, obtained by combining 5 images for each
technical triplicate. Every biological repeat was analyzed separately. A
representative boxplot is shown in the figure, in which the central mark
indicates the median of the experimental triplicate means, and the
bottom and top edges of the box indicate the 25th and 75th percen-
tiles, respectively.
Thioflavin-T assays
All solutions in these experiments were prepared using DNase/RNase-
free water (ultrapure water) and were filtered twice using a 0.22-μm
syringe filter. Preparations were done under sterile conditions and
using sterile filter pipette tips to ensure RNA preservation. Thioflavin T
(ThT) stock solution was freshly prepared from powder (Sigma, CAS
ID: 2390-54-7) at a concentration of 20 mM in DNase/RNase ultrapure
water, followed by 0.22-μm filtration.
Thioflavin-T fibrillation kinetic assays. Purified NCAP protein and its
segments were separately diluted into 20 mM Tris pH 8.0, 300 mM
NaCl buffer at 235 μM concentration. S2hp RNA was diluted by 10 mM
HEPES pH 7.0 buffer to 75 μM concentration. The proteins, RNA and
ThT were then mixed to final concentrations of 300 μM ThT, 30 μM
protein, and 0 or 7.5 μM RNA (as indicated in Fig. 1 and Supplementary
Fig. 2), in 1X PBS pH 7.4. Blank samples containing everything but the
protein were prepared. The reaction was carried out in a black 384-well
clear-bottom plate (NUNC 384) covered with optical film (Corning
Sealing Tape Universal Optical) and incubated in a plate reader (BMG
LABTECH FLUOstar Omega) at 37 °C, with 700 rpm double orbital
shaking for 30 s before each measurement. ThT fluorescence was
measured with excitation and emission wavelengths of 430 and
485 nm, respectively. Measurements were made with technical tripli-
cates for each sample. All triplicate values were averaged, and blank
readings from samples without proteins were averaged and subtracted
from the values of corresponding protein mixtures. The results were
plotted against time. The experiment was repeated at least three times
on different days.
Thioflavin-T endpoint assay. Purified LCD protein segment was dilu-
ted into 20 mM Tris pH 8.0, 300 mM NaCl buffer at 235 μM con-
centration. The proteins and ThT were then mixed to final
concentrations of 300 μM ThT and 100 μM protein, in 1X PBS pH 7.4. A
blank sample containing everything but the protein was prepared and
measured as a buffer control. Fibril formation was carried out in
parafilm-covered PCR tubes, incubated in a floor shaker (Torrey Pines
Scientific Inc, Orbital mixing chilling/heating plate) at 37 °C, with fast
mixing speed for 11 days. 30 μL of the samples were taken out of the
tubes at days 1, 6, and 11of incubation and put in a black 384-well clear-
bottom plate (NUNC 384) covered with optical film (Corning Sealing
Tape Universal Optical) and incubated in a plate reader (BMG LAB-
TECH FLUOstar Omega) at 37 °C, with 700 rpm double orbital shaking
for 30 s before the measurement. ThT fluorescence was measured with
excitation and emission wavelengths of 430 and 485 nm, respectively.
Turbidity assay
All solutions were prepared using DNase/RNase-free water (ultrapure
water) and were filtered twice using a 0.22-μm syringe filter. Prepara-
tions were done under sterile conditions and using sterile filter pipette
tips to ensure RNA preservation. Protein and RNA working solutions
were prepared as described above for the ThT experiment of NCAP
and its segments. Each reaction sample contained 30 μM protein and 0
or 7.5 μM RNA in 1X PBS pH 7.4. Blank samples contained everything
but the protein. The reaction was carried out in a black 384-well clear-
bottom plate (NUNC 384) covered with optical film (Corning Sealing
Tape Universal Optical) and incubated in a plate reader (BMG LAB-
TECH FLUOstar Omega) at 37 °C, with mixing before and between
measurements. Turbidity was measured with absorbance (OD) at
600 nm. Measurements were made with technical triplicates for each
sample. Triplicate values were averaged, and appropriate blank read-
ings (samples without the protein) were averaged and subtracted from
the corresponding readings. The results were plotted against time. The
experiment was repeated at least three times on different days.
Negative stain transmission electron microscopy (TEM)
Samples for negative staining TEM were prepared as described below.
All solutions in these experiments were prepared using DNase/RNase-
free water (ultrapure water) and were filtered twice using a 0.22-μm
syringe filter. Preparations were done under sterile conditions and
using sterile filter pipette tips to ensure RNA preservation. For grid
preparation and screening, 4 μL of each sample was applied directly
onto 400-mesh copper TEM grids with Formvar/Carbon support films
(Ted Pella), glow discharged (PELCO easiGlowxs) for 45 s at 15 mA
immediately before use. Grids were incubated with the samples for
2 min, then the samples were blotted off using filter paper. The grids
were washed three times with water and once with 2% uranyl acetate
solution with blotting after each wash. The grids were then incubated
with 6 μL of uranyl acetate solution for 30–45 s before blotting.
Micrographs were imaged using an FEI Tecnai T12 microscope at room
temperature with an accelerating voltage of 120 kV. Images were
recorded digitally with a Gatan US 1000 CCD camera, using the Digital-
Micrograph® Suite software, and processed in the ImageJ83 software.
NCAP fibrils from PS droplets formed in PBS. NCAP samples with and
without 0.75 μM S2hp and 20 μM ZnCl2 (Supplementary Fig. 2b) were
prepared in PBS as described in the PS method section. Samples were
vigorously scrapped from the bottom of the wells after 6 days of
incubation using a 100 μl pipette tip and used for TEM grid prepara-
tion. A blank control containing 0.75 μM S2hp, 20 μM ZnCl2 and
0.0002% w/v ThS in PBS was imaged as well.
NCAP fibrils in 2 mM Tris pH 8.0, 30 mM NaCl (Supplementary
Fig. 2d). Purified NCAP was diluted to 50 μM final concentration from
its stock solution (made in 20 mM Tris pH 8.0, 300 mM NaCl buffer)
into ultrapure water supplemented with ZnCl2 in 0 (water only) or
20 μM final concentration. Samples were incubated for 3 days with
acoustic resonance mixing at 37 °C using a custom-built 96-well plate
Nature Communications |
(2023) 14:2379
13
Article
https://doi.org/10.1038/s41467-023-37865-3
shaker set to 40 Hz. The samples were then recovered and applied to
the EM grid as described above.
Fibrils of NCAP and its segments in PBS. NCAP (Supplementary
Fig. 2c) and its segments (Fig. 1f) were separately diluted to 235 μM
concentration by 20 mM Tris pH 8.0, 300 mM NaCl. The S2hp RNA was
diluted to 250 μM by 10 mM HEPES, pH 7.0 buffer. The proteins and
RNA were further diluted in 1X PBS pH 7.4 such that each reaction
sample contained 100 μM protein and 0/ 25 μM RNA. Fibril formation
was carried out in parafilm-covered PCR tubes, incubated in a floor
shaker (Torrey Pines Scientific Inc, Orbital mixing chilling/heating
plate) at 37 °C, with fast mixing speed for 6 (LCD and DD-Cterm) to 14
(NCAP) days.
LCD fibril formation with short RNA segments (Fig. 2b). RNA stock
solutions were thawed, then annealed by heating at 95 °C for 3 min and
transferring quickly on the ice. The RNAs were diluted to 1 mM con-
centration by their original buffer of 10 mM HEPES pH 7.0. LCD protein
stock was freshly thawed and added together with the appropriate
RNA solution into 1X PBS to reach a 1:2 protein:RNA molar ratio at
50 μM final concentration of LCD, in 50 μL final volume in a black 384-
well clear-bottom plate (NUNC 384). The plate was covered with
optical film (Corning Sealing Tape Universal Optical) and incubated in
a plate reader (BMG LABTECH FLUOstar Omega) at 37 °C with shaking.
Samples were taken for TEM screening after 12 h (day 1) and 4 days of
incubation.
X-ray fiber diffraction
LCD with and without S2hp vRNA. 1.27 mM purified LCD stock solu-
tion was thawed and dialyzed in a dialysis cassette with a 3.5 kDa cutoff
(Thermo Scientific cat. no. 87724) for 4 h at RT in 20 mM Tris pH 7.4,
50 mM NaCl buffer with or without the addition of S2hp vRNA in 4:1
LCD:S2hp molar ratio (955 μM protein and 236 μM RNA). After dialysis,
the samples were added to a black 384-well clear-bottom plate (NUNC
384), covered with optical film (Corning Sealing Tape Universal Opti-
cal), and incubated in a plate reader (BMG LABTECH FLUOstar Omega)
at 37 °C, with 30 s of 700 rpm double orbital shaking every 5 min for
3 weeks. The fibrils were pelleted and washed three times in water by
centrifugation at 13,000×g for 10 min at RT, then pelleted again and
resuspended in 5 μL of deionized water. Fibrils were aligned by
pipetting 2 μL of the fibril resuspension in a 3 mm gap between two
fire-polished glass rods, positioned end-to-end. After 1 h of drying at
room temperature, another 2 μL of the fibril suspension was applied,
thickening the sample. After another hour of drying, the aligned fibril
sample was transferred to the exterior of a standard crystal mounting
loop. To glue the sample to the loop, the loop was wetted with 50 % v/v
ethylene glycol solution, then touched to the surface of the sample and
immediately plunged in liquid nitrogen. The samples were shipped to
the Advanced Photon Source, beamline 24-ID-E at Argonne National
Laboratory for remote data collection. The sample was kept at 100 K
using a nitrogen cryo-stream. Diffraction patterns were collected on a
Dectris Eiger 16M pixel detector using a 2 s exposure at 100% trans-
mission and 1-degree rotation. The X-ray beam wavelength was
0.9792 Å and impinged on the sample only, avoiding the loop and
ethylene glycol, so these later materials do not contribute to the dif-
fraction pattern. The detector was placed 350 mm from the sample.
Diffraction images were displayed with the ADXV program (Scripps).
LCD with non-specific RNA (antisense siDGCR8-1 RNA). LCD stock
solution was concentrated to 2.2 mM and the buffer was exchanged to
20 mM Tris pH 8.0, 150 mM NaCl in a centrifugal filter with 3 kDa cutoff
(Milliepore Sigma Amicon Ultra cat. no. C82301). Antisense siDGCR8-1
RNA stock solution, stored at −20 °C, was thawed and combined with
the LCD solution in 1:3 LCD:RNA molar ratio. The solution was titrated
to reach a final pH of ~5 as confirmed with pH paper. The final protein
concentration was 283 μM and RNA concentration was 849 μM. The
reaction mixture was incubated in a floor shaker (Torrey Pines Scien-
tific Inc, Orbital mixing chilling/heating plate) at 37 °C, with rapid
mixing speed for 7 days. The fibrils were prepared and mounted as
described above except that the fibrils were aligned with a single
application of 5 μL of the fibril suspension, rather than two smaller
applications. Diffraction was measured at beamline24-ID-C, rather than
24-ID-E. Diffraction patterns were collected on a Dectris Eiger2 16M
pixel detector using a 1 s exposure at 90% transmission and 0.5-degree
sample rotation. The X-ray beam wavelength was 0.9791 Å and impin-
ged on the sample only, avoiding the loop and ethylene glycol, so these
later materials do not contribute to the diffraction pattern. Exposures
were collected at sample-to-detector distances of 200 and 500 mm.
Diffraction images were displayed with the ADXV program (Scripps).
Crystallization of NCAP peptide segments
The NCAP segment 217AALALL222 crystallized in batch just before the
purification by RP-HPLC. The peptide had been deprotected and
cleaved from the resin, triturated with cold diethyl ether, and pre-
cipitated. Most of the product had been collected via filtration, but
some residual peptide remained in the round bottom flask and we
intended to use this residual peptide to check the peptide purity by
analytical HPLC. We dissolved the residual peptide with water, acet-
onitrile, and TFA in a volume ratio of approximately 45:45:10 and
transferred it to a 1 mL glass vial for HPLC injection. The solution was
left in the sample holder and needle-like crystals formed after a week.
Some of these crystals were retained for crystal structure determina-
tion. The bulk of the peptide was further purified, as described above.
Later, we showed we could reproduce the crystals by dissolving
0.75 mg of AALALL in 50 μL of TFA and then diluting with 225 μL of
acetonitrile and 225 μL of water. This was left to sit in an HPLC vial
which had its septum top poked open once with the HPLC injection
needle, and the same crystal form appeared in 3 months.
We screened for additional AALALL crystals using 96-well kits and
purified peptide dissolved at 10 mg/mL concentration in 19.6 mM
LiOH. Crystals were grown by the hanging drop vapor diffusion
method. The UCLA Crystallization Facility set up crystallization trays
with a Mosquito robot dispensing 200 nL drops. Needle-shaped crys-
tals of 217AALALL222 grew at 20 °C in a reservoir solution composed of
30% w/v polyethylene glycol (PEG) 3000 and 0.1 M n-cyclohexyl-2-
aminoethanesulfonic acid (CHES), pH 9.5. The purified NCAP segment
179GSQASS184 was dissolved in water at 100 mg/mL concentration.
Hanging drop crystallization trays were set using 200 nL drops.
Needle-shaped crystals grew at 20 °C using a reservoir solution com-
posed of 1.0 M Na, K tartrate, 0.2 M Li2SO4, and Tris pH 7.0. Needle-
shaped crystals appeared immediately after setting up the tray. The
purified NCAP segment 243GQTVTK248 was dissolved in water at 68 mg/
mL concentration. Hanging drop crystallization trays were set using
200 nL drops. Needle-shaped crystals appeared within 1 day at 20 °C
using a reservoir solution composed of 2.0 M (NH4)2SO4, 0.1 M sodium
HEPES, pH 7.5, and 2% v/v PEG 400.
Structure determination of NCAP peptide segments
Microfocus X-ray beam optics were required to measure crystal dif-
fraction intensities from our crystals since they were needle-shaped
and less than 5 microns thick. We used microfocus beamline 24-ID-E of
the Advanced Photon Source located at Argonne National Laboratory.
Crystals were cooled to a temperature of 100 K. Diffraction data were
indexed, integrated, scaled, and merged using the programs XDS and
XSCALE84. Data collection statistics are reported in Table 1. Initial
phases for AALALL and GSQASS were obtained by molecular replace-
ment with the program Phaser85 using a search model consisting of an
ideal β-strand with sequence AAAAAA. Phases for GQTVTK were
obtained by direct methods using the program ShelxD86. Simulated
annealing composite omits maps57 were calculated using Phenix87.
Nature Communications |
(2023) 14:2379
14
Article
https://doi.org/10.1038/s41467-023-37865-3
Refinement was performed using the program Refmac88. Model
building was performed using the graphics program Coot89. Structure
illustrations were created using PyMOL80. Residue hydrophobicity of
the steric zipper segments was assigned and colored according to the
Kyte and Doolittle hydrophobicity scale embedded in UCSF Chimera90.
10 mM HEPES pH 7 (Gibco cat no.
G12 evaluation in HEK293-ACE2 cells infected with SARS-CoV-2
Lyophilized G12 peptide powder was dissolved in 100 % DMSO (Sigma
cat. no. D2650) to approximately 10 mM, centrifuged at 21,000×g for
30 min to remove large aggregates, then aliquoted and stored at
−20 °C until use. To determine peptide concentrations accurately, the
stock was diluted in UltraPure distilled water (ThermoFisher cat. no.
10977015), and the concentration was measured using the Pierce
Quantitative Fluorometric Peptide Assay (ThermoFisher cat. no.
23290). HEK293-ACE2 cells (ATCC, cat. no. CRL-3216, authenticated
and quality tested by ATCC [https://www.atcc.org/products/crl-3216])
stably over-expressing the human ACE2 receptor91 were cultured in
DMEM (Gibco cat no. 11995-065) supplemented with 10% FBS (Gibco
cat no. 26140-079), 1% penicillin-streptomycin (Gibco cat no. 15140-
15630106), 50 μM
122),
2-mercaptoethanol (Sigma cat no. M3148), and 1 μg/mL puromycin
(Gibco cat no. A1113803) for selection, at 37 °C, 5 % CO2. Cells were
confirmed negative for mycoplasma by PCR using a Universal Myco-
plasma Detection Kit (ATCC cat. no. 30-1012K). The HEK293-ACE2 cells
were plated in 96-well black/clear plates (Greiner Bio-One cat. no.
655090) at 2 × 104 cells per well. The cells were incubated for 1–2 days
at 37 °C, 5% CO2, then exchanged into antibiotic-free media and incu-
bated for an additional day. Cells were then transfected with the
peptide-based inhibitors, either unlabeled (Fig. 5d and Supplementary
Fig. 8; Final peptide concentrations are detailed in the figures), or with
~15 μM of FITC-labeled G12 (Supplementary Fig. 7) by diluting stock
solutions (made in 5% DMSO) into cell culture medium to a 10X con-
centration, and serially diluting from there for dose–response assays
while maintaining similar DMSO concentration in all peptide dosages
(Fig. 5d and Supplementary Fig. 8). 10 μL of 10X peptide diluted in
culture medium was added to 90 μL media in each well, for a final
DMSO concentration of 0.5% in all wells. Finally, Endo-Porter (PEG-
formulation) transfection reagent (GeneTools LLC, Philomath, OR) was
added to each well at a final concentration of 6 μM. Plates were incu-
bated for 2- 4 h, then the cells were infected with SARS-CoV-2 (Isolate
USA-WA1/2020) (BEI Resources) in the UCLA BSL3 High-Containment
Facility91 by adding the virus in 200 μl final volume at an MOI of 0.05
for evaluation of dose dependence antiviral activity with the inhibitor
G12 (Fig. 5d and Supplementary Fig. 8). The uninfected control
received only the base media used for diluting the virus. The plates
were incubated for an additional 24 h at 37 °C, 5% CO2, and fixed with
100% methanol for immunofluorescence assay. Fixed cells were
washed 3 times with PBS pH 7.4 (Gibco cat. no. 10010-023) and incu-
bated with blocking buffer (2% BSA, 0.3% Triton X-100, 5% goat serum,
5% donkey serum, 0.01% NaN3 in PBS) for 2 h at room temperature.
Anti-Spike protein primary antibody was diluted into blocking buffer
and incubated overnight at 4 °C. Either of these primary anti-Spike
protein antibodies was used (depending on availability): BEI Resour-
ces, NIAID, NIH rabbit monoclonal Anti-SARS-Related Coronavirus 2
Spike Glycoprotein S1 Domain (produced in vitro), cat. no. NR-53788,
clone no. 007, Lot: HA14AP3001 (purchased from SinoBiological, cat.
no. 40150-R007), at a 1:100 dilution ratio, or BEI Resources, NIAID, NIH:
Mouse Monoclonal Anti-SARS-CoV S Protein (Similar to 240C), cat. no.
NR-616, Lot: 102204 (purchased from ATCC), at a 1:300 dilution ratio.
Following overnight incubation, cells were washed with PBS and
incubated for one hour at room temperature with AlexaFluor-555
conjugated secondary goat anti-mouse (Abcam cat. no. ab150114, Lot:
GR299321-5), or goat anti-rabbit (Abcam cat. no. ab150078, Lot:
GR302355-2) antibody, diluted at 1:1000. All antibodies used in this
section were validated by their respective vendors. Following
incubation with the secondary antibody, the cells were stained with
10 μg/mL DAPI (ThermoFisher cat. no. D1306) for 10 min, and stored in
PBS for imaging. Plates were imaged using an ImageXpress Micro
Confocal High-Content Imaging System (Molecular Devices, San Jose,
CA) in widefield mode at 10X magnification. 9 sites per well were
imaged, and the percentage of infected cells was quantified using the
MetaXpress multiwavelength cell scoring module. We considered
spike protein-expressing cells as infected and calculated their per-
centage from the total number of cells in the well. Raw values were
exported into Microsoft Excel, and percent-infected cells were then
normalized to an infected culture that was treated with vehicle only.
Statistical analysis was performed using one-way ANOVA in GraphPad
Prism, and IC50 values were estimated (Fig. 5d) using a four-parameter
non-linear fit dose–response curve.
Cytotoxicity assay in HEK293-ACE2 cells (Fig. 5d)
HEK293-ACE2 cells were plated and transfected with peptides follow-
ing the same protocol as used for the viral assays, but following
transfection were incubated at 37 °C and 5% CO2 for 24 h. Peptide
cytotoxicity was then assessed using the CyQUANT LDH Cytotoxicity
Assay (ThermoFisher cat no. C20300) following the manufacturer
protocol. Absorbance was measured at 490 and 680 nm (background
subtraction) using a SpectraMax M5 (Molecular Devices) with Soft-
MaxPro v5.3 software.
Statistics and reproducibility
All turbidity and ThT fibrillation kinetic experiments were repeated
three independent times with technical triplicates. Technical tripli-
cates were averaged and blank subtracted. Representative curves are
presented in the figures. Endpoint ThT measurements of the LCD-only
segment were done using three samples. Each sample was measured
once per every time point. X-ray diffractions of LCD only and LCD+
S2hp fibrils were each collected three times on different days, using
different diffractometers and x-ray sources while showing similar
results. Diffraction of LCD+ non-specific RNA fibrils was collected twice
from different regions of the same loop, showing similar results. EM
micrographs of LCD-only fibrils were captured at least five indepen-
dent times. LCD fibrils with the different vRNA segments were visua-
lized by EM at least two independent times per vRNA type, once of
which with different time points. NCAP with and without S2hp vRNA in
PBS was imaged by EM from two independent samples. Other EM
images were taken from a single sample. PS of the LCD-only segment
with ThS and PS of NCAP with and without different concentrations of
G12 were each performed three independent times with technical tri-
plicates showing similar results. PS of NCAP with ThS was repeated
twice (2nd repeat incubated for 3 days only) showing ThS partitioning
into NCAP’s PS droplets. FITC-labeled G12 was tested on NCAP PS
droplets in vitro once. Antiviral activity of G12 in cells was tested three
independent times with G12 concentrations of over 10 μM showing
inhibition of ~40–60% in viral infectivity. Full dose response of G12 and
its cytotoxicity in cells was tested in triplicated wells. Distribution of
FITC labeled G12 in HEK293-ACE2 cells was tested two independent
times with duplicated wells.
Reporting summary
Further information on research design is available in the Nature
Portfolio Reporting Summary linked to this article.
Data availability
Atomic coordinates that support the findings of this study are available
in the RCSB Protein Data Bank (PDB) under accession numbers: 7LV2
[https://doi.org/10.2210/pdb7LV2/pdb],
7LTU [https://doi.org/10.
7LUX [https://doi.org/10.2210/
2210/pdb7LTU/pdb]
and 7LUZ [https://doi.org/10.2210/
pdb7LUX/pdb]
pdb7LUZ/pdb]. The amino acid sequences of the Nucleocapsid
(form 1),
(form 2),
Nature Communications |
(2023) 14:2379
15
Article
https://doi.org/10.1038/s41467-023-37865-3
proteins of SARS-CoV-2 and SARS-CoV analyzed in this study are
available on UniProtKB, accession numbers: P0DTC9, and P59595
respectively. Amino acid sequences of other coronavirus Nucleocapsid
proteins were accessed from the European Nucleotide Archive [ENA;
https://www.ebi.ac.uk/genomes/virus.html]. Raw EM images, light and
fluorescence microscopy images and fiber diffraction source files
generated in this study have been deposited in the Figshare respiratory
[https://figshare.com/projects/Low_Complexity_Domains_of_the_
at
Nucleocapsid_Protein_of_SARS-CoV-2_Form_Amyloid_Fibrils/162391].
Data for all plots presented in this manuscript are provided with this
paper in the Source Data file. Source data are provided with this paper.
References
1.
Savastano, A., Ibáñez de Opakua, A., Rankovic, M. & Zweckstetter,
M. Nucleocapsid protein of SARS-CoV-2 phase separates into RNA-
rich polymerase-containing condensates. Nat. Commun. 11,
1–10 (2020).
2. Chen, H. et al. Liquid–liquid phase separation by SARS-CoV-2
nucleocapsid protein and RNA. Cell Res. 30, 1143–1145 (2020).
3. Cubuk, J. et al. The SARS-CoV-2 nucleocapsid protein is dynamic,
disordered, and phase separates with RNA. Nat. Commun. 12,
1936 (2021).
Lu, S. et al. The SARS-CoV-2 nucleocapsid phosphoprotein forms
mutually exclusive condensates with RNA and the membrane-
associated M protein. Nat. Commun. 12, 502 (2021).
Perdikari, T. M. et al. SARS‐CoV‐2 nucleocapsid protein phase‐
separates with RNA and with human hnRNPs. EMBO J. 39,
e106478 (2020).
4.
5.
7.
6. Cascarina, S. M. & Ross, E. D. A proposed role for the SARS‐CoV‐2
nucleocapsid protein in the formation and regulation of biomole-
cular condensates. FASEB J. 34, 9832–9842 (2020).
Iserman, C. et al. Genomic RNA elements drive phase separation of
the SARS-CoV-2 nucleocapsid. Mol. Cell 80, 1078–1091.e6 (2020).
8. Carlson, C. R. et al. Phosphoregulation of phase separation by the
SARS-CoV-2 N protein suggests a biophysical basis for its dual
functions. Mol. Cell 80, 1092–1103.e4 (2020).
Ye, Q., West, A. M. V., Silletti, S. & Corbett, K. D. Architecture and
self-assembly of the SARS-CoV-2 nucleocapsid protein. Protein Sci.
29, 1890–1901 (2020).
9.
10. Bai, Z., Cao, Y., Liu, W. & Li, J. The SARS-CoV-2 nucleocapsid protein
and its role in viral structure, biological functions, and a potential
target for drug or vaccine mitigation. Viruses 13, 1115 (2021).
Jack, A. et al. SARS-CoV-2 nucleocapsid protein forms condensates
with viral genomic RNA. PLoS Biol. 19, e3001425 (2021).
11.
12. Wu, Y. et al. RNA-induced liquid phase separation of SARS-CoV-2
nucleocapsid protein facilitates NF-κB hyper-activation and
inflammation. Signal Transduct. Target. Ther. 6, 1–13 (2021).
13. Kato, M. et al. Cell-free formation of RNA granules: low complexity
sequence domains form dynamic fibers within hydrogels. Cell 149,
753–767 (2012).
14. Romero, P. et al. Sequence complexity of disordered protein. Pro-
teins Struct. Funct. Genet. 42, 38–48 (2001).
15. Cascarina, S. M., Elder, M. R. & Ross, E. D. Atypical structural ten-
dencies among low-complexity domains in the Protein Data Bank
proteome. PLoS Comput. Biol. 16, e1007487 (2020).
16. Xiang, S. et al. The LC domain of hnRNPA2 adopts similar con-
formations in hydrogel polymers, liquid-like droplets, and nuclei.
Cell 163, 829–839 (2015).
17. Hughes, M. P. et al. Atomic structures of low-complexity protein
segments reveal kinked β sheets that assemble networks. Science
359, 698–701 (2018).
18. Zhou, X. et al. Mutations linked to neurological disease enhance
self-association of low-complexity protein sequences. Science 377,
eabn5582 (2022).
19. Harrison, A. F. & Shorter, J. RNA-binding proteins with prion-like
domains in health and disease. Biochem. J. 474, 1417–1438 (2017).
20. Kato, M., Zhou, X. & McKnight, S. L. How do protein domains of low
sequence complexity work? RNA 28, 3–15 (2022).
21. Roush, S. W. et al. Historical comparisons of morbidity and mortality
for vaccine-preventable diseases in the United States. J. Am. Med.
Assoc. 298, 2155–2163 (2007).
22. Wootton, J. C. & Federhen, S. Analysis of compositionally biased
regions in sequence databases. Methods Enzymol. 266,
554–571 (1996).
23. Rosenberg, G. M. et al. Bioinformatic identification of previously
unrecognized amyloidogenic proteins. J. Biol. Chem. 298,
101920 (2022).
24. Martin, E. W. & Mittag, T. Relationship of sequence and phase
separation in protein low-complexity regions. Biochemistry 57,
2478–2487 (2018).
25. Tai, C. E. et al The SARS-CoV-2 nucleocapsid protein preferentially
binds long and structured RNAs. Preprint at bioRxiv https://doi.org/
10.1101/2021.12.25.474155 (2021).
26. Goldschmidt, L., Teng, P. K., Riek, R. & Eisenberg, D. Identifying the
amylome, proteins capable of forming amyloid-like fibrils. Proc.
Natl Acad. Sci. USA 107, 3487–3492 (2010).
27. Sawaya, M. R. et al. Atomic structures of amyloid cross-β spines
reveal varied steric zippers. Nature 447, 453–457 (2007).
28. Hatos, A., Tosatto, S. C. E., Vendruscolo, M. & Fuxreiter, M. FuzDrop
on AlphaFold: visualizing the sequence-dependent propensity of
liquid-liquid phase separation and aggregation of proteins. Nucleic
Acids Res. 50, W337–W344 (2022).
29. Gruca, A. et al. Common low complexity regions for SARS-CoV-2
and human proteomes as potential multidirectional risk factor in
vaccine development. BMC Bioinforma. 22, 1–18 (2021).
30. Sievers, S. A. et al. Structure-based design of non-natural amino-
acid inhibitors of amyloid fibril formation. Nature 475, 96–100
(2011).
31. Seidler, P. M. et al. Structure-based inhibitors halt prion-like seeding
by Alzheimer’s disease-and tauopathy-derived brain tissue sam-
ples. J. Biol. Chem. 294, 16451–16464 (2019).
32. Saelices, L. et al. Amyloid seeding of transthyretin by ex vivo cardiac
fibrils and its inhibition. Proc. Natl Acad. Sci. 115, E6741–E6750
(2018).
33. Griner, S. L. et al. Structure-based inhibitors of amyloid beta core
suggest a common interface with tau. Elife 8, e46924 (2019).
34. Leaver-Fay, A. et al. Rosetta3: an object-oriented software suite for
the simulation and design of macromolecules. Methods Enzymol.
487, 545–574 (2011).
35. Luo, F. et al. Atomic structures of FUS LC domain segments reveal
bases for reversible amyloid fibril formation. Nat. Struct. Mol. Biol.
25, 341–346 (2018).
36. Murray, D. T. et al. Structure of FUS protein fibrils and its relevance
to self-assembly and phase separation of low-complexity domains.
Cell 171, 615–627.e16 (2017).
37. Lu, J. et al. CryoEM structure of the low-complexity domain of
hnRNPA2 and its conversion to pathogenic amyloid. Nat. Commun.
11, 4090 (2020).
38. Guenther, E. L. et al. Atomic structures of TDP-43 LCD segments
and insights into reversible or pathogenic aggregation. Nat. Struct.
Mol. Biol. 25, 463–471 (2018).
39. Holubová, M., Štěpánek, P. & Hrubý, M. Polymer materials as pro-
moters/inhibitors of amyloid fibril formation. Colloid Polym. Sci.
299, 343–362 (2021).
40. Cao, Y., Adamcik, J., Diener, M., Kumita, J. R. & Mezzenga, R. Dif-
ferent folding states from the same protein sequence determine
reversible vs. irreversible amyloid fate. J. Am. Chem. Soc. 143,
11473–11481 (2021).
Nature Communications |
(2023) 14:2379
16
Article
https://doi.org/10.1038/s41467-023-37865-3
41. Zhao, M. et al. GCG inhibits SARS-CoV-2 replication by disrupting
the liquid phase condensation of its nucleocapsid protein. Nat.
Commun. 12, 1–14 (2021).
42. Julia, M. L. et al. Laura & and the Center for Viral Systems Biology
outbreak.info. outbreak.info. https://outbreak.info/ (2020).
43. Shu, Y. & McCauley, J. GISAID: global initiative on sharing all
influenza data—from vision to reality. Eurosurveillance 22,
30494 (2017).
phase transition and amyloid formation. Biomolecules 11, 1324
(2021).
64. Shanmugam, N. et al. Herpes simplex virus encoded ICP6 protein
forms functional amyloid assemblies with necroptosis-associated
host proteins. Biophys. Chem. 269, 106524 (2021).
65. Pham, C. L. et al. Viral M45 and necroptosis‐associated proteins
form heteromeric amyloid assemblies. EMBO Rep. 20, e46518
(2019).
44. Elbe, S. & Buckland-Merrett, G. Data, disease and diplomacy:
66. Steain, M. et al. Varicella zoster virus encodes a viral decoy RHIM to
GISAID’s innovative contribution to global health. Glob. Chall. 1,
33–46 (2017).
45. Syed, A. M. et al. Rapid assessment of SARS-CoV-2–evolved var-
iants using virus-like particles. Science 374, 1626–1632 (2021).
46. Ong, S. W. X. et al. Clinical and virological features of SARS-CoV-2
variants of concern: a retrospective cohort study comparing B.1.1.7
(Alpha), B.1.315 (Beta), and B.1.617.2 (Delta). SSRN Electron. J. 75,
e1128–e1136 (2021).
47. Sheikh, A., McMenamin, J., Taylor, B. & Robertson, C. SARS-CoV-2
Delta VOC in Scotland: demographics, risk of hospital admission,
and vaccine effectiveness. Lancet 397, 2461–2462 (2021).
48. Dagpunar, J. Interim estimates of increased transmissibility, growth
rate, and reproduction number of the Covid-19 B.1.617.2 variant of
concern in the United Kingdom. Preprint at medRxiv https://doi.
org/10.1101/2021.06.03.21258293 (2021).
inhibit cell death. PLoS Pathog. 16, e1008473 (2020).
67. Zheng, X. et al. The C-terminal amyloidogenic peptide contributes
to self-assembly of Avibirnavirus viral protease. Sci. Rep. 5,
14794 (2015).
68. Luckey, M. et al. A peptide from the adenovirus fiber shaft forms
amyloid-type fibrils. FEBS Lett. 468, 23–27 (2000).
69. Papanikolopoulou, K. et al. Amyloid fibril formation from sequences
of a natural β-structured fibrous protein, the adenovirus fiber. J.
Biol. Chem. 280, 2481–2490 (2005).
70. Ghosh, A. et al. Self-assembly of a nine-residue amyloid-forming
peptide fragment of SARS corona virus E-protein: mechanism of
self aggregation and amyloid-inhibition of hIAPP. Biochemistry 54,
2249–2261 (2015).
71. Nyström, S. & Hammarström, P. Amyloidogenesis of SARS-CoV-2
spike protein. J. Am. Chem. Soc. 144, 8945–8950 (2022).
49. Fisman, D. N. & Tuite, A. R. Evaluation of the relative virulence of
novel SARS-CoV-2 variants: a retrospective cohort study in Ontario,
Canada. CMAJ 193, E1619–E1625 (2021).
50. Furukawa, Y. & Nukina, N. Functional diversity of protein fibrillar
72. Charnley, M. et al. Neurotoxic amyloidogenic peptides in the pro-
teome of SARS-COV2: potential implications for neurological
symptoms in COVID-19. Nat. Commun. 13, 1–11 (2022). 2022 131.
73. Tetz, G. & Tetz, V. Prion-like domains in eukaryotic viruses. Sci. Rep.
aggregates from physiology to RNA granules to neurodegenerative
diseases. Biochim. Biophys. Acta—Mol. Basis Dis. 1832,
1271–1278 (2013).
51. Ramaswami, M., Taylor, J. P. & Parker, R. Altered ribostasis:
RNA–protein granules in degenerative disorders. Cell 154,
727–736 (2013).
52. Molliex, A. et al. Phase separation by low complexity domains
promotes stress granule assembly and drives pathological fibrilli-
zation. Cell 163, 123–133 (2015).
53. Goedert, M., Eisenberg, D. S. & Crowther, R. A. Propagation of Tau
aggregates and neurodegeneration. Annu. Rev. Neurosci. 40,
189–210 (2017).
54. Eisenberg, D. & Jucker, M. The amyloid state of proteins in human
diseases. Cell 148, 1188–1203 (2012).
8, 1–10 (2018).
74. Consortium, T. U. et al. UniProt: the universal protein knowledge-
base in 2021. Nucleic Acids Res. 49, D480–D489 (2021).
75. Seidler, P. M. Structural and Biochemical Studies of the hsp90
Chaperone Proteins (State University of New York at Buffalo, 2014).
76. Heckman, K. L. & Pease, L. R. Gene splicing and mutagenesis by
PCR-driven overlap extension. Nat. Protoc. 2, 924–932 (2007).
77. Saff, E. B. & Kuijlaars, A. B. J. J. Distributing many points on a sphere.
Math. Intell. 19, 5–11 (1997).
78. Lawrence, M. C. & Colman, P. M. Shape complementarity at pro-
tein/protein interfaces. J. Mol. Biol. 234, 946–950 (1993).
79. Winn, M. D. et al. Overview of the CCP 4 suite and current develop-
ments. Acta Crystallogr. Sect. D Biol. Crystallogr. 67, 235–242 (2011).
80. Schrödinger, L. The PyMOL Molecular Graphics System, Version
55. Haikal, C. et al. The bacterial amyloids phenol soluble modulins
1.2r3pre.
from staphylococcus aureus catalyze alpha-synuclein aggregation.
Int. J. Mol. Sci. 22, 11594 (2021).
81. Schindelin, J. et al. Fiji: an open-source platform for biological-
image analysis. Nat. Methods 9, 676–682 (2012).
56. Javed, I. et al. Accelerated amyloid beta pathogenesis by bacterial
82. Canny, J. A computational approach to edge detection. IEEE Trans.
amyloid FapC. Adv. Sci. 7, 2001299 (2020).
Pattern Anal. Mach. Intell. PAMI-8, 679–698 (1986).
57. Sampson, T. R. et al. A gut bacterial amyloid promotes α-synuclein
aggregation and motor impairment in mice. Elife 9, e53111 (2020).
58. Friedland, R. P. & Chapman, M. R. The role of microbial amyloid in
83. Schneider, C. A., Rasband, W. S. & Eliceiri, K. W. NIH Image to
ImageJ: 25 years of image analysis. Nat. Methods 9, 671–675 (2012).
84. Kabsch, W. XDS. Acta Crystallogr. Sect. D Biol. Crystallogr. 66,
neurodegeneration. PLoS Pathog. 13, e1006654 (2017).
125–132 (2010).
59. Ren, B. et al. Fundamentals of cross-seeding of amyloid proteins: an
85. McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crys-
introduction. J. Mater. Chem. B 7, 7267–7282 (2019).
tallogr. 40, 658–674 (2007).
60. Semerdzhiev, S. A., Fakhree, M. A. A., Segers-Nolten, I. & Blum, C.,
Claessens, M. M. A. E. Interactions between SARS-CoV-2 N-protein
and α-synuclein accelerate amyloid formation. ACS Chem. Neu-
rosci. 13, 143–150 (2022).
61. Li, Y. et al. SARS-CoV-2 impairs the disassembly of stress granules
and promotes ALS-associated amyloid aggregation. Protein Cell 13,
602–614 (2022).
62. Vidic, J. et al. Amyloid assemblies of influenza a virus PB1-F2 protein
damage membrane and induce cytotoxicity. J. Biol. Chem. 291,
739–751 (2016).
63. Salladini, E. et al. Identification of a region in the common amino-
terminal domain of hendra virus P, V, and W proteins responsible for
86. Sheldrick, G. M. Experimental phasing with SHELXC/D/E: combin-
ing chain tracing with density modification. Acta Crystallogr. Sect. D
Biol. Crystallogr. 66, 479–485 (2010).
87. Afonine, P. V. et al. Towards automated crystallographic structure
refinement with phenix.refine. Acta Crystallogr. Sect. D Biol. Crys-
tallogr. 68, 352–367 (2012).
88. Murshudov, G. N. et al. REFMAC5 for the refinement of macro-
molecular crystal structures. Acta Crystallogr. Sect. D Biol. Crys-
tallogr. 67, 355–367 (2011).
89. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and
development of Coot. Acta Crystallogr. Sect. D Biol. Crystallogr. 66,
486–501 (2010).
Nature Communications |
(2023) 14:2379
17
Article
https://doi.org/10.1038/s41467-023-37865-3
90. Pettersen, E. F. et al. UCSF Chimera—a visualization system for
exploratory research and analysis. J. Comput. Chem. 25,
1605–1612 (2004).
91. Garcia, G. et al. Antiviral drug screen identifies DNA-damage
response inhibitor as potent blocker of SARS-CoV-2 replication.
Cell Rep. 35, 108940 (2021).
92. Terwilliger, T. C. et al. Iterative-build OMIT maps: map improvement
by iterative model building and refinement without model bias.
Acta Crystallogr. Sect. D Biol. Crystallogr. 64, 515–524 (2008).
preparation and experimentation: C.E.T., Y.L., Luk.S. Peptide prepara-
tion: C.-T.Z. X-ray crystallography: M.R.S., C.-T.Z, J.L., K.H., G.F., D.C.
Fluorescence and electron Microscopy: E.T.-F., J.T.B., X.C., D.R.B., R.A.,
Y.X.J., H.P., G.M.R., J.L. Computational analysis and peptide self-
assembly modulators design: G.M.R., P.M.S., Y.X.J., E.T-F., Lor.S., K.A.M.
Brightfield image segmentation and shape analysis in MATLAB: L.L. In-
cell assays: J.T.B., G.G. Jr. Writing and figure preparation: E.T.-F., M.R.S.,
J.T.B., F.G., D.S.E. Technical support: D.H.A., Project management: E.T.-
F., M.R.S., R.D., V.A., F.G., and D.S.E.
Acknowledgements
We thank Megan Bentzel, Jose Rodriguez, Meytal Landau, and Mark
Arbing for the discussions. We thank the staff at the Northeastern Col-
laborative Access Team, which is funded by the National Institute of
General Medical Sciences from the National Institutes of Health (P30
GM124165). The Eiger 16M detector on the 24-ID-E beamline is funded by
an NIH-ORIP HEI grant (S10OD021527). The Advanced Photon Source, a
U.S. Department of Energy (DOE) Office of Science User Facility oper-
ated for the DOE Office of Science by Argonne National Laboratory
under Contract No. DE-AC02-06CH11357. Some of this work was also
performed at the Stanford-SLAC Cryo-EM Center (S2C2), which is sup-
ported by the National Institutes of Health Common Fund Transforma-
tive High-Resolution Cryo-Electron Microscopy program (U24
GM129541). The content is solely the responsibility of the authors and
does not necessarily represent the official views of the National Insti-
tutes of Health. The authors also acknowledge the use of instruments at
the Electron Imaging Center for NanoMachines supported by NIH
(1S10RR23057 to ZHZ) and CNSI at UCLA. Mass spectrometry data were
collected on instrumentation maintained and made available through
the support of the UCLA Molecular Instrumentation Center—Mass
Spectrometry Facility in the Department of Chemistry. This material is
based upon work supported by the National Science Foundation under
Grant No. (MCB 1616265), NIH/NIA R01 Grant AG048120, the U.S.
Department of Energy (DOE) Contract No. DOE-DE-FC02-02ER63421,
and by UCLA David Geffen School of Medicine—Eli and Edythe Broad
Center of Regenerative Medicine and Stem Cell Research Award Pro-
gram, Broad Stem Cell Research Center (BSCRC) COVID 19 Research
Award (OCRC #20-73). This study is also supported by the UCLA W.M.
Keck Foundation COVID-19 Research Award and National Institute of
Health awards 1R01EY032149-01, 5U19AI125357-08, 5R01AI163216-02
and 1R01DK132735-01 to V.A. The Human Frontiers Science Project
Organization (HFSPO) (LT000623/2018-L) supported E.T-F. NIH NIGMS
GM123126 grant supported Luk.S. C.-T.Z. was funded by the UCLA Dis-
sertation Year Fellowship.
Author contributions
Constructs design and cloning: P.M.S., Luk.S. Protein preparation and
experimentation: E.T-F, J.T.B., S.L.G., X.C., R.A., J.L., Y.X.J. RNA
Competing interests
D.S.E. is an advisor and equity shareholder in ADRx, Inc. The remaining
authors declare no competing interests.
Additional information
Supplementary information The online version contains
supplementary material available at
https://doi.org/10.1038/s41467-023-37865-3.
Correspondence and requests for materials should be addressed to
David S. Eisenberg.
Peer review information Nature Communications thanks Nicholas Rey-
nolds, Dan Li and the other anonymous reviewer(s) for their contribution
to the peer review of this work. Peer review reports are available.
Reprints and permissions information is available at
http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to jur-
isdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as
long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate if
changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons license and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.
© The Author(s) 2023
Nature Communications |
(2023) 14:2379
18
| null |
10.1111_jbfa.12686.pdf
|
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from third parties. Restrictions apply to the availability
of these data, which were used under license for this study. Data are available from the authors with the permission of
third partie
|
DATA AVAILABILITY STATEMENT The data that support the findings of this study are available from third parties. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the authors with the permission of third parties.
|
Asymmetric trading responses to credit rating announcements from issuer-
Asymmetric trading responses to credit rating announcements from issuer-
versus investor-paid rating agencies
versus investor-paid rating agencies
Quan Pham Minh Nguyen, HX Do, A Molchanov, L Nguyen, NH Nguyen
Publication date
Publication date
01-01-2023
Licence
Licence
This work is made available under the CC BY 4.0 licence and should only be used in accordance with that
licence. For more information on the specific terms, consult the repository record for this item.
Document Version
Document Version
Published version
Citation for this work (American Psychological Association 7th edition)
Citation for this work (American Psychological Association 7th edition)
Nguyen, Q. P. M., Do, H., Molchanov, A., Nguyen, L., & Nguyen, N. (2023). Asymmetric trading responses to
credit rating announcements from issuer- versus investor-paid rating agencies (Version 1). University of
Sussex. https://hdl.handle.net/10779/uos.23495603.v1
Published in
Published in
Journal of Business Finance and Accounting
Link to external publisher version
Link to external publisher version
https://doi.org/10.1111/jbfa.12686
Copyright and reuse:
Copyright and reuse:
This work was downloaded from Sussex Research Open (SRO). This document is made available in line with publisher policy
and may differ from the published version. Please cite the published version where possible. Copyright and all moral rights to the
version of the paper presented here belong to the individual author(s) and/or other copyright owners unless otherwise stated. For
more information on this work, SRO or to report an issue, you can contact the repository administrators at [email protected].
Discover more of the University’s research at https://sussex.figshare.com/
Received: 16 June 2020
Revised: 6 December 2022
Accepted: 9 January 2023
DOI: 10.1111/jbfa.12686
A R T I C L E
Asymmetric trading responses to credit rating
announcements from issuer- versus investor-paid
rating agencies
Quan M. P. Nguyen1
Lily Nguyen4
Nhut H. Nguyen5
Hung Xuan Do2,3
Alexander Molchanov2
1Department of Accounting and Finance,
University of Sussex, Brighton, UK
2School of Economics and Finance, Massey
University, Auckland, New Zealand
3International School, Vietnam National
University, Hanoi, Vietnam
4UQ Business School, University of
Queensland, Brisbane, Queensland, Australia
5Department of Finance, Auckland University
of Technology, Auckland, New Zealand
Correspondence
Alexander Molchanov, School of Economics
and Finance, Massey University, Auckland,
New Zealand.
Email: [email protected]
Abstract
The credit rating industry has traditionally followed the
“issuer-pays” principle. Issuer-paid credit rating agencies
(CRAs) have faced criticism regarding their untimely release
of negative rating adjustments, which is attributed to a
conflict of interests in their business model. An alternative
model based on the “investor-pays” principle is arguably
less subject to the conflict of interest problem. We examine
how investors respond to changes in credit ratings issued
by these two types of CRAs. We find that investors react
asymmetrically: They abnormally sell equity stakes around
rating downgrades by investor-paid CRAs, while abnormally
buying around rating upgrades by issuer-paid CRAs. Our
study suggests that, through their trades, investors capital-
ize on value-relevant information provided by both types of
CRAs, and a dynamic trading strategy taking advantage of
this information generates significant abnormal returns.
K E Y W O R D S
credit ratings, institutional investors, trading strategy
J E L C L A S S I F I C AT I O N
G11, G24
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and
reproduction in any medium, provided the original work is properly cited.
© 2023 The Authors. Journal of Business Finance & Accounting published by John Wiley & Sons Ltd.
J Bus Fin Acc. 2023;1–29.
wileyonlinelibrary.com/journal/jbfa
1
2
1
INTRODUCTION
NGUYEN ET AL.
The credit rating sector has long been dominated by three major issuer-paid credit rating agencies (CRAs): Standard
and Poor’s (S&P), Moody’s Investors Service (Moody’s) and Fitch Ratings. These issuer-paid CRAs extract fees
directly from bond issuers, which might lead to potential conflicts of interest when they provide rating services to
those issuers. Issuer-paid CRAs tend to delay the release of negative ratings (Cornaggia & Cornaggia, 2013; J. He
et al., 2012; Skreta & Veldkamp, 2009) while giving favorable ratings to stocks in their owners’ portfolios (Kedia
et al., 2017). Baghai and Becker (2018) find evidence that issuer-paid CRAs assign higher ratings even to those
issuers who pay them for non-rating services. The lack of timeliness in negative rating adjustments in high-profile
bankruptcies, such as Enron (2001), WorldCom (2002) and Lehman Brothers (2008), is often presented as evidence
of such conflicts. For example, on September 10, 2008—the day Lehman Brothers announced its bankruptcy—S&P
and Moody’s had them rated at A2 and A, respectively, and only adjusted the credit ratings down after the bankruptcy
announcement.
The entry of investor-paid CRAs (e.g., Egan-Jones Ratings [EJR] and Rapid Ratings) has changed the dynamics of the
credit rating industry. These CRAs are paid by the end users of their ratings, such as institutional investors, and the con-
flict of interest problem is potentially alleviated. Extant literature documents significant evidence of high rating quality
of investor-paid CRAs. Cornaggia and Cornaggia (2013) show that Rapid Ratings provides more timely downgrades for
defaulting bonds than Moody’s downgrades, which results in significant loss avoidance for investors. Xia (2014) con-
siders the entry of EJR as a natural experiment to assess issuer-paid CRAs’ reactions to potential competition from a
new player. They find that due to reputational concerns, credit ratings issued by S&P tend to become more responsive
and informative following the EJR entry. Beaver et al. (2006) and Bruno et al. (2016) report that EJR’s credit ratings
are more accurate and timely than Moody’s, even after its successful registration as a nationally recognized statisti-
cal rating organization in December 2007. X. Hu et al. (2019) find corroborating evidence in a non-US setting. Using
the introduction of China Bond Rating (CBR) in 2010, a CRA that combines a public utility model and an investor-paid
model, the authors show that the CBR entry triggers a significant reduction in rating inflation and improvements in
information quality of credit rating announcements by nine traditional issuer-paid CRAs in China.
Given the rise of investor-paid CRAs, the competition they bring about and the information content of their credit
ratings relative to issuer-paid CRAs, it is crucial to understand whether and how financial market participants uti-
lize credit ratings provided by both issuer- and investor-paid CRAs for their benefit. Xia (2014) and Berwart et al.
(2019) find that stocks with downgrade announcements by EJR experience significantly more negative returns than
following downgrades by issuer-paid CRAs, whereas EJR upgrades apparently do not trigger a positive response from
investors. Investigating the reaction of institutional investors to EJR’s rating changes, Bhattacharya et al. (2019) find
that these investors are more responsive to its rating announcements than to other trading signals. They also show
that institutional investors who follow EJR’s credit rating announcements outperform those who ignore these signals.
We contribute to this strand of literature and examine the value relevance of credit rating changes issued by both
types of CRAs.
We argue that investor-paid CRAs cannot completely dominate traditional issuer-paid CRAs that have long-term
positions in the credit rating sector. As argued in previous studies, issuer-paid CRAs only tend to delay negative credit
rating announcements due to the potential conflict of interest (Cornaggia & Cornaggia, 2013; He et al., 2012; Skreta &
Veldkamp, 2009). In contrast, issuer-paid CRAs are likely less conservative in issuing rating upgrades since it would
be in their interest to cater positive ratings to their clients (e.g., Bolton et al., 2012; Griffin et al., 2013). Hence, it
remains unclear whether investors show different trading patterns in responding to negative and positive credit rat-
ing adjustments from issuer- and investor-paid CRAs. The answer to this question is important as it provides a better
understanding of the relevance and viability of different types of CRAs.
We use institutional investors and mutual funds’ changes in stock holdings around rating announcements as a proxy
for market reaction. We consider EJR as a representative of investor-paid CRAs, while the “Big Three” CRAs (S&P,
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
3
Moody’s and Fitch) are representatives of issuer-paid CRAs. We find that institutional investors abnormally decrease
their equity holdings surrounding investor-paid rating downgrades but do not respond to any issuer-paid rating down-
grades. On the contrary, they significantly increase their equity holdings around issuer-paid rating upgrades but
remain unresponsive to investor-paid rating upgrades. These results suggest that institutional investors and mutual
funds consider investor-paid CRAs’ rating downgrades as being timely and informative for their trading as opposed to
issuer-paid CRAs’ rating downgrades. Further, they regard issuer-paid rating upgrades as having more value-relevant
information than investor-paid rating upgrades. In the main analysis, we use quarterly mutual fund (S12) holdings and
quarterly institutional (13F) holdings provided by Thomson Reuters. We also use daily institutional trades provided by
Abel Noser Corporation to measure institutional reactions to credit rating adjustments.1
We then examine whether investors can profit from trading decisions in response to rating changes. We con-
struct and compare four trading strategies: (1) a “dynamic” strategy—selling following investor-paid negative signals
and buying following issuer-paid positive signals, (2) a “naïve” strategy—selling following negative signals and buy-
ing following positive signals from any rating agency, (3) an “EJR-based” strategy—selling following negative signals
and buying following positive signals announced by EJR and (4) an “issuer-paid CRA-based” strategy—selling follow-
ing negative signals and buying following positive signals by any of the issuer-paid CRAs. Following Jagolinzer et al.
(2011), we compute returns for each trading strategy adjusting for common risk factors using the Fama–French five-
factor model. The trading strategy analysis is performed in two steps. First, we construct “notional” trading strategies
to acknowledge the fact that any market player with access to credit ratings can potentially benefit from these strate-
gies. These results also correspond to equally weighted returns of an investor who trades on every signal consistent
with a given strategy. While all four strategies outperform the buy-and-hold strategy, we find that the dynamic strat-
egy produces the highest returns, offering an average difference in annualized risk-adjusted returns of up to 5.02%
over the other three strategies for a 1-month holding period. Second, using aggregate credit rating changes and insti-
tutional investors’ quarterly stock holding changes from S12 and 13F data, we find that all four trading strategies earn
substantially higher abnormal returns than the corresponding notional strategy returns and that the dynamic strategy
consistently exhibits the highest abnormal returns. Finally, since Abel Noser Corporation provides daily trading data
for institutional investors, we use them as an alternative dataset to identify trading strategies based on cumulative net
buy around announcement dates. We thus explicitly acknowledge that an institution can dynamically switch between
strategies and potentially follow multiple strategies at a time. Our results confirm the superiority of the dynamic trad-
ing strategy. More importantly, it outperforms the other three strategies by more than 10% per annum for a 1-month
holding period and up to 7.26% per annum for a 9-month holding period. This outperformance is more than twice the
notional strategies’ corresponding outperformance; hence, they are consistent with the argument that institutional
investors have advanced trading skills and knowledge (Puckett & Yan, 2011) to exploit the informative announcements
in the financial markets.
Our study contributes to the literature in several important ways. First, we add to the knowledge of the relationship
between the quality of credit ratings and market participants’ behavior. The related literature finds that the high qual-
ity of investor-paid CRA ratings creates a reputational concern for issuer-paid CRAs, which motivates them to improve
the overall quality of credit ratings (e.g., Berwart et al., 2019; Bruno et al., 2016; Ramsay, 2011; Xia, 2014). For example,
Xia (2014) finds that following EJR’s appearance, S&P ratings started to reflect credit risks more accurately. Similarly,
Ramsay (2011) discovers that the entry of Rapid Ratings—another investor-paid CRA—motivates major issuer-paid
CRAs to improve the quality of credit ratings. X. Hu et al. (2019) provide evidence of significant improvements in credit
rating informativeness in the China bond market after the introduction of CBR, a combined public utility and investor-
paid CRA. However, the impact of credit rating quality on investors’ behavior remains underexamined. Our study fills
this gap by examining the role of timeliness of credit rating adjustments—a proxy for credit rating quality—in driving
institutional investors’ behavior.
1 We thank the editor and referee for this constructive suggestion.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 4
NGUYEN ET AL.
Second, our findings enrich the understanding of how institutional investors, as professional players, analyze and
react dynamically to negative and positive rating adjustments obtained from different sources over time. Baker and
Mansi (2002) report interesting results regarding the view of institutional investors toward credit ratings. They
find that a majority of institutional investors value credit ratings in their investment decisions and place significant
importance on rating timeliness. Although they also generally agree on the accuracy of ratings in reflecting firms’
creditworthiness, they believe that ratings could either overstate or understate credit risk. Therefore, institutional
investors tend to also rely on their own internal analysis before responding to credit rating news. Cantor et al. (2007)
find from their survey that investment managers in the United States and Europe share remarkably similar usage of
credit ratings to conduct their investment activities. He (2021) finds that transient institutional investors tend to trade
more intensively in low credit rating firms following their earnings announcements. Bhattacharya et al. (2019) find
that institutional investors who follow EJR’s rating announcements significantly focus on EJR rating news rather than
important equity trading signals, such as analyst recommendations, earnings announcements and earnings forecast
revisions. They also find that institutional investors who persistently follow EJR’s credit rating announcements out-
perform those who do not embrace these signals. Our study extends their findings by providing new evidence that
investors with access to rating announcements could dynamically exploit the value-relevant information of negative
and positive rating signals provided by both investor-paid and issuer-paid CRAs in making their trading decisions. Our
results show that while such trading behavior is generally profitable, institutional investors evidently earn the high-
est abnormal profits. Finally, the reported abnormal profits that continue to exist up to at least 6 months suggest
that investors underreact to the information content of credit rating announcements, particularly to negative signals
provided by investor-paid EJR and positive signals given by issuer-paid CRAs.
The remainder of the paper is organized as follows. Section 2 summarizes data collection, variable measurements
and summary statistics. Section 3 presents the methodology and empirical results. Robustness checks are presented
in Section 4. Section 5 concludes.
2
SAMPLE SELECTION, VARIABLE MEASUREMENTS AND SUMMARY STATISTICS
2.1
Sample selection
We consider two quarterly institutional holding databases to extract institutional investors’ trading activities: mutual
fund (S12) holdings and institutional (13F) holdings provided by Thomson Reuters. The S12 holdings database pro-
vides data on mutual fund holdings of US securities at the end of each quarter. The 13F holdings database provides a
similar data structure at the institutional (i.e., investment company or fund family) level.2 Our analysis includes all US
equity mutual funds and institutional investors that have at least 65% of their assets in common stocks (e.g., Amihud
& Goyenko, 2013; Cremers & Petajisto, 2009).3 The final samples include 8566 mutual funds and 8656 institutional
investors.
As mentioned above, we focus on two types of CRAs: investor- and issuer-paid CRAs. EJR is a representative of
investor-paid CRAs, while the “Big Three” represent issuer-paid CRAs. Credit rating data are sourced from Egan-Jones
Rating Company4 and Bloomberg for the period from 1999 to 2017 to match with the S12 and 13F holding data. The
credit rating databases include two types of rating information: rating warning announcements5 and official rating
2 Note that Form 13F is only required for institutional investment managers with more than $100 million in assets under management.
3 We also consider alternative thresholds such as 50%, 60% and 70% as robustness checks. The results are consistent and available upon request.
4 We wish to thank the Egan-Jones Rating Company for sharing its historical rating data.
5 Based on the data availability, there are two types of rating warning announcements: outlook and developing signals. These signals are normally announced
before official rating adjustments.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
5
adjustments.6 The databases also report the date of each credit rating adjustment. As we are interested in corporate
credit ratings, sovereign credit and asset-backed securities ratings are excluded.
2.2
Variable definitions
Since credit ratings are represented by different combinations of letters and numbers (e.g., AAA/Aaa, AA+/Aa1,
AA/Aa2, AA-/Aa3), several prior studies follow Gande and Parsley (2005) to construct a unique “comprehensive credit
rating” (CCR) scale to quantify alphabetic ratings (Alsakka & ap Gwilym, 2012; Chen et al., 2016; Dimitrov et al., 2015;
Drago & Gallo, 2016). Based on the features of credit rating data availability, we follow Joe and Oh’s (2018) rating con-
version scale. The numeric score for letter rating and warning (single) signals are shown in Appendix A.7 In addition,
we also follow the literature (Chen et al., 2016; Vu et al., 2015) to measure the significance of the credit rating event
for firm n at time t as the change in CCR, ΔCCRn,t:
ΔCCRn,t = CCRn,t − CCRn,t−1.
(1)
To match the frequency of fund holding data, we aggregate changes in credit rating adjustment on a quarterly basis.
For instance, in the first quarter of 2010, S&P announces two credit rating adjustments for firm n, a single downgrade
(i.e., −1 notch) on February 1, 2010, and a double downgrade (i.e., −2 notches) on March 2, 2010, and the aggregate
credit rating change by S&P for firm n in the first quarter of 2010 is −3 notches.
We use abnormal mutual fund and institutional investors’ trading as a proxy for investors’ responses, measured by
quarterly abnormal net buy, NBi,n,q.
NBi,n,q = nbi,n,q − average nbi,n,q,
(2)
where nbi,n,q is the quarterly net buy by mutual fund or institutional investor i on stock n measured as dollar stock
holding in quarter q minus quarter q − 1, normalized by the stock’s total market value at the end of the quarter q.8
Average nbi,n,q denotes the average value of nbi,n,q in the period from quarter q − 4 to q − 1 as follows:
average nbi,n,q =
∑−4
k = −1 nbi,n,q+k
4
.
(3)
We follow Chemmanur et al. (2016) to convert NBi,n,q into basis points.
2.3
Control variables
We also follow the related literature (Bernile et al., 2015; Bhattacharya et al., 2019; Henry et al., 2017) to control
for a vector of firm characteristics related to institutional trading activities. The control variables include firm size,
profitability, stock idiosyncratic volatility, Z-score, analyst coverage, interest coverage, firm age, leverage, high-tech
6 Official rating adjustments are basically divided into two types: positive and negative signals. These signals can also include single and multiple events. In our
study, a single event is either a one-notch upgrade or downgrade, and a multiple event is either a multiple-notch upgrade (downgrade) or a combined event of
a rating warning announcement and an official rating adjustment.
7 Gande and Parsley (2005) count positive and negative outlooks as one notch. In our study, to highlight the impacts of official upgrades (downgrades), positive
and negative outlooks are counted as 0.5 notch and positive and negative developments as 0.25.
8 This is to follow the merit of Chemmanur et al. (2009) who estimate institutional net buy based on shares traded and shares outstanding.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 6
NGUYEN ET AL.
dummy and an S&P 500 index inclusion dummy. The descriptions of control variables and their sources are presented
in Appendix B.
2.4
Summary statistics
Table 1 presents summary statistics of mutual funds (S12) and institutional investors (13F). The number of mutual
funds (institutional investors) has gradually increased from 3364 (1751) in 1999 to 4752 (4130) in 2017. The number
of stocks held by mutual funds (institutional investors) has been relatively stable, ranging from 576 (562) in 1999 to
653 (692) in 2017. On average, each institutional investor holds 109 stocks in their portfolio in 1999. The number
gradually increases to 162 in 2017. These figures are almost double those of mutual funds, which are at 52 stocks in
1999 and 99 stocks in 2017. Mutual funds’ (institutional investors’) stock holdings have sharply increased from $301
(267) billion in 1999 to $2573 (1168) billion in 2017. On average, each mutual fund holds $89 million worth of stocks
in 1999, and the amount increases to $541 million in 2017. The figures for institutional investors are $153 million in
1999 and $283 million in 2017.
Table 2 displays summary statistics of credit rating events. The first row of panel A shows the number of unique
firms that each CRA provides credit rating announcements over the sample period of 1999–2017. Despite being
a relatively new player in the credit rating industry, EJR provides credit ratings for 1502 firms, which are only
slightly fewer than S&P (1432 firms) but more than double the coverage by either Moody’s (645) or Fitch (502).
EJR is also the only CRA that provides developing signals, whereas the traditional issuer-paid CRAs do not provide
such service during our sample period.9 We split our rating announcements into negative and positive events and
present them in panel A, sections 1 and 2. There are 2628 (2504), 1172 (546), 355 (172) and 200 (64) negative
(positive) combined events10 assigned by EJR, S&P, Moody’s and Fitch, respectively. In addition, the sample comprises
2013 (1896), 1187 (1342), 370 (541) and 578 (549) solo downgrades (upgrades) and 415 (264), 428 (114), 187
(48) and 163 (70) multiple downgrades (upgrades) announced by EJR, S&P, Moody’s and Fitch, respectively. Panel
A also shows 1648 (1910), 1537 (730), 440 (299) and 278 (80) negative (positive) outlook signals by these CRAs,
respectively.
Panel B of Table 2 presents the distribution of credit rating adjustments. Regarding the total number of rating
events, EJR issues about 20% more rating changes than all issuer-paid CRAs’ events combined. Within each CRA,
EJR has more positive than negative rating announcements. This is opposite to the issuer-paid CRAs, which announce
more negative rating adjustments than positive ones. Regarding the magnitude of rating adjustments, Fitch, on aver-
age, seems to provide the boldest adjustments, compared to the other CRAs. For example, the mean absolute value of
negative rating adjustments is 1.174 for Fitch, while that is 1.041, 1.050 and 1.068 for EJR, S&P and Moody’s, respec-
tively. Negative rating adjustments are generally larger in absolute value than positive rating adjustments. The median
column in panel B suggests that S&P is relatively more conservative in their negative rating adjustments: 50% of their
negative rating events have a median value of 0.5 notch.
Table 3 presents the descriptive statistics of control variables computed around credit rating changes. Observa-
tions are divided into three groups: The first group includes firms rated by EJR and S&P, the second group is for firms
rated by EJR and Moody’s and the third group covers firms rated by EJR and Fitch. The N column shows the number
of fund-firm-quarter observations. The first group has the largest number of observations in both S12 and 13F sam-
ples, followed by groups three and two. The third group includes, on average, larger and older firms. This appears to be
9 EJR derives its “watch” assignments from the difference between the current and projected ratings. No difference between the two results in a “stable”
watch, a higher projected rating results in a “positive” or “POS” watch and a lower projected rating results in a “negative” or “NEG” watch. The absence of a
projected rating results in a “developing” or “DEV” watch or no watch being populated. The addition of a POS or NEG is at the discretion of the analyst or Rating
Committee and usually results from the direction the rate is expected to move over time. See https://www.egan-jones.com/public/download/methodologies/
20210510/EJR_Main_Methodologies_V15a.pdf
10 A combined event is a multiple announcement when a CRA adjusts both credit rating score and outlook (or developing) signal.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
7
)
s
I
I
(
F
3
1
s
r
o
t
s
e
v
n
i
l
a
n
o
i
t
u
t
i
t
s
n
I
)
s
F
M
(
2
1
S
s
d
n
u
f
l
a
u
t
u
M
e
u
l
a
v
k
c
o
t
s
l
a
t
o
T
e
u
l
a
v
k
c
o
t
s
l
a
t
o
T
e
u
l
a
v
k
c
o
t
s
l
a
t
o
T
I
I
h
c
a
e
y
b
d
l
e
h
e
u
l
a
v
k
c
o
t
s
l
a
t
o
T
s
k
c
o
t
s
.
o
N
F
M
h
c
a
e
y
b
d
l
e
h
s
F
M
y
b
d
l
e
h
s
k
c
o
t
s
.
o
N
)
s
n
o
i
l
l
i
b
(
)
s
n
o
i
l
l
i
b
(
s
I
I
y
b
d
l
e
h
I
I
r
e
p
s
k
c
o
t
s
.
o
N
s
I
I
.
o
N
)
s
n
o
i
l
l
i
b
(
)
s
n
o
i
l
l
i
b
(
F
M
r
e
p
s
k
c
o
t
s
.
o
N
s
F
M
o
N
.
3
5
1
0
.
5
8
1
0
.
6
5
1
0
.
7
4
1
0
.
0
7
1
0
.
9
9
1
0
.
6
0
2
0
.
7
2
2
0
.
5
4
2
0
.
4
8
1
0
.
7
5
1
0
.
6
9
1
0
.
7
0
2
0
.
7
2
2
0
.
9
5
2
0
.
4
8
2
0
.
2
6
2
0
.
9
5
2
0
.
3
8
2
0
.
1
1
2
0
.
7
6
2
8
4
3
4
1
3
0
0
3
7
4
3
5
3
4
4
9
4
5
7
5
1
8
6
4
4
5
9
5
4
9
6
5
4
3
6
2
2
7
3
6
8
2
2
0
1
8
0
0
1
3
3
0
1
8
6
1
1
0
2
6
9
0
1
3
2
1
8
2
1
0
3
1
9
3
1
4
4
1
1
4
1
2
4
1
2
4
1
4
3
1
5
3
1
2
4
1
8
4
1
2
5
1
9
5
1
5
6
1
1
6
1
9
5
1
2
6
1
3
4
1
2
6
5
2
1
6
1
1
6
9
2
6
7
3
6
4
5
6
9
6
6
5
6
6
5
6
6
2
4
6
2
4
6
1
5
6
1
7
6
6
7
6
1
9
6
9
9
6
6
0
7
6
9
6
2
9
6
6
5
6
1
5
7
1
5
7
8
1
2
1
0
2
3
4
0
2
6
4
0
2
7
8
1
2
1
0
4
2
9
2
5
2
0
8
7
2
0
6
9
2
3
2
9
2
0
0
9
2
0
7
0
3
2
8
1
3
0
4
3
3
3
9
5
3
3
5
8
3
9
8
9
3
0
3
1
4
9
1
8
2
9
8
0
0
.
2
0
1
0
.
1
9
0
0
.
7
8
0
0
.
8
9
0
0
.
1
3
1
0
.
2
5
1
0
.
3
8
1
0
.
4
9
1
0
.
3
5
1
0
.
6
5
1
0
.
8
0
2
0
.
4
1
2
0
.
1
4
2
0
.
9
1
3
0
.
7
8
3
0
.
7
8
3
0
.
9
1
4
0
.
1
4
5
0
.
9
1
2
0
.
1
0
3
1
8
3
1
5
3
8
4
3
8
9
3
5
2
5
6
0
6
8
1
7
5
4
8
9
1
7
8
8
6
7
5
8
7
2
9
4
4
0
1
9
5
3
1
5
2
7
1
3
9
7
1
4
2
0
2
3
7
5
2
7
5
9
2
5
1
6
9
6
5
7
9
7
2
8
4
8
6
8
7
8
1
9
8
9
8
9
6
9
9
9
0
0
1
8
9
9
9
0
0
1
9
9
7
8
6
7
5
8
0
6
4
1
6
8
1
6
7
2
6
4
4
6
7
5
6
3
5
6
2
4
6
8
2
6
8
2
6
8
3
6
9
4
6
7
5
6
3
7
6
0
8
6
3
8
6
3
7
6
3
5
6
2
4
6
4
6
3
3
6
2
7
3
3
6
8
3
4
8
9
3
2
7
0
4
3
0
0
4
2
0
0
4
3
2
9
3
5
6
3
4
4
9
6
4
3
1
4
4
3
1
1
4
3
3
3
4
8
2
3
4
4
5
2
4
4
5
4
4
0
4
6
4
6
3
8
4
2
5
7
4
7
1
2
4
r
a
e
Y
9
9
9
1
0
0
0
2
1
0
0
2
2
0
0
2
3
0
0
2
4
0
0
2
5
0
0
2
6
0
0
2
7
0
0
2
8
0
0
2
9
0
0
2
0
1
0
2
1
1
0
2
2
1
0
2
3
1
0
2
4
1
0
2
5
1
0
2
6
1
0
2
7
1
0
2
e
g
a
r
e
v
A
i
l
s
c
i
t
s
i
t
a
t
s
s
g
n
d
o
h
)
F
3
1
(
r
o
t
s
e
v
n
i
l
a
n
o
i
t
u
t
i
t
s
n
i
d
n
a
)
2
1
S
(
d
n
u
f
l
a
u
t
u
M
1
E
L
B
A
T
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 8
NGUYEN ET AL.
TA B L E 2 Credit rating sample statistics
Panel A: Rating changes
Egan-Jones
Ratings (EJR)
Standard and Poor
(S&P)
Moody’s Investors
Service (Moody’s)
Fitch Ratings
(Fitch)
Number of firms rated
Section 1: Negative events
Negative developing
Negative outlook
Negative combine event
Single downgrade
Multiple downgrade
Section 2: Positive events
Positive developing
Positive outlook
Positive combine event
Single upgrade
Multiple upgrade
1502
186
1648
2628
2013
415
741
1910
2504
1896
264
1432
–
1537
1172
1187
428
–
730
546
1342
114
645
–
440
355
370
187
–
299
172
541
48
502
–
278
200
578
163
–
80
64
549
70
Panel B: The distribution of rating changes
N
Mean
Std. dev.
P1
P25
Median
P75
P99
EJR negative event
EJR positive event
S&P negative event
S&P positive event
6885
1.041
7315
0.909
4325
1.050
2730
1.024
Moody’s negative event
1352
1.068
Moody’s positive event
1060
0.884
Fitch negative event
1219
1.174
Fitch positive event
762
1.177
0.758
0.740
0.933
0.961
0.736
0.423
1.054
0.867
0.250
0.500
0.250
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
1.000
1.000
0.750
0.500
1.000
1.000
1.000
1.000
1.000
1.250
4.000
1.000
3.750
1.000
5.000
1.000
5.500
1.500
3.500
1.000
2.500
1.000
5.500
1.000
5.000
Note: The table presents credit rating events announced by EJR (investor-paid credit rating agency [CRA]) and S&P, Moody’s
and Fitch (issuer-paid CRAs). Panel A displays the number of firms rated and the number of rating events (negative and pos-
itive separately) announced by each CRA after being merged with COMPUSTAT, CRSP and S12/13F data. Panel B presents
summary statistics for credit rating changes of each CRA, where the magnitude of a rating change is calculated as the total
number of notches by which a rating agency changes a firm’s credit rating.
consistent with EJR’s and Fitch’s policy of rating veteran firms. For instance, the mean market capitalization in the S12
(13F) sample in the third group is $36,874 ($44,837) million, while the number is $30,333 ($37,638) million for group
one and $12,836 ($16,899) million for group two.
The mean Z-scores in the S12 (13F) sample are 2.12 (2.11), 1.72 (1.70) and 1.84 (1.86) for the first, second and third
groups, respectively. These means are relatively close to the conventional threshold of 1.8 but above the risk level of a
financially healthy firm. Leverage ratios are similar across all three groups. Finally, the median interest coverage ratio
is slightly lower for firms in group two than for firms in groups one and three.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
9
TA B L E 3 Descriptive statistics for firms covered in S12 and 13F databases
Panel A: Mutual fund (S12) database
Firms rated by EJR
and S&P
Firms rated by EJR and
Moody’s
Firms rated by EJR
and Fitch
N
Mean Median
N
Mean Median
N
Mean
Median
Ln(MV)
ROA
IDIO_RISK
Z-SCORE
2,808,671 10.32
9.37
1,079,231
2,808,671
0.04
0.04
1,079,231
2,808,671
0.02
0.01
1,079,231
2,808,671
2.12
1.82
1,079,231
ANALYST_COVERAGE 2,808,671
7.15
6.78
1,079,231
Ln(AGE)
2,808,671
3.21
3.33
1,079,231
9.46
0.03
0.02
1.72
6.33
2.99
8.59
1,716,964 10.51
0.04
1,716,964
0.04
0.02
1,716,964
0.01
1.54
1,716,964
1.84
5.84
1,716,964
7.38
3.00
1,716,964
3.31
INTEREST_COVERAGE 2,808,671 15.65
9.26
1,079,231
11.78
7.15
1,716,964 14.39
LEVERAGE
S&P_500
2,808,671
0.33
0.27
1,079,231
2,808,671
0.65
1.00
1,079,231
HIGH_TECH
2,808,671
0.03
0.00
1,079,231
0.33
0.48
0.02
0.31
1,716,964
0.37
0.00
1,716,964
0.76
0.00
1,716,964
0.02
9.65
0.04
0.01
1.63
7.00
3.47
8.59
0.28
1.00
0.00
Panel B: Institutional investors (13F) database
Firms rated by EJR
and S&P
Firms rated by EJR and
Moody’s
Firms rated by EJR
and Fitch
N
Mean Median
N
Mean Median
N
Mean
Median
Ln(MV)
ROA
IDIO_RISK
Z-SCORE
3,180,369 10.54
9.65
1,084,625
3,180,369
0.04
0.04
1,084,625
3,180,369
0.01
0.01
1,084,625
3,180,369
2.11
1.86
1,084,625
ANALYST_COVERAGE 3,180,369
7.44
7.02
1,084,625
Ln(AGE)
3,180,369
3.32
3.40
1,084,625
9.73
0.03
0.02
1.70
6.53
3.05
8.75
2,040,834 10.71
0.04
2,040,834
0.04
0.02
2,040,834
0.01
1.51
2,040,834
1.86
6.00
2,040,834
7.62
3.04
2,040,834
3.44
INTEREST_COVERAGE 3,180,369 16.01
9.80
1,084,625
12.65
7.29
2,040,834 15.12
LEVERAGE
S&P_500
3,180,369
0.35
0.28
1,084,625
3,180,369
0.68
1.00
1,084,625
HIGH_TECH
3,180,369
0.03
0.00
1,084,625
0.33
0.49
0.02
0.31
2,040,834
0.38
0.00
2,040,834
0.76
0.00
2,040,834
0.02
9.87
0.04
0.01
1.66
7.30
3.56
9.26
0.28
1.00
0.00
Note: The table presents the summary statistics of control variables, which are defined in Appendix B. Statistics are computed
around credit rating announcements.
3
MAIN RESULTS
3.1
Institutional responses to issuer- and investor-paid rating adjustments
We now examine institutional investors’ responses to credit rating signals announced by issuer- and investor-paid
CRAs. To ensure that reactions are comparable, we construct three paired samples, which include firms rated by EJR
and each of the major issuer-paid CRAs: EJR and S&P, EJR and Moody’s and EJR and Fitch. We estimate the following
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 10
NGUYEN ET AL.
regression for each of the paired samples:
NBi, n,q = 𝛼 + 𝛽1NEGn,q + 𝛽2POSn,q + 𝛽3NEGn,q∗EJRn,q + 𝛽4POSn,q∗EJRn,q
𝛾kCONTROLS(n,q) +q
1
+ 𝛽5EJR(n,q) +
𝜃qQuarterFEq
∑q
1
∑k
1
+
𝛿iInvestorFEi +
𝜑nFirmFEn + 𝜀i,n,q
i∑
1
n∑
1
,
(4)
∑
∑
q and denote it by
where NBi,n,q, defined in equation (2), denotes mutual fund (institutional investor) i’s abnormal dollar net buy of firm n’s
stock for credit rating adjustments in quarter q. We sum all ΔCCRs, as defined in equation (1), for each firm n in quarter
∑
ΔCCRn,q > 0 and
ΔCCRn,q < 0. Therefore, an increase in NEGn,q (POSn,q) represents an
POSn,q as
absolute increase in aggregate credit rating downgrade (upgrade) for firm n in quarter q. CONTROLSn,q represents a set
of firm-level control variables as described in Table 3. QuarterFEq denotes quarter-specific dummy variables to control
for differences in institutional trading behavior that can be induced by various economic conditions in different quar-
ΔCCRn,q.11 We then define NEGn,q as |
∑
ΔCCRn,q < 0 and zero if
ΔCCRn,q > 0 and zero if
ΔCCRn,q| if
ΔCCRn,q if
∑
∑
∑
ters. InvestorFEi (FirmFEn) is used to control for investor- (firm-) characteristics that are not captured by CONTROLSn,q.
In this model, NEGn,q and POSn,q are interacted with EJRn,q, a dummy variable that equals one for EJR’s credit rating
announcements and zero otherwise. 𝜀i,n,q is a random error.
The results of equation (4) are presented in Table 4. We find significant asymmetries in the abnormal trading of insti-
tutional investors and mutual funds in relation to EJR’s and issuer-paid CRAs’ rating announcements. These results
are robust to the inclusion of control variables and fixed effects. For example, for firms that are rated by EJR and
S&P, columns 1 and 2 show significant increases in mutual funds’ and institutional investors’ net buy of stocks with
an aggregate positive change in S&P’s rating adjustments in a given quarter. The POS coefficient is positive and signifi-
cant across the regression specifications. Its magnitude is also economically meaningful. For example, the 0.0571 basis
point coefficient in column 2 of panel B is equivalent to an average increase of $215,824 in abnormal institutional net
buy over the respective quarter with a one-notch upgrade.12 Institutional investors, however, react significantly less
to positive rating changes issued by EJR. The EJR*POS interaction coefficient is negative in almost every model. The F-
test results for the overall impact of rating upgrades by EJR, that is, the sum of POS and EJR*POS coefficients, indicate
that both mutual funds and institutional investors are unresponsive to EJR’s positive rating changes.
Table 4 shows the opposite results for rating downgrades. Institutional investors and mutual funds apparently find
EJR’s negative rating adjustments more informative than S&P’s announcements. While the NEG coefficient shows no
clear pattern across specifications, the EJR*NEG is negative and statistically and economically significant across the
models. For example, the −0.1368 coefficient of EJR*NEG in column 2 of panel B indicates that a firm receiving a one-
notch rating downgrade by EJR experiences an average decrease of $517,077 in abnormal institutional net buy over
the respective quarter, compared to a similar downgrade by S&P. The F-test results for the overall impact of rating
downgrades by EJR, that is, the sum of NEG and EJR*NEG coefficients, indicate that the total effect of EJR downgrades
is statistically and economically significant.
We find similar asymmetric responses by institutional investors to upgrades and downgrades for firms that are
rated by EJR and Moody’s. For example, a POS coefficient of 0.0931 in column 4 indicates that abnormal institutional
net buy, on average, increases by $156,543 over the quarter in which a one-notch aggregate rating upgrade by Moody’s
takes place. The combined effect of POS + EJR*POS shows that institutional investors do not react to EJR’s upgrades
as opposed to Moody’s. However, the results for rating downgrades support the notion that institutional investors
respond to EJR’s rather than Moody’s downgrades. The EJR*NEG coefficient in column 4 of Panel B is −0.2707, indi-
cating that EJR downgrades are associated with, on average, a decrease of $455,170 in abnormal institutional net
buy, compared to Moody’s downgrades. The significant F-test results for the sum of NEG and EJR*NEG in columns (3)
11 In our analysis, we exclude firm-quarter observations that EJR and the paired issuer-paid CRA have different credit rating signals in a quarter.
12 The increase is calculated by multiplying the POS coefficient of 0.0571 by the average market capitalization (e10.54 = $37,798 million) of firms in the EJR
and S&P group in panel B of Table 3 and dividing the result by 10,000 (since the net buy is in basis points).
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
11
TA B L E 4 Abnormal trading responses to credit rating adjustments—S12 and 13F samples
Panel A: Mutual funds’ abnormal holding changes
EJR versus S&P
EJR versus Moody’s
EJR versus Fitch
(1)
(2)
(3)
(4)
(5)
(6)
Intercept
0.1379***
−0.0856**
0.4049***
−0.3387***
0.1770***
−0.5467***
NEG
POS
(0.0183)
(0.0346)
(0.0445)
(0.0828)
(0.0234)
(0.0481)
0.0105*
−0.0013
−0.004
0.0234
0.0211**
0.0286**
(0.0062)
(0.0073)
(0.024)
(0.0283)
(0.009)
(0.0122)
0.0318***
0.0324***
0.0241**
0.0308**
0.0441***
0.0612***
(0.0087)
(0.0108)
(0.0121)
(0.0123)
(0.0127)
(0.0179)
EJR×NEG
−0.0199***
−0.0103*
−0.0416*
−0.0512*
−0.0139
−0.0187
EJR×POS
−0.0257***
−0.0262**
−0.0123
−0.0152
−0.0424***
−0.0587***
(0.0074)
(0.0057)
(0.0253)
(0.0292)
(0.0102)
(0.0135)
EJR
−0.2294***
−0.2410***
−0.5808***
−0.6357***
−0.2948***
−0.2884***
(0.0097)
(0.012)
(0.0303)
(0.0337)
(0.0138)
(0.0191)
Control variables:
No
Yes
No
Yes
No
Yes
(0.0084)
(0.0099)
(0.0267)
(0.030)
(0.0124)
(0.0165)
F-tests:
NEG + EJR×NEG
−0.0094**
−0.0115**
−0.0456***
−0.0278***
0.0073
0.0100
POS + EJR×POS
0.0061
0.0062
0.0118
0.0156
0.0017
0.0025
(0.0042)
(0.005)
(0.0084)
(0.0096)
(0.0051)
(0.0063)
(0.0044)
(0.0056)
(0.0083)
(0.0095)
(0.0055)
(0.0072)
Fixed effects:
Investor FE
Firm FE
Quarter FE
N
Adj. R2
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
3,582,992
2,808,671
1,273,265
1,079,231
2,291,401
1,716,964
0.002
0.003
0.003
0.004
0.002
0.002
Panel B: Institutional investors’ abnormal holding changes
EJR versus S&P
EJR versus Moody’s
EJR versus Fitch
(1)
(2)
(3)
(4)
(5)
(6)
Intercept
0.6663***
2.0098***
2.4527***
2.7954***
0.2020***
−1.3221***
NEG
POS
(0.0548)
(0.1136)
(0.1483)
(0.2731)
(0.0568)
(0.1378)
−0.0278
−0.0665***
0.0124
0.0511
−0.0383
0.0785**
(0.0219)
(0.0257)
(0.0827)
(0.0959)
(0.0234)
(0.0346)
0.0330*
0.0571**
0.1087***
0.0931***
0.0893**
0.1089**
(0.0181)
(0.0285)
(0.0325)
(0.0224)
(0.0352)
(0.0531)
EJR×NEG
−0.1555***
−0.1368***
−0.2474***
−0.2707***
−0.0083
−0.0624
(0.026)
(0.0305)
(0.0875)
(0.0992)
(0.0273)
(0.0387)
(Continues)
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 12
TA B L E 4
(Continued)
NGUYEN ET AL.
Panel B: Institutional investors’ abnormal holding changes
EJR versus S&P
EJR versus Moody’s
EJR versus Fitch
(1)
(2)
(3)
(4)
(5)
(6)
EJR×POS
0.0071
−0.0288
−0.0956***
−0.0923***
−0.1045***
−0.1373**
EJR
−0.7735***
−1.0401***
−2.6222***
−2.6376***
−0.3948***
−0.3114***
(0.035)
(0.0433)
(0.0276)
(0.027)
(0.039)
(0.0575)
Control variables:
No
Yes
No
Yes
No
Yes
(0.0292)
(0.0342)
(0.0974)
(0.1075)
(0.0336)
(0.0466)
F-tests:
NEG + EJR×NEG
−0.1833***
−0.2034***
−0.2350***
−0.2196***
−0.0467*
0.0162
POS + EJR×POS
0.0401
0.0283
0.0131
0.0008
−0.0151
−0.0285
(0.0145)
(0.0174)
(0.0305)
(0.0334)
(0.0247)
(0.0189)
(0.0265)
(0.0207)
(0.033)
(0.0361)
(0.0173)
(0.0238)
Fixed effects:
Investor FE
Firm FE
Quarter FE
N
Adj. R2
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
4,088,703
3,180,369
1,281,861
1,084,625
2,729,378
2,040,834
0.001
0.001
0.002
0.002
0.001
0.001
Note: The table reports OLS regression results for mutual fund (S12) and institutional investor (13F) abnormal holding changes
in response to credit rating adjustments announced by EJR and issuer-paid CRAs. The dependent variable, defined in equa-
tion (2), is a mutual fund’s (institutional investor’s) abnormal net buy of a stock during a quarter. Based on a firm’s aggregate
credit rating change in a quarter, we define NEG as the absolute value of a negative change and zero otherwise and POS as
the value of a positive change and zero otherwise. Therefore, an increase in NEG (POS) represents an absolute increase in the
firm’s aggregate downgrade (upgrade) in that quarter. EJR is a dummy variable that equals one for EJR’s credit rating announce-
ments and zero otherwise. Detail descriptions of firm-level control variables are described in Appendix B. Standard errors in
parentheses are adjusted for heteroskedisticity and clustering at the firm and quarter levels.
***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.
and (4) indicate that the total effect of EJR downgrades is statistically and economically strong. The results for firms
jointly rated by EJR and Fitch in columns 5 and 6 are less clear. While the POS coefficient is significant, indicating that
investors react to Fitch’s upgrades, EJR*NEG and EJR + EJR*NEG are mostly insignificant. We investigate this further in
Section 3.5.
Overall, the results in Table 4 suggest that mutual funds and institutional investors find that credit rating upgrades
are more informative; hence, they respond accordingly when issued by S&P or Moody’s rather than by EJR. In contrast,
they find that negative rating adjustments are more value-relevant when they are announced by EJR than by S&P or
Moody’s. These findings are consistent with the argument that institutional investors are well-equipped to assess the
informativeness of credit rating announcements. Previous studies have shown that issuer-paid CRAs tend to delay rat-
ing downgrades due to conflict of interests (e.g., Cornaggia & Cornaggia, 2013) but still issue timely rating upgrades
(e.g., Kedia et al., 2017). Brogaard et al. (2019) also find that upgrades issued by issuer-paid CRAs do convey new infor-
mation. In contrast, investor-paid CRAs tend to be more timely in rating downgrade adjustments (e.g., Berwart et al.,
2019; Johnson, 2004).
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
13
3.2
Do CRAs behave the way we assume they do?
The findings in the previous section that institutional investors respond more to positive rating announcements by
major issuer-paid CRAs and to negative rating announcements by the investor-paid EJR suggest a lead-lag in the
timeliness of credit rating announcements between these two types of CRAs. We now empirically examine this.
As before, we separately consider three pairs: EJR and S&P, EJR and Moody’s and EJR and Fitch. For each firm
rated by each pair of CRAs, the credit rating score is adjusted multiple times by two paired CRAs throughout the
sample period. We investigate the lead–lag relationship of each CRA pair for upgrades and downgrades separately.
Based on the announcement timeline and the relative magnitude of consecutive rating adjustments, three scenarios
are possible. First, when one CRA issues a rating adjustment that is relatively larger in magnitude than the subsequent
adjustment announced by the other CRA, the leading CRA is classified as a “major leader.” Second, when one CRA
issues a rating adjustment relatively smaller in magnitude than the subsequent adjustment announced by the other
CRA, the following CRA is classified as a “major confirmer.” Third, if a rating adjustment by one CRA is followed by an
adjustment of the same magnitude by the other CRA, we classify the leading CRA as an “equal magnitude leader.” We
then perform a binominal test with the null hypothesis that the relative frequencies that both CRAs in a pair hold for a
specific role are equal.
In Table 5, section 1 reports the results for negative events, and section 2 shows the results for positive events.
Panels A, B and C present the results for EJR and S&P, EJR and Moody’s and EJR and Fitch, respectively. The results
generally confirm our expectations that EJR issues relatively larger rating adjustments than the issuer-paid CRAs
when these adjustments are downgrades. For example, EJR’s downgrades are larger than S&P’s subsequent down-
grades 56.95% ( = 422/(422 + 319)) of the time, which is statistically higher than 43.05% of the time when S&P plays
the role of a major leader. The comparison is even higher for EJR than Moody’s in Panel B, at 67.14% ( = 141/(141 +
69)) versus 32.86%. When EJR follows S&P or Moody’s after their respective negative rating adjustments, EJR tends to
issue larger negative adjustments more frequently than when the other two CRAs follow EJR’s downgrades with larger
magnitudes. The major confirmer row for negative events confirms these differences statistically. There are no statisti-
cal differences between EJR and the other CRAs in the frequency of being an equal magnitude leader. However, we find
no evidence of EJR’s leading role, compared to Fitch in the issuance of negative signals. Fitch apparently issues larger
negative adjustments more frequently than EJR, although these frequency differences are not statistically significant.
The results for positive events in Section 2 of Table 5 indicate that all three issuer-paid CRAs tend to issue larger
rating upgrades more frequently than EJR. These frequency differences are statistically significant for both cases when
these traditional CRAs are major leaders or major confirmers. There are no significant frequency differences in being
an equal magnitude leader, except for the EJR and Fitch pair where EJR leads Fitch more often when they issue positive
rating adjustments of the same magnitude.
Overall, the findings in this table support the results in Table 4 that EJR’s negative rating announcements are
apparently more timely and value-relevant to institutional investors than those rating downgrades by the other CRAs.
However, the issuer-paid CRAs’ positive rating announcements are valued more by institutional investors than EJR’s
rating upgrades.
3.3
Profitability of asymmetric trading strategies
3.3.1
Notional trading strategies
We now investigate whether a trading strategy based on credit rating signals with the highest information content
can generate superior returns. Credit rating announcements are, in principle, available to all investors—not just
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14
NGUYEN ET AL.
e
u
l
a
v
-
p
h
c
t
i
F
R
J
E
e
u
l
a
v
-
p
s
’
y
d
o
o
M
R
J
E
e
u
l
a
v
-
p
P
&
S
R
J
E
h
c
t
i
F
d
n
a
R
J
E
:
C
l
e
n
a
P
s
’
y
d
o
o
M
d
n
a
R
J
E
:
B
l
e
n
a
P
P
&
S
d
n
a
R
J
E
:
A
l
e
n
a
P
7
9
6
2
0
.
7
1
1
1
0
.
5
2
8
7
0
.
1
0
0
0
0
<
.
1
0
0
0
0
<
.
)
.
%
8
3
3
5
(
2
4
1
)
.
%
2
6
6
4
(
4
2
1
1
0
0
0
0
<
.
)
.
%
6
8
2
3
(
9
6
)
.
%
4
1
7
6
(
1
4
1
2
0
0
0
0
.
)
.
%
5
0
3
4
(
9
1
3
)
.
%
5
9
6
5
(
2
2
4
)
t
(
r
e
d
a
e
l
r
o
j
a
M
)
.
%
6
0
5
5
(
6
3
1
)
.
%
4
9
4
4
(
1
1
1
2
1
0
0
0
.
)
.
%
4
8
8
3
(
2
8
)
.
%
4
1
1
6
(
9
2
1
2
1
0
0
0
.
)
.
%
2
1
4
4
(
4
3
3
)
.
%
8
8
5
5
(
3
2
4
)
1
+
t
(
r
e
m
r
i
f
n
o
c
r
o
j
a
M
)
.
%
5
9
0
5
(
7
0
1
)
.
%
5
0
9
4
(
3
0
1
8
9
4
5
0
.
)
.
%
5
4
7
4
(
5
6
)
.
%
5
5
2
5
(
2
7
7
9
9
2
0
.
)
.
%
5
8
7
4
(
8
7
2
)
.
%
5
1
2
5
(
3
0
3
)
t
(
r
e
d
a
e
l
e
d
u
t
i
n
g
a
m
l
a
u
q
E
s
t
n
e
v
e
e
v
i
t
a
g
e
N
:
1
n
o
i
t
c
e
S
)
.
%
3
7
2
7
(
2
1
1
)
.
%
7
2
7
2
(
2
4
2
1
2
0
0
.
)
.
%
6
5
8
5
(
6
0
1
)
.
%
4
4
1
4
(
5
7
)
.
%
7
7
0
8
(
6
2
1
)
.
%
3
2
9
1
(
0
3
3
7
9
0
0
.
)
.
%
5
2
6
5
(
9
9
)
.
%
5
7
3
4
(
7
7
1
0
0
0
0
<
.
1
0
0
0
0
<
.
)
.
%
8
4
1
6
(
5
6
2
)
.
%
2
5
8
3
(
6
6
1
)
t
(
r
e
d
a
e
l
r
o
j
a
M
)
.
%
8
4
3
6
(
9
9
2
)
.
%
2
5
6
3
(
2
7
1
)
1
+
t
(
r
e
m
r
i
f
n
o
c
r
o
j
a
M
s
t
n
e
v
e
e
v
i
t
i
s
o
P
:
2
n
o
i
t
c
e
S
4
4
5
0
0
.
)
.
%
6
9
1
4
(
0
6
)
.
%
4
0
8
5
(
3
8
6
4
4
9
0
.
)
.
%
4
2
0
5
(
4
0
1
)
.
%
6
7
9
4
(
3
0
1
7
6
2
9
0
.
)
.
%
9
7
9
4
(
5
3
2
)
.
%
1
2
0
5
(
7
3
2
)
t
(
r
e
d
a
e
l
e
d
u
t
i
n
g
a
m
l
a
u
q
E
d
e
t
a
r
s
m
r
i
f
r
o
f
s
t
n
e
m
e
c
n
u
o
n
n
a
r
o
f
e
r
a
C
d
n
a
B
,
A
s
l
e
n
a
P
.
s
t
n
e
m
e
c
n
u
o
n
n
a
g
n
i
t
a
r
t
i
d
e
r
c
e
v
i
t
a
g
e
n
d
n
a
e
v
i
t
i
s
o
p
n
i
s
A
R
C
d
i
a
p
-
r
o
t
s
e
v
n
i
d
n
a
-
r
e
u
s
s
i
l
f
o
e
o
r
e
v
i
t
a
l
e
r
e
h
t
s
w
o
h
s
e
b
a
t
l
s
i
h
T
:
e
t
o
N
n
a
h
t
e
d
u
t
i
n
g
a
m
n
i
r
e
g
r
a
l
y
l
e
v
i
t
a
l
e
r
s
i
t
a
h
t
j
t
n
e
m
t
s
u
d
a
g
n
i
t
a
r
a
s
e
u
s
s
i
A
R
C
e
n
o
n
e
h
w
,
t
s
r
i
F
.
s
o
i
r
a
n
e
c
s
e
e
r
h
t
r
e
d
i
s
n
o
c
e
W
’
.
)
h
c
t
i
F
d
n
a
s
y
d
o
o
M
P,
&
S
(
s
A
R
C
”
e
e
r
h
T
g
i
B
“
e
h
t
f
o
e
n
o
d
n
a
R
J
E
y
b
n
i
r
e
l
l
a
m
s
y
l
e
v
i
t
a
l
e
r
s
i
t
a
h
t
j
t
n
e
m
t
s
u
d
a
g
n
i
t
a
r
a
s
e
u
s
s
i
A
R
C
e
n
o
n
e
h
w
,
d
n
o
c
e
S
”
.
r
e
d
a
e
l
r
o
j
a
m
“
a
s
a
d
e
i
f
i
s
s
a
l
c
s
i
A
R
C
g
n
d
a
e
i
l
e
h
t
,
A
R
C
r
e
h
t
o
e
h
t
y
b
d
e
c
n
u
o
n
n
a
t
n
e
m
t
s
u
d
a
j
t
n
e
u
q
e
s
b
u
s
e
h
t
n
a
y
b
d
e
w
o
l
l
o
f
s
i
A
R
C
e
n
o
y
b
t
n
e
m
t
s
u
d
a
g
n
i
t
a
r
a
f
i
j
,
d
r
i
h
T
”
.
r
e
m
r
i
f
n
o
c
r
o
j
a
m
“
a
s
a
d
e
i
f
i
s
s
a
l
c
s
i
A
R
C
g
n
w
o
i
l
l
o
f
e
h
t
,
A
R
C
r
e
h
t
o
e
h
t
y
b
d
e
c
n
u
o
n
n
a
t
n
e
m
t
s
u
d
a
t
n
e
u
q
e
s
b
u
s
e
h
t
n
a
h
t
e
d
u
t
i
n
g
a
m
j
n
i
(
y
c
n
e
u
q
e
r
f
e
v
i
t
a
l
e
r
e
h
t
d
n
a
s
e
m
i
t
f
o
r
e
b
m
u
n
e
h
t
w
o
h
s
e
w
,
r
i
a
p
h
c
a
e
r
o
F
”
.
r
e
d
a
e
l
e
d
u
t
i
n
g
a
m
l
a
u
q
e
“
n
a
s
a
A
R
C
g
n
d
a
e
i
l
e
h
t
y
f
i
s
s
a
l
c
e
w
,
A
R
C
r
e
h
t
o
e
h
t
y
b
e
d
u
t
i
n
g
a
m
e
m
a
s
e
h
t
f
o
t
n
e
m
t
s
u
d
a
j
.
l
e
o
r
c
i
f
i
c
e
p
s
a
n
i
r
i
a
p
A
R
C
h
c
a
e
f
o
y
c
n
e
u
q
e
r
f
e
v
i
t
a
l
e
r
e
h
t
e
r
a
p
m
o
c
o
t
t
s
e
t
i
l
a
i
m
o
n
b
a
m
o
r
f
s
i
e
u
l
a
v
-
p
.
l
e
o
r
c
i
f
i
c
e
p
s
a
s
d
o
h
A
R
C
a
t
a
h
t
l
)
s
t
e
k
c
a
r
b
s
l
a
n
g
i
s
e
v
i
t
i
s
o
p
d
n
a
e
v
i
t
a
g
e
n
n
i
s
A
R
C
d
i
a
p
-
r
o
t
s
e
v
n
i
d
n
a
-
r
e
u
s
s
i
l
f
o
e
o
r
e
v
i
t
a
l
e
r
e
h
T
5
E
L
B
A
T
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
15
institutions. Therefore, we begin by analyzing “notional” trading strategies available to a hypothetical investor with
timely access to credit ratings.
The first strategy we consider is the “dynamic strategy”—selling following EJR’s negative rating signals and buy-
ing following issuer-paid CRAs’ positive rating signals. This trading strategy is our main interest. The second one is
the ‘naïve strategy’—selling following negative signals and buying following positive signals from any rating agency.
The third strategy is the “EJR-based strategy”– selling following negative signals and buying following positive signals
announced by EJR. The fourth strategy is the “issuer-paid CRA-based strategy”—selling following negative signals and
buying following positive signals issued by any of the “Big Three” CRAs. We also add a passive strategy as an additional
benchmark—investing in the S&P 500 index.
We measure the profitability for each trading strategy as follows. First, we examine various holding periods of k
months (where k = 1, 3, 6, 9 and 12) starting from the rating announcement date t to day t + 5. We follow Jagolinzer
et al. (2011) and estimate abnormal returns after adjusting for common risk factors. Specifically, for each day in the [0,
5] window, risk-adjusted return is the intercept (alpha) from the Fama and French (2015) five-factor model estimated
over a holding period of k months:
(Rn − Rf ) = 𝛼 + 𝛽1
(
Rmkt − Rf
)
+ 𝛽2SMB + 𝛽3HML + 𝛽4RMW + 𝛽5CMA + ei,
(5)
where Rn is the daily return of firm n; Rf is the daily risk-free rate; Rmkt is the CRSP (Center for Research in Secu-
rity Prices) value-weighted market return; SMB, HML, RMW and CMA are size, book-to-market, operating profitability
and investment factors, respectively.13 For notional strategies, we assume that investors trade in accordance with a
credit rating signal, that is, selling (buying) if the signal is negative (positive). Therefore, if the announcement is a rat-
ing downgrade, we multiply daily alphas in equation (5) by (−1) to represent risk-adjusted returns to investors’ sales.
This adjustment does not apply for investors’ purchases following a rating upgrade. We calculate the risk-adjusted
alpha for firm n’s rating announcement t as the simple average of alphas over the [0, 5] window and denote it by αn,t.
We use equal weightings to calculate the firm’s mean alpha for each announcement event as with these hypothetical
transactions we do not have data on investors’ buy and sell values.
The event alphas, αn,t, are then grouped into appropriate trading strategies described above, and a t-test is per-
formed across all rating announcements in a given strategy. We also test the mean difference in the value-weighted
risk-adjusted returns (i.e., weighted by market capitalization) between strategies with a two-sample t-test and report
the results in Table 6. 14 All returns are annualized. We find that all four strategies outperform the buy-and-hold of the
S&P 500 index. In addition, consistent with our expectations, the dynamic strategy yields higher abnormal returns than
all other strategies. Over the 1-month investment horizon, the dynamic strategy outperforms the other three strate-
gies by an annualized risk-adjusted return ranging from 4.22% to 5.02%. Its outperformance is statistically significant
for up to 6 months.
3.3.2
Institutional trading strategies
We now examine trading strategies based on institutional transactions. The returns on notional strategies can be
interpreted as equally weighted returns of an institution trading around every credit rating announcement consis-
tent with a certain strategy. By explicitly considering institutional transactions, we acknowledge that institutions may
13 We thank Kenneth French for sharing data on the five risk factors in his website, http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.
html
14 We adjust firms’ market capitalization for inflation following the merit of Acharya and Pedersen (2005). Specifically, we first calculate the ratio of CRSP
total market value at the end of month m – 1 (relative to the credit event month) to CRSP total market value at the end of 1998 (just before our sample
starts). We then divide a firm’s market capitalization in month m by this ratio before using it as a weight in the t-test. Our (unreported) results are robust when
unadjusted market capitalization is used.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 0.0259
(0.0248)
0.0035
(0.0156)
(0.0034)
−0.0068
(0.0358)
0.0011
(0.0012)
0.0224
(0.0293)
0.0102
(0.025)
0.0327
(0.0435)
0.0248
(0.0248)
0.0024
(0.0157)
16
NGUYEN ET AL.
TA B L E 6 Notional trading strategy profitability
Holding periods
(1) Dynamic
1 month
3 months
6 months
9 months
12 months
0.0824***
0.0655***
0.0462***
0.0317***
(0.0104)
(0.0079)
(0.0061)
(0.0109)
(2) Naïve
0.0356***
0.0349***
0.0203***
0.0141**
(0.0075)
(0.0046)
(0.0038)
(0.0068)
(3) EJR-based
0.0322***
0.0327***
0.0344***
0.0147***
0.0156***
(0.0086)
(0.0041)
(0.003)
(0.0023)
(4) Issuer-paid CRA-based
0.0402***
0.0369***
0.019**
(5) S&P 500 index
(0.0134)
0.0011
(0.0012)
(0.0093)
0.0011
(0.0012)
(0.008)
0.0011
(0.0012)
(0.0012)
0.0132
(0.016)
0.0011
0.0468***
0.0307***
0.0259***
0.0176
(0.0129)
(0.0092)
(0.0072)
(0.0129)
0.0502***
0.0328***
0.0118*
0.017
(0.0135)
(0.0089)
(0.0068)
(0.0112)
0.0422**
0.0286**
0.0272***
0.0185
(0.0169)
(0.0122)
(0.0101)
(0.0194)
0.0813***
0.0644***
0.0451***
0.0306***
(0.0105)
(0.008)
(0.0062)
0.0345***
0.0337***
0.0192***
(0.011)
0.013*
(0.0076)
(0.0048)
(0.004)
(0.0069)
(1)–(2)
(1)–(3)
(1)–(4)
(1)–(5)
(2)–(5)
(3)–(5)
(4)–(5)
0.0311***
0.0316***
0.0333***
0.0136***
0.0145***
(0.0087)
(0.0043)
(0.0032)
(0.0026)
0.0391***
0.0358***
0.0179**
(0.0134)
(0.0094)
(0.0081)
0.0121
(0.016)
(0.0036)
−0.0079
(0.0358)
Note: This table reports and compares annualized risk-adjusted returns on four notional trading strategies. “Dynamic” is a
strategy that sells a stock when it receives an EJR’s negative rating adjustment and buys the stock when its rating is upgraded
by an issuer-paid CRA. The “naïve” strategy is simply to sell (buy) a stock following a negative (positive) signal from any rating
agency. For the “EJR-based” strategy, an investor sells (buys) a stock when EJR announces a downgrade (upgrade) in the stock’s
credit rating. The “issuer-paid CRA-based” strategy involves selling (buying) a stock following a rating downgrade (upgrade)
from an issuer-paid CRA. A buy-and-hold of the S&P 500 index is included as a benchmark strategy. We measure the prof-
itability for each trading strategy as follows. For each day in the time window [0, 5] surrounding each rating announcement
on a firm, risk-adjusted return is the intercept (or alpha) from the Fama–French five-factor model estimated over a holding
period. The firm-event alpha is then calculated as a simple average of the estimated alphas in the assessment window. We mul-
tiply the firm-event alpha by (−1) to represent risk-adjusted returns to investors’ sales following a rating downgrade. Based
on the firm’s rating change, we assign its event alpha to one of the trading strategies. Finally, we use firms’ inflation-adjusted
market capitalization as weights and assess the strategies’ performance using one and two sample t-tests. Standard errors of
the t-test for the mean and difference in means are in parentheses.
***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.
follow multiple strategies at a time, switch in and out of strategies, and may not trade on every signal consistent with a
given strategy. Since holding data in S12 and 13F forms are available on a quarterly basis, we assume that institutional
holding adjustments (and aggregate credit rating changes) happen on the last day of each quarter.15
15 In Section 4.2, we assume that trading occurs on the first day of each quarter. Our findings are robust.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
17
∑
First, given firm n’s aggregate rating change in quarter q,
We form trading strategies and estimate their returns in calendar time at the institutional investor level as follows.
ΔCCRn,q, and institutional investor i’s dollar net buy of the
firm’s stock, nbi,n,q, we classify this firm into a specific strategy. Since we assume that an institutional investor can con-
currently follow multiple strategies, a firm can be assigned to more than one trading strategy. We multiply the stock’s
ΔCCRn,q < 0 and nbi,n,q < 0 to reflect stock returns to sales following a rating downgrade. Next,
returns by (−1) if
for each institutional investor i, we calculate portfolio returns for each strategy using the absolute values of stock net
buy as weights.16 These portfolio returns are computed for each day over an investment horizon of k months starting
∑
at the end of the quarter. We repeat this process for each quarter in our sample period. Finally, the risk-adjusted
return for each strategy is obtained using a pooled cross-sectional time series regression of the Fama–French
five-factor model. We use a two-sample t-test to examine the difference in mean risk-adjusted returns between
strategies.
The strategy alphas are reported in Table 7. Generally, while all four trading strategies provide positive risk-adjusted
profits for up to 12 months after credit rating announcements, the dynamic strategy that mimics the typical institu-
tional response to credit rating adjustments yields the highest returns. For example, its 1-month annualized return is
17.54% and 16.73% for the mutual fund and institutional samples, respectively. While the return magnitude decreases
with the holding period, it significantly outperforms all other strategies for the holding periods of at least 6 months.
Among the other strategies, the strategy following issuer-paid CRAs’ credit announcements, while still significantly
outperforming the buy-and-hold, yields the lowest returns.
Overall, the results in Table 7 are consistent with our expectations that credit rating announcements have valuable
information content and that the most value-relevant announcements are downgrades by the investor-paid EJR and
upgrades by the issuer-paid CRAs. Our findings illustrate that institutional investors that dynamically change their
trading behavior based on the advantages and disadvantages of credit rating information are likely to make abnormal
profits beyond those of naïve trading strategies. Having said that, the prolonged abnormal profits are consistent with
underreaction to credit rating information, especially to the information content of EJR’s credit downgrades and the
issuer-paid CRAs’ credit upgrades.
3.4
Alternative institutional trading data
In the preceding analysis of institutional trading, we rely on S12 and 13F quarterly data. In this section, we conduct
the analysis using daily transaction-level data provided by the Abel Noser Corporation. G. Hu et al. (2018) describe
several important features of Abel Noser’s institutional trading data. The dataset covers at least 12% of the total
CRSP trading volume, 233 million transactions with $37 trillion in traded volume. It also records equity transactions
by a large number of institutions from January 1999 to September 2011.17 Despite its limited availability, the data
on institutional investors’ daily trading activities enable us to better capture their trading responses to credit rating
adjustments.18 We winsorize institutional trading data at the 1st and 99th percentiles to minimize the effect of out-
liers. After matching with our credit rating samples, we find 1126, 1259, 509 and 420 firms rated by EJR, S&P, Moody’s
and Fitch, respectively.
16 As a robustness check, we calculate each strategy’s returns using stock returns and their associated institutional net buy values across all institutional
investors in the quarter. Hence, we have one value-weighted return per strategy per day in a holding horizon. The unreported results are robust in both
statistical significance and economic magnitude.
17 While Abel Noser arguably provides “cleaner” transaction-level data, we rely on S12 and 13F data in the main analysis, as Abel Noser does not provide data
for research purposes after 2011. We note that our results are robust to the choice of the dataset.
18 Due to Abel Noser’s high level of coverage, several prior studies have used these data to investigate institutional trading behavior. G. Hu et al. (2018)
summarize 55 publications that use these data.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 18
NGUYEN ET AL.
TA B L E 7
Trading strategy profitability—S12 and 13F samples
Panel A: Mutual funds’ trading strategy profitability
Holding periods
(1) Dynamic
1 month
3 months
6 months
9 months
12 months
0.1754***
0.1421***
0.1242***
0.0791***
0.0608***
(0.0003)
(0.0015)
(0.0009)
(0.001)
(0.001)
(2) Naïve
0.1398***
0.1114***
0.0974***
0.0766***
0.059***
(0.0099)
(0.0034)
(0.001)
(0.001)
(0.001)
(3) EJR-based
0.1416***
0.1158***
0.1004***
0.0803***
0.0633***
(0.0003)
(0.001)
(0.0003)
(0.0003)
(0.0003)
(4) Issuer-paid CRA-based
0.125***
0.0986***
0.086***
0.0322***
0.0384***
(0.0114)
(0.0025)
(0.0018)
(0.0015)
(0.0023)
(5) S&P 500 index
0.033
0.033
0.033
0.033
(1)–(2)
(1)—(3)
(1)–(4)
(1)–(5)
(2)–(5)
(3)–(5)
(4)–(5)
(0.0453)
(0.0453)
(0.0453)
(0.0453)
0.0356***
0.0307***
0.0268***
0.0025**
(0.0099)
(0.0037)
(0.0011)
0.0338***
0.0263***
0.0238***
(0.0004)
(0.0018)
(0.0007)
(0.0011)
−0.0011
(0.0007)
0.033
(0.0453)
0.0018
(0.0011)
−0.0025
(0.0111)
0.0504***
0.0436***
0.0382***
0.0469***
0.0224***
(0.0114)
(0.0025)
(0.0018)
(0.0016)
(0.0023)
0.1424***
0.1091***
0.0912***
0.0461***
0.0278***
(0.0045)
(0.0041)
(0.0025)
(0.0025)
(0.0025)
0.1068***
0.0784***
0.0643***
0.0436***
0.026***
(0.0109)
(0.0051)
(0.0015)
(0.0015)
(0.0015)
0.1085***
0.0828***
0.0674***
0.0473***
0.0303***
(0.0045)
(0.0039)
(0.0016)
0.092***
0.0656***
0.053***
(0.0172)
(0.0042)
(0.0034)
(0.0015)
−0.0008
(0.0034)
(0.0023)
0.0054
(0.0034)
Panel B: Institutional investors’ trading strategy profitability
Holding periods
(1) Dynamic
1 month
3 months
6 months
9 months
12 months
0.1673***
0.1455***
0.1091***
0.0728***
0.0618***
(0.0139)
(0.0015)
(0.0007)
(0.0004)
(0.0004)
(2) Naïve
0.1175***
0.1143***
0.0898***
0.0649***
0.0535***
(0.0178)
(0.0046)
(0.0014)
(0.0014)
(0.0014)
(3) EJR-based
0.1236***
0.1191***
0.0953***
0.0727***
0.062***
(0.0093)
(0.0009)
(0.0003)
(0.0002)
(0.0003)
(4) Issuer-paid CRA-based
0.1202***
0.0804***
0.0678***
0.0483***
0.0384***
(0.0198)
(0.0062)
(0.0045)
(0.0038)
(0.0057)
(5) S&P 500 index
0.033
0.033
0.033
0.033
0.033
(0.0453)
(0.0453)
(0.0453)
(0.0453)
(0.0453)
(1)–(2)
0.0498**
0.0312***
0.0194***
0.0078***
0.0084***
(0.0226)
(0.0048)
(0.0014)
(0.0014)
(0.0014)
(Continues)
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
TA B L E 7
(Continued)
19
Panel B: Institutional investors’ trading strategy profitability
Holding periods
1 month
3 months
6 months
9 months
12 months
(1)–(3)
(1)–(4)
(1)–(5)
(2)–(5)
(3)–(5)
(4)–(5)
0.0438***
0.0264***
0.0138***
0.0001
(0.0168)
(0.0017)
(0.0005)
(0.0004)
−0.0001
(0.0005)
0.0471**
0.0651***
0.0414***
0.0245***
0.0235***
(0.0204)
(0.0063)
(0.0045)
(0.0038)
(0.0057)
0.1343***
0.1125***
0.0761***
0.0397***
0.0288***
(0.0147)
(0.004)
(0.0024)
(0.0023)
(0.0023)
0.0845***
0.0813***
0.0567***
0.0319***
0.0205***
(0.0183)
(0.0059)
(0.0018)
(0.0018)
(0.0018)
0.0905***
0.0861***
0.0623***
0.0397***
0.0289***
(0.0103)
(0.0039)
(0.0015)
(0.0015)
0.0872***
0.0474***
0.0347***
0.0153**
(0.0298)
(0.0095)
(0.0077)
(0.0077)
(0.0023)
0.0054
(0.0077)
Note: This table reports and compares annualized risk-adjusted returns on four institutional trading strategies. “Dynamic” is a
strategy that sells a stock when it receives an EJR’s negative rating adjustment and buys the stock when its rating is upgraded
by an issuer-paid CRA. The “naïve” strategy is simply to sell (buy) a stock following a negative (positive) signal from any rating
agency. For the “EJR-based” strategy, an investor sells (buys) a stock when EJR announces a downgrade (upgrade) in the stock’s
credit rating. The “issuer-paid CRA-based” strategy involves selling (buying) a stock following a rating downgrade (upgrade)
from an issuer-paid CRA. A buy-and-hold of the S&P 500 index is included as a benchmark strategy. We measure the prof-
itability for each trading strategy as follows. First, based on a firm’s aggregate rating change in a quarter and an institutional
investor’s net buy of the firm’s stock, we classify it to a specific strategy. We multiply the stock’s returns by (−1) if the quar-
terly rating change is negative and the investor exhibits a net sale of the stock to reflect stock returns to sales following a rating
downgrade. Next, for each institutional investor, we calculate portfolio returns for each strategy using the absolute values of
stock net buy as weights. These portfolio returns are computed for each day over an investment horizon starting at the end
of the quarter. We repeat this process for each quarter in our sample period. Finally, the risk-adjusted return for each strat-
egy is obtained using a pooled regression of the Fama–French five-factor model. We use a two-sample t-test to examine the
difference in mean risk-adjusted returns between strategies. Robust standard errors are in parentheses.
***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.
3.4.1
Institutional trading response to credit rating announcements
We investigate abnormal institutional trading surrounding a stock’s credit rating adjustments in the time window [0,
5].19 Day 0 is the date of a credit rating event. We consider institutions’ trading activities up to five days after the
credit rating adjustment to account for investors’ potential gradual reactions while also avoiding confounding effects
that can appear in longer windows.
With detailed transaction data, we calculate institutional investor i’s abnormal net buy of stock n over the [0, 5] day
window around a credit rating announcement, AN_NBi,n,w, as follows:
+5∑
AN_NBi,n,w =
k = 0
(AN_nbi,n,t+k − average ANnbi,n,t),
(6)
19 We also consider two different time windows [−2, 5] and [−2, 1] for robustness. The purpose is to account for institutional investors’ pre-reactions because
of potential information leakage (e.g., Bhattacharya et al., 2019). Our results are robust.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 20
NGUYEN ET AL.
where AN_nbi,n,t is the daily dollar volume bought minus daily dollar volume sold scaled by the stock’s 1-month-lagged
market capitalization as shown in equation (7). average AN_nbi,n,t is the average value of AN_nbi,n,t in the period from
day t − 371 to day t − 6 prior to the announcement date t as shown in equation (8).
AN_nbi,n,t =
BOUGHTi,n,t − SOLDi,n,t
MARKET_CAPi,n,m−1
,
average AN_nbi,n,t =
∑−371
k = −6 AN_nbi,n,t+k
365
.
We then estimate the following model, which is similar to equation (4), for each of the paired samples:
AN_NBi,n,w = 𝛼 + 𝛽1NEGn,t + 𝛽2POSn,t + 𝛽3NEGn,t∗EJRn,t + 𝛽4POSn,t∗EJRn,t
+ 𝛽5EJRn,t +
𝛾kCONTROLSn,t +
t∑
1
𝜃tQuarterFEt
.
+
i∑
1
𝛿iInvestorFEi +
𝜑nFirmFEn + 𝜀i,n,w
k∑
1
n∑
1
(7)
(8)
(9)
Depending on the numeric change in the CCR scale, ΔCCRn,t, for firm n on the adjustment date t, we define NEGn,t as
|ΔCCRn,t| if ΔCCRn,t < 0 and zero if ΔCCRn,t > 0,and POSn,t as ΔCCRn,t if ΔCCRn,t > 0,and zero if ΔCCRn,t < 0. Therefore,
an increase in NEGn,t (POSn,t) represents an absolute increase in credit rating downgrade (upgrade) for firm n at credit
event t. Control variables and fixed effects are described in equation (4).
We present the results in Table 8. The results are qualitatively similar to those based on S12 and 13F data—
institutional investors react asymmetrically to credit rating announcements made by EJR and issuer-paid CRAs.20
Columns 1 and 2 show the results for firms jointly rated by EJR and S&P. Institutional investors’ net buy increases
significantly around S&P’s positive rating adjustments. The POS coefficient is positive and significant across all speci-
fications. The 0.1655 basis point coefficient in column 2 is equivalent to an average increase of $316,914 in abnormal
net buy over [0, 5] days around the S&P’s one-notch rating upgrade announcements. However, the insignificant F-test
results for the overall impact of rating upgrades by EJR, that is, the sum of POS and EJR*POS coefficients, indicate that
institutional investors are unresponsive to EJR’s positive rating changes.
We document opposite results for rating downgrades. The EJR*NEG coefficient is negative and statistically and eco-
nomically significant across all models. For example, the −0.1191 coefficient of EJR*NEG in column 2 shows that a
one-notch downgrade announcement by EJR is equivalent to a decrease of $228,063 in abnormal institutional net
buy over the [0, 5] day window compared to a similar announcement by S&P. The F-test for the overall impact of
EJR downgrades, that is, the sum of NEG and EJR*NEG coefficients, indicates that the effect is strong statistically and
economically.
We find similarly asymmetric responses for firms jointly rated by S&P and Moody’s in columns 3 and 4. For exam-
ple, the POS coefficient of 0.4078 in column 4 indicates that abnormal institutional net buy, on average, increases by
$281,586 over the [0, 5] day window surrounding a credit rating upgrade by Moody’s. The F-test results for the sum
of NEG and EJR*NEG coefficients in column 4 indicate that EJR’s downgrades, on average, are associated with a signif-
icant decrease of $111,378 in abnormal institutional net buy over the [0, 5] day window. The results in columns 5 and
6 do not exhibit any robust and significant difference in the response of institutional investors around credit rating
changes for firms covered by both EJR and Fitch. All coefficients of interest are statistically insignificant. We further
investigate investor reactions to Fitch ratings in Section 3.5.
20 In order to assess the potential impact of changes in the sample period, we perform the analysis of S12 and 13F data on two sub-periods: 1999–2011 (to
match Abel Noser data coverage) and 2012–2017. The results are presented in Tables A11 and A12 in the Online Appendix and are qualitatively similar across
sub-periods.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
21
TA B L E 8 Abnormal trading responses to credit rating adjustments—Abel Noser sample
EJR versus S&P
EJR versus Moody’s
EJR versus Fitch
(1)
(2)
(3)
(4)
(5)
(6)
Intercept
0.3095
−0.4884
−0.3379
−0.4840
0.8414
−0.4015
NEG
POS
(1.8238)
(2.1424)
(4.9528)
(4.7255)
(1.7814)
(2.0034)
0.0470*
0.0538
0.0966
0.0925
0.0285
0.0259
(0.0266)
(0.0408)
(0.1154)
(0.1302)
(0.0233)
(0.0448)
0.2450***
0.1655***
0.3324**
0.4078***
0.074**
0.0818
(0.0300)
(0.0498)
(0.1319)
(0.1387)
(0.0353)
(0.0711)
EJR×NEG
−0.1014***
−0.1191**
−0.1984*
−0.2538*
−0.0295
−0.0636
EJR×POS
−0.2125***
−0.1232**
−0.2723**
−0.4453***
0.0023
−0.0436
(0.0322)
(0.0478)
(0.1204)
(0.1342)
(0.0349)
(0.0568)
(0.0360)
(0.0563)
(0.1387)
(0.1481)
(0.0443)
(0.0807)
EJR
0.1316***
0.1254**
0.2429**
0.4005***
0.0507
0.0744
(0.0365)
(0.0493)
(0.1234)
(0.1282)
(0.0491)
(0.0738)
Control variables:
No
Yes
No
Yes
No
Yes
F-tests:
NEG + EJR×NEG
−0.0544***
−0.0654**
−0.1018**
−0.1613***
−0.0011
−0.0377
POS + EJR×POS
0.0325
0.0423
0.0601
−0.0375
0.0763***
0.0381
(0.0209)
(0.0273)
(0.0468)
(0.0487)
(0.0279)
(0.0369)
(0.0216)
(0.0272)
(0.0496)
(0.0523)
(0.0285)
(0.0393)
Fixed effects:
Investor FE
Firm FE
Quarter FE
N
Adj. R2
Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
No
Yes
429,268
304,731
114,354
93,867
207,587
133,788
0.010
0.004
0.016
0.010
0.006
0.008
Note: The table reports OLS regression results for institutional investors’ abnormal trading around credit rating adjustments
announced by EJR and issuer-paid CRAs. The dependent variable, defined in equation (6), is an institutional investor’s abnor-
mal net buy of a stock over the [0, 5] day window. We define NEG as the absolute value of a rating downgrade and zero
otherwise and POS as the value of a rating upgrade and zero otherwise. Therefore, an increase in NEG (POS) represents an
absolute increase in a firm’s downgrade (upgrade). EJR is a dummy variable that equals one for EJR’s credit rating announce-
ments and zero otherwise. Detail descriptions of firm-level control variables are described in Appendix B. Standard errors in
parentheses are adjusted for heteroskadisticity and clustering at the firm and quarter levels.
***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.
3.4.2
Institutional trading profits in response to credit rating announcements
We now turn to assess institutional trading strategy profitability based on Abel Noser data. First, we use the
Fama–French five-factor model, as shown in equation (5), to estimate risk-adjusted returns, that is, alphas, for each
day in the [0, 5] window around a credit rating announcement. We multiply a daily alpha by (−1) if an institutional
investor’s trades on that day represent a net sale. We calculate the average alpha over the assessment window using
the institutional investor’s absolute daily net trade values as weights. We then assign this firm-institution-event alpha
to different trading strategies based on the institutional net buy over the assessment window and the credit rating
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 22
NGUYEN ET AL.
signal. Finally, we assess the performance of these strategies using firms’ market capitalization as weights in one and
two-sample t-tests.
Strategy trading profits and their significance are presented in Table 9. The results are consistent with those based
on S12 and 13F data. The dynamic strategy that follows EJR’s negative signals and other CRAs’ positive signals earns
the highest returns, compared to the other strategies. For example, for the 1-month investment horizon, the dynamic
strategy outperforms the other three strategies by an annualized value-weighted risk-adjusted return ranging from
10.17% to 10.95%. This outperformance is approximately twice as much as the corresponding outperformance of
notional strategies. Although this outperformance decreases with the investment horizon, it is still statistically signif-
icant for up to 9 months. Among the other three active strategies, following EJR’s signals alone apparently generates
the best returns, whereas following issuer-paid CRAs’ signals only yields the least profits. We also compare the four
trading strategies to a passive strategy—a buy-and-hold annual return of the S&P 500 index. We observe that all
trading strategies outperform the index for up to 6 months, except the strategy following issuer-paid CRAs’ rating
announcements.
3.5
The case of Fitch ratings
We note that investor reactions around credit rating changes for firms jointly rated by EJR and Fitch are different from
those covered by S&P and Moody’s. In the S12 and 13F samples, we observe significant reactions to Fitch upgrades and
no significant reactions to EJR downgrades. We document essentially no significant reactions in Able Noser data. This
has prompted us to investigate this further.21 Fitch has traditionally held a smaller market share relative to Moody’s
and S&P (Becker & Milbourn, 2011; Livingston & Zhou, 2016). This may have influenced both their rating behavior
(Beatty et al., 2019; Hirth, 2014) and investor reaction.
Our empirical analysis suggests that Fitch differs in rating behavior from other issuer-paid CRAs. First, as reported
in Table 5, not only does Fitch lead EJR in positive events (as expected) but is also the only issuer-paid CRA to lead
EJR in negative announcements (although the difference is not statistically significant). Furthermore, our unreported
analysis also shows that Fitch leads S&P and Moody’s in both positive and negative announcements. We believe this is
consistent with Fitch providing more timely rating announcements in order to increase their market share.
Second, we look at the information content of Fitch announcements by constructing two additional trading
strategies: the “Fitch-based strategy”—buying on Fitch upgrades and selling on Fitch downgrades (for the sake
of completeness, we also create “S&P-based strategy” and “Moody’s-based strategy”), and the “modified dynamic
strategy”—buying on credit upgrades by the “Big Three” and selling on Fitch downgrades. Our unreported results show
that the Fitch-based strategy not only outperforms a simple buy-and-hold of the S&P 500 index but also produces bet-
ter returns than the issuer-paid CRA-based strategy, particularly over longer time periods. This suggests that Fitch’s
announcements actually have higher information content than other issuer-paid CRAs. The modified dynamic strategy
is the second-best performing strategy, suggesting that Fitch’s negative announcements have substantial information
content. However, the “dynamic strategy”—buying on positive issuer-paid CRA announcements and selling on EJR’s
negative announcements—yields the best returns, which is consistent with our main hypothesis.
Finally, we investigate institutional investors’ reactions to Fitch’s announcements in greater detail. In the main anal-
ysis, institutions do not appear to react significantly to either positive or negative announcements in the sample of
firms jointly rated by Fitch and EJR, despite evidence that both CRAs’ announcements have significant information
content. We posit that as Fitch leads EJR in negative signals (although insignificantly), the lack of significant reaction
to EJR’s negative announcements may be due to the dilution of investors’ reaction to both Fitch and EJR’s announce-
ments. Investors do not react to Fitch’s announcements in a significant way (even though these announcements have
significant informational content), and this still weakens investors’ reactions to subsequent announcements by EJR.
21 We thank an anonymous referee for this suggestion.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
23
TA B L E 9
Trading strategy profitability—Abel Noser sample
1 month
3 months
6 months
9 months
12 months
(3) EJR-based
0.0609***
0.047***
0.0476***
0.0129
Holding periods
(1) Dynamic
(2) Naïve
(4) Issuer-paid CRA-based
(5) S&P 500 index
(1)–(2)
(1)–(3)
(1)–(4)
(1)–(5)
(2)–(5)
(3)–(5)
(4)–(5)
0.1626***
0.1098***
0.0728***
0.0714**
(0.0223)
0.0534*
(0.0317)
(0.0188)
0.0468*
(0.0267)
(0.0224)
0.0188
(0.0129)
(0.033)
0.0078
(0.0227)
(0.0113)
0.0531*
(0.0297)
−0.0163
(0.0621)
(0.0127)
0.0395
(0.0254)
−0.0163
(0.0621)
0.1092***
0.063*
(0.0388)
(0.0327)
(0.0114)
−0.0213
(0.026)
−0.0163
(0.0621)
0.054***
(0.0171)
0.1017***
0.0628***
0.0253
(0.025)
(0.0227)
(0.0188)
(0.0217)
−0.0012
(0.0265)
−0.0163
(0.0621)
0.0635**
(0.028)
0.0584*
(0.0309)
0.1095***
0.0703***
0.0941***
0.0726**
(0.0306)
(0.0262)
(0.0275)
(0.0287)
0.0297
(0.0496)
0.0008
(0.0451)
0.0049
(0.0909)
0.0184
(0.0463)
−0.0163
(0.0621)
0.0289
(0.0515)
0.0248
(0.1036)
0.0113
(0.0525)
0.1789***
0.1261***
0.0891***
0.0877***
0.046
(0.0232)
(0.0195)
(0.0226)
(0.0332)
0.0697**
0.0631**
0.0351***
0.0241
(0.0323)
(0.0272)
(0.013)
(0.0227)
0.0772***
0.0633***
0.0639***
0.0292
(0.0129)
0.0694*
(0.0376)
(0.0137)
0.0558
(0.0382)
(0.0116)
−0.005
(0.0434)
(0.0218)
0.0151
(0.053)
(0.0497)
0.0171
(0.0451)
0.0212
(0.091)
0.0347
(0.0617)
Note: This table reports and compares annualized risk-adjusted returns on four institutional trading strategies. “Dynamic” is a
strategy that sells a stock when it receives an EJR’s negative rating adjustment and buys the stock when its rating is upgraded
by an issuer-paid CRA. The “naïve” strategy is simply to sell (buy) a stock following a negative (positive) signal from any rating
agency. For the “EJR-based” strategy, an investor sells (buys) a stock when EJR announces a downgrade (upgrade) in the stock’s
credit rating. The “issuer-paid CRA-based” strategy involves selling (buying) a stock following a rating downgrade (upgrade)
from an issuer-paid CRA. A buy-and-hold of the S&P 500 index is included as a benchmark strategy. We measure the prof-
itability for each trading strategy as follows. For each day in the time window [0, 5] surrounding each rating announcement on
a firm, risk-adjusted return is the intercept (or alpha) from the Fama–French five-factor model estimated over a holding period.
We multiply a daily alpha by (−1) if an institutional investor’s trades on that day represent a net sale. Next, we calculate the
average alpha over the assessment window using the institutional investor’s absolute daily net trade values as weights. We
then assign a stock’s event alpha to different trading strategies based on its institutional net buy over the assessment window
and the credit rating signal. Finally, we use firms’ inflation-adjusted market capitalization as weights and assess the strate-
gies’ performance using one and two sample t-tests. Standard errors of the t-test for the mean and difference in means are in
parentheses.
***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 24
NGUYEN ET AL.
To investigate this, we remove negative announcements led by Fitch. Our unreported results are consistent with our
expectations, that is, investors’ reaction to EJR’s negative announcements becomes negative and significant in three
out of four specifications, which is consistent with the results reported in panels A and B of Table 4 for EJR and S&P
and EJR and Moody’s.
4
ROBUSTNESS TESTS
4.1
Hedge trading strategies
Although our findings on trading strategy profits suggest that investors generally underreact to the information con-
tent of credit rating signals, some investors may, in fact, overreact to rating events, which may result in subsequent
profits for contrarian trading strategies (Ellul et al., 2011). Alternatively, CRAs could provide upward-biased credit
ratings to some relationship firms (e.g., Baghai & Becker, 2018), and some institutional investors may be sophisticated
enough to detect these overrated firms and respond in the opposite way around rating announcements, which sub-
sequently earns them abnormal profits. We account for the effect of this potential contrarian trading strategy on the
performance of our main strategies as follows. We follow the steps in Section 3.3.2 to construct portfolios that are
opposite to our main strategies. For example, an institutional investor that increases its net holding of a stock with an
aggregate downgrade by EJR and decreases its net holding of a stock with an aggregate upgrade by issuer-paid CRAs
is considered to be a dynamic contrarian strategy. A dynamic hedge portfolio is then defined as longing the dynamic
portfolio and shorting the dynamic contrarian portfolio. Finally, we estimate the four hedge portfolios’ risk-adjusted
returns using the Fama–French five-factor model and report the results in Table A1. The dynamic hedge portfolio
remains the best performer, and its outperformance relative to the other hedge portfolios is qualitatively similar in
magnitude and statistical significance to the results in Table 7.
4.2
Alternative assumptions on the timing of trades
In the main analysis of quarterly S12 and 13F data, we assume that credit rating adjustments and fund-stock holding
changes happen on the final day of each quarter. In this robustness check, we make an alternative assumption that
these changes occur on the first day of each quarter. Trading strategy profitability is then re-estimated, and the results
are presented in Table A2 in the Online Appendix. The results are consistent with those in the main analysis that the
dynamic strategy significantly outperforms all other strategies considered.
We also consider alternative event windows in the analysis of daily Abel Noser data: [−2, 1] and [−2, 5] trading days.
First, these time windows include the two days prior to credit rating adjustments to control for potential information
leakage before official rating adjustments (Bhattacharya et al., 2019). Second, we also choose short time windows to
control for any effect of clusters of rating signals (e.g., Alsakka & ap Gwilym, 2012; Gande & Parsley, 2005; Vu et al.,
2015). In other words, shorter time windows enable us to avoid any information contamination problems caused by
the appearance of other information in the financial market in longer time windows. The results for the two alter-
native event windows are presented in Table A3 in the Online Appendix and are consistent with the main findings.
Institutional investors still exhibit asymmetric trading behavior to the issuer- and investor-paid credit rating signals,
abnormally buying on issuer-paid CRAs’ positive rating adjustments and abnormally selling on EJR’s negative rating
adjustments in both alternative time windows.
All four active trading strategies earn significant profits in similar patterns as in Table 4. They outperform the buy-
and-hold return of the S&P 500 index for up to a 9-month horizon. Most importantly, the dynamic trading strategy
is the best performer over all other strategies. The robust results of institutional trading strategies constructed sur-
rounding alternative event windows of [−2, 1] and [−2, 5] days are reported in Table A4 in the Online Appendix. We
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
25
also report the results for notional trading strategies for these alternative windows, and the results exhibit similar
patterns as shown in Table A5 in the Online Appendix.
4.3
Raw institutional trading
Our next robustness check analyzes “raw” reactions (i.e., unadjusted for the average of past trading activities) of insti-
tutional investors to credit rating announcements. We perform the analysis on both S12 and 13F quarterly data (Table
A6) and in the [0, 5] day window on Abel Noser daily data (Table A7). The results are highly consistent with the main
findings that institutional investors tend to abnormally sell stocks of firms with EJR’s negative rating announcements
but ignore positive ones; however, their net buy increases substantially surrounding issuer-paid CRAs’ positive rating
announcements.
4.4
Combined issuer-paid CRA
In another robustness check, we treat all three issuer-paid CRAs as a combined issuer-paid CRA. We then investigate
institutional investor’s trading activities surrounding negative and positive rating signals by EJR and the combined
issuer-paid CRA. The results are reported for the S12, 13F and Abel Noser Sample in Table A8 in the Online Appendix.
The results are consistent: Institutional investors tend to abnormally sell stocks surrounding negative signals issued
by EJR and abnormally buy stocks surrounding positive signals issued by the combined issuer-paid CRA.
4.5
Excluding non-trading observations
In our main analysis, abnormal net buy is set at zero if institutional investors have no trading activities surrounding
credit rating adjustments. In this final robustness check, we exclude these non-trading observations. We find robust
results in Table A9 for S12 and 13F data and Table A10 for Abel Noser data in the Online Appendix. The results are
robust. After excluding non-trading observations, institutional investors still have asymmetric responses, abnormally
increasing (decreasing) stock holdings surrounding positive (negative) rating signals by issuer- (investor-) paid CRAs.
Overall, these robustness tests confirm our main findings that institutional investors who have advanced trading skills
selectively react to credit rating signals from different sources based on their relative informational values.
5
CONCLUSION
This study investigates institutional investors’ responses to credit rating adjustments announced by the investor-paid
EJR and the “Big Three” issuer-paid CRAs. In recent years, traditional issuer-paid CRAs have faced criticism regard-
ing lack of timeliness in negative signals in many infamous scandals such as Enron (2001), WorldCom (2002) and
Lehman Brothers (2008). Meanwhile, investor-paid CRAs, particularly EJR, have built a good reputation regarding
the timeliness of their negative rating adjustments. As a result, institutional investors with advanced trading skills and
sophistication (Puckett & Yan, 2011) are likely to dynamically switch between following investor- and issuer-paid CRAs
based on the timeliness of credit rating information.
We document considerable asymmetries in institutional investors’ responses to issuer- and investor-paid CRA
announcements. They react by abnormally selling following EJR’s negative signals and abnormally buying following
issuer-paid CRAs’ positive signals. The results differentiate our paper from the existing literature. Several prior stud-
ies show that institutional investors simply tend to be more sensitive to negative rather than positive signals. Our study
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 26
NGUYEN ET AL.
finds that institutional investors, as professional players, have their own responses to the lack of timeliness criticism
by following investor-paid CRA’s negative signals. They still maintain faith in positive issuer-paid rating announce-
ments due to no evidence of their delays. The results are robust across different databases from which the institutional
investors’ trading activities are extracted.
We also document that a dynamic trading strategy based on selling following the investor-paid CRA’ negative sig-
nals and buying following issuer-paid CRAs’ positive signals produces superior returns. While any investor can take
advantage of these strategies, institutional investors evidently achieve higher returns. Although we document the
highly dynamic behavior of institutions in responding to important market signals, our results imply that market partic-
ipants tend to underreact to positive signals by issuer-paid CRAs and negative signals by investor-paid CRA. Therefore,
the information content of these signals is not fully reflected in prices at the announcement time, thus leading to
opportunities to earn abnormal returns by trading following these signals. As further information in support of CRA
creditworthiness predictions is released, abnormal returns are generally dissipated. Our results are consistent with
this view—abnormal returns decrease as holding periods increase. The difference between dynamic strategy and naïve
strategy returns becomes substantially smaller in the 12-month holding period.
Given that discrepancies in credit rating quality between issuer-paid CRAs and investor-paid CRAs are not limited
only to the US bond market (e.g., X. Hu et al., 2019), we believe there are some interesting avenues for future research
such as whether institutional investors also respond asymmetrically to credit rating announcements by issuer-paid and
investor-paid CRAs in an international setting (e.g., China); if so, whether such asymmetric responses are conditional
on some firm or market level shocks.
ACKNOWLEDGMENTS
We would like to thank Professor Pope, the JBFA editor and an anonymous referee for their constructive and valu-
able comments. Our thanks are also to participants at the New Zealand Finance Colloquium (NZFC) for their useful
comments.
Open access publishing facilitated by Massey University, as part of the Wiley - Massey University agreement via
the Council of Australian University Librarians.
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from third parties. Restrictions apply to the availability
of these data, which were used under license for this study. Data are available from the authors with the permission of
third parties.
ORCID
Alexander Molchanov
https://orcid.org/0000-0003-0133-3811
REFERENCES
Acharya, V., & Pedersen, L. (2005). Asset pricing with liquidity risk. Journal of Financial Economics, 77, 375–410.
Alsakka, R., & ap Gwilym, O. (2012). Foreign exchange market reactions to sovereign credit news. Journal of International Money
and Finance, 31, 845–864.
Amihud, Y., & Goyenko, R. (2013). Mutual fund’s R2 as predictor of performance. Review of Financial Studies, 26(3), 667–694.
Baghai, R. P., & Becker, B. (2018). Non-rating revenue and conflicts of interest. Journal of Financial Economics, 127(1), 94–112.
Baker, H. K., & Mansi, S. A. (2002). Assessing credit rating agencies by bond issuers and institutional investors. Journal of
Business Finance and Accounting, 29(9), 1367–1398.
Beatty, A., Gillette, J., Petacchi, R., & Weber, J. (2019). Do rating agencies benefit from providing higher ratings? Evidence from
the consequences of municipal bond ratings recalibration. Journal of Accounting Research, 57(2), 323–354.
Beaver, W. H., Shakespeare, C., & Soliman, M. T. (2006). Differential properties in the ratings of certified versus non-certified
bond-rating agencies. Journal of Accounting and Economics, 43, 303–334.
Becker, B., & Milbourn, T. (2011). How did increased competition affect credit ratings? Journal of Financial Economics, 101(3),
493–514.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
27
Bernile, G., Sulaeman, J., & Wang, Q. (2015). Institutional trading during a wave of corporate scandals: “Perfect Payday”? Journal
of Corporate Finance, 34(C), 191–209.
Berwart, E., Guidolin, M., & Milidonis, A. (2019). An empirical analysis of changes in the relative timeliness of issuer-paid vs.
investor-paid ratings. Journal of Corporate Finance, 59, 88–118.
Bhattacharya, U., Wei, & Xia, H. (2019). Follow the money: Investor trading around investor-paid credit rating changes. Journal
of Corporate Finance, 58, 68–91.
Bolton, P., Freixas, X., & Shapiro, J. (2012). The credit ratings game. Journal of Finance, 67(1), 85–111.
Brogaard, J., Koski, J. L., & Siegel, A. F. (2019). Do upgrades matter? Evidence from trading volume. Journal of Financial Markets,
43, 54–77.
Bruno, V., Cornaggia, J., & Cornaggia, K. J. (2016). Does regulatory certification affect the information content of credit ratings?
Management Science, 62(6), 1578–1597.
Cantor, R., ap Gwilym, O., & Thomas, S. (2007). The use of credit ratings in investment management in the US and Europe.
Journal of Fixed Income, 17(2), 13–26.
Chemmanur, T. J., He, S., & Hu, G. (2009). The role of institutional investors in seasoned equity offerings. Journal of Financial
Economics, 94(3), 384–411.
Chemmanur, T. J., Li, Y., Xie, J., & Zhu, Y. (2016). Noisy signalling through open market share repurchase progrqams and infor-
mation production by institutions (Working Paper. 2018 Asian Finance Association (AsianFA) Conference, 2016 Financial
Intermediation Research Society (FIRS) conference, 2015 European Finance Association (EFA) conference). https://ssrn.
com/abstract=2543397
Chen, S. S., Chen, H. Y., Chang, C. C., & Yang, S. L. (2016). The relation between sovereign credit rating revisions and economic
growth. Journal of Banking and Finance, 64, 90–100.
Cornaggia, J., & Cornaggia, K. (2013). Estimating the costs of issuer-paid credit ratings. Review of Financial Studies, 26(9), 2229–
2269.
Cremers, K. J. M., & Petajisto, A. (2009). How active is your fund manager? A new measure that predicts performance. Review
of Financial Studies, 22(9), 3329–3365.
Dimitrov, V., Palia, D., & Tang, L. (2015). Impact of the Dodd-Frank act on credit ratings. Journal of Financial Economics, 115,
505–520.
Drago, D., & Gallo, R. (2016). The impact and the spillover effect of a sovereign rating announcement on the euro area CDS
market. Journal of International Money and Finance, 67, 264–286.
Ellul, A., Jotikasthira, C., & Lundblad, C. T. (2011). Regulatory pressure and fire sales in the corporate bond market. Journal of
Financial Economics, 101, 596–620.
Fama, E. F., & French, K. R. (2015). A five-factor asset pricing model. Journal of Financial Economics, 116, 1–22.
Gande, A., & Parsley, D. (2005). New spillovers in the sovereign debt market. Journal of Financial Economics, 75, 691–734.
Griffin, J. M., Nickerson, J., & Tang, D. Y. (2013). Rating shopping or catering? An examination of the response to competitive
pressure for CDO credit ratings. Review of Financial Studies, 26(9), 2270–2310.
He, G. (2021). Credit rating, post-earnings-announcement drift, and arbitrage from transient institutions. Journal of Business
Finance & Accounting, 48, 1434–1467.
He, J., Qian, J., & Strahan, P. E. (2012). Are all ratings created equal? The impact of issuer size on the pricing of mortgage-backed
securities. Journal of Finance, 67(6), 2097–2137.
Henry, D., Nguyen, L., & Pham, V. H. (2017). Institutional trading before dividend reduction announcements. Journal of Financial
Markets, 36, 40–55.
Heron, R. A., & Lie, E. (2009). What fraction of stock option grants to top executives have been backdated or manipulated.
Management Science, 55, 513–525.
Hirth, S. (2014). Credit rating dynamics and competition. Journal of Banking and Finance, 29, 100–112.
Hu, X., Huang, H., Pan, Z., & Shi, J. (2019). Information asymmetry and credit rating: A quasi-natural experiment from China.
Journal of Banking and Finance, 106, 132–152.
Hu, G., Jo, K. M., Wang, Y. A., & Xie, J. (2018). Institutional trading and Abel Noser data. Journal of Corporate Finance, 52, 143–
167.
Jagolinzer, A. D., Larcker, D. F., & Taylor, D. J. (2011). Corporate governance and the information content of insider trades.
Journal of Accounting Research, 49(5), 1249–1274.
Joe, D. Y., & Oh, F. D. (2018). Spillover effects within business groups: The case of Korean chaebols. Management Science, 64(3),
1396–1412.
Johnson, R. (2004). Rating agency actions around the investment-grade boundary. Journal of Fixed Income, 13, 25–38.
Kedia, S., Rajgopal, S., & Zhou, X. (2017). Large shareholders and credit ratings. Journal of Financial Economics, 124, 632–653.
Livingston, M., & Zhou, L. (2016). Information opacity and Fitch bond ratings. Journal of Financial Research, 39(4), 329–357.
Puckett, A., & Yan, X. (2011). The interim trading skills of institutional investors. Journal of Finance, 66(2), 601–633.
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 28
NGUYEN ET AL.
Ramsay, J. (2011). Testimony concerning oversight of the credit rating agencies Post-Dodd-Frank. https://www.sec.gov/news/
testimony/2011/ts072711jr.htm
Skreta, V., & Veldkamp, L. (2009). Ratings shopping and asset complexity: A theory of ratings inflation. Journal of Monetary
Economics, 56, 678–695.
Vu, H., Alsakka, R., & ap Gwilym, O. (2015). The credit signals that matter most for sovereign bond spreads with split rating.
Journal of International Money and Finance, 53, 174–191.
Xia, H. (2014). Can investor-paid credit rating agencies improve the information quality of issuer-paid rating agencies? Journal
of Financial Economics, 111, 450–468.
SUPPORTING INFORMATION
Additional supporting information can be found online in the Supporting Information section at the end of this article.
How to cite this article: Nguyen, Q. M. P., Do, H. X., Molchanov, A., Nhut, L., & Nguyen, N. H. (2023).
Asymmetric trading responses to credit rating announcements from issuer- versus investor-paid rating
agencies. Journal of Business Finance & Accounting, 1–29. https://doi.org/10.1111/jbfa.12686
APPENDIX
APPENDIX A: NUMERIC TRANSFORMATION OF ALPHANUMERICAL RATING CODES
Investment grade
Speculative grade
Credit eventsa
Rating
AAA (Aaa)
AA+ (Aa1)
AA (Aa2)
AA- (Aa3)
A+ (A1)
A (A2)
A- (A3)
BBB+ (Baa1)
BBB (Baa2)
BBB− (Baa3)
Score
22
21
20
19
18
17
16
15
14
13
Rating
BB+ (Ba1)
BB (Ba2)
BB− (Ba3)
B+ (B1)
B (B2)
B− (B3)
CCC+ (Caa1)
CCC (Caa2)
CCC− (Caa3)
CC (Ca)
C
SD, D
Score
12
11
10
9
8
7
6
5
4
3
2
1
Single upgrade
Positive outlook
Positive developing
Stable
Negative developing
Negative outlook
Single downgrade
Score
1
0.5
0.25
0
−0.25
−0.5
−1
aSingle upgrade (downgrade) is a credit rating announcement when a rating agency adjusts the firm’s credit rating by one letter
rating higher (lower; e.g., up from AA+ to AAA or down from AA+ to AA). A positive (negative) outlook is a credit rating review
when a CRA adjusts its short-term expectations about the firm from being stable to positive (negative). A positive (negative)
developing is a credit rating signal when a CRA adjusts its long-term expectations about the firm from being stable to positive
(negative).
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License NGUYEN ET AL.
29
APPENDIX B: FIRM-LEVEL VARIABLE DEFINITIONS AND DATA SOURCES
Variable
Ln(MV)
ROA
Description
Data source
The natural log of total market capitalization in the quarter
CRSP
The ratio of operating income before depreciation to total
COMPUSTAT
assets in the quarter
IDIO_RISK
The standard deviation of residual returns from the
Kenneth R. French &
Fama–French three-factor model using daily stock returns
from day t − 31 to day t − 1
CRSP
Z-SCORE
Alman’s Z-score that presents the probability that a firm will
COMPUSTAT
go into bankruptcy within 2 years
ANALYST_COVERAGE
The average number of analysts covering a firm in the quarter
CRSP
Ln(AGE)
The natural log of number of years since a firm’s first
CRSP
appearance on CRSP database
INTEREST_COVERAGE
The ratio of earnings before interest, tax and depreciation and
COMPUSTAT
amortization to total interest expense in the quarter
LEVERAGE
The ratio of sum of long-term debt and debt in current
COMPUSTAT
liabilities to total assets in the quarter
S&P_500
A binary variable that equals one if a firm is included in the S&P
S&P 500 Index
500 list
HIGH_TECH
A binary variable that equals one if a firm’s Standard Industry
Classification (SIC) code is between 7370 and 7379 (Heron
and Lie, 2009) and zero otherwise
CRSP
14685957, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbfa.12686 by Test, Wiley Online Library on [15/06/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
| null |
10.1371_journal.pcbi.1007774.pdf
| null |
All relevant data are available at the github repository: https://github. com/mdkarcher/BEAST-XML .
|
PDF file not found
|
|
10.1088_1361-6382_ad0749.pdf
|
Data availability statement
All data that support the findings of this study are included within the article (and any supple-
mentary files).
|
Data availability statement All data that support the findings of this study are included within the article (and any supplementary files).
|
Class. Quantum Grav. 40 (2023) 235011 (14pp)
https://doi.org/10.1088/1361-6382/ad0749
Classical and Quantum Gravity
Reconciling absence of vDVZ discontinuity
with absence of ghosts in nonlocal
linearized gravity
D Dalmazi
UNESP—Campus de Guaratinguetá—DFI, CEP, 12516-410 Guaratinguetá, SP,
Brazil
E-mail: [email protected]
Received 12 July 2023; revised 18 October 2023
Accepted for publication 26 October 2023
Published 6 November 2023
Abstract
The modern massive gravity theories resolve a historical tension between the
absence of the so called vDVZ mass discontinuity and the absence of ghosts
via a fine tuned gravitational potential and a sophisticated screening mech-
anism. Those theories have originated the modern covariant bimetric models
which are local, ghost free and cosmologically viable apparently, they contain
a massive plus a massless graviton in the spectrum. It seems hard to solve the
mentioned tension if we do insist in a model with a minimal number of degrees
of freedom, with only one massive spin-2 particle in the spectrum, even if we
allow nonlocal theories. Here we show that this problem can be circumvented
in linearized nonlocal theories by the introduction of exponential terms with
infinite derivatives. The model admits non linear completions via nonlocal
quadratic terms in curvatures. We also investigate the role of the exponential
factors in linearized models where the graviton remains massless and a mass
scale is introduced via nonlocal terms, they are also ghost free and approach
the Einstein–Hilbert theory as we go much above the introduced mass scale.
Keywords: ghosts, nonlocal gravity, massive gravity, vDVZ
1. Introduction
It has been known for several decades [1, 2] that the introduction of a graviton mass term in
the linearized Einstein–Hilbert (LEH) theory leads to incorrect predictions for solar system
−23ev, which is the
tests of general relativity (GR), even if we keep the graviton mass below 10
upper bound obtained from the detection of gravitational waves from black-hole and neutron
stars mergers [3]. In fact, the predictions of massive linearized gravity are irreconcilable with
experiments no matter how small is the graviton mass. Although the massive linearized action
→ 0, this is not true for the exchange amplitude.
continuously approach the LEH theory as mgr
1361-6382/23/235011+14$33.00 © 2023 IOP Publishing Ltd Printed in the UK
1
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
This is the known vDVZ mass discontinuity [1, 2]. It turns out that the graviton mass introduces
another scale in the gravitational theory, the Vainshtein radius [4], which breaks down the
linearized truncation and demands non linear terms which introduce, on their turn, a ghost [5]
in the theory in general.
Quite recently [6, 7], partially motivated by the possible role of a massive graviton in the
late time acceleration of the Universe [8, 9], one has been able to accommodate the absence
of mass discontinuity with the absence of ghosts by fine tuning the graviton potential. Due
to unstable Friedmann–Lemaître–Robertson–Walker (FLRW) solutions, the model of [6] has
been improved giving rise to the bimetric model of [10], see the review works [11–14]. In
the bimetric model the otherwise arbitrary reference metric of [6] acquires its own dynamics
without introducing ghosts. Now we have two dynamic metric tensors. Its particle content
corresponds to a massive plus a massless spin-2 particle. The bimetric model is a ghost free
local field theory, cosmologically viable apparently, see the recent work [15]. In this model
the accelerated expansion of the Universe is due to the interaction of the gravitons, there is
no need of dark energy. A natural extension of [10] corresponds to multi-gravity models with
more than one massive graviton, see [16–19] and references therein.
On the other hand, one might wonder whether there would be another massive gravity
model without ghosts and vDVZ discontinuity while keeping a minimal number of degrees
of freedom with no other particle than the massive graviton. Here we investigate the above
issue from the point of view of linearized (about flat space) nonlocal gravitational mod-
els. In the next section we argue that the incompatibility between absence of ghosts and
absence of vDVZ discontinuity persists even in nonlocal theories. In section 3 however we
show that the use of exponential factors with infinite derivatives may circumvent the prob-
lem. In section 4 we return to massless gravitons but with a mass scale introduced via non-
local terms. The exponential factors allow us to recover the LEH theory as we go above the
introduced mass scale. We also comment on the use of exponentials as a way of introducing
a screening factor in the nonlocal Deser–Woodard models [20]. In section 5 we draw our
conclusions.
2. Absence of ghosts versus mass continuity
Before we analyze nonlocal massive spin-2 theories it is convenient for our purposes to start
with the spin-1 case. The Proca model with a scalar Stueckelberg field is given by:
L
PS (Aµ, ϕ) = − 1
4
µν (A) − m2
F2
2
(
Aµ +
∂µϕ
m
)
2
(
−
Aµ +
)
∂µϕ
m
jµ
.
(1)
In order to avoid divergences in the massless limit at action level we work with conserved
external currents from the start ∂µjµ = 0. The Proca–Stueckelberg theory is invariant under
U(1) transformations: (δAµ; δϕ) = (∂µΛ; −mΛ).
In the case of the usual (non gauge) Proca theory L
PS(Aµ, 0) the equations
of motion lead to (□ − m2)Aµ = jµ and the constraint ∂µAµ = 0 which leads to
P = L
2
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
3 = 2s + 1 independent degrees of freedom (d.o.f.) in D = 4 as expected for a massive
spin s = 1 model. Consequently, we have the exchange amplitude in the Proca theory1
ˆ
ˆ
A
P (j, j′) =
dDx j′
µAµ =
dDx j′
µ
1
□ − m2 jµ
,
(2)
which continuously tend to the Maxwell amplitude as m → 0. Returning to the Proca-
Stueckelberg theory, since we are not interested in correlations functions of the pure gauge
scalar field and assuming ϕ-independent gauge conditions, we can Gaussian integrate over ϕ
in the path integral and derive a massive U(1) invariant nonlocal model that may be called a
nonlocal Proca (NLP) model, namely
L
NLP = − 1
4
Fµν
)
(
□ − m2
□
Fµν − Aµjµ
,
which leads to the equations of motion
)
(
□ − m2
□
∂µFµν = jν
,
(3)
(4)
dDx j′
which continuously flow to the Maxwell equations as m2/□ → 0. The equation (4) can be
made local in the Lorentz gauge ∂µAµ = 0. As in the non gauge local Proca theory, we have
(□ − m2)Aµ = jµ. Thus, the gauge invariant exchange amplitude of th NLP theory coincides
´
P(j, j′) given in (2). They share
with the usual Proca amplitude A
the same particle content too. If we add a gauge fixing term to L
NLP we obtain the propagator
which has only one pole at □ = m2 with positive residue (physical pole in our notation). Notice
that, differently from the Maxwell theory, although the Lorentz gauge is still invariant under
U(1) harmonic transformations, this is not a residual symmetry of the gauge fixed NLP model
because the equations of motion are massive now. Thus, we have D − 1 degrees of freedom as
in the local Proca theory.
NLP(j, j′) =
µAµ = A
Since the factor (□ − m2)/□ in (3) basically replaces a massless pole by a massive one
and the LEH model is the spin-2 analogue of the Maxwell theory, one might think that the
following nonlocal model is the natural s = 2 analogue of (3),
L
NL−LEH = hµν
)µναβ
(
∂2
LEH
)
(
□ − m2
□
hαβ + Tµνhµν
,
(5)
where the external source satisfies ∂µTµν = 0 while ∂2
appears in the linearization of the EH theory about flat space gµν = ηµν + hµν, i.e.
LEH is the differential operator which
formula (2)
is a short notation for AP( j, j ′) =
µ(x)G(x, y)jµ(y) where
1 More precisely,
(□ − m2)G(x, y) = δ(D)(x − y) and for causal reasons G(x, y) is the retarded Green function obtained in momentum
⃗k2 + m2 in semi-circles from
space (Fourier transform) integrating along the real axis surrounding both poles k0 = ±
above. Accordingly, we assume that AP( j, j ′) = 0 in the absence of any of the two sources, so we have neglected the
source independent solution of the Klein–Gordon equation (□ − m2)Aµ = 0 (homogeneous equation). Similarly for
the remaining (spin-2) formulae in the present work, and for m2 = 0.
dDy j ′
√
dDx
´
´
3
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
(√
L
LEH =
)
hh
≡ hµν
□
4
(
)µναβ
∂2
LEH
(∂µhµν)2
2
−gR
□
4
= hµν
hµν − h
h +
hαβ
− 1
2
∂µhµν∂νh
.
(6)
hαβ
,
(7)
(8)
(9)
The model (5) has been suggested from the degravitation perspective in [21, 22], see also
[23–25]. For later comparison it is convenient to decompose the nonlocal LEH model in terms
(s)
of spin-s projection operators P
2 which project rank-2 symmetric tensors into a subspace of
transverse symmetric tensors with 2s + 1 (in the case of D = 4) independent components,
(
∂2
LEH
hµν
)µναβ
(
)
□ − m2
□
hαβ =
(
1
4
hµν
□ − m2
) [
(2)
2
P
− (D − 2) P
(0)
2
]µναβ
where
)
)
(
(
(2)
2
P
(0)
2
P
µναβ
µναβ
=
=
(θµα θνβ + θµβ θνα) − θµνθαβ
1
D − 1
2
θµν = ηµν − ∂µ∂ν
θµνθαβ
□
D − 1
;
,
.
√
The nonlocal linear model (7) can be non linearly completed in terms of
−gR and nonlocal
quadratic terms in curvatures like R(m2/□2)R ; Rµν(m2/□2)Rµν and Cµναβ(m2/□2)Cµναβ
where Cµναβ(g) is the conformal (or Weyl) tensor. The nonlinear completion is not unique
even if we only look at terms which are at most quadratic in hµν, due to total derivatives. Here
we are mainly interested however, in problems which appear before the non linear completion.
Similar to the spin-1 case in the Lorentz gauge, in the de Donder gauge:
∂µhµν − ∂νh/2 = 0
(10)
the equations of motion δSNL−EH = 0 become local: (□ − m2)(hµν − ηµνh/2) = −2 Tµν.
Plugging back its own trace we have
(
)
.
(
)
T
D − 2
So the gauge invariant exchange amplitude is given by
Tµν − ηµν
hµν = −2
□ − m2
ˆ
ˆ
.
(
(11)
)
A
NL−LEH (T, T
′) =
dDx hµνT
′
µν = −2
dDx T
′
µν
1
□ − m2
Tµν − ηµν T
D − 2
,
(12)
which continuously tend to the corresponding massless model (LEH) amplitude as m → 0, thus
solving the vDVZ mass discontinuity problem present in the non gauge local Fierz–Pauli (FP)
[26] theory. In the FP case the exchange amplitude for conserved sources has the same form
of (12) but with D − 2 replaced by D − 1 , which is in D = 4 the famous vDVZ discontinuity
[1, 2]. So similarly to the NLP model (3), the model (7) has no discontinuity as m → 0.
However, there is an important difference with respect to the s = 1 case, a negative one. It is
the fact that the particle content of L
NL−LEH contains a massive spin-0 ghost, see [25], besides
the expected massive spin-2 particle with 5 = 2s + 1 degrees o freedom. Indeed, the symmetric
tensor hµν contains ten components in D = 4 while the gauge condition (10) subtracts only 4
components. As in the spin-1 case there is no residual gauge symmetry, although the gauge
condition (10) is invariant under linearized harmonic reparametrizations (δhµν = ∂µϵν + ∂νϵµ
with □ϵµ = 0) the massive equations of motion are not invariant. So we are left with 10 − 4 = 6
d.o.f.. Later it will become clear here that the sixth mode is a ghost, see also [25].
should try to repeat the same procedure that has led us to L
NL−LEH(hµν) is not a faithful s = 2 analogue of spin-1 NLP theory, perhaps we
NLP(Aµ). Namely, we introduce
Since L
4
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
Stueckelberg fields in the local FP model (s = 2 analogue of the local Proca theory) and integ-
rate over them. Following a Kaluza–Klein dimensional reduction, see e.g. [27], we can go
from the massless LEH theory in D + 1 down to the massive theory in D dimensions by com-
pactfying one spatial dimension into a closed circle. By restricting us to only one massive
mode we end up with a gauge invariant version of the FP model with a vector Aµ and a scalar
ϕ Stueckelberg field. It amounts to replace hµν → hµν + (∂µAν + pνAµ)/m + 2 ∂µ∂νϕ/m2 in
the FP model. Since the coupling to conserved sources is not affected, after some integrations
by parts we have the s = 2 analogue of (1):
L
FPS (h, A, ϕ) = L
LEH
(
− m2
4
)
h2
µν
− h2
+ 2 m Aµjµ − 2 ϕ ∂µjµ − 1
4
µν (A) + hµνTµν
F2
(13)
where the first two terms in (13) correspond to the usual FP model which describes massive
spin-2 particles without ghosts and jµ(h) ≡ ∂αhαµ − ∂µh. By construction (13) is invariant
under linearized reparametrizations and U(1) transformations,
δhµν = ∂µϵν + ∂νϵµ
;
δAµ = −m ϵµ + ∂µΛ ;
δϕ = −m Λ.
(14)
Notice that ϕ appears linearly in (13). Thus, its functional integral would lead to a vanish-
ing linearized scalar curvature ∂µjµ = ∂µ∂νhµν − □h = RL(h) = 0 and we would not be able
to further integrate over the vector field Aµ without fixing the gauge and explicitly breaking
the reparametrization invariance. So in order to derive a nonlocal gauge invariant action for
an unconstrained rank-2 symmetric tensor we follow a different route. We first make a trivial
(invertible) field redefinition with a so far arbitrary real constant k: hµν = Hµν + k ηµν ϕ, lead-
ing to
L
FPS (h, A, ϕ) = L
FPS (H, A, 0) + ϕ
K
2
ϕ + ϕ J
,
where
{
}
K = k (D − 1)
J = k T + k m2 (D − 1) H + [k (D − 2) − 2] ∂µjµ (H) + 2 k m (D − 1) ∂ · A.
k D m2 + [4 − k (D − 2)] □
(15)
(16)
(17)
The action corresponding to (15) is still invariant under (14) but now the redefined rank-2
tensor also transforms under U(1) transformation which looks like a conformal symmetry,
δHµν = ∂µϵν + ∂νϵµ + ηµν k m Λ .
(18)
The Gaussian integral over the scalar field produces a nonlocal gauge fixing type term propor-
tional to (∂ · A)2 even though we have not broken any gauge symmetry. We could make it local
with the choice k = 4/(D − 2), see (16), but we keep k an arbitrary (non vanishing) constant
leading, after integrating over ϕ, to the intermediate nonlocal theory
L
where
I (k) = L
FP (H) + HµνTµν + 2 m Aµjµ (H) − 1
2
− 1
2
[2 k m (D − 1) ∂ · A + R]
−1
K
)
(
F2
µν
[2 k m (D − 1) ∂ · A + R]
R = k T + k m2 (D − 1) H + [k (D − 2) − 2] ∂ · j (H)
.
5
(19)
(20)
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
Although (19) is still invariant under (18) altogether with δAµ = −m ϵµ + ∂µΛ, we are able
now to functionally integrate over Aµ without breaking any gauge symmetry. Remarkably, all
the k dependence disappears and we end up with a simple nonlocal result:
L
C (H, T) = Hµν
)
(
(
□ − m2
2
(2)
2
P
)µναβ
Hαβ + HµνT
µν
,
where we have an effective traceless conserved source :
Tµν = Tµν − θµν
D − 1
T .
(21)
(22)
(
)µναβ
(2)
2
P
defined in (8) and θµν given in (9). If we recall (inside spacetime integ-
with
rals) that −Fµν(□ − m2)Fµν/(4 □) = Aµ(□ − m2)θµνAν/2 and notice that θµν is the spin-1
(2)
analogue of P
2 , it becomes clear that the nonlocal model (21) is indeed the natural spin-2
version of the NLP model (3) rather than (7) which contains a lower spin projection operator.
The model (21) is invariant under linearized reparametrizations and Weyl (conformal) trans-
formations, the latter being inherited from the U(1) symmetry in (14). Redefining the U(1)
parameter we have:
δHµν = ∂µϵν + ∂νϵµ + ηµνΛ .
(23)
(
)µναβ
Regarding non linear completions, there are different ways of writing the nonlocal
(2)
2
P
Lagrangian Hµν□
Hαβ in terms of quadratic truncations of covariant (under gen-
eral coordinate transformations) theories. They differ, in flat space, by total derivatives.
Introducing three arbitrary real parameters (α, β, γ) we can write inside space-time integrals,
supposing now gµν = ηµν + Hµν,
{
[
]
Hµν
(
□ − m2
4
[
√
+
−g
)
(
)
P
(2)
2
µναβ
Hαβ =
(D − 2)
4 (D − 3)
(
β □ − γ m2
□2
)
µναβ
C
(
√
−g
)
−
D
4 (D − 1)
√
−g R
β □ − γ m2
□2
R + α
)
(
□ − m2
□2
Rµν
µν −
R
)
(
□ − m2
□2
R
(
β □ − γ m2
□2
D
R
4 (D − 1)
]
)
µν
R
]}
Cµναβ − Rµν
√
−g
[
R − Rµν
1
□ Rµν + R
1
2 □ R
HH
(24)
The right side of the first line of (24) is enough to reproduce the left side. In the second and third
lines we have total derivatives. The factor (β □ − γ m2)/□2 appears inside a Gauss–Bonnet
like term. The last α-term is also known [28] to be a total derivative in flat space.
The equations of motion from (21) determines the dynamics of a traceless and conserved
nonlocal extension of the linearized Einstein tensor (GL
(
)
)
(
µν) namely,
□ − m2
□
GL
µν
− θµν GL
D − 1
= Tµν − θµν
D − 1
T = Tµν
.
(25)
Regarding the number of degrees of freedom, due to (23), we can impose five gauge
conditions:
∂µHµν = 0
; H = 0
.
(26)
6
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
So we are left with 10 − 5 = 5 d.of. in D = 4 as expected for a massive spin-2 particle.
Substituting (26) in (25) we have a massive equation (□ − m2)Hµν = −2 Tµν. Notice that the
massive equations of motion and the gauge conditions have no residual gauge symmetry. The
gauge invariant exchange amplitude of the linearized conformal model (21) coincides with the
amplitude of the local FP theory,
ˆ
A
C =
ˆ
(
′
µν = −2
dDxHµνT
ˆ
(
= −2
dDxT
′
µν
1
□ − m2
Tµν − θµνT
D − 1
)
dDx
Tµν − ηµνT
D − 1
)
(
1
□ − m2
′
µν
T
′
− θµνT
D − 1
)
= A
FP ,
(27)
just like the exchange amplitude of the NLP theory reproduces the amplitude of the local (non
gauge) Proca model.
Considering the analytic structure of the propagator, after adding gauge fixing terms
(∂µHµν)2 + γ2
2 H2 we can obtain the propagator (suppressing indices)
γ1
2
−1
C =
G
4i
□ − m2 P
(2)
2 + · · ·
,
(28)
where the dots stand for nonlocal differential operators depending upon γ1 and γ2 which vanish
after saturation with conserved and traceless sources. After such saturation we have the two
point amplitude in momentum space:
2 (p) = − i
A
2
∗
µν (p)
T
[
−1
C (p)
G
]µναβ
Tαβ (p) = 2 i
T
∗
µν (p) Tµν (p)
p2 + m2
.
(29)
ij
∑
|Tij
Since pµTµν = 0, in the center of mass pµ = (m, 0, · · · , 0) we have T0µ = 0 and the imagin-
ary part of the residue of A
|2 which is definite positive,
characterizing a physical particle.
2(p) at p2 → −m2 is Rm = 2
In summary, the nonlocal model L
C is essentially equivalent to the usual (non gauge) FP
theory with all its positive and negative features. Namely, we have a ghost free massive spin-2
theory with mass discontinuity, compare (27) with (12). Although we have now a family of
possible covariant non linear completions (24) we must stress that the linearized Weyl sym-
metry will be broken by the self interacting vertices. The same problem occurs in the local
higher derivative topologically massive gravity model of [29, 30] in D = 3.
From the previous analysis of L
C there seems to be a persistent tension
between absence of mass discontinuity and absence of ghosts at linearized level even in non-
local theories. In the next section we suggest another linearized nonlocal gravitational model
which reconciles both features.
NL−LEH and L
3. Nonlocal massive gravity with exponential factors
Both models L
NL−LEH and L
C fit in the following reparametrization invariant form:
L
f (h, T) = hµν
)µναβ
(
∂2
LEH
)
(
□ − m2
□
hαβ − RL
f (□)
2 □2 RL + hµνTµν
,
(30)
where f(□) = 0 and f(□) = −(D − 2)(□ − m2)/[2(D − 1)] respectively. After adding a gauge
fixing term γ (∂µhµ − ∂µh/2)2 we can obtain the corresponding propagator, suppressing
space-time indices again,
7
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
−1
f =
G
4 i
□ − m2 P
(2)
2
−
4i
[(D − 2) (□ − m2) + 2 (D − 1) f (□)]
(0)
2 + · · ·
P
,
(31)
where dots stand for terms which vanish when saturated with conserved sources.
At f(□) = 0 we find a massive pole in the scalar sector with a wrong sign for the residue as
compared to the physical massive pole in the spin-2 sector thus, demonstrating the presence
of a massive scalar ghost (the 6th degree of freedom in D = 4 mentioned before), see detailed
analysis in [25]. The singularity at f(□) = −(D − 2)(□ − m2)/[2(D − 1)] points to the appear-
ance of the conformal (Weyl) symmetry of L
C whose gauge has not been fixed. In this case, as
we have seen before, after fixing all symmetries the propagator is non singular and the particle
content of L
C consists only of massive physical spin-2 particles. An interesting question now
is whether we have some f(□) which allows us to get rid of ghosts and mass discontinuity
simultaneously. If we define
s (□) ≡ −
4
[(D − 2) (□ − m2) + 2 (D − 1) f (□)]
.
(32)
In order to avoid any other propagating mode than the massive graviton, s(□) must be an
−1
analytic function, see (31). Moreover, the propagator G
f will continuously approach the EH
propagator as m → 0 if
s (□) = −
lim
m→0
4
(D − 2) □ .
(33)
Assuming that the graviton mass is the only mass scale present in the propagator, a possible
two parametric solution of the above requirements is given by2
s (□) = − 4
(
)
(
□ + b m2
(D − 2)
1 − e
−c2□2/m4
□2
)
.
(34)
where (b, c) are so far arbitrary real parameters. The function (34) does satisfy the previous
can be neglected as m → 0.
requirements if we assume3 that the exponential factor e
From (32) and (34) we have
−c2□2/m4
f(□) =
(D − 2)□2
(
2(D − 1)(□ + b m2)(
1 − e−c2□2/m4
) − (D − 2)
2(D − 1)
(□ − m2)
.
(35)
So the propagator (31) becomes
−1 =
G
4i
□ − m2 P
(2)
2
− 4i
(
)
(
□ + b m2
(D − 2)
1 − e
−c2□2/m4
□2
)
(0)
2 + · · ·
P
.
(36)
2 Exponentials of the square of the D’Alambertian have appeared before [31] in gravitational models.
3 From a more rigourous standpoint it is important to mention that the definition of the exponential operator itself may
be problematic. Infinite series of differential operators may have severe convergence problems, see for instance [32,
33]. It may also happen that integral representations (like a Fourier transform) are not really equivalent to an infinite
sum of derivatives. Heuristic equivalence arguments usually require the interchange of infinite sums and integrals
which may fail due to the lack of uniform convergence. In particular, the definition of the inverse operator is not
granted as we have tacitly assumed here, see also [34].
8
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
(37)
(38)
The apparent massless pole in the spin-0 sector has vanishing residue and we are only left with
a physical massive spin-2 particle for arbitrary values of (b, c) such that c ̸= 0.
The equations of motion from the linearized model (30) are given by
(
)
□ − m2
□
GL
µν
− θµν
f (□)
□ RL = Tµν
.
In the small mass limit we have from (35),
f (m → 0) =
(D − 2) m2
2 (D − 1)
[
1 − b +
b2m2
□ + O
)]
(
m4
□2
.
Therefore, the equations of motion tend to the usual linearized Einstein equations as m → 0,
we have no vDVZ mass discontinuity. In the opposite limit □/m2 → 0, inverting the operator
in front of the Einstein tensor we have
□
(
)
)
(
GL
µν = b
m2
□
θµνT
D − 1
−
m2
Tµν − θµνT
D − 1
+ O
□2 T
m4
(39)
.
So the long wavelength sources will be suppressed (degravitation) only if b = 0 which is
assumed henceforth in this section. Let us calculate the exchange amplitude in this case. First
we fix the de Donder gauge (10) again. From the trace of the equations of motion at the de
Donder gauge we have
GL = ηµνGL
µν =
(
RL =
1 − e
2 − D
2
−c2□2/m4
)
T ,
which is formally equivalent to a D’Alambertian equation :
(
□
h −
4
(D − 2)
1 − e
−c2□2/m4
□
)
= 0
T
.
(40)
(41)
Neglecting any non trivial solution (harmonic functions) , see footnote on page 3, the quantity
inside the brackets of (41) must vanish identically and h can be algebraically determined in
terms of the trace of the source h = h(T) which altogether with the de Donder gauge eliminates
five d.o.f. in D = 4. So we are left with the expected 5 d.o.f for a massive spin-2 particle in
D = 4. Substituting the trace h(T) back in (37) we solve for hµν and obtain the gauge invariant
exchange amplitude:
)
ˆ
(
A
m2
=
dDxhµνT
′
µν
ˆ
= −2
dDxT
′
µν
(
1
□ − m2
Tµν − ηµνT
D − 1
)
+
(
)
−c2□2/m4 − 1
e
(D − 1) (D − 2) □ ηµνT
,
(42)
Although the first term in (42) with the massive pole corresponds to the FP result, in the small
mass limit, neglecting the exponential factor, we have the usual (m = 0) LEH amplitude with
the (D − 2) factor plus corrections:
(
ˆ
)]
)
(
)
(
[
(
)
A
m2 → 0
= 2
dDx T
′
µν
1
□
Tµν − ηµνT
D − 2
+
m2
□2
Tµν − ηµνT
D − 1
+ O
m4
□3 T
.
(43)
9
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
From (43) and (36) we see that we have been able to accommodate absence of mass discon-
tinuity with absence of ghosts while keeping only massive gravitons as propagating modes.
Regarding the nonlinear completion of (30) with f(□) given in (35), we can write down,
suppressing indices and recalling that b = 0,
L
f (h) = h ∂2
(2)
h P
LEH h − m2
4
[
√
(1 − α) R (g) + α
2 h − (D − 2)
4 (D − 1)
(
Rµν
{
=
) RL
□
RL
−c2□2/m4
e
(
1 − e−c2□2/m4
)]}
1
1
□ Rµν − R
2 □ R
µν − (1 − γ) D
4 (D − 1)
}
R
hh
1
□2 R +
[
(1 − γ) Rµν
1
□2 R
√
−gR
□
−c2□2/m4
e
(
1 − e−c2□2/m4
) R
.
hh
−g
{
−g
{
√
− m2
4
− (D − 2)
4 (D − 1)
γ (D − 2)
4 (D − 3)
C
µναβ 1
□2 Cµναβ
]}
hh
(44)
We have added total derivatives (in the flat space) with the multiplicative arbitrary real con-
□2 Rµν terms lead to bad cosmological evol-
1
stants (α, γ). In [35] it has been shown that Rµν
□2 Cµναβ produces unstable tensor cosmological perturbations, see also
1
ution while Cµναβ
[36] . Notice that we cannot get rid of both terms simultaneously in (44). So in the next section
we replace the massive graviton by a massless one and consider ghost free models where the
nonlocal terms only involve the scalar curvature like for instance: 1
□2 R.
4. Nonlocal massless gravity with exponential factors
Let us consider a class of nonlocal ghost free gravities similar to (44) but containing only
massless gravitons, namely,
L
f0 (h, T) = hµν
)µναβ
(
∂2
LEH
hαβ − RL
f0 (□)
2 □2 RL + hµνTµν
−gR
f0(h, 0) =
√
,
[
(45)
]
The previous model has a natural non linear version L
. Once
again adding gauge fixing terms for the linearized reparametrization invariance we can obtain
the propagator
1 − f0(□)
2□2 R
−1
0 =
G
4i
□ P
(2)
2
−
4i
[(D − 2) □ + 2 (D − 1) f0 (□)]
(0)
2 + · · ·
P
,
(46)
where dots represent again terms which vanish when saturated with conserved sources.
Although, we do not have a massive graviton any more it is natural to introduce a mass scale
→ ∞ similarly to
0/(D − 1). By further imposing absence of
m0 and require that the nonlocal model continuously flow to GR as □/m2
0
the RR model of [37] which corresponds to f0 = m2
ghosts as in the last section we require:
−1
0 =
G
4i
□ P
(2)
2
− 4 i
)
(
□ + b0 m2
0
(D − 2)
which leads to
(
)
1 − e
−c2□2/m4
0
□2
(0)
2 + · · ·
P
,
f0 (□) =
(
2 (D − 1)
(D − 2) □2
) (
□ + b0 m2
0
1 − e
) − (D − 2) □
2 (D − 1)
.
−c2□2/m4
0
10
(47)
(48)
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
0/□ → 0, the Lagrangian (45) as well as the corresponding propagator (47) continuously
As m2
approach the linearized EH theory. In particular, we have
(
f0
0/□ → 0
m2
)
= − (D − 2) b0m2
2 (D − 1)
0
+
(D − 2)
2 (D − 1)
0m4
b2
0
□
· · ·
.
(49)
In the opposite limit □/m2
0
→ 0 we deduce
{
(
f0
□/m2
0
→ 0
)
=
(D−2)
2(D−1)
m4
c2□ − □
0
2
m2
0
b0 c2
− (D−2)
2(D−1)
;
+ · · ·
(
)
□ + · · ·
1 + 1
b2
0c2
(b0 = 0)
;
(b0
̸= 0)
,
(50)
0/□ → 0 and □/m2
→ 0, as far as b ̸= 0, the leading correction to the LEH
In both regimes m2
0/(D − 1)) of [37]. The RR model is
theory is proportional to m2
ruled out by solar system tests, see comments in [38, 39]. So it is important to consider the
effect of the next to leading correction for b ̸= 0 and the case b = 0, this is now in progress.
0 like the RR model (f0 = m2
0
Regarding the equations of motion from L
f0(h, T), they are given by
GL
µν
− θµν
f0 (□)
□ RL = Tµν
.
If we plug its trace in itself we can rewrite it as
(
µν = Tµν − θµνT
GL
D − 1
+
θµν
D − 1
□ + b0 m2
0
□
)
(
1 − e
−c2□2/m4
0
)
T ,
(51)
(52)
Notice that the first two terms on the right side of (52) are independent of the parameters (b0, c)
0/□ → 0 we have
and never degravitate. Moreover, if we neglect the exponential factor as m2
(
)
0
]
[
(53)
GL
µν
= Tµν +
m2
0
□ T ,
1 − b0 c2□
m2
0/□ → 0
m2
0 . On the other hand, in the infra red □/m2
− c2□2
b θµν
D − 1
So if b = 0, we identically recover the EH linearized equations and get rid of the nonlocal
→ 0 we have:
source term already for finite m2
m4 + O( □3
µν = Tµν − θµν
GL
m6 )
T which leads to a traceless Einstein tensor
(D−1)
and vanishing scalar curvature RL = 0 at leading order. Further investigations beyond the lin-
earized truncation about flat space made here must be carried out in order to check the viability
of the full nonlinear version of the nonlocal model presented here. Other nonlocal ghost free
models with only massless gravitons can be found in [40, 41]. They use entire functions of □
which however, affect the ultraviolet behavior of the theory in order to make it renormalizable.
Last, one might consider another application of the exponential factors employed here in
another nonlocal model with massless gravitons. Namely, in the nonlocal Deser–Woodard
, although the function f(□−1R) can be
(DW) model [20], L
fine tuned [42] to fit the ΛCDM cosmological evolution, it is known, see [38] and com-
ments in [39], that the DW model fails in reproducing the phenomenology at the solar
system scale. One might think of introducing a mass scale in the model by the replace-
0) f(□−1R), where the screening function may be given by r =
ment f(□−1R) → r(□/m2
)
(
≡ 1/m0 lies between the cosmological and the
1 − e
0). Assuming that l0
solar system length scales, close to both ends we have respectively r(□/m2
→ 0) = 1 −
0
□2/m4
0/□2. So we do not loose the ΛCDM cosmological evol-
ution in the first case and continuously approach the LEH theory in the second one.
0 + · · · and r(□/m2
R + R f(□−1R)R
→ ∞) → m4
/(□2/m4
−□2/m4
0
DW =
−g
√
[
]
0
11
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
5. Conclusions
It is known [1, 2] that the usual local description of linearized massive gravity in terms of the
paradigmatic FP [26] theory is ghost free but fails in reproducing solar system phenomena
like the gravitational lenses effect, no matter how small is the graviton mass. This is the vDVZ
mass discontinuity problem. In modern massive gravity models, like the bimetric model of
[10], which uses the fine tuned gravitational potential of [6], the nonlinear terms resolve the
incompatibility between the absence of vDVZ discontinuity and the absence of ghosts via
a sophisticated screening mechanism. Moreover the bimetric model is local and apparently
[15] cosmological viable. Its spectrum consists of a massless spin-2 particle in addition to the
massive graviton. It is natural to ask for a kind of minimal solution for the above incompatib-
ility with only a massive spin-2 particle in the spectrum.
−c2□2/m4
The problem is hard to solve in linearized theories about the flat space4. We have argued
that the incompatibility tends to persist also in linearized nonlocal theories. However, as we
have shown, it is possible to reconcile both features with the help of exponential factors with
. See the model (44) whose spectrum consists
an infinite number of derivatives, like e
only of a massive spin-2 particle with positive residue (physical) in the two point amplitude.
The equations of motion of (44) and the corresponding gauge invariant exchange amplitude
continuously flow to the GR results in the massless limit, see (37), (38) and (43). Moreover, we
have shown that they are consistent with 5 = 2s + 1 degrees of freedom in D = 4 as expected.
Using again exponential factors with infinite derivatives, we have introduced in section 4
a linearized nonlocal model, see (45) and (48), with a mass scale m0 but where the particle
content is restricted to a massless spin-2 particle without ghosts as in GR. In (45) the nonlocal
terms only involve the scalar curvature differently from (44) where dangerous terms from the
√
cosmological point of view, see [35, 36], like
□2 Cµναβ also
̸= 0 the leading non-
contribute. The model (45) depends upon two real parameters (b0, c). If b0
0/□ → 0
local term added to LEH becomes proportional to
0/(D − 1) and has been
and □/m2
0
ruled out by solar system phenomenology [38]. On the other hand, if b0 = 0 the leading non-
→ 0.
local term added to LEH becomes proportional to
Therefore, it is important to check the case b = 0 and how the corrections to the leading con-
tribution for b ̸= 0 respond to solar system scale phenomena.
→ 0 like the RR model of [37] which corresponds to f0 = m2
−gR(m4/□3)R in the infra red □/m2
0
0/□2)R in both limits m2
−gRµν 1
√
−gCµναβ 1
□2 Rµν and
−gR(m2
√
√
Another possibility is to follow an on shell approach, as in the case of the RT model
of [44], and try to covariantize the equations of motion (51) with (48), but in this case
one has to introduce auxiliary fields in order to make both sides of (37) transverse, since
[∇µ, □−1] ̸= 0. However, the auxiliary fields may lead to a non equivalent theory, see [45]
and further references and comments in [39]. All those possibilities are in the scope of our
present investigations.
We have also shown at the end of section 4 how the exponential factor could also be used
to soften the Einstein–Hilbert limit of the Deser–Woodard model [20] which might be useful
in reconciling the model with the Solar System size phenomenology. In summary there are
different applications of the exponential factor used here in nonlocal linearized theories which
4 The reader is invited to look at the work [43] where our 4D world is in a brane which is a slice of a 5D AdS bulk
theory. A kinetic term for the longitudinal component of the massive graviton is generated by the bulk theory which
weakens its coupling to matter and gets rid of the vDVZ discontinuity.
12
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
must be exploited beyond the linear truncation and even at linearized level about non flat back-
grounds. Finally, as a word of caution, the reader should be aware of convergence problems
for infinite series of derivatives, see footnote 3 and [32–34].
Data availability statement
All data that support the findings of this study are included within the article (and any supple-
mentary files).
Acknowledgments
We thank two anonymous referees for bringing [16–19, 32–34, 43] to our knowledge. The
author is partially supported by CNPq (Grant 313559/2021-0) and also thanks Alessandro L
R dos Santos for discussions.
ORCID iD
D Dalmazi https://orcid.org/0000-0001-8383-8367
References
[1] van Dam H and Veltman M J G 1970 Massive and massless Yang-Mills and gravitational fields
Nucl. Phys. B 22 397
[2] Zakharov V I 1970 Linearized gravitation theory and the graviton mass JETP Lett. 12 312
[3] Abbott B et al (Virgo, LIGO Scientific Collaboration) 2017 GW170817: observation of gravita-
tional waves from a binary neutron star inspiral Phys. Rev. Lett. 119 161101
[4] Vainshtein A I 1972 To the problem of nonvanishing gravitation mass Phys. Lett. B 39 393
[5] Boulware D G and Deser S 1972 Inconsistency of finite range gravitation Phys. Lett. B 40 227
[6] de Rham C, Gabadadze G and Tolley A J 2011 Resummation of massive gravity Phys. Rev. Lett.
106 231101
[7] Hassan S F and Rosen R A 2012 Resolving the ghost problem in non-linear massive gravity Phys.
Rev. Lett. 108 041101
[8] Riess A G et al (Supernova Search Team) 1998 Astron. J. 116 1009
[9] Perlmutter S et al (Supernova Cosmology Project) 1999 Astrophys. J. 517 565
[10] Hassan S and Rosen R A 2012 Bimetric gravity from ghost-free massive gravity J. High Energy
Phys. JHEP02(2012)126
[11] Hinterbichler K 2012 Theoretical aspects of massive gravity Rev. Mod. Phys. 84 671–710
[12] de Rham C 2014 Massive gravity Living Rev. Relativ. 17 7
[13] Schmidt-May A and von Strauss M 2016 Recent developments in bimetric theory J. Phys. A: Math.
Theor. 49 183001
[14] Heisenberg L 2019 Phys. Rep. 796 1–113
[15] Hogas M and Mortsell E 2021 J. Cosmol. Astropart. Phys. JCAP05(2021)002
[16] Baldacchino O and Schmidt-May A 2017 J. Phys. A: Math. Theor. 50 175401
[17] Albornoz N L G, Schmidt-May A and Strauss M V 2018 Dark matter scenarios with multiple spin-2
fields J. Cosmol. Astropart. Phys. JCAP01(2018)014
[18] Hassan S F and Schmidt-May A 2019 Interactions of multiple spin-2 fields beyond pairwise coup-
lings Phys. Rev. Lett. 122 251101
[19] Molaee Z and Shirzad A 2021 Class. Quantum Grav. 38 065006
[20] Deser S and Woodard R P 2007 Nonlocal cosmology Phys. Rev. Lett. 99 111301
[21] Arkani-Hamed N, Dimopoulos S, Dvali G and Gabadadze G 2002 Non-local modification of gravity
and the cosmological constant problem (arXiv:hep-th/0209227)
[22] Dvali G, Hofmann S and Khoury J 2007 Degravitation of the cosmological constant and graviton
width Phys. Rev. D 76 084006
13
Class. Quantum Grav. 40 (2023) 235011
D Dalmazi
[23] Dvali G 2006 New J. Phys. 8 326
[24] Porrati M 2002 Phys. Lett. B 534 209
[25] Jaccard M, Maggiore M and Mitsou E 2013 Nonlocal theory of massive gravity Phys. Rev. D
88 044033
[26] Fierz M 1939 Force-free particles with any spin Helv. Phys. Acta 12 3
Fierz M and Pauli W 1939 On relativistic wave equations for particles of arbitrary spin in an elec-
tromagnetic field Proc. R. Soc. A 173 211
[27] Khoudeir A, Montemayor R and Urrutia L F 2008 Phys. Rev. D 78 065041
[28] Barvinsky A O 2003 Nonlocal action for long-distance modifications of gravity Phys. Lett. B
572 109–16
[29] Dalmazi D and Mendonça E L 2009 A New spin-2 self-dual model in D = 2+1 J. High Energy
Phys. JHEP09(2009)011
[30] Andringa R, Bergshoeff E A, de Roo M, Hohm O, Sezgin E and Townsend P K 2010 Class.
Quantum Grav. 27 025010
[31] Krasnikov N V 1987 Nonlocal gauge theories Theor. Math. Phys. 73 1184–90
Krasnikov N V 1987 Teor. Mat. Fiz. 73 235
[32] Carlsson M, Prado H and Reyes E G 2016 Differential equations with infinitely many derivatives
and the Borel transform Ann. Henri Poincaré 17 2049–74
[33] Heredia C and Llosa J 2023 Infinite derivatives vs integral operators. The Moeller-Zwiebach para-
dox (arXiv:2302.00684)
[34] Heredia C, Koláˇr I, Llosa J, Torralba F J M and Mazumdar A 2022 Infinite-derivative linearized
gravity in convolutional form Class. Quantum Grav. 39 085001
[35] Cusin G, Foffa S, Maggiore M and Mancarella M 2016 Nonlocal gravity with a Weyl-square term
Phys. Rev. D 93 043006
[36] Cusin G, Foffa S, Maggiore M and Mancarella M 2016 Conformal symmetry and nonlinear exten-
sions of nonlocal gravity Phys. Rev. D 93 083008
[37] Maggiore M and Mancarella M 2014 Non-local gravity and dark energy Phys. Rev. D 90 023005
[38] Belgacem E, Finke A, Frassino A and Maggiore M 2019 Testing nonlocal gravity with lunar laser
ranging J. Cosmol. Astropart. Phys. JCAP02(2019)035
[39] Calcagni G 2021 Non-local gravity Modified Gravity and Cosmology (Springer)
[40] Modesto L 2012 Super-renormalizable quantum gravity Phys. Rev. D 86 044005
[41] Biswas T, Gerwick E, Koivisto T and Mazumdar A 2012 Towards singularity- and ghost-free the-
ories of gravity Phys. Rev. Lett. 108 031101
[42] Deffayet C and Woodard R P 2009 Reconstructing the distortion function for nonlocal cosmology
J. Cosmol. Astropart. Phys. JCAP08(2009)023
[43] Gabadadze G, Older D and Pirtskhalava D 2019 Phys.Rev. D 100 124017
[44] Maggiore M 2014 Phantom dark energy from nonlocal infrared modifications of general relativity
Phys. Rev. D 89 043008
[45] Koshelev N A 2009 Comments on scalar-tensor representation of nonlocally corrected gravity Grav.
Cosmol. 15 220–3
14
| null |
10.1103_physrevd.105.123002.pdf
| null | null |
PHYSICAL REVIEW D 105, 123002 (2022)
Revisiting the dark matter interpretation of excess rates in semiconductors
Peter Abbamonte,1,* Daniel Baxter
Noah Kurinsky,6,∥
Yonatan Kahn,1,3,‡
Bashi Mandava ,1,7,¶ and Lucas K. Wagner1,8,**
Gordan Krnjaic,2,4,5,§
,2,†
1Department of Physics, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801, USA
2Fermi National Accelerator Laboratory, Batavia, Illinois 60510, USA
3Illinois Center for Advanced Studies of the Universe, University of Illinois at Urbana-Champaign,
Urbana, Illinois 61801, USA
4Department of Astronomy and Astrophysics, University of Chicago, Chicago, Illinois 60637, USA
5Kavli Institute for Cosmological Physics, University of Chicago, Chicago, Illinois 60637, USA
6SLAC National Accelerator Laboratory, 2575 Sand Hill Road, Menlo Park, California 94025, USA
7Berkeley Center for Theoretical Physics, University of California, Berkeley, California 94720, USA
8Institute for Condensed Matter Theory, University of Illinois at Urbana-Champaign,
Urbana, Illinois 61801, USA
(Received 15 February 2022; accepted 13 May 2022; published 1 June 2022)
In light of recent results from low-threshold dark matter detectors, we revisit the possibility of a common
dark matter origin for multiple excesses across numerous direct detection experiments, with a focus on the
excess rates in semiconductor detectors. We explore the interpretation of the low-threshold calorimetric
excess rates above 40 eV in the silicon SuperCDMS Cryogenic Phonon Detector and above 100 eV in the
germanium EDELWEISS Surface detector as arising from a common but unknown origin, and demonstrate
a compatible fit for the observed energy spectra in both experiments, which follow a power law of index
α ¼ 3.43þ0.11
−0.06 . Despite the intriguing scaling of the normalization of these two excess rates with
approximately the square of the mass number A2, we argue that the possibility of common origin by
dark matter scattering via nuclear recoils is strongly disfavored, even allowing for exotic condensed matter
effects in an as-yet unmeasured kinematic regime, due to the unphysically large dark matter velocity
required to give comparable rates in the different energy ranges of the silicon and germanium excesses. We
also investigate the possibility of inelastic nuclear scattering by cosmic ray neutrons, solar neutrinos, and
photons as the origin, and quantitatively disfavor all three based on known fluxes of particles.
DOI: 10.1103/PhysRevD.105.123002
I. INTRODUCTION
Direct detection experiments searching for particle dark
matter (DM) with masses below 1 GeV have made signifi-
cant advancements in the last decade, driven by lower
thresholds, improved resolution, and sophisticated analysis
techniques [1]. These experiments are on the forefront of
new technological development, and have demonstrated
sensitivity to individual electron-hole pair creation at the eV
*[email protected]
†
[email protected]
‡
[email protected]
§[email protected]
∥
[email protected]
¶[email protected]
**[email protected]
Published by the American Physical Society under the terms of
license.
the Creative Commons Attribution 4.0 International
Further distribution of this work must maintain attribution to
the author(s) and the published article’s title, journal citation,
and DOI. Funded by SCOAP3.
energy scale [2–7] as well as eV-scale calorimetry enabling
direct energy measurements independent of charge produc-
tion [8–10]. An important distinction between ionization
and calorimetric detectors is that ionization detectors are all
limited by uncalibrated, nonradiogenic backgrounds which
are often referred to as dark rates. A dark rate can in principle
arise from any source that produces anomalous ionization
events in a detector, with an irreducible contribution from
thermal processes at the detector temperature. Substantial
effort is under way to better characterize these dark rates
[11,12]. On the other hand, calorimetric detectors currently
have higher energy thresholds but do not suffer from the dark
rates mentioned above. This complementarity offers an
interesting window on new physics when the two detector
types are taken together, as was previously done in Ref. [13].
In this paper, we continue in the spirit of Ref. [13] by
performing a joint analysis of the two most recent results
from calorimetric semiconductor detectors,
the silicon
SuperCDMS Cryogenic Phonon Detector (SuperCDMS
CPD) [8] and the germanium EDELWEISS Surface detector
2470-0010=2022=105(12)=123002(10)
123002-1
Published by the American Physical Society
PETER ABBAMONTE et al.
PHYS. REV. D 105, 123002 (2022)
(EDELWEISS-Surf) [9]. Both experiments observe a sta-
tistically significant excess event rate above known back-
ground sources near threshold. Our analysis here differs
from Ref. [13] because recent work has sharply constrained
our previously proposed signal models: the plasmon pro-
duction channel from nuclear scattering is only a small part
of the total spectrum from the Migdal effect in solid-state
systems and cannot account for the observed spectral shape
[14,15], and a fast DM subcomponent is excluded by
XENON1T except for a very narrow range of DM velocities
[16]. That said, our approach is similar in that we consider
novel inelastic nuclear scattering channels where the rela-
tionship between the deposited energy Er and the momen-
tum transfer from the DM q differs from Er ¼ q2=ð2mNÞ
(where mN is the mass of the nucleus) expected from free-
particle elastic scattering.1 Indeed, given that the energy
scales of the excess are close to the lattice displacement
energy, many-body effects may be expected to be important
[18], and collective effects do substantially extend the reach
of semiconductor ionization detectors to sub-GeV DM
through the Migdal effect [15], compared to calculations
which assume isolated atom targets [19–23].
to attempt
the detector
This paper is organized as follows. In Sec. II, we review
the recent progress in understanding the persistent excesses
in low-threshold detectors, and perform a combined fit to
the SuperCDMS CPD and Edelweiss-Surf excesses, dem-
onstrating an intriguing consistency in spectral index and
normalization which is suggestive of a possible DM
interpretation. In Sec. III, we use a phenomenological
response, parametrized by the
model of
dynamic structure factor,
to explain both
excesses in the context of inelastic DM-nuclear scattering.
We find that such an interpretation is inconsistent even
allowing for exotic structure factors, largely due to the fact
that the allowed region for the silicon excess rate requires
dark matter masses small enough that they have insufficient
kinetic energy to yield the measured germanium rate at
higher energies. In Sec. IV, we argue that the excess is also
inconsistent with nuclear scattering from known particle
sources, namely cosmic-ray neutrons, photons, and neu-
trinos, as well as secondary interactions. We conclude in
Sec. V with our summary of this puzzling situation: the
calorimetric excesses remain robustly mysterious.
II. COMBINED ANALYSIS OF SEMICONDUCTOR
EXCESSES
We noted in Ref. [13] that there was significant dis-
crepancy at the time among the excesses in the silicon
ionization detectors SuperCDMS HVeV, DAMIC at
SNOLAB, and SENSEI [3,5,24], each of which observed
different single-electron dark rates. SENSEI has since
released new results [4] from a detector operated with
shallow 225 m.w.e. (meters water equivalent) overburden
that reduced their measured single (multiple)-electron dark
rate to 5ð0.05Þ Hz=kg, consistent with the DAMIC single-
electron dark rate of 7 Hz=kg [24] despite the increased
shielding and 6000 m.w.e. overburden at SNOLAB. This
resolved the initial tension mentioned in Ref. [13] and
indicated some unrelated origin for the single (multiple)-
electron dark rate background in the SuperCDMS HVeV
detector of 1700ð13Þ Hz=kg [5,25]. Moreover, recent work
[12] has demonstrated consistency between some of these
dark rates and secondary background processes, such as
from Cherenkov emission, indicating a potential radiogenic
contribution to these backgrounds.
Thus, since there has been much progress toward under-
standing the excesses in ionization detectors, we now focus
exclusively on a common interpretation of semiconductor
calorimetric excesses, which remain mysterious. The
SuperCDMS CPD [8] excess in silicon is analogous to
the earlier EDELWEISS-Surf measurement in germanium
[9] in that it measures the total recoil energy deposited in the
detector Er, regardless of the distribution of the primary
event energy into heat or charge (less any persistent defect
energies which are on the order of 4 eV per defect [26] and
are neglected in this analysis). Both detectors are also
notably operated on the surface with minimal shielding.
Whereas in Ref. [13], we focused primarily on qualitative
arguments to motivate further interest in these excesses, here
we perform a more quantitative analysis of the SuperCDMS
CPD [8] and EDELWEISS-Surf [9] excess rates.
The SuperCDMS CPD result is of particular interest
because its threshold (25 eV) is considerably lower than
that of EDELWEISS-Surf (60 eV). Both detectors measure
an approximately exponential background near threshold
which is likely from noise triggers that are not removed by
the analysis cuts. In the case of EDELWEISS-Surf, a model
for these noise-induced triggers was published in Ref. [9]
and has been incorporated into this analysis directly, with
no free parameters. For SuperCDMS CPD, these triggers
are likely coming from environmental noise, and thus do
strictly follow an exponential in energy. At higher energies,
both detectors are limited by “flat” radiogenic back-
grounds (e.g., Compton scattering [27]) on the order of
105 cts kg−1 day−1 keV−1, as is to be expected for detectors
operating on the surface. However, between these two
distinct features, both detectors observe a statistically
significant excess of events. EDELWEISS explored in
these excess events in
great detail
germanium come from elastic or Migdal2 scattering of
the possibility that
1Inelasticity here refers exclusively to detector response and is
not to be confused with inelastic DM, which is a mass splitting
between different DM states [17].
2Reference [9] uses the isolated atom formalism [19] to
calculate these rates in germanium, which neglects important
collective effects [28]; that said, the isolated atom approach was
the only calculation in the literature at the time of publication.
123002-2
REVISITING THE DARK MATTER INTERPRETATION OF …
PHYS. REV. D 105, 123002 (2022)
TABLE I. Values for the fit parameters of Eq. (1) for each of the cases considered in the text. The p-values are calculated based on the
listed Pearson χ2 values and degrees of freedom (number of bins minus parameters in each fit). Here, larger p-values indicate a better fit.
The poor (but notably nonzero) p-value when including the EDELWEISS-Surf data is driven by the bin just below 100 eV, which is
pulling the power law index up and is maximally sensitive to systematics in the (fixed) noise-induced trigger model, which are not
included in the fit. Of particular note, all three fits give consistent power law indices α, which is positively correlated with both the
overall normalization C and the ratio of the rates κ2
Ge
=κ2
Cκ2
Si
½ðkg dayÞ−1 keVα−1(cid:2) Cκ2
Ge
10.8þ5.6
−2.6 × 1011
4.1 (cid:3) 1.5 × 1011
NA
Si. (NA represents “not applicable.”).
α
½ðkg dayÞ−1 keVα−1(cid:2)
7.9þ5.0
−2.3 × 1012
NA
51.4þ44.2
−25.9 × 1012
3.43þ0.11
−0.06
3.18þ0.10
−0.09
3.84þ0.13
−0.15
κ2
=κ2
Ge
Si
7.2þ0.7
−0.6
NA
NA
χ2=d:o:f:
227.2=174
121.5=121
101.3=52
p-value
0.004
0.471
5 × 10−5
Combined fit (Fig. 1)
SuperCDMS CPD [8] only
EDELWEISS-Surf [9] only
DM particles and found that neither gives a good spectral
match to the data [9].
triggers
for noise
leaking above
We simultaneously fit the digitized SuperCDMS CPD
and EDELWEISS-Surf data to the following model: a flat
background Di for each detector (1 parameter each), a
threshold
model
fðErÞ (consisting of a two-parameter exponential
for
SuperCDMS CPD and a zero-parameter model
for
EDELWEISS-Surf taken directly from Ref. [9] and scaled
by signal efficiency to compare with data), and a power law
component in recoil energy for the excess (with indepen-
dent normalization for each detector and a common power
law index for both, three parameters total):
dRi
dEr
¼ ðCκ2ÞiE−α
r þ Di þ fiðErÞ;
ð1Þ
where i ¼ Si; Ge. We write the normalization of the excess
in the suggestive form ðCκ2Þ because the DM model we
present in Sec. III will contain an overall normalization of
the DM-nucleon cross section proportional to C and a
detector-dependent factor κ2 which will be A2 or Z2 for DM
which couples to nucleons or protons, respectively (here A
is the mass number and Z is the atomic number of the
target). Under the assumption of a common origin between
the two detectors, the common normalization C cancels
yielding only a detector-dependent ratio κ2
Si. This joint
Ge
fit provides a best fit power law index of α ¼ 3.43þ0.11
−0.06 and
¼ 7.2þ0.7
a Ge-to-Si normalization ratio of κ2
−0.6 . The
Ge
results of this fit are shown in Fig. 1 and presented in
Table I.3 We also perform individual fits to each detector
spectrum, by allowing a power law index αi which differs
for each dataset. Notably,
the fits when including the
EDELWEISS-Surf data are worse, which would be sig-
nificantly improved by adding additional fit parameters to
capture the systematic uncertainties present in the high-side
tail of the noise-induced trigger model for that data.
=κ2
Si
=κ2
3A similar analysis was recently performed on excess rates in
sapphire [29], which focused instead on the potential creation of
defects through an exotic power-law nuclear scattering channel.
that
The fact
these two independent datasets from
different collaborations with different sensor technologies
and different target materials (but comparable mK temper-
atures) measure an excess of events at
low energies
following compatible power laws is by itself interesting.4
Independent of their respective rates, this common power
law is potentially indicative of a similar (or even identical)
physical process as the origin of these events in each
detector. Notably, the excess rates in each of these detectors
can also be individually fit to an exponential, rather than a
power law; however, the different energy ranges of the two
excess rates exclude the possibility of a common expo-
nential decay constant. In addition to the common power
law index α ≈ 3.4, the fact that the ratio of the κ2 is
consistent with the ratio of the square of the mass numbers
A of the two targets, ð74=28Þ2 ¼ 7.0, is intriguing given
that the standard benchmark model of spin-independent
DM-nuclear scattering scales precisely in this fashion.
III. DARK MATTER INTERPRETATION
THROUGH EXOTIC STRUCTURE FACTORS
To see whether the observed excess might be consistent
with a DM interpretation yielding a power-law energy
spectrum dR=dω ∝ ω−3.4, here we consider a generic
formalism for calculating the rates for DM-nuclear scatter-
ing on solid-state targets using an empirical parametrization
of the dynamic structure factor that allows for physically
allowed, but nontrivial, collective effects without neces-
sarily requiring a microscopic interpretation.
In what
follows, we will refer to the energy deposited by the
DM as ω rather than Er,
that
collective effects may play a role and that we are not
dealing with the elastic recoil energy of a single isolated
nucleus.
to emphasize the fact
4Curiously, the prediction of the Migdal effect from Ref. [15]
is a power law with index α ¼ 4 at these energies, but the total
rate is inconsistent with the interpretation of these events as
coming from DM scattering through the Migdal effect, and
furthermore the large energy deposit to the electronic system
would likely result in excess ionization yields which are not seen
in the ionization detectors at these energies [7,30].
123002-3
PETER ABBAMONTE et al.
PHYS. REV. D 105, 123002 (2022)
FIG. 1. The efficiency-corrected calorimetric rates (red points) from the silicon SuperCDMS CPD [8] (left) and germanium
EDELWEISS-Surf [9] (right) detectors with statistical error bars are plotted against a global three-component fit [thin black line, Eq. (1)]
consisting of noise triggers above threshold (black dashed, see text for more details), a one-parameter flat component representing
standard radiogenic backgrounds (black dotted), and excess events which are fit to a E−3.43
power law dependence (solid thick blue). For
the excess component, which is considered for this work as arising from inelastic nuclear scattering, a 1σ uncertainty band from the
combined fit is also shown (shaded blue), along with the fit to each dataset separately (dashed blue). In the combined fit, the two datasets
are fit simultaneously, with separate noise trigger and flat background components for each detector but a common power law index. The
detector thresholds are represented by the grey shaded regions at low energy.
r
the DM-detector
Assuming nothing about
system
other than the validity of the Born approximation, the
DM-nuclear scattering rate may be expressed in terms of
the dynamic structure factor [31], which encapsulates the
response of the target to a perturbation of the ion density,
Sðq; ωÞ ¼ 2π
X
jhΨβjnqjΨ0ij2δðω − ωβÞ:
ð2Þ
β
Here, jΨ0i is the ground state of the system, jΨβi runs over
all final states, and nq is the density operator in Fourier
space,
nq ¼
1
p
ffiffiffiffi
V
X
eiq·rj;
j
ð3Þ
where rj are the positions of all the nuclei in the target and
V is the detector volume. When the dynamic structure
factor is isotropic, Sðq; ωÞ ¼ Sðq; ωÞ, the differential DM
scattering rate per unit target mass can be obtained from the
structure factor as [31]
Z
dR
dω
¼
ρχ
mχ
κ2 ¯σn
2μ2
χn
1
2πρT
dqqSðq; ωÞηðv
min
Þ;
ð4Þ
where ρχ ¼ 0.3 GeV cm−3 is the local DM density; ¯σn is
the fiducial DM-proton or DM-nucleon cross section; κ2 ¼
Z2 or A2 depending on whether the DM couples to protons
or all nucleons, respectively; ρT ¼ mNn0 is the target mass
density; ηðv
min is
the minimum DM speed required to deposit energy ω,
Þ is the DM mean inverse speed; and v
min
v
min
ðq; ωÞ ¼
ω
q
þ
q
2mχ
:
ð5Þ
DM
We have assumed a heavy mediator such that the cross
section is independent of q [i.e., F
ðqÞ ¼ 1], both for
simplicity and to more easily make contact with exper-
imental limits making the same assumption [32]. Similarly,
in the DM mass range we will be interested in, there is
insufficient momentum to probe nuclear substructure and
so we also set the nuclear form factor to unity. It is clear
from Eq. (4) that a choice of Sðq; ωÞ fully determines the
spectral shape of the differential scattering rate, given a
choice of DM velocity distribution. The integrated rate
requires further input from the DM interaction strength,
parametrized by κ2 ¯σn.
A first-principles computation of the structure factor is
possible in specific simplified models, including treating
jΨ0i and jΨβi as single-particle harmonic oscillator states
or plane waves [15,18]. However, when considering the
energy deposit to the scattered nucleus alone (in contrast to
the ω−4 electronic energy spectrum noted in footnote 4),
such a model either yields the ordinary flat spectrum of
elastic
a
steeply
, or
falling spectrum dR=dω ∝ exp½−ω2mN=ðq2
ω0Þ(cid:2) when
p
q ≪
, where ω0 ≃ 60 meV is the optical phonon
energy in Si or Ge and q
≃ 2mχv is the maximum
max
momentum transfer. Both of these spectral shapes are
clearly inconsistent with the data.
scattering when q ≃
ffiffiffiffiffiffiffiffiffiffiffiffiffi
2mNω
ffiffiffiffiffiffiffiffiffiffiffiffiffi
2mNω
max
p
To attempt to reproduce the observed power-law spec-
trum, we first suppose that the structure factor is dominated
by a single-quasiparticle excitation, representing a single
scattered nucleus interacting with the surrounding electron
123002-4
REVISITING THE DARK MATTER INTERPRETATION OF …
PHYS. REV. D 105, 123002 (2022)
density. In this case, the dynamic structure factor may be
parametrized as
dR
dω
¼
ρχ
mχ
κ2 ¯σn
2μ2
χn
2A2
q
j2 − nj
FnðωÞ
q0
(cid:3)
(cid:4)
2n
η½v
min
ðωÞ(cid:2);
ð9Þ
(cid:3)
Sðq; ωÞ ¼ 2πn0SðqÞδ
ω −
(cid:4)
:
q2
2mNSðqÞ
ð6Þ
if SðqÞ ¼ 1,
The function SðqÞ is known as the static structure factor and
parametrizes departures from the free-particle dispersion
the dispersion relation is ω ¼
relation:
q2=ð2mNÞ as expected for elastic nuclear recoil, but static
structure factors which differ from unity permit different
dispersions. Furthermore, for any choice of SðqÞ, Sðq; ωÞ
in Eq. (6) automatically satisfies the “f-sum rule”
Z
∞
0
dω
2π
ωSðq; ωÞ ¼
q2
2mN
n0;
ð7Þ
which is a consistency condition on physically realizable
dynamical structure factors imposed by causality and
conservation of mass.
We now make an ansatz for the form of SðqÞ designed
specifically to yield the desired power-law spectrum.
Suppose that the static structure factor is isotropic and
itself follows a power law,
SðqÞ ≈ Aqðq=q0Þn
ðAq ¼ 0.015Þ;
ð8Þ
for mN ¼ 26 GeV in
over a limited range of q around a fiducial momentum value
q0. The prefactor Aq may of course be absorbed into q0, but
is explicitly separated here to better illustrate typical
kinematics: if mχ ¼ 200 MeV, its typical momentum is
q ∼ 200 keV, and in order for Sðq; ωÞ to have support at
ω ¼ 50 eV and q ¼ 200 keV, we must have SðqÞ ¼
q2=ð2mNωÞ ¼ 0.015
silicon.
Indeed, the fact that Aq ≪ 1 (so SðqÞ ≪ 1 for q near q0)
reflects the highly inelastic nature of the scattering inter-
pretation of the excess: much more energy is deposited for a
given momentum transfer than would be expected from
elastic scattering. The free parameters in this model are thus
the momentum scale q0 and the power law index n. We
emphatically do not attempt any microscopic explanation
the
of such a structure factor, but simply note that
(uncalibrated) energy regime we are concerned with here
is just above the typical displacement energy in Si and Ge
required to remove a nucleus from its lattice site, and thus
we might expect qualitatively different behavior than in the
single-phonon or high-energy ballistic recoil
regimes,
perhaps due to binding potential effects which distort the
outgoing wave function, and/or interactions of the charged
recoiling ion with the electron system.
Plugging in the power law ansatz for SðqÞ into Eq. (4),
and rearranging to emphasize the similarities to Eq. (1),
yields
where we have defined the dimension-1 quantity
FnðωÞ ≡
(cid:3)
2mNAqω
qn
0
(cid:4) 1
2−n;
such that the minimum velocity becomes
v
min
ðωÞ ¼
ω
FnðωÞ
þ
FnðωÞ
2mχ
:
ð10Þ
ð11Þ
Note that F0ðωÞ ¼ q for the elastic case with Aq ¼ 1. If
v
ðωÞ is independent of ω (which is approximately true
min
for sufficiently large mχ), then Eq. (9) reduces to
dR
dω
≈ ðCκÞ2ω−α;
ð12Þ
min
and the spectrum is (by construction) exactly a power law
with α ¼ 2n=ðn − 2Þ. Including the effects of ηðv
Þ will
the spectrum for smaller mχ, since less kinetic
distort
energy and less momentum are available for scattering, as
well as for small q0 which pushes the scattering to the high-
velocity tail. Therefore, Eq. (12) is approximate and results
from fitting the full spectrum to a power law. In particular,
taking n ¼ 5ðn ¼ 6Þ yields dR=dω ∝ ω−3.3 ðdR=dω ∝
ω−3Þ up to velocity-suppression effects (which would begin
to increase the effective α, rapidly in the case of Ge).
Figure 2 shows the results of fitting the Si and Ge spectra
dR=dω with a power law, as a function of q0 and mχ with
SðqÞ ¼ 0.015ðq=q0Þn for n ¼ 5, 6. We see that achieving
the index of α ≈ 3.4 preferred by the data is allowed in Si
for a wide range of values for both mχ and q0. The higher
energies of the excess in Ge make the same power law fit
difficult because of the effects of velocity suppression,
yielding a much narrower parameter space which does not
overlap in q0 with the Si best-fit contours except at the
largest DM masses. At this stage, the difficulty of fitting
both spectra simultaneously is clear, at least assuming that
Si and Ge have comparable structure factors.
integrated
Once the power-law dependence of SðqÞ is fixed, the
normalization of the spectrum is also fixed up to the overall
scaling by ¯σn. For the same n ¼ 5ðn ¼ 6Þ dependence of
SðqÞ and taking κ2 ¼ A2, the magenta solid (dashed) line in
Fig. 3 shows the preferred region of ¯σn and mχ which yields
an
for
ω ∈ ½40; 100(cid:2) eV, and also yields a spectrum with the
individual best-fit power law index α ¼ 3.18 for the CPD
data in that energy range. Points on the magenta curves in
Fig. 3 correspond to taking parameters along the corre-
sponding ðmχ; q0Þ magenta contours in Fig. 2. We see that
heavier dark matter masses mχ ≳ 170 MeV are robustly
0.6 Hz=kg
silicon
rate
of
in
123002-5
PETER ABBAMONTE et al.
PHYS. REV. D 105, 123002 (2022)
FIG. 2. Contours of the power-law index α dependence of dR=dω in silicon (left) and germanium (right) as a function of q0 and mχ
with SðqÞ ¼ 0.015ðq=q0Þn for n ¼ 5 (top) and n ¼ 6 (bottom), from Eq. (12). The contours in each panel represent values of mχ and q0
which consistently yield a power law with the labeled value of α, as shown in Eq. (12). The black shaded regions yield zero events at
ω ¼ 75 eV for Si and at ω ¼ 150 eV for Ge (here, ω ≡ Er) and/or a nonmonotonic spectrum, either of which is inconsistent with the
data. The best-fit contours for each, α ¼ 3.18 for Si and α ¼ 3.84 for Ge, are shown in magenta.
excluded by CRESST-III [10], but intriguingly, the pre-
ferred region for lower DM masses mχ ≃ 100 MeV is not
excluded by any nuclear scattering experiment. Note that
including an additional elastic term in the structure factor
which has support at the same values of ω amounts to
taking SðqÞ ¼ 1 in a regime of q distinct from the one
where the inelastic structure factor has support. Since the
sum rule in Eq. (7) fixes the normalization of the structure
factor at all q, such an elastic contribution would only serve
to increase the rate, and therefore in principle this could
push the preferred values of ¯σn slightly lower. However, for
mχ ≲ 400 MeV, DM with velocity below the lab-frame
galactic escape velocity cannot yield an elastic nuclear
recoil in Si with energy above 40 eV, so all contributions to
the observed excess above the exponential noise trigger
must come from the inelastic structure factor.
Taking the same structure factor parameters along the Si
best-fit contours in Fig. 2, the green line in Fig. 3 shows the
region of ¯σn and mχ which yields a total rate of 1.3 Hz=kg
in germanium for ω ∈ ½100; 250(cid:2) eV, with the gradient
indicating the power law dependence of the spectrum. Not
only is the cross section ¯σn inconsistent with Si, but the Ge
spectrum is everywhere too steep to match the best-fit value
of α ¼ 3.84, except near mχ ≃ 500 MeV which is excluded
by several other experiments. It is possible that the cross
sections may be brought into agreement by widely differing
values of Aq (or equivalently q0) between Si and Ge,
but the fact that the allowed region for Si is restricted to
mχ ≲ 170 MeV means that only DM on the high-velocity
tail of the DM distribution has enough kinetic energy to
generate events in the 200 eV range, regardless of the
structure factor. This will always serve to steepen the
power law index beyond what is observed in the data and
renders
the
SuperCDMS CPD and EDELWEISS-Surf data highly
implausible.
the simultaneous DM interpretation of
123002-6
REVISITING THE DARK MATTER INTERPRETATION OF …
PHYS. REV. D 105, 123002 (2022)
varying only by Oð1Þ factors over 10 orders of magnitude
between En ¼ 10 meV and En ¼ 100 MeV [37,38]. Here,
Φ0 ≈ 1 × 10−3 Hz=cm−2 is the approximate CR neutron
flux at sea level. This spectrum translates to a CR neutron
velocity distribution fðvÞ ∝ 1=v2. In the case of elastic
scattering parametrized by a neutron-nucleus cross section
σnN, this leads to an energy spectrum
dR
CRn
dω
(cid:5)
(cid:5)
(cid:5)
(cid:5)
el:
¼
Φ0σnN
mN
ω−1;
ð14Þ
which has the wrong power-law index to match the
observed excess. Moreover, for Si, taking σnN ¼ 4πa2,
where a ¼ 4.2 fm is the neutron scattering length in Si, the
total rate between 40 and 100 eV is ≃0.05 Hz=kg, a factor
of 10 below the measured excess rate. In order to achieve a
power-law spectrum ω−3.18 and an integrated rate of
1 Hz=kg, one would have to postulate a neutron energy
spectrum dΦ=dEn ∝ E−3.18
with a total neutron flux more
than 5 orders of magnitude larger than Φ0 since the
different neutron spectrum implies a different normaliza-
tion for neutrons of the appropriate energy. Even if we
further speculate a peculiar inelastic dispersion from a
nontrivial structure factor SðqÞ, which could reconcile the
spectral index of the excess with the observed log-flat CR
neutron spectrum,
the large observed rate would still
require a substantial additional flux of neutrons that is
not observed.
n
Similar reasoning can rule out nuclear recoils induced
from known fluxes of either incident neutrinos or photons.
While exotic structure factors of the kind considered in
Sec. III can change the spectral shape, the overall nor-
malization is still driven by the total cross section for
photons or neutrinos scattering off an individual nucleus,
σνN ≈
G2
F
4π
Q2
WmNω
max
¼ 2.3 × 10−42 cm2
ð15Þ
σγN ≈
8π
3
Z4α2
m2
N
¼ 1.0 × 10−29 cm2;
ð16Þ
max
where QW ¼ N − Zð1 − 4 sin2 θWÞ is the weak charge of a
nucleus with N neutrons and Z protons, and we have given
the numerical values for silicon ðω
¼ 140 eVÞ. For the
photon cross section, we have assumed coherent Thomson
scattering from x-ray or gamma-ray photons with
Eγ ≪ mN, which dominates over other photonuclear proc-
esses such as Delbrück scattering and resonant scattering
[39]. For the neutrino cross section we have assumed
coherent scattering with ω ≪ Eν and sufficiently low
momentum transfer that the nuclear form factor is approx-
imately unity, a reasonable approximation even for elastic
scattering in the energy range relevant for the excesses. The
incident particle X is
total
R ¼ ΦXNTσXN, where NT ≈ 2 × 1025=kg is the number
rate per unit mass for
FIG. 3. Parameter space for a DM interpretation of the excess
rates in SuperCDMS CPD and EDELWEISS-Surf. The solid
(dashed) magenta contour corresponds to an integrated rate of
0.6 Hz=kg in SuperCDMS CPD for ω ∈ ½40; 100(cid:2) eV, SðqÞ ¼
Aqðq=q0Þn for n ¼ 5 (n ¼ 6) with Aq ¼ 0.015, and q0 chosen
along the best-fit power law contour α ¼ 3.18 from Fig. 2. The
green contour shows the same structure factor applied to
EDELWEISS data, normalized to a total rate of 1.3 Hz=kg for
ω ∈ ½100; 250(cid:2) eV; the color gradient indicates the power law
index, which is everywhere steeper than the best-fit α ¼ 3.84.
Both contours correspond to a DM-nucleon interaction with
κ2 ¼ A2. The mismatched power law indices and DM-nucleon
cross sections between the two experiments indicate the tension
in a DM interpretation. Also included are elastic DM-nucleon
scattering limits from CRESST [10,33], SuperCDMS CPD [8],
and EJ-301 [34].
Even attempting to explain one or the other of the
excesses, rather than both, requires an extremely peculiar
inelastic dispersion ω ∝ q−3 which arises from SðqÞ ∝ q5.
That said, systems with such a dispersion, where the energy
of the excitation decreases with increasing momentum, are
not unheard of; indeed, superfluid helium exhibits this
phenomenology between the maxon and roton regions [35],
as do plasmons in some transition metal dichalcogenides
[36]. Testing this explanation of the excess would require
measuring the structure factor in semiconductors with
neutron scattering, exactly as was done to determine the
structure factor of helium, but with momentum transfers on
the order of q0 and energy deposits in the 40–100 eV
energy range.
IV. RULING OUT KNOWN PARTICLE SOURCES
Since even rather unusual condensed matter effects are
unable to furnish a consistent DM interpretation, and given
that both detectors were operated on the surface, we also
examine the possibility that the excess is due to cosmic-ray
(CR) neutron scattering. The CR neutron spectrum at
ground level is very close to flat in ln En, where En is
the CR neutron energy,
dΦ
d ln En
≈ Φ0;
ð13Þ
123002-7
PETER ABBAMONTE et al.
PHYS. REV. D 105, 123002 (2022)
density of nuclei in Si, ΦX is the flux of particles in
question, and σXN its cross section with nuclei.
For neutrinos, the total flux at the surface is dominated by
keV-volt solar neutrinos, Φν ≈ 5 × 1010 Hz=cm2 [40]. The
largest the rate can possibly be is if all of these neutrinos
contributed to scattering (of course, this would also require a
highly inelastic structure factor), in which case the total rate
would be at most
Rν ≈ 3 × 10−6 Hz=kg;
ð17Þ
clearly ruling out solar neutrinos.
We can estimate the total photon flux from the measured
Compton rate in the SuperCDMS CPD detector, which is
approximately 105 events kg−1 day−1 keV−1 at low ener-
gies. If we conservatively assume this rate is flat out to
1 MeV and integrate over this full range, we get an
integrated rate of ∼103 Hz=kg. The Compton cross section
per electron is approximately σγe ¼ ð8π=3Þα2=m2
e, so the
ratio of nuclear Thomson to Compton cross sections
(accounting for the Z electrons per nucleus) is σγN=σγe ¼
Z3m2
N ¼ 2 × 10−4 in Si. An upper bound on the total
nuclear scattering rate from these photons can be obtained
by rescaling the measured Compton rate, yielding for
silicon
e=m2
Rγ ≲ 0.2 Hz=kg;
ð18Þ
which is close to the observed excess rate. However, the
maximum elastic recoil energy for Eγ ¼ 1 MeV in Si is
77 eV, so to explain the excess with elastic scattering, all of
the photons contributing to the Compton rate must have
energies around or above 1 MeV, and the photon spectrum
must be a power law with the correct index. Including
inelastic structure factors will not improve the situation. In
order to take advantage of the large number of photons at
low energies, we would need SðqÞ ≪ 1, which would
suppress the total rate well below that of the observed
the
excess. Furthermore,
EDELWEISS-Surf detector is actually lower than in the
SuperCDMS CPD data, indicating that the excess rates
scale inversely to the ratio of radiogenic ionization back-
grounds; this fact additionally disfavors a traditional radio-
genic origin of these rates.
the flat background rate of
An alternative possibility is that secondary interactions in
material surrounding the detector may contribute to this low
energy background, such as Cherenkov emission, decay of
metastable states, or thermal events coupling into the
detector via clamps. In the case of Cherenkov emission,
this possibility was excluded in the analysis presented in
Ref. [12]. For the second case, we would expect a Poisson
distribution of events in energy, which does not resemble the
power law we have shown in this work; even a Poisson
distribution with small mean would resemble an exponential
with common decay constant between both detectors, which
as argued in Sec. II is inconsistent with the distinct energy
regimes of the two excesses. For both Cherenkov emission
and metastable states, events would have to be modeled on a
case-by-case basis as in Ref. [12], which involves enough
free parameters that any analysis is fundamentally under-
determined, and thus while it is possible to create a power
law in a limited regime, this explanation would be demon-
strative but not fundamental proof of this mechanism. The
thermal coupling scenario is largely ruled out for athermal
detectors such as SuperCDMS CPD, as thermal events in
surrounding materials can usually be rejected by pulse shape
discrimination [41].
In summary, all NR explanations for the measured
excesses from fluxes of known particles seem rather
implausible, and indeed all could easily be falsified with
additional shielding in future runs of the experiments.
However,
this analysis also reveals the importance of
developing new, lower energy neutron calibration methods
low-q0
which could be used to probe the low-energy,
kinematic regimes considered in this analysis. The methods
used in this paper are useful for excluding possible event
origins based on allowable structure factors, but we stress
the importance of measuring the features of this inelastic
scattering regime.
V. CONCLUSIONS
recoil.
In particular,
We have demonstrated that the SuperCDMS CPD and
EDELWEISS-Surf excess rates can be modeled by a
common power law of index α ≈ 3.4. Using a novel
approach to quantitatively parametrize a physically allow-
able dynamic structure factor which could yield such a
power-law spectrum in the uncalibrated kinematic regime
where inelastic effects may be expected, we argue that these
two excess rates cannot be explained by a common origin
involving inelastic nuclear
the
SuperCDMS CPD silicon data excludes the DM explan-
ation for the EDELWEISS-Surf germanium excess, but
itself could still be consistent with DM of mass ≲200 MeV
scattering through a highly inelastic, novel nuclear recoil
channel. Moreover, the rates from both of these experi-
ments are too high to be explained by nuclear scattering
from any standard backgrounds, including neutrons, solar
neutrinos, or photons. We thus conclude that these excesses
are likely not due to a novel inelastic scattering process as
originally proposed in Ref.
[13], which bolsters the
evidence for detector effects as a likely origin. That said,
our analysis demonstrates the value of exploring compat-
ibility between low-energy experimental excess rates in
widely varying detector environments, which can be a
powerful tool for disentangling complicated new physics at
these energies.
123002-8
REVISITING THE DARK MATTER INTERPRETATION OF …
PHYS. REV. D 105, 123002 (2022)
ACKNOWLEDGMENTS
None of the observations in this paper would be possible
without the experimental results we cite, but also without
private conversations with the collaborations responsible.
We thus want to acknowledge (in alphabetical order) Ray
Bunker, Alvaro Chavarria, Enectali Figueroa-Feliciano,
Lauren Hsu, Paolo Privitera, Florian Reindl, and Belina
von Krosigk. We thank Gordon Baym, Dan Hooper, Rocky
Kolb, and Ben Safdi for useful conversations related to the
content of this paper. We are especially grateful to Julian
Billiard, Juan Collar, Rouven Essig, Juan Estrada, Jules
Gascon, Matt Pyle, Alan Robinson, and Felix Wagner for
their feedback on early drafts of this analysis. We thank the
Gordon and Betty Moore Foundation and the American
Physical Society for the support of the “New Directions in
Light Dark Matter” workshop where the key idea for this
work was conceived. We thank the EXCESS workshop
organizers and participants for continuing the discussion of
these experimental excess rates. P. A. acknowledges sup-
port from the Gordon and Betty Moore Foundation through
is
EPiQS Grant No. GBMF-9452. The work of Y. K.
supported in part by U.S. Department of Energy Grant
No. DE-SC0015655. L. K. W. assisted in the theoretical
modeling and editing of the paper, and was supported by
the U.S. Department of Energy, Office of Science, Office of
Basic Energy Sciences, Computational Materials Sciences
program under Award No. DE-SC0020177. Fermilab is
operated by Fermi Research Alliance, LLC, under Contract
No. DE-AC02-07CH11359 with the U.S. Department of
Energy. This material is based upon work supported by the
U.S. Department of Energy, Office of Science, National
Quantum Information Science Research Centers, Quantum
Science Center. This work was supported in part by the
Kavli Institute for Cosmological Physics at the University
of Chicago through an endowment
from the Kavli
Foundation and its founder Fred Kavli.
[1] EXCESS Workshop 2022: Descriptions of Rising Low-
Energy Spectra, edited by A. Fuss, M. Kaznacheeva, F.
Reindl, and F. Wagner, arXiv:2202.05097.
[2] M. Crisler, R. Essig, J. Estrada, G. Fernandez, J. Tiffenberg,
M. Sofo haro, T. Volansky, and T.-T. Yu (SENSEI Col-
laboration), Phys. Rev. Lett. 121, 061803 (2018).
[3] O. Abramoff, L. Barak, I. M. Bloch, L. Chaplinsky, M.
Crisler, Dawa, A. Drlica-Wagner, R. Essig, J. Estrada, E.
Etzion et al., Phys. Rev. Lett. 122, 161801 (2019).
[4] L. Barak et al., Phys. Rev. Lett. 125, 171802 (2020).
[5] R. Agnese, T. Aralis, T. Aramaki, I. Arnquist, E. Azadbakht,
W. Baker, S. Banik, D. Barker, D. Bauer, T. Binder et al.,
Phys. Rev. Lett. 121, 051301 (2018).
[6] N. Castelló-Mor (DAMIC-M Collaboration), Nucl. Instrum.
[16] R. Harnik, R. Plestid, M. Pospelov, and H. Ramani, Phys.
Rev. D 103, 075029 (2021).
[17] D. Tucker-Smith and N. Weiner, Phys. Rev. D 64, 043502
(2001).
[18] Y. Kahn, G. Krnjaic, and B. Mandava, Phys. Rev. Lett. 127,
081804 (2021).
[19] M. Ibe, W. Nakano, Y. Shoji, and K. Suzuki, J. High Energy
Phys. 03 (2018) 194.
[20] M. J. Dolan, F. Kahlhoefer, and C. McCabe, Phys. Rev. Lett.
121, 101801 (2018).
[21] N. F. Bell, J. B. Dent, J. L. Newstead, S. Sabharwal, and T. J.
Weiler, Phys. Rev. D 101, 015012 (2020).
[22] R. Essig, J. Pradler, M. Sholapurkar, and T.-T. Yu, Phys.
Rev. Lett. 124, 021801 (2020).
Methods Phys. Res., Sect. A 958, 162933 (2020).
[23] C. P. Liu, C.-P. Wu, H.-C. Chi, and J.-W. Chen, Phys. Rev. D
[7] Q. Arnaud et al. (EDELWEISS Collaboration), Phys. Rev.
102, 121303 (2020).
Lett. 125, 141301 (2020).
[8] I. Alkhatib et al., Phys. Rev. Lett. 127, 061801 (2021).
[9] E. Armengaud, C. Augier, A. Benoît, A. Benoit, L. Berg´e, J.
Billard, A. Broniatowski, P. Camus, A. Cazes, M. Chapellier
et al., Phys. Rev. D 99, 082003 (2019).
[10] A. Abdelhameed, G. Angloher, P. Bauer, A. Bento, E.
Bertoldo, C. Bucci, L. Canonica, A. D’Addabbo, X. Defay,
S. Di Lorenzo et al., Phys. Rev. D 100, 102002 (2019).
[11] L. Barak et al., Phys. Rev. Applied 17, 014022 (2022).
[12] P. Du, D. Egana-Ugrinovic, R. Essig, and M. Sholapurkar,
Phys. Rev. X 12, 011009 (2022).
[13] N. Kurinsky, D. Baxter, Y. Kahn, and G. Krnjaic,
companion paper, Phys. Rev. D 102, 015017 (2020).
[14] J. Kozaczuk and T. Lin, Phys. Rev. D 101, 123012
(2020).
[15] S. Knapen, J. Kozaczuk, and T. Lin, Phys. Rev. Lett. 127,
081805 (2021).
[24] A. Aguilar-Arevalo et al. (DAMIC Collaboration), Phys.
Rev. Lett. 123, 181802 (2019).
[25] D. W. Amaral et al. (SuperCDMS Collaboration), Phys.
Rev. D 102, 091101 (2020).
[26] M. Jiang, H. Xiao, S. Peng, G. Yang, Z. Liu, L. Qiao, and X.
Zu, Nanoscale Res. Lett. 13, 133 (2018).
[27] K. Ramanathan et al., Phys. Rev. D 96, 042002 (2017).
[28] J. Kozaczuk and T. Lin, Phys. Rev. D 101, 123012 (2020).
[29] M. Heikinheimo, S. Sassi, K. Nordlund, K. Tuominen, and
N. Mirabolfathi, arXiv:2112.14495.
[30] A. Aguilar-Arevalo et al. (DAMIC Collaboration), Phys.
Rev. Lett. 125, 241803 (2020).
[31] T. Trickle, Z. Zhang, K. M. Zurek, K. Inzani, and S. Griffin,
J. High Energy Phys. 03 (2020) 036.
[32] J. D. Lewin and P. F. Smith, Astropart. Phys. 6, 87 (1996).
[33] G. Angloher et al. (CRESST Collaboration), Eur. Phys. J. C
77, 637 (2017).
123002-9
PETER ABBAMONTE et al.
PHYS. REV. D 105, 123002 (2022)
[34] J. I. Collar, Phys. Rev. D 98, 023005 (2018).
[35] R. Donnelly, J. Donnelly, and R. Hills, J. Low Temp. Phys.
[38] T. Sato and K. Niita, Radiat. Res. 166, 544 (2006).
[39] K. V. Berghaus, R. Essig, Y. Hochberg, Y. Shoji, and M.
44, 471 (1981).
[36] J. van Wezel, R. Schuster, A. König, M. Knupfer, J. van den
Brink, H. Berger, and B. Büchner, Phys. Rev. Lett. 107,
176404 (2011).
[37] M. Gordon, P. Goldhagen, K. Rodbell, T. Zabel, H. Tang, J.
Clem, and P. Bailey, IEEE Trans. Nucl. Sci. 51, 3427 (2005).
Sholapurkar, arXiv:2112.09702.
[40] D. Baxter et al., Eur. Phys. J. C 81, 907 (2021).
[41] H. L. Kim, H. J. Kim, I. Kim, S. R. Kim, Y. D. Kim, Y. H.
Kim, D. H. Kwon, J. A. Jeon, M. H. Lee, M. K. Lee, and
J. H. So, J. Low Temp. Phys. 199, 1082 (2020).
123002-10
| null |
10.1371_journal.pone.0280975.pdf
|
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
|
All relevant data are within the paper and its Supporting Information files.
|
RESEARCH ARTICLE
Biochemical characterization of a GDP-
mannose transporter from Chaetomium
thermophilum
Gowtham Thambra Rajan PremageethaID
Sucharita BoseID
Samuel GrandfieldID
Subramanian RamaswamyID
4, Vinod NayakID
1,2*
2, Lavanyaa Manjunath2, Deepthi Joseph2, Aviv PazID
4,
2, Luis M. Bredeston5, Jeff Abramson4,
1,2,3, KanagaVijayan Dhanabalan1,2,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 Biological Sciences, Purdue University, West Lafayette, Indiana, United States of America, 2 Institute for
Stem Cell Science and Regenerative Medicine, Bengaluru, Karnataka, India, 3 Manipal Academy of Higher
Education, Manipal, Karnataka, India, 4 Department of Physiology, David Geffen School of Medicine at
UCLA, Los Angeles, CA, United States of America, 5 Departamento de Quı´mica Biolo´gica-IQUIFIB, Facultad
de Farmacia y Bioquı´mica, Universidad de Buenos Aires-CONICET, Ciudad Auto´ noma de Buenos Aires,
Junı´n, Argentina
OPEN ACCESS
Citation: Premageetha GTR, Dhanabalan K, Bose
S, Manjunath L, Joseph D, Paz A, et al. (2023)
Biochemical characterization of a GDP-mannose
transporter from Chaetomium thermophilum.
PLoS ONE 18(4): e0280975. https://doi.org/
10.1371/journal.pone.0280975
Editor: Michael Massiah, George Washington
University, UNITED STATES
Received: January 11, 2023
Accepted: April 4, 2023
Published: April 20, 2023
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0280975
Copyright: © 2023 Premageetha et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
* [email protected]
Abstract
Nucleotide Sugar Transporters (NSTs) belong to the SLC35 family (human solute carrier) of
membrane transport proteins and are crucial components of the glycosylation machinery.
NSTs are localized in the ER and Golgi apparatus membranes, where they accumulate
nucleotide sugars from the cytosol for subsequent polysaccharide biosynthesis. Loss of
NST function impacts the glycosylation of cell surface molecules. Mutations in NSTs cause
several developmental disorders, immune disorders, and increased susceptibility to infec-
tion. Atomic resolution structures of three NSTs have provided a blueprint for a detailed
molecular interpretation of their biochemical properties. In this work, we have identified,
cloned, and expressed 18 members of the SLC35 family from various eukaryotic organisms
in Saccharomyces cerevisiae. Out of 18 clones, we determined Vrg4 from Chaetomium
thermophilum (CtVrg4) is a GDP-mannose transporter with an enhanced melting point tem-
perature (Tm) of 56.9˚C, which increases with the addition of substrates, GMP and GDP-
mannose. In addition, we report—for the first time—that the CtVrg4 shows an affinity to bind
to phosphatidylinositol lipids.
Introduction
Glycosylation is the process that adds glycans to lipids and proteins. Most of these glycosyla-
tion reactions occur in the lumen of the endoplasmic reticulum (ER) and Golgi compartments.
The building blocks for glycan biosynthesis are nucleotide sugars (NS), which function as sub-
strates for glycosyltransferases to append sugar residues onto glycoproteins or glycolipids. In
general, NS are synthesized in the cytosol, except CMP-sialic acid [1], and transported across
the ER/Golgi membrane by Nucleotide Sugar Transporters (NST). NSTs function as
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
1 / 15
PLOS ONE Funding: RS thanks support from DBT-B-life grant,
Grant/Award Number: BT/PR5081/INF/156/2012,
DBT-Indo Swedish Grant, Grant/Award Number:
BT/IN/SWEDEN/06/SR/2017-18, ESRF Access
Program of RCB, Grant/Award Number: BT/INF/22/
SP22660/2017. AP and JA were supported by
grant 5R35GM135175-03 from the National
Institute of General Medical Sciences. Scientific-
Technological Cooperation Program MINCyT-
Argentina and DST-India (Grant Award Number IN/
14/09 to LMB and RS). RS thanks support from
SERB, India for Grant/Award Number EMR/2016/
001825. KV thanks support from SERB, India, for a
National post-doctoral fellowship. The funders had
no role in study design, data collection, analysis,
publication decision, or manuscript preparation.
The authors declare that they do not have any
competing interests.
Competing interests: The authors have declared
that no competing interests exist.
GDP-mannose transporter from C. thermophilum
antiporters where they transport nucleotide sugars into the lumen of ER/Golgi in exchange for
nucleoside mono/di-phosphate (NMP/NDP) back to the cytosol for regeneration [2].
NSTs belong to the solute carrier SLC35 family of membrane transporters. This family is
subdivided into seven subfamilies (SLC35A−G) that are further delineated by the specificity of
the sugars they transport [3]. Humans have nine sugars—glucose (Glc), galactose (Gal), N-ace-
tyl glucose (GlcNAc), N-acetyl galactose (GalNAc), glucuronic acid (GlcA), xylose (Xyl), man-
nose (Man), and fucose (Fuc)—conjugated to either GDP or UDP nucleotides. CMP-sialic
acid is the lone monosaccharide available as a nucleotide monophosphate [4]. Although GDP-
mannose is a naturally occurring NS in humans, no NST that transports it is found. Hence, it
provides a unique opportunity to target GDP-mannose transporters for fighting fungal infec-
tions in humans where mannose is the most abundant sugar of the fungal cell wall, which
directly supports the integrity of the cellSince NSTs serve as the primary transporters of NS,
their loss of function has several consequences for human health and disease, resulting in Con-
genital Disorders of Glycosylation (CDG). Two well-documented autosomal recessive disor-
ders linked to NSTs are leukocyte adhesion deficiency syndrome II [5] and Schneckenbecken
dysplasia [6, 7], which results from a loss of function in the GDP-fucose (SLC35A1) and UDP-
sugar transporters (SLC35D1) respectively. Additionally, NSTs have been linked to develop-
mental disorders in invertebrates [8, 9] and pathogenicity and survival of lower eukaryotes
[10]. Thus, a detailed structure and functional analysis are required.
After more than four decades of research on NSTs, the atomic resolution structure of the
GDP-mannose transporter from Saccharomyces cerevisiae was determined in 2017 [11, 12].
More recently, NST structures of the maize CMP-sialic acid transporter [13] and the mouse
CMP-Sialic acid transporter [14] have been determined. Despite their functional and sequence
disparity, these NST structures reveal some common salient features. The crystal structures
reveal that NSTs comprise ten transmembrane (TM) alpha helices where TM 1−5 is related to
TM 6−10 via a pseudo twofold axis. As seen with many transporters, the inverted repeats share
structural homology with little to no sequence similarities. To date, all structures of NSTs
reside in an outward facing conformation (i.e., opening to lumen) where both substrates
(nucleotide sugars and the corresponding NMP) bind to NST in a similar manner.
Lipids play a crucial role in altering NST’s function, stability, conformation dynamics, and
oligomeric state [15], yet no lipid-binding site(s) have been structurally resolved. A key aspect
of NST’s function is its interactions with lipids. Vrg4 from S. cerevisiae prefers short-chain lip-
ids for its function [11]. Despite the similar structural architecture of the NSTs, it remains
unclear the role lipids play in augmenting NSTs function and how minor changes in amino
acid sequences correspond to sugar specificity. In this manuscript, we characterize a GDP-
mannose transporter from Chaetomium thermophilum (CtVrg4) and screened several lipids
for specific protein-lipids interaction. We show CtVrg4 prefers phosphatidylinositol species
such as phosphatidylinositol-(3)-phosphate (PI3P), phosphatidylinositol-(4)-phosphate
(PI4P), and phosphatidylinositol-(5)-phosphate (PI5P).
Results
NSTs have proven to be difficult to express, purify, and structurally resolve. To overcome these
limitations, we adopted the ‘funnel approach’ initially proposed by Lewinson et al., 2008 [16].
We screened 18 NSTs, from different organisms, in an effort to find the ones more amenable
to crystallization and characterization. Of these 18 constructs, we selected Vrg4 from C. ther-
mophilum based on its expression, detergent extraction, and stability (S1 Fig and S1 Table) as a
crystallization target. During the course of our work, the structure of the GDP-mannose trans-
porter from S. cerevisiae (ScVRG4) was determined, causing us to refocus our priorities toward
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
2 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
Fig 1. Sequence alignment of C. thermophilum CtVrg4, S. cerevisiae ScVrg4 and other GDP-mannose transporter
homologs. Clustal Omega [17] was used for multiple sequence alignment of selected GDP-mannose transporters.
Identical residues are highlighted in red, and highly conserved residues (>0.7) are highlighted in blue boxes. The
nucleotide-binding and sugar recognition motifs are highlighted with star and closed circles, respectively, at the
bottom of the alignment. The positions of the transmembrane domains are indicated and colored as blue and red,
corresponding to ScVrg4 and CtVrg4 structures, respectively.
https://doi.org/10.1371/journal.pone.0280975.g001
a detailed biochemical characterization of CtVrg4 [11]. The amino acid sequences of CtVrg4
and ScVrg4 are 53.6% identical (S2 Table), and both have the characteristic FYNN and
GALNK GDP-mannose binding motifs (Fig 1). Due to these similarities, we generated a
homology model for identifying structural components of CtVrg4 function.
Chymotrypsin-cleaved CtVrg4 (cCtVrg4)
Initial crystallization trials were performed using a mosquito crystallization robot where puri-
fied CtVrg4 was equally mixed with 576 commercial crystallization screening conditions. Ini-
tial crystals were identified in 5 conditions and optimized by varying the contents of the
crystallization components. Of these 5 conditions, the optimized condition of 0.07M sodium
citrate pH 4.8, 75mM sodium fluoride, and 25% PEG300 yielded crystal that diffracted to 3.8Å.
Unfortunately, these crystals proved difficult to reproduce and took almost a month for crystal
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
3 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
Fig 2. Chymotrypsin cleaved CtVrg4 purification, crystallization, and diffraction pattern. (a) Size exclusion profile
of chymotrypsin cleaved CtVrg4 compared with full-length CtVrg4 protein. Inset shows SDS-PAGE of chymotrypsin
cleaved CtVrg4 (lane 1) and full-length protein (lane 3), markers in lane 4. (b) CtVrg4 crystal indicated by the arrow.
(c) Diffraction pattern of CtVrg4 crystal.
https://doi.org/10.1371/journal.pone.0280975.g002
formation leading to difficulties in obtaining more detailed characterization of CtVrg4
crystals.
Further analysis of the CtVrg4 crystals by SDS-PAGE showed a significantly reduced
molecular weight (~25 kDa) when compared to full-length purified protein, which has a
molecular weight of 37 kDa (Fig 2A, gel insert) We speculated that CtVrg4 suffered from pro-
teolysis and attempted to mimic this modification through incubation with chymotrypsin
(cCtVrg4). After chymotrypsin induced proteolysis, cCtVrg4 elutes as a monodisperse peak at
75.44mL from the size exclusion column, where the full-length protein elutes at 74ml (Fig 2A).
To test the functionality of cCtVrg4, we reconstituted the protein into liposomes made of
Yeast Polar Lipid (YPL) and carried out transport assay. The proteoliposome transport assay
showed that the cleaved protein is functional with a Km value of 32.07μM for GDP-mannose
(S2 Fig). Additionally, cCtVrg4 crystalized reproducibly but did not diffract to high resolution
for structure determination (Fig 2B and 2C).
AlphaFold2 model of CtVrg4
We used AlphaFold2 to generate a model of the CtVrg4 [18] to aid in biochemical interpreta-
tion and to better assist structural comparisons with known NST structures [11, 13, 14]. The
structural features are very similar and the confidence values of the prediction in the conserved
regions are very high (S3 Fig).
The homology model shows the anticipated ten transmembrane helical structure corre-
sponding to the NSTs fold. Additionally, the model predicted a 34 amino acids long stretch of
disordered region and a short helix at the N-terminal and similarly a 17 amino acids long dis-
ordered region at the C-terminal of CtVrg4. This elongated stretch of disordered region is not
seen in other NSTs (Fig 3A). Superimposition of the Alphafold2 model onto the ScVrg4 apo
structure (PDB-5OGE Chain A) resulted in a good alignment of all the transmembrane helices
with an RMSD value of 1.6Å (299 Cα atom pairs), barring unstructured and short helices at
the terminus (indicated by arrowhead in Fig 3B).
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
4 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
Fig 3. AlphaFold2 model of CtVrg4. (a), AlphaFold model of CtVrg4 with transmembrane helix colored from blue
(N-terminal) to red (C-terminal) and numbered from 1 to 10. (b), Structural comparison of CtVrg4 model (Cyan) and
ScVrg4 apo crystal structure (Maroon). Arrowhead indicates the unaligned structural element of the AlphaFold2
model with respect to the ScVrg4 structure.
https://doi.org/10.1371/journal.pone.0280975.g003
Complementation assay of CtVrg4
The functionality of CtVrg4 was assessed by complementation assay using a hygromycin B-
sensitive yeast strain, NDY5, which lacks the Vrg4-2 gene [19]. Yeast that lacks or with an
inhibited Vrg4 gene show defects in the glycosylation and the outer cell membrane becoming
sensitive to hygromycin. This assay is a good proxy for the measurement of integrity of glycan
structure. The assay shows that both CtVrg4 and ScVrg4 rescue NDY5 in hygromycin
(Fig 4A). Additionally, based on the AlphaFold2 model CtVrg4Δ17(1−368 amino acids)—a
truncation of 17 amino acids from the C-terminal end, which is a disordered loop region
flanking the transmembrane helix 10 is able to rescue NDY5 in hygromycin. This indicates
that the unstructured c-terminus is not necessary for function (Fig 4A).
Based on the ScVrg4 crystal structure (PDB-5OGK Chain A) published by Parker and
Newstead [11], four residues—N220 and N221 form hydrogen bonds with the guanine moiety
and Y28 and Y281 coordinate the ribose sugar (Fig 4B)—were identified for further functional
characterization. In CtVrg4, alanine substitution of these amino acids shows either no (Y54A,
Y310A) or only partial rescue (N245A, N246A) in the NDY5 assay (Fig 4A). More conservative
substitution of Y310F and N246S also had limited ability for rescue. This complementation
assay result agrees well with the in vitro transport assays of ScVrg4 mutants [11, 12] and estab-
lishes CtVrg4 as a GDP-mannose transporter.
Transport kinetics of CtVrg4
Based on the ScVrg4 structure, the hydroxyl moiety of Y310 coordinates the ribose component
of NS. To resolve the significance of the hydroxyl moiety in GDP-mannose binding and trans-
port kinetics, we generated a protein with Y310F mutation. Proteins with CtVrg4 WT and
Y310F mutation were reconstituted into the liposomes. The GDP-mannose IC50 for WT
CtVrg4 is 25.45μM and 47.05μM for Y310F (Fig 5A).
To further characterize CtVrg4 WT and Y310F mutant, a thermal shift assay was performed
to determine the dissociation constant (Kd) in the presence of their substrates, GMP and
GDP-mannose [20]. The Kd values in Y310F mutant are similar for both substrates yielding a
Kd of 139.2μM for GMP and 129.95μM for GDP-mannose. However, the WT protein has a Kd
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
5 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
Fig 4. Functional Characterization of CtVrg4 and its mutants by a Hygromycin B based in-vivo assay: (a)
Complementation assay of NDY5 by CtVrg4 WT and various functional mutants. Transformed yeast cells were serially
diluted and spotted on synthetic agar media in the presence and absence of 100μg/mL of Hygromycin B. (b), Close-up
view of ScVrg4 crystal structure. Equivalent amino acids that were chosen to be mutated in CtVgr4 are highlighted and
shown in magenta stick color representation. GDP-Mannose is shown in yellow. Transmembrane helixes are
numbered.
https://doi.org/10.1371/journal.pone.0280975.g004
of 143.2μM for GMP but is reduced to 74.26μM for GDP-mannose (Fig 5B). Taken together,
these suggests that the hydroxyl group is not critical for binding or transport of GDP-mannose
by CtVrg4. We discuss later our idea that the importance of Y310 may come from its role in
positioning Y54, which is critical for binding to the ribose sugar.
Lipid binding activity and kinetics of CtVrg4 WT and CtVrg4Δ31 construct
Earlier studies showed that 1,2-dimyristoyl-sn-glycerol-3-phosphocholine (DMPC), is essen-
tial for ScVrg4 function [11]. We therefore screened several lipids for specific interaction with
CtVrg4 wildtype protein using an established lipid blot assay [21]. In short, specific lipids are
immobilized on strips (100 pmol) and subsequently bathed with purified protein to determine
if there are protein/lipid interactions. After the incubation period, the strips are washed to
remove nonspecific binding and the presence of protein is detected using antibody that recog-
nizes the protein’s poly-histidine affinity tag fused to the N-terminus. This assay revealed that
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
6 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
Fig 5. Proteoliposome and thermal shift assay of CtVrg4. (a), Representative GDP-mannose IC50 curve for CtVrg4
WT and CtVrg4 Y310F. IC50 values are shown at the top of the graph. (b), Thermal shift assay for CtVrg4 WT and
CtVrg4 Y310F with varying concentrations of GMP and GDP-mannose. Kd values are shown at the bottom of the
graph. Calculated IC50 and Kd values are the means of two independent biological repeats (each done in technical
duplicate or triplicate), errors are indicated as S.D.
https://doi.org/10.1371/journal.pone.0280975.g005
CtVrg4 specifically binds to three phosphatidylinositol lipids—phosphatidylinositol-(3)-phos-
phate (PI3P), phosphatidylinositol-(4)-phosphate (PI4P), and phosphatidylinositol-(5)-phos-
phate (PI5P) (Fig 6A). These results suggest that CtVrg4 is possibly present in the Golgi
membrane.
MD simulation by Parker et al. 2019 [12] predicted two lipid binding sites in the ScVrg4
protein. The first site is at the dimer interface formed by two transmembrane helices, TM5 and
TM10, and the second is at the shallow groove between TM1, TM9, and TM10. Identifying
TM10 as an integral component of lipid binding, we probed its’ role by deleting the last 31
amino acid segments from the C-terminus, including the predicted Golgi retrieval signal
(K355VRQKA), which leaves most of the TM10 buried in the membrane (S3 Fig). This signal
harbors several positively charged amino acids that could potentially bind to negatively
charged phosphatidylinositol species. The new construct (CtVrg4Δ31) still showed binding for
the same set of lipids as the WT protein, indicating the possibility of an additional lipid bind-
ing site.
This surprising result led to further characterization of the protein with CtVrg4Δ31 muta-
tion. Both the proteoliposome and thermal shift assays for the truncated protein construct
showed similar kinetics as the WT protein (Fig 6B and 6C). GDP-mannose IC50 was found to
be 48μM, and the Kd for GMP and GDP-mannose was 63.2μM and 58.5μM, respectively.
These results suggest that the C-terminal 31 residues are not critical for transport.
Discussion
Our efforts to study the structure-function relationship of NST led to the identification of a
thermostable GDP-mannose transporter, CtVrg4. CtVrg4 shares 53.6% sequence identity with
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
7 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
Fig 6. Lipid binding activity and kinetics of CtVrg4Δ31. (a), Lipid blot. Left–lipid binding activity of CtVrg4 WT
and CtVrg4Δ31 protein. Right–legend with lipids present in the corresponding PIP strip spots. (b), Representative
GDP-mannose IC50 curve for CtVrg4Δ31. The IC50 value is shown at the top of the graph. (c), Thermal shift assay for
CtVrg4Δ31 with varying concentrations of GMP and GDP-mannose. Kd values are shown at the bottom of the graph.
Calculated IC50 and Kd values are mean of two independent biological repeats (each done in technical duplicate or
triplicate), errors are indicated as S.D. Lipid binding assay was done in at least two biological replicates.
https://doi.org/10.1371/journal.pone.0280975.g006
ScVrg4 and maintains the conserved substrate binding motifs, FYNN and GALNK, indicative
that CtVrg4 is a GDP-mannose transporter as well. We further confirmed that CtVrg4 is a
GDP-mannose transporter utilizing a hygromycin-based complementation assay and substan-
tiated our finding by in vitro proteoliposome transport assays. The IC50 value of GDP-man-
nose was found to be 25.45μM, which is approximately three times higher than that for ScVrg4
(7.7μM). Our melting point assay in DDM showed CtVgr4 is significantly more stable
(56.9˚C) (Fig 7A) than ScVrg4 (37.9˚C). This is not surprising as CtVrg4 is isolated from a
thermostable fungus.
During crystallization trials, we found that the chymotrypsin cleaved CtVrg4 produced bet-
ter diffraction quality crystals than the full-length protein. Proteoliposome transport assay of
chymotrypsin cleaved CtVrg4 showed that the cleaved protein is functional with a GDP-man-
nose Km value of 32.07 μM, suggesting that limited proteolysis does not affect the core of the
protein. The observation that a shorter construct is still functional suggests it might be worth
reattempting the structure determination, which we did not pursue further after the structure
of ScVrg4 was published. Molecular replacement with the 3.8 Å data did not produce good
solutions. Given the similarity in sequence between CtVrg4 and ScVrg4 (53%), and the mod-
ern structure prediction tools like AlphaFold2, the predicted structure would be a good model
(compared to a 3.8 Å structure).
In ScVrg4, Y281 forms a hydrogen bond with the ribose sugar of GDP-mannose and a π-
stacking interaction with Y28, which along with S269, further coordinates the ribose sugar.
The alanine mutants Y28A, and Y281A of ScVrg4, resulted in the complete abolishment of
transport activity [11]. In our complementation assay, we found that Y310F (equivalent to
Y281 in ScVrg4) conferred a partial rescue, whereas no complementation was observed in the
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
8 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
Fig 7. Thermal stability analysis of CtVrg4. (a), Bar graph of melting temperature (Tm) of CtVrg4 WT, Y310F, and
CtVrg4Δ31. (b), Kd values of GMP/GDP-mannose binding to CtVrg4 WT, Y310F, and CtVrg4Δ31 were estimated by a
thermal shift assay. (c), Electrostatic surface representation of the AlphaFold2 model for CtVrg4 highlighting positively
charged residues in the Golgi retrieval sequence in TM10. The orientation of CtVrg4 in the lipid bilayer was simulated
through the OPM server [22] (d), GMP/GDP-mannose binding to CtVrg4 was estimated by a shift in Tm after the
addition of 25 mM GMP/GDP-mannose to WT, Y310F, and CtVrg4Δ31.
https://doi.org/10.1371/journal.pone.0280975.g007
case of Y310A, Y54A, and Y54F. The Kd for GDP-mannose almost doubled in the Y310F
mutant compared with WT, but the Kd for GMP remained the same (Fig 7B). The IC50 for
GDP-mannose also doubled compared with the WT, further validating the decreased binding
efficiency of GDP-mannose in the Y310F mutation. These results suggest that Y310 also forms
a hydrogen bond with the ribose sugar and helps position the Y54 residue to bind the ribose
sugar, whose substitution is lethal for transport activity.
It was previously reported that ScVrg4 localizes to the Golgi using its C-terminal Golgi
retrieval sequence, which binds to COPI vesicles [23]. CtVrg4 has a Golgi retrieval sequence
(K355VRQKA) like ScVrg4 in the cytoplasmic end of TM10. The Golgi retrieval sequence is
positioned on membrane boundaries to interact with the membrane lipids (Fig 7C). The Golgi
retrieval sequence is rich in positively charged residues, which could be a potential binding site
for negatively charged PIP lipids to bind. However, our results show that CtVrg4Δ31 (devoid
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
9 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
of the last 31 amino acids from the C-terminal end, including the Golgi retrieval sequence) has
the same binding affinity as the full-length protein—suggesting the Golgi retrieval sequence
may not play any role in PIP binding. Further studies are needed to pinpoint the exact location
of PIP binding to the transporter and its structural or functional effect on CtVrg4.
We report that CtVrg4Δ31 is less thermostable than the WT and Y310F mutant with a
3–4˚C lower melting temperature (Fig 7A). Interestingly, both GMP and GDP-mannose bind
to CtVrg4Δ31 and show similar thermal shift values compared with WT and Y310F (Fig 7D).
This result suggests that although overall stability is affected by truncation, the ability of the
substrate to bind and stabilize the core protein is not affected. Our work reported here suggests
that further structure function studies of NSTs are needed in order to understand the mecha-
nism of transport and the role of lipids in localization, stability and NS transport.
Materials and methods
Cloning and expression of CtVrg4 wildtype, mutants, and truncated
constructs in a yeast expression system
The sequence of CtVrg4 from C. thermophilum (NCBI GenBank XM_006692792.1) was iden-
tified through a homology search against the ScVrg4 sequence. CtVrg4 is 385 amino acids
long. The gene was cloned and expressed in a modified pDDGFP yeast expression vector with
an N-terminal 12 histidine tag followed by a SmaI restriction site for homologous recombina-
tion-based gene insertion and a stop codon.
All CtVrg4 mutants were generated via site-directed mutagenesis using the PCR method
and further confirmed by sequencing. Two C-terminal truncated versions of CtVrg4 were con-
structed, CtVrg4Δ17 (constituting amino acids from 1 to 368) and CtVrg4Δ31 (comprising
amino acids from 1 to 354). The cloned genes were expressed in S. cerevisiae haploid strain
FGY217 (MATa, ura3-52, lys2_201, and pep4).
Expression and purification of CtVrg4
The primary cultures of CtVrg4 WT, mutations, and truncations were grown in synthetic
media without uracil in 2% glucose. The primary culture was diluted into a secondary culture
(1L) to the final OD of 0.2 in 0.1% glucose. The culture was induced with 2% galactose at 0.6 to
0.8 OD to express the protein and further grown for 22–24h before harvesting the cells by cen-
trifugation. Cells were resuspended in membrane resuspension buffer (75mM Tris pH 8.0, 150
mM NaCl, and 5% glycerol) and then lysed using a cell disruptor (Constant Systems Ltd) at 28,
32, 36, and 39 kpsi. Membranes were isolated by centrifugation at 200,000 xg for 1.5h. The pro-
tocol for protein purification was adapted from Drew et al., 2008 with modifications [24]. The
membranes were solubilized in dodecyl β-D-maltopyranoside (DDM, Anatrace) at a 1: 0.2 (w/
w) ratio of the membrane to DDM for 2 h in membrane resuspension buffer along with a pro-
tease inhibitor cocktail. For crystallization, the protein was solubilized in 2% Decyl β-D-malto-
pyranoside (DM) for two hours in the membrane resuspension buffer. The solubilized
membrane was centrifuged at 200,000 xg for 30 minutes, and the supernatant was loaded onto
a 5 mL His-Trap FF/HP column (Cytiva). The protein was eluted with 300 mM imidazole. The
eluted fractions were desalted using a Hiprep 26/10 desalting column (Cytiva) to remove the
imidazole from the buffer. The protein was concentrated using a 50 kDa cut-off Amicon device
(Millipore). The concentrated protein was centrifuged at 14,000 xg for 15 mins before injecting
onto a Superdex 200 (10/300) size-exclusion column (Cytiva) pre-equilibrated with liposome
assay buffer (20 mM HEPES pH 7.4, 50 mM KCl, and 2 mM MgSO4) containing 0.1 mM
EDTA and 0.013% DDM.
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
10 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
Chymotrypsin cleavage and crystallization
Purified CtVrg4 was subjected to limited proteolysis using chymotrypsin (50:1 ratio) for 2h at
18˚C. The reaction was stopped with 0.1mM phenylmethylsulfonyl fluoride followed by ultra-
centrifugation at 150,000 xg for 30 minutes. The chymotrypsin cleaved protein eluted at
75.44ml compared to 74ml for the full-length protein on a Superdex 200 (16/600) column
(Cytiva). The cleaved protein’s peak fractions were pooled and concentrated using 50kDa cut-
off Amicon (Millipore).
Purified chymotrypsin cleaved CtVrg4 (12mg/ml) was mixed in a 4:1 (protein/bicelle) ratio
with a 25% (2.8:1) DTPC/CHAPSO bicellar solution for 45 minutes on ice, yielding 9.6 mg/ml
of chymotrypsin cleaved CtVrg4 in 5% bicelles. The best crystals grew to a size of 0.07 mm X
0.07 mm X 0.02 mm at 18˚C in 0.07 M sodium citrate pH 4.8, 75 mM sodium fluoride, and
25% PEG300. The crystals were cryoprotected with 30% PEG400 before flash-cooling in liquid
nitrogen. X-ray diffraction data were collected at the PROXIMA-1 beamline, SOLEIL synchro-
tron source (France), at a wavelength of 0.97857 Å using a 100 K nitrogen stream. The best
crystal diffracted to 3.8 Å resolution.
Protein reconstitution into liposomes
Chymotrypsin cleaved CtVrg4 was reconstituted into yeast polar lipid (YPL, Avanti Polar Lip-
ids) liposomes in 20 mM HEPES pH 7.5, 100 mM KCl, as described by Parker et al., 2017 [11].
CtVrg4 wildtype/Y310F/truncates were reconstituted into liposomes using a modified proto-
col from Majumdar et al., 2019 [25]. Briefly, YPL were suspended in chloroform, dried using a
nitrogen stream, and left in a vacuum desiccator overnight. The lipid film was resuspended at
10 mg/mL in liposome assay buffer, and then bath sonicated to form small multilamellar vesi-
cles. Vesicles were extruded for 10 cycles through 400 nm polycarbonate membranes (Avanti
Polar Lipids). For reconstitution, purified (CtVrg4 wildtype/Y310F/truncate) protein in DDM
(at 8 to 16 mg/mL) was added to the extruded YPL at a final lipid: protein ratio (w/w) of 80:1.
Sodium cholate (Anatrace) was added at a concentration of 0.65−0.75% to this mixture and
incubated for one hour at room temperature, then for a further 30 minutes on ice. As a control,
liposomes without protein were resuspended in the assay buffer containing 0.013% of DDM.
The protein-lipid mixture was passed through an 8.3 mL PD10 column (Cytiva) pre-equili-
brated with 0.5 mg YPL in liposome assay buffer. A fraction volume of 2.8 mL was collected
after the column’s void volume (2.6 mL) and centrifuged at 150,000 xg for 30 minutes. Subse-
quently, the pellet containing proteoliposomes was resuspended with liposome assay buffer,
flash-frozen in liquid N2, and stored at -80˚C. The protein reconstitution into the lipids was
verified by Western blot with anti-His antibody.
Transport assay
CtVrg4 wildtype/Y310F/truncate proteoliposomes were thawed, and the desired concentration
of internal cold substrate (1 mM GDP-mannose) was added and then subjected to six rounds
of freeze-thaw in liquid nitrogen to load the proteoliposomes with the substrate. For chymo-
trypsin-cleaved CtVrg4, the internal concentration of GDP-mannose ranged from 1 to
1000 μM. Unloaded GDP-mannose was removed by ultracentrifugation at 150,000 xg for 30
minutes. The pellets were resuspended in a cold liposome assay buffer. For a typical 50 μl reac-
tion, 10 μl of loaded proteoliposomes containing approximately 4 μg of protein was added to
40 μl of liposome assay buffer containing 0.5 μM [3H]GMP (American Radiolabeled Chemi-
cals, Inc). For the chymotrypsin cleaved protein in proteoliposomes, the concentration of [3H]
GMP was 0.384 μM. The addition of GMP initiated the exchange reaction. For IC50 value
determination, the external competing substrate was varied from 2 μM to 2.5 mM. The
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
11 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
mixture was then incubated at 25˚C for 20 minutes. For chymotrypsin cleaved CtVrg4, the
reaction was performed at room temperature and terminated after 10 min. The uptake of the
radiolabeled substrate was stopped by adding 800 μl of cold water and rapidly filtering onto
0.22 micron mixed cellulose esters filters (Millipore), which were then washed three times with
2 mL of ice-cold water. A liquid scintillation counter (Perkin Elmer) measured the amount of
[3H]GMP transported inside the liposomes. All experiments were done in two biological
repeats, each in technical duplicate or triplicate. Kinetic parameters were calculated by nonlin-
ear fit using the GraphPad Prism software (GraphPad Software, Inc., San Diego, CA, USA).
Hygromycin B-based in vivo assay
For in vivo functional characterization, wildtype/mutant/truncated proteins were expressed in
the yeast NDY5 strain (MAT ura3–52a, leu2–211, vrg4–2), which is a vrg4Δ mutant. The trans-
formants were selected based on the uracil resistance marker. Cells grown overnight were seri-
ally diluted and spotted on the synthetic agar media in the presence and absence of 100 μg/mL
of Hygromycin B. Protein expression was induced with 2% galactose, and phenotype was
observed after three days. For negative control, transformed cells were spotted in 2% glucose
and compared with 2% galactose.
Lipid blot assay
Lipid blot assay was carried out using PIP StripsTM (Echelon Biosciences) following the manu-
facturer’s instructions. Briefly, the strips were blocked overnight with PBST buffer (1x PBS
with 0.05% Tween-20) containing 5% non-fat dry milk at 4˚C with mild rocking. The strips
were then incubated with 50 μg of CtVrg4 WT/CtVrg4Δ31 in PBST for two hours at room
temperature. After three washes with PBST, the strips were incubated with conjugated anti-
polyHistidine−Peroxidase antibody (Sigma A7058, in a 1:2000 ratio) for one hour at room
temperature. After three more washes with PBST, the bound proteins were detected with the
ClarityTM Western ECL kit (Bio-Rad Laboratories).
Thermal shift assay
Following size exclusion chromatography, detergent-solubilized CtVrg4 wildtype/mutant/
truncate (final concentration of 0.5mg/mL) were incubated with 0 mM to 50 mM GMP/GDP-
mannose at room temperature for 15 minutes before measuring the Tm using a Tycho NT.6
instrument (NanoTemper Technologies, Germany). All experiments were done in two biologi-
cal repeats and three technical repeats. Kinetic parameters were calculated by nonlinear fit
using the GraphPad Prism software.
AI prediction and superimposition
The AlphaFold2 structure prediction feature in UCSF ChimeraX was used to predict the struc-
ture of CtVrg4. Superimposition and all protein model figures were generated with UCSF Chi-
meraX [26]. ESPript was used to render sequence similarities and secondary structure
information onto multiple sequence alignments [27].
Supporting information
S1 Fig. A schematic of the funnel approach that was used to screen for crystallizable NSTs.
(TIF)
S2 Fig. Exchange velocities of internal GDP-man with external [3H]GMP in proteolipo-
somes. Values are the means of three independent biological repeats (each done in technical
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
12 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
triplicate). Errors are indicated as SD. Km was calculated by non-linear fit using the GraphPad
Prism software.
(TIF)
S3 Fig. AlphaFold2 model for CtVrg4. Colors represent the predicted confidence value of the
structure. The confidence value decreases as the color changes from red to blue. The figure
shows that the transmembrane regions are predicted with high confidence–in red. The N-ter-
minus region is floppy and is predicted poorly. Note that the C-terminal helix is predicted
with intermediate confidence (green). Amino acid Alanine 27 (A27) and the C-terminal resi-
due (S385) are labeled for reference. A354, is the last residue in the CtVrg4Δ31 construct.
(TIF)
S1 Table. NSTs identified for crystallization through homology search.
(DOCX)
S2 Table. Percentage sequence identity of known and putative GDP-mannose transporters
from different species of fungi—alignment carried out using Clustal Omega.
(DOCX)
S1 Raw image.
(PDF)
Acknowledgments
We thank Dr. Leonard Chavas from SOLEIL Synchrotron (Proxima-1 beamline) for providing
the beamtime for data collection. The Tycho experiment was done in the Chemical Genomics
Facility at Purdue Institute for Drug Discovery and the NIH-funded Indiana Clinical and
Translational Sciences Institute.
Author Contributions
Conceptualization: KanagaVijayan Dhanabalan, Luis M. Bredeston, Jeff Abramson,
Subramanian Ramaswamy.
Data curation: Gowtham Thambra Rajan Premageetha, KanagaVijayan Dhanabalan, Jeff
Abramson.
Formal analysis: Gowtham Thambra Rajan Premageetha, KanagaVijayan Dhanabalan,
Sucharita Bose, Lavanyaa Manjunath, Vinod Nayak.
Funding acquisition: Luis M. Bredeston, Jeff Abramson, Subramanian Ramaswamy.
Investigation: Gowtham Thambra Rajan Premageetha, KanagaVijayan Dhanabalan, Sucharita
Bose, Lavanyaa Manjunath, Deepthi Joseph, Aviv Paz, Samuel Grandfield, Vinod Nayak,
Luis M. Bredeston, Jeff Abramson, Subramanian Ramaswamy.
Methodology: Gowtham Thambra Rajan Premageetha, KanagaVijayan Dhanabalan, Sucharita
Bose, Lavanyaa Manjunath, Vinod Nayak.
Supervision: Luis M. Bredeston, Jeff Abramson, Subramanian Ramaswamy.
Validation: KanagaVijayan Dhanabalan.
Writing – original draft: Gowtham Thambra Rajan Premageetha, KanagaVijayan
Dhanabalan, Jeff Abramson, Subramanian Ramaswamy.
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
13 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
Writing – review & editing: Gowtham Thambra Rajan Premageetha, KanagaVijayan
Dhanabalan, Sucharita Bose, Lavanyaa Manjunath, Deepthi Joseph, Aviv Paz, Vinod
Nayak, Luis M. Bredeston, Jeff Abramson, Subramanian Ramaswamy.
References
1. Coates SW, Gurney T, Sommers LW, Yeh M, Hirschberg CB. Subcellular localization of sugar nucleo-
tide synthetases. J Biol Chem. 1980; 255: 9225–9229. PMID: 6251080
2. Hadley B, Maggioni A, Ashikov A, Day CJ, Haselhorst T, Tiralongo J. Structure and function of nucleo-
tide sugar transporters: Current progress. Computational and Structural Biotechnology Journal. 2014;
10: 23–32. https://doi.org/10.1016/j.csbj.2014.05.003 PMID: 25210595
3. Hadley B, Litfin T, Day CJ, Haselhorst T, Zhou Y, Tiralongo J. Nucleotide Sugar Transporter SLC35
Family Structure and Function. Computational and structural biotechnology journal. 2019; 17: 1123–
1134. https://doi.org/10.1016/j.csbj.2019.08.002 PMID: 31462968
4. Mikkola S. Nucleotide Sugars in Chemistry and Biology. Molecules. 2020; 25: 5755. https://doi.org/10.
3390/molecules25235755 PMID: 33291296
5.
Liu L, Xu YX, Hirschberg CB. The role of nucleotide sugar transporters in development of eukaryotes.
Seminars in Cell and Developmental Biology. 2010; 21: 600–608. https://doi.org/10.1016/j.semcdb.
2010.02.002 PMID: 20144721
6. Hiraoka S, Furuichi T, Nishimura G, Shibata S, Yanagishita M, Rimoin DL, et al. Nucleotide-sugar trans-
porter SLC35D1 is critical to chondroitin sulfate synthesis in cartilage and skeletal development in
mouse and human. Nat Med. 2007; 13: 1363–1367. https://doi.org/10.1038/nm1655 PMID: 17952091
7. Rautengarten C, Quarrell OW, Stals K, Caswell RC, De Franco E, Baple E, et al. A hypomorphic allele
of SLC35D1 results in Schneckenbecken-like dysplasia. Human molecular genetics. 2019 Nov 1; 28
(21):3543–51. https://doi.org/10.1093/hmg/ddz200 PMID: 31423530
8. Cipollo JF, Awad AM, Costello CE, Hirschberg CB. srf-3, a mutant of Caenorhabditis elegans, resistant
to bacterial infection and to biofilm binding, is deficient in glycoconjugates. Journal of Biological Chemis-
try. 2004; 279: 52893–52903. https://doi.org/10.1074/jbc.M409557200 PMID: 15452127
9. Goto S, Taniguchi M, Muraoka M, Toyoda H, Sado Y, Kawakita M, et al. UDP–sugar transporter impli-
cated in glycosylation and processing of Notch. Nat Cell Biol. 2001; 3: 816–822. https://doi.org/10.1038/
ncb0901-816 PMID: 11533661
10. Gaur U, Showalter M, Hickerson S, Dalvi R, Turco SJ, Wilson ME, et al. Leishmania donovani lacking
the Golgi GDP-Man transporter LPG2 exhibit attenuated virulence in mammalian hosts. Experimental
Parasitology. 2009; 122: 182–191. https://doi.org/10.1016/j.exppara.2009.03.014 PMID: 19328787
11. Parker JL, Newstead S. Structural basis of nucleotide sugar transport across the Golgi membrane.
Nature. 2017; 551: 521. https://doi.org/10.1038/nature24464 PMID: 29143814
12. Parker JL, Corey RA, Stansfeld PJ, Newstead S. Structural basis for substrate specificity and regulation
of nucleotide sugar transporters in the lipid bilayer. Nat Commun. 2019; 10: 4657. https://doi.org/10.
1038/s41467-019-12673-w PMID: 31604945
13. Nji E, Gulati A, Qureshi AA, Coincon M, Drew D. Structural basis for the delivery of activated sialic acid
into Golgi for sialyation. Nature Structural and Molecular Biology. 2019; 26: 415–423. https://doi.org/10.
1038/s41594-019-0225-y PMID: 31133698
14. Ahuja S, Whorton MR. Structural basis for mammalian nucleotide sugar transport. eLife. 2019; 8.
https://doi.org/10.7554/eLife.45221 PMID: 30985278
15. Corradi V, Sejdiu BI, Mesa-Galloso H, Abdizadeh H, Noskov SY, Marrink SJ, et al. Emerging Diversity
in Lipid-Protein Interactions. Chem Rev. 2019; 119: 5775–5848. https://doi.org/10.1021/acs.chemrev.
8b00451 PMID: 30758191
16.
Lewinson O, Lee AT, Rees DC. The Funnel Approach to the Precrystallization Production of Membrane
Proteins. Journal of Molecular Biology. 2008; 377: 62–73. https://doi.org/10.1016/j.jmb.2007.12.059
PMID: 18241890
17. Sievers F, Wilm A, Dineen D, Gibson TJ, Karplus K, Li W, et al. Fast, scalable generation of high-quality
protein multiple sequence alignments using Clustal Omega. Molecular systems biology. 2011; 7(1):539.
Tunyasuvunakool K, Adler J, Wu Z, Green T, Zielinski M, Zˇ ı´dek A, et al. Highly accurate protein struc-
ture prediction for the human proteome. Nature. 2021; 596: 590–596. https://doi.org/10.1038/s41586-
021-03828-1 PMID: 34293799
18.
19. Dean N. Yeast glycosylation mutants are sensitive to aminoglycosides. Proc Natl Acad Sci USA. 1995;
92: 1287–1291. https://doi.org/10.1073/pnas.92.5.1287 PMID: 7877969
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
14 / 15
PLOS ONE GDP-mannose transporter from C. thermophilum
20. Vivoli M, Novak HR, Littlechild JA, Harmer NJ. Determination of protein-ligand interactions using differ-
ential scanning fluorimetry. JoVE (Journal of Visualized Experiments). 2014 Sep 13(91):e51809.
https://doi.org/10.3791/51809 PMID: 25285605
21.
22.
Jin W, Ge WP, Xu J, Cao M, Peng L, Yung W, et al. Lipid binding regulates synaptic targeting of PICK1,
AMPA receptor trafficking, and synaptic plasticity. Journal of Neuroscience. 2006 Mar 1; 26(9):2380–
90. https://doi.org/10.1523/JNEUROSCI.3503-05.2006 PMID: 16510715
Lomize MA, Pogozheva ID, Joo H, Mosberg HI, Lomize AL. OPM database and PPM web server:
resources for positioning of proteins in membranes. Nucleic acids research. 2012 Jan 1; 40(D1):D370–
6. https://doi.org/10.1093/nar/gkr703 PMID: 21890895
23. Abe M, Noda Y, Adachi H, Yoda K. Localization of GDP-mannose transporter in the Golgi requires
retrieval to the endoplasmic reticulum depending on its cytoplasmic tail and coatomer. Journal of Cell
Science. 2004; 117: 5687–5696. https://doi.org/10.1242/jcs.01491 PMID: 15494368
24. Drew D, Newstead S, Sonoda Y, Kim H, Von Heijne G, Iwata S. GFP-based optimization scheme for
the overexpression and purification of eukaryotic membrane proteins in Saccharomyces cerevisiae.
Nature protocols. 2008 May; 3(5):784–98. https://doi.org/10.1038/nprot.2008.44 PMID: 18451787
25. Majumder P, Khare S, Athreya A, Hussain N, Gulati A, Penmatsa A. Dissection of Protonation Sites for
Antibacterial Recognition and Transport in QacA, a Multi-Drug Efflux Transporter. Journal of Molecular
Biology. 2019; 431: 2163–2179. https://doi.org/10.1016/j.jmb.2019.03.015 PMID: 30910733
26. Pettersen EF, Goddard TD, Huang CC, Meng EC, Couch GS, Croll TI, et al. UCSF ChimeraX: Structure
visualization for researchers, educators, and developers. Protein Science. 2021 Jan; 30(1):70–82.
https://doi.org/10.1002/pro.3943 PMID: 32881101
27. Robert X, Gouet P. Deciphering key features in protein structures with the new ENDscript server.
Nucleic acids research. 2014 Jul 1; 42(W1):W320–4. https://doi.org/10.1093/nar/gku316 PMID:
24753421
PLOS ONE | https://doi.org/10.1371/journal.pone.0280975 April 20, 2023
15 / 15
PLOS ONE
| null |
10.1186_s12934-021-01548-9.pdf
|
Availability of data and materials
The original contributions presented in the study are publicly available.
The raw sequence data reported in this paper have been deposited in the
Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2017) in
National Genomics Data Center (Nucleic Acids Res 2020), Beijing Institute of
Genomics (China National Center for Bioinformation), Chinese Academy of
Sciences, under accession number CRA003010 that are publicly accessible at
https ://bigd.big.ac.cn/gsa.
|
Availability of data and materials The original contributions presented in the study are publicly available. The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2017) in National Genomics Data Center (Nucleic Acids Res 2020), Beijing Institute of Genomics (China National Center for Bioinformation), Chinese Academy of Sciences, under accession number CRA003010 that are publicly accessible at https ://bigd.big.ac.cn/gsa .
|
Yuan et al. Microb Cell Fact (2021) 20:53
https://doi.org/10.1186/s12934-021-01548-9
Microbial Cell Factories
RESEARCH
Open Access
The role of the gut microbiota
on the metabolic status of obese children
Xin Yuan1, Ruimin Chen1*
, Kenneth L. McCormick2, Ying Zhang1, Xiangquan Lin1 and Xiaohong Yang1
Abstract
Background: The term “metabolically healthy obese (MHO)” denotes a hale and salutary status, yet this connotation
has not been validated in children, and may, in fact, be a misnomer. As pertains to obesity, the gut microbiota has
garnered attention as conceivably a nosogenic or, on the other hand, protective participator.
Objective: This study explored the characteristics of the fecal microbiota of obese Chinese children and adolescents
of disparate metabolic statuses, and the associations between their gut microbiota and circulating proinflamma-
tory factors, such as IL-6, TNF-α, lipopolysaccharide-binding protein (LBP), and a cytokine up-regulator and mediator,
leptin.
Results: Based on weight and metabolic status, the 86 Chinese children (ages 5–15 years) were divided into three
groups: metabolically healthy obese (MHO, n
23), and healthy normal
weight controls (Con, n
21). In the MUO subjects, the phylum Tenericutes, as well as the alpha and beta diversity,
were significantly reduced compared with the controls. Furthermore, Phylum Synergistetes and genus Bacteroides
were more prevalent in the MHO population compared with controls. For the MHO group, Spearman’s correlation
analysis revealed that serum IL-6 positively correlated with genus Paraprevotella, LBP was positively correlated with
genus Roseburia and Faecalibacterium, and negatively correlated with genus Lactobacillus, and leptin correlated posi-
tively with genus Phascolarctobacterium and negatively with genus Dialister (all p < 0.05).
42), metabolic unhealthy obese (MUO, n
=
=
=
Conclusion: Although there are distinct differences in the characteristic gut microbiota of the MUO population
versus MHO, dysbiosis of gut microsystem is already extant in the MHO cohort. The abundance of some metabolism-
related bacteria associates with the degree of circulating inflammatory compounds, suggesting that dysbiosis of gut
microbiota, present in the MHO children, conceivably serves as a compensatory or remedial response to a surfeit of
nutrients.
Keywords: Metabolically healthy obese, Children, 16s rRNA, Gut microbiota
Introduction
The global epidemic of childhood obesity, and the accom-
panying rise in the prevalence of endocrine, metabolic,
and cardiovascular comorbidities, is perhaps the most
impactful and ubiquitous public health disorder of the
modern world [1]. In the context of this pandemic, a
*Correspondence: [email protected]
1 Department of Endocrinology, Fuzhou Children’s Hospital of Fujian
Medical University, NO. 145, 817 Middle Road, Fuzhou 350005, China
Full list of author information is available at the end of the article
distinct group of youth with obesity who are devoid of
metabolic disturbances—so-called “metabolically healthy
obese” (MHO)—have been identified. Obesity notwith-
standing, by definition MHO children retain a favora-
ble metabolic profile, with preserved insulin sensitivity
along with normal blood pressure, glucose homeostasis,
lipids, and liver enzymes. Moreover, their hormonal,
inflammation, and immune profiles are seemingly imper-
vious to obesity [2]. First described in obese adults, the
MHO phenotype has also been extensively studied in
young people with obesity [2]. Arguably, MHO may be
© The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and
the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material
in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material
is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the
permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco
mmons .org/licen ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/publi cdoma in/
zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Yuan et al. Microb Cell Fact (2021) 20:53
Page 2 of 13
a transitional stage to the far more common, more high-
risk, conventional cardio-metabolic obese phenotype.
Regardless of the aforesaid normal biochemical charac-
teristics of MHO, the risk for cardiovascular disease per-
sists since the MHO phenotype may be unstable, thereby
transitory [3, 4].
Among the non-genetic factors associated with obesity,
the gut microbiota has garnered attention as an obesity
regulator given the robust correlations in animal stud-
ies between gut microbiota and body weight. Obese
individuals, whether adults or children, have increased
abundance in Firmicutes in concert with decreased
in Bacteroidetes [5, 6]. The distinctive gut microbiota
prevalent in obese subjects is recognized as promoting
an unhealthy metabolic obese (MUO) phenotype with
attendant comorbidities, such as increased endotoxemia,
intestinal and systemic inflammation, as well as insulin
resistance. An altered gut microbiota has been implicated
in obesity and type 2 diabetes mellitus (T2DM) inso-
far as a decrement in certain species and gene richness
have been linked to adiposity, dyslipidemia, and insulin
resistance [7]. Hence, the clinical repercussions aside, it
is plausible that differences in the gut microbiota could
dictate whether an obese child is metabolically fit (MHO)
or not (MUO) [8, 9].
Obesity and related metabolic disorders are associated
with gut microbiota dysbiosis, disrupted intestinal bar-
rier and chronic inflammation [10]. For instance, obese
Mexican children and adolescents had increased levels of
leptin and C-reactive protein, which were associated with
changes in the gut microbiota [11]. However, the asso-
ciation between gut microbiota and proinflammatory
cytokines, such as IL-6, TNF-α and lipopolysaccharide-
binding protein (LBP), has not been fully investigated in
children of varying metabolic statues. Firstly, this study
examined the metabolic heterogeneity of obese children
as it relates to the composition of the gut microbiota.
And, as a secondary end point, identify metabolic-spe-
cific bacteria which associate with serum inflammatory
factors incriminated in obesity comorbidities.
Results
Study participants
Based on weight status, the metabolically stable cohort
subjects (n = 63) were subdivided as MHO (n = 42) or
Con (n = 21).
The age of the 86 participates ranged from 5.5 to
14.3 years, with a mean of 9.76 ± 1.93 years. There were
65 obese children, of whom 23 were MUO and 42 were
MHO. The BMI of other 21 children were normal. Age,
weight, BMI, BMI-Z, WHtR, SBP, TG and LDL-c in the
MUO group were significantly higher than the Con and
MHO children, and HDL-c in the in the MUO group
were significantly lower than the Con and MHO children
(all p < 0.05, Table 1).
The weight, BMI, BMI-Z, WHR, WHtR, SBP, DBP,
TG, LDL-c, IL-6, TNF-α, LBP and leptin were signifi-
cantly higher in the MHO group than the Con children,
and HDL-c in the MHO group were significantly lower
than the Con group (all p < 0.05). There was no statisti-
cal difference in age, gender, FPG and fasting TC between
MHO and Con (all p > 0.05, Table 1).
Microbiota profiles in different metabolic status subjects
A total of 918,578 sequencing reads were obtained from
86 fecal samples, with an average value of 10,681 counts
per sample. We identified an overall of 146 OTUs, among
which 136 OTU with ≥ 2 counts, and they were grouped
in 9 phylum and 38 families.
Abundance profiling in different metabolic status subjects
Grouping OTUs at phylum level, and applying the
Mann–Whitney U test on the relative abundances of
phyla for the two groups, the relative abundances of phy-
lum Tenericutes was more prevalent in the metabolically
healthy cohorts (Con and MHO children) compared to
the MUO group (p = 0.006, Additional file 1: Table S1
and Fig. 1a).
On OTUs at the genera level, by Mann–Whitney
U-test, including all the genera (merging small taxa with
counts < 10), we identified that genera Anaerostipes, Alis-
tipes, Desulfovibrio, Fusobacterium, Gemmiger, Odori-
bacter, Oscillospira and Parabacteroides were more
prevalent in the metabolically healthy cohorts (Con and
MHO children) versus MUO children, yet the genus
Dorea was more prevalent in MUO (p < 0.05; Fig. 1b,
Table 2).
Alpha‑ and beta‑diversity in different metabolic status
subjects
To assess the overall differences of microbial community
structures in metabolic healthy and MUO subjects, we
measured ecological parameters based on alpha-diver-
sity. The alpha-diversity analysis showed significantly
higher diversity in metabolic healthy subjects (Con and
MHO children) in comparison to MUO participants
(p < 0.05, Fig. 2a, b, Additional file 1: Table S2).
To determine the differences between microbial com-
munity profiles in metabolic healthy and MUO subjects,
we calculated beta-diversity. By Distance method Bray–
Curtis dissimilarities PCoA analysis, the gut microbiota
samples from Con and MHO children were clustered
together and separated partly from the MUO group.
Upon analysis, the first coordinate (Axis.1) explained the
18.6% of the inter sample variance the second coordinate
(Axis.2) explained the 14.5% of the inter sample variance
Yuan et al. Microb Cell Fact (2021) 20:53
Page 3 of 13
Table 1 Anthropometric profiles and laboratory measurements
MUO (n
=
23)
Metabolic healthy subjects
Total (n
=
63)
MHO (n
=
42)
Con (n
=
21)
Age (year)
Male (%)
Weight (kg)
BMI (kg/m2)
BMI-Z
WHR
WHtR
SBP (mmHg)
DBP (mmHg)
FPG (mmol/L)
TC (mmol/L)
TG (mmol/L)
LDL-c (mmol/L)
HDL-c (mmol/L)
Leptin (μg/mL)
TNF-α (pg/mL)
IL-6 (μg/mL)
LBP (μg/mL)
10.96
1.69
±
65.2
61.4
±
27.02
2.81
11.5
2.75
±
0.61
±
±
0.89
0.05
0.04
0.55
±
116.45
8.77
±
5.72
65.09
5.09
4.54
1.62
±
0.67
0.90
0.99
2.65
0.66
±
±
±
±
±
0.24
1.48
1.24
2.70
±
47.50
1.76
±
25.63
±
0.86
34.8 (29.55, 41.20)
0.86
0.06
1.84*
14.6*
4.91*
±
1.53*
8.36*
0.06*
±
5.79
±
0.39
0.62
0.30*
9.32
50.8
±
43.0
±
21.80
1.77
±
±
0.50
±
101.52
62.57
4.87
4.30
0.86
±
±
±
±
±
2.31
0.53*
1.58
2.23
±
48.48
1.65
±
0.30*
1.83
18.77
±
0.93
33.66 (27.01, 38.95)
1.68*
12.4*
3.14*
±
0.60
9.47
54.8
±
49.6
±
24.65
2.74
±
±
0.88
0.05
0.04
0.53
±
105.51
6.96*
±
6.45
±
0.38*
0.57
0.33*
63.81
4.82
4.39
0.93
2.45
0.48
±
±
±
±
±
1.51
3.10
±
53.43
1.86
±
0.30*
1.65
17.88
±
1.04
33.28 (27.75, 41.22)
9.02
2.14
±
42.9
8.5#
1.91#
0.79#
±
0.06#
0.03#
5.51#
3.56#
±
±
0.40
29.9
±
16.11
±
0.16
−
0.84
±
0.43
±
94.48
60.38
4.97
±
±
±
±
1.71
2.03
4.14
0.72
0.69
0.19#
0.54#
0.26#
0.35*#
16.81#
±
0.42*#
1.23
27.18 (22.02, 36.61)*#
±
38.59
0.51
±
±
MUO metabolic unhealthy obese, MHO metabolically healthy obese, Con controls, BMI body mass index, BMI-Z BMI standard deviation Z score, WHR waist-to-hip ratios,
TC total cholesterol, TG triglyceride, LDL-c low-density lipoprotein cholesterol, HDL-c high density lipoprotein cholesterol, LBP lipopolysaccharide-binding protein
*Compared with the MUO group, p < 0.05
# Compared with the MHO group. Data is expressed either as mean
±
SD or median (25th–75th centiles)
in metabolic healthy subjects (Con and MHO children)
in comparison to MUO participants (P = 0.038, Fig. 2e,
Additional file 1: Table S3).
Bacterial taxa differences in different metabolic status
subjects
We next used LEfSe analysis to identify bacteria in which
the relative abundance was significantly increased or
decreased in each phenotypic category. The Con and
MHO children had members of the phylum Tenericutes,
class Deltaproteobacteria, Mollicutes, order Desulfovi-
brionales, RF39, family Christensenellaceae, Odoribacte-
raceae, Porphyromonadaceae, Ruminococcaceae, genera
Anaerostipes, Oscillospira, Odoribacter, Gemmiger, Para-
bacteroides, Alistipes, that were significantly higher than
MUO subjects. Furthermore, the MUO subjects had
members of the genus Fusobacterium that were sig-
nificantly higher than the Con and MHO children (all
p < 0.05, Fig. 3a, b).
Microbiota profiles in obese children with different
metabolic status
Abundance profiling
Grouping OTUs at phylum level, and applying the
Mann–Whitney U test on the relative abundances of
phyla for the MHO and MUO groups, the relative abun-
dance of phylum Tenericutes was more prevalent in the
MHO group compared to the MUO group (p = 0.027,
Table 3 and Fig. 1c).
On OTUs at the genera level, by Mann–Whitney U
analysis, including all the genera (merging small taxa
with counts < 10), we identified that genera Desulfovibrio,
Parabacteroides and Gemmiger were more prevalent in
MHO subjects compared to MUO subjects (p = 0.027,
0.040 and 0.047, respectively; Fig. 1d).
Alpha‑ and beta‑diversity between MHO and MUO subjects
Regarding alpha-diversity, in both the MHO and MUO
group, the analysis exposed significantly higher diversity
in MHO subjects versus MUO participants (all p < 0.05,
Fig. 2c, d, Additional file 1: Table S2).
Regarding beta-diversity, by an unweighted-UniFrac
method, the MHO group was lower than the MUO group
(p = 0.021, Additional file 1: Table S3).
Bacterial taxa differences between MHO and MUO subjects
LEfSe analysis showed MHO subjects had members
of the phylum Tenericutes, class Deltaproteobacte-
ria, Mollicutes, order Desulfovibrionales, RF39, family
Christensenellaceae, Odoribacteraceae, Rikenellaceae,
Yuan et al. Microb Cell Fact (2021) 20:53
Page 4 of 13
a
MUO
MHO
& Con
0.00
0.25
0.50
Relative Abundance
0.75
b
MUO
MHO
& Con
1.00
0.00
0.25
0.50
0.75
1.00
Phylum
Firmicutes
Bacteroidetes
Proteobacteria
Actinobacteria
Fusobacteria
Tenericutes
Verrucomicrobia
TM7
Synergistetes
c
MUO
MHO
Bacteroides
Not_Assigned
Prevotella
Megamonas
Faecalibacterium
Roseburia
Phascolarctobacterium
Ruminococcus
Dialister
Blautia
Sutterella
Bifidobacterium
Alistipes
Lachnospira
Parabacteroides
Streptococcus
Relative Abundance
Acidaminococcus
Bilophila
Fusobacterium
Enterococcus
Desulfovibrio
Citrobacter
Paraprevotella
Odoribacter
Anaerostipes
Eubacterium
Akkermansia
Lactobacillus
Turicibacter
Gemmiger
Oscillospira
Dorea
Coprococcus
Butyricicoccus
Butyricimonas
Lactococcus
SMB53
Lachnobacterium
Cetobacterium
Rothia
Mitsuokella
Holdemania
Catenibacterium
Actinomyces
Weissella
Anaerotruncus
Klebsiella
Clostridium
Megasphaera
Oxalobacter
Morganella
Adlercreutzia
Coprobacillus
Pseudoramibacter_Eubacterium
Comamonas
Granulicatella
Eggerthella
Pyramidobacter
Abiotrophia
Actinobacillus
Aggregatibacter
Leuconostoc
Veillonella
Enterobacter
Haemophilus
Genus
d
MUO
MHO
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
Relative Abundance
Relative Abundance
Phylum
Bacteroidetes
Firmicutes
Proteobacteria
Actinobacteria
Fusobacteria
Tenericutes
Verrucomicrobia
e
MHO
Con
Bacteroides
Not_Assigned
Prevotella
Megamonas
Phascolarctobacterium
Dialister
Sutterella
Faecalibacterium
Bifidobacterium
Alistipes
Parabacteroides
Roseburia
Oxalobacter
Comamonas
Streptococcus
Ruminococcus
Klebsiella
Blautia
Enterobacter
Veillonella
Haemophilus
Acidaminococcus
Oscillospira
Megasphaera
Clostridium
Lachnospira
Granulicatella
SMB53
Bilophila
Dorea
Gemmiger
Fusobacterium
Coprococcus
Citrobacter
Paraprevotella
Eubacterium
Desulfovibrio
Odoribacter
Butyricicoccus
Turicibacter
Coprobacillus
Actinobacillus
Akkermansia
Lactobacillus
Cetobacterium
Lactococcus
Butyricimonas
Anaerostipes
Holdemania
Catenibacterium
Actinomyces
Morganella
Rothia
Lachnobacterium
Aggregatibacter
Leuconostoc
Genus
f
MHO
Con
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
Relative Abundance
Relative Abundance
Phylum
Firmicutes
Bacteroidetes
Proteobacteria
Actinobacteria
Fusobacteria
Tenericutes
Verrucomicrobia
TM7
Synergistetes
Cyanobacteria
Genus
Bacteroides
Not_Assigned
Faecalibacterium
Prevotella
Megamonas
Roseburia
Ruminococcus
Phascolarctobacterium
Blautia
Dialister
Alistipes
Parabacteroides
Gemmiger
Comamonas
Eggerthella
Pyramidobacter
Bifidobacterium
Oscillospira
Sutterella
Coprococcus
Megasphaera
Clostridium
Klebsiella
Lachnospira
Veillonella
Dorea
Streptococcus
Haemophilus
Acidaminococcus
Parvimonas
Granulicatella
Abiotrophia
Bilophila
Desulfovibrio
Fusobacterium
Odoribacter
Paraprevotella
Anaerostipes
Lactobacillus
Akkermansia
Lachnobacterium
Eubacterium
Turicibacter
Butyricimonas
Mitsuokella
Actinobacillus
Christensenella
Pseudoramibacter_Eubacterium
Cetobacterium
SMB53
Enterobacter
Anaerotruncus
Holdemania
Oxalobacter
Citrobacter
Actinomyces
Adlercreutzia
Lactococcus
Weissella
Coprobacillus
Rothia
Aggregatibacter
Leuconostoc
Fig. 1 Bar chart representing Mann–Whitney U-test results on operational taxonomic units (OTUs) grouped in phyla (a, c, e) and in genus (b, d, f)
of the different metabolic status groups. Each column in the plot represents a group, and each color in the column represents the percentage of
relative abundance for each OTU. MUO metabolic unhealthy obese, MHO metabolically healthy obese, Con controls
Yuan et al. Microb Cell Fact (2021) 20:53
Page 5 of 13
Table 2 The mean relative abundance of gut microbiota
with significantly differences in different metabolic status
at genera level
MUO
MHO and Con
Anaerostipes
Odoribacter
Desulfovibrio
Alistipes
Fusobacterium
Dorea
Gemmiger
Oscillospira
Parabacteroides
0.001
0.000
0.000
0.010
0.001
0.012
0.007
0.008
0.007
0.001
0.002
0.003
0.023
0.002
0.005
0.013
0.010
0.020
Z
−
−
−
−
−
−
−
−
−
2.084
2.122
2.142
2.182
2.185
2.288
2.320
2.445
2.552
P value
0.037
0.034
0.032
0.029
0.029
0.022
0.020
0.014
0.011
MUO metabolic unhealthy obese, MHO metabolically healthy obese, Con
controls
Desulfovibrionaceae, Porphyromonadaceae, Ruminococ-
caceae, genus Gemmiger, Parabacteroides that were sig-
nificantly higher than MUO subjects (all p < 0.05, Fig. 3c,
d).
Microbiota profiles in MHO and Con children with different
weight status
Abundance profiling
Grouping OTUs at phylum level, the relative abundances
of phylum Synergistetes was more prevalent in the MHO
group compared to the Con group (p < 0.05, Fig. 1e,
Table 4).
On OTUs at the genera level, including all the genera
(merging small taxa with counts < 10), genera Anaer-
otruncus, Bacteroides, Adlercreutzia and Pyramidobacter
were more prevalent in MHO subjects versus MUO sub-
jects (p < 0.05; Fig. 1f ).
Alpha‑ and beta‑diversity between different weight status
Regarding alpha-diversity, the Shannon diversity index,
Observed OTUs, Faith’s phylogenetic diversity and Pie-
lou’s evenness based on OTU distribution did not reveal
any significant difference between MHO and Con (all
p > 0.05, Additional file 1: Table S2); also, beta-diversity
did not differ significantly between these two groups.
Importantly, none of the comparisons were significantly
different (all p > 0.05) after correction for multiple testing
(Additional file 1: Table S3).
MHO
&Con MUO
Group
MHO
&Con
MUO
Group
MHO
MUO
b
60
40
20
d
60
40
20
1
o
a
h
C
:
x
e
d
n
I
y
t
i
s
r
e
v
i
d
-
a
h
p
l
A
1
o
a
h
C
:
x
e
d
n
I
y
t
i
s
r
e
v
i
d
-
a
h
p
l
A
MHO
&Con MUO
MHO
MUO
MHO
MUO
Group
MHO
&Con
MUO
Group
MHO
MUO
group
MHO
&Con
MUO
a
3.0
2.5
2.0
1.5
1.0
n
o
n
n
a
h
S
:
x
e
d
n
I
y
t
i
s
r
e
v
i
d
-
a
h
p
l
A
c
3
2
1
n
o
n
n
a
h
S
:
x
e
d
n
I
y
t
i
s
r
e
v
i
d
-
a
h
p
l
A
e
0.50
0.25
0.00
-0.25
-0.50
]
%
5
4
1
[
.
2
.
s
i
x
A
-0.6
-0.3
0.0
0.3
0.6
Axis.1 [18.6%]
Fig. 2 Characterization of alpha- and beta-diversity of the gut
microbiota in Con, MUO and MHO groups. The y-axes show the
Shannon index (a, c) and Chao1 richness index (b, d). The x-axes show
the phenotypic categories. Additional data are in Additional file 1:
Table S2. Principal coordinates analysis (PCoA) plot of Con and MHO
children and MUO subjects (e). The plots show the first two principal
coordinates (axes) for PCoA using Bray–Curtis Distance method. MUO
metabolic unhealthy obese, MHO metabolically healthy obese, Con
controls
Bacterial taxa differences in MHO and Con children
of different weight status
LEfSe analysis showed MHO subjects had members of
the phylum Synergistetes, class Synergistia, order Syn-
ergistales, Erysipetotrichales, family Dethiosulfovibrion-
aceae, genus Pyramidobacter were significantly higher
than the Con-, however, the latter had members of the
Yuan et al. Microb Cell Fact (2021) 20:53
Page 6 of 13
a
MH
MUO
c
MHO
e
Con
MHO
b
MH
MUO
d
f
MHO MUO
Con
MHO
Fig. 3 Differential biomarkers associated with different metabolic status. A linear discriminant effect size (LeFse) analysis have been performed (α
value
2.0). MUO metabolic unhealthy obese, MHO metabolically healthy obese, Con controls
0.05, logarithmic LDA score threshold
=
=
Yuan et al. Microb Cell Fact (2021) 20:53
Page 7 of 13
Table 3 The mean relative abundance of gut microbiota
obese subjects with different metabolic status at phylum
level
MHO
MUO
Actinobacteria
Bacteroidetes
Firmicutes
Fusobacteria
Proteobacteria
Tenericutes
Verrucomicrobia
0.012
0.453
0.393
0.006
0.132
0.003
0.001
0.025
0.371
0.321
0.016
0.267
0.000
0.000
z
−
−
−
−
−
−
−
0.783
0.823
0.919
1.494
0.535
2.212
1.480
p value
0.434
0.410
0.358
0.135
0.593
0.027
0.139
MHO, metabolically healthy obese; MUO: metabolic unhealthy obese
Italicized value P < 0.05
Table 4 The mean relative abundance of gut microbiota
with significantly differences in obese subjects with different
metabolic status at genera level
MHO
Con
Actinobacteria
Bacteroidetes
Cyanobacteria
Firmicutes
Fusobacteria
Proteobacteria
Synergistetes
Tenericutes
TM7
Verrucomicrobia
0.012
0.319
0.000
0.572
0.006
0.088
0.000
0.002
0.000
0.001
0.018
0.377
0.000
0.531
0.014
0.057
0.000
0.002
0.000
0.001
MHO, metabolically healthy obese; Con, control
Italicized value P < 0.05
Z
−
−
−
−
−
−
−
−
−
−
1.181
1.006
1.245
0.831
0.324
1.881
1.964
1.408
0.481
0.177
P value
0.238
0.314
0.213
0.406
0.746
0.060
0.050
0.159
0.630
0.859
family Bacteroidaceae, genus Anaerotruncus that were
significantly higher (all p < 0.05, Fig. 3e, f ).
Correlations between inflammatory factors and bacterial
abundance
To evaluate correlations between bacteria and serum
inflammatory factors (IL-6, TNF-α and leptin), Spear-
man’s rho cut-off values were assessed, taking into
account r > 0.4, r < − 0.4 (p < 0.05, Additional file 1:
Table S4).
For MUO subjects, Spearman’s correlation analy-
sis revealed that IL-6 positively correlated with genus
Lactococcus, TNF-α positively correlated with phylum
Bacteroidetes, negatively correlated with genus Citro-
bacter. LBP positively correlated with genus Prevotella,
Odoribacter, and negatively correlated with genus Bifi-
dobacterium, Streptococcus, Roseburia, Clostridium
and Veillonella. Leptin positively correlated with genus
Eubacterium and negatively correlated with genus Fae-
calibacterium and Lachnospira (all p < 0.05, Additional
file 1: Table S4).
For MHO subjects, Spearman’s correlation analy-
sis revealed that serum IL-6 positively correlated with
genus Paraprevotella. LBP positively correlated with
genus Roseburia and Faecalibacterium, and negatively
correlated with genus Lactobacillus. Leptin positively
correlated with phylum Bacteroidetes, Firmicutes, genus
Phascolarctobacterium and negatively correlated with
genus Dialister (all p < 0.05). There was no association
between the bacteria and TNF α at the genus level (all
p > 0.05).
Metabolic pathway predictions
A total of 15 KEGG pathways were generated using the
composition of the fecal microbiota based on PICRUSt2
in the metabolic healthy cohorts (MHO and Con sub-
jects) versus MUO subjects (Fig. 4, Additional file 1:
Table S5). Importantly, the glucose metabolism pathways,
including GDP-mannose biosynthesis and superpathway
of UDP-N-acetylglucosamine-derived O-antigen building
blocks biosynthesis, were increased in metabolic healthy
cohorts and, conversely, the superpathway of fucose and
rhamnose degradation were alternated in the metabolic
healthy cohorts (all p < 0.05). In the comparison between
MHO and MUO subjects, we obtained 3 differential
pathways including superpathway of fucose and rham-
nose degradation, photorespiration, and sucrose degrada-
tion III, which were also observed significantly different
between the metabolic healthy cohorts (MHO and Con
subjects) versus MUO subjects (Fig. 4, Additional file 1:
Table S6). Moreover, 11 differential metabolic pat-
terns differentially expressed resulted in the compari-
son between MHO versus Con (Fig. 4, Additional file 1:
Table S7).
Discussion
Recognized for decades, there is wide-ranging het-
erogeneity among obese individuals as to their risk for
developing metabolic dysfunction and its attendant com-
plications [12]. Also well-established, and which may
contribute to this metabolic heterogeneity, is the fact
those with central obesity are more prone to develop-
ing T2DM and cardiovascular disease than those with
peripheral obesity [13]. In this study, to indirectly address
the issue of fat distribution, we found there were no sig-
nificant differences in WHR and WHtR between the two
obese cohorts, MHO vs. MUO.
A chronic low-grade inflammation, triggered by nutri-
ent surplus, is a constituent of obesity. Adipose-origi-
nated metabolic inflammation develops pari passu with
insulin resistance and, as such, is a key element in the
Yuan et al. Microb Cell Fact (2021) 20:53
Page 8 of 13
a
MHO&Con
b
c
Con
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
0.05
Fig. 4 KEGGs biomarkers associated with the three metabolic statuses. MUO metabolic unhealthy obese, MHO metabolically healthy obese, Con
controls
Yuan et al. Microb Cell Fact (2021) 20:53
Page 9 of 13
metabolic syndrome [14]. In this study, we found there
were no significant differences in serum IL-6, TNF-α,
LBP and leptin between MHO and MUO subjects. It
stands to reason that, besides these cytokines, other
biochemical factors likely contribute to the metabolic
diverseness in obese subjects. Or, perhaps, the concen-
trations of circulating compounds—such as those above-
mentioned—poorly reflect those found in extracellular or
intracellular tissues.
Evidence can be adduced that the gut microbiota is
involved in the aetiology of obesity and obesity-related
complications such as nonalcoholic fatty liver disease,
insulin resistance and T2DM [15, 16]. These disorders
are characterized by alterations in the diversity of the gut
microbiota, and the relative abundance of certain genera.
And bacteria-generated metabolites, translocated from
the gut across a disrupted intestinal barrier, can affect
several metabolic organs, such as the liver and adipose,
thereby contributing to systemic metabolic inflammation
[17].
Recently, several animal studies concluded that an
optimal healthy-like gut microbiota may bestow a more
propitious obese phenotype [18, 19]. For instance, the
abundance of Bacteroidetes and Tenericutes were closely
aligned with bile acid metabolism and obesity-related
inflammation in a murine model of the metabolic syn-
drome [20]. In our study, we corroborate this finding:
reduced abundance of Tenericutes in the MUO group
compared with the metabolically healthy groups (MHO
and Con). Moreover, individuals with diminished insulin
sensitivity had lower abundance of Tenericutes [21]. And,
in animal experiments, administration of hydrogenated
xanthohumol, which mitigates the metabolic syndrome
by altering gut microbiota diversity and abundance,
specifically, a reduction in Bacteroidetes and Teneri-
cutes [20]. These results suggested an important role of
Tenericutes in metabolism. We also observed greater
abundance of Anaerostipes in the MHO and Con cohort,
as well as the alpha and beta diversity. Using separate-
sample Mendelian randomization to obtain estimates
of the associations of 27 genera of gut microbiota with
cardiovascular disease risks, Anaerostipes was identified
as being nominally associated with T2DM [22], and this
effect may be a result of butyrate production [23]. These
results buttress the notion of dysbiosis in the gut micro-
biota of MUO individuals.
To characterize the gut microbiota in obese children of
different metabolic status, we further analyze the MHO
and the MUO groups. The abundance of Tenericutes was
significantly reduced in the MUO group compared with
the metabolic healthy children, indicating that Teneri-
cutes is related to the metabolic state, and the bacterial
imbalance is independent of weight. Previously reported,
the abundance of Parabacteroides was significantly
decreased in obese subjects with metabolic syndrome [6],
and nonalcoholic fatty liver disease [24], and negatively
correlated with weight gain and leptin plasma levels [25].
And germane to our findings, both genera Gemmiger
[26] and Parabacteroides [27] are gut bacteria negatively
associated with obesity and disturbed host metabolism.
In accordance, we found that that the fecal abundance of
these bacteria was significantly higher in the MHO group
compared with MUO.
The genera Parabacteroides are short-chain fatty acids
(SCFAs)-producing bacteria. SCFAs are low molecular
weight molecules produced from fermentation of dietary
fiber or polysaccharides by gut microbiota. Absorbed by
the intestinal epithelium into the blood, they can beget
physiological disorders in the host, such as deranged lipid
metabolism and intestinal environment imbalances [28,
29]. In our determination, alpha and beta diversity were
significantly higher in Con and MHO children compared
with the MUO group, again supporting the notion of dys-
biosis in the unhealthy MUO population.
Notwithstanding that the gut microbiota of obese
individuals with metabolic syndrome may indeed be
unhealthy, is the gut microbiota of the MHO popula-
tion really healthy? We compared the characteristic of
gut microbiota in the Con and MHO children of differ-
ent weights. Even though there was no significant differ-
ence in alpha and beta diversity, the relative abundances
of phylum Synergistetes and genus Bacteroides were ele-
vated in the MHO group compared to the Con children.
Based on a metagenomic approach and bioinformatics
analysis in obese adults, it is plausible that an abundance
of the microbiota taxa Bacteroides could portent the evo-
lution to T2DM [30].
Alterations in gut ecology can propel inflammatory
pathways in several tissues, resulting in glucose intoler-
ance and CVD [31, 32]. In rodents, a disturbance in the
tripartite interactions between the microbiota, bile acids,
and host metabolism, along with the bacterial production
of lipopolysaccharides (LPS, i.e., endotoxemia), can beget
derangements in glucose homeostasis [16, 26]. LBP is an
acute inflammation phase protein that complexes with
LPS and facilitates binding with CD14. In adolescents,
serum LBP robustly correlates positively with indices of
abnormal glucose and lipid metabolism. Herein, we found
that, depending on the metabolic status, the serum levels
of classic proinflammatory factors IL-6, TNF-α, LBP and
leptin were related to the abundance of various fecal bac-
teria. Notably, in MHO children, serum leptin correlated
positively with genus Phascolarctobacterium and nega-
tively with Dialister—the latter genera observed with low
abundance in obese children [33]. And, relevant to our
findings, it is noteworthy that Phascolarctobacterium is
Yuan et al. Microb Cell Fact (2021) 20:53
Page 10 of 13
purportedly a biomarker for adult T2DM [30]. In high
fat diet obese mice with insulin resistance, Prevotella was
deemed as pro-inflammatory and, of note, its abundance
in our study correlated with serum LBP [34]. As illus-
trated in our MHO children and the above-cited studies
in humans, the gut microbiota is a marquee player in pre-
serving normal metabolism despite obesity or, perhaps,
an ephemeral protective microbiota destined to change
with transition to MUO.
Compared to the metabolic healthy cohorts in the
MUO children, several pathways associated with glu-
cose and lipid metabolism pathways, such as fucose and
rhamnose degradation and sucrose degradation III were
increased. Conversely, mannan degradation was mark-
edly decreased. Of interest, serum fucose levels are higher
in the T2DM patients compared to healthy cohorts [35].
Mannan-oligosaccharide in the diet improves the meta-
bolic syndrome in mice, alternatively insulin resistance
and dyslipidemia [36, 37]. We found that bacterial fucose
and rhamnose degradation and sucrose degradation III
were increased in the MUO subjects compared with the
MHO subjects, inferring that the change was independ-
ent of weight. However, insofar as serum levels of fucose
were undetectable, and the dietary intake of sucrose and
mannan were not assessed in our study, future longitudi-
nal studies could conceivably unravel the intricate, pos-
sibly causual, relationships between the gut microbiota,
obesity, and aberrant intermediary host metabolism.
Conclusion
In aggregate, the MUO population had lower alpha- and
beta-diversity, and lower abundance of Tenericutes, inferring
a robust intricate inter-relationship between gut bacterial
ecology and host metabolic state. In the MHO population,
phylum Synergistetes and genus Bacteroides and Phasco-
larctobacterium were more prevalent, and the abundance
of some metabolism-related bacteria correlated with circu-
lating proinflammatory factors, suggesting that compared
to healthy controls, dysbiosis of gut microbiota was already
extant in the MHO children, and conceivably a compensa-
tory or remedial response to a surfeit of nutrients.
Methods
Study population
This study was approved by the Ethics Committee of the
Fuzhou Children’s Hospital of Fujian Medical University
and, in all cases, informed consent was obtained.
The cross-sectional study consisted of participants
managed by Fuzhou Children’s Hospital of Fujian Medi-
cal University from September 2017 to March 2018. This
study was limited to participants who met the following
criteria: (a) ages between 5 to 15 years old, and (b) resi-
dence of Fujian province.
The exclusion criteria were as follows: any endo-
crine disorder, history of antibiotic therapy in the past
3 months prior to the enrollment, chronic gastrointesti-
nal illness or use of gastro-intestinal-related medication,
or diarrheal disease (World Health Organization defini-
tion) in the past 1 month.
Clinical assessment
Height and weight were measured by trained nurses.
BMI-Z scores were calculated based on reference values
of Li et al. [38]. At the end of normal expiration, waist
and hip circumference were measured to the nearest
0.5 cm using standard technique with nonelastic tape.
Waist circumference was measured at a point midway
between the lower border of the ribs and the iliac crest,
and hip circumference was measured at the widest part
of the hip. A waist-to-hip ratio (WHR) was calculated by
waist circumference (cm) divided by hip circumference
(cm) and a waist-to-height ratio (WHtR) by waist cir-
cumference (cm) divided by height (cm).
Laboratory methods
All participants maintained their usual dietary pattern
at least 3 days before blood sampling. After 12 h of fast-
ing, 10 mL venous blood was drawn by registered nurses.
All blood samples were stored at − 80 ℃, and analyzed
within two weeks of sampling. Serum IL-6 was meas-
ured using a commercial ELISA kit (Abcam, UK), with
an 4.4% inter-assay coefficient of variation (CV). Serum
TNF-α levels was measured using a commercial ELISA
kit (Abcam, UK), with inter-assay and intra-assay CVs
of 3.3% and 9%, respectively, and serum leptin assayed
using a commercial ELISA kit (Abcam, UK), with inter-
assay and intra-assay CVs of 2.4% and 2.7%, respectively.
The serum LBP levels were measured using a commercial
ELISA kit (Abnova, Taiwan, China), with inter-assay and
intra-assay CV 9.8–17.8% and 6.1%, respectively. Fasting
plasma glucose (FPG) and plasma lipids, including total
cholesterol (TC), triglyceride (TG), high-density lipopro-
tein cholesterol (HDL-c) and low density lipoprotein cho-
lesterol (LDL-c), were assayed by standard methods using
specific reagents (Beckman Coulter AU5800, USA). Fast-
ing insulin (INS) was determined by a chemiluminescent
immunoassay (IMMULITE 2000, Siemens Healthcare
Diagnostics Products Limited, Germany). Fecal samples
were collected and processed as previously described
[39].
Definition of metabolic unhealthy
Metabolic syndrome parameters were applied accord-
ing to 2019 Expert Committees [40], and MUO was
Yuan et al. Microb Cell Fact (2021) 20:53
Page 11 of 13
defined by the presence of at least one of the following
metabolic traits: (1) FPG ≥ 5.6 mmol/L; (2) systolic blood
pressure ≥ 90th percentile for gender and age; (3) fasting
HDL-C < 1.03 mmol/L; and (4) fasting TG ≥ 1.7 mmol/L.
Genomic DNA extraction and library construction
The microbial community DNA was extracted and
quantified as previously described [39]. Variable regions
V3–V4 of bacterial 16s rRNA gene were amplified with
degenerate PCR primers [39]. Libraries were qualified
by the Agilent 2100 bioanalyzer (Agilent, USA). The
validated libraries were used for sequencing on Illumina
MiSeq platform (BGI, Shenzhen, China) following the
standard pipeline of Illumina, and generating 2 × 300 bp
paired-end reads.
Statistical analysis
Statistical analyses of clinical data were performed using
the Statistical Package for the Social Sciences software
version 23.0 (SPSS Inc. Chicago, IL, USA). The normal-
ity of the data was tested by Kolmogorov–Smirnov test.
Data are expressed as mean ± SD or median (25th–75th
percentiles). Comparisons of the results were assessed
using independent samples t test, Mann–Whitney U
test and Kruskal–Wallis test, depending on the type of
data distribution (e.g., non parametric). Comparison of
rates between two groups was by chi-square. A value of
P < 0.05 was deemed statistically significant.
Statistical analysis of 16s rRNA sequencing data were
performed on alpha- and beta-diversity measurements,
which was done by software QIIME2 (v2019.7) [41].
Kruskal–Wallis Test was adopted for two groups com-
parison. Linear discriminant analysis Effect Size (LEfSe)
Analysis was assessed by software LEFSE [42]. To pre-
dict metagenome functional content from 16S rRNA
gene surveys, Picrust2 [43] have been applied to obtain
the Kyoto Encyclopedia of Genes and Genomes (KEGG)
pathways, and STAMP [44] was used to analyze the dif-
ferential pathways.
Supplementary Information
The online version contains supplementary material available at https ://doi.
org/10.1186/s1293 4-021-01548 -9.
Additional file 1: Table S1. The mean relative abundance of gut micro-
biota in different metabolic status at phylum level. Table S2. Comparison
of alpha-diversity in obese subjects with different metabolic status.
Table S3. Comparison of beta-diversity between different metabolic
status. Table S4. Spearman’s correlation table on OTUs and inflamma-
tory factors in MHO and MUO groups. Table S5. KEGGs biomarkers in
MHO and Con subjects compared with MUO subjects. Table S6. KEGGs
biomarkers in MHO and MUO subjects. Table S7. KEGGs biomarkers in
MHO and Con subjects.
Acknowledgements
The authors are grateful to all the participants.
Authors’ contributions
XY drafted the initial manuscript; RMC conceptualized and designed the study,
and reviewed and revised the manuscript; KLM assisted in data analysis and
manuscript composition; YZ and XHY collected cases; XQL did the laboratory
testing. All authors read and approved the final manuscript.
Funding
This study was supported by Technology Innovation Team Train Project of
Fuzhou Health Committee in China (2016-S-wp1), and sponsored by key
Clinical Specialty Discipline Construction Program of Fuzhou, Fujian, P.R.C.
(201610191) and Fuzhou Children’s Medical Center (2018080310).
Availability of data and materials
The original contributions presented in the study are publicly available.
The raw sequence data reported in this paper have been deposited in the
Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2017) in
National Genomics Data Center (Nucleic Acids Res 2020), Beijing Institute of
Genomics (China National Center for Bioinformation), Chinese Academy of
Sciences, under accession number CRA003010 that are publicly accessible at
https ://bigd.big.ac.cn/gsa.
Ethics approval and consent to participate
This study was reviewed and approved by the Ethics Committee of Fuzhou
Children’s Hospital of Fujian Medical University, and was conducted in
agreement with the Declaration of Helsinki Principles. Informed consent was
obtained from all individual participants included in the study.
Consent for publication
Informed consent for publication was obtained from all individual participants
included in the study.
Competing interests
The authors declare that they have no competing interests.
Author details
1 Department of Endocrinology, Fuzhou Children’s Hospital of Fujian Medical
University, NO. 145, 817 Middle Road, Fuzhou 350005, China. 2 Division of Pedi-
atric Endocrinology and Diabetes, University of Alabama at Birmingham,
Birmingham, AL 35233, USA.
Received: 8 December 2020 Accepted: 18 February 2021
References
1.
The CR, Picture B. Outrunning child obesity trends. BMJ. 2018;363:k4362.
https ://doi.org/10.1136/bmj.k4362 .
2. Damanhoury S, Newton AS, Rashid M, Hartling L, Byrne JLS, Ball GDC.
Defining metabolically healthy obesity in children: a scoping review.
Obes Rev. 2018;19(11):1476–91. https ://doi.org/10.1111/obr.12721 .
3. Greenhill C. Redefining metabolically healthy obesity. Nat Rev Endocrinol.
2018;14(8):440. https ://doi.org/10.1038/s4157 4-018-0046-9.
4. Garcia-Moll X. Obesity and prognosis: time to forget about metabolically
healthy obesity. Eur Heart J. 2018;39(5):407–9. https ://doi.org/10.1093/
eurhe artj/ehx53 5.
5. Gérard P. Gut microbiota and obesity. Cell Mol Life Sci. 2016;73(1):147–62.
https ://doi.org/10.1007/s0001 8-015-2061-5.
6. Gallardo-Becerra L, Cornejo-Granados F, García-López R, Valdez-Lara
A, Bikel S, Canizales-Quinteros S, López-Contreras BE, Mendoza-Vargas
A, Nielsen H, Ochoa-Leyva A. Metatranscriptomic analysis to define
the Secrebiome, and 16S rRNA profiling of the gut microbiome in
obesity and metabolic syndrome of Mexican children. Microb Cell Fact.
2020;19(1):61. https ://doi.org/10.1186/s1293 4-020-01319 -y.
7. Ma Q, Li Y, Li P, Wang M, Wang J, Tang Z, Wang T, Luo L, Wang C, Wang T,
Zhao B. Research progress in the relationship between type 2 diabetes
mellitus and intestinal flora. Biomed Pharmacother. 2019;117:109138.
https ://doi.org/10.1016/j.bioph a.2019.10913 8.
Yuan et al. Microb Cell Fact (2021) 20:53
Page 12 of 13
8. Rial SA, Karelis AD, Bergeron KF, Mounier C. Gut microbiota and metabolic
health: the potential beneficial effects of a medium chain triglyceride diet
in obese individuals. Nutrients. 2016;8(5):281. https ://doi.org/10.3390/
nu805 0281.
Iacobini C, Pugliese G, Blasetti Fantauzzi C, Federici M, Menini S. Meta-
bolically healthy versus metabolically unhealthy obesity. Metabolism.
2019;92:51–60. https ://doi.org/10.1016/j.metab ol.2018.11.009.
9.
10. Yuan X, Chen R, Ouyang Q, Lin X, Ai Z, Zhang Y, Yang X. Novel associations
of serum adropin and lipopolysaccharide-binding protein versus lipid
profiles in childhood obesity. J Pediatr Endocrinol Metab. 2020;33(2):265–
70. https ://doi.org/10.1515/jpem-2019-0329.
11. Nirmalkar K, Murugesan S, Pizano-Zárate ML, Villalobos-Flores LE, García-
González C, Morales-Hernández RM, Nuñez-Hernández JA, Hernández-
Quiroz F, Romero-Figueroa MDS, Hernández-Guerrero C, Hoyo-Vadillo C,
García-Mena J. Gut microbiota and endothelial dysfunction markers in
obese mexican children and adolescents. Nutrients. 2018;10(12):2009.
https ://doi.org/10.3390/nu101 22009 .
12. Viveiros A, Oudit GY. The dual nature of obesity in metabolic pro-
gramming: quantity versus quality of adipose tissue. Clin Sci.
2020;134(18):2447–51. https ://doi.org/10.1042/CS202 01028 .
13. Brandão I, Martins MJ, Monteiro R. Metabolically healthy obesity-
heterogeneity in definitions and unconventional factors. Metabolites.
2020;10(2):E48. https ://doi.org/10.3390/metab o1002 0048.
14. Kacířová M, Zmeškalová A, Kořínková L, Železná B, Kuneš J, Maletínská
L. Inflammation: major denominator of obesity, type 2 diabetes and
Alzheimer’s disease-like pathology? Clin Sci. 2020;134(5):547–70. https ://
doi.org/10.1042/CS201 91313 .
15. Canfora EE, Meex RCR, Venema K, Blaak EE. Gut microbial metabolites in
obesity, NAFLD and T2DM. Nat Rev Endocrinol. 2019;15(5):261–73. https
://doi.org/10.1038/s4157 4-019-0156-z.
16. Pedersen HK, Gudmundsdottir V, Nielsen HB, Hyotylainen T, Nielsen T,
Jensen BA, Forslund K, Hildebrand F, Prifti E, Falony G, Le Chatelier E,
Levenez F, Doré J, Mattila I, Plichta DR, Pöhö P, Hellgren LI, Arumugam
M, Sunagawa S, Vieira-Silva S, Jørgensen T, Holm JB, Trošt K, MetaHIT
Consortium, Kristiansen K, Brix S, Raes J, Wang J, Hansen T, Bork P, Brunak
S, Oresic M, Ehrlich SD, Pedersen O. Human gut microbes impact host
serum metabolome and insulin sensitivity. Nature. 2016;535(7612):376–
81. https ://doi.org/10.1038/natur e1864 6.
17. Tilg H, Zmora N, Adolph TE, Elinav E. The intestinal microbiota fuelling
metabolic inflammation. Nat Rev Immunol. 2020;20(1):40–54. https ://doi.
org/10.1038/s4157 7-019-0198-4.
18. Serino M, Luche E, Gres S, Baylac A, Bergé M, Cenac C, Waget A, Klopp
P, Iacovoni J, Klopp C, Mariette J, Bouchez O, Lluch J, Ouarné F, Monsan
P, Valet P, Roques C, Amar J, Bouloumié A, Théodorou V, Burcelin R.
Metabolic adaptation to a high-fat diet is associated with a change in the
gut microbiota. Gut. 2012;61(4):543–53. https ://doi.org/10.1136/gutjn
l-2011-30101 2.
19. Neyrinck AM, Van Hée VF, Piront N, De Backer F, Toussaint O, Cani PD,
Delzenne NM. Wheat-derived arabinoxylan oligosaccharides with
prebiotic effect increase satietogenic gut peptides and reduce metabolic
endotoxemia in diet-induced obese mice. Nutr Diabetes. 2012;2(1):e28.
https ://doi.org/10.1038/nutd.2011.24.
20. Zhang Y, Bobe G, Revel JS, Rodrigues RR, Sharpton TJ, Fantacone ML,
Raslan K, Miranda CL, Lowry MB, Blakemore PR, Morgun A, Shulzhenko N,
Maier CS, Stevens JF, Gombart AF. Improvements in metabolic syndrome
by xanthohumol derivatives are linked to altered gut microbiota and bile
acid metabolism. Mol Nutr Food Res. 2020;64(1):e1900789. https ://doi.
org/10.1002/mnfr.20190 0789.
21. Benedict C, Vogel H, Jonas W, Woting A, Blaut M, Schürmann A, Ceder-
naes J. Gut microbiota and glucometabolic alterations in response to
recurrent partial sleep deprivation in normal-weight young individu-
als. Mol Metab. 2016;5(12):1175–86. https ://doi.org/10.1016/j.molme
t.2016.10.003.
22. Yang Q, Lin SL, Kwok MK, Leung GM, Schooling CM. The roles of 27 gen-
era of human gut microbiota in ischemic heart disease, type 2 diabetes
mellitus, and their risk factors: a mendelian randomization study. Am J
Epidemiol. 2018;187(9):1916–22. https ://doi.org/10.1093/aje/kwy09 6.
23. Koh A, De Vadder F, Kovatcheva-Datchary P, Bäckhed F. From dietary fiber
to host physiology: short-chain fatty acids as key bacterial metabolites.
Cell. 2016;165(6):1332–45. https ://doi.org/10.1016/j.cell.2016.05.041.
24. Monga Kravetz A, Testerman T, Galuppo B, Graf J, Pierpont B, Siebel S,
Feinn R, Santoro N. Effect of gut microbiota and PNPLA3 rs738409 vari-
ant on nonalcoholic fatty liver disease (NAFLD) in obese youth. J Clin
Endocrinol Metab. 2020;105(10):dgaa382. https ://doi.org/10.1210/cline
m/dgaa3 82.
25. Carbajo-Pescador S, Porras D, García-Mediavilla MV, Martínez-Flórez S,
Juarez-Fernández M, Cuevas MJ, Mauriz JL, González-Gallego J, Nistal E,
Sánchez-Campos S. Beneficial effects of exercise on gut microbiota func-
tionality and barrier integrity, and gut-liver crosstalk in an in vivo model
of early obesity and non-alcoholic fatty liver disease. Dis Model Mech.
2019;12(5):dmm039206. https ://doi.org/10.1242/dmm.03920 6.
26. Brown JM, Hazen SL. Microbial modulation of cardiovascular disease. Nat
Rev Microbiol. 2018;16:171–81. https ://doi.org/10.1038/nrmic ro.2017.
27. Patrone V, Vajana E, Minuti A, Callegari ML, Federico A, Loguercio C, Dallio
M, Tolone S, Docimo L, Morelli L. Postoperative changes in fecal bacterial
communities and fermentation products in obese patients undergo-
ing bilio-intestinal bypass. Front Microbiol. 2016;7:200. https ://doi.
org/10.3389/fmicb .2016.00200 .
28. Jing N, Liu X, Jin M, Yang X, Hu X, Li C, Zhao K. Fubrick tea attenuates
high-fat diet induced fat deposition and metabolic disorder by regulating
gut microbiota and caffeine metabolism. Food Funct. 2020;11(8):6971–
86. https ://doi.org/10.1039/d0fo0 1282c .
29. Hold GL. The gut microbiota, dietary extremes and exercise. Gut.
2014;63(12):1838–9. https ://doi.org/10.1136/gutjn l-2014-30730 5.
30. Wang TY, Zhang XQ, Chen AL, Zhang J, Lv BH, Ma MH, Lian J, Wu YX,
Zhou YT, Ma CC, Dong RJ, Ge DY, Gao SH, Jiang GJ. A comparative study
of microbial community and functions of type 2 diabetes mellitus
patients with obesity and healthy people. Appl Microbiol Biotechnol.
2020;104(16):7143–53. https ://doi.org/10.1007/s0025 3-020-10689 -7.
31. Biassoni R, Di Marco E, Squillario M, Barla A, Piccolo G, Ugolotti E, Gatti C,
Minuto N, Patti G, Maghnie M, d’Annunzio G. Gut microbiota in T1DM-
onset pediatric patients: machine-learning algorithms to classify microor-
ganisms as disease linked. J Clin Endocrinol Metab. 2020;105(9):dgaa407.
https ://doi.org/10.1210/cline m/dgaa4 07.
32. van de Guchte M, Blottière HM, Doré J. Humans as holobionts: implica-
tions for prevention and therapy. Microbiome. 2018;6:81. https ://doi.
org/10.1186/s4016 8-018-0466-8.
33. Wahlström A, Sayin SI, Marschall HU, Bäckhed F. Intestinal crosstalk
between bile acids and microbiota and its impact on host metabolism.
Cell Metab. 2016;24(1):41–50. https ://doi.org/10.1016/j.cmet.2016.05.005.
34. Cai W, Xu J, Li G, Liu T, Guo X, Wang H, Luo L. Ethanol extract of propolis
prevents high-fat diet-induced insulin resistance and obesity in
association with modulation of gut microbiota in mice. Food Res Int.
2020;130:108939. https ://doi.org/10.1016/j.foodr es.2019.10893 9.
35. Al-Aama JY, Al Mahdi HB, Salama MA, Bakur KH, Alhozali A, Mosli HH,
Bahijri SM, Bahieldin A, Willmitzer L, Edris S. Detection of secondary
metabolites as biomarkers for the early diagnosis and prevention of type
2 diabetes. Diabetes Metab Syndr Obes. 2019;17(12):2675–84. https ://doi.
org/10.2147/DMSO.S2155 28.
36. Wang H, Zhang X, Wang S, Li H, Lu Z, Shi J, Xu Z. Mannan-oligosaccharide
modulates the obesity and gut microbiota in high-fat diet-fed mice. Food
Funct. 2018;9(7):3916–29. https ://doi.org/10.1039/c8fo0 0209f .
37. Chen H, Nie Q, Hu J, Huang X, Zhang K, Pan S, Nie S. Hypoglycemic and
hypolipidemic effects of glucomannan extracted from Konjac on type
2 diabetic rats. J Agric Food Chem. 2019;67(18):5278–88. https ://doi.
org/10.1021/acs.jafc.9b011 92.
38. Li H, Zong XN, Ji CY, Mi J. Body mass index cut-offs for overweight and
obesity in Chinese children and adolescents aged 2–18 years. Zhonghua
Liu Xing Bing Xue Za Zhi. 2010;31(6):616–20.
39. Yuan X, Chen R, Zhang Y, Lin X, Yang X. Sexual dimorphism of gut micro-
biota at different pubertal status. Microb Cell Fact. 2020;19(1):152. https ://
doi.org/10.1186/s1293 4-020-01412 -2.
40. Liu C, Wang C, Guan S, Liu H, Wu X, Zhang Z, Gu X, Zhang Y, Zhao Y,
Tse LA, Fang X. The prevalence of metabolically healthy and unhealthy
obesity according to different criteria. Obes Facts. 2019;12(1):78–90. https
://doi.org/10.1159/00049 5852.
41. Bolyen E, Rideout JR, Dillon MR, Bokulich NA, Abnet CC, Al-Ghalith GA,
Alexander H, Alm EJ, Arumugam M, Asnicar F, Bai Y, Bisanz JE, Bittinger K,
Brejnrod A, Brislawn CJ, Brown CT, Callahan BJ, Caraballo-Rodríguez AM,
Chase J, Cope EK, Da Silva R, Diener C, Dorrestein PC, Douglas GM, Durall
DM, Duvallet C, Edwardson CF, Ernst M, Estaki M, Fouquier J, Gauglitz JM,
Yuan et al. Microb Cell Fact (2021) 20:53
Page 13 of 13
Gibbons SM, Gibson DL, Gonzalez A, Gorlick K, Guo J, Hillmann B, Holmes
S, Holste H, Huttenhower C, Huttley GA, Janssen S, Jarmusch AK, Jiang
L, Kaehler BD, Kang KB, Keefe CR, Keim P, Kelley ST, Knights D, Koester I,
Kosciolek T, Kreps J, Langille MGI, Lee J, Ley R, Liu YX, Loftfield E, Lozupone
C, Maher M, Marotz C, Martin BD, McDonald D, McIver LJ, Melnik AV,
Metcalf JL, Morgan SC, Morton JT, Naimey AT, Navas-Molina JA, Nothias
LF, Orchanian SB, Pearson T, Peoples SL, Petras D, Preuss ML, Pruesse E,
Rasmussen LB, Rivers A, Robeson MS 2nd, Rosenthal P, Segata N, Shaffer
M, Shiffer A, Sinha R, Song SJ, Spear JR, Swafford AD, Thompson LR, Torres
PJ, Trinh P, Tripathi A, Turnbaugh PJ, Ul-Hasan S, van der Hooft JJJ, Vargas
F, Vázquez-Baeza Y, Vogtmann E, von Hippel M, Walters W, Wan Y, Wang
M, Warren J, Weber KC, Williamson CHD, Willis AD, Xu ZZ, Zaneveld JR,
Zhang Y, Zhu Q, Knight R, Caporaso JG. Reproducible, interactive, scalable
and extensible microbiome data science using QIIME 2. Nat Biotechnol.
2019;37(8):852–7. https ://doi.org/10.1038/s4158 7-019-0209-9.
42. Segata N, Izard J, Waldron L, Gevers D, Miropolsky L, Garrett WS, Hutten-
hower C. Metagenomic biomarker discovery and explanation. Genome
Biol. 2011;12(6):R60. https ://doi.org/10.1186/gb-2011-12-6-r60.
43. Douglas GM, Maffei VJ, Zaneveld JR, Yurgel SN, Brown JR, Taylor CM,
Huttenhower C, Langille MGI. PICRUSt2 for prediction of metagenome
functions. Nat Biotechnol. 2020;38(6):685–8. https ://doi.org/10.1038/
s4158 7-020-0548-6.
44. Parks DH, Tyson GW, Hugenholtz P, Beiko RG. STAMP: statistical analysis
of taxonomic and functional profiles. Bioinformatics. 2014;30(21):3123–4.
https ://doi.org/10.1093/bioin forma tics/btu49 4.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in pub-
lished maps and institutional affiliations.
• fast, convenient online submission • thorough peer review by experienced researchers in your field• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
| null |
10.1038_s41586-023-05909-9.pdf
|
Data availability
The atomic coordinates and experimental data of RPB_PEW3_R4–
PAWx4, RPB_PLP3_R6–PLPx6, RPB_LRP2_R4–LRPx4, RPB_PLP1_R6–
PLPx6, RPB_PLP1_R6–PLPx6 (alternative conformation 1), RPB_PLP1_
R6–PLPx6 (alternative conformation 2) and RPB_LRP2_R4 (pseudo-
polymeric) have been deposited in the RCSB PDB with the accession
Article numbers 7UDJ, 7UE2, 7UDK, 7UDL, 7UDM, 7UDN and 7UDO, respec-
tively. The Rosetta macromolecular modelling suite (https://www.roset-
tacommons.org) is freely available to academic and non-commercial
users. Commercial licences for the suite are available through the Uni-
versity of Washington Technology Transfer Office. The mass spectrom-
etry proteomics data have been deposited to the ProteomeXchange
Consortium through the PRIDE partner repository with the dataset
identifiers PXD038492 and 10.6019/PXD038492. Source data are pro-
vided with this paper. All protein sequences for the binders described
in this study are provided in Supp
|
Data availability The atomic coordinates and experimental data of RPB_PEW3_R4-PAWx4, RPB_PLP3_R6-PLPx6, RPB_LRP2_R4-LRPx4, RPB_PLP1_R6-PLPx6, RPB_PLP1_R6-PLPx6 (alternative conformation 1), RPB_PLP1_ R6-PLPx6 (alternative conformation 2) and RPB_LRP2_R4 (pseudopolymeric) have been deposited in the RCSB PDB with the accession numbers 7UDJ, 7UE2, 7UDK, 7UDL, 7UDM, 7UDN and 7UDO, respectively. The Rosetta macromolecular modelling suite ( https://www.roset- tacommons.org ) is freely available to academic and non-commercial users. Commercial licences for the suite are available through the University of Washington Technology Transfer Office. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium through the PRIDE partner repository with the dataset identifiers PXD038492 and 10.6019/PXD038492. Source data are provided with this paper. All protein sequences for the binders described in this study are provided in Supplementary Table 2 . Code availability The design scripts and main PDB models, computational protocol for data analysis, experimental data and analysis scripts, all the design models and the results used in this paper can be downloaded from file servers hosted by the Institute for Protein Design: https://files.ipd.uw.edu/pub/2023_modular_peptide_bind- ing_proteins/all_data_modular_peptide_binding_proteins.tar.gz . The code to identify proteins in databases containing any linear combination of amino acid triplets given as an input can be found on GitHub ( https://github.com/tjs23/prot_pep_scan ).
|
De novo design of modular peptide-binding
proteins by superhelical matching
https://doi.org/10.1038/s41586-023-05909-9
Received: 10 April 2022
Accepted: 1 March 2023
Published online: 5 April 2023
Open access
Check for updates
Kejia Wu1,2,3,12, Hua Bai1,2,4,12, Ya-Ting Chang5, Rachel Redler5, Kerrie E. McNally6,
William Sheffler1,2, T. J. Brunette1,2, Derrick R. Hicks1,2, Tomos E. Morgan6, Tim J. Stevens6,
Adam Broerman1,2,7, Inna Goreshnik1,2, Michelle DeWitt1,2, Cameron M. Chow1,2, Yihang Shen8,
Lance Stewart1,2, Emmanuel Derivery6 ✉, Daniel Adriano Silva1,2,9,10 ✉, Gira Bhabha5,
Damian C. Ekiert5,11 & David Baker1,2,4 ✉
General approaches for designing sequence-specific peptide-binding proteins
would have wide utility in proteomics and synthetic biology. However, designing
peptide-binding proteins is challenging, as most peptides do not have defined
structures in isolation, and hydrogen bonds must be made to the buried polar groups
in the peptide backbone1–3. Here, inspired by natural and re-engineered protein–
peptide systems4–11, we set out to design proteins made out of repeating units that
bind peptides with repeating sequences, with a one-to-one correspondence between
the repeat units of the protein and those of the peptide. We use geometric hashing to
identify protein backbones and peptide-docking arrangements that are compatible
with bidentate hydrogen bonds between the side chains of the protein and the
peptide backbone12. The remainder of the protein sequence is then optimized for
folding and peptide binding. We design repeat proteins to bind to six different
tripeptide-repeat sequences in polyproline II conformations. The proteins are
hyperstable and bind to four to six tandem repeats of their tripeptide targets with
nanomolar to picomolar affinities in vitro and in living cells. Crystal structures reveal
repeating interactions between protein and peptide interactions as designed,
including ladders of hydrogen bonds from protein side chains to peptide backbones.
By redesigning the binding interfaces of individual repeat units, specificity can be
achieved for non-repeating peptide sequences and for disordered regions of native
proteins.
A number of naturally occurring protein families bind to peptides with
repeating internal sequences7,9. The armadillo-repeat proteins, which
include the nuclear import receptors, bind to extended peptides with
lysine- and arginine-rich sequences such that each repeat unit in the
peptide fits into a repeat unit or module in the protein5,8. Previous
studies have shown that the specificity of individual protein repeat
units can be re-engineered, which enables broader recognition of
peptide sequences6,11,13,14. Although this approach is powerful, it is
limited to binding peptides in backbone conformations that are com-
patible with the geometry of the armadillo repeat. Tetratricopeptide-
repeat proteins bind to peptides with a variety of sequences and
conformations with lower (micromolar) affinity (for exceptions, see
refs. 15–17) and with deviations in each peptide–protein interaction
register, which complicates engineering for more general peptide
recognition4,9,10.
Design approach
We set out to generalize peptide recognition by modular repeat-protein
scaffolds to arbitrary repeating-peptide backbone geometries. This
requires solving two main challenges: first, building protein struc-
tures with a repeat spacing and orientation matching that of the target
peptide conformation; and, second, ensuring the replacement of pep-
tide–water hydrogen bonds in the unbound state with peptide–protein
hydrogen bonds in the bound state. The first challenge is crucial for
modular and extensible sequence recognition: if individual repeat units
in the protein are to bind individual repeat units on the peptide in the
same orientation, the geometric phasing of the repeat units on protein
and peptide must be compatible. The second challenge is important
for achieving a high binding affinity: in conformations other than the
α- and 310-helix, the NH and C=O groups make hydrogen bonds with
1Department of Biochemistry, University of Washington, Seattle, WA, USA. 2Institute for Protein Design, University of Washington, Seattle, WA, USA. 3Biological Physics, Structure and Design
Graduate Program, University of Washington, Seattle, WA, USA. 4Howard Hughes Medical Institute, University of Washington, Seattle, WA, USA. 5Department of Cell Biology, New York
University School of Medicine, New York, NY, USA. 6MRC Laboratory of Molecular Biology, Cambridge, UK. 7Department of Chemical Engineering, University of Washington, Seattle, WA, USA.
8Department of Computational Biology, Carnegie Mellon University, Pittsburgh, PA, USA. 9Division of Life Science, The Hong Kong University of Science and Technology, Kowloon, Hong Kong.
10Monod Bio, Seattle, WA, USA. 11Department of Microbiology, New York University School of Medicine, New York, NY, USA. 12These authors contributed equally: Kejia Wu, Hua Bai. ✉e-mail:
[email protected]; [email protected]; [email protected]
Nature | Vol 616 | 20 April 2023 | 581
Article
water in the unbound state that need to be replaced with hydrogen
bonds to the protein upon binding to avoid incurring a substantial
free-energy penalty15.
To address the first challenge, we reasoned that a necessary (but
not sufficient) criterion for in-phase geometric matching between
repeating units on the designed protein and repeating units on the
peptide was a correspondence between the superhelices that the two
trace out. All repeating polymeric structures trace out superhelices
that can be described by three parameters: the translation (rise) along
the helical axis per repeat unit; the rotation (twist) around this axis;
and the distance (radius) of the repeat unit centroid from the axis18,19
(Fig. 1a). We generated large sets of repeating-protein backbones that
sampled a wide range of superhelical geometries (see Methods). We
then generated corresponding sets of repeating-peptide backbones by
randomly sampling di-peptide and tri-peptide conformations (avoid-
ing intra-peptide steric clashes), and then repeating these 4–6 times
to generate 8–18-residue peptides. We then searched for matching
pairs of repeat-protein and repeat-peptide backbones, requiring the
rise to be within 0.2 Å, the twist to be within 5° and the radius to differ
by at least 4 Å (the difference in radius is necessary to avoid clashing
between peptide and protein; the peptide can wrap either outside or
inside the protein).
To address the second challenge, we reasoned that bidentate hydro-
gen bonds between side chains on the protein and pairs of backbone
groups or backbone and side-chain groups on the peptide could allow
the burying of sufficient peptide surface area on the protein to achieve
high-affinity binding without incurring a large desolvation penalty20,21.
As the geometric requirements for such bidentate hydrogen bonds are
quite strict, we developed a geometric hashing approach to enable rapid
identification of rigid-body docks of the peptide on the protein that are
compatible with ladders of bidentate interactions. To generate the hash
tables for bidentate side-chain–backbone interactions, Monte Carlo
simulations of individual side-chain functional groups making biden-
tate hydrogen-bonding interactions with peptide backbone and/or
side-chain groups were performed using the Rosetta energy function12,
and a move set consisting of both rigid-body perturbations and changes
to the peptide backbone torsions (Fig. 1b; see Methods for details).
For each accepted (low-energy) arrangement, side-chain rotamer
conformations were built backwards from the functional group to
identify placements of the protein backbone from which the bidentate
interaction could be realized. The results were stored in hash tables:
for each placement, a hash key was computed from the rigid-body
transformation and the peptide backbone and side-chain torsion angles
determining the position of the hydrogen-bonding groups (for exam-
ple, the phi and psi torsion angles for a bidentate hydrogen bond to the
NH and CO groups of the same amino acid), and the chi angles of the
corresponding rotamer were stored in the hash for this key20. Hash
tables were generated for Asn and Gln making bidentate interactions
with the N–H and C=O groups on the backbone of a single residue
or adjacent residues, for Asp or Glu making bidentate interactions
with the N–H groups of two successive amino acids, and for side-
chain–side-chain pi–pi and cation–pi interactions (see Methods).
To identify rigid-body docks that enable multiple bidentate hydrogen
bonds between the repeat protein and the repeat peptide, we took
advantage of the fact that for matching two superhelical structures
along their common axis, there are only two degrees of freedom:
the relative translation and rotation along this axis. For each repeat
protein–repeat peptide pair, we performed a grid search in these two
degrees of freedom, sampling relative translations and rotations in
increments of around 1 Å and 10° (Fig. 1c). For each generated dock,
we computed the rigid-body orientation for each peptide–protein
residue pair, and queried the hash tables to rapidly determine whether
bidentate interactions could be made; docks for which the number of
matches was less than a set threshold were discarded. For the remain-
ing docks, after building the interacting side chains using the chi
582 | Nature | Vol 616 | 20 April 2023
angle information stored in the hash, and rigid-body minimization
to optimize hydrogen-bond geometry, we used Rosetta combina-
torial optimization to design the protein and peptide sequences22,
keeping the residues that were identified in the hash matching fixed,
and enforcing sequence identity between repeats in both the peptide
and the protein (see Methods).
In initial calculations with unrestricted sampling of peptide confor-
mations, designs were generated with a wide range of peptide confor-
mations. Examples of repeat proteins designed to bind to extended
β-strand, polypeptide II and helical peptide backbones, as well as
to a range of less canonical structures, are shown in Extended Data
Fig. 1a–c. Reasoning that proline-containing peptides would incur a
lower entropic cost upon binding than non-proline-containing pep-
tides, we decided to start our experimental characterization with
designs containing at least one proline residue; in most of these designs,
the peptide backbone is in or near the polyproline II portion of the
Ramachandran map. Our design strategy requires matching the twist of
the repeat unit of the peptide with that of the protein, and hence choos-
ing a repeat length of the peptide that generates close to a full 360° turn
requires less of a twist in the repeat protein. For the polyproline helix,
there are roughly three residues per turn, and, probably because of this,
we obtained more designs that target three-residue than two-residue
proline-containing repeat units. We selected for experimental charac-
terization 43 designed complexes with near-ideal bidentate hydrogen
bonds between protein and peptide, favourable protein–peptide inter-
action energies12, interface shape complementary23 and few interface
unsatisfied hydrogen bonds24, and which consistently retained more
than 80% of the interchain hydrogen bonds in 20-ns molecular dyna-
mics trajectories.
Experimental characterization
We obtained synthetic genes encoding the designed proteins with
6×His tags for purification and terminal biotinylation tags for fluores-
cent labelling, expressed the proteins in Escherichia coli and purified
them by Ni-NTA chromatography. Out of 49, 30 were monomeric and
soluble. To assess binding, the target peptides were displayed on the
yeast cell surface25, and binding to the repeat proteins was monitored
by flow cytometry. To obtain readout of the peptide-binding specific-
ity of individual designs, we in parallel used large-scale array-based
oligonucleotide synthesis to generate yeast display libraries encod-
ing all two- and three-residue repeat peptides with eight repeat units
each, and used fluorescence-activated cell sorting (FACS) followed
by Sanger sequencing to identify the peptides recognized by each
designed protein. Many of the designs bound peptides with sequences
similar to those targeted, but the affinity and specificity were both
relatively low, with most of the successes for three-residue repeat units
(Extended Data Table 1a).
On the basis of these results, we sought to increase the peptide
sequence specificity of the computational design protocol, focusing on
the design of binders for peptides with three-residue repeat units. First,
we required that each non-proline residue in the peptide make specific
contacts with the protein, and that the pockets and grooves engaging
side chains emanating from the two sides of the peptide were quite dis-
tinct. Second, after the design stage, we evaluated the change in binding
energy (Rosetta ddG)26 for all single-residue changes to the peptide
repeating unit, and selected only designs for which the design target
sequence made the most favourable interactions with the designed
protein. Third, we used computational alanine scanning to remove
hydrophobic residues on the protein surface that did not contribute to
binding specificity, to decrease non-specific binding27. Fourth, to assess
the structural specificity of the designed peptide-binding interface,
we performed Monte Carlo flexible backbone docking calculations,
starting from large numbers of peptide conformations with super-
helical parameters in the range of those of the proteins, and selected
Article b
C
O
CA
N
C
O
O
CA
N
C
Res i+1
Res i
CA
N
C
CA
OG1
O
N
Thr
ΔZ
ΔZ
90º
(cid:90)
(cid:90)
CA
N
O
C
(cid:92)
(cid:92)
(cid:73)
C
O
CA
N
O
CAC
N
Sample
6D rigid-body
transformation
and backbone
dihedral angles
Sample
side-chain
rotamers
Hash key
= ƒ(6D rigid-body transformation,
,
ΦΨ
)
Hash table value
Residue name
Chi angles
(cid:70)3: 39.6º
(cid:70)2: –94.0º
(cid:70)1: 66.8º
4611686132949431850
Hash key
122369946438
Hash key
not found
Hash key
in table
a
c
d
Fig. 1 | Overview of the procedure for designing modular peptide binders.
a, Like all repeating structures, repeat proteins and peptides form superhelices
with constant axial displacement (ΔZ) and angular twist (ω) between adjacent
repeat units (shown in green and yellow). For in-register binding, the protein
and peptide parameters must match (for some integral multiple of repeat units).
b, Construction of hash tables for privileged residue–residue interactions. Top
row: classes of side-chain–backbone interactions for which hash tables were
built. The side-chain amide group of asparagine or glutamine forms bidentate
interactions with the N–H and C=O groups on the backbone of a single residue
(left) or consecutive residues (middle), or with the backbone N–H group and
side-chain oxygen atom of a serine or threonine residue (right). Second row:
as illustrated for the case of the glutamine–backbone bidentate interaction,
to build the hash table we perform Monte Carlo sampling over the rigid-body
orientation between the terminal amide group and the backbone, and the
backbone torsions φ and ψ, saving configurations with low-energy bidentate
hydrogen bonds. For each configuration, the possible placements for the
backbone of the glutamine are enumerated by growing side-chain rotamers
back from the terminal amide. Third row: from the six rigid-body degrees of
freedom relating the backbones of the two residues, together with the two φ
and ψ torsion angle degrees of freedom, a hash key is calculated using an eight-
dimensional hashing scheme. The hash key is then added to the hash table with
the side-chain name and torsions as the value. CA, α-carbon; OG, γ-oxygen.
c, To dock repeat proteins and repeat peptides with compatible superhelical
parameters, their superhelical axes are first aligned, and the repeat peptide is
then rotated around and slid along this axis. For each of these docks, for each
pair of repeat protein–repeat peptide residues within a threshold distance, the
hash key is calculated from the rigid-body transform between backbones and
the backbone torsions of the peptide residue, and the hash table is interrogated.
If the key is found in the hash table, side chains with the stored identities and
torsion angles are installed in the docking interface. d, The sequence of the
remainder of the interface is optimized using Rosetta for high-affinity binding.
Two representative designed binding complexes are shown to highlight the
peptide-binding groove and the shape complementarity. The magnified views
illustrate hydrophobic interactions (right), salt bridges (middle) and π–π stacks
(left) incorporated during design.
Nature | Vol 616 | 20 April 2023 | 583
a
6
x
P
L
P
6
x
P
R
L
6
x
W
E
P
6
x
P
Y
I
6
x
M
R
P
6
x
W
K
P
b
c
d
e
)
U
A
(
)
I
(
g
o
l
102
101
100
10–1
10–2
102
101
100
10–1
10–2
102
101
100
10–1
10–2
102
101
100
10–1
10–2
102
101
100
10–1
10–2
102
101
100
10–1
10–2
Experimental
FoXS χ2 = 1.95
0 0.05 0.10 0.15 0.20 0.25 0.30
Experimental
FoXS χ2 = 1.16
0 0.05 0.10 0.15 0.20 0.25 0.30
Experimental
FoXS χ2 = 1.48
0 0.05 0.10 0.15 0.20 0.25 0.30
Experimental
FoXS χ2 = 1.62
)
g
e
d
m
(
l
i
a
n
g
s
D
C
0 0.05 0.10 0.15 0.20 0.25 0.30
Experimental
FoXS χ2 = 0.385
0 0.05 0.10 0.15 0.20 0.25 0.30
Experimental
FoXS χ2 = 0.751
100
50
0
−50
50
0
−50
50
0
−50
50
0
−50
50
0
−50
50
0
−50
20 °C
95 °C
20 °C recovery
200
220
240
200
220
240
200
220
240
)
m
n
(
l
a
n
g
s
i
t
e
t
c
O
200
220
240
200
220
240
Kd < 0.5 nM
20 nM
1,300 2,600 3,900 5,200
Kd ≈ 2.5 nM
39 nM
800 1,400 2,000 2,600
Kd ≈ 5.0 nM
47 nM
500
800
1,100 1,400
Kd ≈ 25.0 nM
78 nM
800
1,400 2,000 2,600
Kd ≈ 31.0 nM
78 nM
800 1,400
2,000
2,600
Kd > 40.0 nM
78 nM
4
2
0
3
2
1
0
2
1
0
1.5
1.0
0.5
0
1.5
1.0
0.5
0
0.4
0.2
0
0 0.05 0.10 0.15 0.20 0.25 0.30
200
220
240
600
800
1,000
1,200
q (Å–1)
Wavelength (nm)
Time (s)
Fig. 2 | Biophysical characterization of designed protein–peptide
complexes. a, Computational models of the designed six-repeat version of
protein–peptide complexes. Designed proteins are shown in cartoons and
peptides in sticks. b, Magnified views for single designed protein–peptide
interaction units. Residues interacting across the interface are shown in sticks.
c, Predicted SAXS profiles overlaid on experimental SAXS data points. The
scattering vector q is on the x axis (from 0 to 0.25) and the intensity (I) is on the
y axis on a logarithmic scale. AU, arbitrary units. d, Circular dichroism (CD)
spectra at different temperatures (blue, 20 °C; orange, 95 °C; green, 95 °C
followed by 20 °C). e, Bio-layer interferometry characterization of the binding
of designed proteins to the corresponding peptide targets. Twofold serial
dilutions were tested for each binder and the highest concentration is labelled.
The biotinylated target peptides were loaded onto streptavidin biosensors,
and incubated with designed binders in solution to measure association and
dissociation.
those designs with converged peptide backbones (root-mean-square
deviation (RMSD) < 2.0 between the 20 lowest ddG designs) close to
the design model (RMSD < 1.5) (Extended Data Fig. 1d).
We tested 54 second-round designed protein–peptide pairs using
the yeast flow cytometry assay described above. Forty-two of the
designed proteins were solubly expressed in E. coli, and 16 bound their
targets with considerably higher affinity and specificity than in the first
round (Extended Data Table 1b). We selected six designs with diverse
superhelical parameters and shapes, and a range of target peptides for
more detailed characterization (Fig. 2). As evident in the design models
(Fig. 2a), there is a one-to-one match between the six repeat units in
the protein and in the target peptide (Fig. 2b shows a single unit inter-
action). Small-angle X-ray scattering (SAXS) profiles28,29 were close to
those computed from the design models, suggesting that the proteins
fold into the designed shapes in solution (Fig. 2c and Extended Data
Table 2b). Circular dichroism studies showed that all six were largely
helical and thermostable up to 95 °C (Fig. 2d). Bio-layer interferometry
characterization of binding to biotinylated target peptides immobilized
on Octet sensor chips revealed Kd values ranging from less than 500 pM
(below the instrument level of detection) to around 40 nM; five out
of six had a dissociation half-life of at least 500 s, and for three of the
six there was little dissociation after 2,000 s (Fig. 2e; little decrease in
binding was observed after storage of the proteins for 30 days at 4 °C,
Extended Data Fig. 2). The binding surfaces of several related designs
were subjected to site-saturation mutagenesis (SSM)30 on yeast, and
after the incorporation of one to three enriched substitutions, binding
was observed by flow cytometry using only 10 pM biotinylated cognate
peptide (Extended Data Fig. 3).
Many cell biology approaches31 involve tagging cellular target pro-
teins with a protein or peptide, and then introducing into the same
cell a protein that binds the tag with high affinity and specificity, but
does not bind endogenous targets. A bottleneck in such studies is that
binders obtained from antibody scaffold (scFV or VHH)-based library
screens often do not fold properly in the reducing environment of the
cytosol, resulting in a loss of binding32. We reasoned that our binders
would not have this limitation as they are designed for stability and lack
disulfide bonds. As a proof of concept, we co-expressed the peptide
PLPx6 fused to GFP and its cognate binder, RPB_PLP2_R6, a variant of
RPB_PLP1_R6, fused to both mScarlet and a targeting sequence for the
mitochondrial outer membrane (Fig. 3a). (In the naming convention
584 | Nature | Vol 616 | 20 April 2023
Article
Control
Target peptide
GFP
Target peptide
+ Mito-Tag
Binder
GFP
mScarlet
Cytosolic GFP signal
GFP relocalization to mitochondria
PLPx6–GFP control
c
PLPx6–GFP
PLPx6–GFP and
Mito–RPB_PLP2_R6–mScarlet
IRPx6–GFP control
e
IRPx6–GFP
IRPx6–GFP and
Mito–RPB_LRP2_R6_FW6–mScarlet
PLPx6
GFP
IRPx6
mScarlet
+ Mito-Tag
RPB_PLP2_R6
+ PEX-Tag
RPB_LRP2_R6_FW6
+
=
GFP relocalization
to mitochondria
mScarlet relocalization
to peroxisomes
Multiplexed relocalization
PLPx6–GFP
IRPx6–mScarlet
Merge
a
b
d
f
g
,
6
R
_
2
P
L
P
_
B
P
R
–
o
t
i
M
P,
F
G
–
6
x
P
L
P
6
W
F
_
6
R
_
2
P
R
L
_
B
P
R
–
X
E
P
,
t
e
l
r
a
c
S
m
–
6
x
P
R
I
Fig. 3 | Designed binders function in living cells. a, Experimental design.
U2OS cells co-express the target peptide fused to GFP and a fusion between
the specific binder fused to mScarlet and a mitochondria-targeting sequence
(Mito-Tag). If binding occurs in cells, the GFP signal is relocalized to the
mitochondria, whereas control cells that do not express the binder show a
cytosolic GFP signal. b–e, In vivo binding. Live, spreading U2OS cells expressing
PLPx6–GFP alone (b), IRPx6–GFP alone (d), PLPx6–GFP and Mito–RPB_PLP2_
R6–mScarlet (c) or IRPx6–GFP and Mito–RPB_LRP2_R6_FW6–mScarlet (e) were
imaged by spinning disk confocal microscopy (SDCM). Note that the GFP signal
is cytosolic in the control but relocalized to the mitochondria after co-expression
with the respective binder. f,g, In vivo multiplexing. f, Experimental design.
U2OS cells co-express two target peptides, one fused to GFP and the other to
mScarlet, and their corresponding specific binder fused to mitochondria- or
peroxisome-targeting sequences. If orthogonal binding occurs, GFP and
mScarlet signals should not overlap. g, Live, spreading U2OS cells co-expressing
PLPx6–GFP, IRPx6–mScarlet, Mito–RPB_PLP2_R6 and PEX–RPB_LRP2_R6_FW6
imaged by SDCM. Note the absence of overlap between channels. Images
correspond to maximum intensity z-projections (Δz = 6 µm). Dashed line
indicates the cell outline. Scale bars, 10 µm.
here and throughout the remainder of the text, ‘RPB’ indicates ‘repeat
peptide binder’; ‘PLP’ indicates the intended peptide specificity (for
proline-leucine-proline in this case); ‘2’ indicates the specific mod-
ule designed to bind this peptide unit; and ‘R6’ indicates the number
(six) of repeat units. In peptide names, the sequence ‘PLP’ is followed
by the number of repeats ‘x6’. In protein–peptide complex descrip-
tors, the protein name is specified first, followed by a dash and then the
peptide name.) Although the PLPx6 peptide on its own was diffuse
in the cytosol (Fig. 3b), after co-expression with the binder, it was
relocalized to the mitochondria (Fig. 3c and Extended Data Fig. 2b).
Thus, the PLPx6–RPB_PLP2_R6 pair retains binding activity in cells.
Similar results were obtained for IRPx6–GFP and RPB_LRP2_R6_FW6
(Fig. 3d,e).
If individual repeat units on the designed protein engage indi-
vidual repeat units on the target peptide, the binding affinity should
increase when the number of repeats is increased. We investigated
this with four of our designed systems—in two cases varying the
number of protein repeats while keeping the peptide constant, and
in the other two cases, varying the number of peptide repeats while
keeping the protein constant. Six-repeat versions of RPB_LRP2_R6
and RPB_PEW2_R6 had a higher affinity for eight-repeat LRP and
PEW peptides than did four-repeat versions, without any decrease in
specificity (Extended Data Fig. 4a). Similarly, six-repeat IYP and PLP
peptides had a higher affinity for six-repeat versions of the cognate
designed repeat proteins (RPB_IYP1_R6 and RPB_PLP1_R6) than did
four-repeat versions (Extended Data Fig. 4b). These results are consist-
ent with a one-to-one modular interaction between repeat units on the
protein and repeat units on the peptide, and suggest that a very high
binding affinity could be achieved simply by increasing the number
of interacting repeat units. This ability to vary the affinity by varying
the number of repeats could be useful in many contexts in which com-
petitive binding would be advantageous. For example, when isolating
proteins by affinity purification, a peptide with a larger number of
repeats than that fused to the protein being expressed could be used
for elution.
High-resolution structural validation
To assess the structural accuracy of our design method, we used X-ray
crystallography. We obtained high-resolution co-crystal structures of
three first-round designs (RPB_PEW3_R4–PAWx4, RPB_LRP2_R4–LRPx4,
RPB_PLP3_R6–PLPx6) and one second-round design (RPB_PLP1_R6–
PLPx6) (Fig. 4); and a crystal structure of the unbound first-round
design RPB_LRP2_R4 (Extended Data Fig. 5a; interface side-chain RMSD
values for all crystal structures are in Extended Data Table 2a). In the
crystal structure of RPB_PLP3_R6–PLPx6, the PLP units fit exactly into
the designed curved groove formed by repeating tyrosine, alanine and
tryptophan residues, matching the design model with near atomic
accuracy (Cα RMSD for protein, protein–peptide interface and full
complex: 1.70 Å, 2.00 Å and 1.64 Å, respectively; Fig. 4b and Extended
Data Fig. 5b). In the co-crystal structure of RPB_PEW3_R4–PAWx4, as
in the design model, the PAW units bind to a relatively flat groove
formed by repeating histidine residues and glutamine residues, as
designed (Fig. 4a and Extended Data Fig. 5c, RMSD 2.08 Å between
design and crystal structure over the protein, median RMSD 2.12 Å
over the peptide and interface between crystal and docked peptide
ensemble; Extended Data Table 2a). For RPB_LRP2_R4–LRPx4, flexible
backbone docking converged with the LRP units fitting in between
repeating glutamine residues and phenylalanine residues as designed,
and the peptide arginine side chain sampling two distinct states associ-
ated with parallel and antiparallel protein-binding modes (Extended
Data Fig. 4c). The lowest-energy docked structure was close to the
crystal structure, with Cα RMSD values of 1.15 Å, 0.98 Å and 1.16 Å for
the protein alone, the peptide plus interface and the entire complex,
respectively (Fig. 3c and Extended Data Table 2a). SSM interface foot-
printing results were consistent with the design model and crystal
structure (Extended Data Fig. 6), and an Phe-to-Trp substitution that
increases interactions across the interface substantially increased the
affinity (Extended Data Fig. 3d).
Nature | Vol 616 | 20 April 2023 | 585
a
a
b
b
c
c
RPB_PEW3_R4–PAWx4
RPB_PLP3_R6–PLPx6
RPB_LRP2_R4–LRPx4
d
f
e
90º
6
x
P
L
P
–
6
R
_
1
P
L
P
_
B
P
R
RPB_PLP1_R6–PLPx6
g
RPB_PLP1_R6–PLPx6
Magnified front view
RPB_PLP1_R6–PLPx6
Magnified back view
Fig. 4 | Evaluation of design accuracy by X-ray crystallography.
a–c, Superposition of computational design models (coloured) on
experimentally determined crystal structures (yellow). a, RPB_PEW3_R4–PAWx4.
b, RPB_PLP3_R6–PLPx6. c, RPB_LRP2_R4–LRPx4. d–g, RPB_PLP1_R6–PLPx6,
d, Overview of the superimposition of the computational design model
and the crystal structure. e, A 90° rotation of d. The complex is shown in
surface mode (protein in orange and peptide in yellow) to highlight the shape
complementarity. f, Zoom in on the internal three units from d (front view).
Glutamine residues from the protein in both the design and the crystal structure
are shown as sticks to highlight the accuracy of the designed side-chain-to-
backbone bidentate ladder. g. View from the side opposite to f. Tyrosine
residues from the protein in both the design and the crystal structure are
shown as sticks to highlight the accuracy of the designed polar interactions.
The 2.15-Å crystal structure of the second-round design RPB_PLP1_
R6–PLPx6 highlights key features of the computational design pro-
tocol. The PLPx6 peptide binds to the slightly curved groove mainly
through polar interactions from tyrosine, hydrophobic interactions
from valine and side-chain–backbone bidentate hydrogen bonds from
glutamine, exactly as designed (Fig. 4d–g; RMSD 1.11 Å for the protein–
peptide interface and 1.91 Å for the complex). All interacting side chains
from both the protein side and the peptide side in the computational
design model are nearly perfectly recapitulated in the crystal struc-
ture. This design has near-picomolar binding affinity (Fig. 2d) and high
specificity for the PLP target sequence (Fig. 5a).
We next investigated the specificity of the six designs (Fig. 5a). The
PLPx6, LRPx6, PEWx6, IYPx6 and PKWx6 binders showed almost com-
plete orthogonality in the concentration range from around 5 nM to
40 nM, with each design binding its cognate designed repeat peptide
much more strongly than the other repeat peptides. For example,
PLPx6 binds RPB_PLP1_R6 strongly at 5 nM, but shows no binding sig-
nal to RPB_IYP1_R6 at 40 nM, whereas PEWx6 binds RPB_PEW1_R6 but
not RPB_PKW1_R6 at 20 nM. Some cross-talk was observed between
the PRMx6 and LRPx6 binders, perhaps involving the arginine resi-
due, which makes cation–pi interactions in both designs. We observe
similar interaction orthogonality in cells: the IRPx6 and PLPx6 bind-
ers specifically direct the localization of their cognate peptides
to different compartments when co-expressed in the same cells
(Fig. 3e,f).
586 | Nature | Vol 616 | 20 April 2023
As described thus far, our approach enables the specific binding
of peptides with perfectly repeating structures. To go beyond this
limitation and enable a much wider range of non-repeating pep-
tides to be targeted, we investigated the redesign of a subset of the
peptide-repeat-unit binding pockets to change their specificity. We
broke the symmetry in the designed repetitive binding interface by
redesigning both protein and peptide in one or more repeats of six-
repeat complexes; the rest of the interface was kept untouched to
maintain the binding affinity. After redesign, the peptide backbone
conformation was optimized by Monte Carlo resampling and rigid-body
optimization (see Methods). Designs were selected for experimental
characterization as described above, favouring those for which the
new design had a lower binding energy for the new peptide than the
original peptide.
We redesigned the PLPx6 binder RPB_PLP3_R6 to bind two PEP units
in the third and fourth positions (target binding sequence PLPPLPPEP-
PEPPLPPLP or, more concisely, PLP2PEP2PLP2). The redesigned protein,
called RPB_hyb1_R6, bound the redesigned peptide considerably more
tightly in Octet experiments, whereas the original design favoured
the previous perfectly repeating sequence, resulting in nearly complete
orthogonality (Fig. 5b). We next designed another hybrid starting from
the RPB_IYP1_R6–IYPx6 complex, in which we changed three of the IYP
units to RYP to generate IYP3RYP3, and redesigned the corresponding
binding pockets. The new design, RPB_hyb2_R6, selectively bound the
intended cognate target as well (Fig. 5b). We measured the binding of all
four proteins against all four peptides, and observed high specificity of
the designed repeat proteins for their intended peptide targets (Fig. 5b).
Generalization to native disordered regions
The ability to design hybrid binders against non-repetitive sequences
opens the door to the de novo design of binders against endogenous
proteins. Intrinsically disordered regions have been very difficult to
specifically target using other approaches, but are in principle good
targets, because binding is not complicated by folding. As a proof of
concept, we focused on human ZFC3H1, a 226-kDa protein that together
with MTR4 forms the heterotetrameric poly(A) tail exosome targeting
(PAXT) complex, which directs a subset of long polyadenylated poly(A)
RNAs for exosomal degradation33,34 (Fig. 6a). We designed binders
against ZFC3H1 residues 594–620 (PLP4PEDPEQPPKPPF), which lie
within an approximately 100-residue disordered region (Fig. 6a), by
extending both the protein and the peptide in the PLPx4 designed
complex. On the peptide side, we kept the (PLP)x4 backbone fixed,
and used Monte Carlo sampling with Ramachandran map biases to
model the remaining sequence (PEDPEQPPKPPF); on the protein side,
we extended the PLPx4 design with four additional repeats, designed
binding interactions with each peptide conformer and selected eight
designs for experimental characterization. These eight designs were
expressed, and seven were found to bind the extended target pep-
tide by bio-layer interferometry (Extended Data Fig. 7a). The two
highest-affinity designs—αZFC-high and αZFC-low—were found by
fluorescence polarization to have Kd values of less than 200 nM and
around 1.2 µM, respectively (Fig. 6b,c), somewhat weaker than the syn-
thetic constructs described above. Nevertheless, αZFC-high co-eluted
with a 103-amino-acid segment of the disordered region of ZFC3H1
containing the targeting sequence by size-exclusion chromatography
(SEC) (Fig. 6d), demonstrating that the binder can recognize the target
peptide in the context of a larger protein. αZFC-high specifically pulled
down the endogenous ZFC3H1 from human cell extracts when assessed
by western blot with established antibodies (Fig. 6e, top), whereas
αZFC-low—which has a similar size and surface composition—did not;
αZFC-low hence provides a control for non-specific association (see
Extended Data Fig. 7b for replicates, and Fig. 6f for independent identi-
fication of ZFC3H1 by mass spectrometry). Mass spectrometry revealed
that MTR4 was enriched in the αZFC-high pull-down, demonstrating
Article RPB_PLP1_R6
RPB_LRP1_R6
RPB_PEW1_R6
RPB_IYP1_R6
RPB_PRM1_R6
RPB_PKW1_R6
a
5 nM
5 nM
20 nM
20 nM
40 nM
40 nM
b
6
x
P
L
P
6
x
P
R
L
6
x
W
E
P
6
x
P
Y
I
6
x
M
R
P
6
x
W
K
P
RPB_PLP3_R6–PLPx6
R6ST3–LRPx6
R6PO11–PLPx6
R6n11–PEWx6
R602–IYPx6
1.50
1.25
1.00
0.75
0.50
0.25
0
R6CP33–PRMx6
R6M4–PKWx6
PLPx6
PLP2PEP2PLP2
IYPx6
IYP3RYP3
47 nM
1 μM
3 μM
3 μM
RPB_hy1–PLP2PEP2PLP2
300 nM
1 μM
3 μM
3 μM
RPB_IYP1_R6–IYPx6
l
a
n
g
s
i
t
e
t
c
O
d
e
z
i
l
a
m
r
o
N
3 μM
3 μM
556 nM
556 nM
RPB_hy2–IYP3RYP3
3 μM
3 μM
556 nM
556 nM
Fig. 5 | Designed protein–peptide interaction specificity. a, Left, to assess
the cross-reactivity of each designed peptide binder in Fig. 2 with each target
peptide, biotinylated target peptides were loaded onto bio-layer interferometry
streptavidin sensors and allowed to equilibrate, and the baseline signal was set
to zero. The bio-layer interferometry tips were then placed into a solution
containing proteins at the indicated concentrations for 500 s and washed with
buffer, and dissociation was monitored for another 500 s. The heat map shows
the maximum signal for each binder–target pair (cognate and non-cognate)
normalized by the maximum signal of the cognate designed binder–target pair.
Right, surface shape complementarity of the cognate complexes. The peptides
are in sphere representation. b, Modular pocket sequence redesign generates
binders for peptide sequences that are not strictly repeating. Left, ribbon
Time
diagrams of base designs (rows 1 and 3) and versions with a matching subset of
the protein and peptide modules redesigned. The ribbon diagrams show the
cognate designed and redesigned assemblies; for example, the first row shows
a six-repeat PLP binding design in complex with PLPx6, and the second row the
same backbone with repeat units 3 and 4 redesigned to bind PEP instead of PLP,
in complex with a PLP2PEP2PLP2 peptide. The redesigned peptide and protein
residues are shown in purple sticks and yellow, respectively. Right, orthogonality
matrix. Biotinylated target peptides were loaded onto biosensors, and
incubated with designed binders in solution at the indicated concentrations.
Red rectangle boxes indicate cognate complexes. Octet signal was normalized
by the maximum signal of the cognate designed binder–target pair.
that the binder can recognize the native PAXT complex in a physiologi-
cal context. We also detected in the αZFC-high pull-down, but not in the
αZFC-low pull-down, other binding partners of ZFC3H1 that are present
in the Bioplex 3.0 interactome in multiple cell lines (for example, BUB3
and ZN207)16–18,35,36, and several RNA-binding proteins that probably
associate with PAXT–RNA assemblies (Fig. 6f; see Source Data for the
full proteomics dataset).
Conclusion
Our results show that by matching superhelical parameters between
repeating-protein and repeating-peptide conformations, and incor-
porating specific hydrogen-bonding and hydrophobic interactions
between matched protein and peptide repeats, we can now design
modular proteins that bind to extended peptides with high affinity
and specificity. The strategy should be generalizable to a wide range
of repeating-peptide structures, and the ability to break symmetry by
redesigning individual repeat units opens the door to more general
peptide recognition. Our approach complements existing efforts to
achieve general peptide recognition by redesigning naturally occurring
repeat proteins; an advantage of our method is that a much broader
range of protein conformations and binding-site geometries can be
generated by de novo protein design than by starting with a native
protein backbone. Proteins that bind to repeating or nearly repeating
sequences could have applications as affinity reagents for diseases that
are associated with repeat expansions, such as Huntington’s disease.
Similarly, rigid fusion of protein modules designed to recognize differ-
ent di-, tri- and tetrapeptide sequences, using the approach described
here, provides an avenue to achieving sequence-specific recognition
of entirely non-repeating sequences. The ability to design specific
binders for proteins that contain large disordered regions—shown
here by the specific pull-down of the PAXT complex (Fig. 6)—should
help to unravel the functions of this important but relatively poorly
understood class of proteins, and should reduce our reliance on animal
immunization to generate antibodies, which can also suffer from repro-
ducibility issues. The affinity of around 100 nM that we attained for
Nature | Vol 616 | 20 April 2023 | 587
a
Disordered region containing the target peptide (103 amino acids)
Target peptide (24 amino acids)
...LPPPPQVSSLPPLSQPYVEGLCVSLEPLPPLPPLPPLPPEDPEQPPKPPFADEEEEEEMLLREELLKSLANKRAFKPEETSSNSDPPSPPVLNNSHPVPRSNL...
1
EDGEI
594
ZFC3H1
1
EDGEI
C
C
ZN
MTR4
MTR4
ZN
TPR
1989
TPR
1989
b
e
kDa
250 -
50 -
250 -
150 -
100 -
75 -
50 -
37 -
25 -
20 -
15 -
Inputs
w
o
l
-
C
F
Z
α
i
h
g
h
-
C
F
Z
α
l
o
r
t
n
o
C
His
pull-down
w
o
l
-
C
F
Z
α
i
h
g
h
-
C
F
Z
α
l
o
r
t
n
o
C
ZFC3H1
Tubulin
Specifically
enriched
proteins
-αZFC-high
-αZFC-low
Coomassie
αPLPx6
αZFC-low
αZFC-high
10–10
10–8
10–6
Concentration (M)
d
)
U
A
(
m
n
0
8
2
D
O
160
140
120
100
80
60
40
20
0
c
300
250
200
150
100
)
U
A
(
n
o
i
t
a
z
i
r
a
o
P
l
f
αZFC-high + GFP–ZFC3H1
disordered region
αZFC-high
GFP–ZFC3H1 disordered region
5
7
11
15
9
Elution volume (ml)
13
17
19
Proteins identified
9
34
199
1,249
16
6
77
Beads only
αZFC-low
negative control
αZFC-high
Protein
Description
Control
αZFC-low
αZFC-high
ZFC3H1
PAXT complex
0
0
MTR4
BUB3
PAXT complex
8 (11%)
5 (5.8%)
Mitotic checkpoint
3 (13%)
3 (13%)
ZN207
Mitotic checkpoint
2 (2.7%)
3 (5.4%)
RBM12
RNA processing
4 (3.8%)
6 (6.3%)
RBM26
RNA processing
3 (3.6%)
2 (3.7%)
27 (19%)
35 (37%)
23 (84%)
13 (14%)
43 (43%)
47 (42%)
Fig. 6 | Design of binders to disordered regions of endogenous human
proteins. a, Schematic model of the human PAXT complex composed of a
heterotetramer of ZFC3H1 and MTR4. CC, coiled-coil domain; ZN, Zn-finger
domain. Inset shows the sequence environment of the target sequence.
b, Surface shape complementarity between the target peptide from ZFC3H1
(sphere) and the highest-affinity cognate binder, αZFC-high. c, Fluorescence
polarization binding curves between the indicated ZFC3H1 binders and the
target ZFC3H1 peptide (PLP)4PEDPEQPPKPP. As a negative control, we used the
(PLP)x6 binder, RPB_PLP3_R6 (see Fig. 4). αZFC-high shows a higher binding
affinity to the target peptide than αZFC-low, in contrast with RPB_PLP3_R6,
which shows negligible binding. d, Superdex 200 10/300 GL SEC profiles of
purified αZFC-high, a fusion between GFP and a 103-amino-acid fragment of
the disordered region of ZFC3H1 containing the target sequence (see a), or a
1:1 mix of the two after two hours of incubation. OD280 nm, optical density at
280 nm. e, Top, HeLa cell extracts were subjected to pull-down using the
indicated binders bound to Ni-NTA agarose beads, or naked beads as a control.
Recovered proteins were processed for western blot against endogenous
ZFC3H1 (or tubulin as a loading control). Bottom, Coomassie-stained SDS–
PAGE gel of the samples analysed at the top. These panels are representative
of n = 3 experiments. f, Proteomic analysis of the His-pull-down samples shown
in e. Top, overlap between the proteins identified, setting a threshold of five
peptides for correct identification. Bottom, examples of proteins identified
(number indicates exclusive peptide count; protein coverage is indicated in
parentheses). See Source Data for the full dataset. For gel source data, see
Supplementary Fig. 1.
this endogenous binder is compatible with other cellular applications,
such as enzyme targeting for specific post-translational modifications
in vivo18,35,36, or for imaging probes, in which a trade-off must always be
found between high-affinity interactions for labelling specificity and
low-affinity interactions to avoid perturbing protein function37,38. More
generally, our results reveal the power of computational protein design
for targeting peptides and intrinsically disordered regions that do not
have rigid three-dimensional structures. Because the designed proteins
are expressed at quite high levels and are very stable, we anticipate
that these and further designs for a wider range of target sequences
will have many uses in proteomics and other applications that require
specific peptide recognition.
Online content
Any methods, additional references, Nature Portfolio reporting summa-
ries, source data, extended data, supplementary information, acknowl-
edgements, peer review information; details of author contributions
and competing interests; and statements of data and code availability
are available at https://doi.org/10.1038/s41586-023-05909-9.
588 | Nature | Vol 616 | 20 April 2023
Article
1.
London, N., Movshovitz-Attias, D. & Schueler-Furman, O. The structural basis of peptide–
protein binding strategies. Structure 18, 188–199 (2010).
2. Neduva, V. et al. Systematic discovery of new recognition peptides mediating protein
24. Coventry, B. & Baker, D. Protein sequence optimization with a pairwise decomposable
penalty for buried unsatisfied hydrogen bonds. PLoS Comput. Biol. 17, e1008061
(2021).
interaction networks. PLoS Biol. 3, e405 (2005).
25. Boder, E. T. & Wittrup, K. D. Yeast surface display for screening combinatorial polypeptide
3. Neduva, V. & Russell, R. B. Peptides mediating interaction networks: new leads at last.
libraries. Nat. Biotechnol. 15, 553–557 (1997).
4.
Curr. Opin. Biotechnol. 17, 465–471 (2006).
Ernst, P. & Plückthun, A. Advances in the design and engineering of peptide-binding repeat
proteins. Biol. Chem. 398, 23–29 (2017).
26. Kortemme, T. & Baker, D. A simple physical model for binding energy hot spots in protein–
protein complexes. Proc. Natl Acad. Sci. USA 99, 14116–14121 (2002).
27. Kortemme, T., Kim, D. E. & Baker, D. Computational alanine scanning of protein–protein
5. Andrade, M. A., Petosa, C., O’Donoghue, S. I., Müller, C. W. & Bork, P. Comparison of ARM
interfaces. Sci. STKE 2004, pl2 (2004).
and HEAT protein repeats. J. Mol. Biol. 309, 1–18 (2001).
28. Hura, G. L. et al. Comprehensive macromolecular conformations mapped by quantitative
6. Reichen, C. et al. Computationally designed armadillo repeat proteins for modular peptide
SAXS analyses. Nat. Methods 10, 453–454 (2013).
recognition. J. Mol. Biol. 428, 4467–4489 (2016).
29. Hura, G. L. et al. Robust, high-throughput solution structural analyses by small angle X-ray
7. Conti, E. & Kuriyan, J. Crystallographic analysis of the specific yet versatile recognition of
distinct nuclear localization signals by karyopherin α. Structure 8, 329–338 (2000).
scattering (SAXS). Nat. Methods 6, 606–612 (2009).
30. Siloto, R. M. P. & Weselake, R. J. Site saturation mutagenesis: methods and applications in
8. Conti, E., Uy, M., Leighton, L., Blobel, G. & Kuriyan, J. Crystallographic Analysis of the
protein engineering. Biocatal. Agric. Biotechnol. 1, 181–189 (2012).
recognition of a nuclear localization signal by the nuclear import factor karyopherin α.
Cell 94, 193–204 (1998).
Zeytuni, N. & Zarivach, R. Structural and functional discussion of the tetra-trico-peptide
repeat, a protein interaction module. Structure 20, 397–405 (2012).
9.
31. Helma, J., Cardoso, M. C., Muyldermans, S. & Leonhardt, H. Nanobodies and recombinant
binders in cell biology. J. Cell Biol. 209, 633–644 (2015).
32. Moutel, S. et al. NaLi-H1: a universal synthetic library of humanized nanobodies providing
highly functional antibodies and intrabodies. eLife 5, e16228 (2016).
10. D’Andrea, L. D. & Regan, L. TPR proteins: the versatile helix. Trends Biochem. Sci 28, 655–662
33. Foucher, A.-E. et al. Structural analysis of Red1 as a conserved scaffold of the RNA-targeting
(2003).
MTREC/PAXT complex. Nat. Commun. 13, 4969 (2022).
11. Ernst, P. et al. Structure-guided design of a peptide lock for modular peptide binders. ACS
34. Meola, N. et al. Identification of a nuclear exosome decay pathway for processed transcripts.
Chem. Biol. 15, 457–468 (2020).
Mol. Cell 64, 520–533 (2016).
12. Alford, R. F. et al. The Rosetta all-atom energy function for macromolecular modeling and
35. Huttlin, E. L. et al. Dual proteome-scale networks reveal cell-specific remodeling of the
design. J. Chem. Theory Comput. 13, 3031–3048 (2017).
human interactome. Cell 184, 3022–3040 (2021).
13. Hansen, S. et al. Structure and energetic contributions of a designed modular peptide-
binding protein with picomolar affinity. J. Am. Chem. Soc. 138, 3526–3532 (2016).
14. Reichen, C., Hansen, S. & Plückthun, A. Modular peptide binding: from a comparison of
natural binders to designed armadillo repeat proteins. J. Struct. Biol. 185, 147–162 (2014).
15. Fleming, P. J. & Rose, G. D. Do all backbone polar groups in proteins form hydrogen bonds?
Protein Sci. 14, 1911–1917 (2005).
16. Schweppe, D. K., Huttlin, E. L., Harper, J. W. & Gygi, S. P. BioPlex display: an interactive suite
for large-scale AP–MS protein–protein interaction data. J. Proteome Res. 17, 722–726 (2018).
17. Rhys, G. G. et al. De novo designed peptides for cellular delivery and subcellular localisation.
Nat. Chem. Biol. 18, 999–1004 (2022).
18. Brunette, T. J. et al. Exploring the repeat protein universe through computational protein
design. Nature 528, 580–584 (2015).
19. Hicks, D. R. et al. De novo design of protein homodimers containing tunable symmetric
protein pockets. Proc. Natl Acad. Sci. USA 119, e2113400119 (2022).
20. Shimoni, L. & Glusker, J. P. Hydrogen bonding motifs of protein side chains: descriptions
of binding of arginine and amide groups. Protein Sci. 4, 65–74 (1995).
21. Fallas, J. A. et al. Computational design of self-assembling cyclic protein homo-oligomers.
Nat. Chem. 9, 353–360 (2017).
22. Leman, J. K. et al. Macromolecular modeling and design in Rosetta: recent methods and
frameworks. Nat. Methods 17, 665–680 (2020).
23. Kuroda, D. & Gray, J. J. Shape complementarity and hydrogen bond preferences in
protein–protein interfaces: implications for antibody modeling and protein–protein
docking. Bioinformatics 32, 2451–2456 (2016).
36. Ramirez, D. H. et al. Engineering a proximity-directed O-GlcNAc transferase for selective
protein O-GlcNAcylation in cells. ACS Chem. Biol. 15, 1059–1066 (2020).
37. Kumari, A., Kesarwani, S., Javoor, M. G., Vinothkumar, K. R. & Sirajuddin, M. Structural
insights into actin filament recognition by commonly used cellular actin markers. EMBO J.
39, e104006 (2020).
38. Brunette, T. J. et al. Modular repeat protein sculpting using rigid helical junctions. Proc. Natl.
Acad. Sci. USA 117, 8870–8875 (2020).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution
4.0 International License, which permits use, sharing, adaptation, distribution
and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons licence,
and indicate if changes were made. The images or other third party material in this article are
included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your
intended use is not permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a copy of this licence,
visit http://creativecommons.org/licenses/by/4.0/.
© The Author(s) 2023
Nature | Vol 616 | 20 April 2023 | 589
Methods
Generation of DHR scaffolds
Each designed helical repeat (DHR) scaffold is formed by a helix-loop-
helix-loop topology that is repeated four or more times18,35,36. The heli-
ces range from 18 to 30 residues and the loops from 3 to 4 residues. The
DHR design process goes through backbone design, sequence design
and computational validation by energy landscape exploration. To
match the peptides, the designs were required to have a twist (omega)
between 0.6 and 1.0 radians, a radius of 0 to 13 Å and a rise between
0 and 10 Å. The geometry of a repeat protein can be described by the
radius of the super-helix, the axial displacement and the twist37,38.
The backbone is designed using Rosetta fragment assembly
guided by motifs21. Backbone coordinates are built up through 3,200
Monte Carlo fragment assembly steps with fragments taken from a
non-redundant set of structures from the Protein Data Bank (PDB). After
the insertion of each fragment, the rigid-body transform is propagated
to the downstream repeats. The score that guides fragment assembly
is composed of Van der Waal interactions, packing, backbone dihedral
angles and residue-pair-transform (RPX) motifs21. RPX motifs are a
fast way to measure the full-atom hydrophobic packability of the
backbone before assigning side chains. After design, backbones are
screened for native-like features. The loops are required to be within
0.4 Å of a naturally occurring loop or rebuilt. Structures with helices
above 0.14 Å appear bent and kinked and are discarded. And poorly
packed structures in which fewer than four helices are in contact with
each other are filtered.
The sequence is designed using Rosetta for each backbone that
passes filtering. Design begins in a symmetrical mode in which each
repeat is identical using the RepeatProteinRelax mover. Core residues
are restricted to be hydrophobic and surface residues hydrophilic
using the layer design task operators. Sequence is biased toward natu-
ral proteins with a similar local structure using the structure profile
mover. After the symmetrical design is complete, the N-terminal and
C-terminal repeats are redesigned to eliminate exposed hydrophobics.
Designs with poor core packing as measured by Rosetta Holes < 0.5
are then filtered39.
The designs are computationally validated using the Rosetta ab initio
structure prediction on Rosetta@home40. Rosetta ab initio verifies that
the design is a lower-energy state than the thousands of alternative
conformations sampled. Simulating a protein using Rosetta@home
can take several days on hundreds of CPUs. To speed this up, we used
machine learning to filter designs that were most likely to fail37,38.
Backbone generation of curved repeat-protein monomers in
polyproline II conformation
A second round of designs was made to ensure that the distance
between helices matches the 10.9 Å. distance between prolines in
the polyproline II conformation. To design these backbones, we used
atom-pair constraints between the first helix of each repeat. The
atom-pair constraints were set to 10.9 Å with a tolerance of 0.5 Å. For
these designs, we found the topologies that most efficiently produced
structures that matched the atom-pair constraints had a helix length
of 20 or 21 residues and a loop range of three residues.
Design of peptide binders
Modular peptide docking and hashing. To construct hash tables
storing the pre-computed privileged residue interactions, we first
surveyed the non-redundant PDB database and extracted the intended
interacting residues as seeds. For each seeding interaction residue
pair, random perturbations were applied to search for alternative
relative conformations of the interacting residues. In the case of
the side-chain–backbone bidentate interactions, random rigid-body
perturbations were applied to the backbone residues, with a random
set of Euler angles drawn from a normal distribution with 0° as the mean
and 60° as the standard deviation, as well as a random set of transla-
tion distances in three-dimensional (3D) space drawn from a normal
distribution with 0 Å as the mean and 1 Å as the standard deviation. At
the same time, the backbone torsion angles Φ and Ψ of the backbone
residue were randomly modified to values drawn from a Ramachandran
density plot based on structures from the PDB database. The trans-
formed set of residues losing the intended interactions were discarded.
The transformed residues keeping the interactions will be collected.
Then, the side chains of the side-chain residues were replaced with
all reasonable rotamers, to further diversify the samples of the sets
of interacting residues. Finally, the geometry relationship of each set
of residues keeping the intended interactions was subjected to an 8D
hash function (6D rigid-body transformation plus two torsion angles),
and represented with a 64-bit unsigned integer as the key of an entry
in the hash table. The identity and the side-chain torsion angles (Χs)
of the side-chain residues were treated as the value of the entry in the
hash table. Similar processes were used to build different hash tables
for various interactions, with minor alterations. For example, for pi–pi
and cation–pi interactions, only a 6D hash function was used, because
there is no need for the perturbation and consideration of the back-
bone torsions. For Asn, Gln, Asp or Glu interacting with two residues on
the backbone, a 10D hash table was applied for representing the geom-
etry relationship, and, in these cases, the geometries of the N–H and
C=O groups on the backbone were treated as 5D rays.
To sample repeat peptides that match the superhelical parameters
of the DHRs, we randomly generate a set of backbone torsion angles
φ and ψ, for example, [φ1, ψ1, φ2, ψ2, φ3, ψ3] for repeats of tripeptide.
If any pair of φ and ψ angles gets a high Rosetta Ramachandran score
above the threshold of −0.5, it means that this pair of torsion angles
is likely to introduce intra-peptide steric clashes, and in these cases
we randomly regenerate a new pair of φ and ψ angles until they are
reasonable according to the Rosetta Ramachandran score. Next,
we set the backbone torsion angles of the repeat peptide using this set
of φ and ψ angles repetitively across the eight repeats. And we calcu-
late the superhelical parameters using the 3D coordinates of adjacent
repeat units of the repeat peptide. The repeat peptides matching the
superhelical parameters of any one of the curated DHRs are saved for
the docking step.
To dock cognate repeat proteins and repeat peptides, with matching
superhelical parameters, they are first aligned to the z axis by their own
superhelical axes. In the next step, a 2D grid search (rotation around
and translation along the z axis) is carried out to sample compatible
positions of the repeat peptide in the binding groove of the repeat
protein. Once a reasonable dock is generated without steric clash, the
relevant hash function is used to iterate through all potential peptide–
protein interacting residue sets, to calculate the hash keys. If a hash key
exists in the hash table, the interacting side-chain identities and torsion
angles will be pulled out immediately and installed on all equivalent
positions of this repeat peptide–repeat protein docking conformation.
The docked peptide–DHR pair is saved for the interface design step if
the peptide–DHR hydrogen-bond interactions are satisfied.
Design of the peptide-binding interface. If a single dock was accepted
with the designed repetitive peptide–DHR hydrogen bond, the peptide
was first trimmed to the exact same repeat number as the DHR (for
example, four-repeat or six-repeat). After that, for both peptide and
DHR sides, each amino acid was set linked to its corresponding amino
acids on the same position in each repeat unit. This was to make sure
that all of the following design steps would be carried out with the exact
same symmetry inside both the DHR and the peptide.
During our design cycles, the interface neighbour distance is set as
9 Å as the whole designable range around the DHR–peptide binding
interface, and 11 Å as the whole minimization range. Three rounds of
full hydrophobic FastDesign21 followed by hydropathic FastDesign
were carried out, with each hydrophobic or hydrophilic FastDesign
Article repeating twice. The Rosetta score function beta_nov16 was chosen in
all design cycles. In the produced complex, the peptide itself with an
averaged score (three calculations were carried out) larger than 20.0
or a complex score larger than −10.0 were rejected directly.
After the preliminary design was done, we performed two types
of sanity check to further optimize the designed peptide sequence,
as well as the designed DHR interface. Specifically, for the peptide
side, in the tripeptide repeat units, every two amino acids other than
proline were scanned for a possible mutation to all twenty amino acids
except cysteine, unless a certain originally designed peptide amino
acid is making the hashed side-chain–backbone hydrogen bond, or
side-chain–side-chain hydrogen bond, or side-chain–side-chain–
backbone hydrogen bond with the DHR interface. The DDG (binding
energy for the peptide–DHR complex) was compared before and after
this peptide side mutation; and the mutation was accepted if the delta
DDG (DDG_after – DDG_before) was larger than 1.0. Similarly, we also
checked the designed DHR interface by mutation. The whole DHR was
scanned. For the designed hydrophobic amino acids that were originally
hydrophilic, a delta DDG of −5.0 was set as the threshold to be accepted
as a necessary design that made enough binding contribution. For
the designed hydropathic amino acids, a delta DDG of −2.0 was used
as the threshold.
For experimental characterization, we selected designed complexes
with near-ideal bidentate hydrogen bonds between protein and peptide,
favourable protein–peptide interaction energies (DDG ≤ −35.0), inter-
face shape complementarity (Iface_SCval ≥ 0.65), tolerable interface
unsatisfied hydrogen bonds (Iface_HbondsUnsatBB ≤ 2, Iface_Hbonds-
UnsatSC ≤ 4) and low peptide apo energies (ScoreRes_chainB ≤ 0.9).
Forward docking. As for the selected designed complexes from
our round-two experiments, forward docking was performed to
ensure the specificity in silico. For each designed complex, 10,000
arbitrary peptide conformations were generated as above, using the
designed sequence. The same docking protocol was conducted as
described in the docking stage, against the untouched designed DHR.
FastRelax41 was then performed for the 10,000 docks, and the DDG
versus peptide-backbone RMSD was plotted to check the conver-
gence of the complex. Only the ‘converged’ complexes were selected
for experimental characterization; for example, (i) peptide backbone
RMSD < 2.0Å among the top 20 designs with the lowest DDG during
forward docking; and (ii) the averaged peptide backbone of the top 20
designs was close to the original design model (RMSD < 1.5 Å).
Preparation of SSM libraries. We performed SSM studies for some
of the designed peptide–protein binding pairs to gain a better under-
standing of the peptide-binding modes, and to search for improved
peptide binders. For each designed repeat protein, we ordered a SSM
library covering the central span of 65 amino acids within the whole
repeat protein, owing to the chip DNA size limitation. This span roughly
equals one and a half repeating units, across three helices. The chip
synthesized DNA oligos for the SSM library were then amplified and
transformed to EBY100 yeast together with a linearized pETCON3
vector including the encoding regions of the rest of the designed repeat
protein. Each SSM library was subjected to an expression sort first, in
which the low-quality sequences due to chip synthesizing defects or
recombination errors were filtered out. The collected yeast population,
which successfully expresses the designed repeat-protein mutants,
will be regrown, and subjected to the next round of peptide-binding
sorts. The next-generation sequencing results of this yeast population
will also serve as the reference data for SSM analysis. The next round
of without-avidity peptide-binding sorts used various concentrations
of the target peptide, depending on the initial peptide-binding abili-
ties, ranging from 1 nM to 1,000 nM. The peptide-bound yeast popula-
tions were collected and sequenced using the Illumina NextSeq kit. The
mutants were identified and compared to the mutants in the expression
libraries. Enrichment analysis was used to identify beneficial mutants
and provide information for interpreting the peptide-binding modes.
For each mutant, its enrichment value is calculated by dividing its ratio
in the peptide-bound population by its ratio in expression population.
The enrichment value is then subjected to a log10 transformation, and
plotted in heat maps for the SSM analysis.
Design of binders against endogenous targets. To evaluate
which endogenous proteins could at present be targeted with our
method (Fig. 6), we developed Python code to search databases for
sub-sequences that match permutations of the set of amino acid triplets
for which we designed binders in this study (that is, LRP PEW PLP IYP
PKW IRP LRT LRN LRQ RRN PSR PRQ). This code can be accessed freely
(https://github.com/tjs23/prot_pep_scan). We then ranked all outputs
to find the longest sub-sequence possible, and manually inspected
the candidates to find sub-sequences landing in disordered regions.
Doing this analysis on the human proteome suggested that ZFC3H1
could be a good target for two main reasons: (1) this protein possesses
the sequence (PLP)x4 within a large disordered domain, with down-
stream sequence (PEDPEQPPKPPF) within the reach of our binder
design method; and (2) this protein is well studied, and—in particular—
commercial, highly specific and validated antibodies exist against it.
Synthetic gene constructs
All genes in this work were ordered from either Integrated DNA Tech-
nologies (IDT) or GenScript. For both the first- and the second-round
designs, a His tag containing a TEV protease cleavage site and short
linkers were added to the N terminus of protein sequences. For the
protein lacking a tryptophan residue, a single tryptophan was added
to the short N-terminal linker following the TEV protease cleavage
site to help with the quantification of protein concentration by
A280. The protein sequence along with the linker (MGSSHHHHHH
HHSSGGSGGLNDIFEAQKIEWHEGGSGGSENLYFQSG or LEHHHHHH)
was reverse-translated into DNA using a custom Python script that
attempts to maximize the host-specific codon adaptation index42 and
IDT synthesizability, which includes optimizing whole-gene and local
GC content as well as removing repetitive sequences. Finally, a TAATCA
stop codon was appended to the end of each gene. Genes were deliv-
ered cloned into pET-29b+ between NdeI and XhoI restriction sites. For
the second-round designs, the designed amino acid sequences were
inserted directly into pET-29b+ between Ndel and Xhol restriction sites.
For the disordered region of ZFC3H1, the 103 amino acids contain-
ing the key targeting sequence (LPPPPQVSSLPPLSQPYVEGLCVSLEPLP
PLPPLPPLPPEDPEQPPKPPFADEEEEEEMLLREELLKSLANKRAFKPEETS
SNSDPPSPPVLNNSHPVPRSNL) was cloned into a customized vector
with sfGFP at the N terminus and His6 at the C terminus with a linker
(GGSGSG) in between.
Protein expression and purification
Proteins were transformed into Lemo21(DE3) E. coli from New England
Biolabs (NEB) and then expressed as 50-ml cultures in 250-ml flasks
using Studiers M2 autoinduction medium with 50 µg ml−1 kanamy-
cin. The cultures were either grown at 37 °C for around 6–8 h and then
around 18 °C overnight (around 14 h), or at 37 °C for the entire time
(around 14 h). Cells were pelleted at 4,000g for 10 min, after which
the supernatant was discarded. Pellets were resuspended in 30 ml lysis
buffer (25 mM Tris-HCl pH 8, 150 mM NaCl, 30 mM imidazole, 1 mM
PMSF, 0.75% CHAPS, 1 mM DNase and 10 mM lysozyme, with Thermo
Fisher Scientific Pierce protease inhibitor tablet). Cell suspensions
were lysed by microfluidizer or sonication, and the lysate was clarified
at 20,000g for around 30 min. The His-tagged proteins were bound to
Ni-NTA resin (Qiagen) during gravity flow and washed with a wash buffer
(25 mM Tris-HCl pH 8, 150 mM NaCl and 30 mM imidazole). Protein
was eluted with an elution buffer (25 mM Tris-HCl pH 8, 150 mM NaCl
and 300 mM imidazole). For the first-round designs, the His tag was
removed by TEV cleavage, followed by IMAC purification to remove
TEV protease. The flowthrough was collected and concentrated before
further purification by SEC or fast-performance liquid chromatography
on a Superdex 200 increase 10/300 GL column in Tris-buffered saline
(TBS; 25 mM Tris pH 8.0 and 150 mM NaCl).
Circular dichroism
Circular dichroism spectra were measured with an AVIV Model 420
DC or Jasco J-1500 circular dichroism spectrometer. Samples were
0.25 mg ml−1 in TBS (25 mM Tris pH 8.0 and 150 mM NaCl), and a 1-mm
path-length cuvette was used. The circular dichroism signal was
converted to mean residue ellipticity by dividing the raw spectra by
N × C × L × 10, in which N is the number of residues, C is the concentra-
tion of protein and L is the path length (0.1 cm).
SEC with multi-angle light scattering
Purified samples after the initial SEC run were pooled then concentrated
or diluted as needed to a final concentration of 2 mg ml−1 and 100 µl of
each sample was then run through a high-performance liquid chroma-
tography system (Agilent) using a Superdex 200 10/300 GL column.
These fractionation runs were coupled to a multi-angle light scattering
detector (Wyatt) to determine the absolute molecular weights for each
designed protein as described previously21.
SAXS
SAXS was collected at the SIBYLS High Throughput SAXS Advanced
Light Source in Berkeley, California43,44. Beam exposures of 0.3 s for
10.2 s resulted in 33 frames per sample. Data were collected at low
(around 1.5 mg ml−1) and high (around 2–3 mg ml−1) protein concentra-
tions in SAXS buffer (25 mM Tris pH 8.0, 150 mM NaCl and 2% glycerol).
The SIBYLS website (SAXS FrameSlice) was used to analyse the data
for high- and low-centration samples and average the best dataset. If
there was obvious aggregation over the 33 frames, only the data points
before aggregation arose were used in the Gunier region; otherwise,
all data were included for the Gunier region. All data were used for the
Porod and Wide regions. The averaged file was used with scatter.jar to
remove data points with outlier residuals in the Gunier region. Finally,
the data were truncated at 0.25 q. This dataset was then compared to
the predicted SAXS profile based on the design model using the FoxS
SAXS server (FoXS Server: Fast X-Ray Scattering n.d.), and the volatil-
ity ratio (Vr) was calculated to quantify how well the predicted data
matched the experimental data. Proteins with a Vr of less than 2.5 were
considered to be folded to the designed quaternary shape.
Bio-layer interferometry
Bio-layer interferometry binding data were collected in an Octet RED96
(ForteBio) and processed using the instrument’s integrated software.
To measure the affinity of peptide binders, N-terminally biotinylated
(biotin-Ahx) target peptides with a short linker (GGS) were loaded onto
streptavidin-coated biosensors (SA ForteBio) at 50–100 nM in bind-
ing buffer (10 mM HEPES (pH 7.4), 150 mM NaCl, 3 mM EDTA, 0.05%
surfactant P20 and 0.5% non-fat dry milk) for 120 s. Analyte proteins
were diluted from concentrated stocks into the binding buffer. After
baseline measurement in the binding buffer alone, the binding kinet-
ics were monitored by dipping the biosensors in wells containing the
target protein at the indicated concentration (association step) and
then dipping the sensors back into baseline buffer (dissociation).
Yeast surface display
Saccharomyces cerevisiae EBY100 strain cultures were grown in
C-Trp-Ura medium and induced in SGCAA medium following the pro-
tocol in ref. 45. Cells were washed with PBSF (phosphate-buffered saline
(PBS) with 1% BSA) and labelled with biotinylated designed proteins
using two labelling methods: with-avidity and without-avidity labelling.
For the with-avidity method, the cells were incubated with biotinylated
RBD, together with anti-Myc fluorescein isothiocyanate (FITC, Miltenyi
Biotec) and streptavidin–phycoerythrin (SAPE, Thermo Fisher Scien-
tific). The SAPE in the with-avidity method was used at one-quarter of
the concentration of the biotinylated RBD. The with-avidity method
was used in the first few rounds of screening against the repeat-peptide
library to fish out weak binder candidates. For the without-avidity
method, the cells were first incubated with biotinylated designed
proteins, washed and then secondarily labelled with SAPE and FITC.
Crystallization and structure determination
RPB_PEW3_R4–PAWx4. Purified RPB_PEW3_R4 protein + PAWx4 pep-
tide at a concentration of 36 mg ml−1 was used to conduct sitting-drop,
vapour-diffusion crystallization trials using the JCSG Core I-IV screens
(NeXtal Biotechnologies). Crystals of RPB_PEW3_R4–PAWx4 grew from
drops consisting of 100 nl protein plus 100 nl of a reservoir solution
consisting of 0.1 M MES pH 5.0 and 30% (w/v) PEG 6000 at 4 °C, and
were cryoprotected by supplementing the reservoir solution with 5%
ethylene glycol. Native diffraction data were collected at APS beamline
23-ID-D, indexed to P212121 and reduced using XDS46 (Supplementary
Table 1). The structure was phased by molecular replacement using
Phaser46. A set of around 50 of the lowest-energy predicted models
from Rosetta were used as search models. Several of these models
gave clear solutions, which were adjusted in Coot47 and refined using
PHENIX48. Model refinement in P212121 initially resulted in unacceptably
high values for Rfree – Rwork. Refinement was therefore first performed in
lower-symmetry space groups (P1 and P21). In the late stages of refine-
ment, these P1 and P21 models were refined against the P212121, which
ultimately yielded acceptable, albeit somewhat higher, R-factors.
RPB_PLP3_R6–PLPx6. Purified RPB_PLP3_R6 protein + PLPx4 peptide
at a concentration of 70 mg ml−1 was used to conduct sitting-drop,
vapour-diffusion crystallization trials using the JCSG Core I-IV screens
(NeXtal Biotechnologies). Crystals of RPB_PLP3_R6-PLPx6 grew from
drops consisting of 100 nl protein plus 100 nl of a reservoir solution
consisting of 2.4 M (NH4)2SO4 and 0.1 M sodium citrate pH 4 at 18 °C,
and were cryoprotected by supplementing the reservoir solution with
2.2 M sodium malonate pH 4. Native diffraction data were collected
at APS beamline 23-ID-D, indexed to I422 and reduced using XDS49
(Supplementary Table 1). The structure was phased by molecular
replacement using Phaser46. A set of around 28 of the lowest-energy
predicted models from Rosetta were used as search models. Several
of these models gave clear solutions, which were adjusted in Coot47
and refined using PHENIX48.
RPB_LRP2_R4–LRPx4. Purified RPB_LRP2_R4 protein + LRPx4 peptide
at a concentration of 21.4 mg ml−1 was used to conduct sitting-drop,
vapour-diffusion crystallization trials using the JCSG Core I-IV screens
(NeXtal Biotechnologies). Crystals of RPB_LRP2_R4–LRPx4 grew from
drops consisting of 100 nl protein plus 100 nl of a reservoir solution
consisting of 0.1 M HEPES pH 7 and 10% (w/v) PEG 6000 at 18 °C, and
were cryoprotected by supplementing the reservoir solution with 25%
ethylene glycol. Native diffraction data were collected at APS beamline
23-ID-B, indexed to P32 2 1 and reduced using XDS49 (Supplementary
Table 1). The structure was phased by molecular replacement using
Phaser46. The coordinates of apo-RPB_LRP2_R4 from the proteolysed or
filament structure were used as a search model. The resulting model was
adjusted in Coot47 and refined using PHENIX48. Like the apo structure,
this crystal structure of RPB_LRP2_R4 also contained infinitely long
filaments in the crystal, this time with peptide bound.
RPB_PLP1_R6–PLPx6. Purified RPB_PLP1_R6 protein + PLPx6 peptide
at a concentration of 143 mg ml−1 was used to conduct sitting-drop,
vapour-diffusion crystallization trials using the JCSG Core I-IV screens
(NeXtal Biotechnologies). Crystals of RPB_PLP1_R6–PLPx6 grew from
drops consisting of 100 nl protein plus 100 nl of a reservoir solution
consisting of 0.2 M NaCl and 20% (w/v) PEG 3350 at 4 °C, and were cryo-
protected by supplementing the reservoir solution with 15% ethylene
glycol. Native diffraction data were collected at APS beamline 23-ID-B,
Article indexed to H32 and reduced using XDS49 (Supplementary Table 1). The
structure was phased by molecular replacement using Phaser46. A set of
around 230 of the lowest-energy predicted models from Rosetta were
used as search models. Several of these models gave clear solutions,
which were adjusted in Coot47 and refined using PHENIX48. In the later
stages of refinement, two copies of the 6xPLP peptide were built into
clearly defined electron density in the asymmetrical unit. The first
copy adopts the expected location based on the design, and makes the
designed interactions with RPB_PLP1_R6. The density for this peptide
and the final atomic model (19 amino acid residues) are slightly longer
than the peptide used in crystallization (18 residues); this is probably
due to ‘slippage‘ or misregistration of the peptide relative to the R6PO11
in many unit cells, resulting in density longer than the peptide itself.
A second copy of the peptide lies across a twofold symmetry axis at
around 50% occupancy, resulting in the superposition of this peptide
with a symmetry-derived copy of itself running in the opposite direc-
tion. Despite this, the locations of each Pro or Leu side-chain unit were
reasonably well defined. However, it seems unlikely that the binding of
the peptide at this second site would occur readily in solution.
RPB_PLP1_R6, alternative conformation 1. Purified RPB_PLP1_R6
protein + PLPx6 peptide at a concentration of 166 mg ml−1 was used
to conduct sitting-drop, vapour-diffusion crystallization trials using
the JCSG Core I-IV screens (NeXtal Biotechnologies). Crystals of
RPB_PLP1_R6-PLPx6 grew from drops consisting of 100 nl protein plus
100 nl of a reservoir solution consisting of 0.02 M CaCl2, 30% (v/v) MPD
and 0.1 M sodium acetate pH 4.6 at 18 °C, and were cryoprotected by
supplementing the reservoir solution with 5% MPD. Native diffrac-
tion data were collected at APS beamline 23-ID-B, indexed to P22121
and reduced using XDS49 (Supplementary Table 1). The structure was
phased by molecular replacement using Phaser46, using the coordinates
for R6PO11 (alternative conformation 1) as a search model. The model
was adjusted in Coot47 and refined using PHENIX48. In the later stages
of refinement, one copy of the 6xPLP peptide was model at a site of
crystal contact, where it is sandwiched between adjacent subunits in
a way that is likely to only be bound in the crystal lattice.
RPB_PLP1_R6, alternative conformation 2. Purified RPB_PLP1_R6
protein + PLPx6 peptide at a concentration of 166 mg ml−1 was used
to conduct sitting-drop, vapour-diffusion crystallization trials using
the JCSG Core I-IV screens (NeXtal Biotechnologies). Crystals of RPB_
PLP1_R6-PLPx6 grew from drops consisting of 100 nl protein plus 100 nl
of a reservoir solution consisting of 40% (v/v) MPD and 0.1 M sodium
phosphate-citrate pH 4.2 at 18 °C, and were cryoprotected by supple-
menting the reservoir solution. Native diffraction data were collected
at APS beamline 23-ID-B, indexed to P22121 and reduced using XDS49
(Supplementary Table 1). Initial attempts to phase by molecular replace-
ment using Phaser46 and around 500 predicted models from Rosetta
and RoseTTAfold failed to yield any clear solutions. Similarly, several
thousand truncations of these models (containing all combinations
of 1, 2, 3, 4 or 5 of the 6 repeat units) also failed to give clear solutions.
To try to identify correct but low-scoring solutions in the output of
these trials, we ran SHELXE autobuilding and density modification on
a large number of these potential solutions. Ultimately, we were able to
identify an MR solution with two out of six repeats correctly placed that
allowed the autobuilding of a polyalanine model and an interpretable
map, which could be further improved by iterative rounds of rebuilding
in Coot47 and refinement using PHENIX48. Ultimately, the final model
revealed that in this crystal form and a similar crystallization condi-
tion (RPB_PLP1_R6, alternative conformation 1, above), RPB_PLP1_R6
adopted an alternative fold.
RPB_LRP2_R4. Purified RPB_LRP2_R4–LRPx4 protein at a concentra-
tion of 33 mg ml−1 was used to conduct sitting-drop, vapour-diffusion
crystallization trials using the JCSG Core I-IV screens (NeXtal
Biotechnologies). Crystals of RPB_LRP2_R4 grew from drops consisting
of 100 nl protein plus 100 nl of a reservoir solution consisting of 0.2 M
K2HPO4 and 20% (w/v) PEG 3350 at 18 °C, and were cryoprotected by
supplementing the reservoir solution with 15% ethylene glycol. Native
diffraction data were collected at APS beamline 23-ID-B, indexed to P32
2 1 and reduced using XDS49 (Supplementary Table 1). The structure
was phased by molecular replacement using Phaser46. A set of around
50 of the lowest-energy predicted models from Rosetta, as well as a
variety of truncated models, were used as search models. Several of
these models gave clear solutions, which were adjusted in Coot47 and
refined using PHENIX48. Four helical-repeat modules were present in the
asymmetrical unit. However, unexpectedly, side-chain densities for all
four repeats were very similar to one another and matched the sequence
of the internal helical repeats, but not the N- and C-terminal capping
repeats, which are slightly different from the internal ones. In addition,
these four repeat units pack tightly against adjacent, symmetry-related
molecules such that they form an ‘infinitely long’ repeat protein run-
ning throughout the crystal. Careful examination of the the junction
between each repeat unit revealed no clear breaks in electron density;
the density for the backbone is continuous through the asymmetrical
unit, and continuous with the symmetry-related molecules near the
N terminus and C terminus of the molecule in the asymmetrical unit.
Rather than truly forming an infinitely long polymer, we suspect that
proteolytic cleavage of the RPB_LRP2_R4 (either during purification or
crystallization) led to the removal of the N- and/or C-terminal caps in
many molecules, which could allow the internal repeats from separate
molecules to polymerize to form fibres in the crystal. Heterogeneity
in these cleavage products and how they assemble into the crystal
lattice (misregistration) could consequently explain the ‘continuous’
filaments of this repeat protein that we observe in these crystals.
Cell studies
Plasmids. For expression in cells, constructs were synthesized by
Genescript and cloned into a modified pUC57 plasmid (GenScript)
allowing mammalian expression under a EF1a promoter. Target pep-
tides were cloned as C-terminal fusions with a linker (GAGAGAGRP)
followed by EGFP. Binders were expressed as fusions with an N-terminal
Mito-Tag—the first 34 residues of the Mas70p protein, shown to effi-
ciently relocalize proteins to mitochondria in mammalian cells50 —and
a C-terminal mScarlet tag51. Plasmids encoding the GFP-tagged peptide
and the mScarlet-tagged binder were then cotransfected into cells.
Alternatively, for an in vivo demonstration of the multiplexed bind-
ing between different peptides and their cognate binders (Fig. 3f,g),
bicistronic plasmids were generated expressing the binder flanked with
a Mito-Tag followed by a stop codon, then an internal ribosome entry
site (IRES) sequence and the target peptide tagged with EGFP. Alter-
natively, the binder was flanked with a PEX tag—the first 66 residues
of human PEX3, targeting to peroxisomes52—and the target peptide
was tagged with mScarlet. Cells were then cotransfected with both
bicistronic plasmids to express all four proteins.
Cells. U2OS FlipIn Trex cells (a gift from S. C. Blacklow) and HeLa FlpIn
Trex cells (a gift from S. Bullock), were cultured in DMEM (Corning)
supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin–
streptomycin (Gibco) at 37 °C with 5% CO2. Cells were transfected with
Lipofectamine 3000 (Invitrogen) according to the manufacturer’s
instructions, and imaged after one day of expression. Cell lines were
not authenticated. Cells were routinely screened for mycoplasma by
DAPI staining.
Live-cell imaging. For live-cell imaging (Fig. 3), U2OS FlipIn Trex cells
were plated on glass-bottom dishes (World Precision Instruments,
FD35) coated with fibronectin (Sigma, F1141, 50 µg ml−1 in PBS), for 1 h at
37 °C in DMEM-10% serum. Medium was then changed to Leibovitz’s L-15
medium (Gibco) supplemented with 20 mM HEPES (Gibco) for live-cell
imaging. Imaging was performed using a custom spinning disk confocal
instrument composed of a Nikon Ti stand equipped with a perfect focus
system, a fast Z piezo stage (ASI) and a PLAN Apo Lambda 1.45 NA 100×
objective, and a spinning disk head (Yokogawa CSUX1). Images were
recorded with a Photometrics Prime 95B back-illuminated sCMOS
camera run in pseudo global shutter mode and synchronized with the
spinning disk wheel. Excitation was provided by 488 and 561 lasers
(Coherent OBIS mounted in a Cairn laser launch) and imaged using
dedicated single-bandpass filters for each channel mounted on a Cairn
Optospin wheel (Chroma 525/50 for GFP and Chroma 595/50 for mScar-
let). To enable fast 4D acquisitions, an FPGA module (National Instru-
ment sbRIO-9637 running custom code) was used for hardware-based
synchronization of the instrument, in particular to ensure that the piezo
z stage moved only during the readout period of the sCMOS camera.
The temperature was kept at 37 °C using a temperature control chamber
(MicroscopeHeaters.Com). The system was operated by Metamorph.
Immunofluorescence. For immunofluorescence of mitochondria
(Extended Data Fig. 2b), U2OS FlpIn Trex cells (a gift from S. C. Black-
low) were spread on glass-bottom dishes coated with fibronectin as
above. Cells were washed with PBS then fixed in 4% PFA for 20 min at
room temperature. After fixation, cells were washed with PBS and then
permeabilized with 0.1% Triton X-100 in PBS for 5 min at room tem-
perature. Cells were washed again with PBS and blocked in 1% BSA in
PBS for 15 min. Cells were then incubated with TOM20 antibody (Santa
Cruz, sc-17764, used at 1:200 dilution), diluted in 1% BSA in PBS, for
1 h at room temperature. Cells were washed three times with PBS and
then incubated with DAPI (Roche, 10236276001) and anti-mouse Alexa
Fluor 488, diluted at 1:400 in 1% BSA in PBS, for 1 h at room temperature.
Cells were washed a final three times in PBS and then imaged using the
spinning disk confocal described above.
Pull-down of endogenous proteins from extracts using designed
binders. For the pull-down of endogenous ZFC3H1 from human cell
extracts, HeLa FlpIn Trex cells were lysed in lysis buffer (25 mM HEPES,
150 mM NaCl, 0.5% Tx100, 0.5% NP-40 and 20 mM imidazole, pH 7.4,
supplemented with Roche EDTA-free protease inhibitor tablets). The
lysate was incubated on ice for 10 min to continue lysis and then spun
at 4,000g for 15 min at 4 °C. The supernatant was incubated with
pre-washed Ni-NTA agarose (Qiagen, 30210 318/AV/01) for 1 h with
rocking at 4 °C to remove or reduce proteins in the lysate that bind to
the resin non-specifically. For each condition, 50 µl of fresh Ni-NTA
agarose resin was washed twice in lysis buffer. Equimolar amounts of
purified His-tagged binder, or as a control an equal volume of buffer,
was added to the Ni-NTA agarose. The pre-cleared HeLa lysate was split
evenly between the three conditions. An input was taken of each condi-
tion, and the tubes were incubated for 2 h at 4 °C with rocking. Beads
were then washed twice in lysis buffer and twice in wash buffer (25 mM
HEPES, 150 mM NaCl and 20 mM imidazole pH 7.4). Proteins were then
eluted from the beads in elution buffer (25 mM HEPES, 150 mM NaCl
and 500 mM imidazole, pH 7.4). Inputs and elutions were run on a
NuPage 3-8% Tris-Acetate gel (Invitrogen, EA0375) and transferred to a
nitrocellulose membrane using the iBlot system (Thermo Fisher Scien-
tific). Membranes were blocked in 5% (w/v) milk in TBS-TWEEN (10 mM
Tris-HCl, 120 mM NaCl and 1% (w/v) TWEEN20, pH 7.4) for 30 min at
room temperature with gentle shaking. Rabbit anti-ZFC3H1 (Sigma,
HPA007151, used at 1:250) and mouse anti-α-tubulin 488 (Clone DMA1,
Sigma T6199, directly labelled with Abberior STAR 488, NHS ester lead-
ing to a 4.5 dye/antibody degree of labelling, and used at 0.1 µg ml−1
final concentration) were diluted in 1% (w/v) milk in TBS-TWEEN and
incubated with the membrane overnight at 4 °C with gentle shaking.
The membrane was washed three times in TBS-TWEEN then incubated
with goat anti-rabbit Alexa 555 (Invitrogen, A32732, 1:2,000) for 1 h at
room temperature with gentle shaking. The membrane was washed
twice with TBS-TWEEN, followed by a final wash with TBS-TWEEN
with 0.001% SDS. Membranes were imaged using a ChemiDoc system
(BioRad). Alternatively, the same samples were analysed using 4–12%
Bis-Tris gels (Invitrogen NP0323BOX) and stained with InstantBlue
Coomassie stain (Sigma ISB1L). Note that αZFC-high was also able
to pull down endogenous ZFC3H1 from human cell extracts when
50 mM rather than 150 mM NaCl was used in all buffers (Extended
Data Fig. 7b).
Mass spectrometry. Each line of the polyacrylamide gel presented in
Fig. 6c was cut into six pieces (1–2 mm) and prepared for mass spec-
trometric analysis by manual in situ enzymatic digestion (the gel area
containing the binder was omitted from the analysis to avoid saturation
of the detector by overabundance of binder peptides). In brief, the
excised protein gel pieces were placed in a well of a 96-well microtitre
plate and destained with 50% (v/v) acetonitrile and 50 mM ammonium
bicarbonate, reduced with 10 mM DTT and alkylated with 55 mM io-
doacetamide. After alkylation, proteins were digested with 6 ng µl−1
trypsin (Promega) and 0.1% Protease Max (Promega) overnight at 37 °C.
The resulting gel pieces were extracted with ammonium bicarbonate
(100 µl, 100 mM) and ammonium bicarbonate/acetonitrile (50/50,
100 µl) before being dried down by vacuum. Clean-up of peptide digests
was carried out with HyperSep SpinTip P-20 (Thermo Fisher Scien-
tific) C18 columns, using 80% acetonitrile as the elution solvent before
being dried down again. The resulting peptides were extracted in 0.1%
(v/v) trifluoroacetic acid acid and 2% (v/v) acetonitrile. The digest was
analysed by nano-scale capillary liquid chromatography–tandem mass
spectrometry (LC–MS/MS) using an Ultimate U3000 HPLC (Dionex,
Thermo Fisher Scientific) to deliver a flow of 250 nl min−1. Peptides
were trapped on a C18 Acclaim PepMap100 5 µm, 100 µm × 20 mm
nanoViper (Thermo Fisher Scientific) before separation on a PepMap
RSLC C18, 2 µm, 100 A, 75 µm × 75 cm EasySpray column (Thermo Fisher
Scientific). Peptides were eluted on a 90-min gradient with acetonitrile
and interfaced using an EasySpray ionization source to a quadrupole
Orbitrap mass spectrometer (Q-Exactive HFX, Thermo Fisher Scien-
tific). Mass spectrometry data were acquired in data-dependent mode
with a top-25 method; high-resolution full mass scans were performed
(R = 120,000, m/z 350–1,750), followed by higher-energy collision dis-
sociation with a normalized collision energy of 27%. The corresponding
tandem mass spectra were recorded (R = 30,000, isolation window
m/z 1.6, dynamic exclusion 50 s). LC–MS/MS data were then searched
against the Uniprot human proteome database, using the Mascot search
engine programme (Matrix Science)53. Database search parameters
were set with a precursor tolerance of 10 ppm and a fragment ion mass
tolerance of 0.1 Da. One missed enzyme cleavage was allowed and vari-
able modifications for oxidation, carboxymethylation and phospho-
rylation. MS/MS data were validated using the Scaffold programme
(Proteome Software)54. All data were in addition interrogated manually.
To generate the Venn diagram in Fig. 6f, we considered a threshold of
minimum five peptides to consider that a protein had been identified.
The mass spectrometry proteomics data have been deposited to the
ProteomeXchange Consortium through the PRIDE55 partner repository
with the dataset identifiers PXD038492 and 10.6019/PXD038492. See
also Source Data for the annotated full dataset.
Reporting summary
Further information on research design is available in the Nature Port-
folio Reporting Summary linked to this article.
Data availability
The atomic coordinates and experimental data of RPB_PEW3_R4–
PAWx4, RPB_PLP3_R6–PLPx6, RPB_LRP2_R4–LRPx4, RPB_PLP1_R6–
PLPx6, RPB_PLP1_R6–PLPx6 (alternative conformation 1), RPB_PLP1_
R6–PLPx6 (alternative conformation 2) and RPB_LRP2_R4 (pseudo-
polymeric) have been deposited in the RCSB PDB with the accession
Article numbers 7UDJ, 7UE2, 7UDK, 7UDL, 7UDM, 7UDN and 7UDO, respec-
tively. The Rosetta macromolecular modelling suite (https://www.roset-
tacommons.org) is freely available to academic and non-commercial
users. Commercial licences for the suite are available through the Uni-
versity of Washington Technology Transfer Office. The mass spectrom-
etry proteomics data have been deposited to the ProteomeXchange
Consortium through the PRIDE partner repository with the dataset
identifiers PXD038492 and 10.6019/PXD038492. Source data are pro-
vided with this paper. All protein sequences for the binders described
in this study are provided in Supplementary Table 2.
Code availability
The design scripts and main PDB models, computational protocol for
data analysis, experimental data and analysis scripts, all the design
models and the next-generation-sequencing results used in this paper
can be downloaded from file servers hosted by the Institute for Protein
Design: https://files.ipd.uw.edu/pub/2023_modular_peptide_bind-
ing_proteins/all_data_modular_peptide_binding_proteins.tar.gz. The
code to identify proteins in databases containing any linear combina-
tion of amino acid triplets given as an input can be found on GitHub
(https://github.com/tjs23/prot_pep_scan).
39. Sheffler, W. & Baker, D. RosettaHoles: rapid assessment of protein core packing for
structure prediction, refinement, design, and validation. Protein Sci. 18, 229–239
(2009).
40. Bradley, P., Misura, K. M. S. & Baker, D. Toward high-resolution de novo structure prediction
for small proteins. Science 309, 1868–1871 (2005).
41. Tyka, M. D. et al. Alternate states of proteins revealed by detailed energy landscape
mapping. J. Mol. Biol. 405, 607–618 (2011).
42. Sharp, P. M. & Li, W. The codon adaptation index—a measure of directional synonymous
codon usage bias, and its potential applications. Nucleic Acids Res. 15, 281–1295
(1987).
43. Dyer, K. N. et al. High-throughput SAXS for the characterization of biomolecules in solution:
a practical approach. Methods Mol. Biol. 1091, 245–258 (2014).
44. Williams, C. J. et al. MolProbity: more and better reference data for improved all-atom
structure validation. Protein Sci. 27, 293–315 (2018).
45. Chao, G. et al. Isolating and engineering human antibodies using yeast surface display.
Nat. Protoc. 1, 755–768 (2006).
46. McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674
(2007).
47. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot.
Acta Crystallogr. D 66, 486–501 (2010).
48. Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular
structure solution. Acta Crystallogr. D 66, 213–221 (2010).
49. Kabsch, W. XDS. Acta Crystallogr. D 66, 125–132 (2010).
50. Kessels, M. M. & Qualmann, B. Syndapins integrate N-WASP in receptor-mediated
endocytosis. EMBO J. 21, 6083–6094 (2002).
51. Bindels, D. et al. mScarlet: a bright monomeric red fluorescent protein for cellular
imaging. Nat. Methods 14, 53–56 (2017).
52. Fakieh, M. H. et al. Intra-ER sorting of the peroxisomal membrane protein Pex3 relies on
its luminal domain. Biol. Open 2, 829–837 (2013).
53. Perkins, D. N., Pappin, D. J. C., Creasy, D. M. & Cottrell, J. S. Probability-based protein
identification by searching sequence databases using mass spectrometry data.
Electrophoresis 20, 3551–3567 (1999).
54. Keller, A., Nesvizhskii, A. I., Kolker, E. & Aebersold, R. Empirical statistical model to estimate
the accuracy of peptide identifications made by MS/MS and database search. Anal. Chem.
74, 5383–5392 (2002).
55. Perez-Riverol, Y. et al. The PRIDE database resources in 2022: a hub for mass spectrometry-
based proteomics evidence. Nucleic Acids Res. 50, D543–D552 (2022).
Acknowledgements We thank B. Wicky, A. Ljubetic and I. Lutz for advice on the split luciferase
assay for the second-round design screening; C. Xu for help troubleshooting experiments;
T. Schlichtharle for discussion; L. Cao for advice on bio-layer interferometry; H. Pyles for advice
on circular dichroism and DHR proteins; R. Hegde for the suggestion to target disordered
regions of endogenous proteins; and K. Van Wormer and A. Curtis Smith for laboratory support
during COVID-19. This work was supported by the Audacious Project at the Institute for Protein
Design (D.B., K.W., M.D., D.A.S. and A.B.); the Michelson Found Animals Foundation grant
number GM15-S01 (L.S., K.W. and D.B.); the National Institute on Aging grant 5U19AG065156-02
(D.R.H., K.W. and D.B.); the National Institute of General Medical Sciences grant R35GM128777
(D.C.E.); the Howard Hughes Medical Institute (D.B., W.S. and H.B.); the Open Philanthropy
Project Improving Protein Design Fund (Y.-T.C., R.R., C.M.C., G.B., D.C.E. and D.B.); the Donald
and Jo Anne Petersen Endowment for Accelerating Advancements in Alzheimer’s Disease
Research (T.J.B. and D.B.); a donation from AMGEN to the Institute for Protein Design (I.G.);
the Medical Research Council (MC_UP_1201/13 to E.D., T.E.M. and T.J.S.); the Human Frontier
Science Program (CDA00034/2017-C to E.D.); and a Sir Henry Wellcome Postdoctoral
Fellowship (220480/Z/20/Z to K.E.M.).
Author contributions K.W., D.A.S. and D.B. designed the research. D.A.S. and D.B. developed
the preliminary computational method and hash database. W.S. contributed to the
development of the hash database. K.W. updated the computational method with help
from D.A.S. and H.B. H.B. updated the hash database to be more general. Y.S. helped and
contributed to the first development of the hash database. K.W. and T.J.B. designed the
polyproline II DHR scaffold library using the method developed by D.R.H. K.W. designed
the binders with help from H.B. H.B. and K.W. performed the yeast screening, expression
and binding experiments with help from I.G. for the first-round design characterization.
K.W. performed bio-layer interferometry and Octet assays for the second-round design
characterization. H.B. constructed and screened SSM libraries. Y.-T.C., R.R., G.B. and D.C.E.
solved the structures of RPB_PEW3_R4–PEWx4, RPB_PLP3_R6–PLPx6, RPB_LRP2_R4–LRPx4
and RPB_PLP1_R6–PLPx6. K.E.M. designed and performed all cell experiments in this work,
in particular the multiplex binding assay and the demonstration of the endogenous binder for
ZFC3H1. E.D. identified ZFC3H1 as a good target for the development of an endogenous
binder with help from T.J.S. T.E.M. performed mass spectrometry analysis. A.B. helped with
the modular binding assay. M.D. and C.M.C. helped with preparing protein samples for
crystallography. All authors analysed data. L.S., D.A.S. and D.B. supervised research. K.W.
and D.B. wrote the manuscript with input from the other authors. All authors revised the
manuscript.
Competing interests The authors declare no competing interests, except as follows. K.W.,
H.B., D.R.H., T.J.B., K.E.M., T.J.S., T.E.M., Y.-T.C., R.R., G.B., D.C.E., L.S., E.D., D.A.S., W.S., I.G. and
D.B. are co-inventors on a patent application entitled ‘De novo designed modular peptide
binding proteins by superhelical matching’ (63/381,109, filed 26 October 2022).
Additional information
Supplementary information The online version contains supplementary material available at
https://doi.org/10.1038/s41586-023-05909-9.
Correspondence and requests for materials should be addressed to Emmanuel Derivery,
Daniel Adriano Silva or David Baker.
Peer review information Nature thanks Vikas Nanda and the other, anonymous, reviewer(s) for
their contribution to the peer review of this work. Peer reviewer reports are available.
Reprints and permissions information is available at http://www.nature.com/reprints.
Extended Data Fig. 1 | Examples of computationally designed model
geometry and convergence of backbone docking. a–c, Examples of repeat
proteins computationally designed to bind to extended beta strand (a),
polypeptide II (b) and helical peptide backbones (c). d, Monte Carlo flexible
backbone docking calculations after design to assess the structural specificity
of the designed peptide-binding interface. It started from large numbers of
peptide conformations randomly generated with superhelical parameters in
the range of those of the proteins (usually 10,000–50,000 trajectories), and
selected those designs with converged peptide backbones (RMSD < 2.0 among
the top 20 designs with lowest DDG) close to the design model (RMSD < 1.5).
Green dots shown in the above example plot represent the converged designs
picked by this threshold.
Article Extended Data Fig. 2 | Comparison of binding affinities from freshly made
and 30-day-old samples, and mitochondria immunostainings in control
U2OS cells. a, Little decrease in binding observed for designs RPB_PLP1_R6 and
RPB_PEW1_R6 30-day-old in 4 °C. Bio-layer interferometry characterization of
binding of designed proteins to the corresponding peptide targets. Twofold
serial dilutions were tested for each binder, and the full tested concentration is
labelled. The biotinylated target peptides were loaded onto the streptavidin
(SA) biosensors, and incubated with designed binders in solution to measure
association and dissociation. b, Mitochondria immunostainings in control
U2OS cells. Wild-type U2OS cells were spread onto fibronectin coverslips as
in Fig. 3, then fixed and processed for immunofluorescence using TOM20
antibodies as a marker of mitochondria. Note that mitochondria appearance in
these control cells is similar to that observed upon overexpression of designed
binders fused to mitochondria-targeting sequences (Fig. 3). suggesting that
these constructs do not affect mitochondria shape. Scale bar, 10 µm.
Extended Data Fig. 3 | See next page for caption.
Article Extended Data Fig. 3 | SSMs libraries are constructed and screened for
enhancing the peptide-binding abilities of designed repeat-peptide
binders. a, A schematic illustration of the mutagenesis region within the
designed repeat protein, and the principles of the yeast surface display assay
for peptide binding analysis. In short, the biotinylated repeat peptides (a six-
repeat of LRP peptide is shown as an example) are synthesized and can be
detected by SAPE, while the expression of designed protein on yeast surface
are monitored by FITC-conjugated anti-Myc antibody. A double high signal of
both PE and FITC, using flow cytometry, indicates the valid peptide-binding
events. b, The SSM libraries are first subjected to expression sorting (left), in
which there is no targeted peptide added. The yeast populations, which display
well expressed SSM mutants, will show above threshold FITC signals, are
collected (green box) for next-generation sequencing, and are regrown for
the next rounds of sorting. In the next round sorting, the targeted peptide is
incubated with the yeast library, and labelled by both FITC and SAPE (right).
The FITC+PE+ population is collected for analysis (orange box). c, By using
next-generation sequencing, enrichment analysis for each mutation is carried
out, and a heat map for all mutations is generated. In this heat map, using a
designed LRP binder SSM library as an example, the red shades indicate
enrichment with incubating with the targeted peptide, and the blue shades
indicate depletion. Several mutations show exceptional enhancement of the
LRP repeat peptide-binding ability, such as F93W, H102S and others. d, Using
the SSM library, we can markedly enhance the peptide-binding abilities of the
designed peptide binder. Three example yeast display assays titrating the
peptide concentrations are shown here. The top row of each example is using
the originally designed peptide binder, and the bottom row is using the peptide
binder containing the combinations of the best mutations discovered in the
SSM library screenings. An approximately 1,000-fold increase of the peptide-
binding ability can be achieved with the assistance of SSM libraries. Note, the
ratio of yeast population in the upper right quadrant indicates the peptide-
binding ability.
Extended Data Fig. 4 | Comparison of binding affinities when changing
repeat numbers from either binder or peptide side. and top five flexible
backbone docks for the four-repeat LRP binder RPB_LRP2_R4–LRPx4.
a, Six-repeat versions of RPB_LRP2_R6 and RPB_PEW2_R6 had higher affinity
for eight-repeat LRP and PEW peptides than four-repeat versions without any
decrease in specificity in yeast surface display. Biotinylated repeat proteins
(the six-repeat versions RPB_LRP2_R6 and RPB_PEW2_R6 and the four-repeat
versions RPB_LRP2_R4 and RPB_PEW2_R4) were detected by SAPE, and the
expression of the designed repeat peptide on yeast surface was monitored by
FITC-conjugated anti-Myc antibody. Serial dilutions were tested for each
binder, and the full tested concentration is labelled. b, Six-repeat IYP and PLP
peptides had higher affinity for six-repeat versions of the cognate designed
repeat proteins (RPB_IYP1_R6 and RPB_PLP1_R6) than four-repeat versions
by bio-layer interferometry. The full tested concentration is labelled. The
biotinylated target peptides were loaded onto the streptavidin (SA) biosensors,
and incubated with designed binders in solution to measure association and
dissociation. The dissociation rate was markedly increased when testing
against the six-repeat peptides as compared to the four-repeat peptides,
indicating a much tighter binding event. c, Top five complex PDBs for RPB_
LRP2_R4–LRPx4 from the flexible docking generated ensemble. Green, pink
and grey are the ones closest to the crystal structure (shown in yellow) with
RMSD over the peptide and the binding residues ≈ 0.03 Å, whereas the cyan
dock RMSD = 3.89 Å.
Article Extended Data Fig. 5 | Crystal structures of the unbound RPB_LRP2_R4,
bound RPB_PLP3_R6–PLPx6 and bound RPB_PEW3_R4 and its top five
flexible backbone docks. a, Crystal structure of the unbound first-round
design RPB_LRP2_R4 (yellow) aligned with the design model (cyan). b, Crystal
structure of the first-round complex RPB_PLP3_R6–PLPx6 (yellow) aligned
with the design model (cyan). As is shown here, the peptide PLP units fit exactly
into the designed curved groove formed by repeating tyrosine, alanine and
tryptophan residues matching the design model with near atomic accuracy,
with Cα RMSD of 1.70 Å for the binder apo, 2.00 Å for the peptide neighbour
interface and 1.64 Å for the whole complex. c, Co-crystal structure of RPB_
PEW3_R4–PAWx4. The PAW units bind to a relatively flat groove formed by
repeating histidine residues and glutamine residues as designed (shown as
sticks). d, Top five complex PDBs for RPB_PEW3_R4–PAWx4 from the flexible
docking generated ensemble. Green, pink and grey are the ones closest to the
crystal structure (shown in yellow) with RMSD over the peptide and the binding
residues ≈ 0.03 Å, whereas the cyan dock RMSD = 3.89 Å.
Extended Data Fig. 6 | SSM binding interface footprinting results were
consistent with the design model and crystal structure. a, Using a PPL
repeat-peptide binder as an example, a heat map presenting enrichment
analysis for each mutation is generated. In each cell, the red colour indicates
enrichment, and the blue colour indicates depletion. Wild-type sequences are
indicated in the cells labelled with amino-acid one-letter codes. The mutants
missing in the expression library are labelled with asterisks. Two positions
(109Q and 156Q) are highlighted as examples showing conserved positions.
Almost all mutations other than the wild type in these two positions are greatly
depleted. b, Illustration shows the SSM region (orange), and the two conserved
positions (109Q and 156Q in yellow).
Article Extended Data Fig. 7 | Characterization of ZFC3H1 binders. a, Bio-layer
interferometry screening for the seven endogenous ZFC3H1 binders. Twofold
serial dilutions were tested for each binder, and the full tested concentration is
labelled. The biotinylated target 24-amino-acid peptides (PLPPLPPLPPLPPEDP
EQPPKPPF) were loaded onto the streptavidin (SA) biosensors, and incubated
with designed binders in solution to measure association and dissociation.
The two tightest binders (αZFC_93 and αZFC_97, renamed αZFC-high and
αZFC-low, respectively) were selected for further fluorescence polarization
characterization and cell assays. b, Characterization of ZFC3H1 binders for
pull-down of endogenous target: Hela cell extracts were subjected to
pull-down using the indicated binders bound to Ni-NTA agarose beads, or
naked beads as control. Recovered proteins were processed for western blot
against endogenous ZFC3H1 (or tubulin as a loading control). Two completely
independent experiments are shown. These experiments are repeats of the
experiment presented in Fig. 6e, albeit at a different salt concentration, namely
50 mM instead of 150 mM. For gel source data, see Supplementary Fig. 1.
Extended Data Table 1 | Summaries of first- and second-round experimental characterization
a, First-round experimental characterization summary. It is clearly shown that among the binders, most of them bound peptides with sequences similar to those targeted but not the same;
and peptides with three-residue repeat units were targeted more successfully (19 in total) than those with two-residue repeat units (2 in total). b, Second-round experimental characterization
summary. In total, 54 second-round designed protein–peptide pairs were tested. Forty-two of the designed proteins were solubly expressed in E. coli, 25 were monomerically dispersed by SEC
and 16 bound their targets with considerably higher affinity and specificity than in the first round.
Article Extended Data Table 2 | Interface side-chain heavy-atom RMSD calculations and SAXS Vr calculations
a, Interface side-chain heavy-atom RMSD calculation for four co-crystal complexes and design models. The interface heavy-atom RMSD calculations using Pymol align with cycles=0 (iRMSD
for short) was applied to all four crystal-design complexes. For the first-round designs, for example, the values are averaged over five top designs for RPB_PEW3_R4–PAWx4, because the design
models were not fully converged (as stated in the main text). For RPB_LRP3_R4–LRPx4, because the final two models were sampling two distinct arginine rotamers as stated in the main text, we
calculated the iRMSD for these two, respectively. The closest one was shown above with asterisks, and the further one as (full_iRMSD = 5.29, inter_iRMSD = 5.16). For all four pairs, we inspected
both the full-repeat RMSD with internal-repeat RMSD (N-terminal and C-terminal caps excluded) here, owing to the potential lever-arm effect. b, Structural validation of six-repeat peptide
binders by SAXS volatility ratio (Vr) calculation.
| null |
10.1161_CIRCRESAHA.121.319314.pdf
| null |
Sources of Funding This work was funded by National Institutes of Health grants HL139819 and HL141256 to K. Walsh, HL152174 to S. Sano and K. Walsh, T32 HL007284 to N.W. Chavkin, and American Heart Association grant 20POST35210098 to M.A. Evans.
|
PDF file not found
|
|
10.1371_journal.pwat.0000213.pdf
|
Data Availability Statement: All data supporting
this study is free access data and the sources and
collections details are provided as supplementary
information accompanying this paper. The code
used for the analysis and results figures is also
made available at: https://github.com/J-Marcal/
WSF_IneqAnalysis.
|
All data supporting this study is free access data and the sources and collections details are provided as supplementary information accompanying this paper. The code used for the analysis and results figures is also made available at: https://github.com/J-Marcal/ WSF_IneqAnalysis .
|
RESEARCH ARTICLE
Assessing inequalities in urban water security
through geospatial analysis
Juliana Marc¸ alID
1,2*
Jan HofmanID
1,2*, Junjie Shen3, Blanca Antizar-Ladislao4,5, David Butler6,
1 Water Innovation and Research Centre (WIRC), Department of Chemical Engineering, University of Bath,
Bath, United Kingdom, 2 Water Informatics in Science and Engineering (WISE) Centre for Doctoral Training,
University of Bath, Bath, United Kingdom, 3 University Library, University of Bath, Bath, United Kingdom,
4 Isle Utilities Ltd., London, United Kingdom, 5 Department of Civil, Environmental and Geomatic
Engineering, University College London, London, United Kingdom, 6 Centre for Water Systems, Department
of Engineering, University of Exeter, Exeter, United Kingdom
* [email protected] (JM); [email protected] (JH)
Abstract
Water security, which is key for sustainable development, has been broadly investigated
through different spatial scales, time frames and perspectives, as a multi-dimensional con-
cept. Fast growth and the diversity of the urban environment add to the challenges of reach-
ing good levels of water security in cities. Yet, few studies have focused on evaluating the
heterogeneous distribution of water security in urban areas, which is a key step to highlight
where inequalities in large cities are present and how to best guide interventions. The objec-
tive of this research is to investigate the spatial heterogeneity of urban water security as well
as quantifying inequalities using the new assessment presented in this paper. A holistic indi-
cator-based evaluation framework to intra-urban sectors of the city of Campinas in Brazil is
applied, followed by an inequality analysis to describe the distribution of water security
aspects. A spatial correlation analysis is then carried out to identify patterns for high inequal-
ity indicators. Results show that even though Campinas has established good overall water
security conditions, spatial heterogeneity is still noticeable in the urban area. Quantification
of inequality by the Theil index highlighted aspects, such as vegetation cover, social green
areas, and wastewater collection, that are inequitably distributed in the urban area. The sub-
sequent analysis of spatial patterns exposed areas on the outskirts of the city where infra-
structure challenges and social vulnerability coincide. This novel approach has been
therefore successfully validated in a city in Brazil, and it has been demonstrated that our
water security assessment framework identifies what are the main water security challenges
and where they are in the city. For the first time we show that associating spatial and
inequality analysis with conventional evaluation of urban water security has the potential to
help target areas in need and tackle specific water security issues in the urban area. This is
crucial to inform urban planning and policy making for a sustainable and inclusive urban
water management strategy.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Marc¸al J, Shen J, Antizar-Ladislao B,
Butler D, Hofman J (2024) Assessing inequalities
in urban water security through geospatial analysis.
PLOS Water 3(2): e0000213. https://doi.org/
10.1371/journal.pwat.0000213
Editor: Venkatramanan Senapathi, Alagappa
University, VIET NAM
Received: February 17, 2023
Accepted: December 13, 2023
Published: February 1, 2024
Copyright: © 2024 Marc¸al et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All data supporting
this study is free access data and the sources and
collections details are provided as supplementary
information accompanying this paper. The code
used for the analysis and results figures is also
made available at: https://github.com/J-Marcal/
WSF_IneqAnalysis.
Funding: This study was conducted as part of the
Water Informatics Science and Engineering (WISE)
Centre for Doctoral Training (CDT), funded by the
UK Engineering and Physical Sciences Research
Council, grant number EP/L016214/1. JM is
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
1 / 25
supported by a research studentship from this
CDT. The content is solely the responsibility of the
authors. The funders had no role in study design,
data collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Assessing inequalities in urban water security
Introduction
Urban areas around the world are facing increasing water security challenges associated with
rapid growth and climate change. In 2022, we saw cities around the globe experience extreme
weather, particularly severe droughts [1] with significant impacts on water availability affecting
food and energy production and human well-being [2]. Additionally, urban areas are an intri-
cate system of water and other infrastructures that coexist and interact in heterogeneous
spaces. This heterogeneity and complexity increase with the size of cities, alongside pressures
on the water system and resources. These conditions reinforce the need to investigate urban
water security especially from a multi-dimensional perspective, considering the different
aspects involved but also its dependence on space and time.
With several definitions, perspectives, approaches and assessment methodologies, water
security is acknowledged as a broad concept and has been object of interest of scholars for
decades [3–7]. The UN considers water security as the “capacity of a population to safeguard
sustainable access to adequate quantities of acceptable quality water for sustaining livelihoods,
human well-being, and socio-economic development, for ensuring protection against water-
borne pollution and water-related disasters, and for preserving ecosystems in a climate of peace
and political stability” [8]. This all-encompassing and well-accepted definition [7] provides an
interpretation that includes not only supply and accessibility but also environmental, hazard,
economic, social and well-being elements.
In the urban context, rapid growth and governance issues may lead to opportunities, infra-
structure and services to be unevenly distributed in the urban area [9]. As a consequence, the
benefits of city life may not be equally available for all, leading to varying water security experi-
ences for its inhabitants. The marginalisation of people in informal settlements and slums,
inequality, insufficiency and urban poverty compromise water security [10, 11]. Therefore, in
an urban environment, certain areas and communities can be more vulnerable to water-
related issues [6]. It is thus very important to develop policies that consider the spatial hetero-
geneity of the urban area. Being spatially explicit allows the identification of city districts or
areas that require strategies for increasing water security. Incorporating a spatial approach to
urban water security evaluation can help identify inequalities and provide information to iden-
tify areas at risk, helping to establish effective policies to protect the most vulnerable people,
making sure that no one is left behind [12].
Previous studies have been interested in the question water security for whom? [13, 14]
through investigation of the spatial distribution of different water-related aspects. At a global
level, Gain et al. [15] highlighted the importance of spatial and temporal assessment of perfor-
mances to identify specific needs and persistent problems in different countries. Doeffinger
and Hall [14] worked on evaluating water security across states and counties in the United
States, showing evidence of how spatial analysis can reveal the heterogeneity across the coun-
try. The work by Stuart et al. [16] discovered geographical patterns and the spatial heterogene-
ity of water insecurity in rural Uganda as well as their implications for community water
interventions. In terms of urban water security, the study by Tholiya and Chaudhary [17] pro-
vides a geospatial assessment of water supply services in Pune in India. While the investigation
highlights the differences that were found within the city boundary, the evaluation focuses on
water supply performance indicators. Other water security related aspects such as water infra-
structure inequalities [18], ecological security [19], alternative water supply [20] and domestic
water consumption [21] have also been spatially investigated in the literature, showing the
importance of looking within the traditional boundaries as a way to capture disparities.
Although the importance of studying a smaller scale has been highlighted by different
authors [11, 14, 20, 22], few works in the literature have assessed urban water security
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
2 / 25
PLOS WATER Assessing inequalities in urban water security
holistically at intra-city level. The study by Mukherjee et al. [22] provided an evaluation at
micro-level for 16 administrative regions in Kolkata, India focusing on availability, accessibil-
ity quality and risks as components of an urban water security index. Assefa et al. [23] devel-
oped a domestic water security framework applied to the city of Addis Ababa in Ethiopia,
subdivided into ten administrative regions. The authors included water supply, sanitation and
hygiene indicators in their assessment and the analysis showed considerable disparities in
domestic water security within the city, indicating opportunities for local development. How-
ever, these studies tend to focus on the drinking water safety aspects of urban water security
and lack the explicit incorporation of a spatial approach to their analysis. An in-depth and
holistic evaluation of urban water security accounting for spatial patterns and inequality mea-
sure is not found in the literature.
In this study we present an urban water security assessment that explicitly accounts for the
spatial distribution and patterns of water security elements. The main contributions are two-
fold: (1) we explore the spatial variability of water security from an intra-urban perspective fol-
lowing a framework that includes not only water supply and accessibility but also
environmental, hazards, economic, social and well-being elements and (2) we further explore
the heterogeneity of urban water security by including an inequality measure in the analysis.
In this way we investigate the diversity of the urban area by downscaling the assessment to
urban districts and neighbourhoods, and visualising how the results are distributed in the area.
This provides a more detailed vision of the city and allows the investigation of where inequali-
ties lie. We investigate the ‘what’ and ‘where’ of the water security challenges in the urban area.
This could lead to important information to help establish priorities for either monitoring or
acting upon local issues, potentially leading to more equality and inclusiveness for water secu-
rity in a city. We offer an exploratory analysis of such approach by using the city of Campinas
in Brazil as a case study.
The paper is structured as follows: the next section describes the methods used in the devel-
opment of the assessment framework, including the dimensions considered and the corre-
sponding indicators, as well as the context of the city of Campinas and how data was obtained
for the case study. We also present the data processing and analysis methods that are used in
the framework. This is followed by a section presenting the results of the application of the
framework to the city of Campinas, where we discuss the findings and highlight how inequali-
ties emerge from the qualitative and quantitative analysis of spatial variation. Finally, we pro-
vide some perspectives on the approach and end with concluding remarks.
Materials and methods
Assessment framework
Based on the analysis of gaps in existing water security assessment frameworks reported in lit-
erature [11, 24–27], an indicator based framework was created to evaluate urban water secu-
rity. The choice and classification of indicators was guided by the United Nations definition
[28] of water security—considered as an interdisciplinary, holistic and well-accepted view of
the concept [7, 24, 29]. Indicators were divided into different hierarchical levels: first the four
dimensions, following the UN water security infographic [8], then categories characterised by
one or more indicators. The aspects included in the framework are presented in the Table 1
that also provide references of works adopting similar variables to the assessment of water
security. Dimension A: Drinking water and human well-being encompasses some of the funda-
mental aspects of water security such as having enough water in terms of quantity and quality
available for basic needs. We also include in this dimension measures to indicate access to
basic urban water services such as piped drinking water and wastewater collection at the
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
3 / 25
PLOS WATER Table 1. List of selected indicators used in the framework.
Dimension
Category
Indicator
Measure
Assessing inequalities in urban water security
A Drinking water and well-
being
A1 Water quantity
A1.1 Water demand [26, 30, 31]
Domestic water consumption (L/cap/day)
A1.2 Water availability [32, 33]
A1.3 Diversity of sources [25, 34, 35]
Ratio between the average flow of renewable freshwater resources
and population (in m3/cap/year)
Shannon Index accounting for the proportion of water coming
from different sources
A1.4 Reserve/storage capacity [25, 34] Storage volume in terms of days of supply
A1.5 Water stress [26, 36, 37]
Freshwater withdrawn as a percentage of the total available
A2 Water quality
A2.1 Drinking water quality [23, 25,
37]
Proportion of drinking water samples meeting local standards
A3 Accessibility to
services
A3.1 Piped water coverage [23, 36,
38]
Percentage of population with access to residential piped water
supply
A3.2 Sewage coverage [23, 25]
A3.3 Affordability [23, 34, 39]
Percentage of population with access to residential wastewater
collection network
Proportion of the household budget spent on water and sanitation
services
A4 Infrastructure
reliability
A4.1 Service discontinuity [39–41]
Proportion of households affected by supply discontinuity
A4.2 Service reliability [41, 42]
Ratio of the number of sewer corrective maintenance operations to
the extension of the sewage network
A4.3 Metering level [36, 43, 44]
Percentage of households with metered water
A4.4 Water loss [36, 37, 45]
Percentage of produced water lost in distribution
A5 Public health and
well-being
A5.1 Incidence of water-borne
diseases [26, 36, 41]
Occurrence of gastrointestinal diseases in number of cases per year
per 100.000 people
A5.2 Recreational opportunities [46]
Area of the sector contained within a radius of 2 kilometres from a
social green area
B Ecosystems
B1 Environment
B1.1 Green areas [36, 43]
Proportion of area covered by vegetation
C Water related hazards and
climate change
B1.2 Environmental safety
Incidence of vector-borne diseases (cases per year per 100.000
people)
B2 Pollution control
B2.1 Groundwater quality [30, 47, 48] Assessment based on pollutants concentration, according to local
standards
B2.2 Surface water quality [11, 41, 48] Assessment based on local standards to protection of aquatic life
B2.3 Wastewater treatment rate [36,
49, 50]
Percentage of collected wastewater treated before discharge
B3 Usage efficiency
B3.1 Energy usage efficiency [37, 51]
B3.2 Wastewater reuse (recycling)
[36, 43, 52]
Energy consumption by the removal efficiency of wastewater
treatment plants
Ratio of wastewater reused to wastewater treated
B4 Solid waste
B4.1 Solid waste collection [30, 48]
Coverage of door-to-door solid waste collection
B4.2 Recyclable waste collection [48]
Coverage of door-to-door recyclable waste collection
C1 Water-hazards
C1.1 Flood frequency [25, 26, 36]
Flood occurrences over a decade
C1.2 Drought frequency [43, 47]
Drought occurrences over a decade
C1.3 Flood- prone areas [47, 53]
Percentage of area susceptible to flooding
C1.4 People living under hazardous
zones [52]
Percentage of people living under hazardous zones
C2 Preparedness
C2.1 Risk Management [25, 54, 55]
Qualitative measurement to evaluate disaster prevention and risk
management
C2.2 Urban drainage [26, 37, 56]
Storm drains coverage
C2.3 Paved streets
Pavement coverage
C2.4 Drainage investment [55]
Percentage of budget destined to rainwater management
C3 Climate change
C3.1 Greenhouse gas emissions [43,
48, 57]
Emission of greenhouse gases expressed in tonnes of CO2
equivalent per capita
C3.2 Temperature increase [43, 58]
Average annual temperature increase
C3.3 Extreme rainfall events
Number of extreme rain events over a decade
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
(Continued )
4 / 25
PLOS WATER Table 1. (Continued)
Dimension
D Economic and social
development
Category
D1 Governance
Assessing inequalities in urban water security
Indicator
Measure
D1.1 Communication and access to
information [59]
Qualitative assessment over effective government communication
and information access
D1.2 Public participation
opportunities [26, 60]
D1.3 Equality and non-
discrimination
Qualitative assessment on significant participation opportunities
Qualitative assessment on representation diversity in decision
making groups
D1.4 WASH investment [31, 49]
Percentage of the GDP invested in water and sanitation
D1.5 Water self-sufficiency [37, 49]
Proportion of water withdrawal taken from within own territory
D1.6 Organisational structure [26, 47,
61]
Qualitative assessment on organisational structure.
D2 Social aspects
D2.1 Literacy rate [30]
D2.2 Population density [56, 62]
D2.3 Inequality [30]
Percentage of population more than 15 years old that is able to read
Population density in the urban area (inhabitants per km2)
Gini coefficient, representing the degree of inequality in the
distribution of income
D2.4 Income [62]
Ratio of average income and minimum wage
D2.5 Informal dwellings [26, 30]
Percentage of population living in informal settings
D2.6 Gender equality
Ratio of average income from households headed by women and
men
D3 Economic
development
D3.1 Per capita GDP [63]
Ratio of GDP and population
D3.2 Water productivity
Ratio of GDP and total freshwater withdrawal
https://doi.org/10.1371/journal.pwat.0000213.t001
household, as well as measures of how reliable these services are in the urban area. Finally, we
consider the safeguard of health and well-being [8] in the city by including indicators of the
incidence of water-borne diseases and access to social green spaces. The status of water
resources, pollution-related aspects (including wastewater treatment), vegetation cover, effi-
ciency of resource use and solid waste management are grouped under dimension B: Ecosys-
tems. Dimension C: Water related hazards and climate change includes water hazards,
resilience and protection infrastructure as well as indicators related to changing climate.
Finally, social, economic and governance aspects of water use are included under dimension
D: Economic and social development.
Once populated, since originally expressed in different units, the indicators were normal-
ised between 0 and 1 following thresholds based on references from the literature and regional
values [23, 47]. Detailed information on the measures for each indicator and the normalisation
procedure is presented as supplementary material (see S1 File). Scores range from 1 to 0, with
desirable characteristics given ‘1’ and undesirable values, ‘0’. In order to calculate sub-indexes,
the indicators are aggregated first by category and then by dimension, using the arithmetic
mean of the indicators’ scores.
Study area
The framework was applied to the city of Campinas in Brazil (see Fig 1A), the third most popu-
lous municipality in São Paulo state with an estimated population in 2020 of 1,213,792 people
in a territory of 794,571 km2 [64]. One of the richest cities in Brazil, Campinas has gone
through an accelerated urbanisation process in the last decade. Campinas, as many other cities
in Brazil, is challenged by fast growth and urbanisation—between 1990 and 2018, the popula-
tion of Campinas grew by 70% [64] and the urban area increased by 72% [65]. It has nonethe-
less resources to monitor its infrastructure and potential to improve its urban water security.
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
5 / 25
PLOS WATER Assessing inequalities in urban water security
Fig 1. Study area. Location of the municipality of Campinas, Brazil and its territorial division: (A) Country and State
and (B) Campinas territorial units. Country and State basemaps source: IBGE (Brazilian Institute of Geography and
Statistics) https://www.ibge.gov.br/geociencias/organizacao-do-territorio/malhas-territoriais/15774-malhas.html [67,
68] available under open license. Territorial units basemap source: Campinas geospatial database from the Campinas
Municipal Council https://informacao-didc.campinas.sp.gov.br/metadados.php [69], freely available to use.
https://doi.org/10.1371/journal.pwat.0000213.g001
In addition, Campinas has five water treatment plants and, located at the meeting of three
river basins, it has a collection system divided into 15 sewer catchments relying on over 20
wastewater treatment plants to serve the urban area [66], which makes this city an interesting
case study for geospatial analysis of water security.
The municipality recognises 77 territorial units within the urban perimeter and eight in the
rural area [70]. These territorial units are defined by the city’s development plan [70, 71] as the
smallest territorial divisions (Fig 1B) that configure portions of the urban space that maintain
a significant degree of homogeneity in terms of patterns or use of land and socio-economic
characteristics [71]. Already used by the local government, considering these sectors would
facilitate communication with stakeholders, therefore, we adopted these as spatial units for
application of the framework and study of urban security distribution in the urban area.
Data collection and processing
To quantify the indicator variables, secondary data were collected from reliable official data-
bases, government agencies and organisations. Sources such as activity reports from the local
water utility [72], surveys from the Brazilian National Institute of Statistics [64], municipal
diagnostic reports [66], etc, were used for data collection. The use of public data renders the
process transparent and reproducible by other parties. The data used in the application ranged
between 2010 and 2014 as a consequence of availability. We have chosen to take a snapshot in
time to have a consistent relationship between indicators. Using too large a time range could
lead to an inconsistent view of the situation. The data sources and time frame can be found in
the supplementary material (see S1 File), along with further details on data collection and
normalisation.
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
6 / 25
PLOS WATER Assessing inequalities in urban water security
Data were collected for the city scale and when possible, to sectors within the city. Nonethe-
less, data were not always available at the scale of the sectors. In these cases, data were gathered
at the smallest possible intra-urban scale and then transformed to the scale of the territorial
units for the calculation of the sub-indexes (level of categories and dimensions). This transfor-
mation to the required sector scale was carried out using free and open-source software QGIS
(version 3.16). Data analysis, normalisation, aggregation, and visualisation was carried out
using GeoPandas (version 0.10.2) package for Python. To deal with missing data, a spatial
interpolation using the k-nearest neighbours’ method was carried out using the Scikit-learn
(version 1.1.1) Python machine learning library. Once the data for all the indicators have been
represented in the same scale, sub-indexes were calculated and urban water security maps for
each category and dimensions were created to convey their spatial variability.
Data analysis
The number of divisions inside the city boundaries for the original data scale was considered
as the sample size (n) for that measure. For example, an indicator where only one measure was
available for the entire city boundary had a sample size of 1, while indicators for which data
were available at a small scale, and specific measurements were available for all territorial units
had a sample size of 77. The sample size was important to study the distribution of data. A
minimum of five points was required for inequality analysis.
The Theil entropy index [73], a measure of regional disparities, was adopted as an inequal-
ity measure and calculated for the indicators across the sectors. This index measures an entro-
pic distance between groups and an ideal state of equality, where all regions would have the
same income, for example. It ranges between 0 (for ideal equality) and 1, with higher values
indicating higher inequality. Usually adopted to measure economic inequality—used by the
OECD to evaluate inequality in terms of productivity (GDP per worker at place of work) and
GDP per capita for instance [74]—the Theil index can be employed to measure any variable of
interest, from income inequality, to carbon intensity disparities across countries [75] and
inequality in access to improved water source [76]. It is calculated according to Eq 1.
Theil ¼
� �
yi
m
yi
m
ln
1
n
Xn
i¼1
ð1Þ
with N as the sample size, yi the indicator (variable of interest) in the sector and μ the mean
across the regions. The analysis of inequality is carried out at the indicator level in order to
investigate what causes the observed variation in each dimension, but only when a sample size
equal or larger than 5 is available. Indicators with higher levels of inequality were selected for
an analysis of spatial autocorrelation. This allows us to evaluate how the score of an indicator
in a sector correlates with neighbouring observations and to investigate the existence of pat-
terns in the geographical distribution of the indicators.
The global spatial correlation is a measure of aggregation of an attribute in the entire study
area. Derived from the Pearson correlation coefficient, the statistic used is Moran’s I [77]. The
null hypothesis tested is that a certain attribute is randomly distributed in the study area and
the computation of an empirical p-value allows us to reject or accept the null hypothesis. A sta-
tistically significant p-value (we adopt p = 0.05) indicates a spatial distribution of the variable
more spatially clustered than expected if the values were randomly allocated. Similar to corre-
lation coefficients, the Moran’s I can be positive or negative, between -1 to 1, with the higher
correlation strength to values closest to 1 in absolute value. The positive spatial correlation
indicates tendency to clustering of similar values while a negative coefficient, the clustering of
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
7 / 25
PLOS WATER Assessing inequalities in urban water security
dissimilar values. The global Moran’s I statistic is given by Eq 2.
P
i
I ¼
P
i
n
P
jwij
P
P
jwijzizj
iz2
i
ð2Þ
with n the number of observations (spatial units, indexed by i and j), zi the standardised value
of the variable of interest at location i, and wij the spatial weight (i-th row and j-th column).
Following the analysis of global spatial correlation, a further spatial analysis of local correla-
tion was carried out. Using local Moran’s I (or LISA—Local Indicators of Spatial Association),
we can identify clusters where unusual values are concentrated in space. Areas where values
are above or below the mean are clustered and four situations can be identified: two when
regions with high/low indicators are surrounded by regions with similar values (High/High
and Low/Low, HH and LL respectively) and two when regions with high/low indicators are
close to regions with opposite values (High/Low and Low/High, HL and LH, respectively)
[78]. Derived from Moran’s I, the local Moran’s Ii is given by Eq 3:
zi
m2
wijzj; where m2 ¼
Ii ¼
X
P
;
ð3Þ
j
i
iz2
n
with n the number of units, zi the standardised value of the variable of interest at location i,
and wij the spatial weight (i-th row and j-th column). The spatial correlation analysis was car-
ried out using PySAL: Python Spatial Analysis Library (version 2.6.0).
The code used for the data analysis and result figures presented in this paper is available at:
https://github.com/J-Marcal/WSF_IneqAnalysis.
Inclusivity in global research
Additional information regarding the ethical, cultural, and scientific considerations specific to
inclusivity in global research is included in the supporting information (see S2 File).
Results and discussion
Urban water security evaluation
The task of populating the list of indicators revealed different levels of data availability and
granularity for the city of Campinas. Several indicators only had values for the entire city, espe-
cially for water quantity, climate change and governance. This process allowed us to audit the
accessibility of free data for this case study and to note the impacts on the following assess-
ment. Data at a small scale may be further available within stakeholders’ organisations, how-
ever, for transparency reasons only freely accessible data were used in this study.
Most the of granular data available issued from a decennial national survey carried out by
the Brazilian Institute of Geography and Statistics [64]. Incorporating small scale monitoring
to the local agenda and making that information available is important to better investigate
certain aspects, especially in terms of governance and risks and climate change. Information
such as temperature differences in the urban space can provide insights on urban heat island
fluctuations for instance. These have been found to be related to urbanisation pattern and hav-
ing influence on public health [79], therefore, detailed information on spatial distribution of
temperature in urban areas can prompt public action and help improve different dimensions
of water security. Nevertheless, small-scale free information from the state or municipality was
difficult to find. Data for some indicators, such as diversity of sources (A1.3), metering level
(A4.3) and water loss (A4.4) (Dimension A), were only available for the city scale, therefore, all
the sectors received the same score and a study of inequality in the city was not possible. This
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
8 / 25
PLOS WATER Assessing inequalities in urban water security
was also the case for several aspects of dimensions C and D, for which data at a small scale was
less available. This hinders the assessment on the urban water security heterogeneity since it is
difficult to conclude if this is related to homogeneity of the urban area or if there was not
enough data to translate the existing variability.
The results of the assessment at city and sector scales are presented to each of the four
dimensions in Figs 2–5. These show the scores attributed for the city as bars and the scores cal-
culated for sectors as circular markers. The size of the circular marker indicates the population
living in each sector. The scores range from 1 to 0, with desirable characteristics given ‘1’ and
undesirable values, ‘0’. This visualisation shows the interest of our framework since it high-
lights the dispersion existent within the studied area for high scoring indicators, such is the
case of affordability (A3.3) and access to wastewater collection (A3.2).
When aggregating the categories for the four dimensions for the sectors in the city, the spa-
tial distribution of the results can be visualised, as seen in Fig 6. Different scores are visibly dis-
tributed in the urban area, given an indication of existing spatial inequalities of water security.
These results show less differentiation for dimensions C and D, for which granular data was
less available. Nonetheless, even with the challenge of data availability, adding the spatial
dimension to water security assessment allowed us to show, for all four dimensions consid-
ered, some variability in the aggregated scores. The results support the need to investigate
inequality within the city boundary rather than considering the average value for the entire
urban area.
The results for the Drinking water and human well-being dimension (A) show that, in gen-
eral, few districts have a below average score (Fig 6A), while diversity can be observed within
the municipality when investigating separate categories and indicators (Fig 2). Water quantity
Fig 2. Results of assessment for city and sector scales for Drinking water and human well-being (Dimension A).
https://doi.org/10.1371/journal.pwat.0000213.g002
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
9 / 25
PLOS WATER Assessing inequalities in urban water security
Fig 3. Results of assessment for city and sector scales for Ecosystems (Dimension B).
https://doi.org/10.1371/journal.pwat.0000213.g003
Fig 4. Results of assessment for city and sector scales for Water related hazards and climate change (Dimension C).
https://doi.org/10.1371/journal.pwat.0000213.g004
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
10 / 25
PLOS WATER Assessing inequalities in urban water security
Fig 5. Results of assessment for city and sector scales for Economic and social development (Dimension D).
https://doi.org/10.1371/journal.pwat.0000213.g005
(A1) was found to be the most concerning category for the case study, with the lowest scores in
the dimension, and water stress (A1.5) being the main challenge for the city (see Fig 2). The
high concentration of people and economic activities in the region, associated with decreasing
water availability over the years results in constant pressure in the basin’s water resources and
a low score for the city. The region has faced water crises in 2014 and 2016, while the available
water quantity is a continuous concern of local organisations [80].
Regarding accessibility to services (A3), Campinas has been able to establish very good con-
ditions in the urban area. Yet, it is possible to see markers with low score, representing sectors
where challenges are still present as shown in Fig 2. Data on sewage coverage (A3.2) for
instance, showed some deficiency in the infrastructure of certain sectors in the outskirts of the
city. For the last decade a plan to achieve universal sanitation has been implemented by the
water utility [66]: for the time scale of this study, 83% of the population had access to sewage
collation, a percentage that increased to 94% in 2020 [72]. According to the Sustainable Cities
Program, Campinas has achieved the goals for water supply and sewage collection and treat-
ment from the SDG 6 but still faces challenges regarding water loss [81].
In terms of reliability of services (infrastructure reliability (A4)), measures of non-scheduled
maintenance services (service reliability (A4.2)) were found for the different sewage collection
systems, allowing visualisation of some variability between the sectors, especially highlighting
low scores in the outskirts and south of the municipality. As for public health and well-being
(A5), with little incidence of gastrointestinal infections (incidence of water-borne diseases
(A5.1)) throughout the territory, the main component leading to diversity in this category was
accessibility to green social areas (recreational opportunities (A5.2)). A very dispersed set of
results showed an unequal distribution of scores, with districts in the centre having good
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
11 / 25
PLOS WATER Assessing inequalities in urban water security
Fig 6. Spatial distribution of water security. Aggregated results for (A): Drinking water and human well-being (B): Ecosystems (C):
Water related hazards and climate change and (D): Economic and social development. Labels on the maps show the highest and lowest
scores found for each dimension. Territorial units basemap source: Campinas geospatial database from the Campinas Municipal Council
https://informacao-didc.campinas.sp.gov.br/metadados.php [69], freely available to use.
https://doi.org/10.1371/journal.pwat.0000213.g006
access to parks and gardens and therefore high scores while sectors at the outskirts of the city
received low scores.
The heterogeneity of scores was more prominent for the dimension Ecosystems (B)(see Fig
6B) that also had the lower score, ranging between 0.34 and 0.74 for the urban sectors. Investi-
gation of the categories of this dimension showed that indicators related to green coverage and
environmental diseases, from the Environment (B1) category, presented relative low average
scores and high dispersion within the city boundary (Fig 3). Campinas, as many other urban
areas in tropical and subtropical regions, faces challenges with environmental safety (B1.2)—or
water-vectored—diseases, such as dengue fever. These are related to high population density,
irregular supply, waste management, etc [82, 83]. The results also demonstrate challenges
regarding green coverage (green areas (B1.1)). These are common to the urban context, due to
the urbanisation process and high urbanisation rate in the city (in Campinas, of about 98%)
[64].
In terms of the pollution control (B2) category, intra-city granular data for groundwater and
surface water quality (B2.1 and B2.2) were not available, and therefore, little differentiation was
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
12 / 25
PLOS WATER Assessing inequalities in urban water security
observed for these aspects. As for wastewater treatment rate (B2.3), data from wastewater col-
lection systems allowed us to verify diversity within the city. For the time scale analysed,
improvement was required in some sectors, especially in the south of the city. However, sub-
stantial investment has taken place in the last decade which improves the score for this indica-
tor. The wastewater treatment rate in the city increased from 72% in 2010 to 89% in 2020, with
the water utility goal expected to be reaching 100% by 2025 [72].
A reuse water station, using membrane bioreactor (MBR) technology, is installed and in
operation since 2012 in the south of the city. For this area, high removal efficiency is accompa-
nied by high energy consumption, leading in some sectors to relatively low scores for the
energy usage efficiency (B3.1) indicator [72] (Fig 3). Other districts that have their wastewater
treated by energy demanding activated sludge and aerated ponds technology, also had lower
scores for this indicator. As to wastewater reuse (B3.2), the practice is still limited due to legis-
lation restrictions, resulting in a very low score overall. However, with a second water reuse
station inaugurated in 2021, there is great potential to improve usage efficiency in the city of
Campinas for the next decade [72].
As for dimension C: Water related hazards and climate change, in terms of water hazards
(C1), Campinas did not face any drought during the decade preceding the evaluation date
[84], and, although it has faced several flood events, the proportion of flood prone areas varies
considerably in the sectors (see Fig 4). As for preparedness (C2), a wide distribution of drainage
infrastructure and people living in hazardous areas was found. Nonetheless, due to lack of
available granular data for other indicators in the dimension, possible existing spatial variation
was attenuated and rendered virtually invisible in the final visualisation map (see Fig 6C).
Related to the SDG 13—urgent action to combat climate change and its impacts [85], the
scores of dimension C are supported by the results found in the Sustainable Cities Program of
which Campinas has taken part since 2012 [81]. This program monitors participant cities in
Brazil and evaluates them in terms of the Sustainable Development Index, adopting SDG indi-
cators. According to their results, Campinas scores highly in terms of climate change perfor-
mance, which also included greenhouse gas emissions and strategies for risk management and
prevention of natural disasters.
For dimension D—Economic and social development, the spatial distribution of the aggre-
gated score was similar to dimension C. It is less noticeable but still exists (see Figs 5 and 6D).
This is expected in view of data collection challenges and low sample sizes obtained for some
indicators in these dimensions: the lack of data granularity prevents the grasp of urban
inequalities. Governance (D1) aspects in particular were only feasible at the city scale and
therefore, no distinction is made for the sectors. Granularity was available for social aspects
(D2) indicators and therefore, it was possible to observe a distribution of scores in the city for
this category (see Fig 5). Gender equality (D2.6) results showed low scores throughout the
municipality with only few sectors with a scores above 0.5. This was confirmed by the a similar
low score received by the city of Campinas in the Sustainable Cities Program [81] for the SDG
5—Achieve gender equality and empower all women and girls, considering participation of
women in decision making positions, wage inequality among others, major challenges were
identified in order to achieve this specific goal. Interestingly, the score for income inequality
(D2.3) was smaller for the city than for the sectors, an indication that the sectors are somewhat
homogeneous, but differences can be found between them. This is supported by the results of
average income (D2.4) that show a great dispersion of results (see Fig 5). As for economic devel-
opment (D3) indicators, data were available only for the city scale, and translated the favour-
able economic position of the city—Campinas is a relatively wealthy city with one of the
highest GDPs of the state [64].
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
13 / 25
PLOS WATER Assessing inequalities in urban water security
The use of granular data and spatial visualisation clearly highlights the intra-urban variabil-
ity for the different water security aspects. Similar to the results of Tholiya and Chaudhary [17]
on the performance water supply services and Doeffinger and Hall [14] on sub-national water
security assessment, the geospatial visualisation demonstrates the heterogeneity of the studied
area. This helps to expose vulnerable regions, and therefore, could inform and support effec-
tive decision making.
Assessing inequality
The inequality of the water security indicators is measured in terms of the Theil entropy index.
Results are presented in Fig 7. This figure shows the results of the inequality index against the
scores for the sectors, with the ideal setting being high scores and low inequality index (0
would be ideal equality)—the bottom right quadrant, where most indicators are placed for
Campinas.
Among the indicators from dimension A, data for recreational opportunities (A5.2) show a
high inequality score (see Fig 7). Recreational green areas are important for well-being and life
quality in urban spaces, nevertheless, intense urbanisation can often neglect this aspect. Cam-
pinas, in 2010, had 23 parks and other public green spaces for a population of 1,080,113 people
[64], nonetheless, these were concentrated in certain areas and according to the local Environ-
mental Office, 70% of the districts had no local social green area [86]. In our study we consider
the proximity of people to these areas, but we still find almost 20% of districts with no public
Fig 7. Inequality vs scores quadrant plot of inequality indexes versus scores for the analysed indicators.
https://doi.org/10.1371/journal.pwat.0000213.g007
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
14 / 25
PLOS WATER Assessing inequalities in urban water security
green area within a 30-minute walk. Considering the distance to these local areas also has an
effect on the distribution of the results. Even so, the presence of a range of scores shows
inequality and consequently different levels of well-being resulting from the access to green
areas. The disparity is being addressed by the local government—a municipal Green Plan,
established in 2016, targets the deficit of social green areas and aims to implement linear parks
in the city [86].
For the accessibility to services category (A3), very high scores were obtained overall, with
water supply coverage (A3.1) specially clustered with low inequality index (see Fig 7) associated
with high scores, indicating a very favourable situation for the city—99.5% of the urban popu-
lation is connected to the drinking water supply network [66]. The results for wastewater col-
lection coverage (A3.2), on the other hand, show a higher dispersion and larger range of scores.
Over 80% of the urban population had access to sewage collection in 2010 [72], and a Sanita-
tion Program is in place aiming to provide the entire city with this service [66]. Nonetheless,
the data set in this study shows areas, especially at the urban edges, where the population still
lacks sewage connection, relying on individual solutions [66].
Data are especially unequal for green areas (B1.1) and environmental safety (B1.2), for
which data on the occurrence of environmental safety diseases are not only scattered but also
tending to low scores, resulting in the highest inequality index of the dimension. In 2010,
Campinas faced a large dengue fever epidemic with the majority of cases in health centres in
the Northwest area of the city [87]. In this study, low scores were attributed to several districts
based on data from 61 local health centres, which, overall contributed to the resulting low and
disperse score of the indicator and, therefore, of dimension B. Despite that, Campinas has
resources to carry out prevention and warning actions and in 2015 the municipal government
established a committee for combating arbovirus infections (such as dengue, yellow and Zika
fevers) and coordinate prevention and response actions between different stakeholders [88].
Also a concerning aspect for the dimension B, the overall percentage of green areas (B1.1)
to the total area is low in Campinas and in addition, the data show an unequal distribution
regarding vegetation coverage, with specially low percentage in the city centre. This is closely
related to the urbanisation process, high urbanisation rate (about 98%) [64] and population
density [86]. Since 2013, the municipality has worked on the recovery of green areas by plant-
ing trees and improving the inspection to promote natural regeneration [86]. In contrast, solid
waste collection (B4.1) presented a very clustered and high score result, with lower inequality
index (see Fig 7).
As for dimension C, flood-prone areas (C1.3) and presence of storm drains (urban drainage
(C2.2)) presented average scores and the highest inequality results for the dimension. The
flood-prone areas (C1.3) are often related to insufficient drainage systems, increase of imper-
meable areas and occupation of valleys [66]. In terms of urban drainage (C2.2), data show that
only 57% of the public roads have underground storm drains in the urban area [89]. Even if
one argues that not all roads need storm water drains due to the geography of the watersheds,
the results still show an important variation in the urban zone that can increase the vulnerabil-
ity of certain areas. The other indicators analysed for this dimension (paved streets (C2.3) and
people living in hazardous zones (C1.4)) are located at the bottom right quadrant, showing an
overall good score and low inequality measure. This is compatible with the situation in Campi-
nas, where a total of 2% of the of the households living at risk according to the municipal civil
defence [89] and the majority of the streets in the urban area are paved (95% [66]).
Concerning social aspects (D2), literacy rate (D2.1) presented the highest overall score
amongst the analysed indicators of dimension D. Literacy is crucial for the understanding of
water issues and therefore the success of collective action. With a very clustered data set (low
inequality index, as seen in Fig 7), the analysis shows a very favourable and consistent situation
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
15 / 25
PLOS WATER Assessing inequalities in urban water security
for Campinas, yet, when considering the large number of inhabitants of the city, in 2010 the
number of people above 15 years old who were not able to read and write was over 28 thou-
sand people [64]. Since 2014 a campaign to end illiteracy has been carried out by the munici-
pality, showing great progress in the last decade: the illiteracy rate dropped 46% by 2019 [90].
In terms of income, analysis of the Gini Coefficient (inequality (D2.3)) showed that income
inequality inside the districts (comparing incomes inside the same sector) resulted in a rather
clustered data set. Interestingly, the results for average income (D2.4) in the city showed a
more spread-out behaviour with higher inequality index. This indicates that, while inside the
sectors a more homogeneous situation in terms of income may be found, different sectors are
living different realities: results showed an average income ranging between 2.5 and 35 mini-
mum wages [64]. The lowest average incomes were found to be in the south, southwest and
north edges of the city, somewhat coinciding with areas where deficit of infrastructure was
observed in the other dimensions.
The population living in informal settlements (informal dwellings (D2.5)), considered in the
assessment of SDG 11- Make cities and human settlements inclusive, safe, resilient and sustain-
able, is identified by the Sustainable Cities Program as a big challenge for Campinas [81]. The
results in this study showed a generally clustered data set for this indicator (D2.5). This is due
to the vast majority of districts having no or a small percentage of people living in such settings
and therefore, high scores for this indicator. Nonetheless, the outliers in this case are signifi-
cant: a few districts, especially in the south of the city, have higher proportions with up to 80%
of the residents living in informal settlements [64]. These areas are classified as highly vulnera-
ble by the São Paulo Social Vulnerability Index, an assessment tool to identify areas most vul-
nerable to poverty [91].
Another social aspect that deserves attention is gender equality (D2.6), with low scores
across the city (see Fig 7). Related to SDG 5—Achieve gender equality and empower all women
and girls, this indicator shows great challenges for the city of Campinas (SDG 5 in Campinas
received the lowest score in the Sustainable Cities Program evaluation [81]), translating the
inequality of incomes in households headed by women and men. The present analysis placed
this indicator in the bottom left quadrant of Fig 7 indicating a deficient and considerably uni-
form situation with low scores and low dispersion and inequality measures.
The quantification of inequality for water security indicators provides a valuable tool for
decision making. It raises flags on which indicators show a wide, non-uniform distribution in
the urban area. In addition, including this aspect allows us to quantitatively consider water
security equity in the city, informing decision makers on aspects that require action to tackle
inequalities.
Spatial variation. Dimension A, on drinking water and human well-being showed impor-
tant variability for certain aspects such as access to recreational areas (A5.2) and wastewater
collection (A3.2). The results from the spatial analysis showed some overlay between the low
scoring regions for these indicators (Fig 8).
Wastewater collection (Sewage coverage(A3.2)) scores showed a significant positive spatial
correlation, with a Moran’s I value of 0.522 and p-value of 0.001. This indicates a tendency of
similar values being clustered in space. The results for local spatial correlation analysis showed
the spatial association around each individual sector. For (A3.2), sectors with high scores for
wastewater collection, near neighbourhoods that also have a high score (high/high score), are
located in the city centre as seen in Fig 8A. This area is therefore composed of a group of sec-
tors that have a very good infrastructure in terms of wastewater collection, while a cluster of
low scoring areas near other low areas (LL) are found in the northern and southern outskirts
of the city (see Fig 8A). The deficient areas (Low/low association, or cold spots) identified
make up 8% of the urban area analysed and take in 4% of the population of Campinas. These
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
16 / 25
PLOS WATER Assessing inequalities in urban water security
Fig 8. Moran cluster maps for dimension A. (A): Moran cluster map for indicator (A3.2) Sewage coverage and (B): Moran cluster
map for indicator (A5.2) Recreational opportunities. Territorial units basemap source: Campinas geospatial database from the
Campinas Municipal Council https://informacao-didc.campinas.sp.gov.br/metadados.php [69], freely available to use.
https://doi.org/10.1371/journal.pwat.0000213.g008
results are in agreement with the diagnostics obtained in the Municipal Sanitation Plan of
2013 [66], particularly with respect to the neighbourhoods that lack sewage collection infra-
structure. The cluster in the south of the city encompasses vulnerable neighbourhoods charac-
terised by high population density, low income, and informal settings. The north cluster units
do not include informal settlements and, although not as socially vulnerable as the ones in the
south cluster, consist of isolated urban patches across the rural area. This entails certain infra-
structure shortcomings such as households relying on individual solutions for wastewater
management.
A similar trend is found for recreational opportunities (A5.2) (see Fig 8B), for which low/
low areas (cold spots) are situated on the suburbs (covering 8.9% of the urban area and near
5% of the population) while more central areas appear as a high scoring cluster. The develop-
ment of green social areas was found to be associated with public or private interests during
the urbanisation process of the city [92]. That led to parks and other social green areas being
located in more developed areas, where there was interest of capital, contributing to the
observed inequalities. For this indicator, a negative local association is observed: a unit with
low score, that is, a sector with little access to green social areas but surrounded by districts
with accessible parks and other recreational opportunities. Despite that, overall, the indicator
shows a positive global spatial correlation (similar regions tending to cluster) with Moran’s I of
0.659 and p-value of 0.001.
Contrary to the trends observed with the indicators belonging to dimension A, areas with a
low score for green areas (B1.1) appear clustered in the centre of the municipality (see Fig 9A).
With a positive overall correlation (Moran’s I of 0.303 and p-value of 0.002), the local analysis
showed clusters of low/low association (cold spots) in the highly urbanised and dense city cen-
tre. The areas included in this cluster house 17% of the urban population and count for 8% of
the area. This situation, connected to the urbanisation rate and process in the city, is closely
related to the environmental pressures and the need to increase green areas in the city. The
most recent municipal plans for conservation and recovery of native vegetation targets urban
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
17 / 25
PLOS WATER Assessing inequalities in urban water security
Fig 9. Moran cluster maps for dimension B. (A): Moran cluster map for indicator (B1.1) Green areas and (B): Moran cluster map
for indicator (B1.2) Environmental safety. Territorial units basemap source: Campinas geospatial database from the Campinas
Municipal Council https://informacao-didc.campinas.sp.gov.br/metadados.php [69], freely available to use.
https://doi.org/10.1371/journal.pwat.0000213.g009
green areas as well as plans for the construction of several linear parks within the urban area
have been announced [86].
In terms of environmental safety (B1.2), areas with low scores (high incidence of environ-
mental related diseases, such as dengue fever) show tendency to gather (positive spatial Moran
´s I of 0.467, p-value of 0.001) in the northwest and south of the city (see Fig 9B). The contrib-
uting factors to the observed clustering pattern may be related to heterogeneity of infrastruc-
ture, land occupation or life habits [88]. The clusters of low scores are consistent with the areas
of high incidence of dengue fever identified by Johansen et al. [82] when analysing the rela-
tionship between social inequality and dengue fever incidence in Campinas. They emphasise
the expansion of the peri-urban areas as a cause of spatial segregation and inequality in the
access to urban resources and services.
The low values of dispersion and inequality observed for dimension C is also translated to
the analysis of spatial autocorrelation. The indicator on the presence of storm drains (urban
drainage (C2.2)) presents a data set with a low tendency to cluster with Moran´s I of 0.193 (p-
value of 0.015). A small area of cold spots is observable in the northern outskirts of the city in a
small area corresponding to 2.3% of the urban area housing a little over 0.33% of the urban
population (see Fig 10A). The region is also an area of low/low scores association for wastewa-
ter collection (A3.2) and recreational opportunities (A5.2), indicating a set of challenges in the
area.
As for dimension D, income (D2.4) showed higher values of dispersion and Theil index.
Presenting a positive tendency to cluster (Moran´s I of 0.575, p-value of 0.001), a large cold
spot (low/low) is found in the southwest of the city (see Fig 10B), covering 25% of the area and
23% of the urban population. The area in the south of the city is also part of the identified low/
low scoring association clusters for wastewater collection (A3.2) and recreational opportunities
(A5.2). This is a region where informal settings (D2.5) are predominant and communities are
highly social vulnerable according to the São Paulo Social Vulnerability Index [91]. It is also
noticeable that the identified cluster of high income areas (high/high score association)
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
18 / 25
PLOS WATER Assessing inequalities in urban water security
Fig 10. Moran cluster maps for dimensions C and D. (A): Moran cluster map for indicator (C2.2) Urban drainage and (B): Moran
cluster map for indicator (D2.4) Income. Territorial units basemap source: Campinas geospatial database from the Campinas
Municipal Council https://informacao-didc.campinas.sp.gov.br/metadados.php [69], freely available to use.
https://doi.org/10.1371/journal.pwat.0000213.g010
presents some overlap with high scoring areas for access to recreational areas (A5.2) (as seen
in Fig 8B). These overlays and regional disparities identified show that areas rarely face one
specific water security challenge. As the dimensions of water security are interconnected, so
are the challenges and advantages brought by infrastructure, policies, and management strate-
gies. Therefore, a holistic and in-depth water security evaluation is crucial for sustainable
urban water management.
Perspectives on the approach
We applied a holistic framework to assess water security in the city of Campinas, Brazil, and
investigate the heterogeneity of its aspects in the urban context. This was done by incorporat-
ing inequality and spatial analysis to the assessment in order to reveal what the challenges are
and how they are distributed in the urban area. This study was also presented to experts and
water professionals in the field that have provided valuable feedback and perspectives on this
approach.
Although data availability is viewed as a challenge to such detailed analysis, the potential of
this downscaled assessment lies in the visual component which enables identifying what and
where the urban water security problems are. This is considered to be an important asset to
communication with policymakers. A follow up on possible solutions for local issues and their
costs could then lead to regenerative actions. This type of approach can help raise ‘red flags’ in
terms of what areas are being overlooked and realities that are getting lost in averages.
There is also potential in learning from within the city: sharing experiences and successes
between different sectors or neighbourhoods on local initiatives, as well between stakeholders
from different areas, equivalent to city-to-city learning [93].
The work for such detailed analysis is more labour-intensive than traditional water security
assessment frameworks, but the involvement of stakeholders, when applying such approach
can help obtain data and determine priorities. It is also important to consider the flexibility in
terms of choice of indicators: this approach can be carried out for any indicator, depending on
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
19 / 25
PLOS WATER Assessing inequalities in urban water security
local issues and data availability. This flexibility should still be guided by the concept of water
security and the different dimensions and aspects involved. Certain indicators adopted in this
study, especially the inclusion of solid waste management, drew attention to its importance in
water management within urban areas.
Including the inequality index as an extra measure and the spatial analysis to assess water
security is therefore an asset to reveal hidden issues and tackle inequality in a local and targeted
manner.
Conclusions
Including spatial and inequality analysis deepens the assessment of water security in the urban
context. Downscaling the water security assessment presents both an opportunity and a chal-
lenge. Increasing the granularity of the evaluation allows incorporating the spatial dimension
in the assessment and therefore investigating inequalities within urban boundaries. On the
other hand, large data availability is required for evaluation.
In general, adding the spatial component to water security assessment enriches the evalua-
tion allowing identification of spatial inequalities. The hierarchical approach allows each level
to be uncovered to investigate where the differences lay. Challenges can then be pinpointed,
and solutions proposed. For that, information at a smaller scale is key. Downscaling water
security assessment is therefore a way to also audit the accessibility of data. Since large scale
data can mask variability, downscaled assessment has the potential to encourage small-scale
monitoring in urban areas, which, in turn, can promote the analysis of water security inequali-
ties. Including measures of inequality in the urban water security assessment helps to identify
aspects for which the city has reached an overall positive situation and where important differ-
ences still linger. This will then create incentives and opportunities to leave no-one behind.
The presented case study analysis allowed identification of local challenges for Campinas.
While infrastructure challenges still remain in sectors in the north and south of the city, the
highly urbanised centre lacks green coverage. Despite being a rich city, income inequality is
present and the connection between economic and social vulnerability with other aspects of
water security was identified. There is potential to achieve a more sustainable water cycle, espe-
cially in terms of wastewater reuse. Actions by the municipality, such as the Sanitation Pro-
gram, show great effort to ensure equitable water services in the city. The assessment for
Campinas represents a snapshot in time, with more recent data having been delayed due to the
COVID-19 pandemic. Incorporating the temporal aspect in the analysis would allow compari-
son of the progress of the city in each water security dimension. This would be a valuable con-
tribution for future work.
Ultimately, the proposed assessment delivers a visual tool to communicate regional dispari-
ties and challenges in the urban area. This can help facilitate communication with different
stakeholders by including what and where in the outcomes of the urban water security
assessment.
Supporting information
S1 File. Normalisation of indicators. Description of indicators, metrics, data sources, and
normalisation thresholds.
(PDF)
S2 File. Inclusivity in global research. PLOS Inclusivity in global research questionnaire.
(DOCX)
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
20 / 25
PLOS WATER Assessing inequalities in urban water security
Author Contributions
Conceptualization: Juliana Marc¸al, Jan Hofman.
Data curation: Juliana Marc¸al.
Formal analysis: Juliana Marc¸al.
Investigation: Juliana Marc¸al, Jan Hofman.
Methodology: Juliana Marc¸al, Jan Hofman.
Supervision: Junjie Shen, Blanca Antizar-Ladislao, David Butler, Jan Hofman.
Validation: Juliana Marc¸al, Junjie Shen, Blanca Antizar-Ladislao, David Butler, Jan Hofman.
Visualization: Juliana Marc¸al, Jan Hofman.
Writing – original draft: Juliana Marc¸al.
Writing – review & editing: Juliana Marc¸al, Junjie Shen, Blanca Antizar-Ladislao, David But-
ler, Jan Hofman.
References
1.
Intergovernmental Panel on Climate Change (IPCC). Climate Change 2021: The Physical Science
Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel
on Climate Change [Internet]; 2021. Available from: https://www.ipcc.ch/report/ar6/wg1/ [cited 2023 Oct
11].
2.
Toreti A, Bavera D, Acosta Navarro J, Cammalleri C, de Jager A, Di Ciollo C, et al. Drought in Europe
August 2022 [Internet]. Publications Office of the European Union; 2022. Available from: https://edo.jrc.
ec.europa.eu/documents/news/GDO-EDODroughtNews202208_Europe.pdf [cited 2023 Oct 11].
3. Bakker K. Water Security: Research Challenges and Opportunities. Science. 2012; 337(6097):914–
915. https://doi.org/10.1126/science.1226337 PMID: 22923564
4. Cook C, Bakker K. Water security: Debating an emerging paradigm. Global Environmental Change.
2012; 22(1):94–102. https://doi.org/10.1016/j.gloenvcha.2011.10.011
5. Allan JV, Kenway SJ, Head BW. Urban water security—what does it mean? Urban Water Journal.
2018; 15(9):899–910. https://doi.org/10.1080/1573062X.2019.1574843
6. Hoekstra AY, Buurman J, van Ginkel KCH. Urban water security: A review. Environmental Research
Letters. 2018; 13(5):053002. https://doi.org/10.1088/1748-9326/aaba52
7. Marc¸al J, Antizar-Ladislao B, Hofman J. Addressing Water Security: An Overview. Sustainability. 2021;
13(24). https://doi.org/10.3390/su132413702
8. United Nations | UN Water. What is Water Security? Infographic [Internet]; 2013. Available from: https://
www.unwater.org/publications/water-security-infographic/ [cited 2023 Oct 11].
9. United Nations—Department of Economic and Social Affairs. Urbanization: Expanding Opportunities
but Deeper Divides. In: World Social Report 2020: Inequality in a Rapidly Changing World. United
Nations publication; 2020. p. 107–126.
10. Bahri A. Integrated Urban Water Management [Internet]; 2012. Available from: https://www.gwp.org/
globalassets/global/toolbox/publications/background-papers/16-integrated-urban-water-management-
2012.pdf [cited 2023 Oct 11].
11. Asian Development Bank (ADB). Asian water development outlook 2016: Strengthening water security
in Asia and the Pacific [Internet]; 2016. Available from: https://www.adb.org/sites/default/files/
publication/189411/awdo-2016.pdf [cited 2023 Oct 11].
12. United Nations | UN Water. Sustainable Development Goal 6: Synthesis Report on Water and Sanita-
tion [Internet]; 2018. Available from: https://sustainabledevelopment.un.org/content/documents/
19901SDG6_SR2018_web_3.pdf [cited 2023 Oct 11].
13. Baldwin DA. The concept of security. Review of International Studies. 1997; 23(1):5–26. https://doi.org/
10.1017/S0260210597000053
14. Doeffinger T, Hall JW. Assessing water security across scales: A case study of the United States.
Applied Geography. 2021; 134:102500. https://doi.org/10.1016/j.apgeog.2021.102500
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
21 / 25
PLOS WATER Assessing inequalities in urban water security
15. Gain AK, Giupponi C, Wada Y. Measuring global water security towards sustainable development
goals. Environmental Research Letters. 2016; 11(12). https://doi.org/10.1088/1748-9326/11/12/124015
16. Stuart E, Stoler J, Pearson AL, Asiki G. Spatial heterogeneity of household water insecurity in rural
Uganda: implications for development. Water International. 2023; 48(2):282–301. https://doi.org/10.
1080/02508060.2023.2183641
17.
Tholiya JJ, Chaudhary N. Urban Water Security: Geospatial Assessment For Water Supply Services In
Pune, India. Urban Water Journal. 2022; 00(00):1–13. https://doi.org/10.1080/1573062X.2022.
2047736
18. Wakhungu MJ, Abdel-Mottaleb N, Wells EC, Zhang Q. Geospatial Vulnerability Framework for Identify-
ing Water Infrastructure Inequalities. Journal of Environmental Engineering. 2021; 147(9):04021034.
https://doi.org/10.1061/(ASCE)EE.1943-7870.0001903
19. Qiu M, Zuo Q, Wu Q, Yang Z, Zhang J. Water ecological security assessment and spatial autocorrela-
tion analysis of prefectural regions involved in the Yellow River Basin. Scientific Reports. 2022; 12(1).
https://doi.org/10.1038/s41598-022-07656-9 PMID: 35332130
20. Bichai F, Ryan H, Fitzgerald C, Williams K, Abdelmoteleb A, Brotchie R, et al. Understanding the role of
alternative water supply in an urban water security strategy: an analytical framework for decision-mak-
ing. Urban Water Journal. 2015; 12(3):175–189. https://doi.org/10.1080/1573062X.2014.895844
21. Medina-Rivas CM, Rodrı´guez-Tapia L, Morales-Novelo JA, Revollo-Ferna´ ndez DA. Spatial inequality
of domestic water consumption in Mexico city. Water Resources and Economics. 2022; 40:100210.
https://doi.org/10.1016/j.wre.2022.100210
22. Mukherjee S, Sundberg T, Sikdar PK, Schu¨tt B. An Integrated Quantitative Assessment of Urban Water
Security of a Megacity in the Global South. Frontiers in Water. 2022; 4. https://doi.org/10.3389/frwa.
2022.834239
23. Assefa Y, Babel M, Susˇnik J, Shinde V. Development of a Generic Domestic Water Security Index, and
Its Application in Addis Ababa, Ethiopia. Water. 2018; 11(1):37. https://doi.org/10.3390/w11010037
24. Chapagain K, Aboelnga HT, Babel MS, Ribbe L, Shinde VR, Sharma D, et al. Urban water security: A
comparative assessment and policy analysis of five cities in diverse developing countries of Asia. Envi-
ronmental Development. 2022; 43:100713. https://doi.org/10.1016/j.envdev.2022.100713
25.
26.
Jensen O, Wu H. Urban water security indicators: Development and pilot. Environmental Science & Pol-
icy. 2018; 83(September 2017):33–45. https://doi.org/10.1016/j.envsci.2018.02.003
van Ginkel KCH, Hoekstra AY, Buurman J, Hogeboom RJ. Urban Water Security Dashboard: Systems
Approach to Characterizing the Water Security of Cities. Journal of Water Resources Planning and
Management. 2018; 144(12):04018075. https://doi.org/10.1061/(ASCE)WR.1943-5452.0000997
27. Babel MS, Shinde VR, Sharma D, Dang NM. Measuring water security: A vital step for climate change
adaptation. Environmental Research. 2020; 185. https://doi.org/10.1016/j.envres.2020.109400 PMID:
32222634
28. United Nations—UN Water. Water Security & the Global Water Agenda. The UN-Water analytical brief
[Internet]; 2013. Available from: https://www.unwater.org/publications/water-security-and-global-water-
agenda [cited 2023 Oct 11].
29. Aboelnga HT, El-Naser H, Ribbe L, Frechen FB. Assessing Water Security in Water-Scarce Cities:
Applying the Integrated Urban Water Security Index (IUWSI) in Madaba, Jordan. Water. 2020; 12
(5):1299. https://doi.org/10.3390/w12051299
30. Carden K, Armitage N. Assessing urban water sustainability in South Africa—not just performance mea-
surement. Water SA. 2013; 39(3):345–350. https://doi.org/10.4314/wsa.v39i3.1
31.
32.
Zhang JY, Wang LC. Assessment of water resource security in Chongqing City of China: What has
been done and what remains to be done? Natural Hazards. 2015; 75(3):2751–2772. https://doi.org/10.
1007/s11069-014-1460-5
Jia X, Li C, Cai Y, Wang X, Sun L. An improved method for integrated water security assessment in the
Yellow River basin, China. Stochastic Environmental Research and Risk Assessment. 2015; 29
(8):2213–2227. https://doi.org/10.1007/s00477-014-1012-2
33. Vo¨ ro¨smarty CJ, McIntyre PB, Gessner MO, Dudgeon D, Prusevich A, Green P, et al. Global threats to
human water security and river biodiversity. Nature. 2010; 467(7315):555–561. https://doi.org/10.1038/
nature09440 PMID: 20882010
34. Shrestha S, Aihara Y, Bhattarai AP, Bista N, Kondo N, Futaba K, et al. Development of an objective
water security index and assessment of its association with quality of life in urban areas of developing
countries. SSM—Population Health. 2018; 6:276–285. https://doi.org/10.1016/j.ssmph.2018.10.007
PMID: 30480077
35. Beck L, Chesterfiled C, Brown RR, Dunn G, de Haan F, Lloyd S, et al. Beyond benchmarking: A water
sensitive cities index. OzWater. 2016;.
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
22 / 25
PLOS WATER Assessing inequalities in urban water security
36. Batten J. Sustainable Cities Water Index. Which cities are best placed to harness water for future suc-
cess? [Internet]. ARCADIS; 2016. Available from: https://www.oieau.fr/eaudoc/system/files/33715.pdf
[cited 2023 Oct 11].
37.
van Leeuwen CJ, Frijns J, van Wezel A, van de Ven FHM. City Blueprints: 24 Indicators to Assess the
Sustainability of the Urban Water Cycle. Water Resources Management. 2012; 26(8):2177–2197.
https://doi.org/10.1007/s11269-012-0009-1
38. Nie R, Tian Z, Wang J, Zhang H, Wang T. Water security sustainability evaluation: Applying a multi-
stage decision support framework in industrial region. Journal of Cleaner Production. 2018; 196:1681–
1704. https://doi.org/10.1016/j.jclepro.2018.06.144
39. Sinyolo S, Mudhara M, Wale E. Water security and rural household food security: empirical evidence
from the Mzinyathi district in South Africa. Food Security. 2014; 6(4):483–499. https://doi.org/10.1007/
s12571-014-0358-0
40.
Jepson WE, Wutich A, Colllins SM, Boateng GO, Young SL. Progress in household water insecurity
metrics: a cross-disciplinary approach. Wiley Interdisciplinary Reviews: Water. 2017; 4(3):e1214.
https://doi.org/10.1002/wat2.1214
41. PRI Project Sustainable Development. Canadian Water Sustainability Index (CWSI) [Internet]; 2007.
Available from: https://publications.gc.ca/Collection/PH4-38-2007E.pdf [cited 2023 Oct 11].
42.
Jepson W. Measuring ‘no-win’ waterscapes: Experience-based scales and classification approaches to
assess household water security in colonias on the US–Mexico border. Geoforum. 2014; 51:107–120.
https://doi.org/10.1016/j.geoforum.2013.10.002
43. Aboelnga HT, Ribbe L, Frechen Fb, Saghir J. Urban Water Security: Definition and Assessment Frame-
work. Resources. 2019; 8(4):178. https://doi.org/10.3390/resources8040178
44. Padowski JC, Carrera L, Jawitz JW. Overcoming Urban Water Insecurity with Infrastructure and Institutions.
Water Resources Management. 2016; 30(13):4913–4926. https://doi.org/10.1007/s11269-016-1461-0
45. Yang F, Shao D, Xiao C, Tan X. Assessment of urban water security based on catastrophe theory. Water
Science and Technology. 2012; 66(3):487–493. https://doi.org/10.2166/wst.2012.182 PMID: 22744677
46. Marttunen M, Mustajoki J, Sojamo S, Ahopelto L, Keskinen M. A Framework for Assessing Water Secu-
rity and the Water–Energy–Food Nexus—The Case of Finland. Sustainability. 2019; 11(10):2900.
https://doi.org/10.3390/su11102900
47. Babel M, Shinde VR. A framework for water security assessment at basin scale. APN Science Bulletin.
2018; 8(1):27–32. https://doi.org/10.30852/sb.2018.342
48. Essex B, Koop SHA, Van Leeuwen CJ. Proposal for a National Blueprint Framework to Monitor Prog-
ress on Water-Related Sustainable Development Goals in Europe. Environmental Management. 2020;
65(1):1–18. https://doi.org/10.1007/s00267-019-01231-1 PMID: 31797037
49. Khan S, Guan Y, Khan F, Khan Z. A Comprehensive Index for Measuring Water Security in an Urbaniz-
ing World: The Case of Pakistan’s Capital. Water. 2020; 12(1):166. https://doi.org/10.3390/w12010166
50. Yin S, Dongjie G, Weici S, Weijun G. Integrated assessment and scenarios simulation of urban water
security system in the southwest of China with system dynamics analysis. Water Science and Technol-
ogy. 2017; 76(9):2255–2267. https://doi.org/10.2166/wst.2017.333 PMID: 29144284
51. Krause M. AquaRating: An international standard for assessing water and wastewater services. Water
Intelligence Online. 2015; 14. https://doi.org/10.2166/9781780407401
52.
Liu H, Jia Y, Niu C, Gan Y, Xu F. Evaluation of regional water security in China based on dualistic water
cycle theory. Water Policy. 2018; 20(3):510–529. https://doi.org/10.2166/wp.2017.062
53. Xiao Sc, Li Jx, Xiao Hl, Liu Fm. Comprehensive assessment of water security for inland watersheds in
the Hexi Corridor, Northwest China. Environmental Geology. 2008; 55(2):369–376. https://doi.org/10.
1007/s00254-007-0982-5
54. Milman A, Short A. Incorporating resilience into sustainability indicators: An example for the urban
water sector. Global Environmental Change. 2008; 18(4):758–767. https://doi.org/10.1016/j.gloenvcha.
2008.08.002
55. ARUP. The City Water Resilience Approach [Internet]; 2019. Available from: https://www.arup.com/
perspectives/publications/research/section/the-city-water-resilience-approach [cited 2023 Oct 11].
56. Shao D, Yang F, Xiao C, Tan X. Evaluation of water security: an integrated approach applied in Wuhan
urban agglomeration, China. Water Science and Technology. 2012; 66(1):79–87. https://doi.org/10.
2166/wst.2012.147 PMID: 22678203
57. Dou M, Shi Y, Li G. Optimized urban water security regulation schemes driven by industrial develop-
ment pattern. Water Policy. 2019; 21(3):676–691. https://doi.org/10.2166/wp.2019.198
58. Yomo M, Mourad KA, Gnazou MDT. Examining Water Security in the Challenging Environment in
Togo, West Africa. Water. 2019; 11(2):231. https://doi.org/10.3390/w11020231
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
23 / 25
PLOS WATER Assessing inequalities in urban water security
59. ARUP|The Rockefeller Foundation. City Resilience Index [Internet]; 2018. Available from: https://www.
arup.com/perspectives/publications/research/section/city-resilience-index [cited 2023 Oct 11].
60.
Iribarnegaray MA, Copa FR, Gatto D’Andrea ML, Arredondo MF, Cabral JD, Correa JJ, et al. A compre-
hensive index to assess the sustainability of water and sanitation management systems. Journal of
Water, Sanitation and Hygiene for Development. 2012; 2(3):205–222. https://doi.org/10.2166/washdev.
2012.005
61. OECD. OECD Water Governance Indicator Framework. OECD Studies on Water. Paris: OECD Pub-
lishing; 2018. p. 49–105.
62. Chen L, Shi J. Analysis and predication of urban water security: a case study of Chengdu City, China.
IOP Conference Series: Earth and Environmental Science. 2016;39(1):012027. https://doi.org/10.1088/
1755-1315/39/1/012027
63. Su Y, Gao W, Guan D. Integrated assessment and scenarios simulation of water security system in
Japan. Science of The Total Environment. 2019; 671:1269–1281. https://doi.org/10.1016/j.scitotenv.
2019.03.373
64.
65.
IBGE—Instituto Brasileiro de Geografia e Estatistica [Brazilian Institute of Geography and Statistics].
IBGE—Cidades e Estados [IBGE—Cities and States] [Internet];. Available from: https://www.ibge.gov.
br/cidades-e-estados/sp/campinas.html [cited 2023 Oct 11].
Tramontin V, Feltran-Barbieri R, Barbosa L, Oliveira M, Matsumoto MM, Caccia L, et al. Natural infra-
structure in Campinas’ water system, São Paulo State [Internet]. WRI Brasil and ICLEI; 2022. Available
from: https://iclei.org/wp-content/uploads/2022/09/natural-infrastructure-english-komprimiert.pdf [cited
2023 Oct 11].
66. Prefeitura Municipal de Campinas, Secretaria Municipal do Verde e do Desenvolvimento Sustenta´vel
[Campinas City Hall, Municipal Department of Green and Sustainable Development]. Plano Municipal
de Saneamento Ba´sico—Produto 1: Diagno´stico, Caracterizac¸ão e Ana´ lise Crı´tica [Municipal Plan of
Basic Sanitation—Deliverable 1: Diagnostics, Characterization and Critical Analysis] [Internet]; 2013.
Available from: https://www.campinas.sp.gov.br/arquivos/meio-ambiente/plano-saneamento/p1-
diagnostico.pdf [cited 2023 Oct 11].
67.
68.
IBGE—Instituto Brasileiro de Geografia e Estatistica [Brazilian Institute of Geography and Statistics].
Malha Municipal | IBGE—Brasil | Unidades da Federac¸ão [Municipal Network | IBGE—Brazil | Federa-
tion Units] [Base layer];. Available from: https://geoftp.ibge.gov.br/organizacao_do_territorio/malhas_
territoriais/malhas_municipais/municipio_2022/Brasil/BR/BR_UF_2022.zip [cited 2023 Oct 11].
IBGE—Instituto Brasileiro de Geografia e Estatistica [Brazilian Institute of Geography and Statistics].
Malha Municipal | IBGE—São Paulo | Municı´pios [Municipal Network | IBGE—São Paulo | Municipali-
ties] [Base layer];. Available from: https://geoftp.ibge.gov.br/organizacao_do_territorio/malhas_
territoriais/malhas_municipais/municipio_2022/UFs/SP/SP_Municipios_2022.zip [cited 2023 Oct 11].
69. Prefeitura de Campinas, Diretoria de Informac¸ão, Documentac¸ão e Cadastro (DIDC) [Campinas City
Hall, Directorate of Information, Documentation and Registration (DIDR)]. Metadados Geoespaciais—
PD2018 Unidades Territoriais Ba´sicas (UTB) e Unidades Territoriais Rurais (UTR) [Geospatial Meta-
data—PD2018 Basic Territorial Units (BTU) and Rural Territorial Units (RTU)] [Base layer];. Available
from: https://informacao-didc.campinas.sp.gov.br/download_shp/pd2018_utbs.zip [cited 2023 Oct 11].
70. Prefeitura Municipal de Campinas [Campinas City Hall]. Plano Diretor Estrate´gico—Prefeitura de Cam-
pinas Caderno de Subsı´dios [Strategic Master Plan—Campinas City Hall, Subsidies Booklet] [Internet];
2017. Available from: https://planodiretor.campinas.sp.gov.br/sites/planodiretor.campinas.sp.gov.br/
files/20170412_caderno_site.pdf [cited 2023 Oct 11].
71. Prefeitura Municipal de Campinas [Campinas City Hall]. Plano Diretor e da Polı´tica de Desenvolvimento
do Municı´pio [Master Plan and Development Policy of the Municipality]; 2006. Available from: https://
saude.campinas.sp.gov.br/seplan/publicacoes/planodiretor2006/pd2006vfinal.htm [cited 2023 Oct 11].
72. SANASA—Sociedade de Abastecimento de A´ gua e Saneamento [SANASA—Water Supply and Sani-
tation Society]. Relato´rio de Sustentabilidade 2021 [Sustainability Report 2021] [Internet]; 2021. Avail-
able from: https://www.sanasa.com.br/conteudo/conteudo1.aspx?f=S&flag=-PTSR [cited 2023 Oct 11].
73.
Theil H. Economics and information theory. Amsterdam: North-Holland Publishing Company; 1967.
74. OECD. OECD Regions and Cities at a Glance 2018 [Internet]; 2018. Available from: https://www.oecd-
ilibrary.org/content/publication/reg_cit_glance-2018-en [cited 2023 Oct 11].
75. Sinha A. Inequality of Carbon Intensities Across OECD Countries. Energy Procedia. 2015; 75:2529–
2533. https://doi.org/10.1016/j.egypro.2015.07.275
76. Sinha A, Rastogi SK, et al. Inequality in access to improved water source: a regional analysis by Theil
Index. Theoretical Economics Letters. 2015; 5(06):683. https://doi.org/10.4236/tel.2015.56079
77. Moran PAP. The Interpretation of Statistical Maps. Journal of the Royal Statistical Society Series B
(Methodological). 1948; 10(2):243–251. https://doi.org/10.1111/j.2517-6161.1948.tb00012.x
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
24 / 25
PLOS WATER Assessing inequalities in urban water security
78. Rey SJ, Arribas-Bel D, Wolf LJ. Geographic Data Science with Python [Internet]; 2020. Available from:
https://geographicdata.science/book/intro.html [cited 2023 Oct 11].
79. Heaviside C, Macintyre H, Vardoulakis S. The Urban Heat Island: Implications for Health in a Changing
Environment. Current Environmental Health Reports. 2017; 4(3):296–305. https://doi.org/10.1007/
s40572-017-0150-3 PMID: 28695487
80. Agência das Bacias PCJ [PCJ Basins Agency]. Relato´ rio De Situac¸ão Dos Recursos Hı´dricos 2020,
Ugrhi 05—Bacias Hidrogra´ficas Dos Rios Piracicaba, Capivari e Jundiaı´ [Water Resources Situation
Report 2020, Ugrhi 05—Piracicaba, Capivari and Jundiaı´ River Basins] [Internet]; 2020. Available from:
https://agencia.baciaspcj.org.br/wp-content/uploads/Relat%C3%B3rio_situa%C3%A7%C3%A3o-
2020-2019.pdf [cited 2023 Oct 11].
Instituto Cidades Sustenta´veis [Sustainable Cities Institute]. I´ndice de Desenvolvimento Sustenta´ vel
das Cidades—Brasil [City Sustainable Development Index—Brazil] [Internet];. Available from: https://
idsc.cidadessustentaveis.org.br/ [cited 2023 Oct 11].
81.
82.
Johansen IC, Carmo RL, Alves LC. Desigualdade social intraurbana: implicac¸ões sobre a epidemia de
dengue em Campinas, SP, em 2014 [Intraurban social inequality: Implications on the dengue fever
epidemy in Campinas, SP, in 2014]. Cadernos Metro´ pole. 2016; 18:421–440.
83. Prefeitura Municipal de Campinas, Secretaria Municipal do Verde, Meio Ambiente e Desenvolvimento
Sustentavel [Campinas City Hall, Municipal Department of Green, Environment and Sustainable Devel-
opment]. Programa Cidades Sustentaveis—Relato´rio 2013/2020 [Sustainable Cities Program—2013/
2020 Report]. Campinas; 2020. Available from: https://www.campinas.sp.gov.br/arquivos/meio-
ambiente/cics/relatorio-programa-cidades-sustentaveis_v2.pdf [cited 2023 Oct 11].
84. Universidade Federal de Santa Catarina [Santa Catarian Federal University]. Atlas brasileiro de desas-
tres naturais: 1991 a 2012. Volume São Paulo [Brazilian atlas of natural disasters: 1991 to 2012. São
Paulo Volume] [Internet]. Floriano´ polis: CEPED UFSC; 2013. Available from: https://s2id.mi.gov.br/
paginas/atlas/ [cited 2023 Oct 11].
85. United Nations. Sustainable Development Goals—Welcome to the Sustainable Development Goal indi-
cators website [Internet];. Available from: https://unstats.un.org/sdgs/ [cited 2023 Oct 11].
86. Prefeitura Municipal de Campinas, Secretaria Municipal Do Verde, Meio Ambiente E Desenvolvimento
Sustenta´vel [Campinas City Hall, Municipal Department of Green, Environment and Sustainable Devel-
opment]. Plano Municipal Do Verde—Diagno´ stico [Green Municipal Plan—Diagnostic] [Internet]. Cam-
pinas; 2015. Available from: https://www.campinas.sp.gov.br/arquivos/meio-ambiente/diagnostico_
final_atualizado_22_12.pdf [cited 2023 Oct 11].
87. Prefeitura Municipal de Campinas, Secretaria Municipal de Saude, Coordenadoria de Vigilaˆ ncia em
Sau´ de [Campinas City Hall, Municipal Health Department, Health Surveillance Coordination]. Informe
Dengue: Situac¸ão atual e risco do pro´ ximo verão [Dengue fever brief: Current situation and risk of next
summer] [Internet]. Campinas; 2010. Available from: https://saude.campinas.sp.gov.br/vigilancia/
informes/InformeDengueOutubro2010.pdf [cited 2023 Oct 11].
88. Prefeitura de Campinas [Campinas City Hall]. Plano Municipal De Contingência Para o Enfrentamento
Das Arboviroses Urbanas [Municipal Contingency Plan for Combating Urban Arbovirosis] [Internet].
Campinas; 2021. Available from: https://dengue.campinas.sp.gov.br/sites/dengue.campinas.sp.gov.br/
files/PLANO%20MUNICIPAL%20DE%20CONTING%C3%8ANCIA%20PARA%20O%
20ENFRENTAMENTO%20DAS%20ARBOVIROSES%20URBANAS%20-%20Ano%202021-2022_%
20final.pdf [cited 2023 Oct 11].
89. Sistema Nacional de Informac¸ões sobre Saneamento—SNIS [National Sanitation Information System].
Se´ rie Histo´rica [Historic Series] [Internet]; 2022. Available from: http://app4.mdr.gov.br/serieHistorica/
[cited 2023 Oct 11].
90. Prefeitura Municipal de Campinas [Campinas City Hall]. Campinas lanc¸a a campanha Fevereiro Violeta
contra o analfabetismo [Campinas launches the Violet February campaign against illiteracy] [Internet];
2019. Available from: https://portal.campinas.sp.gov.br/noticia/35750 [cited 2023 Oct 11].
91. Prefeitura Municipal de Campinas [Campinas City Hall]. Plano Municipal de Recursos Hı´dricos—Vol-
ume 1: Panorama e estado dos recursos hidricos [Municipal Plan for Water Resources—Volume 1:
Overview and status of water resources] [Internet]; 2016. Available from: https://www.campinas.sp.gov.
br/arquivos/meio-ambiente/vol-1-diagnostico.pdf [cited 2023 Oct 11].
92. Cisotto MF. Natureza e Cidade: relac¸ões entre os fragmentos florestais e a urbanizac¸ão em Campinas
(SP) [Nature and City: relations between forest fragments and urbanization in Campinas (SP)] [Master’s
Thesis]. Universidade Estadual de Campinas; 2009.
93. Dieperink C, Koop SHA, Witjes M, Van Leeuwen K, Driessen PPJ. City-to-city learning to enhance
urban water management: The contribution of the City Blueprint Approach. Cities. 2023; 135:104216.
https://doi.org/10.1016/j.cities.2023.104216
PLOS Water | https://doi.org/10.1371/journal.pwat.0000213 February 1, 2024
25 / 25
PLOS WATER
| null |
10.1371_journal.pone.0283795.pdf
|
Data Availability Statement: Data and codes are
available at: DOI:10.17605/OSF.IO/PS4RM.
|
Data and codes are available at: DOI:10.17605/OSF.IO/PS4RM.
|
RESEARCH ARTICLE
The Polish adaptation of the measurements
of rule-governed behaviors: Generalized
Pliance Questionnaire, Generalized Tracking
Questionnaire and Generalized Self-Pliance
Questionnaire
Joanna DudekID
1☯*, Maria Cyniak-Cieciura2☯, Paweł OstaszewskiID
3☯
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 Faculty of Psychology in Warsaw, Center for Behavioral Research in Decision Making, SWPS University,
Warsaw, Poland, 2 Advanced Clinical Studies and Therapy Excellence Center, Institute of Psychology,
SWPS University, Warsaw, Poland, 3 Center for Behavioral Research in Decision Making, Institute of
Psychology, SWPS University, Warsaw, Poland
☯ These authors contributed equally to this work.
* [email protected]
OPEN ACCESS
Abstract
Citation: Dudek J, Cyniak-Cieciura M, Ostaszewski
P (2023) The Polish adaptation of the
measurements of rule-governed behaviors:
Generalized Pliance Questionnaire, Generalized
Tracking Questionnaire and Generalized Self-
Pliance Questionnaire. PLoS ONE 18(4): e0283795.
https://doi.org/10.1371/journal.pone.0283795
Editor: Marco Innamorati, Universita degli Studi
Europea di Roma, ITALY
Received: November 4, 2022
Accepted: March 19, 2023
Published: April 5, 2023
Copyright: © 2023 Dudek et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Data and codes are
available at: DOI:10.17605/OSF.IO/PS4RM.
Funding: The research was supported financially
by the Faculty of Psychology in Warsaw, SWPS
University. The funders had no role in study
design, data collection and analysis, decision to
publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
In some circumstances rule-governed behavior, a behavior that is governed by verbal rules
instead of environmental consequences, may be beneficial for human beings. At the same
time, rigid rule following is associated with psychopathology. Thus measurement of rule-
governed behavior may be of special use in a clinical setting. The aim of this paper is to
assess the psychometric properties of Polish adaptations of three questionnaires measuring
generalized tendency to engage in various types of rule-governed behaviors: Generalized
Pliance Questionnaire (GPQ), Generalized Self-Pliance Questionnaire (GSPQ), General-
ized Tracking Questionnaire (GTQ). A forward-backward method was used for translation.
Data was collected from two samples: general population (N = 669) and university students
(N = 451). To measure the validity of the adapted scales the participants filled in a set of
self-assessed questionnaires: Satisfaction with Life Scale (SWLS), Depression, Anxiety,
and Stress Scale– 21 (DASS-21), General Self-Efficacy Scale (GSES), Acceptance and
Action Questionnaire–II (AAQ-II), Cognitive Fusion Questionnaire (CFQ), Valuing Question-
naire (VQ) and Rumination—Reflection Questionnaire (RRQ). The exploratory and confir-
matory analyses confirmed the unidimensional structure of each of the adapted scales. All
of those scales presented good reliability (internal consistency measured with Cronbach
Alpha) and item-total correlations. The Polish versions of questionnaires presented signifi-
cant correlations in the expected directions with relevant psychological variables in line with
the original studies. The measurement occurred invariant across both samples as well as
gender. The results provide evidence that Polish versions of GPQ, GSPQ and GTQ present
sufficient validity and reliability to be used in the Polish-speaking population.
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
1 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
Introduction
To provide a behavior-analytic account of complex human behavior such as thinking and
problem solving, in 1966 Skinner [1] introduced the term rule-governed behavior (RGB),
defined as behavior that is under the control of rules and instructions, in contrast to contin-
gency-shaped behaviors which are under control of direct contingencies in the environment.
The functional behavioral analysis of the concept was first proposed almost 20 years later by
Zettle and Hayes [2] and elaborated in detail within the framework of the Relational Frame
Theory [3].
Beginning with Skinner, researchers emphasize that the ability to generate and follow verbal
rules may be beneficial especially in contexts where learning through direct experience is dan-
gerous (e.g. look both ways before crossing the street) or contingencies are delayed (e.g. attend
classes and study to get the diploma). Thus, RGB helps people to achieve goals, learn from the
experience of others, and cope with events before they occur [4]. However, under certain cir-
cumstances, RGB can also produce undesired consequences, such as insensitivity to real envi-
ronmental contingencies, rigidly following verbal rules despite their effectiveness or even
harmful outcome of rule following, or persistent avoidance [5–7].
Therefore, the concept of RGB has become particularly important in the domain of clinical
behavior analysis, as it provides both explanation of the development of a number of psycho-
pathological symptoms [5–8], and helps to develop psychotherapeutic interventions, such as
Acceptance and Commitment Therapy (ACT) with its focus on the psychological flexibility
model [9]. To support both basic and clinical research on the RGB, reliable and valid methods
of assessing the behaviors are required. The aim of the present paper is to present the valida-
tion process of the three recently developed questionnaires measuring two functional types of
RGB, generalized pliance (and self-pliance) and generalized tracking [10–12] in a Polish
population.
Different functional types of RGB were first introduced by Zettle and Hayes [2], who distin-
guished pliance and tracking as two most fundamental functional classes of RGB, and a third
type, augmenting, operating together with the two former classes by verbally changing the
reinforcing or punishing strength of consequences included in the rules.
Pliance is defined as a functional class of rule-governed behavior under the control of history
of multiple interactions in which the speaker provides the listener with the reinforcement contin-
gent on the correspondence between the rule (e.g. do not touch hot pot) and the relevant behav-
ior (refraining from touching the pot). An example of reinforcement in such circumstances may
be praising the individual (e.g. great that you did not touch the pot [2, 3, 13]). Taking into
account that the listener and the speaker may be the same person [14], under some circum-
stances those rules can also be generated by the individual and then called self-rules [10, 11].
Pliance, being the first type of rule-following developed [13], over-generalizes at some point
in the child’s development. Yet, social interactions lead to contextualizing pliance (so that the
child can distinguish when it is appropriate) and establishing tracking to help her to recognize
natural consequences of her behavior [5, 7].
Lack of such learning history may lead to generalized pliance [11]. Generalized pliance can
be problematic when it becomes the main source of impact on human behavior, as socially
mediated consequences are less predictable and controllable than other types of consequences
which may lead to lower contact with sources of positive reinforcement. Individuals displaying
generalized pliance may be particularly insensitive to direct contingencies (e.g. a person
believes that in order to obtain social approval she needs to start smoking, so she does that
ignoring the negative consequences of smoking). As the child develops and gains fluency in
relational framing, plys become more abstract (e.g. a person believes she needs to be a ‘good’
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
2 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
in order to be loved [15]) which may increase the likelihood social approval to be the main
source of reinforcement.
As many of social rules treat aversive private events as events that need to be controlled or
avoided, individuals displaying generalized pliance are more likely to engage in rigid patterns of
behaviors and a tendency to engage in experiential avoidance—attempts to avoid, control or get
rid of unwanted internal experiences even when doing so is harmful for the individual [5, 16].
Considering that generalized pliance may lead to losing the contact with the sources of posi-
tive reinforcement and behaviors being controlled by negative reinforcement (avoidance) it is
considered a risk factor for developing psychopathology (e.g. [5, 6, 8] and psychological inflex-
ibility–difficulties in engaging in meaningful actions due to the presence of unpleasant internal
experiences that the person wants to avoid, get rid of or control [4].
In child development, pliance is seen as a condition to develop tracking [13]. In contrast to
pliance, tracking is sensitive to the direct environmental contingencies, so that if they change
the individual is changing the behavior accordingly [2] and it may be regarded as a flexible
rule-governed behavior [12]. Tracking is a functional class of behaviors under the control of
the history of multiple exemplars in which following the rule leads to natural consequences
derived from the way the world is arranged (e.g. following the rule “when it is cold, wear a
warm coat” leads to feeling warm even when it is cold outside [2, 3, 13]). Generalized tracking
is a pattern of behaviors that may be developed when an individual has been exposed to multi-
ple interactions in which she has been encouraged to observe and describe functional relation-
ships among events, e.g. recognize natural consequences of her behavior. The individual
engaged in tracking, behaving both as speaker and listener, learns to establish functional rela-
tionships among events and adjust her behavior accordingly [12]. Despite the interest in rule-
governed behaviors especially in the area of contextual behavioral science, there are a number
of limitations to experimental research investigating pliance and tracking [17]. One of the pro-
posed explanations involves noting that both pliance and tracking are listener-oriented con-
cepts [18] and therefore they cannot be produced by the speakers as the participants’ personal
learning history may influence their performance more than the experimental rules [11, 12].
Thus, researchers proposed alternative strategy of measuring pliance and tracking by devel-
oping self-report measures, which explore the perspective of the listener and investigate the
individual’s learning history and personal experience with formulating and following rules
[11, 12]. A measure of generalized pliance, Generalized Pliance Questionnaire (GPQ, [11]),
self-pliance, Generalized Self-Pliance Questionnaire (GSPQ, [10]), and a measure of general-
ized tracking, Generalized Tracking Questionnaire (GTQ, [12]), were created.
Empirical evidence with the GPQ shows that generalized pliance is connected to various
aspects of psychological inflexibility and measures of distress [11]. Higher scores in the GPQ
were predictive of lower levels of mindfulness and sensitivity to changing contingencies [19].
Generalized pliance was positively correlated with repetitive negative thinking, dysfunctional
attitudes, difficulties in valued living, and negatively correlated with life satisfaction [11].
Generalized tracking measured with GTQ was negatively correlated with generalized
pliance, experiential avoidance, tendency to ruminate and emotional symptoms and positively
correlated with valued living, life satisfaction and general self-efficacy and a wide range of
executive functions [12].
Measuring self-reported patterns of rule-governed behaviors, such as the GPQ and GTQ,
may lead to development of research on complex human behavior by broadening the knowl-
edge on rule-governed behaviors, its impact and development (e.g. differences across age, gen-
der, and cultures). These instruments may also be used to explain variability of results in
experimental analyses, predict development of psychopathology and behavioral rigidity, and
eventually, to analyze mediators and moderators of psychological interventions (e.g.
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
3 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
Acceptance and Commitment Therapy). Thus, it seems important to provide researchers and
practitioners in the area of clinical behavior analysis and contextual behavioral science with
appropriate questionnaires in countries and cultures in which people speak other languages
than English or Spanish, such as Polish. The strategy of adopting already existing question-
naires with good psychometric qualities seems a good step, because it allows multilingual and
multicultural comparisons of the same phenomena. We hope that the adaptation of GPQ.
GSPQ, and GTQ into Polish language will aid in the development of contextual behavioral sci-
ence in general, allow for multicultural analysis of the RGB, and most of all, provide a growing
number of researchers and practitioners in Poland with valid questionnaires measuring RGB,
widening the scope of both basic and applied research conducted.
Materials and methods
Scales’ adaptation
The original items of the Generalized Pliance Questionnaire (GPQ18–18 items, and its shorter
version—GPQ9), Generalized Tracking Questionnaire (GTQ– 11 items), and Generalized
Self-Pliance Questionnaire (GSPQ– 12 items) were translated to Polish by three independent
translators, who were practitioners in third wave cognitive-behavioral approaches and/or
experts in contextual behavioral science. Then, three independent versions of the Polish items
were presented to three independent judges–scientists and practitioners in the area of contex-
tual and behavioral science as well as third wave cognitive-behavioral therapy. Final versions
of items’ translations were chosen based on the judges’ opinions, back-translated to English
and accepted by the author of the original questionnaires. The original instructions and
response scales were kept (see [10–12], thus each item was rated on a scale from 1 to 7 (with 7
= always true and 1 = never true). These versions of the scales were used in the study aiming at
verifying their psychometric properties.
Participants
The psychometric properties of Polish versions of GPQ18, GPQ9, GTQ, and GSPQ were
checked in two independent studies and different samples described below.
Sample A–students’ sample
The participants were recruited via university SONA research panel. Students received credits
for taking part in the study in accordance with the university policy. The study was closed after
the data from at least 400 participants was collected. The only exclusion criteria were being less
than 18 years old. A total of 451 people completed the study: 371 women (82.3%) and 80 men
(17.7%) in the age of 18–64 (M = 27.61, SD = 8.27). Most of them had a secondary educational
level (N = 245, 54.3%), 206 (40.7%) declared higher education. Most of the participants were
living in the city with more than 500 000 residents (N = 252, 55.9%) and the least—in the rural
area (N = 48, 10.6%). A total of 87 people (19.3%) were living in a city with less than 100 000
residents and 64 people (14.2%) in a city with 100–500 000 residents.
Sample B–a general sample
The participants were recruited by the Pollster research panel (https://pollster.pl/), the only
exclusion criteria were being less than 18 years old. The data collection was stopped when data
from at least 600 participants was collected. A total of 669 people completed the study: 333
women (49.8%) and 336 men (50.2%) in the age of 18–65 (M = 40.96, SD = 13.49). Most of
them had a secondary educational level (N = 360, 53.8%), 272 (40.7%) declared higher
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
4 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
education and 37 (5.5%) finished only primary school. Most of the participants were living in
the city with less than 100 000 residents (N = 254, 38%), 25.7% (N = 172)—in the city with
100–500 000 residents, 18.7% (N = 125) in the rural area, and 17.6% (N = 118) in a city with
more than 500 000 residents.
Procedure
The study was conducted online between March—June 2021. The participants were asked to
fill in a set of self-report questionnaires: a short demographic questionnaire, the Polish ver-
sions of GPQ18, GTQ and GSPQ and a few other measures aiming at verifying GPQ18,
GPQ9, GTQ and GSPQ’s validity (their short description is presented below). All of the partic-
ipants signed an informed consent. The study was conducted following the Declaration of Hel-
sinki and received a positive opinion from the local Ethics Committee (Nr 5/2021).
Other measures
Satisfaction with Life Scale. The Polish version of the Satisfaction with Life Scale (SWLS
[20, 21]) was used to measure self-perceived well-being. It consists of five items. Participants
rated each item on a 7-point scale, ranging from 7 = strongly agree to 1 = strongly disagree.
Higher scores indicate a greater level of life satisfaction. Medium to large negative correlations
were expected between SWLS and GPQ18, GPQ9, GSPQ, and positive—with GTQ.
Depression, Anxiety, and Stress Scale– 21. The Polish version of Depression, Anxiety,
and Stress Scale– 21 (DASS-21 [22, 23]) was used to measure the level of depression, anxiety
and stress symptoms. It consists of 21 items and a 4-point Likert-type scale (3 = applied to me
very much, or most of the time; 0 = did not apply to me at all). It contains three subscales:
Depression, Anxiety, and Stress with higher scores indicating higher levels of symptoms.
Medium to strong positive correlations were expected between DASS21 and GPQ18, GPQ9,
and GSPQ, as well as negative with GTQ.
General Self-Efficacy Scale. The Polish version of General Self-Efficacy Scale (GSES [24,
25]) was used to measure self-perceived self-efficacy. It comprises ten items assessed on a
4-point Likert scale (1 = not at all true, 4 = exactly true), which enable to calculate a general
score (the higher the score, the higher the level of general self-efficacy). Medium negative cor-
relations were expected between GSES and GPQ18, GPQ9, GSPQ, and positive with GTQ.
Acceptance and Action Questionnaire–II. The Polish version of the Acceptance and
Action Questionnaire-II (AAQ-II [26, 27]) was used to measure psychological inflexibility. It
consists of seven statements. Participants rate each statement on a 7-point scale, ranging from
1 = never true to 7 = always true. Higher scores indicate higher psychological inflexibility.
Medium positive correlations were expected between AAQ-II and GPQ18, GPQ9, GSPQ, and
negative with GTQ.
Cognitive Fusion Questionnaire. The Polish version of Cognitive Fusion Questionnaire
(CFQ [28, 29]) was used to measure the level of cognitive fusion. The CFQ consists of seven
items with a 7-point Likert-type scale (7 = always true; 1 = never true). Medium to strong posi-
tive correlations were expected between the CFQ and GPQ18, GPQ9, GSPQ, and negative–
with GTQ.
Valuing Questionnaire. The Polish version of Valuing Questionnaire (VQ [28, 30]) was
used to measure the general valued living during the past week. VQ consists of ten items and a
6-point Likert scale (6 = completely true; 0 = not at all true). It contains two subscales: Progress
(defined as enactment of values, including clear awareness of what is personally important,
and perseverance), as well as Obstruction (defined as disruption of valued living due to avoid-
ance of unwanted experience and distraction from values). It was expected that the VQ
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
5 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
Progress scale will be negatively correlated with GPQ18, GPQ9, GSPQ, and positively with
GTQ. Positive correlations were expected between VQ Obstruction scale and GPQ18, GPQ9,
GSPQ, and negative with GTQ.
Rumination—Reflection Questionnaire. The Polish version of Rumination—Reflection
Questionnaire– 12 (RRQ12 [31, 32]) was used to measure the level of focus on one’s own expe-
riences (rumination) motivated by fear, and the involvement in getting to know oneself (reflec-
tion) motivated by curiosity. RRQ12 consists of twelve items assessed with a 5-point Likert
scale (1 –I strongly disagree, 5 –I strongly agree) and contains two subscales–Rumination and
Reflection. It was expected that RRQ Rumination scale would be positively correlated with
GPQ18, GPQ9, GSPQ, and negatively with GTQ. Negative correlations were expected between
RRQ Reflection scale and GPQ18, GPQ9, GSPQ, and positive with GTQ.
The reliability coefficients (Cronbach Alphas) of the scales were satisfactory and are pre-
sented in Table 3.
Statistical analyses
There were no missing values within GPQ18, GTQ and GSPQ items in both samples. A cross-
validation procedure was applied with the analyses done on a data from a students’ sample A
and then replicated in a general sample B. Exploratory factor analysis (EFA) was conducted
only on sample A and confirmatory factor analysis (CFA) only on sample B. The measurement
invariance across samples (A and B) and gender was checked. Then, the corrected item-total
correlations and Cronbach Alpha coefficients were calculated in two samples respectively.
Finally, to provide information about the validity of the scales, r-Pearson correlations between
GPQ18, GPQ9, GTQ, GSPQ and other measures were calculated (in two samples respectively).
All the analyses were conducted with the use of SPSS v. 25, FACTOR v.11.05.01 [33] and R
lavaan package [34].
Results
Exploratory factor analysis
The EFA was conducted with the use of FACTOR v. 11.05.01 based on the data from sample A.
Data included in the analyses was categorical and according to Mardia’s test it did not meet
the assumptions of the multivariate normal distribution, due to the exceeded kurtosis values
(The results of Mardia’s test for the GPQ18 items: b = 31.68, Z(1140) = 23.81, p = 1.00 for
skewness and b = 445.22, Z = 33.72, p < .001 for kurtosis. For the GPQ9 items: b = 5.10, Z
(165) = 23.81, p = 1.00 for skewness and b = 118.02, Z = 14.35, p < .001 for kurtosis. For the
GTQ items: b = 13.48, Z(286) = 1013.46, p = 1.00 for skewness and b = 202.88, Z = 37.59, p <
.001 for kurtosis. For the GSPQ items: b = 10.90, Z(364) = 818.99.46, p = 1.00 for skewness and
b = 208.94, Z = 23.71, p < .001 for kurtosis). Data was analyzed with the use of robust diago-
nally weighted least squares (RDWLS) extraction method with polychoric correlations and
robust Promin rotation [33]. The number of dimensions was determined by means of the opti-
mal implementation of parallel analysis (PA [35]). The Unidimensional Congruence (UniCo),
Explained Common Variance (ECV), and Mean of Item Residual Absolute Loadings (MIR-
EAL) indexes were used to assess the unidimensionality. Values larger than .95 and .85 in
UniCo and ECV, respectively, as well as a value lower than .30 for the MIREAL suggest that
data can be treated as essentially unidimensional [36]. In all the analyses the 95% bootstrap
confidence intervals were estimated based on 500 samples.
GPQ18. The Bartlett’s statistic was statistically significant (5102.1(153), p < .001), and the
result of the Kaiser-Meyer-Olkin (KMO) test was satisfactory (.93, 95%CI [.90, .93]). The PA
suggested extracting one factor accounting for 55.08% of variance (eigenvalue = 9.10). Table 1
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
6 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
shows that factor loadings were high for all the items: from .50 (item 2) to .86 (item 13). The
UniCo, ECV and MIREAL values suggest that the data of the GPQ-18 can be treated as unidi-
mensional (UniCo = .97 (95% CI [.96, .99], ECV = .90, (95%CI [.86, .90], MIREAL = .22, 95%
CI [.19, .24]).
GPQ9. The Bartlett’s statistic was statistically significant (2375.6(36), p < .001), and the
result of the Kaiser-Meyer-Olkin (KMO) test was satisfactory (.89, 95%CI [.85, .90]). The PA
suggested extracting one factor accounting for 59.96% of variance (eigenvalue = 4.99). Table 1
shows that factor loadings were high for all the items: from .49 (item 1) to .87 (item 4). The
UniCo, ECV and MIREAL values suggest that the data of the GPQ-9 can be treated as unidi-
mensional (UniCo = .95 (95% CI [.92, .98], ECV = .84, (95%CI [.81, .87], MIREAL = .28, 95%
CI [.24, .32]).
GTQ. The Bartlett’s statistic was statistically significant (2742.6(55), p < .001), and the
result of the Kaiser-Meyer-Olkin (KMO) test was satisfactory (.92, 95%CI [.88, .92]). The PA
suggested extracting one factor accounting for 60.71% of variance (eigenvalue = 6.06). Table 1
shows that factor loadings were high for all the items: from .65 (item 10) to .79 (item 7). The
UniCo, ECV and MIREAL values suggest that the data of the GTQ can be treated as unidimen-
sional (UniCo = .98 (95% CI [.97, .99], ECV = .88, (95%CI [.86, .91], MIREAL = .22, 95% CI
[.17, .24]).
GSPQ. The Bartlett’s statistic was statistically significant (2762.5(66), p < .001), and the
result of the Kaiser-Meyer-Olkin (KMO) test was satisfactory (.91, 95%CI [.87, .92]). The PA
suggested extracting one factor accounting for 59.69% of variance (eigenvalue = 6.07) 58.74%
of variance (eigenvalue = 6.14). Table 1 shows that factor loadings were high for all the items:
from .56 (item 7) to .82 (item 5). The UniCo, ECV and MIREAL values suggest that the data of
the GSPQ can be treated as unidimensional (UniCo = .98 (95% CI [.98, .99], ECV = .89, (95%
CI [.87, .91], MIREAL = .19, 95% CI [.14, .20]).
Summarizing, the EFA results suggest that all the scales measure unidimensional latent con-
structs and the one-factor solutions explain a significant portion of variance in each case.
Confirmatory factor analysis
The CFA was conducted with the use of the R lavaan package to analyze the fit of the one-fac-
tor model of the GPQ18, GPQ9, GTQ and GSPQ in a general sample B. A weighted least
squares–mean (WLSM) estimation method with polychoric correlations was utilized. Good-
ness of fit was evaluated using the robust chi-square test, robust root-mean-square error of
approximation (RMSEA), robust comparative fit index (CFI), the robust Tucker-Lewis index
(TLI), and standardized root-mean-square residual (SRMR; the last one only in CFA). Hu and
Bentler [37] proposed the following criteria of good model fit: RMSEA�.10; SRMR�.08;
CFI�.90; TLI�.95, however recently the controversies to their application to categorical data
have been raised [38]. Shi and Maydeu-Olivares [39] showed that SRMR is less sensitive to the
choice of estimator, thus it is also reported. Standardized factor loading estimates are shown in
Figs 1–4.
GPQ18. The one-factor model exhibited a non-satisfactory fit due to the high value of
robust RMSEA: robust χ2(135) = 2729.54, p < .001; robust RMSEA = .108 (95% CI: [.105,
.112]), SRMR = .066, robust CFI = .983, robust TLI = .980. Therefore, we decided to analyse
modification indices (MI) with a minimum value = 10. MI are univariate score tests that reflect
the improvement of model fit after allowing some of the parameters to be free. After careful
examination of MI we decided to modify the model allowing the error terms between items 16
and 17 (MI = 97.52), 10 and 11 (MI = 95.12), 4 and 5 (MI = 91.93), as well as 1 and 2
(MI = 71.56) to correlate. The decision was based on the high MI values as well as semantic
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
7 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
Table 1. The Polish version of GPQ18, GPQ9, GTQ and GSPQ items’ factor loadings and corrected item-total
correlations.
Items
GPQ18
Factor 1
Corrected item-total
correlation
1. Mo´j nastro´j zależy od tego, co myślą o mnie moi przyjaciele.
2. Bardzo przejmuję się tym, co myślą o mnie moi przyjaciele.
3. Czuję, że moja praca nie jest warta wysiłku, jeżeli inni ludzie jej nie
doceniają.
4. To dla mnie bardzo ważne, aby czuć się akceptowanym przez innych.
5. Żeby czuć się szczęśliwym, potrzebuję być doceniany/a przez innych ludzi.
6. Moje poczucie własnej wartości zależy od tego, co inni ludzie o mnie myślą i
mo´wią.
7. Moim gło´wnym celem w życiu jest bycie rozpoznawanym i poważanym
przez otaczających mnie ludzi.
8. Duży wpływ na moje decyzje mają opinie innych oso´b.
9. Bardzo przejmuję się tym, aby prezentować idealny obraz siebie.
10. To co robię nic by nie znaczyło, gdyby inni nie mogli tego zobaczyć.
11. Ciężka praca ma wartość tylko wtedy, gdy inni ludzie ją dostrzegają.
12. Jest dla mnie ważne, aby inni ludzie mieli w głowie mo´j pozytywny obraz.
13. Potrzebuję aprobaty ze strony innych oso´b, aby czuć się dobrze ze sobą.
14. Nie mogę zawieść oczekiwań innych oso´b wobec mnie.
15. Przed podjęciem decyzji potrzebuję, aby inni ludzie rozumieli moje
powody.
16. Przy podejmowaniu decyzji bardziej doceniam rady innych niż własne
zdanie.
17. Przed zrobieniem czegoś ważnego proszę innych o poradę.
18. Obawa przed krytyką powstrzymuje mnie od robienia ro´żnych rzeczy.
GPQ9
1. Bardzo przejmuję się tym, co myślą o mnie moi przyjaciele.
2. To dla mnie bardzo ważne, aby czuć się akceptowanym przez innych.
3. Żeby czuć się szczęśliwym, potrzebuję być doceniany/a przez innych ludzi.
4. Moje poczucie własnej wartości zależy od tego, co inni ludzie o mnie myślą i
mo´wią.
5. Duży wpływ na moje decyzje mają opinie innych oso´b.
6. To co robię nic by nie znaczyło, gdyby inni nie mogli tego zobaczyć.
7. Ciężka praca ma wartość tylko wtedy, gdy inni ludzie ją dostrzegają.
8. Potrzebuję aprobaty ze strony innych oso´b, aby czuć się dobrze ze sobą.
9. Przy podejmowaniu decyzji bardziej doceniam rady innych niż własne
zdanie.
GTQ
1. Kiedy dostrzegam, że coś nie działa, pro´buję czegoś innego.
2. Lubię dowiadywać się, jak coś działa i wyciągać swoje własne wnioski.
3. Łatwo dostosowuję się do zmian.
4. Potrafię znaleźć nowe rozwiązania problemo´w.
5. Podejmuję decyzje w oparciu o swoje doświadczenie, a nie o to, co mo´wią
inni.
6. Lubię pro´bować ro´żnych podejść, aby zobaczyć kto´re jest lepsze.
7. Jestem dobry/a w znajdywaniu bardziej skutecznych sposobo´w
wykonywania zadań.
8. Kiedy zauważam, że coś nie działa, szybko zmieniam swo´j sposo´b
postępowania.
.681
.497
.615
.740
.782
.844
.695
.796
.607
.645
.627
.776
.857
.670
.700
.654
.588
.719
.493
.758
.812
.867
.745
.663
.652
.859
.578
.679
.690
.668
.778
.692
.723
.789
.745
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
.629
.444
.569
.661
.703
.764
.657
.757
.560
.586
.561
.727
.796
.631
.657
.591
.544
.674
.428
.659
.715
.758
.706
.582
.560
.779
.511
.614
.632
.596
.702
.634
.654
.711
.676
(Continued )
8 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
Table 1. (Continued)
Items
Factor 1
Corrected item-total
correlation
9. Z łatwością uczę się na konsekwencjach swoich działań.
10. Kiedy zdam sobie sprawę, że nie miałem/am racji, zmieniam swo´j sposo´b
myślenia i działania.
11. Podejmuję decyzje na bazie uzyskanych wcześniej wyniko´w.
GSPQ
1. Czuję, że tracę kontrolę nad życiem, jeżeli moje osobiste sprawy nie są w
idealnej ro´wnowadze.
2. Potrzebuję kontrolować swoje lęki, żeby nie czuć się słabym/ą.
3. Szukam odpowiedzi na wszystko, aby nie czuć się głupim/ą.
4. Denerwuje mnie, że nie mogę robić wszystkiego w ten sam sposo´b.
5. Czuję się pogubiony/a, jeżeli nie jestem w stanie wykonać zaplanowanych
działań.
6. Muszę dobrze traktować innych ludzi, aby nie czuć się złym człowiekiem.
7. Jeżeli nie trzymam się mocno w jednej pozycji, czuję się słaby/a.
8. Potrzebuję mieć w życiu porządek, aby nie czuć, że tracę kontrolę.
9. Kiedy muszę spędzić dużo czasu na jednej aktywności, zarzucam sobie, że
nie poświęcam czasu innym rzeczom.
10. Czuję zdezorientowany/a, gdy nie mogę podążać za swoją rutyną.
11. Aby być usatysfakcjonowanym/ą, muszę wszystko zrobić perfekcyjnie, tak
jak sobie tego życzę.
12. Czuję się tak, jakbym gubił/a się w swoim życiu, jeżeli nie spełniam swoich
własnych oczekiwań.
https://doi.org/10.1371/journal.pone.0283795.t001
.750
.645
.721
.749
.699
.678
.643
.816
.576
.562
.793
.625
.731
.673
.705
.682
.552
.621
.684
.638
.632
.586
.741
.542
.510
.706
.581
.650
.599
.642
equivalence of the paired items. The constraint was released one by one and resulted in the
acceptable model fit: robust χ2(131) = 1937.53, p < .001; robust RMSEA = .090 (95% CI: [.087,
.094]), SRMR = .056, robust CFI = .988, robust TLI = .986. The same procedure was followed
in the case of GPQ9 and GSPQ, which is described below.
GPQ9. The one-factor model exhibited a non-satisfactory fit due to the high value of
robust RMSEA: robust χ2(27) = 810.56, p < .001; robust RMSEA = .129 (95% CI: [.122, .137]),
SRMR = .064, robust CFI = .984, robust TLI = .979. Again, based on modification indices (MI),
the error terms between items 10 and 11 (MI = 123.45), and 4 and 5 (MI = 75.76) were allowed
to correlate. This improved the model fit: robust χ2(25) = 340.64, p < .001; robust RMSEA =
.082 (95% CI: [.074, .90]), SRMR = .044, robust CFI = .994, robust TLI = .992.
GTQ. The one-factor model exhibited a satisfactory fit: robust χ2(44) = 680.31, p < .001;
robust RMSEA = .089 (95% CI: [.083, .095]), SRMR = .045, robust CFI = .992, robust TLI =
.991.
GSPQ. The one-factor model exhibited a non-satisfactory fit due to the high value of
robust RMSEA: robust χ2(54) = 854.54, p < .001; robust RMSEA = .097 (95% CI: [.092, .103]),
Fig 1. Standardized solution of the GPQ18 one-factor model in a Polish sample.
https://doi.org/10.1371/journal.pone.0283795.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
9 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
Fig 2. Standardized solution of the GPQ9 one-factor model in a Polish sample.
https://doi.org/10.1371/journal.pone.0283795.g002
SRMR = .055, robust CFI = .985, robust TLI = .982. one more time, based on modification
indices (MI), the error terms between items 1 and 2 (MI = 75.93), 11 and 12 (MI = 60.53) were
allowed to correlate. This improved the model fit: robust χ2(52) = 553.44, p < .001; robust
RMSEA = .078 (95% CI: [.072, .084]), SRMR = .048, robust CFI = .991, robust TLI = .989.
Summarizing, the results of CFA generally confirmed the one-factor solution obtained in
original studies for each scale. In the case of three scales (GPQ18, GPQ9 and GSPQ) the origi-
nal model presented a non-acceptable fit and some modifications were applied to obtain at
least acceptable fit of the model. The final solutions present non-significant chi-square statis-
tics (however, because of the big sample sizes this particular statistic is not a reliable source
about the model fit, see [40]), satisfactory robust CFI, robust TLI and SRMR and at least
acceptable values of robust RMSEA. Factor loadings were moderately or strongly related to
their purported latent factor in the case of each scale.
Measurement invariance across samples (A and B) and gender
Metric, scalar and strict invariance across both samples (A and B) and gender were conducted.
The relative fits of four increasingly restrictive models were compared: the multigroup baseline
model (allowing factor loadings to vary across groups while the factor structure was identical
across groups (i.e., configural invariance), the metric invariance model (placing equality con-
straints on factor loadings across groups), the scalar invariance model (placing equality con-
straints on factor loadings and item intercepts), and the strict invariance model (placing
equality constraints on factor loadings, item intercepts and residuals). The models were com-
pared taking into account the differences in robust RMSEA (ΔRMSEA), CFI (ΔCFI), and TLI
(ΔTLI) indexes between nested models, with ΔCFI being regarded as least affected by the
model complexity and sample size [41]. Although the chi-square statistics and their differences
between the models are also presented, due to the big sample sizes they should not be treated
as decisive; a ΔCFI, ΔTLI and ΔRMSEA less than .01 indicated invariance (see [40–43]).
Fig 3. Standardized solution of the GTQ one-factor model in a Polish sample.
https://doi.org/10.1371/journal.pone.0283795.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
10 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
The results of the analyses are presented in Table 2. Baseline models present well-fit in the
case of each scale. When it comes to the invariance, it can be concluded that metric, scalar and
strict invariance are supported across both samples A and B in the case of each scale except the
GPQ9. At the same time, the measurement can be treated as invariant regardless of the gender
in the case of each scale.
Validity
The bivariate r-Pearson correlations were calculated between GPQ18, GPQ9, GTQ and GSPQ
themselves and between adapted scales and other tools measuring life satisfaction, the level of
depression, anxiety and stress, the level of perceived general self-efficacy, psychological inflexi-
bility, cognitive fusion, general valued living as well as rumination and reflection. The results
are presented in Table 3. They generally support the convergent and divergent validity of
adapted scales and are consistent across the samples (with differences not being tested
directly).
The correlation between GPQ18 and GPQ9 were very high in both samples, suggesting that
both tools measure the same construct (accordingly with the expectations). Pliance (measured
by both GPQ18 and GPQ9) was strongly related to self-pliance, which was also consistent with
a priori hypotheses. Tracking was weakly (and negatively) or non-significantly related to
pliance and self-pliance, showing that these constructs reflect different and strongly indepen-
dent rule-governed behaviors.
Pliance (as measured by GPQ18 and GPQ9) and self-pliance was moderately or strongly
and positively related to the level of depressive, anxiety and stress symptoms, psychological
inflexibility, cognitive fusion, obstruction of valued living and rumination. They showed weak
positive or non-significant relationships with reflection, and progress in valued living. The
relationship with general self-efficacy and life satisfaction was weak and negative or non-signif-
icant. These least results seem less expected, suggesting that possibly life satisfaction and self-
efficacy are less affected by the level of presented pliance and self-pliance (the hypothesis
worth further testing).
Tracking showed to be positively and moderately or strongly related to life satisfaction, gen-
eral self-efficacy and progress in valued living and weakly with reflection. It was also negatively
related (at least moderately) with depressive, anxiety and stress symptoms, psychological
inflexibility, cognitive fusion, obstruction of valued living and rumination. The results are gen-
erally consistent with set hypotheses and those obtained in a study of original scale [12].
Discussion
The aim of the study was to evaluate the psychometric properties of Polish adaptations of the
questionnaires measuring rule-governed behaviors such as pliance, self-pliance and tracking.
The EFA and CFA analyses corroborated unidimensional structure with high factor load-
ings found in the original validation studies for each of the measurements: GTQ [12], GSPQ
Fig 4. Standardized solution of the GSPQ one-factor model in a Polish sample.
https://doi.org/10.1371/journal.pone.0283795.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
11 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
Table 2. Measurement invariance across samples (A and B) and gender.
Model
χ2 (df)
GPQ18
χ2 diff
p diff
RMSEA
Measurement invariance across samples A and B
∆RMSEA
CFI
∆CFI
TLI
∆TLI
GPQ9
GTQ
GSPQ
Baseline model
Metric invariance
Scalar invariance
Strict invariance
Baseline model
Metric invariance
Scalar invariance
Strict invariance
Baseline model
Metric invariance
Scalar invariance
Strict invariance
Baseline model
Metric invariance
Scalar invariance
Strict invariance
293.83 (262)
375.51 (279)
539.35 (296)
554.75 (314)
310.61 (262)
447.51 (279)
523.78 (296)
545.62 (314)
77.72 (50)
90.02 (58)
208.73 (66)
213.70 (75)
84.76 (50)
117.92 (58)
169.63 (66)
180.73 (75)
Baseline model
63.07 (82)
Metric invariance
133.90 (92)
Scalar invariance
Strict invariance
169.08 (102)
181.70 (113)
Baseline model
60.10 (82)
Metric invariance
106.37 (92)
Scalar invariance
Strict invariance
116.34 (102)
133.10 (113)
Baseline model
Metric invariance
Scalar invariance
Strict invariance
Baseline model
Metric invariance
Scalar invariance
Strict invariance
130.56 (104)
185.09 (115)
235.92 (126)
240.29 (138)
129.82 (104)
158.13 (115)
186.00 (126)
193.24 (138)
.060
< .001
.169
.044
.045
.055
.054
-.001
-.010
.001
Measurement invariance across gender
< .001
< .001
.011
.046
.051
.054
.053
-.005
-.003
.001
Measurement invariance across samples A and B
.372
< .001
.523
.054
.050
.077
.072
.004
-.027
.005
Measurement invariance across gender
.003
< .001
.025
.057
.060
.068
.065
-.003
-.008
.003
Measurement invariance across samples A and B
< .001
< .001
.097
.029
.042
.046
.045
-.013
-.004
.001
Measurement invariance across gender
.007
.037
.012
.029
.035
.035
.035
-.006
.000
.000
Measurement invariance across samples A and B
.006
< .001
.904
.041
.045
.050
.047
-.004
-.005
.003
Measurement invariance across gender
.282
< .001
.600
.040
.039
.041
.039
.001
-.002
.002
26.88
328.71
23.60
46.85
157.13
34.55
8.66
248.85
8.11
23.18
114.62
19.01
35.98
66.56
17.37
24.44
19.24
24.14
26.38
90.02
6.23
13.18
48.03
10.18
.991
.990
.985
.985
.991
.988
.985
.985
.991
.990
.975
.975
.990
.987
.980
.979
.996
.991
.988
.987
.996
.994
.993
.992
.991
.987
.983
.984
.991
.990
.988
.988
-.001
-.005
.000
.003
.003
.000
.001
.015
.000
.003
.007
.001
-.005
-.003
-.001
-.002
-.001
-.001
-.004
-.004
.001
-.001
-.002
.000
.990
.989
.984
.985
.989
.987
.985
.985
.986
.988
.972
.976
.985
.983
.978
.980
.995
.989
.987
.988
.995
.992
.993
.992
.988
.986
.982
.984
.988
.989
.988
.989
.001
.005
-.001
.002
.002
.000
.016
.016
-.004
.002
.005
.005
-.002
.006
.002
-.001
.003
-.001
.001
.002
.004
-.002
-.001
.001
-.001
Note: GPQ18 –the 18-item version of Generalized Pliance Questionnaire, GPQ9 –the 9-item version of Generalized Pliance Questionnaire, GTQ–Generalized Tracking
Questionnaire, GSPQ—Generalized Self-Pliance Questionnaire.
https://doi.org/10.1371/journal.pone.0283795.t002
[10] and both versions of GPQ: GPQ18 and GPQ9 [11]. Model fit to data was acceptable. All
of the adapted scales presented good reliability (internal consistency) and item-total
correlations.
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
12 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
Table 3. Descriptive statistics and correlations between GPQ18, GPQ9, GTQ, GSPQ and other measures’ scales in both samples.
Scales
SWLS
DASS Depression
DASS Anxiety
DASS Stress
GSES
AAQ—II
CFQ
VQ Progress
VQ Obstruction
RRQ Rumination
RRQ Reflection
GPQ18
GPQ9
GTQ
GSPQ
SWLS
DASS_D
DASS_A
DASS_S
GSES
AAQ
CFQ
VQP
VQO
RRQ_RU
RRQ_RE
GPQ18
GPQ9
GTQ
GSPQ
α
M (SD)
Skewness (SE)
Kurtosis (SE)
GPQ18
GPQ9
GTQ
GSPQ
Sample A
.873
.913
.884
.888
.886
.933
.954
.830
.847
.825
.820
.934
.884
.903
.903
.896
.920
.874
.903
.927
.941
.961
.829
.858
.844
.694
.934
.890
.910
.899
22.42 (6.09)
6.51 (5.38)
5.40 (5.01)
8.29 (5.14)
29.92 (4.57)
24.09 (10.37)
27.62 (10.28)
16.89 (6.12)
12.51 (6.72)
20.47 (5.12)
22.80 (4.79)
71.12 (17.53)
35.45 (9.03)
53.47 (8.95)
47.90 (12.36)
19.50 (6.09)
6.25 (5.35)
4.43 (4.45)
7.04 (4.95)
29.40 (4.87)
23.27 (10.42)
23.82 (10.79)
15.99 (6.16)
11.82 (6.92)
18.61 (5.69)
19.31 (4.17)
67.47 (18.05)
33.82 (9.48)
53.31 (9.39)
46.72 (12.16)
.23 (.12)
.74 (.12)
.84 (.12)
.37 (.12)
-.21 (.12)
.23 (.12)
-.07 (.12)
-.27 (.12)
.13 (.12)
-.20 (.12)
-.27 (.12)
.02 (.12)
-.01 (.12)
.07 (.12)
-.28 (.12)
-.34 (.10)
.76 (.10)
1.13 (.10)
.48 (.10)
-.70 (.10)
.35 (.10)
.25 (.10)
-.22 (.10)
.24 (.10)
-.09 (.10)
.37 (.10)
.04 (.09)
.03 (.09)
.11 (.09)
-.04 (.09)
Sample B
-.22 (.23)
-.36 (.23)
-.24 (.23)
-.62 (.23)
.95 (.23)
-.67 (.23)
-.57 (.23)
-.38 (.23)
-.68 (.23)
-.27 (.23)
-.72 (.23)
.14 (.23)
.20 (.23)
.03 (.23)
.27 (.23)
.07 (.19)
-.23 (.19)
.80 (.19)
-.44 (.19)
2.06 (.19)
-.59 (.20)
-.64 (.20)
-.11 (.20)
-.57 (.20)
-.55 (.20)
.47 (.20)
.11 (.19)
-.05 (.19)
.08 (.19)
.24 (.19)
-.226**
.312**
.319**
.342**
-.267**
.446**
.437**
-.119*
.343**
.422**
.040
.008
.298**
.279**
.289**
-.108**
.429**
.4398**
.042
.406**
.350**
.121**
-.206**
.295**
.293**
.312**
-.240**
.410**
.395**
-.113*
.320**
.397**
.029
.967**
.000
.280**
.261**
.268**
-.112**
.405**
.361**
.022
.388**
.321**
.096*
.974**
.353**
-.354**
-.295**
-.302**
.620**
-.338**
-.263**
.432**
-.296**
-.213**
.238**
-.199**
-.176**
.322**
-.185**
-.141**
-.129**
.581**
-.234**
-.184**
.454**
-.220**
-.217**
.130**
.007
.001
-.231**
.402**
.430**
.468**
-.277**
.543**
.567**
-.071
.436**
.455**
.035
.595**
.534**
-.174**
-.104**
.440**
.401**
.489**
-.105**
.523**
.589**
.077
.527**
.508**
.270**
.594**
.550**
.067
Note: M–mean, SD–standard deviation, SE–standard error, α –Cronbach Alpha, SWLS–Satisfaction with Life Scale, DASS Depression–the Depression scale of
Depression, Anxiety, and Stress Scale, DASS Anxiety–the Anxiety scale of Depression, Anxiety, and Stress Scale, DASS Stress–the Stress scale of Depression, Anxiety,
and Stress Scale, GSES–General Self-efficacy Scale, AAQ–II–Acceptance and Action Questionnaire–II, CFQ–Cognitive Fusion Questionnaire, VQ Progress–the
Progress scale of Valuing Questionnaire, VQ Obstruction–the Obstruction scale of Valuing Questionnaire, RRQ Rumination–the Rumination scale of Rumination–
Reflection Questionnaire, RRQ Reflection–the Reflection scale of Rumination–Reflection Questionnaire, GPQ18 –the 18-item version of Generalized Pliance
Questionnaire, GPQ9 –the 9-item version of Generalized Pliance Questionnaire, GTQ–Generalized Tracking Questionnaire, GSPQ—Generalized Self-Pliance
Questionnaire;* p < .05, ** p < .01
https://doi.org/10.1371/journal.pone.0283795.t003
Polish versions of questionnaires presented significant correlations in the expected direc-
tions with relevant psychological variables in line with the original studies [10–12]. Regarding
ACT processes, pliance (measured by Polish versions of GPQ9 and GPQ18) and self-pliance
(GSPQ) was positively related to, cognitive fusion and obstruction of valued living, and not
related or negatively correlated with progress in valued living. This last score is slightly differ-
ent from the results obtained by Ruiz and colleagues [11] and by Ruiz and Sua´rez-Falco´n and
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
13 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
colleagues [10], who showed consistent negative correlations between GPQ9, GPQ18 and
GSPQ and progress in valued living. In our study, only in the student sample GPQ18 and
GPQ9 were negatively correlated with progress in valued living, whereas in the general sample,
as well as for GSPQ there were no significant correlations.
Pliance and self-pliance were negatively related to psychological flexibility as measured by
AAQ-II [26, 27] similarly to the results obtained by Ruiz and Sua´rez-Falco´n and colleagues
[10] and Ruiz and colleagues [11] who also used AAQ-II in their studies. Despite AAQ-II have
been questioned as a precise and adequate measure of psychological flexibility [44, 45], the
result is in line with the results obtained by researchers that used different measurements with
greater construct validity (e.g. CompACT [46]). Pliance and self-pliance were positively related
with rumination and emotional symptoms in line with the original validation studies [10, 11]
and showed weak and positive or non-significant relationship with reflection. Finally, they pre-
sented weak and negative or non-significant correlations with life satisfaction and self-efficacy.
In contrast to the original studies, GPQ18 and GPQ9 in the general sample were not related to
life satisfaction.
Finally pliance, self-pliance were positively correlated in both samples, yet pliance and self-
pliance and tracking were negatively correlated only in the student sample, which is contrary
to expectations and needs further replication.
Summarizing, pliance and self-pliance seem negatively related to psychological flexibility.
Although the results of original studies suggest pliance may be a process leading to lower psy-
chological flexibility and the lower level of life satisfaction, the latter was not found in the Pol-
ish sample.
Tracking, as measured by the Polish version of GTQ, was positively related to progress in
valued living and negatively related with psychological inflexibility, cognitive fusion, obstruc-
tion to valued living. Moreover, it was positively correlated with life satisfaction, general self-
efficacy and reflection and negatively with rumination and emotional symptoms in line with
the original validation study [12]. The results support the idea that tracking, i.e. being in direct
contact with contingencies and following the real consequences of behavior, is a process which
may support living a valuable life, the feeling of greater self-efficacy, which may lead to greater
satisfaction with life and, possibly, general health.
Despite indicating the validity of the Polish adaptations of GPQ, GSPQ and GTQ, the pres-
ent study should be complemented by further research. Although in general correlations with
other measures are in line with original validation studies, there may be questions regarding
lack or very weak correlation of pliance with life satisfaction. At the same time, the relationship
between tracking and life satisfaction as well as self-efficacy was consistent in the original and
Polish study. The explanation may be the cultural differences between Polish and South Amer-
ican societies. The main difference between pliance and tracking is apparent source of rein-
forcement for rule-following: in pliance it is social or arbitrary, in tracking–non arbitrary. For
some reason following the behavior considered as socially approved in Poland seems to have
little impact on life satisfaction while it may still have detrimental effects for individuals’ gen-
eral functioning (because of negative relationship with psychological flexibility). Differences
between Irish as well as Columbian adolescents in pliance and psychological inflexibility has
been recently reported by Stapleton et al. [47]. Further longitudinal research allowing for
cause-effect conclusions should determine the relationship between pliance, tracking, life satis-
faction and self-efficacy in cross-cultural studies.
The present study has some more limitations. First of all, the validity of the measurements
should be tested in clinical samples due to its potential utility in research and practice in the
area of psychopathology. Secondly, the validity of GTQ should also be confirmed in future
studies including its relationship with executive functions. Further studies should include
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
14 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
more precise and adequate measurements of psychological flexibility. Furthermore, the stabil-
ity of the results needs to be established in additional study. The study has been conducted
online and we have not compared the mode of administration of measurements (e.g. online vs
pen-and-paper methods might be regarded by participants as providing different levels of ano-
nymity, and this may lead to differences in how socially biased responding would be [48]).
Finally, the samples were not representative of the Polish general and student’s population.
In conclusion, this study contributes with the GPQ, GSPQ and GTQ to the Polish language
and they appear to be adequate to be used in Polish samples which may accelerate the develop-
ment of research on pliance, self-pliance and tracking in Polish language and support clini-
cians working with clients.
Acknowledgments
We would like to thank our colleagues: Krystyna Pomorska, Jan Topczewski, Lidia Baran and
Monika Suchowierska-Stephany, who participated in the preparation of Polish translations of
the questionnaires.
Author Contributions
Conceptualization: Joanna Dudek, Maria Cyniak-Cieciura, Paweł Ostaszewski.
Data curation: Joanna Dudek, Maria Cyniak-Cieciura.
Formal analysis: Maria Cyniak-Cieciura.
Funding acquisition: Joanna Dudek.
Investigation: Joanna Dudek, Maria Cyniak-Cieciura.
Methodology: Joanna Dudek, Maria Cyniak-Cieciura, Paweł Ostaszewski.
Project administration: Joanna Dudek, Maria Cyniak-Cieciura.
Resources: Joanna Dudek, Maria Cyniak-Cieciura.
Supervision: Paweł Ostaszewski.
Visualization: Maria Cyniak-Cieciura.
Writing – original draft: Joanna Dudek, Maria Cyniak-Cieciura, Paweł Ostaszewski.
Writing – review & editing: Joanna Dudek, Maria Cyniak-Cieciura, Paweł Ostaszewski.
References
1. Skinner BF. An operant analysis of problem solving. In Kleinmuntz B, editor. Problem solving: research,
method and theory. New York: Wiley; 1966. p.225–57.
2.
Zettle RD, Hayes SC. Rule-Governed Behavior: A Potential Theoretical Framework for Cognitive-
Behavioral Therapy. In: Advances in cognitive–behavioral research and therapy Academic
Press.1982, 73–118.
3. Barnes-Holmes D, O’;Hora D, Roche B, Hayes SC, Bissett RT, Lyddy F. Understanding and Verbal
Regulation. In: Hayes SC, Barnes-Holmes D, Roche B, editors. Relational Frame Theory: A Post-Skin-
nerian Account of Human Language and Cognition. Boston, MA: Springer US; 2001. p. 103–17. Avail-
able from: https://doi.org/10.1007/0-306-47638-X_6
4. Hayes SC, Strosahl KD, Wilson KG. Acceptance and commitment therapy: An experiential approach to
behavior change. New York, NY, US: Guilford Press; 1999.
5.
Luciano C, Valdivia-Salas S, Ruiz F. The self as the context for rule-governed behavior. In In McHugh,
Stewart (Eds.), The self and perspective taking: Research and applications, Context Press, Oakland,
2012. p. 143–60.
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
15 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
6. McAuliffe D, Hughes S, Barnes-Holmes D. The Dark-Side of Rule Governed Behavior: An Experimental
Analysis of Problematic Rule-Following in an Adolescent Population With Depressive Symptomatology.
Behav Modif. 2014 Jul 1; 38(4):587–613. https://doi.org/10.1177/0145445514521630 PMID: 24509717
7.
To¨ rneke N, Luciano C, Salas SV. Rule-governed behavior and psychological problems. Int J Psychol
Psychol Ther. 2008; 8:141–56.
8. Monestès JL, Villatte M, Stewart I, Loas G. Rule-Based Insensitivity and Delusion Maintenance in
Schizophrenia. Psychol Rec. 2014 Jun 1; 64(2):329–38.
9. Hayes SC, Strosahl KD, Wilson KG. Acceptance and commitment therapy: The process and practice of
mindful change, 2nd ed. New York NY, US: Guilford Press; 2012.
10. Ruiz FJ, Sua´rez-Falco´ n JC, Segura MA, Garcı´a-Martı´n MB, Henao AM, Bedoya-Valderrama L. Devel-
opment of the Generalized Self-Pliance Questionnaire. Paper presented at: The ACBS World Confer-
ence 18; 2020 Jul 14–16, online.
11. Ruiz FJ, Sua´rez-Falco´ n JC, Barbero-Rubio A, Flo´rez CL. Development and initial validation of the Gen-
eralized Pliance Questionnaire. J Context Behav Sci. 2019 Apr 1; 12:189–98.
12. Ruiz FJ, Garcı´a-Martı´n MB, Sua´rez-Falco´n JC, Bedoya-Valderrama L, Segura-Vargas MA, Peña-Var-
gas A, et al. Development and initial validation of the Generalized Tracking Questionnaire. PLOS ONE.
2020 Jun 11; 15(6):e0234393. https://doi.org/10.1371/journal.pone.0234393 PMID: 32525904
13. Hayes SC, Zettle RD, Rosenfarb I. Rule-Following. In: Hayes SC, editor. Rule-Governed Behavior:
Cognition, Contingencies, and Instructional Control. Boston, MA: Springer US; 1989. p. 191–220.
Available from: https://doi.org/10.1007/978-1-4757-0447-1_6
14. Skinner BF. Verbal behavior. New York: Appleton-Century-Crofts.1957.
15. Monestès JL, Greville WJ, Hooper N. Derived insensitivity: Rule-based insensitivity to contingencies
propagates through equivalence. Learn Motiv. 2017; 59:55–63.
16. Hayes SC, Wilson KG, Gifford EV, Follette VM, Strosahl K. Experiential avoidance and behavioral disor-
ders: A functional dimensional approach to diagnosis and treatment. J Consult Clin Psychol. 1996;
64:1152–68.
17. Kissi A, Hughes S, Mertens G, Barnes-Holmes D, De Houwer J, Crombez G. A Systematic Review of
Pliance, Tracking, and Augmenting. Behav Modif. 2017 Sep 1; 41(5):683–707. https://doi.org/10.1177/
0145445517693811 PMID: 28776429
18. Hayes SC, Hayes LJ. The Verbal Action of the Listener as a Basis for Rule-Governance. In: Hayes SC,
editor. Rule-Governed Behavior: Cognition, Contingencies, and Instructional Control Boston, MA:
Springer US; 1989 p. 153–90. Available from: https://doi.org/10.1007/978-1-4757-0447-1_5
19. O’Connor M, Byrne P, Ruiz FJ, McHugh L. Generalized Pliance in Relation to Contingency Insensitivity
and Mindfulness. Mindfulness. 2019 May 1; 10(5):833–40.
20. Diener E, Emmons RA, Larsen RJ, Griffin S. The Satisfaction With Life Scale. J Pers Assess. 1985 Feb
1; 49(1):71–5. https://doi.org/10.1207/s15327752jpa4901_13 PMID: 16367493
21.
22.
Jankowski KS. Is the shift in chronotype associated with an alteration in well-being? Biol Rhythm Res.
2015 Mar 4; 46(2):237–48.
Lovibond PF, Lovibond SH. The structure of negative emotional states: comparison of the Depression
Anxiety Stress Scales (DASS) with the Beck Depression and Anxiety Inventories. Behav Res Ther.
1995 Mar; 33(3):335–43. https://doi.org/10.1016/0005-7967(94)00075-u PMID: 7726811
23. Makara-Studzińska M, Tyburski E, Załuski M, Adamczyk K, Mesterhazy J, Mesterhazy A. Confirmatory
Factor Analysis of Three Versions of the Depression Anxiety Stress Scale (DASS-42, DASS-21, and
DASS-12) in Polish Adults. Front Psychiatry 2022 Jan 4; 12:770532. https://doi.org/10.3389/fpsyt.2021.
770532 PMID: 35058818
24. Schwarzer R, Jerusalem M. (1995). Generalized Self-Efficacy scale. In Weinman J, Wright S, Johnston
M, Measures in health psychology: A user’s portfolio. Causal and control beliefs, Windsor, UK : NFER-
NELSON. 1995, (pp. 35–37).
25.
Juczyński Z. Narze¸dzia pomiaru w promocji i psychologii zdrowia. Pracownia Testo´ w Psychologicznych
Polskiego Towarzystwa Psychologicznego. Warszawa. 2009.
26. Bond FW, Hayes SC, Baer RA, Carpenter KM, Guenole N, Orcutt HK, et al. Preliminary Psychometric
Properties of the Acceptance and Action Questionnaire–II: A Revised Measure of Psychological Inflexi-
bility and Experiential Avoidance. Behav Ther. 2011 Dec 1; 42(4):676–88. https://doi.org/10.1016/j.
beth.2011.03.007 PMID: 22035996
27. Kleszcz B, Dudek JE, Białaszek P, Bond FW. The Psychometric Properties of the Polish Version of the
Acceptance and Action Questionnaire-II (AAQ-II). Stud Psychol. 2018 Jun 5; 56(1):1–19.
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
16 / 17
PLOS ONE The Polish adaptation of the measurements of rule-governed behaviors: GPQ, GTQ and GSPQ
28. Baran L, Hyla M, Kleszcz B. Elastyczność psychologiczna. Polska adaptacja narzędzi pomiarowych dla
praktyko´ w i badaczy Katowice: Wydawnictwo Uniwersytetu Śląskiego; 2019. Available from: https://
rebus.us.edu.pl/handle/20.500.12128/12945
29. Gillanders DT, Bolderston H, Bond FW, Dempster M, Flaxman PE, Campbell L, et al. The Development
and Initial Validation of the Cognitive Fusion Questionnaire. Behav Ther. 2014 Jan 1; 45(1):83–101.
https://doi.org/10.1016/j.beth.2013.09.001 PMID: 24411117
30. Smout M, Davies M, Burns N, Christie A. Development of the Valuing Questionnaire (VQ). J Context
Behav Sci. 2014 Jul 1; 3(3):164–72.
31. Słowińska A, Zbieg A, Oleszkowicz A. Kwestionariusz Ruminacji-Refleksji (RRQ) Paula D. Trapnella i
Jennifer D. Campbell—polska adaptacja metody. Paul D Trapnell and Jennifer D Campbell’s Rumina-
tion-Reflection Questionnaire (RRQ)—polish adaptation of the measure. 2014; Available from: https://
repozytorium.ukw.edu.pl///handle/item/1371
32.
33.
Trapnell PD, Campbell JD. Private self-consciousness and the five-factor model of personality: Distin-
guishing rumination from reflection. J Pers Soc Psychol. 1999; 76:284–304. https://doi.org/10.1037//
0022-3514.76.2.284 PMID: 10074710
Lorenzo-Seva U, Ferrando PJ. FACTOR: a computer program to fit the exploratory factor analysis
model. Behav Res Methods. 2006 Feb; 38(1):88–91. https://doi.org/10.3758/bf03192753 PMID:
16817517
34. Rosseel Y. lavaan: An R Package for Structural Equation Modeling. J Stat Softw. 2012 May 24; 48:1–
36.
35.
36.
Timmerman ME, Lorenzo-Seva U. Dimensionality assessment of ordered polytomous items with paral-
lel analysis. Psychol Methods. 2011 Jun; 16(2):209–20. https://doi.org/10.1037/a0023353 PMID:
21500916
Ferrando PJ, Lorenzo-Seva U. Assessing the Quality and Appropriateness of Factor Solutions and Fac-
tor Score Estimates in Exploratory Item Factor Analysis. Educ Psychol Meas. 2018 Oct; 78(5):762–80.
https://doi.org/10.1177/0013164417719308 PMID: 32655169
37. Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria
versus new alternatives. Struct Equ Model Multidiscip J. 1999 Jan 1; 6(1):1–55.
38. Xia Y, Yang Y. RMSEA CFI, and TLI in structural equation modeling with ordered categorical data: The
story they tell depends on the estimation methods. Behav Res Methods. 2019 Feb 1; 51(1):409–28.
39. Shi D, Maydeu-Olivares A. The Effect of Estimation Methods on SEM Fit Indices. Educ Psychol Meas.
2020 Jun; 80(3):421–45. https://doi.org/10.1177/0013164419885164 PMID: 32425213
40. Cheung GW, Rensvold RB. Evaluating Goodness-of-Fit Indexes for Testing Measurement Invariance.
Struct Equ Model Multidiscip J. 2002 Apr 1; 9(2):233–55.
41. Meade AW, Johnson EC, Braddy PW. Power and sensitivity of alternative fit indices in tests of measure-
ment invariance. J Appl Psychol. 2008 May; 93(3):568–92. https://doi.org/10.1037/0021-9010.93.3.568
PMID: 18457487
42. Chen FF. Sensitivity of Goodness of Fit Indexes to Lack of Measurement Invariance. Struct Equ Model
Multidiscip J. 2007 Jul 31; 14(3):464–504.
43. Xu H, Tracey TJG. Use of multi-group confirmatory factor analysis in examining measurement invari-
ance in counseling psychology research. Eur J Couns Psychol [Internet]. 2017 Feb 14 [cited 2022 Sep
18];6(1). Available from: https://www.psycharchives.org/en/item/ebd7c42d-bfdb-4399-b6f4-
b54708edb8d0
44. Doorley JD, Goodman FR, Kelso KC, Kashdan TB. Psychological flexibility: What we know, what we do
not know, and what we think we know. Soc Personal Psychol Compass. 2020; 14(12):e12566.
45.
Tyndall I, Waldeck D, Pancani L, Whelan R, Roche B, Dawson DL. The Acceptance and Action Ques-
tionnaire-II (AAQ-II) as a measure of experiential avoidance: Concerns over discriminant validity. J Con-
text Behav Sci. 2019; 12:278–84.
46. Waldeck D, Pancani L, Tyndall I. An examination of the construct validity of the Generalized Pliance
Questionnaire. J Context Behav Sci. 2019; 11:50–4.
47. Stapleton A, Ruiz FJ, McHugh L. Comparative investigation of adolescents’ generalized pliance and
psychological inflexibility across cultural contexts. Psychol Rec. 2022; 72:105–9.
48. Ward P, Clark T, Zabriskie R, Morris T. Paper/Pencil Versus Online Data Collection. J Leis Res. 2014
Mar 1; 46(1):84–105.
PLOS ONE | https://doi.org/10.1371/journal.pone.0283795 April 5, 2023
17 / 17
PLOS ONE
| null |
10.1038_s41467-023-40247-4.pdf
| null |
Data availability The raw reads, and complete pathogen genomes generated in this study have been deposited in the Sequence Read Archive (SRA) and NCBI GenBank, respectively, under BioProject accession codes PRJNA824010 and PRJNA436552 . Sample metadata (collection date, state, age, sequencing machine, sequencing batch, etc.), metagenomic read classification data for all samples and controls, viral genome assembly data, reference sequence accession numbers, and RT-qPCR results generated in this study are provided in the Supplementary Data 1 file. Code availability Open source software used in this study is available at https://github. com/broadinstitute/viral-ngs 84 (i.e., pipelines for viral genomic analyses; v2.1.8) and at https://github.com/bpetros95/lassa-metagenomics 94 (i.e., code for statistical analyses; developed in R v4.1.1 with packages bda v15.2.5, mediation v4.5.0, ROCR v1.0-11, stats v4.1.1, and tidyverse v2.0.0). Information about the Microsoft Premonition metagenomics pipeline is available at https://microsoft.com/premonition . Individuals can access the pipeline ahead of its public release by clicking the 'Contact us for availability' button and mentioning this work or by emailing Simon Frost at [email protected].
|
Article
https://doi.org/10.1038/s41467-023-40247-4
Metagenomic surveillance uncovers diverse
and novel viral taxa in febrile patients from
Nigeria
Received: 18 January 2023
A list of authors and their affiliations appears at the end of the paper
Accepted: 10 July 2023
Check for updates
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
Effective infectious disease surveillance in high-risk regions is critical for
clinical care and pandemic preemption; however, few clinical diagnostics are
available for the wide range of potential human pathogens. Here, we conduct
unbiased metagenomic sequencing of 593 samples from febrile Nigerian
patients collected in three settings: i) population-level surveillance of indivi-
duals presenting with symptoms consistent with Lassa Fever (LF); ii) real-time
investigations of outbreaks with suspected infectious etiologies; and iii)
undiagnosed clinically challenging cases. We identify 13 distinct viruses,
including the second and third documented cases of human blood-associated
dicistrovirus, and a highly divergent, unclassified dicistrovirus that we name
human blood-associated dicistrovirus 2. We show that pegivirus C is a com-
mon co-infection in individuals with LF and is associated with lower Lassa viral
loads and favorable outcomes. We help uncover the causes of three outbreaks
as yellow fever virus, monkeypox virus, and a noninfectious cause, the latter
ultimately determined to be pesticide poisoning. We demonstrate that a local,
Nigerian-driven metagenomics response to complex public health scenarios
generates accurate, real-time differential diagnoses, yielding insights that
inform policy.
Infectious diseases place a large, global burden on human health.
There are hundreds of known human pathogens, which differ in their
pathogenesis, epidemiology, and therapeutic vulnerabilities. More-
over, the detection of emerging pathogens has accelerated, driven by
ecological, environmental, and sociodemographic factors1 as well as
increased surveillance and diagnostic testing2. Accurate and timely
diagnosis is essential for both clinical care and mitigation of further
transmission. However, clinical diagnosis remains a challenge, as many
pathogens present with highly overlapping sets of non-specific
symptoms (e.g., fever, swollen lymph nodes, or malaise), and the
presence of one pathogen does not preclude the presence of others
(bluntly phrased by John Hickam: “patients can have as many diseases
as they damn well please”)3,4. In low- and middle-income countries
(LMICs), the disease burden is often the highest, but molecular diag-
nostics are limited. Consequently, misdiagnosis with common
pathogens such as malaria or typhoid fever, or the failure to receive a
diagnosis, occurs frequently in LMICs5–8.
The rapid determination of all species in a sample through
metagenomic analysis9–11 can identify potential causal agents of febrile
illness in an unbiased, high-throughput manner. Metagenomics,
alongside more sensitive approaches such as virome capture
sequencing12, can thus transform diagnostic microbiology13 and out-
break responses14. The development of genomics infrastructure in
Africa has enabled the continent to lead in the characterization of
numerous emerging SARS-CoV-2 variants15–19 and holds promise for
the genomic interrogation of endemic pathogens20. Because genomics
remains relatively expensive and requires technical expertise to both
generate and analyze the data, it cannot be readily applied to every
sample, necessitating an understanding of the most valuable applica-
tions of metagenomics in real-world settings.
e-mail: [email protected]; [email protected]; [email protected]
Nature Communications |
(2023) 14:4693
1
Article
https://doi.org/10.1038/s41467-023-40247-4
To evaluate the utility of metagenomic sequencing for pathogen
surveillance and detection, we genomically characterized viral infec-
tions in plasma samples collected for three distinct use cases over 4
years (2017–2020) in Nigeria (Fig. 1). Nigeria has multiple factors that
make it a meaningful country to study the efficacy of metagenomics in
infectious disease surveillance, including a high burden of infectious
disease, sequencing capacity at the African Centre of Excellence for
Genomics of Infectious Diseases (ACEGID), and a strong partnership
between ACEGID and national public health institutions, especially the
Nigerian Centre for Disease Control (NCDC). Here, we report the
results of (i) a study of suspected Lassa Fever (LF) cases, where we
examine Lassa virus (LASV), non-LASV viral etiologies, and cases of co-
infections; (ii) rapid investigations of three outbreaks suspected of
infectious etiologies; and (iii) metagenomic diagnosis of clinically
challenging cases. We report the strengths and limitations of, as well as
the insights derived from, sequencing technologies in each of these
settings and provide suggestions on the most effective strategies to
leverage metagenomics for disease diagnosis and pathogen detection.
Results
Metagenomics requires stringent experimental processes and
bioinformatic filtering criteria to accurately detect pathogens
The scale and complexity of metagenomic sequencing data, as well as
the risk of contamination or pathogen misassignment, necessitate
strict experimental and computational protocols to ensure that
detected microbes are truly present. We developed procedures that
greatly reduce the chance of calling false positives by (i) using both
identifying intersample
negative
and positive
controls,
(ii)
contamination, and (iii) developing stringent bioinformatic proce-
dures that prioritize specificity over sensitivity (Fig. 1). Because our
protocols evolved over the course of the study, we outline our
recommendations and the proportion of the 593 total samples
sequenced via metagenomics to which each procedure was applied
(Supplementary Table 1).
Experimentally, we developed procedures to both mitigate the
risk of and identify potential cases of contamination occurring in the
laboratory. First, we extracted plasma samples in batches alongside
non-template controls (i.e., water controls) for 574 (96.8%) samples.
We designed batches to minimize the cases where samples known to
be positive for a particular pathogen, such as Lassa virus (LASV), were
extracted or sequenced with samples known to lack the pathogen.
Before synthesizing cDNA or preparing sequencing libraries, we
added a negative control (i.e., RNA isolated from K562 lymphoblast
cells) and a positive control (i.e., RNA from viral seed stock spiked
into RNA isolated from K562 lymphoblast cells or RNA from a pre-
viously sequenced plasma sample known to contain a specific virus)
for 585 (98.7%) and 509 (85.8%) samples, respectively. At this stage,
we also added sample-specific RNA spike-ins using the External RNA
Controls Consortium (ERCC) sequences for each of 508 (85.7%)
samples, including all samples in batches of 12 or more, increasing
the probability of detecting any downstream cross contamination21.
We sequenced the majority of samples with combinatorial dual
indexes (CDIs), although we used unique dual indexes (UDIs) for the
one batch sequenced on the NovaSeq 6000 system (99 or 16.7% of
samples) to minimize the risk of misclassification due to index
hopping.
Cohorts
mpox
YF
?
Plasma
Blood
e
c
n
e
c
s
e
r
o
u
F
l
Threshold
Copies per reaction (Ct)
560 individuals with:
● clinical suspicion for
LASV
● demographic, clinical,
and co-infection data
109 individuals in 3 clusters
with:
8 individuals with:
●
● acute febrile illness
● related symptoms and
presentation dates
presentations
RT-qPCR pathogen panel
TT
Evolved Sequencing Pipeline from
ranga et al. 2016,
Mat
J. Vis. Exp.
Negative
controls
Positive
control
Samples
Unique RNA
spike-in
Sample
Sequencing
Analysis
Filtering
QC
AGTCCCTGAATAA CGA
Metagenomic sequencing
Fig. 1 | Overview of the study design. We conducted RT-qPCR on 670 plasma
samples, followed by metagenomic sequencing of 593 of the samples, received
from (i) individuals suspected to have Lassa Fever (LF; caused by Lassa virus, LASV),
collected from teaching hospitals with clinical expertise in viral hemorrhagic fevers;
(ii) suspected infectious disease outbreaks, collected by the Nigerian Centre for
Disease Control (NCDC) and other regional clinics; and (iii) individuals with unusual
or nonspecific clinical manifestations from regional clinics. We used a
metagenomic pipeline inspired by Matranga et al.21 with additional negative (i.e.,
water and K562 cells) and positive controls (i.e., K562 cells spiked with known viral
genetic material), as well as External RNA Controls Consortium (ERCC) RNA spike-
ins. We use metagenomics to identify putative causes of Lassa-like illness, to assess
the role of co-infection in LASV outcomes, to determine the relationships between
clinically similar acute illnesses, and to diagnose individuals with nonspecific pre-
sentations. QC, quality control. YF, yellow fever. Created with BioRender.com.
Nature Communications |
(2023) 14:4693
2
Article
https://doi.org/10.1038/s41467-023-40247-4
Table 1 | Samples collected from Nigerian patients with symptoms of Lassa Fever (LF)
Number of
Samples
415
95
25
13
5
LASV RT-qPCR Hospital or Public Health Agency
State
Year
LASV Genomes
Non-LASV Genomes
Positive
Irrua Specialist Teaching Hospital
Edo
2017-18
220 reported in Siddle et al.14 37 reported in
this study
Negative
Positive
Positive
Positive
Irrua Specialist Teaching Hospital
Edo
2018
2 reported in this study
11 reported in this study
Alex Ekwueme Federal University Teaching
Hospital Abakaliki
Ebonyi 2019-20 10 reported in this study
0 reported in this study
Federal Medical Centre
Ondo
2019
3 reported in this study
0 reported in this study
Nigerian Centre for Disease Control
Kebbi
2019
2 reported in this study
0 reported in this study
Amidst a 2017–2018 surge of LF in Nigeria, we generated 220 complete Lassa virus (LASV) genomes from the sequencing of 415 LASV-positive samples from the Irrua Specialist Teaching Hospital
(ISTH) and reported the LASV genomes in Siddle et al.14. Here, we generate non-LASV genomes, as well as additional LASV genomes from 4 other cohorts.
Computationally, we chose universal, strict filtering criteria to
analyze the resulting data. We first discarded samples that displayed
evidence of potential cross-contamination via the ERCC spike-ins (7 of
560 samples; Supplementary Fig. 1A). We then ensured that the
expected viral genomic material was identified in the positive controls
via the metagenomic classification tool Microsoft Premonition22
(Supplementary Table 2). Next, to call a virus present in a sample, we
required it to have (i) at least 5 reads assigned to it by Microsoft Pre-
monition; (ii) a greater percent of reads assigned to it than assigned to
the same species in any (a) extraction-batch-specific non-template
controls, (b) sequence-batch-specific positive controls, excluding the
spiked in viral genomic material, and (c) sequence-batch-specific
negative controls; and (iii) genome assembly of Microsoft Premonition
hits with a threshold of at least 10% of the reference genome size
(Supplementary Data 1, Supplementary Fig. 2). Thus, we combined a
highly sensitive, but less specific, probabilistic classification tool with a
highly specific, but less sensitive contig assembly step to assign
pathogens to samples.
sensitive
We assessed the sensitivity and specificity of our metagenomic
pipeline relative to clinical RT-qPCR testing status by using data from
the cohort of individuals suspected of LF. A positive Lassa virus (LASV)
clinical test was defined as the amplification of either the GPC gene or
the L gene via the commercially available Altona assay23,24. Prior clinical
RT-qPCR status is an imperfect ground truth, as (i) genome degrada-
tion can occur between clinical testing and subsequent sequencing
and (ii) RT-qPCR can yield false negative results for samples containing
highly diverse viruses, such as LASV. Moreover, we expect PCR to be
to target-specific
more
than metagenomics due
amplification25,26. Nevertheless, we found that the Premonition-based
thresholds yielded a sensitivity of 91.7% and a specificity of 91.6%; the
additional requirement of contig assembly reduced sensitivity to 35.4%
but increased specificity to minimally 96.8% (Supplementary Fig. 1B).
The imperfect specificity was attributable to 3 samples that were RT-
qPCR-negative but positive via sequencing. Two of these samples
yielded complete, identical LASV genomes (98% and 99% complete),
while the third sample yielded a partial genome. We extensively
queried these samples and re-tested them via RT-qPCR (Supplemen-
tary Note, Supplementary Fig. 3), ultimately concluding that they were
most likely diagnostic false negatives, a known challenge in LASV
molecular detection27,28. In summary, our metagenomic protocols
demonstrated high specificity for identifying pathogens in a given
sample.
Metagenomics identifies Lassa virus co-infections of prognostic
significance as well as viral etiologies of Lassa-like illness
We first used our metagenomic approach on 560 samples collected
from population-level surveillance of individuals with symptoms con-
sistent with LF, a viral hemorrhagic fever caused by LASV that is
endemic to West African countries. We analyzed 458 RT-qPCR-positive
and 95 RT-qPCR-negative samples to identify viral co-infections of
prognostic significance, uncover viral etiologies of LF-like clinical
syndromes in Nigeria, and characterize LASV diversity. The samples
were collected between 2017 and 2020, span patients seen in 15 of
36 states and the Federal Capital Territory, and include 220 samples
from which we previously reported LASV genomes14 (Table 1).
We analyzed the metagenomics reads for other viral pathogens
present in our LASV-positive samples, using the filters described above
to prioritize specificity over sensitivity. We found that 7.8% (36/458) of
LASV patients had a viral co-infection with at least one of the following
viruses: hepatitis B, hepatovirus A, human blood-associated dicis-
trovirus (HuBDV), human immunodeficiency virus 1 (HIV-1), measles,
parvovirus B-19, pegivirus C, and an unclassified dicistrovirus that we
propose to name human blood-associated dicistrovirus 2 (HuBDV-2)
(Fig. 2a). One sample was multiply co-infected with both hepatitis B
and pegivirus C (Supplementary Data 1). We additionally identified
viruses in 13.7% (13/95) of the RT-qPCR-negative samples, including
LASV as previously discussed, as well as anellovirus, hepatitis B, HIV-1,
and pegivirus C (Fig. 2a). One LASV-negative sample was multiply co-
infected, with anellovirus, LASV (i.e., this sample was the PCR false
negative that produced a partial genome), and pegivirus C.
Because co-infections were common among LASV-positive sam-
ples, we investigated whether they played a role in LASV outcomes. We
analyzed the most frequent co-infections (i.e., pegivirus C, HIV-1, and
clinically diagnosed malaria) alongside demographic information (i.e.,
age, sex, and pregnancy status), clinical covariates (i.e., diagnostic Ct
and ribavirin treatment status), and outcomes (i.e., survived or
deceased) for 400 LASV-positive individuals (Table 2). We conducted
univariate logistic regression and found that diagnostic Ct value
(p < 0.001) and receipt of ribavirin (p = 0.01) were significantly asso-
ciated with outcomes, while age (p = 0.06) and co-infection with
pegivirus C (p = 0.18) trended towards an association (Table 2,
Fig. 2b–d, Supplementary Fig. 4A–E). Meanwhile, malaria co-infections,
which were identified in 101 individuals, were not associated with
outcomes (p = 0.76).
We conducted multivariate analyses with the four variables that
were associated with LASV outcomes at p < 0.25. Prior literature sug-
gests that these variables interact with outcomes and with one another
in complex ways29–33. For example, Ct is a measure of the interplay
between the host immune system and the virus, which may be affected
by age34 or co-infections, but Ct cannot be affected by ribavirin treat-
ment since Ct is measured at the time of diagnosis before treatment is
begun. We developed a causal directed acyclic graph35 (DAG; Fig. 2e),
informed by our univariate analyses and previous work29–33, and con-
ducted multivariable linear and logistic regression. Age and pegivirus
co-infection were significant predictors of Ct (Fig. 2e, Table 3, Sup-
plementary Fig. 4G); however, they were not associated with the out-
come when controlling for Ct (Fig. 2e, Table 3, Supplementary Fig. 4F).
We therefore concluded that the effect of age and of pegivirus co-
infection status on the outcome is mediated by Ct36. We determined
that the average causal mediation effects of age (p = 2 × 10−16) and of
pegivirus co-infection status (p = 0.02) on outcome were significant via
bootstrapping (Supplementary Table 3, Supplementary Fig. 4H, I).
Nature Communications |
(2023) 14:4693
3
Article
a.
https://doi.org/10.1038/s41467-023-40247-4
Positive
0
1
1
4
2
2
1
1
25
R
C
P
s
u
r
i
V
a
s
s
a
L
Negative
1
2
0
3
0
0
0
0
5
Percent of Cases
0.05
0.04
0.03
0.02
0.01
0.00
b.
0.8
l
a
i
r
a
a
M
h
t
i
w
n
o
i
t
r
o
p
o
r
P
0.6
0.4
0.2
0.0
Anelloviridae Hepatitis_B Hepatovirus_A
HuBDV
Assembled Viral Genome
c.
HIV_1
HuBDV_2
Measles Parvovirus_B19 Pegivirus_C
d.
91/135
10/14
I
1
−
V
H
h
t
i
w
n
o
i
t
r
o
p
o
r
P
0.05
0.04
0.03
0.02
0.01
0.00
i
s
u
r
i
v
g
e
P
h
t
i
w
n
o
i
t
r
o
p
o
r
P
2/346
1/54
0.100
0.075
0.050
0.025
24/346
1/54
0.000
Survived
Deceased
Lassa Outcome
Survived
Deceased
Lassa Outcome
Survived
Deceased
Lassa Outcome
e.
Importantly, we confirmed that there was no relationship between
pegivirus C and LASV detection, i.e., due to competition for sequen-
cing reads (Fig. 2a; Supplementary Fig. 4J). Though we cannot exclude
the possibility of unknown or unmeasured confounding variables, we
computed the mediational E-value37, which is the risk ratio that an
unmeasured confounder would need to have with both the dependent
and the independent variable to completely explain away the observed
relationships. Unmeasured confounders with risk ratios of at least 1.77,
1.41, and 2.48 would be needed to fully explain the observed rela-
tionships between Ct and outcome, age and Ct, and pegivirus co-
infection and Ct, respectively. In summary, our analyses suggest that
older individuals have higher viral loads and thus poorer outcomes,
Nature Communications |
(2023) 14:4693
4
Article
https://doi.org/10.1038/s41467-023-40247-4
Fig. 2 | Metagenomics identifies Lassa virus co-infections with prognostic
implications as well as viral etiologies of Lassa-like illness. a Metagenomics
identifies Lassa virus (LASV) and non-LASV pathogens in 553 individuals presenting
with symptoms of Lassa Fever (LF). Percent (color scale) and number (reported in
box) of RT-qPCR-positive (458 samples) or RT-qPCR-negative (95 samples) cases
containing the following non-LASV pathogens, which were each found in at least
one sample: anelloviridae, hepatitis B, hepatovirus A, human immunodeficiency
virus 1 (HIV_1), human blood-associated dicistrovirus (HuBDV), HuBDV-2, measles,
parvovirus B19, and pegivirus C. b–d The proportion of surviving or deceased
LASV-positive individuals who were co-infected with malaria (B), HIV-1 (c), or
pegivirus C (d). e Causal directed acyclic graph of hypothesized relationships
between ribavirin treatment, age, pegivirus C co-infection status, LASV cycle
threshold (Ct) value, and outcomes. Arrows are annotated with adjusted p-values
produced via multivariate linear (age + pegivirus → Ct; p = 0.0007 for age and
p = 0.023 for pegivirus) and logistic (age + Ct + pegivirus + ribavirin → outcome;
p = 1.85 × 10−12 for Ct) regression models. ***p < 0.001. *p < 0.05. n.s. not significant.
Table 2 | Univariate logistic regression models identify predictors of LASV outcomes
Variable
Demographics
Age
Sex
Pregnant
Clinical data
Mean Ct
Ribavirin
Outcome
Co-infections
Malaria
HIV-1
Pegivirus C
No. (%) with Data
Median (IQR) or N (%)
Univariate P-value
Unadjusted OR (95% CI)
380 (95.0%)
398 (99.5%)
94 (57%)
391 (97.8%)
386 (96.5%)
400 (100%)
149 (37.3%)
400 (100%)
400 (100%)
31 (21.8–45)
165 (41.5%) female
4 (4.3%) of females
36.9 (31.5–40.8)
257 (66.6%) treated
346 (86.5%) survived
101 (67.8%)
3 (0.8%)
25 (6.3%)
0.06
0.32
0.36
2.79 × 10−14***
0.01*
NA
0.76
0.34
0.18
1.01 (1.00–1.03)
0.74 (0.40–1.33)
3.00 (0.14–26.45)
0.81 (0.76–0.85)
0.48 (0.27–0.87)
NA
1.21 (0.38–4.60)
3.25 (0.15–34.45)
0.25 (0.01–1.24)
No. (%), number (percent) of cases with available data. IQR interquartile range. OR (95% CI), odds ratio (95% confidence interval). CIs, ORs, and unadjusted p-values generated via univariate logistic
regression. ***p < 0.001. *p < 0.05. NA not applicable.
while those co-infected with pegivirus C have lower viral loads and thus
more favorable outcomes.
Next, we further investigated the genome sequences of several
pathogens identified in the LASV-positive and LASV-negative samples,
beginning with LASV itself, which is highly genetically diverse. Its dis-
tinct viral lineages segregate geographically in Nigeria14, though most
available genome sequences are from the southwestern region. Our
work generated 17 new high-quality (>90% of the genome assembled)
LASV genomes, 15 from PCR-positive cases and two from PCR-negative
cases. We observed phylogenetic clustering of these samples by geo-
graphic origin, consistent with previous descriptions of geographic
structure in LASV diversity in Nigeria (Fig. 3). Most of our genomes,
including those from the PCR-negative samples, were of lineage II, and
clustered according to their sampling site (Irrua in the southwestern
cluster and Ebonyi in the southeastern cluster). Two genomes from
samples obtained in northwestern Nigeria clustered with lineage III
genomes but formed a distinct sub-clade, highlighting the extent of
unsampled diversity in this poorly studied lineage.
We also more closely examined our multiple hepatitis B, HIV-1,
and pegivirus C genomes. All three hepatitis B genomes, from one
LASV-positive and two LASV-negative individuals, were classified as
subtype E, the predominant circulating genotype in Western and
Central Africa38. At least two of the seven HIV-1 genomes, from four
LASV-positive and three LASV-negative samples, were recombinant
(Supplementary Table 4). We constructed a phylogenetic tree with our
28 complete pegivirus C genomes from 23 LASV-positive and five
LASV-negative individuals and the other 130 annotated sequences
available in NCBI GenBank. The Nigerian genomes cluster with other
African genomes, in particular those from Ghana and Cameroon, the
nearest countries represented in the tree (Supplementary Fig. 5).
Finally, we report the first four Nigerian genomes of dicis-
troviruses, all of which were found in LASV-positive samples. Dicis-
troviruses have primarily been described in arthropods39–43, though
the poorly characterized human blood-associated dicistrovirus
(HuBDV) was first discovered in a febrile Peruvian patient in 201844.
Here, we assembled the second complete HuBDV genome and another
partial genome. Moreover, we assembled two additional unclassified
dicistroviridae genomes, which were >96% identical to sequences
produced from febrile Tanzanian children45 and highly divergent from
the HuBDV genomes (Fig. 4). We designate the clade that includes our
two unclassified genomes and the three Tanzanian genomes as human
blood-associated dicistrovirus 2 (HuBDV-2; Fig. 4). Our identification
of unlinked cases of HuBDV and HuBDV-2 suggests that these viruses
may be circulating more broadly than known in Nigeria.
Cluster investigations yield genomic insights that inform public
health interventions
Genome sequencing has successfully identified the etiologies of dis-
ease outbreaks and determined the relationships between cases within
a cluster13,46–48. We investigated three separate outbreaks via the ana-
lysis of 109 plasma samples collected by the NCDC. We tested all
samples using an RT-qPCR-based common pathogens panel (Supple-
mentary Table 5; Supplementary Data 1) and conducted subsequent
metagenomic sequencing on a subset of samples for outbreak
characterization.
The first cluster investigation consisted of 71 samples col-
lected in 2017 from patients suspected to have mpox, caused by
monkeypox virus (MPXV). MPXV re-emerged in Nigeria over the
same calendar year, after 40 years of absence, and sequencing of
early cases suggested spillover from a local reservoir, rather than
importation, as the source49. Here, we conducted diagnostics and
sequencing from plasma samples rather than lesion swabs, which
are heterogeneous samples that can be difficult to collect from
those with few or no visible lesions50. Though plasma is a more
standardized sample type, the degree to which MPXV genetic
material is detectable in plasma is unknown. Of our 71 plasma
samples, 35 were positive for MPXV by qPCR (Supplementary
Table 6), indicating a minimum sensitivity of 49% for plasma
testing (as not all patients were certain to have MPXV). We selected
five MPXV-positive plasma samples—those with the highest
sequencing library quantification values—for unbiased sequencing
as well as hybrid capture with pan-viral target enrichment probes
Nature Communications |
(2023) 14:4693
5
Article
https://doi.org/10.1038/s41467-023-40247-4
(Methods). Unbiased metagenomics yielded 30 or fewer aligned
read pairs for each sample, while hybrid capture yielded up to
20,000 aligned read pairs (Supplementary Fig. 6). We produced
contigs capable of determining that the 5 samples belonged to the
IIb clade (i.e., the clade responsible for the 2022 multinational
outbreak), consistent with other outbreak reports49. We could not
assemble complete genomes via either metagenomics or hybrid
capture, likely due in part to the large genome size, reduced viral
loads in the blood relative to lesions51, and the Illumina MiSeq’s
sequencing capacity.
Table 3 | Multivariate linear and logistic regression models
identify predictors of LASV outcomes
P-value
Independent
variable
Age + Pegivirus → Ct
Regression
coefficient (β)
Standard
error of β
95% CI (β)
Age
0.0007***
−0.066
0.019
Pegivirus
3.314
1.447
0.023*
P-value
Independent
variable
Age + Pegivirus + Ct + Ribavirin → Outcome
Regression
coefficient (β)
Age
Pegivirus
Ct
0.945
0.758
1.85 × 10−12***
Ribavirin
0.336
0.001
−0.334
−0.209
−0.364
−0.10
to (−0.02)
0.48–6.15
95% CI (OR)
0.98–1.02
0.03–4.11
0.76–0.86
0.33–1.47
Odds
ratio (OR)
1.00
0.72
0.81
0.70
95% CI 95% confidence interval, OR odds ratio, CIs, ORs, and p-values generated via multivariate
linear and logistic regression and adjusted for covariates. ***p < 0.001. *p < 0.05. NA not
applicable.
The second cluster investigation consisted of eight samples sus-
pected to contain yellow fever virus (YFV), collected in 2020 from
Ebonyi, Edo, and Oyo states. YFV is the etiological agent of YF and also
re-emerged in Nigeria in 2017 after a 40-year absence52. Previously, we
reported YFV in a 2018 cluster with symptoms suggestive of LF and
demonstrated that the cases were more closely related to con-
temporary Senegalese YFV genomes than to historical Nigerian
sequences53. After confirming YFV was found in all eight samples via
RT-qPCR, we sought to characterize the genomic ancestry of the 2020
outbreak. We produced two complete YFV genomes, which belonged
to the West Africa clade (Supplementary Fig. 7) and were >98% similar
to sequences from the Nigerian 2018 YFV outbreak53, suggesting
cryptic transmission and persistence of the 2018 YFV strain. These data
contributed to the NCDC’s and World Health Organization’s (WHO)
efforts to accelerate vaccination campaigns and train local healthcare
workers in the diagnosis and treatment of YF54.
Finally, we received 30 samples in November 2020 from a cluster
in Benue, Nigeria, that presented with headache, diarrhea, vomiting,
and abdominal pain. The samples were negative for all pathogens in
the RT-qPCR panel, and metagenomic sequencing of 12 samples failed
to identify an infectious etiology. The NCDC ultimately expanded its
differential diagnosis to include environmental causes, and the out-
break was determined to be due to pesticide poisoning55,56. While
metagenomics of a single sample type cannot rule out an infectious
cause, this investigation emphasizes that it can aid public health
departments in updating their prior probabilities of specific diagnoses.
Metagenomics identifies viral infections in undiagnosed, severe
clinical cases
In the clinical setting, metagenomic sequencing offers an alternative to
the enumeration of single-pathogen diagnostic tests, which can
Fig. 3 | Lassa virus genetic diversity. Maximum likelihood phylogenetic tree of 17
new genomes (dark blue) alongside 622 published complete S segment coding
sequences. Tips are colored by the country of sample origin, and the tree is rooted
in the Pinneo sequence (1979). The area highlighted in gray, containing the majority
of the new genomes (10/17), is shown in more detail on the left. The asterisk
denotes the two RT-qPCR-negative samples that yielded complete genomes. The
scale bar denotes substitutions per site. Bootstrap values are shown on key nodes.
Nature Communications |
(2023) 14:4693
6
Article
https://doi.org/10.1038/s41467-023-40247-4
NIGERIA-LASV0352-EDO-2018-HuBDV2
NIGERIA-LASV0380-EDO-2018-HuBDV2
4 8
4 6
Tanzania_Dicistroviridae-3__MH536111.1_
1 0 0
Tanzania_Dicistroviridae-1__MH536109.1_
8 2
9 2
Tanzania_Dicistroviridae-2__MH536110.1_
5 9
Hubei_picorna-like_virus__NC_033227.1_
Human blood-associated dicistrovirus 2
1 9
5 1
7 8
4 3
6 6
9 9
9 7
5 9
8 9
9 3
9 1
9 5
5 9
Acute_bee_paralysis_virus__NC_002548.1_
1 0 0
Israel_acute_paralysis_virus_of_bees__NC_009025.1_
Solenopsis_invicta_virus_1__NC_006559.1_
Cricket_paralysis_virus__NC_003924.1_
Drosophila_C_virus__NC_001834.1_
Anopheles_C_virus__NC_030115.1_
Goose_dicistrovirus__NC_029052.1_
Bat_cripavirus__KX644942.1_
Human_blood-associated_dicistrovirus__KY973643.1_
1 0 0
NIGERIA-LASV0386-EDO-2018-HuBDV
Black_queen_cell_virus__NC_003784.1_
Himetobi_P_virus__NC_003782.1_
Triatoma_virus__NC_003783.1_
Homalodisca_coagulata_virus-1__NC_008029.1_
Aphid_lethal_paralysis_virus__NC_004365.1_
Rhopalosiphum_padi_virus__AF022937.1_
1 0 0
Mud_crab_dicistrovirus__NC_014793.1_
Taura_syndrome_virus__NC_003005.1_
0.2 substitutions per site
Fig. 4 | Dicistrovirus RdRp (RNA-dependent RNA polymerase) genetic diversity.
Maximum likelihood phylogenetic tree with 3 new sequences (green) alongside 21
published sequences. Generated from 2540-bp RdRp gene alignment. Bootstrap
values for key nodes are shown. The clade that we name human blood-associated
dicistrovirus 2 (HuBDV-2) is labeled.
require multiple samples and ultimately be costly and time-
consuming57. Moreover, in Nigeria and other LMIC settings, even
large hospitals currently only have the capacity to test for a small set of
pathogens. We received eight plasma samples from individuals with
clinical presentations consistent with an infectious etiology but with-
out evidence of any commonly circulating pathogens, collected in
2019–2020 from Ondo, Lagos, and Ebonyi states. Clinical and demo-
graphic metrics for these cases were highly varied (Supplementary
Table 7).
We first screened the eight patient samples against the RT-qPCR
common pathogens panel (Supplementary Table 5; Supplementary
Data 1) and failed to identify any positive hits. Via unbiased meta-
genomic sequencing, we identified viruses that are plausible candi-
dates for illness in two patients. In a third sample, we detected
Pegivirus C, a common infection in healthy individuals58 that is unli-
kely to be the cause of the clinical syndrome. No plausible pathogenic
viral taxa were detected in the remaining five samples. Here, we
describe the clinical and genomic features of the cases with a puta-
tive diagnosis.
We identified reads mapping to Enterovirus B in the plasma of a
child presenting with fever and seizures. We assembled a genome of
Coxsackievirus-B3 (CV-B3; Fig. 5a), which is associated with both gas-
trointestinal illness and more serious manifestations, including myo-
carditis and meningitis59,60. The genome was most similar to a CV-B3
genome from Japan (82% pairwise sequence identity), though the VP1
gene was most closely related to a partial genome from Nigeria (88%
pairwise sequence identity to GQ496547.1)61.
We detected type IB hepatovirus A (HAV; Fig. 5b) in another child
presenting with left-sided weakness, generalized lymphadenopathy,
hepatosplenomegaly, and a head CT scan with evidence of a right
hemispheric stroke. HAV, the causal agent of hepatitis A, is transmitted
fecal-orally, typically presents with acute gastrointestinal manifesta-
tions, and rarely causes death62. This patient’s symptoms are not
consistent with the textbook presentation of hepatitis A, though cases
associated with HAV have been
of neurological
documented63–66. We thus interpret the metagenomic sequencing
results with caution, as it is possible that HAV is an incidental finding.
However, we only identified HAV in 1 of our 592 other samples, sug-
gesting that it is an uncommon co-infection and lending support to the
possibility that this patient presented with an unusual manifestation
of HAV.
sequelae
Discussion
Here, we describe a highly specific metagenomic sequencing protocol,
which we use to investigate viral etiologies of fever in Nigeria in three
contexts (Fig. 1). Nigeria’s high infectious disease burden, including
endemic (e.g., malaria), emerging (e.g., Dengue virus) and re-emerging
(e.g., LASV) pathogens, advanced sequencing capacity, and robust
public health system make it a compelling place to study the role of
metagenomics in infectious disease surveillance.
Our genomic investigations uncovered 13 distinct viruses using a
single pipeline, informing public and patient health. Our MPXV inves-
tigation demonstrated the benefit of targeted approaches (e.g., qPCR
and hybrid capture) when a pathogen is suspected while also
Nature Communications |
(2023) 14:4693
7
Article
https://doi.org/10.1038/s41467-023-40247-4
a.
1 0 0
8 6
1 0 0
1 0 0
9 7
6 8
1 0 0
1 0 0
1 0 0
1 0 0
1 0 0
1 0 0
1 0 0
1 0 0
1 0 0
1 0 0
6 8
0.04 substitutions per site
b.
1 0 0
0.4 substitutions per site
1 0 0
1 0 0
1 0 0
1 0 0
9 3
9 8
b0013-Nigeria-201 9
MF678311-Australia-2009
KJ489414-France-1993
MK652138-USA-2018
MG451802-UK-2016
MF678314-Australia-2015
MF678304-Australia-2012
Australia
JX476169-India-2009
KR107057-India-2010
JX476166-India-2010
MZ396301-Nepal-2017
India
China
AY673831-USA-1956
U57056-USA
JN048469-USA
AF231763-USA
AF2317650-USA
AY752946-USA
NC_038307-RefSeq
M88483-Sweden
M16572-Sweden
JN048468-USA
MK012537-USA
AY752944-USA
AY752945-USA
KJ025083-China
KX981987-China-2014
JQ040513-China-2007
AF231764-USA
M33854-Germany
MF678302-Australia-2008
JX312064-USA-2008
MG181943-Brazil-2015
KT229611-China-2013
KT229612-China-2013
KT819575-Uganda-2010
EU140838-India
Japan
AB279733-Japan-1992
AB279734-Japan-1995
JQ655151-SouthKorea-2011
AB279732-Japan-1990
EU011791-India
FJ360735-India-1997
DQ991030-India
FJ360731-India-2007
FJ360732-India-2008
FJ360734-India-1999
FJ360730-India-2006
MN062167-USA-2018
FJ360733-India-1995
DQ991029-India
AY644670-SierraLeone
AY644676-France
Gabon
HQ246217.1-1980
LC128713-Thailand-2000
M20273-USA
NC_001489-RefSeq
M14707-USA
DQ646426-Russia
M16632-USA
KX523680-China-1988
KF569906-China-2012
KP879217-USA
MK829707-USA-2013
KP879216-USA
MT181522-UK-2019
KX035096-USA-2013
LC515201-Gabon-2016
b0009-Nigeria
LASV0198-EDO-2018-Nigeria
KY003229-SouthAfrica-2011
MG546668-Iran-2017
MH577313-USA-2017
MH577314-USA-2018
KX228694-Egypt-2014
MN832785-Ireland-2018
USA
X83302-Italy
MN832786-Ireland-2018
HM769724-Argentina-2006
EU526088-Uruguay
EU526089-Uruguay
KJ427799-Italy-2013
K02990-Belgium
JQ425480-USA
HQ437707-Russia-2007
EU251188-Russia
MN062166-USA-2018
KC182588-Mexico-2009
OK625565-Haiti-2016
MG049743-Brazil-2017
MN062164-USA-2017
MN062165-USA-2018
KC182590-Mexico-2009
KC182587-Mexico-2009
KC182589-Mexico-2009
Asia
Fig. 5 | The genetic diversity of pathogens identified in undiagnosed, severe
clinical cases. a Coxsackievirus B3 (CV-B3) genetic diversity. Maximum likelihood
phylogenetic tree with one new sequence (pink) alongside 63 full-length, published
sequences. Generated from whole-genome alignment (7447 bp). Bootstrap values
for key nodes are shown. b Hepatovirus A genetic diversity. Maximum likelihood
phylogenetic tree with two new sequences (red) alongside 105 full-length, pub-
lished sequences. Generated from whole-genome alignment (7736 bp). Bootstrap
values for key nodes are shown.
Nature Communications |
(2023) 14:4693
8
Article
https://doi.org/10.1038/s41467-023-40247-4
demonstrating that MPXV can be detected and subtyped from plasma
samples. Additionally, we identified the poorly described HuBDVs,
highlighting the need for further research while emphasizing the
importance of metagenomics in detecting uncommon pathogens.
Indeed, the identification of unlinked Nigerian cases of HuBDV-1 and
HuBDV-2, previously identified solely in Peru and Tanzania, respec-
tively, suggests that human dicistrovirus infections may be more
widespread than previously suspected. Meanwhile, HIV and hepatitis B
are major causes of morbidity and mortality, both globally and in
Nigeria67–70, and were found in multiple individuals in our LASV-
negative cohort, representing worthy candidates for follow-up testing
of patients with symptoms of LF. We also uncovered the possible
protective effect of pegivirus C in LASV infection. Over 5% of the LASV-
positive cohort was co-infected with pegivirus C, which is consistent
with its estimated prevalence of 7–12% in healthy West African blood
donors71,72. Our causal mediation analysis suggests that pegivirus C
contributes to beneficial LASV outcomes via the mediation of LASV
viral load. While this finding is consistent with favorable prognostic
reports from hepatitis C, HIV, and Ebola virus patients co-infected with
pegivirus C32,33,58, we emphasize the need for further epidemiological
and mechanistic research.
Our phylogenetic reconstructions also produced actionable
public health insights. Our finding that 2020 YFV cases were descen-
dants of 2018 Nigerian cases indicated the presence of cryptic trans-
mission and prompted the NCDC and the National Primary Health Care
Development Authority (NHPCDA) to accelerate their vaccination
efforts. On the other hand, our pegivirus C genomes were interspersed
with those from Cameroon, Sierra Leone, and Uganda, emphasizing
that transmission patterns in Nigeria are a result of both importation
and internal circulation. Finally, we identified undersampled viral
diversity, both by sequencing LASV samples from Kebbi, which form a
clade within lineage III, and by generating the first complete coxsack-
ievirus B3 genome on the African continent.
Metagenomic sequencing is a powerful diagnostic platform but
requires careful analysis and interpretation. By using multiple experi-
mental controls followed by strict computational thresholds, we
achieved a high specificity for pathogen identification. Nevertheless,
the molecular detection of a pathogen does not establish causality nor
fulfill Koch’s postulates73. This is particularly important in individual
cases, where one cannot rely on statistical enrichment
(e.g.,
case–control comparisons) or pseudo-replication (e.g., cluster inves-
tigations). For example, we found HAV in a child lacking the traditional
hepatitis A presentation, expanding rather than narrowing the differ-
ential. Moreover, we failed to identify a pathogen in some samples.
While some cases were truly negative for an infectious etiology, such as
those from individuals with pesticide poisoning55,56, metagenomic
sensitivity is limited by biological and technical factors. Some patho-
gens are undetectable in specific tissue compartments or disease
stages. Additionally, technical challenges limit the sensitivity of
metagenomic (vs. amplicon-based) sequencing for certain pathogens
or sample types12, particularly with certain technologies (e.g., lower-
throughput sequencing machines, which are more widely available in
LMICs). Finally, non-viral pathogens, which we do not consider here28,
require exploration to fully eliminate microbial etiologies.
Practical barriers currently prohibit the widespread use of meta-
genomics for diagnosis, making it most suitable as a complement and
necessary prerequisite to the development of molecular assays. Rou-
tine surveillance of undiagnosed cases through metagenomics can
highlight the pathogens to prioritize for diagnostic capacity building in
a given cohort. We found metagenomics to be particularly valuable in
cluster investigations, where multiple instances of detection increase
diagnostic certainty, and the resulting genomic data enables the study
of transmission and development of policy measures74,75. For hospi-
talized patients needing a diagnosis, we advocate for a tiered
approach, where point-of-care, multiplexed diagnostics are made
available in clinical settings while academic and public health part-
nerships are established so that negative samples can be rapidly
investigated via unbiased sequencing.
Here, we offer real-time insights into the etiologies of febrile ill-
ness and the genetic diversity of circulating pathogens in Nigeria. As
we move beyond the SARS-CoV-2 pandemic, the genomic infra-
structure established in LMICs20 presents an unprecedented oppor-
tunity to use infectious disease genomics in a thoughtful manner to
maximize the benefit to human health.
Methods
Patient recruitment and ethics statement
We obtained samples through studies reviewed and approved by
institutional review boards (IRB) at multiple sites, including Irrua
(Nigeria), Redeemer’s University
Specialist Teaching Hospital
(Nigeria), Harvard University (Cambridge, Massachusetts), and the
National Health Research Ethics Committee (Nigeria). The specific
cohorts covered by each IRB are described below.
Institutional review boards of ISTH (Irrua, Nigeria), Redeemer’s
University, and Harvard University (Cambridge, Massachusetts)
assessed and approved the study before the start of research activities.
De-identified clinical samples and demographic and clinical data were
collected under (i) a waiver of consent, approved by the ISTH Research
Ethics Committee, or (ii) under the written informed consent of par-
ticipants for participation in a separate study that analyzed human
genetic material. The waiver of consent enables the analysis of
pathogen genomic data and de-identified demographic and clinical
data but not the analysis of human genetic material. For the purposes
of the work in this manuscript, the sample sets are equivalent in terms
of data availability.
ISTH is a federal teaching hospital and LF specialist center located
in an area of high LASV endemicity. ISTH treats hundreds of LF patients
each year and tests thousands of patient samples for LASV, including
from patients presenting to ISTH and from samples sent by doctors
elsewhere in Nigeria. Because ISTH is a National Centre of Excellence
for LF management, suspected patients are referred to the hospital for
management from both private practices and surrounding hospitals
and clinics. Among patients presenting to ISTH, LF was considered as a
possible cause of undiagnosed acute febrile illness in patients with (a)
fever ≥38 °C and no improvement after 2 days of antimalarials or
antibiotics, or (b) fever ≥38 °C with at least one LF-associated symp-
tom: bleeding from mucosal surfaces or injection sites, deafness,
conjunctivitis, facial edema, hypotension, spontaneous abortion, sei-
zures, encephalopathy, or acute kidney injury. Plasma was isolated
from a venous blood draw collected from all suspected cases for
diagnostic testing.
In addition to samples collected at ISTH, ACEGID at Redeemer’s
University received samples from multiple sites suspected of an
infectious etiology. Samples in clinical excess (e.g., samples from
from Federal
individuals with critical, undiagnosed conditions)
Teaching Hospital Abakaliki (FETHA) and Federal Medical Center
(FMC) Owo were received via a study approved by the National Health
Research Ethics Committee (NHREC, Nigeria) under a waiver of con-
sent. Samples from individuals in case clusters were received from the
Nigerian Centre for Disease Control (NCDC). As a regulatory body for
public health in Nigeria, NCDC collects samples, some of which are
sent to ACEGID for rapid sequencing in the context of public health
emergencies. All samples received contained plasma isolated from
venous blood draws.
RNA extraction and screening by qPCR
Prior to RT-qPCR testing, suspected LASV samples were inactivated
with Buffer AVL (Qiagen), and RNA was extracted using the QIAmp
Viral Mini extraction kit (Qiagen). At ISTH, patients meeting the criteria
for suspected LASV were tested using 2 RT-qPCR assays, one targeting
Nature Communications |
(2023) 14:4693
9
Article
https://doi.org/10.1038/s41467-023-40247-4
the GPC gene (RealStar LASV RT-PCR Kit 1.0 CE, Altona Diagnostics,
Hamburg, Germany) and a second targeting the LASV L segment23,27. At
Redeemer’s University, samples suspected of LASV infection were
tested using either the RealStar® Lassa Virus RT-PCR Kit 2.0 targeting
the L and GPC genes in one assay or an in-house assay adopted from
Nikisins et al. 27. Samples not suspected of LASV virus were tested for
YFV, Chikungunya virus (CHKV), West Nile Virus (WNV), Zika virus
(ZIKV), O’nyong-nyong virus (ONNV), Ebola virus (EBOV), Dengue,
flaviviruses, and alphaviruses using an RT-qPCR common pathogens
panel. Primers were adopted from previous work (Supplementary
Table 5)76–80.
Suspected MPXV samples underwent DNA extraction using the
Qiagen DNeasy kit and were tested via qPCR using previously pub-
lished primers81.
Samples that were negative for LASV when tested at ISTH but
which assembled a partial or complete LASV genome were re-tested at
the Broad Institute using a previously published primer set27.
Metagenomic sequencing
Unbiased metagenomic sequencing was performed from extracted
nucleic acids as previously described14. Briefly, we used TurboDNase
treatment to remove DNA from all samples except those positive for
MPXV by qPCR. We synthesized double-stranded cDNA using random
hexamer priming. Sequencing libraries were constructed using the
Nextera XT library preparation kit (Illumina) and sequenced on an
Illumina instrument with 100-bp, paired-end sequencing. For MPXV
samples, we additionally performed targeted enrichment with a pan-
viral probe set targeting 356 viral species as previously described82.
Samples were prepared and sequenced at either the Broad Institute or
Redeemer’s University. Metagenomic sequencing data from LASV-
positive cases collected from ISTH were previously reported14, but the
non-LASV reads were not analyzed.
RNA-based controls, including commercially purchased RNA from
K562 cells (negative control) and RNA from K562 cells, spiked with
Ebola virus RNA (Makona variant; positive control), were added prior
to cDNA synthesis. For one batch each, we used extracted RNA from a
previously sequenced sample known to contain LASV or mumps virus
as a positive control (Supplementary Tables 1 and 2).
Genomic data analysis
Samples with fewer than 1000 total reads were discarded. We also
discarded samples with low ERCC spike-in purities, defined as the
number of reads assigned to the major ERCC spike-in divided by the
total number of reads assigned to any ERCC spike-in. For each ERCC
spike-in, we determined the mean and standard deviation of its purity
scores across samples and batches. Samples with greater than 100
total reads assigned to any ERCC spike-in, for which purity was both
<99% and less than three standard deviations below the mean for that
spike-in, were discarded as previously described83.
metagenomics
We then analyzed the sequencing reads using the Microsoft Pre-
pipeline22
(https://microsoft.com/
monition
premonition) with default settings to assign reads to viral taxa. This
pipeline uses an alignment-based approach (e.g., using k-mers) to map
sequences against a large reference database, rather than filtering out
human reads a priori, coupled with a statistical model to assign
probabilities to the assignment of individual reads to taxonomic levels.
Access to the pipeline is via a web interface, with cloud-based pro-
cessing of sequence datasets on the Microsoft Azure platform, allow-
ing rapid generation and retrieval of results. Viral hits were filtered to
remove those with less than five reads. Samples were required to have
a greater percentage of reads assigned to a particular virus than the
percentage of reads assigned to that virus across all batch-specific
controls. We attempted to assemble complete genomes for all
remaining viral hits. For genome assembly, we used the viral-ngs
pipeline84 (version v2.1.8; https://github.com/broadinstitute/viral-ngs).
For most viruses, we performed reference-based assembly using the
RefSeq genome of each virus (Supplementary Data 1). We performed
de novo assembly with reference-genome-guided refinement84 for the
following genetically diverse viruses: LASV, Enterovirus B, HIV-1 (e.g.,
for samples lacking a sufficiently similar reference genome for
reference-based assembly), and pegivirus C. Hits that assembled a
genome of at least 10% of the reference genome length were retained
for downstream analysis. For segmented viruses, we required 10% of
the full genome length (i.e., the sum of individual segment lengths) to
be assembled. Bacterial and eukaryotic taxa were not considered.
We noticed that >50% of samples with reads mapped to Pegivirus
A also had reads mapped to Pegivirus C. In all such cases, we could not
assemble a Pegivirus A genome; for the majority of the samples, we
assembled a Pegivirus C genome. Therefore, we attempted the
assembly of both Pegivirus A and Pegivirus C for all samples meeting
the reads-based thresholds for Pegivirus A, regardless of any reads
mapping to Pegivirus C. We only assembled Pegivirus C genomes
across all cases. This highlights a fundamental challenge of metage-
nomic classification—that highly related species can be misclassified—
but provides support for our combined approach.
Finally, we manually filtered the results to remove known con-
taminants (e.g., the reverse transcriptase of murine leukemia virus)
and to group distinct taxa that were identified within the same family.
Specific torque teno viruses were grouped with the anelloviridae
family, and the unclassified Tanzanian dicistroviridae sequences were
grouped together with our highly related, unclassified dicistroviridae
sequences and designated HuBDV-2.
Causal mediation analysis
Outcomes data and associated clinical covariates were collected at
ISTH and de-identified by clinicians. Missing data were not imputed,
though, for cases with missing pregnancy data, we assumed that
females <10 years old and >60 years old were not pregnant. We also
assumed that individuals who were not admitted to the hospital sur-
vived and additionally did not receive IV-administered ribavirin. The
analyzed Ct values were the average of the Ct values for the L segment
and S segment for samples tested via a multi-target RT-qPCR test. If
only one Ct value existed, either due to failed amplification of one
target or the use of a single-target RT-qPCR test, the single Ct value was
used instead. We assessed the relationship between each variable and
LASV outcome using univariate logistic regression, generating p-values
and unadjusted odds ratios (Table 2). We decided a priori that any
variable associated with the outcome at p < 0.25 in the univariate ana-
lysis would be included in the multivariate logistic regression models.
We fit the linear and logistic regression models (Table 3) to our
data using the stats package (version 4.1.1) in R (version 4.1.1). The
causal mediation analyses were performed using the Baron & Kenny
framework36 and the mediation package (version 4.5.0; Supplementary
Table 3). Mediational E-values were calculated using the website cre-
ated by Mathur et al.85 with the contrast of interest in the exposure set
as 1 for pegivirus co-infection status and 10 years for age.
Viral genotyping
Viral subtyping was carried out using several pathogen-specific tools
settings: Hepatitis A (https://www.rivm.nl/mpf/
with default
typingtool/hav/), Enterovirus B (https://www.rivm.nl/mpf/typingtool/
enterovirus/), HIV (Stanford University HIV Drug Resistance Database;
https://hivdb.stanford.edu/hivdb/), and Hepatitis B (https://www.
genomedetective.com/app/typingtool/hbv/).
Phylogenetic reconstruction
We constructed maximum likelihood phylogenetic trees for multiple
pathogens. For LASV, we used all sequences with greater than 90%
unambiguous length generated in this work. We downloaded from
NCBI GenBank all available S segment sequences (June 23, 2022).
Nature Communications |
(2023) 14:4693
10
Article
https://doi.org/10.1038/s41467-023-40247-4
Sequences were filtered to retain only those sequences with complete
coding sequences (CDS) from either H. sapiens or M. natalensis hosts.
Due to the poor coverage of the region between the GPC and NP CDS
regions, we extracted and concatenated the two CDS from the S seg-
ment for subsequent analysis and performed a multiple sequence
alignment of the concatenated sequences using MAFFT86. We esti-
mated a maximum-likelihood phylogeny with IQ-TREE v2.0.387,88 using
a general time reversible nucleotide-substitution model with a gamma
distribution of rate variation among sites and 1000 iterations of
ultrafast bootstrapping. We rooted the tree on the Pinneo
sequence (1979).
For the dicistroviruses, we downloaded from NCBI GenBank
21 sequences from multiple species, which we aligned with our 3 study
sequences using MAFFT86. The RdRp gene was extracted using Gen-
eious Prime v2023.0.4 (www.geneious.com). We estimated a
maximum-likelihood phylogeny with IQ-TREE v1.6.1289 with a TVM +
F + G4 nucleotide-substitution model and ultrafast bootstrapping90,91.
For pegivirus C, we downloaded from NCBI GenBank all available
full-length, properly annotated sequences (February 28, 2023;
130 sequences), which we aligned with our 28 study sequences from
individuals suspected of LF using MAFFT86. We estimated a maximum-
likelihood phylogeny with IQ-TREE v1.6.1289 with a GTR + F + I + G4
nucleotide-substitution model and ultrafast bootstrapping90,91.
For YFV, we downloaded from the YFV Phylogenetic Typing Tool92
representative full-length sequences from African countries, which we
aligned with our 2 study sequences using MAFFT86. We estimated a
midpoint-rooted maximum-likelihood phylogeny with IQ-TREE
v1.6.1289 with a GTR + F + I nucleotide-substitution model and ultra-
fast bootstrapping90,91.
For coxsackievirus-B3, we downloaded from NCBI GenBank all
available full-length sequences (March 26, 2023; 63 sequences), which
we aligned with our study sequence using MAFFT86. We estimated a
maximum-likelihood phylogeny with IQ-TREE v1.6.1289 with a GTR +
F + I + G4
ultrafast
bootstrapping90,91.
nucleotide-substitution
model
and
For hepatovirus A, we downloaded from NCBI GenBank all avail-
able full-length sequences (March 26, 2023; 105 sequences), which we
aligned with our 2 study sequences using MAFFT86. We estimated a
maximum-likelihood phylogeny with IQ-TREE v1.6.1289 with a GTR +
F + I + G4
ultrafast
bootstrapping90,91.
nucleotide-substitution
model
and
v15.2.5, mediation v4.5.0, ROCR v1.0–11, stats v4.1.1, and tidyverse
v2.0.0). Information about the Microsoft Premonition metagenomics
pipeline is available at https://microsoft.com/premonition. Individuals
can access the pipeline ahead of its public release by clicking the
“Contact us for availability” button and mentioning this work or by
emailing Simon Frost at [email protected].
References
1.
Jones, K. E. et al. Global trends in emerging infectious diseases.
Nature 451, 990–993 (2008).
2. Gire, S. K. et al. Emerging disease or diagnosis? Science 338,
750–752 (2012).
3. Devi, P. et al. Co-infections as modulators of disease outcome:
minor players or major players? Front. Microbiol. 12, 664386 (2021).
4. Mani, N., Slevin, N. & Hudson, A. What three wise men have to say
about diagnosis. Br. Med. J. 343, d7769 (2011).
5. Crump, J. A. & Kirk, M. D. Estimating the burden of febrile illnesses.
6.
7.
8.
9.
PLoS Negl. Trop. Dis. 9, e0004040 (2015).
Prasad, N., Sharples, K. J., Murdoch, D. R. & Crump, J. A. Community
prevalence of fever and relationship with malaria among infants and
children in low-resource areas. Am. J. Trop. Med. Hyg. 93,
178–180 (2015).
Stoler, J. & Awandare, G. A. Febrile illness diagnostics and the
malaria-industrial complex: a socio-environmental perspective.
BMC Infect. Dis. 16, 683 (2016).
Roddy, P. et al. Quantifying the incidence of severe-febrile-illness
hospital admissions in sub-Saharan Africa. PLoS ONE 14,
e0220371 (2019).
Li, C.-X. et al. Unprecedented genomic diversity of RNA viruses in
arthropods reveals the ancestry of negative-sense RNA viruses. Elife
4, e05378 (2015).
10. Stremlau, M. H. et al. Discovery of novel rhabdoviruses in the blood
of healthy individuals from West Africa. PLoS Negl. Trop. Dis. 9,
e0003631 (2015).
11. Wu, F. et al. Author Correction: a new coronavirus associated with
human respiratory disease in China. Nature 580, E7 (2020).
12. Houldcroft, C. J., Beale, M. A. & Breuer, J. Clinical and biological
insights from viral genome sequencing. Nat. Rev. Microbiol. 15,
183–192 (2017).
13. Chiu, C. Y. & Miller, S. A. Clinical metagenomics. Nat. Rev. Genet.
20, 341–355 (2019).
Trees were visualized with FigTree v1.4.493 and are midpoint-
rooted unless otherwise specified. Large clades with no Nigerian or
study sequences were collapsed and labeled with location information.
14. Siddle, K. J. et al. Genomic analysis of Lassa virus during an increase
in cases in Nigeria in 2018. N. Engl. J. Med. 379, 1745–1753 (2018).
15. Tegally, H. et al. Detection of a SARS-CoV-2 variant of concern in
South Africa. Nature 592, 438–443 (2021).
Reporting summary
Further information on research design is available in the Nature
Portfolio Reporting Summary linked to this article.
16. Tegally, H. et al. Sixteen novel lineages of SARS-CoV-2 in South
Africa. Nat. Med. 27, 440–446 (2021).
17. Wilkinson, E. et al. A year of genomic surveillance reveals how the
Data availability
The raw reads, and complete pathogen genomes generated in this
study have been deposited in the Sequence Read Archive (SRA) and
NCBI GenBank, respectively, under BioProject accession codes
PRJNA824010 and PRJNA436552. Sample metadata (collection date,
state, age, sequencing machine, sequencing batch, etc.), metagenomic
read classification data for all samples and controls, viral genome
assembly data, reference sequence accession numbers, and RT-qPCR
results generated in this study are provided in the Supplementary
Data 1 file.
Code availability
Open source software used in this study is available at https://github.
com/broadinstitute/viral-ngs84 (i.e., pipelines for viral genomic analyses;
v2.1.8) and at https://github.com/bpetros95/lassa-metagenomics94 (i.e.,
code for statistical analyses; developed in R v4.1.1 with packages bda
SARS-CoV-2 pandemic unfolded in Africa. Science 374,
423–431 (2021).
18. Viana, R. et al. Rapid epidemic expansion of the SARS-CoV-2 Omi-
cron variant in southern Africa. Nature 603, 679–686 (2022).
19. Olawoye, I. B. et al. Emergence and spread of two SARS-CoV-2
variants of interest in Nigeria. Nat. Commun. 14, 811 (2023).
20. Belman, S., Saha, S. & Beale, M. A. SARS-CoV-2 genomics as a
springboard for future disease mitigation in LMICs. Nat. Rev.
Microbiol. 20, 3 (2022).
21. Matranga, C. B. et al. Unbiased deep sequencing of RNA viruses
from clinical samples. J. Vis. Exp. https://doi.org/10.3791/
54117 (2016).
22. Microsoft Premonition. Microsoft Research https://www.microsoft.
com/en-us/research/project/project-premonition/ (2016).
23. Sylvanus Okogbenin, D Asogun, Ephraim Ogbaini-Emovon, George
Akpede, Stephan Günther. Establishment of diagnostic tools for
detection of Lassa virus & experience from the Irrua Specialist
Nature Communications |
(2023) 14:4693
11
Article
https://doi.org/10.1038/s41467-023-40247-4
Teaching Hospital, Edo State, Nigeria. https://cepi.net/wp-content/
uploads/2019/01/3.1a-Danny-Asogun_Diagnostics.pdf.
24. Boisen, M. L. et al. Field validation of recombinant antigen immu-
noassays for diagnosis of Lassa fever. Sci. Rep. 8, 5939 (2018).
25. Ferreira, C., Otani, S., Aarestrup, F. M. & Manaia, C. M. Quantitative
PCR versus metagenomics for monitoring antibiotic resistance
genes: balancing high sensitivity and broad coverage. FEMS
Microbes 4, xtad008 (2023).
26. Wang, H. et al. Variations among viruses in influent water and
effluent water at a wastewater plant over one year as assessed by
quantitative PCR and metagenomics. Appl. Environ. Microbiol. 86,
24 (2020).
27. Nikisins, S. et al. International external quality assessment study for
molecular detection of Lassa virus. PLoS Negl. Trop. Dis. 9,
e0003793 (2015).
47. Kafetzopoulou, L. E. et al. Metagenomic sequencing at the epi-
center of the Nigeria 2018 Lassa fever outbreak. Science 363,
74–77 (2019).
48. Adams, G. et al. Viral Lineages in the 2022 RSV surge in the United
States. N. Engl. J. Med. 388, 1335–1337 (2023).
49. Faye, O. et al. Genomic characterisation of human monkeypox virus
in Nigeria. Lancet Infect. Dis. 18, 246 (2018).
50. Thornhill, J. P. et al. Monkeypox virus infection in humans across 16
countries—April-June 2022. N. Engl. J. Med. 387, 679–691 (2022).
51. Mailhe, M. et al. Clinical characteristics of ambulatory and hospi-
talized patients with monkeypox virus infection: an observational
cohort study. Clin. Microbiol. Infect. https://doi.org/10.1016/j.cmi.
2022.08.012 (2022)
52. Barrett, A. D. T. The reemergence of yellow fever. Science 361,
847–848 (2018).
28. Ashcroft, J. W. et al. Pathogens that cause illness clinically indis-
53. Ajogbasile, F. V. et al. Real-time metagenomic analysis of undiag-
tinguishable from Lassa Fever, Nigeria, 2018. Emerg. Infect. Dis. 28,
994–997 (2022).
nosed fever cases unveils a yellow fever outbreak in Edo State,
Nigeria. Sci. Rep. 10, 3180 (2020).
29. Duvignaud, A. et al. Lassa fever outcomes and prognostic factors in
Nigeria (LASCOPE): a prospective cohort study. Lancet Glob. Health
9, e469–e478 (2021).
30. Okokhere, P. et al. Clinical and laboratory predictors of Lassa fever
outcome in a dedicated treatment facility in Nigeria: a retrospective,
observational cohort study. Lancet Infect. Dis. 18, 684–695 (2018).
31. McCormick, J. B. et al. Lassa fever. Effective therapy with ribavirin.
N. Engl. J. Med. 314, 20–26 (1986).
32. Williams, C. F. et al. Persistent GB virus C infection and survival in
HIV-infected men. N. Engl. J. Med. 350, 981–990 (2004).
33. Lauck, M. et al. GB virus C coinfections in West African Ebola
patients. J. Virol. 89, 2425–2429 (2015).
34. Hartley, M.-A. et al. Predicting Ebola severity: a clinical prioritization
score for Ebola virus disease. PLoS Negl. Trop. Dis. 11,
e0005265 (2017).
35. Lipsky, A. M. & Greenland, S. Causal directed acyclic graphs. J. Am.
Med. Assoc. 327, 1083–1084 (2022).
36. Baron, R. M. & Kenny, D. A. The moderator-mediator variable dis-
tinction in social psychological research: conceptual, strategic, and
statistical considerations. J. Pers. Soc. Psychol. 51, 1173–1182 (1986).
37. Smith, L. H. & VanderWeele, T. J. Mediational E-values: approximate
sensitivity analysis for unmeasured mediator-outcome confound-
ing. Epidemiology 30, 835–837 (2019).
38. Who, W. Global hepatitis report 2017. (World Health Organization,
Geneva, 2017).
54. Emily Henderson, B. S. WHO supports Nigeria in responding to the
yellow fever outbreak amidst a global pandemic. News-Medical.net
https://www.news-medical.net/news/20211217/WHO-supports-
Nigeria-in-responding-to-the-yellow-fever-outbreak-amidst-a-
global-pandemic.aspx (2021).
55. Chukwurah, F. Nigeria: Pesticides caused death of nearly 300 villa-
gers | DW | 09.10.2021. (Deutsche Welle (www.dw.com)).
56. Onyedinefu, G. Strange disease in Benue suspected to be chemical
poisoning, not Coronavirus—NCDC. https://businessday.ng/
breaking-news/article/strange-disease-in-benue-suspected-to-be-
chemical-poisoning-not-coronavirus-ncdc-2/.
57. Gu, W. et al. Rapid pathogen detection by metagenomic next-
generation sequencing of infected body fluids. Nat. Med. 27,
115–124 (2021).
58. Yu, Y., Wan, Z., Wang, J.-H., Yang, X. & Zhang, C. Review of human
pegivirus: prevalence, transmission, pathogenesis, and clinical
implication. Virulence 13, 324–341 (2022).
59. Brouwer, L., Moreni, G., Wolthers, K. C. & Pajkrt, D. World-wide pre-
valence and genotype distribution of enteroviruses. Viruses 13 (2021).
60. Tariq, N. & Kyriakopoulos, C. Group B Coxsackie Virus. in StatPearls
(StatPearls Publishing, 2022).
61. Faleye, T. O. C. et al. Non-polio enteroviruses in faeces of children
diagnosed with acute flaccid paralysis in Nigeria. Virol. J. 14,
175 (2017).
62. Hepatitis A. https://www.who.int/news-room/fact-sheets/detail/
39. Bonning, B. C. The Dicistroviridae: an emerging family of inverte-
hepatitis-a.
brate viruses. Virol. Sin. 24, 415 (2009).
40. Culley, A. I., Lang, A. S. & Suttle, C. A. Metagenomic analysis of
coastal RNA virus communities. Science 312, 1795–1798 (2006).
41. Victoria, J. G. et al. Metagenomic analyses of viruses in stool sam-
ples from children with acute flaccid paralysis. J. Virol. 83,
4642–4651 (2009).
42. Cheng, R. L., Li, X. F. & Zhang, C. X. Novel dicistroviruses in an
unexpected wide range of invertebrates. Food Environ. Virol 13,
423–431 (2021).
43. Wamonje, F. O. et al. Viral metagenomics of aphids present in bean
and maize plots on mixed-use farms in Kenya reveals the presence
of three dicistroviruses including a novel Big Sioux River virus-like
dicistrovirus. Virol. J. 14, 188 (2017).
44. Phan, T. G. et al. Sera of Peruvians with fever of unknown origins
include viral nucleic acids from non-vertebrate hosts. Virus Genes
54, 33–40 (2018).
63. Lee, J.-J., Kang, K., Park, J.-M., Kwon, O. & Kim, B.-K. Encephalitis
associated with acute hepatitis a. J. Epilepsy Res. 1, 27–28 (2011).
64. Chonmaitree, P. & Methawasin, K. Transverse myelitis in acute
hepatitis a infection: the rare co-occurrence of hepatology and
neurology. Case Rep. Gastroenterol. 10, 44–49 (2016).
65. Cam, S., Ertem, D., Koroglu, O. A. & Pehlivanoglu, E. Hepatitis A
virus infection presenting with seizures. Pediatr. Infect. Dis. J. 24,
652–653 (2005).
66. Dollberg, S., Hurvitz, H., Reifen, R. M., Navon, P. & Branski, D. Seizures
in the course of hepatitis A. Am. J. Dis. Child. 144, 140–141 (1990).
67. HIV Country Profiles. https://cfs.hivci.org/.
68. Global Viral Hepatitis: Millions of People are Affected. https://www.
cdc.gov/hepatitis/global/index.htm (2021).
69. HIV. https://www.who.int/data/gho/data/themes/hiv-aids.
70. Awofala, A. A. & Ogundele, O. E. HIV epidemiology in Nigeria. Saudi
J. Biol. Sci. 25, 697–703 (2018).
45. Cordey, S. et al. Detection of dicistroviruses RNA in blood of febrile
Tanzanian children. Emerg. Microbes Infect. 8, 613–623 (2019).
46. Gu, W., Miller, S. & Chiu, C. Y. Clinical Metagenomic Next-
Generation Sequencing for Pathogen Detection. Annu. Rev. Pathol.
14, 319–338 (2019).
71. Li, C. et al. GB virus C and HIV-1 RNA load in single virus and co-
infected West African individuals. AIDS 20, 379–386 (2006).
72. Tao, I. et al. Screening of hepatitis G and Epstein-Barr viruses among
voluntary non remunerated blood donors (VNRBD) in Burkina Faso,
West Africa. Mediterr. J. Hematol. Infect. Dis. 5, e2013053 (2013).
Nature Communications |
(2023) 14:4693
12
Article
https://doi.org/10.1038/s41467-023-40247-4
73. Falkow, S. Molecular Koch’s postulates applied to bacterial patho-
genicity—a personal recollection 15 years later. Nat. Rev. Microbiol.
2, 67–72 (2004).
74. Armstrong, G. L. et al. Pathogen Genomics in Public Health. N. Engl.
J. Med. 381, 2569–2580 (2019).
75. Petros, B. A. et al. Multimodal surveillance of SARS-CoV-2 at a
university enables development of a robust outbreak response
framework. Med https://doi.org/10.1016/j.medj.2022.09.
003 (2022).
76. Vina-Rodriguez, A. et al. A novel Pan-Flavivirus detection and
identification assay based on RT-qPCR and microarray. Biomed.
Res. Int. 2017, 4248756 (2017).
77. Domingo, C. et al. Advanced yellow fever virus genome detection in
point-of-care facilities and reference laboratories. J. Clin. Microbiol.
50, 4054–4060 (2012).
78. Faye, O. et al. One-step RT-PCR for detection of Zika virus. J. Clin.
Virol. 43, 96–101 (2008).
79. Esposito, D. L. A. & Fonseca, B. A. L. da. Sensitivity and detection of
chikungunya viral genetic material using several PCR-based
approaches. Rev. Soc. Bras. Med. Trop. 50, 465–469
(2017).
80. Sasmono, R. T. et al. Performance of simplexa dengue molecular
assay compared to conventional and SYBR green RT-PCR for
detection of dengue infection in Indonesia. PLoS ONE https://doi.
org/10.1371/journal.pone.0103815 (2014).
81. Li, Y., Olson, V. A., Laue, T., Laker, M. T. & Damon, I. K. Detection of
monkeypox virus with real-time PCR assays. J. Clin. Virol. 36,
194–203 (2006).
82. Metsky, H. C. et al. Capturing sequence diversity in metagenomes
with comprehensive and scalable probe design. Nat. Biotechnol.
37, 160–168 (2019).
83. Lemieux, J. E. et al. Phylogenetic analysis of SARS-CoV-2 in Boston
highlights the impact of superspreading events. Science 371,
6529 (2021).
84. Park, D. et al. viral-ngs: genomic analysis pipelines for viral
sequencing. https://viral-ngs.readthedocs.io/en/latest/index.
html (2015).
85. Mathur, M. B., Ding, P., Riddell, C. A. & VanderWeele, T. J. Web site
and R package for computing E-values. Epidemiology 29,
e45–e47 (2018).
86. Katoh, K. & Standley, D. M. MAFFT multiple sequence alignment
software version 7: improvements in performance and usability.
Mol. Biol. Evol. 30, 772–780 (2013).
87. Nguyen, L.-T., Schmidt, H. A., von Haeseler, A. & Minh, B. Q. IQ-
TREE: a fast and effective stochastic algorithm for estimating
maximum-likelihood phylogenies. Mol. Biol. Evol. 32,
268–274 (2015).
88. Hoang, D. T., Chernomor, O., von Haeseler, A., Minh, B. Q. & Vinh, L.
S. UFBoot2: improving the ultrafast bootstrap approximation. Mol.
Biol. Evol. 35, 518–522 (2018).
89. Trifinopoulos, J., Nguyen, L.-T., von Haeseler, A. & Minh, B. Q. W-IQ-
TREE: a fast online phylogenetic tool for maximum likelihood ana-
lysis. Nucleic Acids Res. 44, W232–W235 (2016).
90. Minh, B. Q., Nguyen, M. A. T. & von Haeseler, A. Ultrafast approx-
imation for phylogenetic bootstrap. Mol. Biol. Evol. 30,
1188–1195 (2013).
91. Kalyaanamoorthy, S., Minh, B. Q., Wong, T. K. F., von Haeseler, A. &
Jermiin, L. S. ModelFinder: fast model selection for accurate phy-
logenetic estimates. Nat. Methods 14, 587–589 (2017).
92. Vilsker, M. et al. Genome Detective: an automated system for virus
identification from high-throughput sequencing data. Bioinfor-
matics 35, 871–873 (2019).
93. Rambaut, A. FigTree v1.4.4. http://tree.bio.ed.ac.uk/software/
figtree/ (2018).
94. Petros, B. bpetros95/lassa-metagenomics: lassa_metagen-
omics_v1.0. https://doi.org/10.5281/zenodo.8020941 (2023).
Acknowledgements
Access to the Microsoft Premonition metagenomics pipeline was
made freely available for this study for research purposes only. This
work was supported by grants from the National Institute of Allergy
and Infectious Diseases (U54HG007480 and U01AI151812 to C.T.H.
and P.C.S., U19-AI110818 to P.C.S.), the World Bank (ACE019 and
ACE-IMPACT to C.T.H.), and the National Institute of General
Medical Sciences (T32GM007753 and T32GM144273 to B.A.P.).
P.C.S. is an investigator supported by the Howard Hughes Medical
Institute (HHMI). This work is made possible by support from Flu
Lab and a cohort of generous donors through TED’s Audacious
Project, including the ELMA Foundation, MacKenzie Scott, the Skoll
Foundation, and Open Philanthropy. The content is solely the
responsibility of the authors and does not necessarily represent the
official views of the National Institutes of Health.
Author contributions
Conceptualization: K.J.S., P.C.S., and C.T.H. Methodology: J.U.O.,
B.A.P., K.J.S., P.E.E., O.A., S.B.M., P.D.I., A.P., P.N., C.T., J.Q., and
S.F.S. Software: S.D.W.F., E.K.J., A.P., D.P., C.T., K.J.S., B.A.P.,
P.E.O., J.U.O., and P.N. Formal analysis: J.U.O., B.A.P., P.E.O.,
S.B.M., K.J.S., P.N., and C.T. Investigation: J.U.O., B.A.P., P.E.E.,
O.A., P.N., S.B.M., P.D.I., I.O., A.P., O.S.G., J.O.A., E.A.U., A.P.E.,
O.B., M.A., P.N., C.T., J.Q., L.S., N.O., N.A.A., K.O., O.O., C.A., N.A.,
O.A., S.O., P.O.O., O.A.F., I.K., C.I., K.J.S., P.C.S., and C.T.H. Data
curation: J.U.O., P.E.O., B.A.P., K.J.S., and S.B.M. Resources: I.O.,
J.O.A., E.A.U., A.P.E., O.B., M.A., L.S., N.O., N.A.A., K.O., O.O., C.A.,
N.A., O.A., S.O., P.O.O., C.I., C.T.H., D.P., and A.A.L. Writing—ori-
ginal draft: J.U.O., B.A.P., K.J.S., and P.E.O. Writing—review and
editing: all authors; Visualization: B.A.P., K.J.S., J.U.O., and P.E.O.
Supervision: O.A.F., I.K., K.J.S., P.C.S., and C.T.H. Funding: P.C.S.
and C.T.H.
Competing interests
P.C.S. is a co-founder and shareholder of Sherlock Biosciences and
Delve Bio, a Board member and shareholder of Danaher Corpora-
tion, and has filed IP related to genomic sequencing and diagnostic
technologies. S.D.W.F., E.K.J., and A.P. are employees of Microsoft
Corporation. S.D.W.F. is a co-founder of DiosSynVax Ltd. and has
filed IP relating to antiviral vaccine technologies, including candi-
dates for Lassa virus. The remaining authors declare no competing
interests.
Additional information
Supplementary information The online version contains
supplementary material available at
https://doi.org/10.1038/s41467-023-40247-4.
Correspondence and requests for materials should be addressed to
Katherine J. Siddle, Pardis C. Sabeti or Christian T. Happi.
Peer review information Nature Communications thanks Shannon
Bennett, Tommy Tsan-Yuk Lam, and the other anonymous reviewer(s) for
their contribution to the peer review of this work. A peer review file is
available.
Reprints and permissions information is available at
http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
Nature Communications |
(2023) 14:4693
13
Article
https://doi.org/10.1038/s41467-023-40247-4
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as
long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons licence, and indicate if
changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons licence and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this licence, visit http://creativecommons.org/
licenses/by/4.0/.
© The Author(s) 2023
Judith U. Oguzie 1,2,19, Brittany A. Petros
3,4,5,6,19, Paul E. Oluniyi
Parvathy Nair9, Opeoluwa Adewale-Fasoro1,2, Peace Damilola Ifoga1,2, Ikponmwosa Odia10, Andrzej Pastusiak11,
Otitoola Shobi Gbemisola2, John Oke Aiyepada10, Eghosasere Anthonia Uyigue10, Akhilomen Patience Edamhande10,
Osiemi Blessing10, Michael Airende10, Christopher Tomkins-Tinch 3,12, James Qu3, Liam Stenson3,
Stephen F. Schaffner
Nelson Adedosu14, Oluwafemi Ayodeji14, Ahmed A. Liasu14, Sylvanus Okogbenin10, Peter O. Okokhere10, Daniel J. Park
Onikepe A. Folarin 1,2, Isaac Komolafe1,2, Chikwe Ihekweazu15, Simon D. W. Frost11,16, Ethan K. Jackson11,
Katherine J. Siddle 3,17,20
13, Kingsley Ojide13, Onwe Ogah13, Chukwuyem Abejegah14,
3,
1,2,7,19, Samar B. Mehta8, Philomena E. Eromon2,
3, Nicholas Oyejide2, Nnenna A. Ajayi
3,9,12,18,20 & Christian T. Happi
, Pardis C. Sabeti
1,2,10,18,20
1Department of Biological Sciences, Faculty of Natural Sciences, Redeemer’s University, Ede, Osun State, Nigeria. 2African Centre of Excellence for Genomics
of Infectious Diseases (ACEGID), Redeemer’s University, Ede, Osun State, Nigeria. 3Broad Institute of Harvard and MIT, Cambridge, MA, USA. 4Harvard-MIT
Program in Health Sciences and Technology, Cambridge, MA 02139, USA. 5Harvard/MIT MD-PhD Program, Boston, MA 02115, USA. 6Systems, Synthetic, and
Quantitative Biology PhD Program, Department of Systems Biology, Harvard Medical School, Boston, MA 02115, USA. 7Chan Zuckerberg Biohub, San
Francisco, CA, USA. 8Department of Medicine, University of Maryland Medical Center, Baltimore, MA, USA. 9Howard Hughes Medical Institute, Chevy Chase,
MD, USA. 10Irrua Specialist Teaching Hospital, Irrua, Edo State, Nigeria. 11Microsoft Premonition, Redmond, WA, USA. 12Department of Organismic and
Evolutionary Biology, Harvard University, Cambridge, MA, USA. 13Alex Ekwueme Federal University Teaching Hospital, Abakaliki, Nigeria. 14Federal Medical
Center, Owo, Ondo State, Nigeria. 15Nigeria Center for Disease Control, Abuja, Nigeria. 16London School of Hygiene and Tropical Medicine, London, UK.
17Department of Molecular Microbiology and Immunology, Brown University, Providence, RI, USA. 18Department of Immunology and Infectious Diseases,
Harvard T.H. Chan School of Public Health, Harvard University, Boston, MA, USA. 19These authors contributed equally: Judith U. Oguzie, Brittany A. Petros, Paul
E. Oluniyi. 20These authors jointly supervised this work: Katherine J. Siddle, Pardis C. Sabeti, Christian T. Happi.
[email protected]; [email protected]
e-mail: [email protected];
Nature Communications |
(2023) 14:4693
14
|
Could not heal snippet
|
10.1016_j.jbc.2023.105075.pdf
|
Data availability
All data described within the article are contained in the
document. The mass spectrometry proteomics data have been
deposited to the ProteomeXchange Consortium via the PRIDE
(73) partner repository with the dataset identifier PXD042589.
Any further information and requests for resources and re-
agents should be directed to and will be fulfilled by the Lead
Contact, Vamsi K. Mootha ([email protected]).
| null |
RESEARCH ARTICLE
Lipoylation is dependent on the ferredoxin FDX1 and
dispensable under hypoxia in human cells
Received for publication, December 22, 2022, and in revised form, June 23, 2023 Published, Papers in Press, July 20, 2023,
https://doi.org/10.1016/j.jbc.2023.105075
Pallavi R. Joshi1,2,3, Shayan Sadre1,2,3
From the 1Broad Institute, Cambridge, Massachusetts, USA; 2Department of Molecular Biology, Howard Hughes Medical Institute,
Massachusetts General Hospital, Boston, Massachusetts, USA; 3Department of Systems Biology, Harvard Medical School, Boston,
Massachusetts, USA
, Xiaoyan A. Guo1,2,3, Jason G. McCoy1,2,3, and Vamsi K. Mootha1,2,3,*
Reviewed by members of the JBC Editorial Board. Edited by George M. Carman
Iron–sulfur clusters
(ISC) are essential cofactors
that
participate in electron transfer, environmental sensing, and
catalysis. Amongst the most ancient ISC-containing proteins
are the ferredoxin (FDX) family of electron carriers. Humans
have two FDXs- FDX1 and FDX2, both of which are localized
to mitochondria, and the latter of which is itself important for
ISC synthesis. We have previously shown that hypoxia can
eliminate the requirement for some components of the ISC
biosynthetic pathway, but FDXs were not included in that
study. Here, we report that FDX1, but not FDX2, is dispensable
under 1% O2 in cultured human cells. We find that FDX1 is
essential for production of the lipoic acid cofactor, which is
synthesized by the ISC-containing enzyme lipoyl synthase.
While hypoxia can rescue the growth phenotype of either
FDX1 or lipoyl synthase KO cells, lipoylation in these same
cells is not rescued, arguing against an alternative biosynthetic
route or salvage pathway for lipoate in hypoxia. Our work re-
veals the divergent roles of FDX1 and FDX2 in mitochondria,
identifies a role for FDX1 in lipoate synthesis, and suggests that
loss of lipoic acid can be tolerated under low oxygen tensions in
cell culture.
Iron–sulfur clusters (ISCs) are ancient cofactors believed to
have first formed in primordial oceans under anaerobic condi-
tions (1, 2). Common forms of ISCs are the 2Fe–2S and 4Fe–4S
clusters, which can perform one-electron transfer reactions,
catalyze dehydration reactions, activate aliphatic substrates, and
stabilize proteins (1–3). In most eukaryotes, ISCs are assembled
via the mitochondrial ISC pathway, which begins with the
synthesis of a 2Fe–2S followed by incorporation into ISC pro-
teins or reductive coupling with another cluster to form a 4Fe–
4S cluster (4–6). ISC synthesis in mitochondria is initiated by
loading a ferrous (Fe2+) iron onto the scaffold protein ISCU (4,
7). ISCU associates with the cysteine desulfurase NFS1 and its
cofactors that together catalyze the conversion of a cysteine to
an alanine and transfer a sulfur group in the form of a persulfide
to ISCU in the presence of frataxin (FXN) (4). For sulfur release
from the persulfide to make a 1Fe–1S cluster, the input of two
electrons is required (4). One of these electrons is donated by the
* For correspondence: Vamsi K. Mootha, [email protected].
iron itself and the other is donated by the electron carrier
ferredoxin 2 (FDX2) (4, 8). To achieve a final 2Fe–2S product,
ISCU is believed to dimerize (4, 5).
To date, over 60 ISC-containing human proteins have been
discovered that localize to the nucleus, cytosol, or mitochon-
dria (7, 9). In the cytosol and nucleus, ISC proteins participate
in reactions such as the breakdown of xanthine (XDH) and
synthesis of the molybdenum cofactor (MOCS1A), or DNA
maintenance (MUTYH, NTHL1) and replication (POLD1,
POLE1) (10–12). In mitochondria, ISC-containing proteins
include subunits of complexes I (CI), II (CII), and III (CIII) of
oxidative phosphorylation (OXPHOS), lipoyl synthase (LIAS),
as well as FDXs (3, 5, 13).
FDXs are versatile single electron carriers found in all do-
mains of life (14, 15). Because the midpoint potential of their
ISC is finely tuned by the local protein environment and sol-
vent exposure, organisms can harbor multiple FDXs that
simultaneously function in distinct cellular reactions (14, 16).
Humans and other chordates have two FDXs (FDX1 and
FDX2), both of which contain a 2Fe–2S cluster (17), localize to
the mitochondrial matrix (17), and are reported to receive
electrons from a mitochondrial NADPH-dependent ferredoxin
(17–19). Foundational studies ascribed
reductase (FDXR)
specific roles for the two FDXs (Fig. 1A), with FDX1 func-
tioning primarily in sterol synthesis pathways by donating
electrons to various cytochrome P450 proteins such as
CYP11A1 (19, 20), and FDX2 functioning in the more ancient
role of electron donation to the ISC machinery as well as the
synthesis of heme A, which is required for complex IV (CIV) of
the electron transport chain (8, 17, 21, 22) (Fig. 1A). Whether
FDXs contribute to other mitochondrial pathways remains an
open question (Fig. 1A).
Previous work from our laboratory has demonstrated in
yeast, worms, and human cells that loss of the ISC machinery
protein FXN, which is mutated in Friedreich’s ataxia, can be
buffered by hypoxia (23). In this same study, we found that
much of the brain pathology of an shFXN mouse model of
Friedreich’s ataxia could also be prevented when mice are
breathing 11% oxygen (23). In addition, we discovered that
core ISC biosynthetic machinery (such as NFS1 and ISCU) is
always essential, but the electron donor systems, FDX2 and
FDXR, were not tested (23). A subsequent low/high oxygen
J. Biol. Chem. (2023) 299(9) 105075 1
© 2023 THE AUTHORS. Published by Elsevier Inc on behalf of American Society for Biochemistry and Molecular Biology. This is an open access article under the CC
BY license (http://creativecommons.org/licenses/by/4.0/).
Lipoylation is dispensable under hypoxia
Figure 1. FDX1, but not FDX2, is dispensable for growth in hypoxic conditions. A, current model of human FDXs and their main functions. B, three-day
proliferation assay of K562 cells edited with control (CTRL), FDXR, FDX1, or FDX2 CRISPR guides. Cells were grown in 21% O2, 1% O2, or treated with 75 μM of
HIF-activator FG-4592 in 21% O2. C, three-day proliferation assay of HepG2 cells edited with control, FDX2, or FDX1 guides and grown in 21% or 1% O2. D,
immunoblots for FDXR, FDX1, FDX2, select OXPHOS subunits, lipoylated PDH and KGDH, and control proteins ACTIN and HSP60 on lysates of edited
K562 cells used for proliferation assay. E, immunoblots for FDX1, FDX2, select OXPHOS subunits, lipoylated PDH and KGDH, and control proteins ACTIN and
TOM20 on lysates of edited HepG2 cells used for proliferation assay. All bar plots show mean ± SD of three independent experiments. ns = p > 0.05, *p ≤
0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001. Two-way ANOVA with Bonferroni’s post-test. FDX, ferredoxin; HIF, hypoxia-inducible factor; KGDH, α-keto-
glutarate dehydrogenase; OXPHOS, oxidative phosphorylation; PDH, pyruvate dehydrogenase.
CRISPR screen from our laboratory broadened the spectrum
of mitochondrial proteins, which, like FXN, may be dispens-
able under low oxygen tensions (24). The screening results
confirmed the dispensability of FXN, but no other core ISC
components scored, including FDX2 or FDXR (24). Curiously,
the screen did uncover FDX1 as potentially dispensable in low
O2 (24). An earlier genome-wide CRISPR galactose death
screen designed to discover proteins required for oxidative
phosphorylation identified both FDX1 and FDXR but not
FDX2 (25). A recent report implicated FDX1 as an upstream
regulator of LIAS (26), which generates the lipoic acid cofactor
required for many tricarboxylic acid (TCA) cycle enzymes
including pyruvate dehydrogenase (PDH) (27–29).
The combined evidence from these studies motivated us to
investigate the roles of FDX1, FDX2, and FDXR in ISC syn-
thesis and lipoate metabolism, as well as to evaluate their re-
quirements under hypoxia. We also find that FDX1 is required
for lipoate synthesis, and in addition, report that FDX1, but
2 J. Biol. Chem. (2023) 299(9) 105075
not FDX2, is dispensable under low oxygen tensions. Sur-
prisingly, lipoate levels are not rescued in either FDX1 or LIAS
KO cells under hypoxia. Hence, the loss of lipoate appears to
be tolerated in low oxygen tensions in cultured human cells.
Results
FDX1, but not FDX2, is dispensable for growth in hypoxia
We began by testing whether FDXR, FDX1, or FDX2 are
dispensable in low oxygen. While our prior study showed that
FXN was unique amongst tested ISC assembly machinery
components, we did not assess FDXR and FDX2 at that time
(23). Although these two proteins did not score in our previous
low/high O2 screen, FDX1 did (ranking 96 amongst 20,113
targeted genes) (24). We thus utilized CRISPR/Cas9-mediated
gene editing to knockout FDX1, FDX2, or FDXR in K562 cells.
At 21% O2 (normoxia), the cells grew with varying growth
defects (Fig. 1B). However, when cells were grown continu-
ously in 1% O2 (hypoxia), FDX1 and FDXR KO growth rate
was similar to the control KO cells, although FDXR KO cells
still showed a mild deficit (Fig. 1B).
A natural question is whether the rescue by hypoxia is
mediated by the hypoxia-inducible factor (HIF) transcriptional
response, which is activated by low oxygen and upregulates
numerous pathways required for cell survival under hypoxia
(30). In normoxia, HIF is hydroxylated by the prolyl hydrox-
ylase enzymes (prolyl hydroxylase domain) and subsequently
ubiquitinated by VHL, which targets the protein for protea-
somal degradation (30, 31). We previously showed that HIF
activation, unlike hypoxia, was not sufficient to rescue FXN KO
cell growth (23). We therefore treated our control and KO
cells with the prolyl hydroxylase domain inhibitor FG-4592 in
normoxic conditions (Fig. 1B) (32). Expression of the canonical
HIF target BNIP3L was indeed increased by this drug as
confirmed by quantitative PCR (qPCR) (Fig. S1) (33), but the
drug was not sufficient to ameliorate the growth defects of our
FDXR, FDX1, and FDX2 KO K562 cell lines (Fig. 1B).
We could extend this observation to a second cell type,
HepG2, which unlike K562 cells in our hands did not exhibit a
baseline growth defect in 1% O2. We repeated our growth
assay experiments with control, FDX1, and FDX2 KO cells in
21% and 1% oxygen tensions and once again found that FDX1,
but not FDX2, was dispensable under low oxygen tensions,
whereas FDXR KO growth was also rescued in this cell line
(Figs. 1C and S2). HepG2 cells lacking FDX2 were not viable in
normoxia, as could be seen by the global depletion of proteins
and large aggregates visible by Ponceau staining of nitrocel-
lulose membranes (Fig. S3); in hypoxia, these phenotypes were
milder and FDX2 loss was more tolerable, although the cells
still grew significantly slower than controls (Figs. 1C and S3).
these studies demonstrate that FDX1 is
dispensable in two different cell types when grown in hypoxic
conditions, and that this effect cannot be recapitulated by
forced stabilization of HIF in K562 cells. In contrast, FDX2 KO
cell phenotypes were in-line with previous findings that
knockouts of core ISC biosynthetic genes NFS1, ISCU, and
LYRM4 are not rescued by hypoxia (23).
Collectively,
Lipoylation is dispensable under hypoxia
Contrary to growth, lipoate depletion in FDX1 KO cells is not
rescued by hypoxia
Given that hypoxia rescues FXN deficiency in an HIF-
independent manner by restoring ISC levels (23), we asked if
hypoxia can also restore the biochemical defects in FDX1 KOs
by a similar mechanism. We performed immunoblotting on cell
lysates and sought
to determine whether the knockouts
exhibited biochemical phenotypes of ISC deficiency within the
mitochondria, such as the loss of ISC-containing respiratory
chain complexes CI and CII, or a reduction in lipoate synthesis
because of loss of the ISC-containing LIAS protein, and whether
these defects could be restored under low O2 (5, 7, 23, 34, 35).
Knockout of FDX1 or FDX2 in either cell line resulted in
loss of CI, CII, and CIV to varying degrees (Fig. 1, D and E).
Surprisingly, while FDX2 purportedly has a role in the syn-
thesis of heme A, the cofactor found in CIV (17, 21), FDX1 loss
caused a greater depletion of CIV, which was rescued in
HepG2 but not K562 cells under hypoxia (Fig. 1, D and E). A
recent study published during the preparation of this report
confirmed that FDX1 plays a role in heme A synthesis,
although the exact degree of FDX2 involvement remains un-
clear (36). Additional deficiencies in CI and CII were also seen
with FDX1 and FDX2 loss and rescued under hypoxia in FDX1
(and not FDX2) KO HepG2 cells (Fig. 1, D and E). The addi-
tion of FG-4592 to KO K562 cells did not rescue any of these
defects seen in 21% O2 (Fig. 1D).
We next performed immunoblotting for lipoate. Four en-
zymes in the TCA cycle (PDH, α-ketoglutarate dehydrogenase
[KGDH], branched-chain α-ketoacid dehydrogenase, and 2-
oxoadipate dehydrogenase) as well as the glycine cleavage
system H protein use lipoic acid as a cofactor (27), and
simultaneous immunoblotting for lipoylation of PDH and
KGDH can be used as a readout of lipoate steady-state levels
(37, 38). FXN KO cells have reduced lipoylation of PDH and
KGDH, and this deficit was restored under hypoxia because of
the restoration of ISC availability (23). Recent reports have
indicated that loss of FDX1 leads to near complete ablation of
PDH and KGDH lipoylation (26, 36). We confirmed these
results in both K562 and HepG2 cells (Fig. 1, D and E). We in
addition found in our knockouts that lipoate levels were more
depleted in FDX1 KO cells compared with the FDXR or FDX2
KO samples (Fig. 1, D and E). However, in contrast to previous
observations with FXN KO cells, we found that lipoylation was
not rescued under low oxygen tensions in FDX1 KO cells
(Fig. 1, D and E) (23). These results indicate that the lip-
oylation is neither restored by hypoxia nor required for the
growth of FDX1 KO cells in hypoxia.
Proteomic profiles of FXN, FDXR, FDX1/2, and LIAS KO cells in
normoxia and hypoxia
We next sought to use a more global approach to get a sense
of how FDX1 and FDX2 differentially affect protein expression
in normoxia and hypoxia. We performed quantitative prote-
omics in HepG2 cells grown in normoxia and hypoxia to (i) gain
a broad and systematic understanding of the downstream
cellular changes induced by loss of either FDX, (ii) to explore the
J. Biol. Chem. (2023) 299(9) 105075 3
Lipoylation is dispensable under hypoxia
relationship between FDX1 and the lipoate synthesis pathway,
and (iii) to define the proteomic responses to hypoxia. Alongside
control, FDX1, FDX2, and FDXR KO cells, we also chose to
study FXN KO cells to compare our FDX datasets with cells
suffering a deficiency of ISC synthesis that can be rescued by
hypoxia, as well as LIAS KO cells to compare with cells affected
by a defect in the lipoate synthesis pathway. The KO cells were
confirmed by immunoblot analysis (Fig. S4) (28, 29).
In total, we could quantify the abundance level of 7692
proteins in FDX1, FDX2, FDXR, LIAS, FXN, and control KO
HepG2 samples in duplicate across the two oxygen tensions
(Table S1). Principal component analysis revealed strong
separation of samples by oxygen tension (principal component
1, explaining 42% of variance) (Fig. 2A). When we focused this
analysis only on the normoxic samples, we found that LIAS
and FDX1 KO samples clustered closely together (along with
FXN and FDXR KOs), whereas FDX2 KO samples segregated
from the rest (Fig. 2A).
FDX1 and FDX2 KO exhibit distinct proteomic profiles
Our analyses suggest very different proteomic responses to
FDX1 versus FDX2 loss. We considered the impact of these
knockouts on all mitochondrial pathways, using an inventory
of 149 MitoPathways from MitoCarta3.0 (39). We plotted the
cumulative distribution of log fold changes caused by each KO
Figure 2. Proteomics of HepG2 KO cells highlights divergent roles for FDX1 and FDX2. A, principal component analysis (PCA) of 7692 proteins detected
in duplicate HepG2 cell samples edited with control (CTRL), FXN, FDXR, FDX1, FDX2, or LIAS guides grown in 21% (normoxia) or 1% O2 (hypoxia). Principal
components calculated for all samples together or normoxic samples separately. B, cumulative distribution functions of 149 MitoCarta MitoPathways in
FDX1 or FDX2 KO samples compared with controls in normoxia. Labeled are those pathways achieving a false discovery rate (FDR) <0.0002. C, volcano plots
highlighting log2 fold changes and corresponding FDR for all proteins in FDX1 and FDX2 KO samples compared with controls in normoxia. Complex IV
subunits are shown in blue; gray horizontal line denotes FDR = 0.01 D, distribution of log2 fold changes for all proteins in FDX1, FDX2, and LIAS KO samples
compared with controls in normoxia. E, volcano plots depicting log2 fold changes and corresponding FDR for all proteins in all KO samples compared with
controls in normoxia. Selected proteins with significant FDR are highlighted in blue. Gray horizontal line denotes FDR = 0.01. FDX, ferredoxin.
4 J. Biol. Chem. (2023) 299(9) 105075
when compared with the control KO in normoxia for all
proteins as well as for those in each mitochondrial pathway,
where pathways with a false discovery rate (FDR) <0.0002 are
colored (Fig. 2B). Of note,
larger proportions of proteins
involved in mitochondrial pathways were downregulated
compared with the overall change in protein expression level
in FDX1 but not FDX2 KO cells (Fig. 2B).
CI subunits and ISC-containing proteins were amongst the
most significantly downregulated pathways in both knockouts,
whereas fatty acid oxidation and vitamin metabolism were
enriched specifically with FDX2 loss (Fig. 2B). Consistent with
our immunoblot analysis, we also saw CIV as a significantly
depleted pathway in FDX1 but not FDX2 KO samples
(Fig. 2B).
To further understand whether the downregulation of the
CIV pathway in FDX1 KO cells was attributable to a few
proteins with dramatic negative changes in their expression
levels, or to a general depletion of all CIV-associated peptides,
we examined both log fold change and significance of differ-
ential protein expression in FDX1 and FDX2 KO samples
compared with controls in normoxia and focused on the dis-
tribution of proteins annotated in the CIV pathway (Fig. 2C).
We saw a clear significant depletion of almost all detected
CIV-associated proteins (shown in blue) in FDX1 but not
FDX2 KO samples (Fig. 2C). This result agrees with our
immunoblot studies and recent reports that implicate FDX1 in
heme A synthesis (36). Our proteomics observations further
solidify the idea that the heme A synthesis pathway has been
misannotated and is likely attributable to FDX1 not FDX2.
Although FDX1 and FDX2 are sequence paralogs (17), the
proteomic and immunoblot analyses suggest divergent func-
tion. We therefore tested if gentle overexpression of FDX2
could functionally complement lipoate deficiency in FDX1 KO
cells. We analyzed growth and lipoate production of control
and FDX1 KO cells that were simultaneously overexpressing
GFP, FDX2, or a guide-resistant FDX1 complementary DNA
(cDNA) (Fig. S5, A and B). We confirmed localization of these
constructs by analyzing protein collected from whole cell and
mitochondrial lysates (Fig. S5C). We found that only when the
KO cells overexpressed guide-resistant FDX1 did they recover
growth or lipoate production (Fig. S5, A and B). We conclude
that the two FDXs cannot substitute for one another in lipoate
synthesis, consistent with a recent study reporting that only
FDX1 donates electrons for LIAS SAM catalysis (36).
FDX1 KO and LIAS KO display similar proteomic profiles
Because we saw closer clustering of FDX1 and LIAS KO
samples to each other in normoxia compared with other
knockouts, we visualized the log fold changes of proteins in
FDX1, FDX2, and LIAS KO samples compared with controls in
normoxia (Fig. 2D). The scatter plots revealed a linear corre-
lation between FDX1 and LIAS KO samples and not between
the FDX1 and FDX2 KO samples (Fig. 2D). These results
indicate that LIAS and FDX1 loss result in similar changes
across the proteome, whereas the consequences of FDX1 loss
and FDX2 loss are clearly more distinct.
Lipoylation is dispensable under hypoxia
In addition, we analyzed the differential protein expression
across all knockouts in normoxia and labeled those proteins
with the most significant changes in each knockout (Fig. 2E).
We noted that the spread of the significance level of protein
differential expressions provided a broad view of how loss of
each protein was affecting the HepG2 cellular proteome, with
FXN and FDXR ablation clearly having few significant effects,
FDX1 and LIAS loss producing similar ranges of significance
to each other, and FDX2 loss causing the most dramatic
“eruption” in number of proteins differentially expressed with
high significance (Fig. 2E).
It
The lack of significant proteomic responses to FDXR loss
was surprising.
is unknown whether this arises from
incomplete ablation of the protein because FDXR has many
isoforms (40), though our proteomics analysis reveals that the
target protein is strongly depleted. Alternatively, an unknown
oxidoreductase could compensate for FDXR loss. Other
striking and notable changes include the strong increase in
GDF15 (a marker of the integrated stress response) (41, 42)
with LIAS deletion and the significant depletion in sulfite ox-
idase abundance with loss of FDX1 (Fig. 2E) (43).
LIAS requires both ISC-binding sites for stability
Our proteomics observations that FDX1 loss phenocopies
LIAS loss led us to initially suspect that the mechanism of
lipoate depletion in FDX1 KO cells could be due to destabi-
lization of LIAS. However, our proteomics indicated that LIAS
abundance actually rises in FDX1 KO cells, consistent with the
protein being stabilized by FDX1 loss (Fig. 2D). LIAS is a
radical SAM enzyme with two 4Fe–4S clusters (reducing and
auxiliary) (29, 44, 45). Protein modeling and sequence align-
ment studies from our laboratory and others indicate FDX1
(but not FDX2) interacts with LIAS at the site of the reducing
ISC (Figs. 3A and S6) (36). FDX1 donates electrons to this
reducing cluster (36), creating a radical species that activates
the octanoate precursor that then abstracts sulfur atoms from
the auxiliary cluster to form the mature lipoate (29, 36, 45, 46).
Prior studies have shown that the regeneration of the auxiliary
cluster is important for the stability of LIAS (29, 44, 45), and
therefore, cells with defective mitochondrial ISC synthesis or
trafficking have low LIAS levels (34, 35, 38). We evaluated the
impact of mutating one or both ISC-binding sites on LIAS
protein stability by overexpressing different mutant constructs
of the protein in WT K562 cells. We engineered and provided
cells with cDNA of either a (i) GFP control, (ii) WT LIAS, (iii)
auxiliary ISC-binding site mutant (aux C→A), (iv) reducing
ISC-binding site mutant (red C→A), (v) or a double mutant of
both reducing and auxiliary ISC-binding sites (aux C→A, red
C→A) (Fig. 3B). Expression of these constructs was well
tolerated (Fig. S7), and immunoblot analysis on whole cell and
mitochondrial lysates indicated that both mutants resulted in
decreased LIAS stability, though the auxiliary site was more
critical. Mutation of both sites resulted in an almost complete
loss of the protein (Fig. 3B). Because loss of FDX1 did not
result in loss of LIAS protein (Fig. 2D), it is unlikely that ISCs
on LIAS are destabilized by FDX1 loss.
J. Biol. Chem. (2023) 299(9) 105075 5
Lipoylation is dispensable under hypoxia
Figure 3. FDX1 loss stabilizes LIAS in an ISC-depleted cell. A, top, a multiple sequence alignment of FDX1 and FDX2 homologs from several eukaryotic
organisms, highlighting key FDX1 residues in green that are absent in FDX2 (N.B., nomenclature for FDX1 and FDX2 is inverted in Drosophila melanogaster
compared with other eukaryotes). Bottom, the top ranked interaction model from AlphaFold analysis of LIAS (pink, surface) and FDX1 (blue and green,
cartoon). FDX2 (gray, cartoon) has been structurally aligned with FDX1. The green residues in FDX1 that are divergent in FDX2 form part of the interface with
LIAS. B, immunoblots for FLAG, LIAS, and control proteins HSP60 and ACTIN on whole cell lysate (W.C.L) and isolated mitochondrial (MITO) lysates of
K562 cells overexpressing (O/E) GFP or four different LIAS constructs with 1× FLAG tags on the C-terminal end. Constructs expressed were either WT LIAS or
LIAS with cysteine to alanine mutations in either the auxiliary cluster site (AUX C-A FLAG), the reducing cluster site (RED C-A FLAG), or both the auxiliary and
reducing cluster site (AUX RED C-A FLAG). C, immunoblots for FXN, FDX1, LIAS, FECH, POLD1, lipoylated PDH and KGDH, and control proteins ACTIN and
TOM20 on lysates from K562 cells edited with control (CTRL) or FXN guides on the background of prior editing with control (CTRL) or FDX1 guides. D,
proposed model of LIAS turnover in the absence of FDX1. If turnover of the LIAS enzyme is halted by eliminating FDX1, then LIAS is no longer dependent on
the ISC pool for its stability. Double asterisk indicates band of interest. FDX, ferredoxin; ISC, iron–sulfur cluster; LIAS, lipoyl synthase.
FDX1 and FXN loss have opposing effects on LIAS stability
During each catalytic cycle, the auxiliary cluster on LIAS
loses two sulfurs that must be reloaded, and therefore, LIAS is
dependent on an adequate supply of ISCs (29, 34, 38, 45). We
were curious if this dynamic was affected by FDX1 loss. We
edited K562 cells with control or FDX1 guides and then
simultaneously used control or FXN guides to induce ISC
depletion. Cells were then grown in normoxia and hypoxia and
collected for immunoblotting. As expected, we saw that LIAS
levels were depleted with FXN loss (likely because of loss of the
auxiliary ISC) and stabilized with FDX1 loss (Figs. 3C and S8).
However, LIAS levels trend toward being restored in FXN/
FDX1 double KOs (Figs. 3C and S8) (p = 0.16), suggesting that
the additional KO of FDX1 restored LIAS protein in a FXN KO
background. We validated that the restorative phenotype was
not caused by a general buffering of ISC loss by confirming
that other ISC proteins FECH and POLD1 were still depleted
in FDX1/FXN double KO cells (Fig. 3C). We in addition
observed that the lipoate depletion in FXN/FDX1 double KO
cells was not rescued under low oxygen (Fig. 3C), indicating
6 J. Biol. Chem. (2023) 299(9) 105075
hypoxia could no longer restore lipoate in an FXN KO cell
when FDX1 was absent. In addition, LIAS levels under all
conditions were similarly boosted by hypoxia (Figs. 3C and S8).
Our results are consistent with a model whereby FDX1 loss
prevents the LIAS enzyme from achieving a catalytically active
state that leads to ISC loss (Fig. 3D).
Loss of lipoate synthesis is tolerated in multiple cell types
under hypoxia
Given the shared proteomic signatures between FDX1 and
LIAS KO cells in normoxia (Fig. 2, A and D), we next analyzed
the proteomes of these KO cells in hypoxia. Volcano plots
reveal striking changes in the proteome following FDX1 loss in
normoxia, but these “eruptions” were attenuated in hypoxia
(Fig. 4A). We observe a similar pattern in LIAS KO cells, where
many proteins are differentially expressed with high levels of
significance in 21% oxygen but not in 1% oxygen (Fig. 4A).
We validated some of
the key proteomic changes by
immunoblot analysis. Again, we confirmed our earlier obser-
vation with FDX1 KO cells (Fig. 1, D and E) that lipoylation
was not rescued by hypoxia in LIAS KO cells (Fig. 4B).
Immunoblot analysis validated our proteomics results, and
markers of the integrated stress response (GDF15, ASNS) (41,
42, 47), and oxidative stress (GLRX2) (48), were rescued by
hypoxia (Fig. 4C). In addition, both genetic KOs were tolerated
under hypoxic conditions, in that their growth improved with
exposure to low oxygen and was no longer significantly
different from the control KO cells (Fig. 4D). We further
confirmed this growth phenotype in K562 cells (Fig. S9A) and
validated in both cell lines that depletion of signal in our lip-
oate immunoblots was not attributable to any downstream loss
of the E2 subunits in PDH (DLAT) or KGDH (DLST), which
are modified with the lipoate cofactor (Figs. 4B and S9B) (27).
We used two different methods to interrogate the activity of
lipoate-containing KGDH, in our HepG2 FDX1 and LIAS KO
cell lines in normoxia and hypoxia. (Fig. 4E). First, we per-
formed a KGDH activity assay using extracts from control and
KO cells grown in normoxia and hypoxia. The activity of
KGDH was lost in FDX1 and LIAS KO cells and was not
rescued by hypoxia (Fig. 4E). Second, we performed per-
meabilized cell seahorse experiments and found that feeding
α-ketoglutarate (α-KG) to FDX1 KO or LIAS KO cells did not
result in an increase in oxygen consumption rate over baseline
regardless of ambient oxygen tension (Fig. 4F).
In the absence of lipoate-containing enzymes, a shift to-
ward glycolytic metabolism is expected (28, 49, 50). Indeed,
FDX1 KO and LIAS KO cells consumed more glucose and
produced more lactate in normoxia relative to our control
cells (Fig. 4, G and H). Under hypoxia, FDX1 and LIAS KO
cells exhibited rates of glucose consumption and lactate
production equivalent to those of control cells (Fig. 4, G and
H). However, the KO cells did not consume more glucose or
produce more lactate in hypoxia than in normoxia, whereas
control KO cells did increase their glucose consumption and
lactate production under low oxygen tensions, as expected
(Fig. 4, G and H) (31, 33, 49).
Lipoylation is dispensable under hypoxia
confirm that
the
these
results
Collectively,
lipoate-
containing enzyme KGDH is not able to function without
lipoylation of its E2 subunit, and that in the absence of lipoate,
cells shift from oxidative phosphorylation to glycolysis both in
normoxia and hypoxia. Although the mechanism by which
lipoylation is
hypoxia is allowing cells to tolerate loss of
currently not known, it does not appear to be due to a simple
shift from oxidative metabolism to glycolysis as the KO cells
achieve this even in normoxia.
Discussion
Here, we have explored the functions of the mitochondrial
FDXs and their requirements in hypoxia, and in the process,
have discovered a key role for FDX1 as a partner for LIAS in
lipoic acid production. Surprisingly, we find that loss of FDX1
and LIAS, as well as lipoylation, is all tolerated in two different
human cell types grown in hypoxic conditions.
Our discovery that FDX1 is required for the lipoate syn-
thesis pathway is richly supported by two recent reports (26,
36) as well as other contemporary work (51). Here, we confirm
reports that FDX1 is required for lipoate synthesis, and that
knocking it out stabilizes the enzyme LIAS (26, 36). Using a
genetic strategy, we find that FDX1 is upstream of LIAS and
that the loss of FDX1 stabilizes LIAS protein in cells with
reduced ISC synthesis, possibly by preventing the LIAS
enzyme from entering its catalytic cycle (Fig. 3, C and D). The
catalytic cycle of LIAS involves an offloading and reloading of
the auxiliary 4Fe–4S cluster (29, 36). In the absence of FXN,
the deficiency of ISCs leads to a destabilization of LIAS (Fig. 3,
C and D). FDX1 activity is necessary to initiate LIAS transition
to an unstable state until the ISC is replenished. When FDX1 is
missing, LIAS is no longer undergoing turnover and is stabi-
lized (Figs. 3, C and D and S8).
Our proteomics analysis identifies cellular pathways that
appear to lie downstream of the mitochondrial FDXs. Of note,
sulfite oxidase (a Moco-dependent enzyme in sulfur meta-
bolism) (42) was uniquely depleted upon loss of FDX1 (Fig. 2E)
and would be potentially interesting for further exploration.
We are also intrigued by the modest effect of FDXR loss in
HepG2 cells in normoxia (Fig. 2E). The differences between
FDXR KO and FDX1 or FDX2 KO cells strongly call into
question the idea that FDXR is the sole electron donor for this
pathway, as loss of either downstream FDX causes larger
proteomic changes than loss of the reductase (Fig. 2E) (8).
These data could imply the availability of alternative electron
donors that can complement the need for FDXR function or
that very low levels of FDXR may persist because of incom-
plete genetic ablation of alternative transcripts that are not
detected by our proteomics (40). We also complete the vali-
dation of all members of the ISC machinery and confirm that
only FXN is fully dispensable under low oxygen tensions (with
FDXR rescue being somewhat cell line specific) (Figs. 1, B and
C and S2A) (23). The reasons underlying this unique shared
property for these specific proteins remain unknown.
One of the most surprising findings from our work is that in
multiple cell lines, loss of lipoate can be tolerated in hypoxia
J. Biol. Chem. (2023) 299(9) 105075 7
Lipoylation is dispensable under hypoxia
Figure 4. Lipoate synthesis is dispensable under low oxygen tensions. A, volcano plots depicting log2 fold changes and corresponding FDR for all
proteins in FDX1 and LIAS KO samples compared with controls in normoxia (21% O2) and hypoxia (1% O2). FDX1 and LIAS are shown in blue, gray horizontal
line denotes FDR = 0.01. B, immunoblots for FDX1, LIAS, lipoylated PDH and KGDH, E2 subunit proteins of PDH (DLAT) and KGDH (DLST) enzyme complexes,
and control proteins ACTIN and TOM20 on lysates of HepG2 cells edited with control (CTRL), FDX1, or LIAS guides and grown in 21% or 1% O2. Double
asterisk indicates band of interest. C, immunoblots for GDF15, ASNS, GLRX2, and control proteins ACTIN and TOM20 on lysates of HepG2 cells edited with
control (CTRL), FDX1, or LIAS guides and grown in 21% or 1% O2. D, three-day proliferation assay of HepG2 cells edited with control (CTRL), FDX1, or LIAS
guides and grown in 21% or 1% O2. E, KGDH activity assayed by a KGDH enzyme activity kit using HepG2 cells edited with control, FDX1, or LIAS guides and
grown in 21% or 1% O2. F, bar plot displaying fold change in oxygen consumption rate (OCR) as assessed via permeabilized cell seahorse assays run at 21%
and 1% O2 on HepG2 cells edited with control (CTRL), FDX1, or LIAS guides. Oxygen consumption rates following injection of α-KG were normalized to
baseline readings per well. Dashed line indicates normalized baseline. G, YSI measured glucose consumption over a 3-day period in HepG2 cells edited with
control (CTRL), FDX1, or LIAS guides and grown in 21% or 1% O2, normalized to final cell count on day 3. H, YSI measured lactate production over a 3-day
period in HepG2 cells edited with control (CTRL), FDX1, or LIAS guides and grown in 21% or 1% O2, normalized to final cell count on day 3. All bar plots show
mean ± SD of three independent experiments. ns = p > 0.05, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001. Two-way ANOVA with Bonferroni’s post-test.
α-KG, α-ketoglutarate; FDR, false discovery rate; KGDH, α-ketoglutarate dehydrogenase; LIAS, lipoyl synthase; PDH, pyruvate dehydrogenase.
(Figs. 4, B and D and S9, A, and B). Lipoate is an ancient
cofactor found in all three domains of life and in humans is
critical for the function of key TCA cycle enzymes including
PDH (28, 52). Previous work in Escherichia coli has shown that
expression of the oxygen-labile enzyme pyruvate formate lyase
is sufficient to maintain viability without lipoate in anaerobic
conditions (52, 53). Some bacteria and parasitic organisms are
8 J. Biol. Chem. (2023) 299(9) 105075
also able to salvage lipoate from the media (27, 54–56). In
Saccharomyces cerevisiae, previous reports have shown that
mutants with defects in lipoate synthesis can grow on
fermentative media but not on respiratory media (38, 57). In
the current work, we find that lipoylation is almost completely
ablated with loss of either FDX1 or LIAS in both oxygen
alternative
the
tensions,
existence of
arguing
against
biosynthetic routes or salvage pathways (Fig. 4B). In our prior
low/high O2 CRISPR screen, many additional genes encoding
proteins required for lipoate synthesis also scored, including
members of type II fatty acid synthesis pathway (required to
generate the octanoate precursor for lipoate) (24, 27, 37),
NFU1 (necessary for repair of the auxiliary cluster on LIAS),
and BOLA3 (the putative assembly factor of LIAS) (24, 35, 38,
58). It is notable that multiple components of PDH, including
DLAT, the lipoate-containing subunit, also scored highly in
that screen (24, 27, 59).
Our results indicate that for cells lacking FDX1 or LIAS,
growth in hypoxia continues despite the absence of lipoylation.
Although the mechanism is unknown, it does not appear to be
a simple buffering by hypoxia-driven increase in glycolysis
because even in normoxia, the KO cell lines achieve a defini-
tive shift toward glycolysis (Fig. 4, G and H). Hypoxia appears
to create a state that is tolerant of lipoate deficiency, as stress
responses evident in the proteome are restored to control
levels (Fig. 4C). Future studies will be required to determine
how hypoxia allows cells to tolerate loss of lipoylation. Mu-
tations in various components of the lipoate synthesis pathway
(including LIAS) are associated with debilitating mitochondrial
diseases (60). It will be interesting to determine whether these
conditions are exacerbated by hyperoxia—as we have shown
for CI deficiency or FXN deficiency—and conversely, whether
they may benefit from “hypoxia therapy (23, 61).”
Experimental procedures
Data analysis
All bar plots were analyzed by two-way ANOVA with
Bonferroni’s post-test, with a threshold of p ≤ 0.05. All data
were analyzed in PRISM (GraphPad Software, Inc) with the
appropriate multiple comparisons, and all immunoblots were
analyzed initially and exported from ImageStudio (LI-COR
Biosciences). Each graphed dot in bar plots represents a single
data point.
Cell lines and culturing
K562 (female), human embryonic kidney 293T (female), and
HEPG2 (male) cells were obtained from American Type Cul-
ture Collection and cultured in Dulbecco’s modified Eagle’s
medium (DMEM) (Gibco) with 25 mM glucose, 10% fetal
bovine serum (nondialyzed; Invitrogen), 4 mM glutamine,
1 mM sodium pyruvate, 50 μg/ml uridine, and 100 U/ml
penicillin/streptomycin under 5% CO2 at 37 (cid:1)C. Cell lines were
checked by American Type Culture Collection profiling before
purchase. Cells were tested to ensure the absence of myco-
plasma by PCR-based assay once every 3 months. Cells were
passaged every 2 to 3 days. Adherent cells were washed with
PBS (Invitrogen) and dissociated using TrypLE (Gibco). For
experiments involving hypoxia, cells were placed in Coy O2
control dual hypoxia chambers maintained at 37 (cid:1)C, 1% O2,
and 5% CO2 with appropriate humidity control. Cells were
treated with 75 μM FG-4592 (1:1000 dilution from a 75 mM
(Selleck Chemicals) or dimethyl sulfoxide (Fisher
stock)
Scientific).
Lipoylation is dispensable under hypoxia
Individual single-guide RNAs were cloned into pLenti-
CRISPRv2 (Addgene; catalog no.: 52961) (62), containing a
puromycin or a hygromycin selection cassette. For studying
growth of FXN/FDX1 double KO cells, K562 cells previously
infected with prDA_186 (Addgene; catalog no.: 133458),
bearing guides against a control locus or FXN, were used for
some replicates, whereas repeated infection using guides with
two different selection cassettes were used for others. For
overexpression assays, cDNAs were either purchased from
ORIGENE or custom synthesized from IDT. When necessary,
1× FLAG or 1× FLAG + 1× MYC tags were added to the C
terminus. Constructs were cloned in pLYS6 bearing a
neomycin selection cassette, using the NheI and EcoRI sites.
All plasmids were verified by sequencing. pMD2.G (Addgene;
catalog no.: 12259) and psPAX2 (Addgene; catalog no.: 12260)
were used for lentiviral packaging.
Lentivirus production
About 2.5 × 106 or 6.25 × 106 human embryonic kidney
293T cells were seeded in 5 or 10 ml in a T25 cm2 flask or a
10 cm dish (one lentivirus per flask). The following day, the
cells were transfected with 1 (or 2) ml of transfection mixture.
The transfection mixture contained 25 (or 50) μl Lipofect-
amine 2000 (Thermo Fisher Scientific), 3.75 (7.5) μg psPAX2,
2.5 (5) μg pMD2.G, 5 (10) μg of lentiviral vector of interest,
and Opti-MEM medium (Gibco) up to 1 (2) ml. The mixture
was incubated at room temperature (RT) for 20 min before
adding it to cells. Six hours following transfection, the media
were replaced with fresh DMEM. Two days after transfection,
media were collected, filtered through a 0.45 μM filter, and
stored at −80 (cid:1)C.
Infection
K562 cells were seeded at 5 × 105 cells/ml in 2 ml per well in
a 6-well plate the day of infection. Cells were infected with
virus, and polybrene was added at 1:1000 final volume (Invi-
trogen). Cells were incubated overnight before being selected
with puromycin (2 μg/ml final concentration) (Gibco), genet-
icin (500 μg/ml) (Gibco), or hygromycin (Gibco) (250 μg/ml)
for 48 h.
HepG2 cells were seeded at 6 × 105 cells per well the day
before infection in a 6-well plate. Cells were then infected with
virus, and polybrene was added at 1:2000 final volume (Invi-
trogen). Cells were incubated overnight before being selected
with puromycin (3 μg/ml final concentration) (Gibco) for 48 h.
Polyacrylamide gel electrophoresis and protein
immunoblotting
About 2 to 5 × 106 K562 or 3 to 6 × 106 HepG2 cells were
harvested, washed in cold PBS, and lysed for 10 to 25 min on
ice in radioimmunoprecipitation lysis buffer (Thermo Fisher),
1× HALT protease and phosphatase inhibitor
(Thermo
Fisher), and Pierce Universal Nuclease for Cell Lysis (Thermo
Fisher). Lysates were further clarified by centrifugation for
10 min at 10,000g at 4 (cid:1)C. Supernatant was collected into fresh
tubes, and protein concentration was measured with the Pierce
J. Biol. Chem. (2023) 299(9) 105075 9
Lipoylation is dispensable under hypoxia
660 nm protein assay (Thermo Fisher). About 30 μg was
loaded per well in Novex Tris–glycine 4 to 20% or 10 to 20%
gels (Life Technologies). Gels were run for 50 min at 200 V
and transferred onto a nitrocellulose membrane, 0.45 μM (Bio-
Rad). Membranes were stained with Ponceau S to check for
adequate loading. Membranes were then blocked for 1 to 2 h
with Odyssey Blocking Buffer (LI-COR Biosciences) at RT.
Afterward, membranes were incubated overnight at 4 (cid:1)C with
a solution of primary antibody diluted in 3% bovine serum
albumin in Tris-buffered saline with Tween-20 (TBS-T) +
0.05% N3. The next day, membranes were washed at RT five
times in TBS-T for 5 min. The membrane was incubated with
goat anti-rabbit or antimouse conjugated to IRDye800 or
IRDye680 (LI-COR Biosciences) in a 1:1 solution of Odyssey
blocking buffer (LI-COR Biosciences) and TBS-T. Membranes
were incubated for 1 h at RT and then washed three times in
TBS-T for 10 min each. Membranes were then scanned for
infrared signal using the Odyssey Imaging System (LI-COR
Biosciences). Band intensities were analyzed with Image Studio
LITE (LI-COR Biosciences).
Antibodies
Antigen
Catalog number
Vendor
FXN
OXPHOS
FDX1
FDX2
LIAS
Tubulin
Actin
Actin
FLAG
DYKDDDDK Tag
HSP60
TOM20
FDXR
Lipoic acid
Lipoic acid
POLD1
FECH
DLAT
DLST
GDF15
ASNS
GLRX2
ab175402
Ab110411
12592-1-AP
HPA043986
11577-1-AP
MA5-16308
Ab8227
8H10D10
F3165-2MG
2368
ab45134
Sc-17764
Sc-374436
437695
Ab58724
15646-1-AP
14466-1-AP
12362
5556
27455-1-AP
14681-1-AP
13381-1-AP
Abcam
Abcam
ProteinTech
Atlas
ProteinTech
Invitrogen
Abcam
Cell Signaling
Sigma
Cell Signaling
Abcam
Santa Cruz Biotechnology
Santa Cruz Biotechnology
EMD Millipore
Abcam
ProteinTech
ProteinTech
Cell Signaling
Cell Signaling
ProteinTech
ProteinTech
ProteinTech
Proliferation assays
Cell proliferation assays were performed between 8 and
10 days following lentiviral infection. Cells were seeded at an
initial density of 1.5 × 105 cells/ml (K562) or 2.5 × 105 cells per
well in a 6-well plate (HepG2) and cultured for 3 days in either
21% or 1% oxygen tensions. Viable cell numbers were then
determined using a Vi-Cell Counter (Beckman).
qPCR
About 2.5 to 3.5 × 106 cells were collected per sample and
snap frozen in liquid nitrogen and stored at −80 (cid:1)C until use.
Cells were then thawed on ice, and RNA was extracted using
the QIAGEN RNeasy mini kit and DNASE-I digested before
murine leukemia virus reverse transcription with random
primers (Promega). qPCR was performed using TaqMan
10 J. Biol. Chem. (2023) 299(9) 105075
technology (Life Technologies) using probes HS00188949_m1
(BNIP3L) and HS00472881_m1 (PUM1).
Proteomics
sgCTRL, sgFXN, sgFDXR, sgFDX1, sgFDX2, and sgLIAS
HepG2 cells were grown for 6 days in 21% or 1% oxygen
conditions in 150 mm plates. Cells were washed four times in
ice-cold PBS, scraped into fresh ice-cold PBS, and spun down
at 300g for 5 min at 4 (cid:1)C in a microcentrifuge. The remaining
PBS was siphoned off, and the cell pellets were snap frozen in
liquid nitrogen and stored at −80 (cid:1)C until the time of sample
submission to the Thermo Fisher Scientific Center for Multi-
plexed Proteomics (Harvard).
Sample preparation for mass spectrometry
Samples for protein analysis were prepared essentially as
previously described (63, 64). Following lysis, protein precipi-
tation, reduction/alkylation, and digestion, peptides were
quantified by micro–bicinchoninic acid assay and 100 μg of
peptide per sample were labeled with TMTpro reagents
(Thermo Fisher) for 2 h at RT. Labeling reactions were
quenched with 0.5% hydroxylamine and acidified with TFA.
Acidified peptides were combined and desalted by Sep-Pak
(Waters).
Basic pH reversed-phase separation
Tandem mass tag (TMT)–labeled peptides were solubilized
in 5% acetonitrile (ACN)/10 mM ammonium bicarbonate, pH
8.0, and 300 μg of TMT-labeled peptides was separated by an
Agilent 300 Extend C18 column (3.5 μm particles, 4.6 mm ID
and 250 mm in length). An Agilent 1260 binary pump coupled
with a photodiode array detector (Thermo Scientific) was used
to separate the peptides. A 45 min linear gradient from 10% to
40% ACN in 10 mM ammonium bicarbonate pH 8.0 (flow rate
of 0.6 ml/min) separated the peptide mixtures into a total of 96
fractions (36 s). A total of 96 fractions were consolidated into
24 samples in a checkerboard fashion, acidified with 20 μl of
10% formic acid, and vacuum dried to completion. Each
sample was desalted via Stage Tips and redissolved in 5%
formic acid/5% ACN for LC-MS3 analysis.
Liquid chromatography separation and tandem mass
spectrometry (LC–MS3)
Proteome data were collected on an Orbitrap Eclipse mass
spectrometer (ThermoFisher Scientific) coupled to a Proxeon
EASY-nLC 1200 LC pump (ThermoFisher Scientific). Frac-
tionated peptides were separated using a 120 min gradient at
500 nl/min on a 35 cm column (i.d. 100 μm, Accucore, 2.6 μm,
150 Å) packed in-house. High-field asymmetric-waveform ion
mobility spectrometry was enabled during data acquisition
with compensation voltages set as −40, −60, and −80 V (65).
MS1 data were collected in the Orbitrap (120,000 resolution;
maximum injection time of 50 ms; automatic gain control
[AGC] 4 × 105). Charge states between 2 and 5 were required
for MS2 analysis, and a 120 s dynamic exclusion window was
used. Top 10 MS2 scans were performed in the ion trap with
collision-induced dissociation fragmentation (isolation win-
dow of 0.5 Da; Turbo; normalized collision energy of 35%;
maximum injection time of 35 ms; AGC 1 × 104). An on-line
real-time search algorithm (Orbiter) was used to trigger MS3
scans for quantification (66). MS3 scans were collected in the
Orbitrap using a resolution of 50,000, normalized collision
energy of 45%, maximum injection time of 200 ms, and AGC
of 3.0 × 105. The closeout was set at two peptides per protein
per fraction (66).
Data analysis
Raw files were converted to mzXML, and monoisotopic
peaks were reassigned using Monocle (67). Searches were
performed using the Comet search algorithm against a human
database downloaded from UniProt in May 2021. We used a
50 ppm precursor ion tolerance, 1.0005 fragment ion toler-
ance, and 0.4 fragment bin offset for MS2 scans. TMTpro on
lysine residues and peptide N termini (+304.2071 Da) and
carbamidomethylation of cysteine residues (+57.0215 Da) were
set as static modifications, whereas oxidation of methionine
residues (+15.9949 Da) was set as a variable modification.
Each run was filtered separately to 1% FDR on the peptide-
spectrum match level. Then proteins were filtered to the target
1% FDR level across the combined dataset. For reporter ion
quantification, a 0.003 Da window around the theoretical m/z
of each reporter ion was scanned, and the most intense m/z
was used. Reporter ion intensities were adjusted to correct for
isotopic impurities of the different TMTpro reagents accord-
ing to the manufacturer’s specifications. Peptides were filtered
to include only those with a summed signal-to-noise ≥120
across 12 TMT channels. The signal-to-noise measurements
of peptides assigned to each protein were summed for a given
protein. These values were normalized so that the sum of the
signal for all proteins in each channel was equivalent, thereby
accounting for equal protein loading.
Proteins that did not have a valid readout in any of the 24
channels were filtered out. To correct for differences caused by
separate experiment runs, the python package pyComBat
(version 0.3.2) was run on the log2-transformed data before
projecting the corrected values back into linear space. Proteins
with differential abundance across conditions were determined
with the R package EdgeR (version 3.36.0) with exact testing,
and the Benjamini–Hochberg multiple testing correction was
applied to control for FDRs. Proteins with FDR of lower than
0.01 were considered to have significantly differential abun-
dances between conditions. The downstream pathway
enrichment analysis was completed using GSEA (version 4.2.3)
PreRanked (68, 69) with a list of significant proteins ranked by
their corresponding log2-fold changes as input. Candidate
pathways for the enrichment analysis were taken from the
Human MitoPathways 3.0 database (39).
Protein modeling
Docking predictions between LIAS and FDX1 and FDX2
(https://github.com/
through the AlphaFold2_mmseqs2
were
sokrypton/ColabFold)
using ColabFold
obtained
Lipoylation is dispensable under hypoxia
notebook (56–58). Runs were performed using pdb70 tem-
plates, alignments were through MMseqs2 in unpaired +
paired mode, and num_recycles was set to 3. The top five
models for each run were structurally compared between each
other, and the predicted alignment error plots were used to
assess the likelihood of the predicted interface (70–72).
Mitochondria isolation
About 5 × 107 cells were harvested, washed in PBS, and
either snap frozen and stored in −80 (cid:1)C before proceeding to
the next step or washed immediately after with 10 ml MB
buffer (210 mM mannitol, 70 mM sucrose, 10 mM Hepes–
KOH at pH 7.4, 1 mM EDTA, and protease/phosphatase in-
hibitor). Cells were resuspended in 1 ml of MB buffer sup-
plemented with 1× HALT protease phosphatase inhibitor
(ThermoFisher Scientific) and transferred to 2 ml glass ho-
mogenizer (Kontes). Cells were broken with (cid:3)35 strokes of a
large pestle on ice. MB + protease/phosphatase was added up
to 6 ml. The samples were then centrifuged at 2000g for 5 min,
and the pellet was discarded. The supernatant was then
centrifuged again at 13,000g for 10 min at 4 (cid:1)C. The mito-
chondrial pellets were washed with MB buffer once and
resuspended in radioimmunoprecipitation lysis buffer with
protease inhibitor (1:100×) and universal nuclease (1:1000×).
Enzyme activity assay
The KGDH enzyme activity kit was purchased from Sigma
(MAK189). Sample processing and activity assays were carried
out as per kit instructions. Briefly, 1 × 106 cells were pelleted
and washed once in PBS. Cells were then lysed in assay buffer
for 10 min on ice in normoxia or hypoxia before being clarified
by centrifugation at 4 (cid:1)C for 5 min at 10,000g. The supernatant
was collected, aliquoted, and snap frozen in liquid nitrogen
before being stored in −80 (cid:1)C for future use. On the day of the
assay, samples were thawed on ice and aliquoted into a 96-well
clear flat bottom plate (Corning). Kit-provided enzyme-spe-
cific developer and substrate reaction mixes were added to
each sample, and the plate was then placed into a Cytation 5
instrument (BioTek). Absorbance was measured at 450 nm
every minute at 37(cid:1) for up to 2 h. Measurement analysis was
then calculated as described in kit instructions.
Permeabilized cell seahorse measurements
Oxygen consumption rate studies were conducted in a
Seahorse XFe96 Analyzer at 21% or 1% O2 tensions. All ex-
periments were conducted at 37 (cid:1)C at pH 7.2. HepG2 cells
were seeded the day before in standard media conditions in the
provided seahorse cell culture microplate at 2.5 × 104 cells/
well. The Seahorse cartridge was hydrated with 200 μl per well
of Seahorse XF Calibrant (Agilent) and placed in a 37 (cid:1)C
incubator overnight at either 21% or 1% oxygen. The following
day, wash buffer, seahorse media, and injectable media were
prepared in 1× MAS buffer (70 mM sucrose, 220 mM
mannitol, 10 mM KH2PO4, 5 mM MgCl2, 2 mM Hepes, and
1 mM EGTA; pH 7.2). For hypoxia experiments, 1X MAS
buffer was placed in the hypoxia glovebox overnight. About
J. Biol. Chem. (2023) 299(9) 105075 11
Lipoylation is dispensable under hypoxia
10% fatty-acid free bovine serum albumin was added to both
the wash buffer and seahorse media at a final concentration of
0.2%, and 0.5 M ADP was supplemented to the seahorse media
at a final concentration of 4 mM along with 1 nM XF Plasma
Membrane Permeabilizer (Agilent).Cells were washed twice
with the wash buffer before being replated in Seahorse Media.
After five to six baseline measurements, cells were injected
with α-KG at a final concentration of 10 mM, followed by
injections of oligomycin (4 μM final), and then piericidin/
antimycin (5 μM final). For assays performed under hypoxia,
edge wells on the plate were injected with 0.1 M sodium sulfite
solution for oxygen tension calibration. For data analysis, the
second measurement postinjection with α-KG (representative
of three measurements) was divided by the last baseline
measurement to normalize the data. Four technical replicates
from each biological replicate were plotted in the bar plot.
Glucose uptake and lactate release measurements
Glucose and lactate concentrations were measured using an
automatic glucose and lactate analyzer YSI 2900 Series. Cells
were seeded in 6-well dishes with 3 ml standard culture media
as mentioned previously. At the end of 3 days, 500 μl of media
were collected from each well and centrifuged at 300g in a
microcentrifuge for 4 min to pellet any cell debris. About 200
μl of media were loaded onto a 96-well flat bottom clear plate
(Corning). The YSI program read triplicate measurements
from each well. Cells from each well were simultaneously
counted using a Vi-Cell Counter (Beckman) to obtain final cell
count. For data analysis, the glucose consumed values were
subtracted from the established glucose values in the DMEM
(25 mM). Lactate values were added to the known lactate
values in the DMEM (0 mM). These concentrations were then
divided by final cell count, the number of days for the exper-
iment (3), and finally this corrected value was multiplied by the
amount of media volume per well. Triplicate values from three
biological replicates for each condition were plotted for the bar
plots shown in the figure.
Data availability
All data described within the article are contained in the
document. The mass spectrometry proteomics data have been
deposited to the ProteomeXchange Consortium via the PRIDE
(73) partner repository with the dataset identifier PXD042589.
Any further information and requests for resources and re-
agents should be directed to and will be fulfilled by the Lead
Contact, Vamsi K. Mootha ([email protected]).
Supporting information—This article contains supporting informa-
tion (70–72).
Acknowledgments—We thank Thomas Hercher, Owen Skinner,
Tsz-Leung To, Joshua Meisel, and all members of the Mootha
laboratory for fruitful discussions and feedback. This work has
been supported by the Friedriech’s Ataxia Research Alliance.
12 J. Biol. Chem. (2023) 299(9) 105075
Author contributions—P. R. J. and V. K. M. conceptualization; P. R. J.,
S. S., and X. A. G. validation; P. R. J. and X. A. G. formal analysis; P. R. J.
and J. G. M. investigation; V. K. M. resources; X. A. G. data curation; P.
R. J. and V. K. M. writing–original draft; P. R. J., S. S., X. A. G., J. G. M.,
and V. K. M. writing–review & editing; P. R. J., S. S., X. A. G., and J. G.
M. visualization; V. K. M. supervision; V. K. M. funding acquisition.
Funding and additional information—P. R. J. is supported by the
National Science Foundation Graduate Research Fellowship Pro-
gram. V. K. M. is an investigator of the Howard Hughes Medical
Institute.
Conflict of interest—V. K. M. is on the scientific advisory board of
Janssen Pharmaceuticals and 5AM Ventures. V. K. M. is listed as an
inventor on a patent application filed by Massachusetts General
Hospital on the use of hypoxia as a therapy for mitochondrial and
degenerative diseases.
false discovery rate; FDX,
Abbreviations—The abbreviations used are: α-KG, α-ketoglutarate;
ACN, acetonitrile; AGC, automatic gain control; cDNA, comple-
mentary DNA; CI, complex I; CII, complex II; CIII, complex III;
CIV, complex IV; DMEM, Dulbecco’s modified Eagle’s medium;
FDR,
ferredoxin
reductase; FXN, frataxin; HIF, hypoxia-inducible factor; ISC, iron–
sulfur cluster; KGDH, α-ketoglutarate dehydrogenase; LIAS, lipoyl
synthase; PDH, pyruvate dehydrogenase; qPCR, quantitative PCR;
RT, room temperature; TBS-T, Tris-buffered saline with Tween-20;
TCA, tricarboxylic acid; TMT, tandem mass tag.
ferredoxin; FDXR,
References
1. Beinert, H. (2000) Iron-sulfur proteins: ancient structures, still full of
surprises. J. Biol. Inorg. Chem. 5, 2–15
2. Imlay, J. A. (2006) Iron-sulphur clusters and the problem with oxygen.
Mol. Microbiol. 59, 1073–1082
3. Brzóska, K., Meczy(cid:1)nska, S., and Kruszewski, M. (2006) Iron-sulfur cluster
proteins: electron transfer and beyond. Acta Biochim. Pol. 53, 685–691
4. Srour, B., Gervason, S., Monfort, B., and D’Autréaux, B. (2020) Mecha-
nism of iron–sulfur cluster assembly: in the intimacy of iron and sulfur
Encounter. Inorganics 8, 55
5. Stehling, O., and Lill, R. (2013) The role of mitochondria in cellular iron-
sulfur protein biogenesis: mechanisms, connected processes, and diseases.
Cold Spring Harb. Perspect. Biol. 5, a011312
6. Weiler, B. D., Brück, M. C., Kothe, I., Bill, E., Lill, R., and Mühlenhoff, U.
(2020) Mitochondrial [4Fe-4S] protein assembly involves reductive [2Fe-
2S] cluster fusion on ISCA1-ISCA2 by electron flow from ferredoxin
FDX2. Proc. Natl. Acad. Sci. U. S. A. 117, 20555–20565
7. Lill, R., and Freibert, S. A. (2020) Mechanisms of mitochondrial iron-
sulfur protein biogenesis. Annu. Rev. Biochem. 89, 471–499
8. Yan, R., Adinolfi, S., and Pastore, A. (2015) Ferredoxin, in conjunction
with NADPH and ferredoxin-NADP reductase, transfers electrons to the
IscS/IscU complex to promote iron-sulfur cluster assembly. Biochim.
Biophys. Acta 1854, 1113–1117
9. Andreini, C., Banci, L., and Rosato, A. (2016) Exploiting bacterial operons
to illuminate human iron–sulfur proteins. J. Proteome Res. 15, 1308–1322
10. Paul, V. D., and Lill, R. (2015) Biogenesis of cytosolic and nuclear iron-
sulfur proteins and their role in genome stability. Biochim. Biophys. Acta
1853, 1528–1539
11. Hänzelmann, P., Hernández, H. L., Menzel, C., García-Serres, R., Huynh,
B. H., Johnson, M. K., et al. (2004) Characterization of MOCS1A, an
oxygen-sensitive iron-sulfur protein involved in human molybdenum
cofactor biosynthesis. J. Biol. Chem. 279, 34721–34732
12. Nishino, T., and Okamoto, K. (2000) The role of the [2Fe-2s] cluster
centers in xanthine oxidoreductase. J. Inorg. Biochem. 82, 43–49
13. Read, A. D., Bentley, R. E., Archer, S. L., and Dunham-Snary, K. J. (2021)
Mitochondrial iron-sulfur clusters: structure, function, and an emerging
role in vascular biology. Redox Biol. 47, 102164
14. Hall, D. O., Cammack, R., and Rao, K. K. (1971) Role for ferredoxins in
the origin of life and biological evolution. Nature 233, 136–138
15. Eck, R. V., and Dayhoff, M. O. (1966) Evolution of the structure of
ferredoxin based on living relics of primitive amino acid sequences. Sci-
ence 152, 363–366
16. Buckel, W., and Thauer, R. K. (2018) Flavin-based electron bifurcation,
ferredoxin, flavodoxin, and anaerobic respiration with protons (Ech) or
NAD(+) (rnf) as electron acceptors: a historical review. Front. Microbiol.
9, 401
17. Sheftel, A. D., Stehling, O., Pierik, A. J., Elsässer, H. P., Mühlenhoff, U.,
Webert, H., et al. (2010) Humans possess two mitochondrial ferredoxins,
Fdx1 and Fdx2, with distinct roles in steroidogenesis, heme, and Fe/S
cluster biosynthesis. Proc. Natl. Acad. Sci. U. S. A. 107, 11775–11780
18. Ziegler, G. A., and Schulz, G. E. (2000) Crystal structures of adrenodoxin
reductase in complex with NADP+ and NADPH suggesting a mechanism
the electron transfer of an enzyme family. Biochemistry 39,
for
10986–10995
19. Ziegler, G. A., Vonrhein, C., Hanukoglu, I., and Schulz, G. E. (1999) The
structure of adrenodoxin reductase of mitochondrial P450 systems:
electron transfer for steroid biosynthesis. J. Mol. Biol. 289, 981–990
20. Ewen, K. M., Ringle, M., and Bernhardt, R. (2012) Adrenodoxin—a ver-
satile ferredoxin. IUBMB Life 64, 506–512
21. Barros, M. H., Carlson, C. G., Glerum, D. M., and Tzagoloff, A. (2001)
Involvement of mitochondrial ferredoxin and Cox15p in hydroxylation of
heme O. FEBS Lett. 492, 133–138
22. Cai, K., Tonelli, M., Frederick, R. O., and Markley, J. L. (2017) Human
mitochondrial ferredoxin 1 (FDX1) and ferredoxin 2 (FDX2) both bind
cysteine desulfurase and donate electrons for iron-sulfur cluster biosyn-
thesis. Biochemistry 56, 487–499
23. Ast, T., Meisel, J. D., Patra, S., Wang, H., Grange, R. M. H., Kim, S. H.,
et al. (2019) Hypoxia rescues frataxin loss by restoring iron sulfur cluster
biogenesis. Cell 177, 1507–1521.e116
24. Jain, I. H., Calvo, S. E., Markhard, A. L., Skinner, O. S., To, T. L., Ast, T.,
et al. (2020) Genetic screen for cell fitness in high or low oxygen high-
lights mitochondrial and lipid metabolism. Cell 181, 716–727.e11
25. Arroyo, J. D., Jourdain, A. A., Calvo, S. E., Ballarano, C. A., Doench, J. G.,
Root, D. E., et al. (2016) A genome-wide CRISPR death screen identifies
genes essential for oxidative phosphorylation. Cell Metab. 24, 875–885
26. Tsvetkov, P., Coy, S., Petrova, B., Dreishpoon, M., Verma, A., Abdusa-
mad, M., et al. (2022) Copper induces cell death by targeting lipoylated
TCA cycle proteins. Science 375, 1254–1261
27. Solmonson, A., and DeBerardinis, R. J. (2018) Lipoic acid metabolism and
mitochondrial redox regulation. J. Biol. Chem. 293, 7522–7530
28. Cronan, J. E. (2020) Progress in the enzymology of the mitochondrial
diseases of lipoic acid requiring enzymes. Front. Genet. 11, 510
29. Booker, S. (2022) The biosynthesis of
destruction, and rebirth. FASEB J. 36
lipoic acid: a saga of death,
30. Semenza, G. L. (2003) Targeting HIF-1 for cancer therapy. Nat. Rev.
Cancer 3, 721–732
31. Majmundar, A. J., Wong, W. J., and Simon, M. C. (2010) Hypoxia-
inducible factors and the response to hypoxic stress. Mol. Cell 40,
294–309
32. Rabinowitz, M. H. (2013) Inhibition of hypoxia-inducible factor prolyl
hydroxylase domain oxygen sensors: tricking the body into mounting
orchestrated survival and repair responses. J. Med. Chem. 56, 9369–9402
33. Denko, N. C., Fontana, L. A., Hudson, K. M., Sutphin, P. D., Raychaud-
huri, S., Altman, R., et al. (2003) Investigating hypoxic tumor physiology
through gene expression patterns. Oncogene 22, 5907–5914
34. Crooks, D. R., Maio, N., Lane, A. N., Jarnik, M., Higashi, R. M., Haller, R.
G., et al. (2018) Acute loss of iron-sulfur clusters results in metabolic
reprogramming and generation of lipid droplets in mammalian cells. J.
Biol. Chem. 293, 8297–8311
Lipoylation is dispensable under hypoxia
NFU1 and BOLA3 cause a fatal deficiency of multiple respiratory chain
and 2-oxoacid dehydrogenase enzymes. Am. J. Hum. Genet. 89, 486–495
36. Schulz, V., Basu, S., Freibert, S.-A., Webert, H., Boss, L., Mühlenhoff, U.,
et al. (2022) Functional spectrum and specificity of mitochondrial ferre-
doxins FDX1 and FDX2. Nature Chem. Biol. 19, 206–217
37. Nowinski, S. M., Solmonson, A., Rusin, S. F., Maschek, J. A., Bensard, C.
L., Fogarty, S., et al. (2020) Mitochondrial fatty acid synthesis coordinates
oxidative metabolism in mammalian mitochondria. Elife 9, e58041
38. Melber, A., Na, U., Vashisht, A., Weiler, B. D., Lill, R., Wohlschlegel, J. A.,
et al. (2016) Role of Nfu1 and Bol3 in iron-sulfur cluster transfer to
mitochondrial clients. Elife 5, e15991
39. Rath, S., Sharma, R., Gupta, R., Ast, T., Chan, C., Durham, T. J., et al.
(2020) MitoCarta3.0: an updated mitochondrial proteome now with sub-
organelle localization and pathway annotations. Nucleic Acids Res. 49,
D1541–D1547
40. Cruz-Garcia, L., O’Brien, G., Sipos, B., Mayes, S., Tichý, A., Sirák, I., et al.
(2020) In vivo validation of alternative FDXR transcripts in human blood
in response to ionizing radiation. Int. J. Mol. Sci. 21, 7851
41. Kim, K. H., and Lee, M. S. (2021) GDF15 as a central mediator for in-
tegrated stress response and a promising therapeutic molecule for
metabolic disorders and NASH. Biochim. Biophys. Acta Gen. Subj. 1865,
129834
42. Wang, D., Day, E. A., Townsend, L. K., Djordjevic, D., Jørgensen, S. B.,
and Steinberg, G. R. (2021) GDF15: emerging biology and therapeutic
applications for obesity and cardiometabolic disease. Nat. Rev. Endocrinol.
17, 592–607
43. Garrett, R. M., Johnson, J. L., Graf, T. N., Feigenbaum, A., and Rajago-
palan, K. V. (1998) Human sulfite oxidase R160Q: identification of the
mutation in a sulfite oxidase-deficient patient and expression and char-
acterization of the mutant enzyme. Proc. Natl. Acad. Sci. U. S. A. 95,
6394–6398
44. Hendricks, A. L., Wachnowsky, C., Fries, B., Fidai, I., and Cowan, J. A.
(2021) Characterization and reconstitution of human lipoyl synthase
(LIAS) supports ISCA2 and ISCU as primary cluster donors and an or-
dered mechanism of cluster assembly. Int. J. Mol. Sci. 22, 1598
45. McCarthy, E. L., and Booker, S. J. (2017) Destruction and reformation of
an iron-sulfur cluster during catalysis by lipoyl synthase. Science 358,
373–377
46. McLaughlin, M. I., Lanz, N. D., Goldman, P. J., Lee, K. H., Booker, S. J.,
and Drennan, C. L. (2016) Crystallographic snapshots of sulfur insertion
by lipoyl synthase. Proc. Natl. Acad. Sci. U. S. A. 113, 9446–9450
47. Pakos-Zebrucka, K., Koryga, I., Mnich, K., Ljujic, M., Samali, A., and
Gorman, A. M. (2016) The integrated stress response. EMBO Rep. 17,
1374–1395
48. Beer, S. M., Taylor, E. R., Brown, S. E., Dahm, C. C., Costa, N. J.,
Runswick, M. J., et al. (2004) Glutaredoxin 2 catalyzes the reversible
oxidation and glutathionylation of mitochondrial membrane thiol pro-
teins: implications for mitochondrial redox regulation and antioxidant
DEFENSE. J. Biol. Chem. 279, 47939–47951
49. Li, X., Yang, Y., Zhang, B., Lin, X., Fu, X., An, Y., et al. (2022) Lactate
metabolism in human health and disease. Signal Transduct. Target. Ther.
7, 305
50. Burr, S. P., Costa, A. S., Grice, G. L., Timms, R. T., Lobb, I. T., Freisinger,
P., et al. (2016) Mitochondrial protein lipoylation and the 2-oxoglutarate
dehydrogenase complex controls HIF1α stability in aerobic conditions.
Cell Metab. 24, 740–752
51. [preprint] Dreishpoon, M. B., Bick, N. R., Petrova, B., Warui, D. M.,
Cameron, A., Booker, S. J., et al. (2023) FDX1 regulates cellular protein
lipoylation through direct binding to LIAS. bioRxiv. https://doi.org/10.
1101/2023.02.03.526472
52. Spalding, M. D., and Prigge, S. T. (2010) Lipoic acid metabolism in mi-
crobial pathogens. Microbiol. Mol. Biol. Rev. 74, 200–228
53. Cronan, J. E., Zhao, X., and Jiang, Y. (2005) Function, attachment and
synthesis of lipoic acid in Escherichia coli. Adv. Microb. Physiol. 50,
103–146
35. Cameron, J. M., Janer, A., Levandovskiy, V., Mackay, N., Rouault, T. A.,
Tong, W. H., et al. (2011) Mutations in iron-sulfur cluster scaffold genes
54. Hermes, F. A., and Cronan, J. E. (2009) Scavenging of cytosolic octanoic
acid by mutant LplA lipoate ligases allows growth of Escherichia coli
J. Biol. Chem. (2023) 299(9) 105075 13
Lipoylation is dispensable under hypoxia
strains lacking the LipB octanoyltransferase of lipoic acid synthesis. J.
Bacteriol. 191, 6796–6803
55. Allary, M., Lu, J. Z., Zhu, L., and Prigge, S. T. (2007) Scavenging of the
cofactor lipoate is essential for the survival of the malaria parasite Plas-
modium falciparum. Mol. Microbiol. 63, 1331–1344
56. Deschermeier, C., Hecht, L. S., Bach, F., Rützel, K., Stanway, R. R.,
Nagel, A., et al. (2012) Mitochondrial lipoic acid scavenging is essential
liver stage development. Cell Microbiol. 14,
for Plasmodium berghei
416–430
57. Hermes, F. A., and Cronan, J. E. (2013) The role of the Saccharomyces
cerevisiae lipoate protein ligase homologue, Lip3, in lipoic acid synthesis.
Yeast 30, 415–427
58. Warui, D. M., Sil, D., Lee, K.-H., Neti, S. S., Esakova, O. A., Knox, H. L.,
et al. (2022) In vitro demonstration of human lipoyl synthase catalytic
activity in the presence of NFU1. ACS Bio Med. Chem. Au 2, 456–468
59. Rowland, E. A., Snowden, C. K., and Cristea, I. M. (2018) Protein lip-
oylation: an evolutionarily conserved metabolic regulator of health and
disease. Curr. Opin. Chem. Biol. 42, 76–85
60. Mayr, J. A., Feichtinger, R. G., Tort, F., Ribes, A., and Sperl, W. (2014)
Lipoic acid biosynthesis defects. J. Inherit. Metab. Dis. 37, 553–563
61. Jain, I. H., Zazzeron, L., Goli, R., Alexa, K., Schatzman-Bone, S., Dhillon,
H., et al. (2016) Hypoxia as a therapy for mitochondrial disease. Science
352, 54–61
62. Sanjana, N. E., Shalem, O., and Zhang, F. (2014) Improved vectors and
genome-wide libraries for CRISPR screening. Nat. Methods 11, 783–784
63. Li, J., Cai, Z., Bomgarden, R. D., Pike, I., Kuhn, K., Rogers, J. C., et al.
(2021) TMTpro-18plex: the Expanded and complete set of TMTpro re-
agents for sample Multiplexing. J. Proteome Res. 20, 2964–2972
65. Schweppe, D. K., Prasad, S., Belford, M. W., Navarrete-Perea, J., Bailey,
D. J., Huguet, R., et al. (2019) Characterization and optimization of
multiplexed quantitative analyses using high-field asymmetric-waveform
ion mobility mass spectrometry. Anal. Chem. 91, 4010–4016
66. Schweppe, D. K., Eng, J. K., Yu, Q., Bailey, D., Rad, R., Navarrete-Perea, J.,
et al. (2020) Full-featured, real-time database searching platform enables
fast and accurate multiplexed quantitative proteomics. J. Proteome Res.
19, 2026–2034
67. Rad, R., Li, J., Mintseris, J., O’Connell, J., Gygi, S. P., and Schweppe, D. K.
(2021) Improved monoisotopic mass estimation for deeper proteome
coverage. J. Proteome Res. 20, 591–598
68. Subramanian, A., Tamayo, P., Mootha, V. K., Mukherjee, S., Ebert, B. L.,
Gillette, M. A., et al. (2005) Gene set enrichment analysis: a knowledge-
based approach for interpreting genome-wide expression profiles. Proc
Natl Acad Sci U S A 102, 15545–15550
69. Mootha, V. K., Lindgren, C. M., Eriksson, K.-F., Subramanian, A., Sihag,
S., Lehar, J., et al. (2003) PGC-1α-responsive genes involved in oxidative
phosphorylation are coordinately downregulated in human diabetes. Nat
Genet 34, 267–273
70. Mirdita, M., Schütze, K., Moriwaki, Y., Heo, L., Ovchinnikov, S., and
Steinegger, M. (2022) ColabFold: making protein folding accessible to all.
Nat. Methods 19, 679–682
71. Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O.,
et al. (2021) Highly accurate protein structure prediction with AlphaFold.
Nature 596, 583–589
72. Johansson-Åkhe, I., and Wallner, B. (2022) Improving peptide-protein
docking with AlphaFold-multimer using forced sampling. Front. Bio-
inform. 2, 959160
64. Navarrete-Perea, J., Yu, Q., Gygi, S. P., and Paulo, J. A. (2018) Streamlined
tandem mass tag (SL-TMT) protocol: an efficient strategy for quantitative
(Phospho)proteome profiling using tandem mass tag-synchronous pre-
cursor selection-MS3. J. Proteome Res. 17, 2226–2236
73. Perez-Riverol, Y., Bai, J., Bandla, C., García-Seisdedos, D., Hewapathir-
ana, S., Kamatchinathan, S., et al. (2022) The PRIDE database resources in
2022: a hub for mass spectrometry-based proteomics evidences. Nucleic
Acids Res. 50, D543–D552
14 J. Biol. Chem. (2023) 299(9) 105075
| null |
10.1112_blms.12639.pdf
| null | null |
ETH Library
The proportion of derangements
characterizes the symmetric and
alternating groups
Journal Article
Author(s):
Poonen, Bjorn; Slavov, Kaloyan
Publication date:
2022-08
Permanent link:
https://doi.org/10.3929/ethz-b-000544180
Rights / license:
Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International
Originally published in:
Bulletin of the London Mathematical Society 54(4), https://doi.org/10.1112/blms.12639
This page was generated automatically upon download from the ETH Zurich Research Collection.
For more information, please consult the Terms of use.
Received: 12 July 2021
Revised: 18 October 2021
Accepted: 14 November 2021
DOI: 10.1112/blms.12639
R E S E A R C H A R T I C L E
Bulletin of the London
Mathematical Society
The proportion of derangements characterizes
the symmetric and alternating groups
Bjorn Poonen1
Kaloyan Slavov2
Abstract
Let 𝐺 be a subgroup of the symmetric group 𝑆𝑛. If the
proportion of fixed-point-free elements in 𝐺 (or a coset)
equals the proportion of fixed-point-free elements in 𝑆𝑛,
then 𝐺 = 𝑆𝑛. The analogue for 𝐴𝑛 holds if 𝑛 ⩾ 7. We give
an application to monodromy groups.
M S C ( 2 0 2 0 )
20B35 (primary), 11A63, 14E20, 14G15, 20B10 (secondary)
1Department of Mathematics,
Massachusetts Institute of Technology,
Cambridge, Massachusetts, USA
2Department of Mathematics, ETH
Zürich, Zurich, Switzerland
Correspondence
Kaloyan Slavov, Department of
Mathematics, ETH Zürich, Rämistrasse
101, 8006 Zürich, Switzerland.
Email: [email protected]
Funding information
National Science Foundation,
Grant/Award Numbers: DMS-1601946,
DMS-2101040; Simons Foundation,
Grant/Award Numbers: #402472,
#550033; SNSF
1
INTRODUCTION
1.1
Derangements in permutation groups
Motivated by an application to monodromy groups, we prove the following.
Theorem 1.1. Let 𝐺 be a subgroup of the symmetric group 𝑆𝑛 for some 𝑛 ⩾ 1. Let 𝐶 be a coset of 𝐺
in 𝑆𝑛. If
|{𝜎 ∈ 𝐶 ∶ 𝜎 has no f ixed points}|
|𝐶|
=
|{𝜎 ∈ 𝑆𝑛 ∶ 𝜎 has no f ixed points}|
|
|𝑆𝑛
,
(1)
then 𝐺 = 𝐶 = 𝑆𝑛.
© 2022 The Authors. Bulletin of the London Mathematical Society is copyright © London Mathematical Society. This is an open access
article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in
any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
Bull. London Math. Soc. 2022;54:1439–1447.
wileyonlinelibrary.com/journal/blms
1439
1440
POONEN and SLAVOV
Elements of 𝑆𝑛 with no fixed points are called derangements. Let 𝐷𝑛 be the number of derange-
ments in 𝑆𝑛. The right side of (1) is
𝐷𝑛
𝑛!
=
𝑛∑
𝑖=0
(−1)𝑖
𝑖!
;
see [10, Example 2.2.1], for instance. When the denominator of 𝐷𝑛∕𝑛! in lowest terms is 𝑛!, the
conclusion of Theorem 1.1 follows immediately, but controlling gcd(𝐷𝑛, 𝑛!) in general is nontrivial.
Our proof requires an irrationality measure for 𝑒, divisibility properties of 𝐷𝑛, and a bound on the
orders of primitive permutation groups.
Remark 1.2. The proof shows also that for 𝑛 ⩾ 5, if 𝐶 is not necessarily a coset but just any subset
of 𝑆𝑛 having the same size as 𝐺, then (1) implies that 𝐺 is 𝐴𝑛 or 𝑆𝑛. In fact, we prove that if a
subgroup 𝐺 of 𝑆𝑛 has order divisible by the denominator of 𝐷𝑛∕𝑛!, then 𝐺 is 𝐴𝑛 or 𝑆𝑛.
Remark 1.3. We also prove an analogue of Theorem 1.1 in which both appearances of 𝑆𝑛 on the
right side of (1) are replaced by the alternating group 𝐴𝑛 for some 𝑛 ⩾ 7; see Theorem 5.1. But there
are counterexamples for smaller alternating groups. For example, the order 10 dihedral group in
𝐴5 has the same proportion of derangements as 𝐴5, namely 4∕10 = 24∕60.
1.2
Application to monodromy
Let 𝔽𝑞 be the finite field of 𝑞 elements. Let 𝑓(𝑇) ∈ 𝔽𝑞[𝑇] be a polynomial of degree 𝑛. Birch and
Swinnerton–Dyer [2] define what it means for 𝑓 to be ‘general’ and estimate the proportion of
field elements in the image of a general 𝑓:
|𝑓(𝔽𝑞)|
𝑞
= 1 −
𝑛∑
𝑖=0
(−1)𝑖
𝑖!
+ 𝑂𝑛(𝑞−1∕2).
More generally, let 𝑓 ∶ 𝑋 → 𝑌 be a degree 𝑛 generically étale morphism of schemes of finite
type over 𝔽𝑞, with 𝑌 geometrically integral. The geometric and arithmetic monodromy groups 𝐺
and 𝐴 are subgroups of 𝑆𝑛 fitting in an exact sequence
1 ⟶ 𝐺 ⟶ 𝐴 ⟶ Gal(𝔽𝑞𝑟 ∕𝔽𝑞) ⟶ 1
for some 𝑟 ⩾ 1; see [4, Section 4] for an exposition. Let 𝐶 be the coset of 𝐺 in 𝐴 mapping to the
Frobenius generator of Gal(𝔽𝑞𝑟 ∕𝔽𝑞). Let 𝑀 be a bound on the geometric complexity of 𝑋 and 𝑌.
Assume that 𝑌(𝔽𝑞) ≠ ∅, which is automatic if 𝑞 is large relative to 𝑀. Then the Lang–Weil bound
implies
|𝑓(𝑋(𝔽𝑞))|
|𝑌(𝔽𝑞)| =
|{𝜎 ∈ 𝐶 ∶ 𝜎 has at least one fixed point}|
|𝐶|
+ 𝑂𝑛,𝑀(𝑞−1∕2);
see [4, Theorem 3], for example. In particular, if 𝐺 = 𝑆𝑛, then
|𝑓(𝑋(𝔽𝑞))|
|𝑌(𝔽𝑞)| = 1 −
𝑛∑
𝑖=0
(−1)𝑖
𝑖!
+ 𝑂𝑛,𝑀(𝑞−1∕2).
(2)
(3)
THE PROPORTION OF DERANGEMENTS CHARACTERIZES THE SYMMETRIC AND ALTERNATING GROUPS
1441
We prove a converse, that an estimate as in (3) on the proportion of points in the image implies
that the geometric monodromy group of 𝑓 is the full symmetric group 𝑆𝑛:
Corollary 1.4. Given 𝑛 and 𝑀, there exists an effectively computable constant 𝑐 = 𝑐(𝑛, 𝑀) such that
for any 𝑓 ∶ 𝑋 → 𝑌 as above, with deg 𝑓 = 𝑛 and the complexities of 𝑋 and 𝑌 bounded by 𝑀, if
|𝑓(𝑋(𝔽𝑞))|
|𝑌(𝔽𝑞)| = 1 −
𝑛∑
𝑖=0
(−1)𝑖
𝑖!
+ 𝜖, where |𝜖| < 1
𝑛!
− 𝑐𝑞−1∕2,
then 𝐺 = 𝑆𝑛.
Proof. Combine (2) and Theorem 1.1.
□
Remark 1.5. We originally proved Corollary 1.4 in order to prove a version of [8, Theorem 1.9],
about specialization of monodromy groups, but later we found a more natural argument.
1.3
Structure of the paper
The proof of Theorem 1.1 occupies the rest of the paper, which is divided in sections according to
the properties of 𝐺. Throughout, we assume that 𝐺, 𝐶, and 𝑛 are such that (1) holds. The cases
with 𝑛 ⩽ 4 can be checked directly, so assume that 𝑛 ⩾ 5 and 𝐺 ≠ 𝑆𝑛.
2
PRIMITIVE PERMUTATION GROUPS
The proportion of derangements in 𝐴𝑛 is given by the inclusion–exclusion formula; it differs
from 𝐷𝑛∕𝑛! by the nonzero quantity ±(𝑛 − 1)∕𝑛!. The proportion for 𝑆𝑛 is the average of the
proportions for 𝐴𝑛 and 𝑆𝑛 − 𝐴𝑛, so the proportion for 𝑆𝑛 − 𝐴𝑛 also differs from 𝐷𝑛∕𝑛!. Thus
𝐺 ≠ 𝐴𝑛.
Suppose that 𝐺 is primitive, 𝑛 ⩾ 5, and 𝐺 ≠ 𝐴𝑛, 𝑆𝑛. The main theorem in [9]1 gives |𝐺| < 4𝑛.
On the other hand, 𝐷𝑛∕𝑛! is close to 1∕𝑒 and hence cannot equal a rational number with small
denominator; this will show that |𝐺| is at least about
𝑛!. These will give a contradiction for large
𝑛. We now make this precise.
√
Let 𝑎 = |{𝜎 ∈ 𝐶 ∶ 𝜎 has no fixed points}| and 𝑏 = |𝐶| = |𝐺|, so 𝑎 ⩽ 𝑏 = |𝐺| < 4𝑛. Then
|
|
|
|
𝑎
𝑏
−
|
|
|
|
1
𝑒
|
|
|
|
𝐷𝑛
𝑛!
=
−
|
|
|
|
1
𝑒
<
1
(𝑛 + 1)!
.
No rational number with numerator ⩽ 4 is within 1∕6! of 1∕𝑒, so 𝑎 ⩾ 5. By the main result of [7]
(see also [1]),
|
|
𝑒 −
|
|
|
|
|
|
𝑏
𝑎
>
log log 𝑎
3𝑎2 log 𝑎
.
1 This is independent of the classification of finite simple groups. Using the classification, [6] gives better bounds.
1442
POONEN and SLAVOV
Combining the two displayed inequalities yields
1
(𝑛 + 1)!
>
|
|
|
|
𝑎
𝑏
−
|
|
|
|
1
𝑒
=
|
|
𝑒 −
|
|
𝑎
𝑏𝑒
|
|
|
|
𝑏
𝑎
>
1
𝑏𝑒
⋅ log log 𝑎
3𝑎 log 𝑎
>
log log 4𝑛
3𝑒(4𝑛)2 log 4𝑛
;
(4)
the last step uses that 𝑎, 𝑏 < 4𝑛 and that log log 𝑥
𝑥 log 𝑥
𝑛 ⩽ 41.
is decreasing for 𝑥 ⩾ 5. Inequality (4) implies
Let 𝑑𝑛 be the denominator of the rational number 𝐷𝑛
. Then 𝑑𝑛 ∣ 𝑏, so 𝑑𝑛 ⩽ 𝑏 < 4𝑛. For
11 < 𝑛 ⩽ 41, the inequality 𝑑𝑛 < 4𝑛 fails. For 𝑛 ⩽ 11, a Magma computation [5] shows that there
are no degree 𝑛 primitive subgroups 𝐺 ≠ 𝐴𝑛, 𝑆𝑛 for which 𝑑𝑛 ∣ 𝑏.
= 𝑎
𝑏
𝑛!
3
IMPRIMITIVE BUT TRANSITIVE PERMUTATION GROUPS
Suppose that 𝐺 is imprimitive but transitive. Then 𝐺 preserves a partition of {1, … , 𝑛} into 𝑙 subsets
of equal size 𝑘, for some 𝑘, 𝑙 ⩾ 2 with 𝑘𝑙 = 𝑛. The subgroup of 𝑆𝑛 preserving such a partition has
order (𝑘!)𝑙𝑙! (it is a wreath product 𝑆𝑘 ≀ 𝑆𝑙). Thus |𝐺| divides (𝑘!)𝑙𝑙!.
|𝐺| = 𝐷𝑛
For a prime 𝑝, let 𝜈𝑝 denote the 𝑝-adic valuation. Since 𝑎
, every prime 𝑝 ∤ 𝐷𝑛 satisfies
𝜈𝑝(𝑛!) ⩽ 𝜈𝑝(|𝐺|) ⩽ 𝜈𝑝((𝑘!)𝑙𝑙!) ⩽ 𝜈𝑝(𝑛!). Thus for every prime 𝑝 ∤ 𝐷𝑛, the inequality 𝜈𝑝((𝑘!)𝑙𝑙!) ⩽
𝜈𝑝(𝑛!) is an equality. The third of the three following lemmas will prove that this is impossible for
𝑛 ⩾ 5.
𝑛!
Lemma 3.1. Let 𝑘, 𝑙 ⩾ 2 and let 𝑝 be a prime. The inequality
𝜈𝑝((𝑘!)𝑙𝑙!) ⩽ 𝜈𝑝((𝑘𝑙)!)
(5)
is an equality if and only if at least one of the following holds:
∙ 𝑘 is a power of 𝑝;
∙ there are no carry operations in the 𝑙-term addition 𝑘 + ⋯ + 𝑘 when 𝑘 is written in base 𝑝 (in
particular, 𝑙 < 𝑝).
Proof. Let 𝑠𝑝(𝑘) denote the sum of the 𝑝-adic digits of a positive integer 𝑘; then 𝜈𝑝(𝑘!) =
Thus equality in (5) is equivalent to equality in
𝑙 + 𝑠𝑝(𝑘𝑙) ⩽ 𝑙𝑠𝑝(𝑘) + 𝑠𝑝(𝑙).
We always have
𝑙 + 𝑠𝑝(𝑘𝑙) ⩽ 𝑙 + 𝑠𝑝(𝑘)𝑠𝑝(𝑙) ⩽ 𝑙𝑠𝑝(𝑘) + 𝑠𝑝(𝑙);
the first follows from 𝑠𝑝(𝑘𝑙) ⩽ 𝑠𝑝(𝑘)𝑠𝑝(𝑙), and the second is simply
(𝑠𝑝(𝑘) − 1)(𝑙 − 𝑠𝑝(𝑙)) ⩾ 0.
Thus equality in (6) is equivalent to equality in both inequalities of (7).
𝑘−𝑠𝑝(𝑘)
𝑝−1
.
(6)
(7)
THE PROPORTION OF DERANGEMENTS CHARACTERIZES THE SYMMETRIC AND ALTERNATING GROUPS
1443
The second inequality of (7) is an equality if and only if either 𝑘 is a power of 𝑝 or 𝑙 < 𝑝; in
each case, we must check when equality holds in the first inequality (7), that is, when 𝑠𝑝(𝑘𝑙) =
𝑠𝑝(𝑘)𝑠𝑝(𝑙). If 𝑘 is a power of 𝑝, then it holds. If 𝑙 < 𝑝, then it holds if and only if 𝑠𝑝(𝑘𝑙) = 𝑙𝑠𝑝(𝑘),
which holds if and only if there are no carry operations in the 𝑙-term addition 𝑘 + ⋯ + 𝑘 when 𝑘
□
is written in base 𝑝.
The following lemma will help us produce primes 𝑝 not dividing 𝐷𝑛.
Lemma 3.2. For 0 ⩽ 𝑚 ⩽ 𝑛, we have 𝐷𝑛
≡ (−1)𝑛−𝑚𝐷𝑚 (mod 𝑛 − 𝑚). In particular,
𝐷𝑛
≡ ±1
(mod 𝑛)
𝐷𝑛
≡ ±1
(mod 𝑛 − 2)
𝐷𝑛
≡ ±2
(mod 𝑛 − 3).
Proof. Reduce each term in 𝐷𝑛 modulo 𝑛 − 𝑚; most of them are 0.
(8)
(9)
(10)
□
Lemma 3.3. Let 𝑘, 𝑙 ⩾ 2. Set 𝑛 = 𝑘𝑙 and assume 𝑛 > 4. Then there exists a prime 𝑝 ∤ 𝐷𝑛 such that
𝜈𝑝((𝑘!)𝑙𝑙!) < 𝜈𝑝(𝑛!).
Proof. Case 1. 𝑙 ⩾ 3 and 𝑛 − 2 is not a power of 2.
Let 𝑝 ⩾ 3 be a prime with 𝑝 ∣ 𝑛 − 2. By (9), 𝑝 ∤ 𝐷𝑛, so 𝜈𝑝((𝑘!)𝑙𝑙!) = 𝜈𝑝(𝑛!). Apply Lemma 3.1.
If 𝑘 is a power of 𝑝, then 𝑝 divides 𝑘, which divides 𝑛, so 𝑝 ∣ 𝑛 − (𝑛 − 2) = 2, contradicting 𝑝 ⩾
3. Otherwise, there are no carry operations in the 𝑙-term addition 𝑘 + ⋯ + 𝑘 in base 𝑝. This is
impossible because the last digit of 𝑛 is 2 (since 𝑝 ∣ 𝑛 − 2 and 𝑝 ⩾ 3) and 𝑙 ⩾ 3.
Case 2. 𝑙 = 2.
Then 2 ∣ 𝑛. By (8), 2 ∤ 𝐷𝑛. By Lemma 3.1, 𝑘 is a power of 2 (since 𝑙 < 2 is violated). Thus 𝑛 = 2𝑘
is a power of 2.
Since 𝑛 ⩾ 5, there exists a prime 𝑝 ∣ 𝑛 − 3. Since 𝑛 is a power of 2, this implies 𝑝 ⩾ 5. By (10),
𝑝 ∤ 𝐷𝑛. Apply Lemma 3.1. Note that 𝑘 is not a power of 𝑝, since 𝑘 is a power of 2 and 𝑝 ≠ 2.
Therefore, there are no carry operations in 𝑘 + 𝑘 = 𝑛, so the last digit of 𝑛 is even. But 𝑝 ∣ 𝑛 − 3
and 𝑝 ⩾ 5, so the last digit of 𝑛 is 3.
Case 3. 𝑙 = 3 and 𝑛 − 2 is a power of 2.
Then 3 ∣ 𝑛. By (8), 3 ∤ 𝐷𝑛. By Lemma 3.1, 𝑘 must be a power of 3 (since 𝑙 < 3 is violated). Then
𝑛 = 3𝑘 is a power of 3, contradicting the fact that 𝑛 is even.
Case 4. 𝑙 > 3 and 𝑛 − 2 is a power of 2.
In particular, 𝑛 = 𝑘𝑙 > 6. Then 𝑛 − 3 is not a power of 3, because otherwise we would have a
solution to 3𝑢 = 2𝑣 − 1 with 𝑢 > 1, whereas the only solution in positive integers is (𝑢, 𝑣) = (1, 2)
(proof: 3 ∣ 2𝑣 − 1, so 𝑣 is even, so 2𝑣∕2 − 1 and 2𝑣∕2 + 1 are powers of 3 that differ by 2, so they are
1 and 3).
Let 𝑝 ≠ 3 be a prime divisor of 𝑛 − 3. Then 𝑝 ⩾ 5. Apply (10) and Lemma 3.1. If 𝑘 is a power of
𝑝, then 𝑝 ∣ 𝑛, so 𝑝 ∣ 𝑛 − (𝑛 − 3) = 3, contradicting 𝑝 ≠ 3. Therefore, there are no carry operations
1444
POONEN and SLAVOV
in the 𝑙-term addition 𝑘 + ⋯ + 𝑘. This is impossible, since the last digit of 𝑘𝑙 is 3 (since 𝑝 ∣ 𝑛 − 3
□
and 𝑝 ⩾ 5) and 𝑙 > 3.
4
INTRANSITIVE PERMUTATION GROUPS
Suppose that 𝐺 is intransitive. Then 𝐺 embeds in 𝑆𝑢 × 𝑆𝑣 ⊂ 𝑆𝑛 for some 𝑢, 𝑣 ⩾ 1 with 𝑢 + 𝑣 = 𝑛.
Consider a prime 𝑝 ∣ 𝑛. By (8), 𝑝 ∤ 𝐷𝑛. Then, analogously to the second paragraph of Section 3,
𝜈𝑝(𝑛!) ⩽ 𝜈𝑝(|𝐺|) ⩽ 𝜈𝑝(𝑢! 𝑣!) ⩽ 𝜈𝑝(𝑛!), so 𝜈𝑝(𝑢!) + 𝜈𝑝(𝑣!) = 𝜈𝑝(𝑛!); equivalently, 𝑠𝑝(𝑢) + 𝑠𝑝(𝑣) =
𝑠𝑝(𝑛). So there are no carry operations in 𝑢 + 𝑣. Let 𝑒 = 𝜈𝑝(𝑛), so the last 𝑒 base 𝑝 digits of 𝑛
are zero; then the same holds for 𝑢 and 𝑣. In other words, 𝑝𝑒 ∣ 𝑢, 𝑣 as well. Since this holds for
each 𝑝 ∣ 𝑛, we conclude that 𝑛 ∣ 𝑢, 𝑣. This contradicts 0 < 𝑢, 𝑣 < 𝑛.
This completes the proof of Theorem 1.1.
5
ALTERNATING GROUP
Theorem 5.1. Let 𝐺 be a subgroup of the symmetric group 𝑆𝑛 for some 𝑛 ⩾ 7. Let 𝐶 be a coset of 𝐺
in 𝑆𝑛 having the same proportion of fixed-point-free elements as 𝐴𝑛. Then 𝐺 = 𝐴𝑛.
Remark 5.2. For 𝑛 ⩽ 6, the subgroups of 𝑆𝑛 other than 𝐴𝑛 for which some coset has the same
proportion as 𝐴𝑛, up to conjugacy, are:
∙ the order 4 subgroup of 𝑆4 generated by (1423) and (12)(34);
∙ the order 4 subgroup of 𝑆4 generated by (34) and (12)(34);
∙ the order 8 subgroup of 𝑆4;
∙ the subgroups of 𝑆5 of order 5, 10, or 20;
∙ the order 36 subgroup of 𝑆6 generated by (1623)(45), (12)(36), (124)(365), and (142)(365);
∙ the order 36 subgroup of 𝑆6 generated by (13)(25)(46), (14)(36), (154)(236), and (145)(236).
The proof of Theorem 5.1 follows the proof of Theorem 1.1; we highlight only the differences.
The proportion of fixed-point-free elements in 𝐴𝑛 is 𝐸𝑛∕𝑛!, where 𝐸𝑛
≔ 𝐷𝑛 + (−1)𝑛−1(𝑛 − 1).
5.1
Primitive permutation groups
Suppose 𝐺 ≠ 𝐴𝑛. The first paragraph of Section 2 shows that 𝐺 ≠ 𝑆𝑛. For 7 ⩽ 𝑛 ⩽ 13, we use
Magma to check Theorem 5.1 for each primitive subgroup of 𝑆𝑛. So assume 𝑛 ⩾ 14. Define 𝑎 and
𝑏 as in Section 2. We have
|
|
|
|
𝑎
𝑏
−
|
|
|
|
1
𝑒
=
|
|
|
|
𝐸𝑛
𝑛!
−
|
|
|
|
1
𝑒
⩽
|
|
|
|
𝐸𝑛 − 𝐷𝑛
𝑛!
|
|
|
|
+
|
|
|
|
𝐷𝑛
𝑛!
−
|
|
|
|
1
𝑒
<
𝑛 − 1
𝑛!
+
1
(𝑛 + 1)!
=
𝑛2
(𝑛 + 1)!
.
No 𝑎∕𝑏 with 𝑎 < 5 is within 152∕16! of 1∕𝑒, so 𝑎 ⩾ 5. Inequality (4) with 1∕(𝑛 + 1)! replaced by
𝑛2∕(𝑛 + 1)! implies 𝑛 ⩽ 49.
Let 𝑒𝑛 be the denominator of 𝐸𝑛∕𝑛!, so 𝑒𝑛 divides |𝐺|, which is less than 4𝑛. But for 13 < 𝑛 ⩽ 49,
the inequality 𝑒𝑛 < 4𝑛 fails.
THE PROPORTION OF DERANGEMENTS CHARACTERIZES THE SYMMETRIC AND ALTERNATING GROUPS
1445
5.2
blocks of equal size
Imprimitive permutation groups that preserve a partition into
To rule out imprimitive permutation groups that preserve a partition into 𝑙 blocks of size 𝑘, we
argue as in Section 3, but with Lemma 3.3 replaced by the following.
Lemma 5.3. Let 𝑘, 𝑙 ⩾ 2. Set 𝑛 = 𝑘𝑙 and assume 𝑛 > 6. Then there exists a prime 𝑝 ∤ 𝐸𝑛 such that
𝜈𝑝((𝑘!)𝑙𝑙!) < 𝜈𝑝(𝑛!).
Proof of Lemma 5.3. For each integer 𝑛 ∈ (6, 30], we check directly that there exists a prime 𝑝 ∈
(𝑛∕2, 𝑛] such that 𝑝 ∤ 𝐸𝑛. Assume from now on that 𝑛 > 30.
Suppose the statement is false. Then whenever a prime 𝑝 satisfies 𝑝 ∤ 𝐸𝑛, (5) is an equality and
Lemma 3.1 applies.
By using 𝐷𝑛
≡ (−1)𝑛−𝑠𝐷𝑠 (mod 𝑛 − 𝑠) and 𝐸𝑛 = 𝐷𝑛 + (−1)𝑛−1(𝑛 − 1), we obtain
𝐸𝑛
𝐸𝑛
𝐸𝑛
𝐸𝑛
≡ 2(−1)𝑛
(mod 𝑛)
≡ 4(−1)𝑛−1
(mod 𝑛 − 3)
≡ 6(−1)𝑛
(mod 𝑛 − 4)
≡ (−1)𝑛−124 × 3
(mod 𝑛 − 5)
(11)
(12)
(13)
(14)
Case 1. 𝑛 − 4 is a power of 2.
Then 𝑛 − 3 is not a power of 3 because otherwise, we have a solution to 3𝑢 − 1 = 2𝑣 with 𝑢 ⩾ 3;
working modulo 4 shows that 𝑢 is even, and factoring the left side leads to a contradiction. Let
𝑝 ≠ 3 be a prime with 𝑝 ∣ 𝑛 − 3. Since 𝑛 − 3 is odd, 𝑝 ⩾ 5. By (12), 𝑝 ∤ 𝐸𝑛, so we have one of the
conclusions of Lemma 3.1.
If 𝑘 is a power of 𝑝, then 𝑝 ∣ 𝑘 ∣ 𝑛, which, combined with 𝑝 ∣ 𝑛 − 3 gives 𝑝 = 3, a contradiction.
Suppose that there is no carry in 𝑘 + ⋯ + 𝑘 (𝑙 terms). This sum has last digit 3 in base 𝑝, so
𝑙 = 3, so 3 ∣ 𝑛, and hence 3 ∤ 𝐸𝑛 by (11). Apply Lemma 3.1 for the prime 3. Since 𝑙 < 3 is violated,
we deduce that 𝑘 is a power of 3. Then 𝑛 = 𝑘𝑙 is also a power of 3, but this contradicts the fact that
𝑛 is even.
Case 2. 𝑛 − 3 is a power of 2 and 𝑙 ≠ 2, 4.
Then 𝑛 − 4 is odd and is not a power of 3. Let 𝑝 ≠ 3 be a prime with 𝑝 ∣ 𝑛 − 4. Then 𝑝 ⩾ 5, so
𝑝 ∤ 𝐸𝑛 by (13). If 𝑘 is a power of 𝑝, then 𝑝 ∣ 𝑘 ∣ 𝑛, which contradicts 𝑝 ∣ 𝑛 − 4 since 𝑝 ⩾ 5. If there
are no carry operations in the 𝑙-term addition 𝑘 + ⋯ + 𝑘 (which has last digit 4 in base 𝑝), then
𝑙 = 2 or 𝑙 = 4, contrary to assumption.
Case 3. 𝑙 = 3.
Then 3 ∣ 𝑛, hence 3 ∤ 𝐸𝑛 by (11). Apply Lemma 3.1 for the prime 3. Since 3 < 𝑙 is violated, 𝑘 is a
power of 3. Then 𝑛 = 𝑘𝑙 is also a power of 3. Then 𝑛 − 4 is odd and not divisible by 3. Let 𝑞 be a
prime with 𝑞 ∣ 𝑛 − 4. Then 𝑞 ⩾ 5, and hence 𝑞 ∤ 𝐸𝑛 by (13). Since 𝑘 is a power of 3, it is not a power
of 𝑞. So there is no carry in 𝑘 + 𝑘 + 𝑘 in base 𝑞. But this sum has last digit 4 in base 𝑞, which is
a contradiction.
1446
POONEN and SLAVOV
Case 4. 𝑙 ≠ 2, 4.
By the previous cases, we may assume in addition that 𝑛 − 4 and 𝑛 − 3 are not powers of 2 and
𝑙 ≠ 3.
Let 𝑝 ≠ 2 be a prime with 𝑝 ∣ 𝑛 − 3. Then 𝑝 ∤ 𝐸𝑛 by (12). Since the 𝑙-term addition 𝑘 + ⋯ + 𝑘
has last digit 3 and 𝑙 ≠ 3, there is some carry. Therefore 𝑘 is a power of 𝑝. Then 𝑝 ∣ 𝑘 ∣ 𝑛, which,
combined with 𝑝 ∣ 𝑛 − 3, gives 𝑝 = 3. In particular, 3 ∣ 𝑛.
Let 𝑞 ≠ 2 be a prime with 𝑞 ∣ 𝑛 − 4. Since 3 ∣ 𝑛, we have 𝑞 ≠ 3 so 𝑞 ⩾ 5. By (13), 𝑞 ∤ 𝐸𝑛. If 𝑘 is
a power of 𝑞, then 𝑞 ∣ 𝑛, hence 𝑞 ∣ 4 — contradiction. Therefore there is no carry in the 𝑙-term
addition 𝑘 + ⋯ + 𝑘 in base 𝑞. This sum has last digit 4 and 𝑙 ≠ 2, 4, so this case is impossible.
Case 5. 𝑙 = 2 or 𝑙 = 4.
Then 𝑛 is even, so 𝑛 − 3 and 𝑛 − 5 are odd.
Subcase 5.1: 𝑛 − 3 is not a power of 3.
Let 𝑝 ≠ 3 be a prime such that 𝑝 ∣ 𝑛 − 3. Then 𝑝 ⩾ 5 and 𝑝 ∤ 𝐸𝑛 by (12). If 𝑘 is a power of 𝑝,
then 𝑝 ∣ 𝑘 ∣ 𝑛, giving 𝑝 = 3, which is a contradiction. However, there is carry in the 𝑙-term addition
𝑘 + ⋯ + 𝑘 because the sum has last digit 3, and 𝑙 is 2 or 4.
Subcase 5.2: 𝑛 − 3 is a power of 3 but 𝑛 − 5 is not a power of 5.
Let 𝑝 ≠ 5 be a prime with 𝑝 ∣ 𝑛 − 5. Then 𝑝 ⩾ 7 and we apply the argument of subcase 5.1: an
𝑙-term sum 𝑘 + ⋯ + 𝑘 cannot have last digit 5 in base 𝑝.
Subcase 5.3: 𝑛 − 3 = 3𝑎 and 𝑛 − 5 = 5𝑏 for some 𝑎, 𝑏 ⩾ 1.
Then 3𝑎 − 5𝑏 = 2, so 𝑎 = 3 and 𝑏 = 2 by [3, Theorem 4.06]. This contradicts 𝑛 > 30.
□
5.3
Intransitive subgroups
As in Section 4, 𝐺 embeds in 𝑆𝑢 × 𝑆𝑣 ⊂ 𝑆𝑛 for some 𝑢, 𝑣 ⩾ 1 with 𝑢 + 𝑣 = 𝑛. Write 𝑛 = 2𝑠𝑚, where
𝑠 ⩾ 0 and 2 ∤ 𝑚. The argument in Section 4 for odd 𝑝 with 𝐸𝑛 in place of 𝐷𝑛 and (11) in place of
(8) implies 𝑚 ∣ 𝑢, 𝑣. Thus 𝑠 ⩾ 1.
If 𝑠 = 1, then 𝑛 = 2𝑚, so 𝑢 = 𝑣. This case is covered in Section 5.2.
Suppose that 𝑠 ⩾ 2. Then 4 ∣ 𝑛, so (11) implies that 𝐸𝑛∕2 is odd. Using 𝑎
, we obtain
𝜈2(𝑛!∕2) ⩽ 𝜈2(|𝐺|) ⩽ 𝜈2(𝑢!𝑣!) ⩽ 𝜈2(𝑛!). If the last inequality is an equality, then the same argu-
ment used in Section 4 shows that 𝜈2(𝑢) = 𝜈2(𝑣) = 𝜈2(𝑛); combining this with 𝑚 ∣ 𝑢, 𝑣 shows
that 𝑛 ∣ 𝑢, 𝑣, a contradiction. Therefore the first two inequalities must be equalities, so 𝜈2(𝑢!𝑣!) =
𝜈2(𝑛!) − 1; equivalently, 𝑠2(𝑢) + 𝑠2(𝑣) = 𝑠2(𝑛) + 1. This means there is exactly one carry opera-
tion in 𝑢 + 𝑣 in base 2. This is possible only when 2𝑠−1 ∣ 𝑢, 𝑣. Also, 𝑚 ∣ 𝑢, 𝑣, so 𝑛∕2 ∣ 𝑢, 𝑣, so again
𝑢 = 𝑣, and this case is covered in Section 5.2.
|𝐺| = 𝐸𝑛∕2
𝑛!∕2
A C K N O W L E D G E M E N T S
We thank Andrew Sutherland for useful discussions concerning Section 2 and specifically for
drawing our attention to [7]. We thank Michael Bennett and Samir Siksek for suggesting refer-
ences for the solution of 3𝑎 − 5𝑏 = 2. We also thank the referees for comments.
B.P. was supported in part by National Science Foundation grants DMS-1601946 and DMS-
2101040 and Simons Foundation grants #402472 and #550033. K.S. was supported by NCCR
SwissMAP of the SNSF.
Open access funding provided by Eidgenossische Technische Hochschule Zurich.
THE PROPORTION OF DERANGEMENTS CHARACTERIZES THE SYMMETRIC AND ALTERNATING GROUPS
1447
J O U R N A L I N F O R M A T I O N
The Bulletin of the London Mathematical Society is wholly owned and managed by the London
Mathematical Society, a not-for-profit Charity registered with the UK Charity Commission.
All surplus income from its publishing programme is used to support mathematicians and
mathematics research in the form of research grants, conference grants, prizes, initiatives for
early career researchers and the promotion of mathematics.
R E F E R E N C E S
1. H. Alzer, On rational approximation to e, J. Number Theory 68 (1998), no. 1, 57–62. MR1492888.
2. B. J. Birch and H. P. F. Swinnerton-Dyer, Note on a problem of Chowla, Acta Arith. 5 (1959), 417–423. MR113844.
3. J. L. Brenner and L. L. Foster, Exponential Diophantine equations, Pacific J. Math. 101 (1982), no. 2, 263–301.
MR675401.
4. A. Entin, Monodromy of hyperplane sections of curves and decomposition statistics over finite fields, Int. Math.
Res. Not. 2021 (2021), no. 14, 10409–10441. MR4285725.
5. W. Bosma, J. Cannon, and C. Playoust, The Magma algebra system. I. The user language, J. Symbolic Comput.
24 (1997), no. 3-4, 235–265. Computational algebra and number theory (London, 1993). Magma is available at
http://magma.maths.usyd.edu.au/magma/. MR1484478.
6. A. Maróti, On the orders of primitive groups, J. Algebra 258 (2002), no. 2, 631–640. MR1943938.
7. T. Okano, A note on the rational approximations to e, Tokyo J. Math. 15 (1992), no. 1, 129–133. MR1164191.
8. B. Poonen and K. Slavov, The exceptional locus in the Bertini irreducibility theorem for a morphism, Int. Math.
Res. Not. 2022 (2022), no. 6, 4503–4513. MR4391895.
9. C. E. Praeger and J. Saxl, On the orders of primitive permutation groups, Bull. Lond. Math. Soc. 12 (1980), no. 4,
303–307. MR576980.
10. R. P. Stanley, Enumerative combinatorics. Volume 1, 2nd ed., Cambridge Stud. Adv. Math., vol. 49, Cambridge
Univ. Press, Cambridge, 2012. MR2868112.
| null |
10.1007_s00440-023-01195-8.pdf
|
Data availability Data sharing not applicable to this article as no datasets were generated or analysed during
the current study.
|
Data availability Data sharing not applicable to this article as no datasets were generated or analysed during
|
Probability Theory and Related Fields
https://doi.org/10.1007/s00440-023-01195-8
Biased 2 × 2 periodic Aztec diamond and an elliptic curve
Alexei Borodin1 · Maurice Duits2
Received: 15 April 2022 / Revised: 15 December 2022 / Accepted: 14 January 2023
© The Author(s) 2023
Abstract
We study random domino tilings of the Aztec diamond with a biased 2 × 2 periodic
weight function and associate a linear flow on an elliptic curve to this model. Our main
result is a double integral formula for the correlation kernel, in which the integrand is
expressed in terms of this flow. For special choices of parameters the flow is periodic,
and this allows us to perform a saddle point analysis for the correlation kernel. In these
cases we compute the local correlations in the smooth disordered (or gaseous) region.
The special example in which the flow has period six is worked out in more detail, and
we show that in that case the boundary of the rough disordered region is an algebraic
curve of degree eight.
Mathematics Subject Classification Primary 60D05 · Secondary 60G55
Contents
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
1 Introduction .
.
.
.
.
2 Preliminaries .
.
.
.
.
.
3 Main results
.
.
.
4 The flow .
.
.
. .
5 Proofs of the main results
.
.
.
.
A Example: torsion point of order six .
.
.
.
.
B Computation of torsion points .
.
. .
.
. .
References .
. .
. .
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
B Maurice Duits
[email protected]
Alexei Borodin
[email protected]
1 Department of Mathematics, Massachusetts Institute of Technology, 77 Massachusetts Ave.,
Cambridge, MA 02139, USA
2 Department of Mathematics, Royal Institute of Technology, Lindstedtsvägen 25, 10044 Stockholm,
Sweden
123
A. Borodin, M. Duits
1 Introduction
Domino tilings of the Aztec diamond, originally introduced in [12], form a popular
arena for various interesting phenomena of integrable probability. A domino tiling of
the Aztec diamond can be viewed as a perfect matching, also called dimer configura-
tion, on the Aztec diamond graph. This is a particular bipartite subgraph of the square
lattice (cf. Fig. 2). By putting weights on the edges of the Aztec diamond graph, one
defines a probability measure on the set of all perfect matchings, and hence all domino
tilings, by saying that the probability of having a particular matching is proportional to
the product of the weights of the edges in that matching. In recent years, several works
have appeared on domino tilings of the Aztec diamond where the weights are doubly
periodic. That is, they are periodic in two independent directions, and we will use the
notation k × (cid:2) to indicate that they are k-periodic in one direction and (cid:2)-periodic in
the other. In this paper, we will study a particular example of a 2 × 2 doubly periodic
weighting that is a generalization of the model studied in [1, 2, 7, 8, 11, 20]. The dif-
ference is that we introduce an extra parameter that induces a bias towards horizontal
dominos, and we refer to this model as the biased 2 × 2 periodic Aztec diamond. The
model considered in [1, 2, 7, 8, 11, 20] will be referred to as the unbiased 2×2 periodic
Aztec diamond.
Doubly periodic weightings lead to rich behavior when the size of the Aztec dia-
mond becomes large. The Aztec diamond can be partitioned into three regions: frozen,
rough disordered (or liquid) and smooth disordered (or gaseous). They are character-
ized by the different local limiting Gibbs measures that one expects in these regions
[22]. The difference between the smooth and disordered regions is that the dimer-
dimer correlations decay exponentially with their distance in the smooth disordered
region and polynomially in the rough region. The three regions are clearly visible in
Fig. 1 where we have plotted a sample of our model for a large Aztec diamond.
From general arguments, that go back to [21], we know that the correlation functions
in our model are determinantal. In order to perform a rigorous asymptotic study,
one aims to find an expression for the correlation kernel that is amenable for an
asymptotic analysis. For the unbiased 2 × 2 periodic Aztec diamond, a double integral
representation was first found in [7] (more precisely, they were able to find the inverse
Kasteleyn matrix [21]). Based on this expression, the boundary between the smooth
and rough disordered region has been studied extensively in [1, 2, 20]. Unfortunately,
it is not obvious how the expression in [7] extends to the biased generalization that we
consider in this paper. Instead, we follow the approach of [5].
In [5] the authors studied probability measures on particle configurations given
by products of minors of block Toeplitz matrices. The biased 2 × 2 periodic Aztec
diamond can be viewed as a special case of such a probability measure. The main
result of [5] is an explicit double integral formula for the correlation kernel, provided
one can find a Wiener–Hopf factorization for the product of the matrix-valued symbols
for the block Toeplitz matrices. That Wiener–Hopf factorization can in principle be
found by carrying out an iterative procedure, in which the total number of iterations
is of the same order as the size of the Aztec diamond. In certain special cases, such as
the unbiased 2 × 2 periodic Aztec diamond [5] and a family of 2 × k periodic weights
[4], the procedure is periodic, and after a few iterations one ends up with the same
123
Biased 2 × 2 periodic Aztec diamond...
parameters that one started with. This means that the Wiener–Hopf factorization has
a rather simple form, and after inserting that expression in the double integral formula
one obtains a suitable starting point for a saddle point analysis [5, 11]. However,
generically, the iteration in [5] is too complicated to find simple expressions for the
Wiener–Hopf factorization, and other ideas are needed.
The biased 2 × 2 periodic Aztec diamond is the simplest doubly periodic case in
which it is difficult to trace the flow in [5]. Our first main result is that the Wiener–
Hopf factorization can alternatively be computed by following a linear flow on an
explicit elliptic curve. This flow is rather simple and consists of repeatedly adding
a particular point on the elliptic curve. For generic parameters, one expects the flow
to be ergodic, but for special choices the flow will be periodic. We will identify a
few explicit examples of these periodic cases, and perform an asymptotic study in the
smooth disordered region for the general periodic situation.
The reason why the iterative procedure in our case is linearizable on an elliptic
curve can be traced back to [26]. In that work it was shown how an isospectral flow on
certain quadratic matrix polynomials, obtained by repeatedly moving the right divisor
of the polynomial with a given spectrum to the left side, is linearizable on the Jacobian
(or the Prym variety) of the corresponding spectral curve. The main goal of [26] was
to describe the dynamics of certain discrete analogs of classical integrable systems in
terms of Abelian functions. Some of the key ideas used in that work had previously
originated in constructing the so-called finite gap solutions of integrable PDEs, see
their book-length exposition [3] with historic notes and references therein. The matrix
case, which was most relevant for [26], had been originally developed in [9, 10, 16,
23, 24].
While our situation does not exactly fit into the formalism of [26], similar ideas
do apply, and they led us to the linearization. We hope that they will also help with
studying more general tiling models.
To conclude, let us mention that in [11] it was shown that the double periodicity
leads to matrix-valued orthogonal polynomials. For the unbiased 2 × 2 periodic Aztec
diamond, these matrix-valued orthogonal polynomials have a particularly simple struc-
ture. Somewhat surprisingly, they even have explicit integral expressions that lead to
an explicit double integral representation for the correlation kernel. The expression in
[11] was re-derived in [5]. For the biased model it is interesting to see what the flow on
the elliptic curve implies for the matrix-valued orthogonal polynomials, and if explicit
expressions can be given in general and/or for the periodic case. Furthermore, it is
interesting to compare our results with [6], in which matrix orthogonal polynomials
were studied using an abelianization based on the spectral curve for the orthogonality
weight.
2 Preliminaries
In this section we will introduce the dimer model that we are interested in, discuss
several standard different representations from the literature and recall the determi-
nantal structure of the correlation functions for a corresponding point processes. In
our discussion we repeat necessary definitions from earlier works, in particular of [5,
123
A. Borodin, M. Duits
Fig. 1 A sampling of the biased
doubly periodic for a large Aztec
diamond. The West and South
dominos are colored yellow, and
the North and East dominos are
colored blue. The three different
regions are clearly visible, with
the smooth disordered region in
the middle, surrounded by the
rough disordered region and
frozen regions in the corners
(colour figure online)
Fig. 2 The left picture is the bipartite graph GN , with N = 4, and the right picture is a perfect mathching
of GN
11, 17, 19], and we will make specific references to those works at several places to
refer the reader for more details. We refer to [15] for a general introduction to random
tilings.
2.1 A doubly periodic dimer model
For N ∈ N define a bipartite graph GN = (BN ∪ WN , EN ), with black vertices
(cid:4)
(cid:2)(cid:3)
(cid:5)
BN =
1
2
− N + j + k, − 1
2
− j + k
| j = 0, . . . , N − 1,
k = 0, . . . , N
,
and white vertices
1
2
− N + j + k, 1
2
− j + k
(cid:4)
| j = 0, . . . , N ,
(cid:5)
k = 0, . . . , N − 1
,
(cid:2)(cid:3)
WN =
123
Biased 2 × 2 periodic Aztec diamond...
Fig. 3 The weights on the edges of GN
and with edges EN between black and white vertices that are neighbors in the lattice
graph (i.e., that have a difference of (±1, 0) or (0, ±1)). This gives the graph on the
left of Fig. 2. The picture on the right of Fig. 2 is a perfect matching of this bipartite
graph, also called a dimer configuration. A dimer model is a probability distribution
on the space of all perfect matchings M of this graph GN such that the probability of
a particular matching M is proportional to
P(M) ∼
(cid:6)
e∈M
w(e),
where w : E → (0, ∞) is a weight function.
In this paper, we will consider the weight functions defined as is shown in Fig. 3.
There are two parameters α, a ∈ (0, 1]. The vertical and horizontal edges with a black
vertex on top or on the right all have weights a and 1, respectively. For vertical edges
with a black vertex on the bottom and horizontal edges with a black vertex on the left,
the weight depends on the coordinates of that black vertex. These weights are given by
aα and α, or by a/α and 1/α, depending on the coordinates of the black vertex in that
+ k for an even k, then the weights
edge. If the vertical coordinate of that vertex is 1
2
+ k for an odd k, then we
are aα and α. If the vertical coordinate of that vertex is 1
2
have the same weight, but with α replaced by 1/α. The distribution of the weights
is thus two periodic in two different directions; edges whose coordinates differ by a
multiple of (2, 2) or (2, −2) have the same weight.
Note that only the parameter α is responsible for the double periodicity. Indeed, for
α = 1 the weights no longer depend on the position of the black vertex in the center
in Fig. 3. We will be particularly interested in the doubly periodic situation and thus
in the case 0 < α < 1. The effect of the extra parameter a is that all the vertical edges
are given an extra factor a. If 0 < a < 1, this makes them less likely, and the model is
biased towards horizontal edges. As we will see, adding this parameter has a profound
effect on the integrable structure of this model. Moreover, we will see that the special
case a = 1, studied by several authors [5, 8, 11], is a very particular point.
An alternative way of representing matchings is by drawing dominos. Indeed, each
matching is equivalent to a domino tiling by drawing rectangles around the matched
vertices as is shown in Fig. 4. The dominos tile a planar region known as the Aztec
diamond. We distinguish between four different types of dominos called the West,
East, North and South dominos. The West dominos are the vertical dominos with a
black vertex on the bottom, the East dominos are the vertical dominos with a black
123
A. Borodin, M. Duits
Fig. 4 The right picture is the domino representation of the dimer configuration on the left
Fig. 5 The DR paths on a domino tiling
vertex on the top, the North dominos are the horizontal dominos with a black vertex
on the right and, finally, the South dominos are the horizontal dominos with a black
vertex on the left. In Fig. 4 these four types of dominos are the furthermost ones in the
corresponding corners.
Note that the weighting that we will consider is such that all North dominos have
weight 1 and all East dominos have weight a. The weight of a West domino is either
aα if the vertical coordinate of the lower left corner is even, or a/α if that coordinate
is odd. Similarly, the weight of a South domino is either α if the vertical coordinate
of the lower left corner is even, and 1/α if that coordinate is odd. For small a > 0
we expect to see more South and North dominos, as the West and East domino have
small weight.
123
Biased 2 × 2 periodic Aztec diamond...
Fig. 6 The left figure shows the underlying graph G p. The right figure shows the graph G p and a collection
of non-intersecting paths starting in (0, − j) and ending in (2N , − j) for j = 0, . . . M, with N = 4 and
M = 5
2.2 Non-intersecting paths
A useful alternative representation, that is easily obtained from the dominos, is the
representation by DR-paths [18, 19, 29]. By drawing an upright path across each West
domino, a down-right path across each East domino, a horizontal across each South
domino and nothing on a North domino, we obtain the picture given in Fig. 5. There
are four paths leaving from the lower left side of the Aztec diamond and ending at
the lower right side. The paths also cannot intersect. Clearly, the paths determine the
location of the East, West and South dominos, and therewith the entire tiling. One
can therefore represent each dimer configuration with a collection of non-intersecting
paths.
Instead of looking directly at the DR paths, however, we will consider closely related
interpretation in terms of non-intersecting paths on a different graph. The reason for
this is two-fold. First, the DR paths are rather uneven in length. The bottom path is
much shorter than the top path. The second reason is that it turns out to be useful to
add paths so that we have an infinite number of them. The auxiliary paths will have
no effect on the model, but will give a very convenient integrable structure.
We start with a directed graph G p = ({0, 1, . . . , 2N }×Z, E p) where we draw edges
between the following vertices (we use the index p in G p and E p to distinguish this
graph from the bipartite graph in the dimer representation):
(2 j, k) → (2 j + 1, k),
(2 j + 1, k) → (2 j + 2, k),
(2 j, k) → (2 j + 1, k + 1),
(2 j + 2, k + 1) → (2 j + 2, k).
A part of the graph is shown in Fig. 6. We then fix starting points (0, − j) for j =
0, . . . , M, and endpoints (2N , − j) for j = 0, . . . , M and consider collections of
paths in the directed graph that connect the starting points with the endpoints, such
that no paths have a vertex in common (i.e., they never intersect).
Note that if M ≥ 2N − 2 only the N top paths and the N − 1 bottom paths are non-
trivial, but any path in between is, due to the non-intersecting condition, necessarily
a straight line. In fact, even the top N and bottom N − 1 paths have parts where they
123
A. Borodin, M. Duits
Fig. 7 From the non-intersecting paths on the graph G p to the DR paths. The middle picture is obtained
by removing the horizontal steps from the paths and the graph G p. In the second transformation (m, u) (cid:8)→
(m, u − m) we obtain the rotated DR paths
are necessarily straight lines. Indeed, in the region between the lines (m, −N + m/2)
and (m, −M + N + m/2) for m = 0, . . . , 2N , all the paths are necessarily horizontal.
The connection with the dimer models is the following: If we remove all paths
below the line (m, −N + m/2) then the configuration that remains is equivalent to the
DR paths for the domino tilings of Aztec diamond. Indeed, by further removing all
horizontal parts (m, u) → (m +1, u) for odd m and concatenating the result, we obtain
the picture in the middle of Fig. 7. The coordinate transform (m, u) (cid:8)→ (m, u − m)
maps the middle picture to the DR-paths shown on the right of Fig. 7.
The next step is to put a probability measure on the collection of non-intersecting
paths that is consistent with the dimer model from Sect. 2.1. To make the correspon-
dence, we note that each up-right diagonal edge in the graph G p corresponds to a West
domino, each vertical edge to an East domino, and each horizontal edge (after remov-
ing the auxiliary horizontal edges at the odd steps) corresponds to a South domino. A
careful comparison with the weights for the dimer models leads us to assigning weights
to the underlying directed graph as follows: the horizontal edges (m, u) → (m + 1, u)
for odd m are auxiliary and have weight 1, the vertical edges correspond to East
dominos and have weight a, the horizontal edges (m, u) → (m + 1, u) for even m
correspond to South dominos and have weight α if u is even and weight 1/α if u is
odd, and, finally, the up-right edges (m, u) → (m + 1, u + 1) for m even have weight
aα if u is even and weight a/α if u is odd. This is also represented in the following
finite weighted graph that is the building block for the rest of G p:
(2 j, 2k)
aα
α
a
α
(2 j, 2k − 1)
1
α
a
a
1
1
1
(2 j + 2, 2k)
(2 j + 2, 2k − 1)
Then the probability of having a particular configuration of non-intersecting paths
is proportional to the product of the weights of all the edges in the corresponding
dimer/domino configuration.
123
Biased 2 × 2 periodic Aztec diamond...
2.3 A determinantal point process
Let us now assign a point process to the above collections of paths. We place points
on these paths by taking the lowest possible vertex on each vertical section (including
those of length 0), as indicated in the right panel of Fig. 6,
(m, u j
m)
for j = 1, . . . , M, m = 0, . . . , 2N ,
= u j
2N
where u j
0
dimer configuration, so do the points (m, u j
turns the set of points with coordinates (m, u j
Z.
= − j +1 and M ≥ N . Since the top N paths uniquely determine the
m). Further, our probability measure also
m ) into a point process on {0, 1, . . . , 2N }×
We stress that we are only interested in the points (m, u j
m) with j ≤ N − m/2 + 1,
as it is those that determine the tiling. The other points are auxiliary and only added
for convenience. Indeed, by a theorem of Lindström-Gessel-Viennot (see, e.g., [14,
25]) the probability of a given point configuration is proportional to
2N(cid:6)
m=1
det Tm(u j
m−1
, uk
m
)M
j,k=1
,
where Tm are the transition matrices defined by
[Tm(2k1 − (cid:2)1, 2k2 − (cid:2)2)]1
(cid:2)1,(cid:2)2=0
(cid:7)
= 1
2πi
Am(z)
dz
zk2−k1+1
,
for k1, k2 ∈ Z, and Am(z) given by
(cid:8)
Am(z) =
Ae(z), if m is even,
Ao(z), if m is odd,
with
Ao(z) =
(cid:9)
(cid:10)
,
α aαz
a
1
α
α
Ae(z) =
1
1 − a2/z
(cid:9)
(cid:10)
.
1 a
a
z 1
We will also use the notation
(1)
123
A. Borodin, M. Duits
A(z) =
2N(cid:6)
m=1
Am(z).
By the Eynard-Mehta theorem (see, e.g., [13]), the point process is determinantal,
meaning that there exists a kernel
K N ,M : ({0, 1, . . . , 2N } × Z) × ({0, 1, . . . , 2N } × Z) → C,
(2)
such that, for any (mk, uk) ∈ {0, . . . , 2N } × Z and k = 1, . . . , n,
P(there are points at (mk, uk),
= det
K N ,M ((m j , u j ), (mk, uk))
(cid:11)
k = 1, . . . , n)
(cid:12)
n
.
j,k=1
Now we recall that we are only interested in the top N paths, and thus we will restrict
u j to be in {−N + 1, . . . , 0}. Then the marginal densities are independent of M as
long as M is sufficiently large and
K N ,M ((m1, u1), (m2, u2))
= lim
M→∞
K N ,M ((m1, u1), (m2, u2)) = K N ((m1, u1), (m2, u2)).
In [5] a double integral formula for the correlation kernel K N was given. That formula
involves a solution to a Wiener–Hopf factorization.
Proposition 2.1 [5, Theorem 3.1] Suppose that we can find a factorization
A(z) = A−(z)A+(z)
with 2 × 2 matrices A±(z) such that
1. A
2. A
±1
+ (z) are analytic in |z| < 1 and continuous in |z| ≤ 1,
±1
− (z) are analytic in |z| > 1 and continuous in |z| ≥ 1,
(cid:9)
(cid:10)
3. A−(z) ∼
1 0
0 1
as z → ∞.
Then the kernel K N ,M has the pointwise limit K N as M → ∞ given by
(cid:11)
K N ((m, 2x − j), (m(cid:10), 2x (cid:10) − j (cid:10)))
+ 1
(2πi)2
(cid:7)
(cid:7)
⎛
⎝
2N(cid:6)
|w|=ρ1
⎞
⎠ wx (cid:10)
zx
|z|=ρ2
j=m(cid:10)+1
dzdw
z(z − w)
,
A j (z)
(cid:13)
1m(cid:10)<m
2πi
(cid:12)
1
j, j=0
= −
⎞
m(cid:6)
|z|=1
j=m(cid:10)+1
A j (z)
dz
zx−x (cid:10)+1
A j (w)
⎠ A+(w)−1 A−(z)−1
(3)
⎛
⎝
m(cid:6)
j=1
123
Biased 2 × 2 periodic Aztec diamond...
where |a|2 < ρ1 < ρ2 < 1/|a|2, 1m(cid:10)<m = 1 if m(cid:10) < m and 0 otherwise, and the
integration contours are positively oriented.
Remark 2.2 Proposition 2.1 is only part of Theorem 3.1 in [5]. Indeed, the kernel in
(3) is called Ktop in [5]. We note that here we already shifted coordinates compared
to [5]. Also, in the formulation of Theorem 3.1 in [5] one needs a second factorization
A(z) = ˜A−(z) ˜A+(z). However, all that is needed for Proposition 2.1 is the existence
of such a factorization, and that is guaranteed by Theorem 4.8 in [5].
2.4 The Wiener–Hopf factorization
The question remains how to find a Wiener–Hopf factorization that is explicit enough
to be able to use (3) as a starting point for asymptotic analysis. The idea for finding a
Wiener–Hopf factorization is simple (see also [5, Sect. 4.4]). Write
where
A(z) =
1
(1 − a2/z)N
(P(z))N ,
P(z) =
(cid:10) (cid:9)
(cid:9)
α aαz
a
1
α
α
(cid:10)
.
1 a
a
z 1
Then in the first step we look for a Wiener–Hopf factorization of the form
P(z) = P0,−(z)P0,+(z),
and then write
where
(P(z))N = P0,−(z)(P1(z))N −1 P0,+(z),
P1(z) = P0,+(z)P0,−(z).
Next, we compute a factorization for P1(z) = P1,−(z)P1,+(z) and set P2(z) =
P1,+(z)P1,−(z). At each step in the procedure we thus construct a new matrix val-
ued function Pk+1(z) = Pk,+(z)Pk,−(z) constructed by switching the order of the
Wiener–Hopf factorization
Pk(z) = Pk,−(z)Pk,+(z).
(4)
The result is that we find a Wiener–Hopf factorization for A(z) of the form
A(z) =
(cid:3)
1
(1 − a2/z)N
P0,−(z) · · · PN −1,−(z)
(cid:4) (cid:3)
PN −1,+(z) · · · P0,+(z)
(cid:4)
.
123
A. Borodin, M. Duits
An important point is that this procedure defines a flow
P0(z) (cid:8)→ P1(z) (cid:8)→ P2(z) (cid:8)→ . . .
and to obtain explicit representations for the correlation kernel in (3) we need to have
a sufficiently detailed description of this flow. As was pointed out in [5, Sect. 4], there
is a general procedure to capture this flow. Generically, the description in [5] of the
flow is rather difficult to control, but for specific values it can be written explicitly.
Indeed, for a = 1, the double integral formula of [11] could be reproduced. See also
[4] for other cases where it was tractable. It is important to note that in the cases of both
[11] and [4], the flow was periodic, which is of great help, in particular for asymptotic
analysis. For the model that we consider in this paper, however, it appears difficult to
control this flow for a < 1, and the point of the present paper is to give an alternative
more tangible description. We will show that the flow is equivalent to translations on
an explicit elliptic curve. This will also help us to track other choices of parameters
for which the flow is periodic.
3 Main results
We now present our main results. All proofs will be postponed to Sect. 5.
3.1 An elliptic curve
Consider an elliptic curve E (over R) defined by the equation
y2 = x 2 + 4x(x − a2)(x − 1/a2)
(a + 1/a)2(α + 1/α)2
,
(5)
where α and a are the parameters from the dimer model in Sect. 2.1. One easily verifies
that the curve crosses the x-axis precisely three times, once at the origin and at two
further intersection points in (−∞, 0). The elliptic curve has therefore two connected
components, and one of those, denoted by E−, lies entirely in the left half plane. Note
also that (0, 0), (a2, a2) and (a−2, a−2) are the intersection points of the curve with
the line y = x. The point (a−2, a−2) will be of particular interest to us.
It is well known that an elliptic curve carries an Abelian group structure, and we
can add points on the curve. The point at infinity serves as the identity. We will be
interested in a linear flow on the curve that is constructed by repeatedly adding the
point (a−2, a−2) starting from the initial parameters (x0, y0) = (−1, − 1−α2
). That
1+α2
is, we consider the flow
(cid:18)
(x j+1, y j+1) = σ (x j , y j ),
(cid:20)
(cid:19)
(x0, y0) =
,
−1, − 1−α2
1+α2
123
Biased 2 × 2 periodic Aztec diamond...
Fig. 8 The flow on the elliptic curve. At each step we add the point (a−2, a−2). This can be geometrically
represented by drawing a straight line through (a−2, a−2) and (x j , y j ). This line intersects the curve at a
unique third point in E−. The point (x j+1, y j+1) is then obtained from the intersection point by flipping
the sign of the second coordinate
where
σ (x, y) = (x, y) + (a−2, a−2),
and + represents addition on the elliptic curve. The flow can be nicely illustrated by
the geometric description of the group addition on the curve. Starting from (x j , y j )
we compute (x j+1, y j+1) as follows: the straight line passing though (x j , y j ) and
(a−2, a−2) passes through a third point and (x j+1, y j+1) is the reflection of that point
with respect to the x axis (in other words, we flip the sign of the y-coordinate). See
also Fig. 8. It can happen that the line through (x j , y j ) and (a−2, a−2) is tangent to E−
at point (x j , y j ). In that case, (x j+1, y j+1) is just the reflection of the (x j , y j ) with
respect to the x axis. Note that the initial point (x0, y0) lies on the oval E−, and from
the geometric interpretation it is easy to see that every point (x j , y j ) is on the oval E−.
Our first main result is that this flow uniquely determines the correlations for the
biased Aztec diamond as described in Sects. 2.1–2.4 above. But before we explain that,
we first discuss properties of the flow that will be of interest to us. For generic choices
of the parameters one can expect the flow to be ergodic on E−, but for certain special
parameters (a−2, a−2) will be a torsion point. In those cases the flow is periodic. This
distinction has important implications for our asymptotic analysis of the tiling model.
We will therefore discuss a few examples in which (a−2, a−2) is a torsion point.
First, if we assume that α = 1, then our dimer model is an example of a Schur
process [27], and we know that simpler double integral formulas for its correlation
123
A. Borodin, M. Duits
kernel can be given. This should mean that our flow has a particularly simple structure.
Indeed, for α = 1, the oval E− reduces to a singleton E− = {(−1, 0)}, and the flow
is constant. This can also be seen directly, from the fact that the two factors in the
definition of P(z) commute.
The second case of interest is the unbiased case where a = 1. In that case,
(a−2, a−2) = (a2, a2), and the elliptic curve is tangent to the line y = x at that
point. For general a > 0 we have the relation (a2, a2) + (a−2, a−2) = (0, 0) and thus,
for a = 1, we have 2(a−2, a−2) = (0, 0). It is also clear that (0, 0) is a point of order
2, and thus (a−2, a−2) is of order 4. This implies that our flow is periodic and returns
to its initial point after 4 steps. For completeness, we compute the flow explicitly:
(cid:9)
−1, − 1 − α2
1 + α2
(cid:9)
(cid:10)
− 1
α2
, 0
(cid:8)→
(cid:10)
(cid:20)
(cid:19)
−α2, 0
(cid:8)→
(cid:8)→
(cid:9)
−1, − 1 − α2
1 + α2
(cid:8)→
(cid:10)
(cid:10)
(cid:9)
−1, 1 − α2
1 + α2
.
(6)
See the left panel of Fig. 9 for an illustration.
The next example we would like to discuss is that of an order six torsion point. This
happens when
a2 =
α
α2 + α + 1
.
(7)
The flow on the elliptic curve is given by:
(cid:10)
(cid:9)
(cid:8)→
(cid:9)
−1, − 1 − α2
1 + α2
(cid:9)
− 1
α2
(cid:8)→
, 1 − α
α2 + α3
(cid:10)
(cid:9)
(cid:8)→
−α2,
−α2,
(cid:10)
(cid:8)→
−α2 + α3
1 + α
(cid:9)
− 1
α2
, − 1 − α
α2 + α3
(cid:10)
(cid:8)→
(cid:10)
(cid:8)→
α2 − α3
1 + α
(cid:9)
−1, − 1 − α2
1 + α2
(cid:10)
(cid:9)
−1, 1 − α2
1 + α2
(cid:10)
.
(8)
Indeed, after six steps we have returned to our initial point. This case is illustrated on
the right panel of Fig. 9.
We found the relation (7) by computing the division polynomial of order 6 and
requiring that (a−2, a−2) is a zero of this polynomial. In fact, this provides a recipe
for deriving relations between a and α such that (a−2, a−2) is a torsion point of order
m. We recall the notion of division polynomials in Appendix B and provide such
relations for m = 4, 5, 6, 7, 8.
123
Biased 2 × 2 periodic Aztec diamond...
Fig. 9 The picture on the left illustrates the flow in case (a−2, a−2) is a torsion point of order four. The
picture on the right shows the flow in case that point has order six
3.2 Correlation kernel
To explain the connection between the flow on the elliptic curve and the Wiener–Hopf
factorization in Proposition 2.1 we define functions a, b, d : E− → (0, ∞) by
⎧
⎪⎨
⎪⎩
a(x, y) = a(a2+1)(α2+1)
b(x, y) = − 1
αax
2aαx(x−1/a2)
d(x, y) =
(a2+1)(α+1/α)(y−x)
y−x
1−a2x
2
,
.
,
Since x < 0 for (x, y) ∈ E−, these functions are well-defined with no poles and take
strictly positive values. Consider the maps
P− : (x, y) (cid:8)→ b(x, y)
(cid:10) (cid:25)
(cid:9)
a(x, y) 0
1
0
(cid:26) (cid:9)
(cid:10)
,
1
0
0 a(x, y)
1 1
a2
z 1
and
P+ : (x, y) (cid:8)→
(cid:25)
1
0 a2
0
α2 d(x, y)
(cid:26) (cid:9)
1 a2z
1 1
(cid:10) (cid:9)
1
0
0 d(x, y)
(cid:10)
.
The first main result of this paper is that the factorization (4) is given by
Pk,±(z) = P±(σ k(x, y)).
We will discuss this claim at length in Sect. 4 in a slightly more general setup and we
refer to that section for more details. The claim is then a special case of Theorem 4.6.
Of important to us now is that it, together with Proposition 2.1, implies the following.
123
Theorem 3.1 The correlation kernel K N from Proposition 2.1 can be written as
A. Borodin, M. Duits
(cid:11)
K N ((2m + ε, 2x − j), (2m(cid:10) + ε(cid:10), 2x (cid:10) − j (cid:10)))
(cid:13)
= −
12m(cid:10)+ε(cid:10)<2m+ε
2πi
(cid:7)
Ae(z)−ε(cid:10)(P(z))m−m(cid:10)
|z|=1
(cid:7)
(cid:12)
1
j, j (cid:10)=0
Ao(z)ε zm−x−m(cid:10)+x (cid:10)
dz
(z − a2)m−m(cid:10)
z
Ae(w)−ε(cid:10)
P(w)N −m(cid:10)
P+(w)−1 P−(z)−1 P(z)m Ao(z)ε
+ 1
(2πi)2
|w|=ρ1
|z|=ρ2
wx (cid:10)+N −m(cid:10) (z − a2)N −m
zx+N −m(w − a2)N −m(cid:10)
dzdw
z(z − w)
,
×
where
P−(z) =
N −1(cid:6)
j=0
b(σ j (x, y))
(cid:10) (cid:25)
(cid:9)
a(σ j (x, y)) 0
1
0
1 1
a2
z 1
(cid:26) (cid:9)
1
0
0 a(σ j (x, y))
(9)
(cid:10)
(10)
and
P+(z) =
N −1(cid:6)
j=0
(cid:25)
1
0 a2
0
α2 d(σ N −1− j (x, y))
(cid:26) (cid:9)
1 a2z
1 1
(cid:10) (cid:9)
1
0
0 d(σ N −1− j (x, y))
(cid:10)
,
(11)
and the contours of integration are counterclockwise oriented circles with radii ρ1
and ρ2 such that |a|2 < ρ1 < ρ2 < 1/|a|2.
A proof of this theorem is given in Sect. 5.1.
If (a−2, a−2) is a torsion point of order d, the flow (x, y) (cid:8)→ σ (x, y) is periodic,
and the double integral formula can be rewritten in a useful way.
Corollary 3.2 Assume that (a−2, a−2) is a torsion point of order d. Define
and
(d)
− (z) = P0,−(z) · · · Pd−1,−(z),
P
(d)
+ (z) = Pd−1,+(z) · · · P0,+(z).
P
Then we can rewrite (9) as
(cid:11)
Kd N ((2m + ε, 2x − j), (2m(cid:10) + ε(cid:10), 2x (cid:10) − j (cid:10)))
(cid:12)
1
j, j (cid:10)=0
= −
12m(cid:10)+ε(cid:10)<2m+ε
2πi
(cid:7)
(cid:7)
|z|=1
(cid:7)
+ 1
(2πi)2
Ae(w)−ε(cid:10)
|w|=ρ1
|z|=ρ2
Ae(z)−ε(cid:10)(P(z))m−m(cid:10)
Ao(z)ε zm−x−m(cid:10)+x (cid:10)
dz
(z − a2)m−m(cid:10)
z
123
Biased 2 × 2 periodic Aztec diamond...
P(w)d N −m(cid:10) (P
(d)
+ (w))−N (P
wx (cid:10)+d N −m(cid:10)(z − a2)d N −m
zx+d N −m(w − a2)d N −m(cid:10)
×
(d)
− (z))−N P(z)m Ao(z)ε
dzdw
z(z − w)
,
(12)
where |a|2 < ρ1 < ρ2 < 1/|a|2.
Note that in (9) we have replaced the size of the Aztec diamond N by d N . This is
not necessary and the upcoming analysis can also be performed for the general case.
Since the difference will only involve non-essential cumbersome bookkeeping, we
feel that working with d N instead of N makes for a cleaner presentation.
3.3 Asymptotics
The representation of the correlation kernel in (12) is a good starting point for an
asymptotic study. We will compute the microscopic process in the limit N → ∞ near
the point
(2dT , 2X ) = (2d(cid:11)N τ (cid:12), 2(cid:11)d N ξ (cid:12)),
0 < τ < 1, − 1
2
< ξ − τ/2 < 0.
(13)
That is, we consider the limiting behavior of the correlation kernel
(cid:11)
Kd N
(cid:3)
(2dT + 2m + ε, 2X + 2x − j) ,
(cid:3)
2dT + 2m(cid:10) + ε(cid:10), 2X + 2x (cid:10) − j (cid:10)
(cid:4)(cid:4)(cid:12)
1
j, j (cid:10)=0
(14)
as N → ∞, with m, m(cid:10) ∈ Z fixed. Note that the first coordinate of the point (13) is a
multiple of 2d and the second coordinate is a multiple of 2. This restriction is made
for clarity purposes and is not necessary. Note also that any finite shift from (13) can
be absorbed into the variables 2m + ε, 2m(cid:10) + ε(cid:10), 2x + j +2x (cid:10) + j (cid:10) in (14).
3.3.1 The spectral curve
To perform the asymptotic analysis it is convenient to diagonalize the matrices P(w),
P(z), P
(d)
+ (w) and P
(d)
− (z).
The spectral curve det(P(z) − λ) = 0 can be easily computed:
λ2 −
(cid:10)
(cid:9)
α + 1
α
(1 + a2)λ + (1 − a2z)
(cid:10)
(cid:9)
1 − a2
z
= 0.
(15)
The curve has branch points at z = 0, z = ∞, and at the zeros of the discriminant:
R(z) :=
(cid:10)
2
(cid:9)
α + 1
α
(1 + a2)2 − 4(1 − a2z)
(cid:10)
(cid:9)
1 − a2
z
= 0.
(16)
These zeros are negative and will be denoted by x1 and x2, ordered as x1 < x2 < 0.
With these points, we define a Riemann surface R consisting of two sheets R j =
123
A. Borodin, M. Duits
C\ ((−∞, x1) ∪ (x2, 0)), that we connect in the usual crosswise manner along the
cuts (−∞, x1) and (x2, 0). The sheets have 0 and ∞ as common points. See also
Fig. 10. We will write z( j) to indicate the point z on the sheet R( j). Then we define
the square root (R(z))1/2 on R such that (R(z(1)))1/2 > 0 for z(1) > 0. The spectral
curve (15) then defines a meromorphic function on R given by
(cid:9)
α + 1
α
(R(z))1/2,
(1 + a2) + 1
2
λ(z) = 1
2
(17)
(cid:10)
with poles at 0 and ∞, and zeros at (a±2)(2). The restrictions of λ to R( j) will be
denoted by λ j , i.e., λ j (z) = λ(z( j)).
Next, consider the spectral curves for P
(d)
− and P
(d)
+ ,
det(P
det(P
(d)
− (z) − μ) = μ2 − μ Tr P
(d)
+ (z) − ν) = ν2 − ν Tr P
(d)
− (z) + det P
(d)
+ (z) + det P
(d)
− (z) = 0,
(d)
+ (z) = 0,
(18)
(19)
These spectral curves factorize (15) in the following way.
Lemma 3.3 The equations (18), (19) for μ and ν define meromorphic functions on R
such that
(λ(z))d = μ(z)ν(z),
(20)
for z ∈ R. Then μ has a zero at (a2)(2) and a pole at 0, both of the same order d, and
ν has a zero at (a−2)(2) and a pole at ∞, both of the same order d.
With E(z) defined by
(cid:9)
E(z) =
aα(1 + z)
aα(1 + z)
λ1(z) − α(a2 + 1) λ2(z) − α(a2 + 1)
(cid:10)
,
we have
and
(cid:9)
(cid:9)
P(z) = E(z)
(d)
− (z) = E(z)
P
(cid:10)
λ1(z)
0
μ1(z)
0
λ2(z)
0
0 μ2(z)
E(z)−1,
(cid:10)
E(z)−1,
(d)
+ (z) = E(z)
P
(cid:9)
ν1(z)
0
(cid:10)
0
ν2(z)
E(z)−1.
Here μ j (z) = μ(z( j)) and ν j (z) = ν(z( j)) for z ∈ C\ ((−∞, x1) ∪ (x2, 0)).
123
(21)
(22)
(23)
(24)
Biased 2 × 2 periodic Aztec diamond...
Fig. 10 The two sheeted Riemann surface R. The dashed lines represent the cycles C1 and C2
The proof of this lemma will be given in Sect. 5.2
One particular consequence of this lemma is that we can simultaneously diagonalize
(d)
± (z). In the following theorem we use this to rewrite the correlation kernel
P(z) and P
in (12).
Theorem 3.4 Assume (a−2, a−2) is a torsion point of order d. Set, with E(z) as in
(15),
F(z) =
⎧
⎪⎪⎨
⎪⎪⎩
E(z)
E(z)
(cid:10)
(cid:10)
(cid:9)
(cid:9)
1 0
0 0
0 0
0 1
E(z)−1, z ∈ R1,
E(z)−1, z ∈ R2.
(25)
Then,
(cid:11)
(cid:13)
= −
Ae(z)−ε(cid:10)
12m(cid:10)+ε(cid:10)<2m+ε
2πi
(cid:7)
Kd N ((2dT + 2m + ε, 2X + 2x − j), (2dT + 2m(cid:10) + ε(cid:10), 2X + 2x (cid:10) − j (cid:10)))
F(z)Ao(z)ελ(z)m−m(cid:10) zm−x−m(cid:10)+x (cid:10)
(z − a2)m−m(cid:10)
wx (cid:10)−m(cid:10)
zx−m
dwdz
z(z − w)
F(w)F(z)Ao(z)ε λ(z)m
λ(w)m(cid:10)
(2πi)2
γ (1)
1
(w − a2)m(cid:10)
(z − a2)m
(z − a2)d(N −T )
(w − a2)d(N −T )
∪γ (2)
γ (1)
1
2
μ(w)N −T
μ(z)N −T
wd(N −T )+X
zd(N −T )+X
ν(z)T
ν(w)T
Ae(w)−ε(cid:10)
+ 1
γ (1)
2
(cid:7)
∪γ (2)
2
∪γ (2)
2
×
(cid:12)
1
j, j (cid:10)=0
dz
z
,
(26)
2
is a counterclockwise oriented contour inside the contour γ (1)
2
where γ (1,2)
γ (1)
1
that goes around (a2)(1) and the cut [x2, 0], and γ (2)
contour on the sheet R2 inside the contour γ (2)
also Fig. 11.
are the unit circles with counterclockwise orientation on the sheets R1,2,
on the sheet R1
is a counterclockwise oriented
that goes around the cut [x2, 0]. See
1
2
The proof of this Theorem will be given in 5.3.
Note that Ae(w)−1 is analytic at w = a2 (even though Ae(w) is not). Moreover,
λ(w)−m(cid:10)μ(w)N −T has a zero at w = (a2)(2) of order d(N − T ) − m(cid:10), and this
123
A. Borodin, M. Duits
Fig. 11 The contours of integration in (26). The blue contour represents γ1 and the orange contours are the
unit circles on the two different sheets (colour figure online)
zero cancels the pole at w = (a2)(2) in the double integral in (26). The contour γ (2)
1
therefore does not have to go around (a2)(2).
By passing to the eigenvalues and spectral curves we in fact are essentially looking
at a scalar problem, instead of a matrix-valued one.
Remark 3.5 We note that the spectral curve det (P(z) − λ) = 0 and the elliptic curve
E in (5) are related. Indeed, (5) can be written as
(cid:19)
(cid:20)
det
P(x) − 1
2
(a2 + 1)(α + 1/α) (1 + y/x)
= 0.
In other words, the elliptic curve E equals the spectral curve after changing the spectral
variable.
3.3.2 Saddle point equation and classification of different regions
The representation (26) is a very good starting point for asymptotic analysis. To illus-
trate this we will perform a partial asymptotic study, based on a saddle point analysis.
We note that a similar analysis has been given in [4, 11]. An interesting feature is that
our analysis will depend on the torsion d, but in such a way that we can treat all values
of d simultaneously.
To perform a saddle point analysis of (26) we need to find the saddle points and the
contours of steepest descent/ascent for the action defined by
(cid:13)(z; τ, ξ ) = (1 − τ ) log μ(z) − τ log ν(z) + d(1 − τ + ξ ) log z − d(1 − τ ) log(z − a2). (27)
This is a multi-valued function, but the differential
(cid:13)(cid:10)(z)dz
is single valued on R. Its zeros are the saddle points for Re (cid:13), and we will be especially
interested in them. Let C1 be the cycle on R defined by connecting the segments (x1, x2)
on R1 and R2 at the end points x1 and x2. Similarly, let C2 be the cycle that combines
the copies of (0, ∞) on both sheets.
123
Biased 2 × 2 periodic Aztec diamond...
Fig. 12 Both pictures represent a partitioning of the region into the frozen region, the rough disordered region
and the smooth disordered region. The picture on the left uses the natural coordinates (τ, ξ ) corresponding
to the point process associated with the non-intersecting paths. The picture on the right corresponds to the
coordinates for the original dimer model. In both pictures we have a2 = α/(1 + α + α2) and α = 1
2
Proposition 3.6 The differential (cid:13)(cid:10)(z)dz has simple poles at 0, (a2)(1), (1/a2)(2) and
∞. There are four saddle points (i.e., the critical points where (cid:13)(cid:10)(z)dz = 0) counted
according to multiplicity. There are at least two distinct saddle points on the cycle C1.
There are always two saddle points on the cycle C1, but it is the location of the two
other saddle points that determines the phase at the point (τ, ξ ). We say that (τ, ξ ) is
• in the frozen region, if we have two distinct saddle points on the cycle C2;
• in the smooth disordered region, if we have four distinct saddle points on the cycle
• in the rough disordered region, if there is a saddle point in the upper half plane of
C1;
R1 or R2;
• on the boundary between the rough and smooth disorderd regions, when this saddle
point from the upper half plane coalesces with its complex conjugate on the cycle
C1;
• on the boundary between the rough and frozen regions, when the saddle point from
the upper half plane coalesces with its complex conjugate on the cycle C2.
We note that the terminology rough, smooth and frozen goes back to at least [22]. See
also Fig. 12 for a partition of the Aztex diamond in the different regions. In that work
also the alternatives gaseous for smooth disordered and liquid for rough disordered
were mentioned. In the subsequent literature both these terms have been used. We
chose to use terminology frozen, rough and smooth disordered. The difference between
these regions is in the decay of the local correlations for the local Gibbs measure. In
the frozen region, the randomness disappears. In the rough disordered region, the
correlations between two points decay polynomially in their distance, whereas in the
smooth disorder regions these correlations decay exponentially. Our list above suggests
that these different behavior can be characterized in terms of the location of the two
123
remaining saddle points. The following theorem justifies this characterization for the
smooth disordered (or gaseous) region.
Theorem 3.7 Let (τ, ξ ) be in the smooth disordered region. Then
A. Borodin, M. Duits
lim
N →∞
[Kd N ((2dT + 2m + ε, 2X + 2x − j),
(2dT + 2m(cid:10) + ε(cid:10), 2X + 2x (cid:10) − j (cid:10))
(cid:12)
1
j, j (cid:10)=0
= −
12m(cid:10)+ε(cid:10)<2m+ε
2πi
12m(cid:10)+ε(cid:10)≥2m+ε
2πi
+
(cid:13)
γ (2)
2
(cid:13)
γ (1)
2
Ae(z)−ε(cid:10)
Ae(z)−ε(cid:10)
F(z)Ao(z)ελ(z)d(m−m(cid:10)) zm−x−m(cid:10)+x (cid:10)
(z − a2)m−m(cid:10)
F(z)Ao(z)ελ(z)d(m−m(cid:10)) zm−x−m(cid:10)+x (cid:10)
(z − a2)m−m(cid:10)
dz
z
dz
z
.
(28)
Note that from (28) we see that the limiting mean density in the smooth disordered
region is given by
Kd N ((2dT +2m +ε, 2X +2x − j), (2dT +2m +ε, 2X +2x − j (cid:10))
(cid:12)
1
j, j (cid:10)=0
(cid:11)
lim
N →∞
= 1
2πi
(cid:13)
Ae(z)−ε F(z)Ao(z)ε dz
z
,
γ (1)
2
and the right-hand side is independent of (T , X ) (as long as it is in the smooth disor-
dered region).
It is also not difficult to see that the right-hand side of (28) decays exponentially with
the distance between (m, x) and (m(cid:10), x (cid:10)). Indeed, for m and m(cid:10) fixed, the right-hand
side is the (x − x (cid:10))-th Fourier coefficient of a function that is analytic in an annulus.
Such coefficients decay exponentially with a rate that is determined by the width of
the annulus. More generally, the exponential decay follows from a steepest descent
analysis for the right-hand side of (28).
The proof of Theorem 3.7 will be given in Sect. 5.5 and it is based on a saddle point
analysis of the integral representation (26). We are confident that such a saddle point
analysis can be carried out similarly for the rough disordered and frozen regions. Since
it requires non-trivial effort and since a full asymptotic study is not the main focus of
this paper, we do not perform such an analysis here.
3.4 The boundary of the rough disordered region
We will now show that the boundary of the rough disordered region is an algebraic
curve and discuss how this curve can be found explicitly in particular cases.
We start with the following proposition.
123
Biased 2 × 2 periodic Aztec diamond...
Proposition 3.8 With (cid:13) as in (27) and R(z) = a2(z − x1)(z − x2)/z as in (16) we
have
(cid:13)(cid:10)(z) = d(1 − τ )a2 zγ1 + γ2 + γ3 R(z)1/2
(z − a2)z R(z)1/2
− d(1 − τ )
z − a2
+ d(1 − τ + ξ )
z
,
− dτ
γ1 + γ2z + γ3z R(z)1/2
(z − a−2)z R(z)1/2
where γ1, γ2 and γ3 are real constants determined by
⎧
⎪⎨
⎪⎩
(cid:19)
(cid:27)
1
d
γ1 = − 1
2
γ2 + a2γ1 = − 1
2
γ3 = 1
,
2
d−1
j=0 a(σ j (x, y))
(a2 + 1)(α + 1/α),
(cid:20)
1/2
(cid:19)
(cid:27)
1
d
d−1
k=0
1
a(σ k (x,y))
(cid:20)
1/2
,
(29)
(30)
and the square root is taken such that R(z)1/2 is meromorphic on R and R(z(1))1/2 > 0
for z > 0.
The proof of this proposition will be given in Sect. 5.6.
By inserting the constants (29) into (cid:13)(cid:10)(z), multiplying by (z −a2)(z −a−2)R(z)1/2,
and re-organizing the equation so that all terms with R(z)1/2 are on the right, we see
that (cid:13)(cid:10)(z) = 0 can be written as
(1 − τ )a2(γ1z + γ2)(z − a−2) − τ (γ1 + γ2z)(z − a2)
(cid:19)
(1 − τ )a2γ3(z − a−2) − τ γ3(z − a2)z
= −R(z)1/2
+(1 − τ + ξ )(z − a2)(z − a−2) − (1 − τ )z(z − a−2)
(cid:20)
.
(31)
Before we proceed, note that z = a−2 and z = a2 are two solutions that we just
introduced by multiplying by (z − a2)(z − a−2) and are not saddle points.
By squaring both sides of (31) and multiplying by z we find a polynomial equation
of degree 6 in z with coefficients that are quadratic functions of τ and ξ . Since z = a±2
are solutions that we are not interested in, we are left with an equation of degree four.
There are four solutions to this equation, and each of them corresponds to exactly one
point on the surface. This confirms that we indeed have four saddle points, which was
part of the statement in Proposition 3.6.
This also allows to write an equation for the rough disordered boundary. Indeed,
the coefficients of this fourth degree equation will be quadratic expressions in τ and
ξ . We have a third order saddle in case the discriminant vanishes. The discriminant of
a polynomial of degree four is a polynomial in its coefficients of degree six. Thus, the
discriminant is a polynomial in τ and ξ of degree twelve. In the explicit cases that we
tried, we found, with the help of computer software, that this degree twelve curve can
be factorized into a curve of degree eight and remaining factors that are not relevant.
This also matches with the findings of [7] and [5] for the special case a = 1. We have,
however, only been able to verify that this holds numerically in special cases (one of
123
A. Borodin, M. Duits
them we will discuss in Appendix A) and do not have a proof that it holds generally.
We leave this as an interesting open problem and post the following conjecture:
Conjecture 3.9 The boundary of the rough disordered region is an algebraic curve in
τ and ξ of degree eight.
Remark 3.10 There is another way of parametrizing the boundary. Indeed, on the two
components of the boundary of the rough region we have a coalescence of saddle
points on the cycles C1 or C2. This means that we have a double zero of the differential
(cid:13)(cid:10)(z)dz. This gives a way of parametrizing these curves. Indeed, (cid:13)(cid:10)(z) = (cid:13)(cid:10)(cid:10)(z) = 0
for z ∈ C1 or C2 gives a linear system of equations for μ and ξ that can be easily
solved.
Another interesting consequence of (29) is that the saddle point equation (cid:13)(cid:10)(z) = 0
only depends on the order d of the torsion via the constant γ1 in (30). However, it is
even possible to replace this with another expression that does not involve d:
Lemma 3.11 The constants γ1 and γ2 from (30) are related via
(cid:13)
x2
x1
γ1
xd x
(x − a−2)
√
R(x)
= −γ2
(cid:13)
x2
x1
d x
(x − a−2)
√
,
R(x)
(32)
√
where
R(x) > 0 for x ∈ (x1, x2).
The proof of this lemma will ve given in Sect. 5.7.
By replacing the equation for γ1 in (30) by (32) we see that we have eliminated
the dependence on d from the saddle point equation, and the saddle point equation
makes sense for general parameters a and α. Although the arguments that we provide
in this paper use the torsion at several places, it is natural to conjecture that the saddle
point analysis and its consequences can be extended in this way. In particular, we
conjecture the characterization of the different phases in Sect. 3.3.2 and Theorem 3.7
to hold under this extension. We leave this as an open problem.
3.5 Overview of the rest of the paper and the proofs
In the remaining part of this paper we will prove the main results. In Sect. 4 we will
show that the linear flow on the elliptic curve can be used to find a Wiener–Hopf
factorization in Proposition 2.1. We will do this in a more general setup than only
for the biased Aztec diamond. In Sect. 5 we will return to the biased Aztec diamond
and prove Theorem 3.1 in Sect. 5.1, which is by then just an identification of the
parameters in the discussion of Sect. 4. Then Lemma 3.3 and Theorem 3.4 are proved
in Sects. 5.2 and 5.3, respectively. The saddle point analysis starts with proving Lemma
3.6 in Sect. 5.4. After that, we perform a saddle point analysis in Sect. 5.5 and prove
Theorem 3.7. Proposition 3.8 is proved in Sect. 5.6 and Lemma 3.11 in Sect. 5.7. In
Appendix A we work out the example where (a−2, a−2) is a torsion point of order six.
We compute the boundary of the rough disordered region, and we provide numerical
results supporting the saddle point analysis of Sect. 5.5. Finally, in Appendix B we will
show how the notion of division polynomials can be used to find algebraic relations
between a and α so that (a−2, a−2) is a torsion point of order d.
123
Biased 2 × 2 periodic Aztec diamond...
4 The flow
In this section we introduce a flow on a space of matrices that will give a Wiener–Hopf
factorization in the correlation kernel. We prove that this flow is equivalent to a linear
flow on an elliptic curve using translations by a fixed point on that curve.
4.1 The space
First we have to define the space of matrices that we work on. To this end, we first
introduce
S =
(cid:8)(cid:9)
a11
a21 + b21/z
(cid:10)
a12 + b12z
a22
| a11, a22, a12, a21, b12, b21 > 0
.
(33)
(cid:28)
Clearly, the determinant det P(z) of any P ∈ S is a rational function in z with poles
at z = 0 and z = ∞ and no other. Also, det P(z) will have two zeros z1 and z2, and
we will assume that
0 < z1 < 1 < z2.
Then the winding number of det P(z) with respect to the unit circle equals zero.
The flow that we will define on S will be such that det P(z) and Tr P(z) will be
invariant under it. We therefore introduce the sets
S(z1, z2, c1, c2) = {P(z) ∈ S | Tr P(z) = 2c1, det P(z) = −c2(z − z1)(z − z2)/z}
for c1, c2 > 0 and 0 < z1 < 1 < z2.
Naturally, c1, c2 and z1, z2 be expressed in terms of ai j and bi j . Indeed,
(cid:8)
c1 = a11+a22
,
c2 = a21b12,
2
(34)
and z1, z2 are the solutions to det P(z) = 0. Equivalently, z1 and z2 can be obtained
from the following equations:
(cid:8)
z1z2 = a12b21
a21b12
c2(z1 + z2) = a11a22 − (a21a12 + b12b21),
,
(35)
which, combined with the condition 0 < z1 < 1 < z2, determine z1 and z2 uniquely.
We also note that the condition 0 < z1 < 1 < z2 is equivalent to requiring det P(1) >
0, because det P(z) → −∞ for z ↓ 0 and z → +∞. In terms of ai j and bi j this
means that the condition is equivalent to
a11a22 > (a12 + b12)(a21 + b21).
123
Note that this also shows that right-hand side of the second equation in (35) is positive,
as it should be.
It should also be noted that c1, c2, z1 and z2 cannot take arbitrary values. For
A. Borodin, M. Duits
instance, we have the following result.
Lemma 4.1 We have
c2
1
≥ c2(
√
√
z1 +
z2)2.
Proof The proof follows after inserting (34) and (35) into
− c2(z1 + z2 + 2
√
z1z2)
c2
1
− c2(
√
√
z1 +
giving the result.
=
z2)2 = c2
1
(a11 + a22)2
4
+ b12b21 − 2
(a11 − a22)2
4
=
− a11a22 + a12a21
(cid:29)
a12a21b12b21
+ (
√
a12a21 −
(cid:29)
b12b21)2 ≥ 0,
(36)
(cid:15)(cid:16)
As we will see later, the inequality (36) is sufficient to ensure that S(z1, z2, c1, c2) (cid:17)=
∅. We will give an explicit parametrization of S(z1, z2, c1, c2) in terms of part of an
elliptic curve. But first, let us define a flow on S(z1, z2, c1, c2).
4.2 Definition of the flow
We will be interested in factorization of the matrices in S of a particular form. Start
by introducing the sets
S− =
(cid:10) (cid:9)
(cid:8)(cid:9)
a 0
0 1
and
(cid:10) (cid:9)
1 1
z1
z 1
b 0
0 1
(cid:10)
(cid:28)
| a > 0, b > 0, 0 < z1 < 1
,
(cid:8)(cid:9)
(cid:10) (cid:9)
S+ =
1 0
0 c
(cid:10) (cid:9)
(cid:10)
1 0
0 d
1 z
z2
1 1
(cid:28)
| c > 0, d > 0, z2 > 1
.
It is straightforward to verify that if Q+ ∈ S+ and Q− ∈ S− then Q+ Q− ∈ S and
Q− Q+ ∈ S.
Proposition 4.2 Let P ∈ S(z1, z2, c1, c2). Then there exist unique Q± ∈ S± such that
P = Q− Q+.
Proof Note that
(cid:10) (cid:9)
(cid:10) (cid:9)
1 0
0 c
(cid:10) (cid:9)
1 z
z2
1 1
(cid:10) (cid:9)
(cid:10)
1 0
0 d
=
(cid:25)
ab + ac acd + abdz
z2
cd + bdz1
c + bz1
z2
z
(cid:26)
.
b 0
0 1
1 1
z1
z 1
(cid:10) (cid:9)
(cid:9)
a 0
0 1
123
Biased 2 × 2 periodic Aztec diamond...
To find Q± we have to solve
(cid:25)
ab + ac acd + abdz
z2
cd + bdz1
c + bz1
z2
z
(cid:26)
=
(cid:9)
a11
a21 + b21/z
(cid:10)
.
a12 + b12z
a22
By comparing the coefficients on both sides we obtain six equations for the six
unkowns a, b, c, d, z1 and z2. The parameters z1, z2 can be found from the condi-
tion det P(z1) = det P(z2) = 0. Then finding the remaining equation gives
⎧
⎪⎪⎪⎨
⎪⎪⎪⎩
,
a = a11z1
a21z1+b21
b = b21
,
z1
c = a21,
d = a12(a21z1+b21)
a11a21z1
.
This determines the factorization P = Q− Q+ uniquely.
(37)
(cid:15)(cid:16)
Because of the special structure of S± we have uniqueness of the factorization. How-
ever, for our purposes we need an additional degree of freedom by adding a diagonal
factor. Indeed, if P = Q− Q+ then P− = Q− D and P+ = D−1 Q+ for any diago-
nal matrix D also provides a factorization of P such that P+ P− = D−1 Q+ Q− D ∈
S(z1, z2, c1, c2). We will use this additional degree of freedom by requiring that
P+ P− = P− P+ + O(1),
z → ∞.
In other words, we require that the leading term in the asymptotic behavior fo P− P+
and P+ P− match. In order to achieve this, we define
(cid:9)
1 0
0 ab
(cid:10)
,
D =
(38)
where a, b are the parameters in Q−.
Definition 4.3 Define the map s : S(z1, z2, c1, c2) → S(z1, z2, c1, c2) as follows: for
P ∈ S(z1, z2, c1, c2) let P = Q− Q+ be the unique factorization from Proposition 4.2
and take P+ = D−1 Q+ and P− = Q− D where D is defined by (38). Then set
s(P) = P+ P−.
The flow on S(z1, z2, c1, c2) that we wish to study is then defined by iterating the
map s, i.e., the flow is defined by the recurrence
(cid:8)
Pk+1 = s(Pk),
P0 = P ∈ S(z1, z2, c1, c2).
k ≥ 0,
It turns out it is rather complicated to keep track of this dynamics, and our goal is to
describe this dynamics in a way that it is easier to grasp.
123
A. Borodin, M. Duits
4.3 Translations on an elliptic curve
Consider the elliptic curve E over R defined by (with c1, c2 > 0 and 0 < z1 < 1 < z2
as before)
(cid:30)
(x, y) ∈ R2 | c2
1
E =
(y2 − x 2) = c2x(x − z1)(x − z2)
.
(cid:31)
We also assume, cf. Lemma 4.1, that
√
≥ (
z1 +
√
z2)2.
c2
1
c2
(39)
This inequality implies that we have three points on the curve whose y coordinate
is zero, (0, 0), (−t1, 0) and (−t2, 0), with t1, t2 > 0. Moreover, the curve E has two
connected components
E± = {(x, y) ∈ E | ±x ≥ 0} ,
one in the left half plane and the other in the right half plane. It will also be important
for us that the lines y = ±x lie above and below E−, meaning that y2 − x 2 < 0 and
thus |y|/|x| < 1. Indeed, the lines y = ±x intersect E at most at three points, and we
already established that these points are on E+. This implies that E− has to lie fully
below or above each of these lines and since (−t1, 0) and (−t2, 0) lie below the line
y = −x and above the line y = x, so does E−. See also Fig. 13.
There is a classical construction of addition on an elliptic curve which we will use.
We can add two points (x1, x1), (x2, y2) ∈ E as follows: generically, the line through
(x1, y1) and (x2, y2) intersects the elliptic curve at exactly one point (x3, −y3). Then
we define (x1, y1) + (x2, y2) = (x3, y3). One exception is when (x2, y2) = (x1, y1)
(in which case the addition becomes a doubling of the point), but this can be defined
by continuity. The other exception is (x1, y1) + (x1, −y1) which we define to be the
point at infinity. The addition turns E into group with the point at infinity as zero.
We will be mostly interested in translation by (z2, z2) on E. Observe that if (x, y) ∈
E− then (x, y) + (z2, z2) ∈ E−. We will define the translation operator
σ : E− → E− : (x, y) (cid:8)→ (x, y) + (z2, z2).
It is not hard to put this into a concrete formula. Since it will be useful to have this
formula at hand, and in order to simplify arguments later, we include it in the following
lemma.
Lemma 4.4 We have
(cid:9)
σ (x, y) =
z2(x − z1)(y − x)
(x − z2)(x + y)
, z2(y − x)(x 2 + y(z1 − z2) − z1z2)
(x − z2)2(x + y)
(cid:10)
,
for all (x, y) ∈ E−.
123
Biased 2 × 2 periodic Aztec diamond...
Fig. 13 An example with parameter z1 = 1
always have an oval in the left half plane. In case we have equality, the oval has shrunk to a point
/c2 = 7. Under strict inequality in (39) we
2 , z2 = 2 and c2
1
Proof The line through the point (x, y) and (z2, z2) is given by the formula Y =
λ(X − z2) + z2 where λ = y−z2
. By substituting this into the equation for E, moving
x−z2
all terms to the right-hand side and collecting the coefficient of X 2 we obtain
−λ2c2
1
+ c2
1
− c2(z1 + z2),
and this equals −c2 times the sum of the three zeros of the resulting cubic equation
for X . In other words, after setting (x ∗, −y∗) = (x, y) + (z2, z2) we have
−c2(x ∗ + z2 + x) = −c2
1
λ2 + c2
1
− c2(z1 + z2).
Thus,
x ∗ =
c2
1
c2
(λ2 − 1) + z1 − x =
(cid:9)
c2
1
c2
(y − x)(x + y − 2z2)
(x − z2)2
(cid:10)
+ z1 − x.
Now use the fact that (x, y) ∈ E to find
x ∗ =
=
(x − z1)x(x + y − 2z2)
(x − z2)(x + y)
(x − z1)
(x − z2)(x + y)
+ z1 − x
(x(x + y − 2z2) − (x − z2)(x + y)) = z2(x − z1)(y − x)
(x − z2)(x + y)
.
123
Inserting this back into y∗ = λ(x ∗ − z2) + z2 we find
(cid:9)(cid:9)
y∗ = z2
(cid:9)
= z2
(y − z2)(x − z1)(y − x)
(x − z2)2(x + y)
(y − z2)(x − z1)(y − x)
(x − z2)2(x + y)
− y − z2
x − z2
(cid:10)
+ x − y
x − z2
and further simplification shows
A. Borodin, M. Duits
(cid:10)
(cid:10)
+ 1
,
y∗ = z2(y − x) ((y − z2)(x − z1) − (x + y)(x − z2))
(cid:3)
z2(y − x)
(x − z2)2(x + y)
x 2 + y(z1 − z2) − z1z2
(cid:4)
.
= −
(x − z2)2(x + y)
By flipping the sign of y∗ we thus obtain the statement.
(cid:15)(cid:16)
4.4 Equivalence of the flows
Our main point is that the flows s and σ from Definition 4.3 and Lemma 4.4 are
equivalent. We start with the following.
Proposition 4.5 The map π : (0, ∞) × E− → S(z1, z2, c1, c2) defined by
(cid:25)
(cid:26)
(cid:4)
π(u, (x, y)) =
(cid:3)
1 − y
c1
(cid:19)
x
1 − z1z2
c2
x z
u
(cid:20)
u(z − x)
(cid:4)
(cid:3)
1 + y
c1
x
(40)
is well-defined and a bijection.
Proof First, since x < 0 and |y| < |x| for (x, y) ∈ E− we see that all entries
and coefficients of π(u, (x, u)) are positive and thus π(u, (x, y)) ∈ S. To see that
π(u, (x, y)) ∈ S(z1, z2, c1, c2) we have to check that the defining equations match.
To this end, we note that
Tr π(u, (x, y)) = 2c1,
and
det π(u, (x, y)) = c2
1
=
(cid:10)
(cid:9)
1 − y2
x 2
(cid:3)
x 2 − y2
c2
1
(cid:10)
(cid:9)
1 − z1z2
x z
− c2(z − x)
(cid:4)
+ c2x(x − z1)(x − z2)
x 2
− c2(z − z1)(z − z2)
z
.
(41)
Hence,
123
det π(u, (x, y)) = −c2(z − z1)(z − z2)/z
Biased 2 × 2 periodic Aztec diamond...
if and only if (x, y) ∈ E− (note that we already observed that x < 0). Therefore,
π(u, (x, y)) ∈ S(z1, z2, c1, c2).
To establish that π is a bijection we construct the inverse map as follows. It is
not difficult to see that any matrix from the general space S can be written as in the
right-hand side of (40) after choosing c1, c2, z1, z2 as in (34) and (35) and u, x, y as
⎧
⎪⎨
⎪⎩
u = b12,
x = − a12
(cid:19)
b12
y = a12
b12
,
a11−a22
a11+a22
(cid:20)
.
By the assumptions ai j > 0 and bi j > 0 we see that u, c1, c2, z1z2 > 0, hence x < 0
and |y| < |x|. We still need to verify that (x, y) lies on the elliptic curve. But this
follows from the computation of the determinant (41). Indeed, since the determinant
matches with det P(z) we must have that (x, y) ∈ E. Since we already know that
(cid:15)(cid:16)
x < 0 we find (x, y) ∈ E−, and we have thus proved the statement.
We now come to the key point of this section.
Theorem 4.6 For any (u, (x, y)) ∈ (0, ∞) × E− we have π(u, σ (x, y)) =
s(π(u, (x, y))).
Proof Since π is a bijection, there must exist (u(cid:10), (x (cid:10), y(cid:10))) ∈ (0, ∞) × E− such that
s(π(u, (x, y))) = π(u(cid:10), (x (cid:10), y(cid:10))). We first compute s(π(u, (x, y))). Note that from
Proposition 4.2 and (37) we have π(u, (x, y)) = Q− Q+ with
⎧
⎪⎪⎪⎨
⎪⎪⎪⎩
a = uc1(x−y)
c2(x−z2) ,
b = − c2z2
,
xu
c = c2
,
u
d = −ux(x−z2)
c1(x−y)
.
(42)
We note that since (x, y) is a point on the elliptic curve, we can rewrite d as
d = uc1(x + y)
c2(x − z1)
.
Now we can compute
s(π(u, (x, y)) = P+(z)P−(z) = D−1 Q+(z)Q−(z)D =
(cid:25)
ab + bdz1
z2
c + cdz1
az
(cid:26)
.
a2b + abdz
z2
ac + cd
To find (u(cid:10), (x (cid:10), y(cid:10))) such that s(π(u, (x, y)) = π(u(cid:10), (x (cid:10), y(cid:10))) we argue as follows.
From (40) we see that u(cid:10) is the coefficient of z in the 12-entry. This gives u(cid:10) = u, so
the parameter u is unchanged under the flow.
Then x (cid:10) is the zero of the 12-entry viewed as a linear function in z and thus
x (cid:10) =
−z2a
d
= z2(y − x)(x − z1)
(x + y)(x − z2)
.
123
A. Borodin, M. Duits
Next, by looking at the 22-entry of P+ P− we find
c(a + d) = c1
(x − z1)(x − y) + (x − z2)(x + y)
(x − z1)(x − z2)
.
By solving for y(cid:10) from the 22-entry of π(u(cid:10), (x (cid:10), y(cid:10))), cf. (40), we find
(cid:9)
(cid:9)
(cid:9)
y(cid:10) =
=
=
(cid:10)
− 1
x (cid:10)
c(a + d)
c1
(x − z1)(x − y) + (x − z2)(x + y)
(x − z1)(x − z2)
(x − z1)(x − y) + (x − z2)(y + z1)
(x − z1)(x − z2)
= z2(x 2 + y(z1 − z2) − z1z2)(y − x)
(x + y)(x − z2)2
.
(cid:10)
− 1
(cid:10)
z2(y − x)(x − z1)
(x + y)(x − z2)
z2(y − x)(x − z1)
(x + y)(x − z2)
Thus, (x (cid:10), y(cid:10)) matches with (z2, z2) + (x, y) from Lemma 4.4 as desired.
(cid:15)(cid:16)
4.5 Wiener–Hopf factorizations
Let P(z) ∈ S with S as defined in (33) and n ∈ N. In this paragraph we will show
how the flows above can be used to find an explicit Wiener–Hopf factorization
(P(z))n+1 = P−(z)P+(z).
First of all, as also discussed in Sect. 2.4, with Pk(z) = sk(P(z)) and Pk(z) =
Pk,−(z)Pk,+(z) as in Definition 4.3 we can take
and
P−(z) = P0,−(z)P1,−(z) · · · Pn,−(z),
P+(z) = Pn,+(z)Pn−1,+(z) · · · P0,+(z).
Then, by Theorem 4.6 we can obtain an explicit representation in terms of the flow on
the elliptic curve. To this end, we first define the functions (cf. (42))
⎧
⎨
⎩
a(x, y) = uc1(x−y)
c2(x−z2) ,
b(x, y) = − c2z2
.
xu
Using the parametrizaton for P(z) as in (40) we then have, by Theorem 4.6,
(cid:10)
(cid:10) (cid:9)
(cid:10) (cid:9)
(cid:9)
a(σ j (x, y)) 0
1
0
1 1
z1
z 1
1
0
0 a(σ j (x, y))
.
(43)
Pj,−(z) = b(σ j (x, y)))
123
Biased 2 × 2 periodic Aztec diamond...
Hence,
P0,−(z)P1,−(z) · · · Pn,−(z) =
n(cid:6)
b(σ j (x, y)))
(cid:9)
a(σ j (x, y)) 0
1
0
×
n(cid:6)
j=0
(cid:10) (cid:9)
j=0
(cid:10) (cid:9)
1 1
z1
z 1
(cid:10)
1
0
0 a(σ j (x, y))
(44)
For future reference, we note that the constant pre-factor is of no interest to us and
will cancel out in the integrand for the double integral formula of Proposition 2.1 for
the correlation kernel. It is thus the evolution of a(σ j (x, y)) that is of importance.
Next, define the function
d(x, y) =
−ux(x − z2)
c1(x − y)
.
Then we have
Pj,+(z) =
Hence,
(cid:9)
1
0 c2
0
z2u2 d(σ j (x, y))
(cid:10) (cid:9)
1 z
z2
1 1
(cid:10) (cid:9)
1
0
0 d(σ j (x, y))
(cid:10)
.
(45)
Pn,+(z)Pn−1,+(z) · · · P0,+(z) =
(cid:9)
n(cid:6)
j=0
1
0 c2
z2u2 d(σ j (x, y))
(cid:10) (cid:9)
(cid:10) (cid:9)
1 z
z2
1 1
1
0
0 d(σ j (x, y))
(cid:10)
.
(46)
5 Proofs of the main results
We now return to the model of the biased doubly periodic Aztec diamond from Sect. 2.1
and prove our main results.
5.1 Proof of Theorem 3.1
Proof of Theorem 3.1 We recall from Proposition 2.1 that we are interested in finding
a factorization for
where
A(z) =
1
(1 − a2/z)N
(P(z))N ,
P(z) =
(cid:10) (cid:9)
(cid:9)
α aαz
α
1
α
a
(cid:10)
.
1 a
a
z 1
123
Comparing this with the setting of Sect. 4.4 we see that we have the special case
A. Borodin, M. Duits
⎧
⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎩
z1 = a2,
z2 = 1/a2,
c1 = 1
2
c2 = a2,
u = aα,
(a2 + 1)(α + 1/α),
(47)
and thus the elliptic curve can be written as
y2 − x 2 = 4x(x − a2)(x − 1/a2)
(a + 1/a)2(α + 1/α)2
.
The flow starts with the initial parameters (x0, y0) = (−1, − 1−α2
1+α2
straightforward consequence of the factorization of Sect. 4.5.
). The theorem is a
(cid:15)(cid:16)
5.2 Proof of Lemma 3.3
Proof of Lemma 3.3 It is readily verified that (22) holds. An important observation is
that
(cid:19)
(d)
− (z)P
P
(d)
+ (z)
(cid:20)
2
= (P(z))2d =
(cid:19)
(d)
− (z)
P
(cid:19)
(cid:20)
2
(cid:20)
2
.
(d)
+ (z)
P
(d)
− (z) and P
This implies that P(z)d , P
ously diagonalizable. Hence, we can write P
note that we can rewrite (23) and (24) as
(cid:10)
(cid:9)
(d)
+ (z) commute,1 and therefore are simultane-
(d)
± (z) as in (23) and (24). Furthermore,
E(z)−1 P
(d)
− (z)E(z) =
, E(z)−1 P
(d)
+ (z)E(z) =
μ1(z)
0
0 μ2(z)
(cid:9)
ν1(z)
0
(cid:10)
,
0
ν2(z)
(48)
with E(z) as in (21). Now the entries of E(z) and E(z)−1 are meromorphic functions
for z ∈ C\ ((−∞, x1] ∪ [x2, 0]). From (48) we then see that μ1,2 and ν1,2 are also
meromorphic for z ∈ C\ ((−∞, x1] ∪ [x2, 0]). Now, on the cuts (−∞, x1] ∪ [x2, 0]
we have
E+(z) = E−(z)
(cid:9)
(cid:10)
0 1
1 0
,
where E±(z) = limε↓0 E(z ± εi). This implies that, for z ∈ (−∞, x1) ∪ (x2, 0), we
have
μ1,±(z) = μ2,∓(z),
ν1,±(z) = ν2,∓(z),
1 We are grateful to Tomas Berggren for reminding us of this fact.
123
Biased 2 × 2 periodic Aztec diamond...
where μ j,± = limε↓0 μ j (z + εi) and ν j,± = limε↓0 ν j (z + εi). Therefore, we see that
the functions μ defined by μ(z( j)) = μ j (z) and, similarly, ν defined ν(z( j)) = ν j (z)
extend to meromorphic functions on R.
Clearly, μ and ν must satisfy (20).
What remains is the statement on the zeros and poles of ν and μ. By (19), any pole
(d)
(d)
+ (z) and/or det P
+ (z) can only possibly have
of ν is a pole of Tr P
(d)
+ (z) has exactly one pole which is at z = ∞ of degree
a pole at z = ∞, and det P
d, we see that ν has a pole at the branch point z = ∞ of degree d and no other. The
(d)
zeros of ν can then be determined from the zeros of det P
+ (z), and this shows that
the only possible locations of the zeros are z = (a−2)(1) and z = (a−2)(2), where the
sum of the orders equals d. By (20) and the fact that λ has no zero at z = (a−2)(1), it
follows that ν has a zero at z = (a−2)(2) of order d. The poles and zeros of μ can be
(cid:15)(cid:16)
determined analogously.
(d)
+ (z). Since Tr P
5.3 Proof of Theorem 3.4
Proof of Theorem 3.4 Note that by (25) we can rewrite the spectral decomposition (22)
as
P(w) = F(w(1))λ(w(1)) + F(w(2))λ(w(2)),
and, similarly for P+(w),
(d)
+ (w) = F(w(1))ν(w(1)) + F(w(2))ν(w(2)).
P
Combining this with F(w(1))F(w(2)) = O (the zero matrix), we see that
P(w)d(N −T )(P
P(w)−m(cid:10)
= F(w(1))λ(w(1))d(N −T )−m(cid:10)ν(w(1))−N + F(w(2))λ(w(2))d(N −T )−m(cid:10) ν(w(2))−N
= F(w(1))λ(w(1))−m(cid:10)μ(w(1))N −T ν(w(1))−T
(d)
+ (w))−N
+F(w(2))λ(w(2))−m(cid:10)μ(w(2))N −T ν(w(2))−T .
(49)
In the same way,
(P
(d)
− (z))−N P(z)dT P(z)m
= F(z(1))λ(z(1))mμ(z(1))T −N ν(z(1))T + F(z(2))λ(z(2))mμ(z(2))T −N ν(z(2))T .
(50)
By substituting (49) and (50) in the double integral of (12) (with adjusted parameters)
and inserting
(P(z))m−m(cid:10) = F(z(1))λ(z(1))m(cid:10)−m + F(z(2))λ(z(2))m(cid:10)−m
in the single integral one obtains the statement.
(cid:15)(cid:16)
123
A. Borodin, M. Duits
5.4 Proof of Proposition 3.6
Proof of Proposition 3.6 One can easily see that (cid:13)(cid:10)(z)dz has simple poles at 0 and ∞.
On the first sheet (cid:13)(cid:10)(z) takes the form
(1 − τ )μ(cid:10)
1
μ1(z)
(z)
−
τ ν(cid:10)
(z)
1
ν1(z)
+ d(1 − τ + ξ )
z
− d(1 − τ )
z − a2
,
(51)
and we see that we have a simple pole at (a2)(1). On the second sheet we can use the
relations ν1(z)ν2(z) = const · (z − a2)d and μ1(z)μ2(z) = const · (z − a−2)d /zd , to
deduce that (cid:13)(cid:10)(z) takes the form
−
(1 − τ )μ(cid:10)
1
μ1(z)
(z)
+
τ ν(cid:10)
(z)
1
ν1(z)
+ dξ
z
− dτ
z − a−2
,
(52)
and the pole at (a2)(2) gets canceled at the cost of a new simple pole at (a−2)(2). Thus,
(cid:13)(cid:10)(z)dz has four simple poles at said locations and thus also four zeros (since R(z)
is of genus 1).
We now show that there are at least two saddle points in C1, which can be done
using the same argument as in [11, proof of Proposition 6.4]. The point is that one can
show that
(cid:7)
C1
(cid:13)(cid:10)(z)dz = 0.
(53)
Indeed, since ν1(z) and μ1(z) are real-valued for z ∈ (x1, x2), so is (cid:13)(cid:10)(z(1,2)) by (51)
and (52), and thus
(cid:7)
Im
C1
(cid:13)(cid:10)(z)dz = 0.
As for the real part, note that
(cid:7)
C1
(cid:13)(cid:10)(z)dz =
(cid:13)
x2
x1
(cid:13)(cid:10)(z(1))dz −
(cid:13)
x2
x1
(cid:13)(cid:10)(z(2))dz
and, since Re (cid:13) is single-valued on R,
(cid:13)
x2
x1
Re
(cid:13)(cid:10)(z(1))dz = Re (cid:13)(x2) − Re (cid:13)(x1) = Re
(cid:13)
x2
x1
(cid:13)(cid:10)(z(2))dz.
Therefore, also the real part of the left-hand side of (53) vanishes. By combining this
with the fact that (cid:13)(cid:10)(z)dz is real-valued and continuous on C1, we see that (cid:13)(cid:10)(z)dz
must change sign at least two times. This means that there are at least two zeros of
(cid:13)(cid:10)(z)dz. (Note that this argument does not work on C2 since (cid:13)(cid:10)(z)dz has two poles
(cid:15)(cid:16)
on C2.)
123
Biased 2 × 2 periodic Aztec diamond...
Fig. 14 The first deformation of contours. The contours γ (1)
contour ˜γ (1)
is a deformation of the contour γ (1)
. By deforming the contour like this, we pick up a residue
at z = w. Note also that each blue contour can be deformed through the cuts and be entirely, or partly, on
the second sheet. The orange contour is allowed to pass the cuts provided one does not pass through the
origin while deforming (colour figure online)
remain untouched. The blue
and γ (2)
1
, γ (2)
2
1
2
1
5.5 Asymptotic analysis in the smooth phase
We will work out the asymptotic analysis in the smooth phase. We prepare the proof
of Theorem 3.7 by first performing two steps:
1. a preliminary deformation of paths.
2. a qualitative description of the paths of steepest descent and ascent leaving from
the saddle points.
After these steps, the asymptotic analysis follows by standard arguments.
5.5.1 A preliminary deformation
We will need the following lemma on the asymptotic behavior of the integrand in (26)
near the poles at 0 and ∞.
Lemma 5.1 We have that
λ(w)−m(cid:10)wx−m(w − a2)m(cid:10)μ(w)N −T ν(w)−T wd(T −N )+X (w − a2)d(N −T )
O(|w|X +x (cid:10)−dT /2−m(cid:10)/2),
w → ∞,
O(|w|X +x (cid:10)−d(N −T )/2−m(cid:10)/2), w → 0.
=
(cid:18)
(54)
Proof The behavior near w = ∞ follows readily after observing
⎧
⎪⎨
⎪⎩
λ(w) = O(|w|1/2),
μ(w) = O(1),
ν(w) = O(|w|d/2),
as w → ∞.
123
A. Borodin, M. Duits
Similarly, the behavior near w = 0 follows after observing
⎧
⎪⎨
⎪⎩
λ(w) = O(|w|−1/2),
μ(w) = O(|w|−d/2),
ν(w) = O(1),
as w → 0.
(cid:15)(cid:16)
It is important to observe that we are considering (τ, ξ ) in the parallellogram defined
by τ = 0, τ = 1, ξ = τ/2 and ξ = (τ −1)/2. By (13) this means that for any x (cid:10), m(cid:10) ∈ Z
we have that
X − d(N − T )/2
(cid:10) + x (cid:10) − m(cid:10)/2 > 0, X − dT /2 + x (cid:10) − m(cid:10)/2 < 0,
(55)
for N sufficiently large, and thus the left-hand side of (54) has no poles (and no
residues) for either w = 0 or w = ∞.
We proceed with the first contour deformation. The contours γ (1)
2
is deformed to the contour ˜γ (1)
1
and γ (2)
, γ (2)
1 will
2
be untouched, but the contour γ (1)
that goes around the
1
cut (−∞, x1) in clockwise direction, as indicated in Fig. 14. While deforming we pick
up possible residues at the pole at w = ∞ and at w = z for z ∈ γ (1)
. As mentioned
2
above, with our choice of parameters there is no pole at w = ∞. The pole at w = z
has a residue for z ∈ γ (1)
, and this gives us a contribution:
2
1
2πi
(cid:13)
γ (1)
2
Ae(z)−ε(cid:10)
F(z)Ao(z)ελ(z)m−m(cid:10) zm−x−m(cid:10)+x (cid:10)
(z − a2)m−m(cid:10)
.
dz
z
This means that we can rewrite (26) as
(cid:11)
Kd N ((2dT + 2m + ε, 2X + 2x − j), (2dT + 2m(cid:10) + ε(cid:10), 2X + 2x (cid:10) − j (cid:10)))
(cid:12)
1
j, j (cid:10)=0
= −
+
(cid:13)
12m(cid:10)+ε(cid:10)<2m+ε
2πi
12m(cid:10)+ε(cid:10)≥2m+ε
2πi
(cid:7)
γ (2)
2
(cid:13)
γ (1)
2
(cid:7)
Ae(z)−ε(cid:10)
Ae(z)−ε(cid:10)
F(z)Ao(z)ελ(z)m−m(cid:10) zm−x−m(cid:10)+x (cid:10)
(z − a2)m−m(cid:10)
F(z)Ao(z)ελ(z)m−m(cid:10) zm−x−m(cid:10)+x (cid:10)
(z − a2)m−m(cid:10)
dz
z
+ 1
(2πi)2
˜γ (1)
1
(w − a2)m(cid:10)
(z − a2)m
∪γ (2)
γ (1)
1
2
μ(w)N −T
μ(z)N −T
×
Ae(w)−ε(cid:10)
∪γ (2)
2
F(w)F(z)Ao(z)ε λ(z)m
λ(w)m(cid:10)
ν(z)T
ν(w)T
wd(N −T )+X
zd(N −T )X
(z − a2)d(N −T )
(w − a2)d(N −T )
dz
z
wx (cid:10)−m(cid:10)
zx−m
dwdz
z(z − w)
.
(56)
This finishes the preliminary deformation.
Before we continue to the steepest descent analysis, we mention that by (54) and
(55), the integrand with respect to w has no pole at w = 0. This means that we can
deform the contour γ (2)
to lie partly or even entirely on the first sheet. The integrand
1
123
Biased 2 × 2 periodic Aztec diamond...
with respect to z does have poles at z = 0 and z = ∞ and thus, the contours γ (1)
and
1
γ (2)
2 may be deformed over the surface R but cannot pass through the origin (without
picking up a residue).
5.5.2 Description of the paths of steepest descent/ascent
By definition, we have four saddle points in the cycle C1, and in the interior of the
smooth region these are distinct and simple. By viewing Re (cid:13) as a function on the
cycle C1, these saddle points will correspond to the locations of the two local minima
and two local maxima of Re (cid:13). We will denote the saddles associated to the local
minima by s1 and s3, and the local maxima by s2 and s4. We take the indexing such
that when traversing the cycle C1 starting from x1 to x2 on R1 and then from x2 to
x1 on R2, the first saddle point one encounters is s1, then s2 and so on. Note also that
both local minima are neighbors to both local maxima (on the cycle C1) and therefore
Re (cid:13)(s1,3) < Re (cid:13)(s2,4).
We proceed by giving a description of the contours of steepest descent and ascent
for Re (cid:13) leaving from these four saddles. Since each saddle point is simple, there will
be two paths of steepest descent and two path of steepest ascent leaving from them. It
is straightforward that the segment of C1 between s2 j and s2 j±1 is a path of steepest
descent for Re (cid:13) leaving from s2 j and a path of steepest ascent leaving from s2 j±1.
What remains, is to identify the paths of steepest descent leaving from s1 and s3 and
the paths of steepest ascent from s2 and s4. These paths will continue in the lower and
upper half planes of the sheets R j and they are further characterized by the condition
that
(cid:13)
z
s j
Im
!
(cid:13)(cid:10)(z)dz
= 0.
It is important to note that, even though (cid:13)(cid:10)(z) is single-valued, (cid:13)(z) is a multi-valued
function, and we cannot replace the condition simply with Im (cid:13)(z) = Im (cid:13)(s j ).
Indeed, because of the logarithmic terms the imaginary part Im (cid:13)(z) jumps whenever
we cross a cut (which we did not specify) for a logarithm. The real part Re (cid:13)(z),
however, is single-valued, and this will be important. We will also need the behavior
near the logarithmic singularities of Re (cid:13)(z) at z = 0, z = (a2)(1), z = (a−2)(2) and
at z = ∞: from (27) (see also (51) and (52))
Re (cid:13)(0) = Re (cid:13)(∞) = −∞,
and
Re (cid:13)((a2)(1)) = Re (cid:13)((a−2)(2)) = +∞.
(57)
(58)
By analyticity of (cid:13)(cid:10)(z) the paths are a finite union of analytic arcs and ultimately
have to end up at some special points. The only options for such special points are other
saddle points or the poles of (cid:13)(cid:10). It takes only a short argument to exclude possibility
123
A. Borodin, M. Duits
Fig. 15 An illustration of the hypothetical case that the two saddle points s2 and s4 connect to the same
saddle point (1/a2)(2)
. In this case, the four paths together form a contractible curve and enclose the
(shaded) region that contains s3 but not the cycle C2. This means that the steepest descent paths leaving s3
will have to cross the paths from s2 or s4, which is not possible
that they will connect to another saddle point. Indeed, since Re (cid:13)(s1,3) < Re (cid:13)(s2,4)
it is impossible to connect s2 j to s2 j±1 in this way. Moreover, it is obvious that a path
of steepest descent (or ascent) from the global minimum (or maximum) cannot be
connected to any other saddle point, hence s1 cannot be connected to s3 and s2 cannot
be connected to s4. We conclude so far that the path of steepest descent leaving from
s1,3 and the paths of steepest ascent from s2,4 will have to end up at the four simple
poles of (cid:13). From (57) we further deduce s1,3 connect to ∞ and 0, and from (57) we
find that s2,4 connect to the simple poles at (a2)(1) and (1/a2)(2).
Observe that none of these paths can cross each other, since by analyticity of (cid:13)
such a crossing necessarily would be a saddle point (which we already excluded) or a
z (cid:13)(cid:10)(s)ds is constant on the cycles, the paths can
pole. For the same reason, since Im
never cross the cycles C1 and C2 as the point where it would cross would necessarily
be a saddle point. The only point the paths have in common with the cycles are the
saddle point at C1 they started at and the pole at C2 they end in. Hence, a path that starts
at a saddle point on R1 and continues in the upper half plane of R1 always remain in
the union of the upper half plane of R1 and the lower half plane of R2 glued together
along the cuts (−∞, x1) ∪ [x1, x2]. This important property shows, in particular, that
steepest ascent/descent paths do not wind around the poles of (cid:13)(cid:10)(z).
"
Next we argue that the paths of steepest ascent leaving from s2 and s4 cannot end
in the same pole. Indeed, if they would, then all these four paths together would form
a closed loop that is contractible and hence cuts the Riemann surface into two parts,
as illustrated in Fig. 15. The cycle C2 lies fully in one of the parts. But s1 and s3 are in
different parts, and hence there must be one of them that is in a part that is different
from the part that contains the cycle C2. The steepest descent path that leaves that
saddle point has to cross the closed loop in order to end up at a pole on C2, which
is not possible, and we arrive at a contradiction. This means that the steepest ascent
paths from s2 and s4 have to end up at different poles, one saddle connects to (a2)(1)
and the other to (1/a2)(2). A similar argument shows that one of the saddle points s1
and s3 connect, via steepest descent paths, to 0 and the other to ∞.
Let us summarize our findings above:
Proposition 5.2 The steepest descent paths leaving from s1 and s3 and steepest ascent
path from s2 and s4 form simple closed loops on R, such that no two loops intersect,
123
Biased 2 × 2 periodic Aztec diamond...
Fig. 16 The seven pictures illustrate the possible locations (schematically) of the paths of steepest descent
and ascent leaving from the four saddle points on the cycle C1 in the smooth region. In a and b we have
all four saddle points on the first sheet, in pictures c–f we have three saddle points on the first sheet and
in picture g we have one point on the first sheet. It is also possible that all four saddle points are on the
second sheet, and in that case the picture is similar to that of a and b with the two sheets switched (but
keeping the poles a±2 in place and slightly adjusting the contours accordingly). Similarly, for the case
of three saddle point on the second sheet. All pictures can be reconstructed started from the picture in a
by continuous deformations. For example, b can be obtained by moving the right most saddle point (and
the orange contour) in a over the cycle C1, first passing the branch point x1 to the second sheet and then
passing the branch point x2 back to the first sheet to become the left most saddle point at (b). The pictures
(c) and (d) can be obtained from (a) by moving the right most and the left most points respectively to the
second sheet, etc. We did not check whether all configurations indeed occur and perhaps some cases can
be excluded, but our arguments hold for any of the above configurations
each loop intersects each cycle C1 and C2 exactly once, and it does so at a saddle point
for Re (cid:13) in C1 and a pole for (cid:13)(cid:10)(z) in C2.
See Fig. 16 and its caption for an illustration.
5.5.3 Proof of Theorem 3.7
Now we are ready for the
Proof of Theorem 3.7 The starting point is the representation of the kernel after the
preliminary deformation as given in (56). To prove the result, all that is needed is to
123
A. Borodin, M. Duits
show that the double integral tends to zero as N → ∞. This is rather straightforward
after one has realized that the contours of the preliminary deformation strongly resem-
ble the paths of steepest descent and ascent for the saddle point s j . Indeed, the two
contours ˜γ (1)
can be deformed to go through the saddle points s1 and s3 and
1
follow the paths of steepest descent, and the contours γ (1)
can be deformed
2
to the path of steepest ascent ending in z = (a2)(1) and z = (1/a2)(2) respectively.
During this deformation, no additional residues are being picked up, and standard
saddle point arguments show that there exists c > 0 such that
and γ (2)
2
and γ (2)
1
(cid:7)
(cid:7)
˜γ (1)
1
∪γ (2)
1
γ (1)
2
∪γ (2)
2
= O(exp(−N c)),
as N → ∞. This finishes the proof.
(cid:15)(cid:16)
5.6 Proof of Proposition 3.8
Before we come to the proof of Proposition 3.8 we need a few lemmas. We use the
notation (cid:11)x(cid:12) for the largest integer smaller than x.
Lemma 5.3 There exists polynomials p, ˜p with real coefficients and of degree at most
(cid:11) d
(cid:12) and q(0) =
2
˜q(0) = 0, such that
(cid:12), and polynomials q, ˜q with real coefficients, of degree at most (cid:11) d−1
2
ν(z) = p(z) + q(z)(R(z))1/2,
μ(z) = ˜p(1/z) + ˜q(1/z)(R(z))1/2,
(59)
(60)
where R(z) is as in (16) and the square root (R(z))1/2 is such that (R(z))1/2 for z > 0
on R1.
Proof From (17) and (48) we then find that
ν(z) = p(z) + q(z)(R(z))1/2,
for some rational functions p(z) and q(z) with real coefficients. It remains to show
that p and q are in fact polynomials in z of said degree.
By computing the trace of P
(d)
+ (z) we have
Tr P
(d)
+ (z) = ν1(z) + ν2(z) = 2 p(z),
(d)
and thus p(z) is a polynomial. The degree of Tr P
+ (z) can also be estimated from
above. Indeed, for any matrices A j,1, A j,2 for j = 1, . . . , d of the same dimensions
such that
A j,2 A j+1,2 = O,
j = 1, . . . , d − 1,
123
Biased 2 × 2 periodic Aztec diamond...
we have that
d(cid:6)
Tr
(A j,1 + z A j,2)
j=1
is a polynomial of degree at most (cid:11) d
2
for some constant c j , and this shows that p(z) has degree at most (cid:11) d
2
(cid:12). In the case of P
(d)
+ , we have A j,2 = c j
(cid:12) as stated.
Finally, let us consider q(z). We have
(cid:9)
(cid:10)
0 1
0 0
det P
(d)
+ (z) = ν1(z)ν2(z) = p(z)2 + R(z)(q(z))2.
Since the left-hand is a polynomial of degree d, and p(z)2 is a polynomial of degree
at most d, R(z)(q2(z))2 is a polynomial of degree d. Hence, the rational function q
must be a polynomial of degree at most (cid:11) d−1
(cid:12). Moreover, since R(z) has a simple
2
pole at z = 0, the polynomial q(z) must have a zero at z = 0.
The statement for μ follows in the same way.
(cid:15)(cid:16)
Lemma 5.4 We have |λ1(z)| > |λ2(z)|, |μ1(z)| > |μ2(z)|, and |ν1(z)| > |ν2(z)| for
z ∈ C \ ((−∞, x1) ∪ (x2, 0]).
Proof The proof is the same for all three cases, so we only prove that |μ1(z)| > |μ2(z)|.
To this end, we note that μ2(z)/μ1(z) is analytic on C \ ((−∞, x1] ∪ [x2, 0]). It has
a zero at z = a2 and a possible pole at z = 0. However, from (60) it follows that the
singularity at z = 0 is removable. Moreover, μ2(z)/μ1(z) → 1 as z → ∞. From (60)
it also follows that |μ2(z)/μ1(z)| = 1 for z ∈ (−∞, x1) ∪ (x2, 0). By the maximum
modulus principle we must have either |μ2(z)/μ1(z)| > 1 or |μ2(z)/μ1(z)| < 1, for
z ∈ C\((−∞, x1) ∪ (x2, 0]). Since μ2(a2) = 0, we conclude that |μ2(z)/μ1(z)| < 1.
(cid:15)(cid:16)
We also need the behavior of μ near the branch point at ∞.
Lemma 5.5 With
we have
(cid:14) =
d−1(cid:6)
j=0
a(σ j (x, y))b(σ j (x, y))
(61)
μ1(z) = (cid:14)
⎛
⎜
⎝1 + a
z1/2
⎛
⎝
d−1$
j=0
a(σ j (x, y))
d−1$
k=0
⎞
⎞
1/2
1
a(σ k(x, y))
⎠
⎟
⎠ + O(z−1), (62)
123
A. Borodin, M. Duits
and
μ2(z) = (cid:14)
⎛
⎜
⎝1 − a
z1/2
⎛
⎝
d−1$
j=0
a(σ j (x, y))
d−1$
k=0
⎞
⎞
1/2
1
a(σ k(x, y))
⎠
⎟
⎠ + O(z−1),
as z → ∞ along the positive real axis, and the square root is taken such that z1/2 > 0.
Proof A simple computation gives
(d)
− (z) = (cid:14)
P
(cid:25)(cid:25)
1
0
(cid:27)
d−1
j=0 a(σ j (x, y))
1
(cid:26)
(cid:25)
(cid:27)
+ a2
z
d−1$
k=0
1
0
(cid:26) (cid:9)
k−1
j=0 a(σ j (x, y))
1
0
0
a(σk (x, y))−1 0
(cid:10) (cid:25)
(cid:27)
(cid:26)
d−1
j=k+1 a(σ j (x, y))
1
1
0
(cid:26)
+ O(z−2)
,
Tr P
(d)
− (z) = (cid:14)
as z → ∞. Hence,
⎛
⎝2 + a2
z
⎛
⎝2 + a2
z
= (cid:14)
d−1$
d−1$
j=0, j(cid:17)=k
k=0
⎛
⎝
d−1$
j=0
⎞
a(σ j (x, y))
a(σ k(x, y))
+ O(z−2)
⎠
a(σ j (x, y))
d−1$
k=0
1
a(σ k(x, y))
⎞
⎞
− d
⎠ + O(z−2)
⎠ ,
as z → ∞. Since det P
(d)
− (z) = (cid:14)2(1 − a2/z)d , we find
μ1,2 = (cid:14)
⎛
⎜
⎝1 ± a
z1/2
⎛
⎝
d−1$
j=0
a(σ j (x, y))
d−1$
k=0
⎞
⎞
1/2
1
a(σ k(x, y))
⎠
⎟
⎠ + O(z−1),
as z → ∞. It remains to determine whether μ1 or μ2 comes with the plus sign. Since
μ1(z) > μ2(z), we see that μ1 comes with the plus sign and μ2 with the minus sign.
(cid:15)(cid:16)
Now we are ready for the
Proof of Proposition 3.8 By (59) we have
ν1(z) = p(z) + q(z)
(cid:29)
R(z),
ν2(z) = p(z) − q(z)
(cid:29)
R(z),
with R(z) = a2(z − x1)(z − x2)/z and the square root is taken so that
R(z) > 0 for
z > 0. Here p(z) is a polynomial of degree at most (cid:11)d/2(cid:12) and q is a polynomial of
degree (cid:11)(d − 1)/2(cid:12) with a zero at z = 0.
√
123
Biased 2 × 2 periodic Aztec diamond...
Observe that ν(cid:10)
1
(z)ν2(z) can be written as
(cid:9)
ν(cid:10)
1
(z)ν2(z) =
p(cid:10)(z) + q(cid:10)(z)
√
= r1(z) + r2(z)
R(z)
z
√
(cid:29)
√
R(z) + q(z)R(cid:10)(z)
R(z)
R(z)
2
,
(cid:10)
( p(z) − q(z)
(cid:29)
R(z))
(63)
where
and
r1(z) = 2zq(cid:10)(z) p(z)R(z) + zq(z) p(z)R(cid:10)(z) − 2 p(cid:10)(z)q(z)z R(z),
r2(z) = 2zp(cid:10)(z) p(z) − 2zq(cid:10)(z)q(z)R(z) + zq(z)2 R(cid:10)(z).
Since q(0) = 0 and R(cid:10)(z) has double pole at z = 0, r1 and r2 are polynomials and
R(z) in
r2(0) = 0. The degree of r1 and r2 is at most d. By replacing
the derivation above we also find
R(z) by −
√
√
ν(cid:10)
2
(z)ν1(z) =
−r1(z) + r2(z)
√
R(z)
z
√
R(z)
.
Therefore, we can write
2r1(z) =
(cid:3)
ν(cid:10)
1
(z)ν2(z) − ν(cid:10)
2
(z)ν1(z)
(cid:4)
(cid:29)
z
R(z),
and
2r2(z) = z
(cid:3)
ν(cid:10)
1
(z)ν2(z) + ν(cid:10)
2
(z)ν1(z)
(cid:4)
(64)
Since ν2(z) has a zero of order d at z = a−2, this means that both r1 and r2 have a
zero of order d − 1 at z = a−2. This implies that ν(cid:10)
1
(z)ν2(z) can be written as
ν(cid:10)
1
(z)ν2(z) = d
(z − a−2)d−1
(cid:3)
γ1 + γ2z + γ3z
√
z
R(z)
√
(cid:4)
R(z)
and thus
ν(cid:10)
(z)
1
ν1(z)
=
ν(cid:10)
(z)ν2(z)
1
ν1(z)ν2(z)
= d
γ1 + γ2z + γ3z
√
(z − a−2)z
√
R(z)
R(z)
,
(65)
where γ j ∈ R, for j = 1, 2, 3, are some real constants.
By a similar reasoning, one can show that
μ(cid:10)
(z)
1
μ1(z)
= da2 z ˜γ1 + ˜γ2 + ˜γ3
√
(z − a2)z
√
R(z)
R(z)
,
(66)
123
for some real parameters ˜γ j , for j = 1, 2, 3.
The next step is to compute the values of the constants γ j , ˜γ j for j = 1, 2, 3. To
this end, add (65) and (66) to obtain
A. Borodin, M. Duits
(γ1 + γ2z)(z − a2) + a2( ˜γ1z + ˜γ2)(z − a−2)
λ(cid:10)
(z)
1
λ1(z)
d
=
μ(cid:10)
(z)
1
μ1(z)
(cid:3)
+d
+
ν(cid:10)
(z)
1
ν1(z)
= d
zγ3(z − a2) + a2 ˜γ3(z − a−2
(z − a2)(z − a−2)z
√
(z − a2)(z − a−2)z
√
R(z)
(cid:4)
)
R(z)
√
R(z)
.
(67)
On the other hand, we easily compute from (15)–(17) that
λ(cid:10)
(z)
1
λ1(z)
=
λ(cid:10)
(z)λ2(z)
1
λ1(z)λ2(z)
(cid:3)
(z2 − 1)
= −
(cid:3)
z2 R(cid:10)(z)
= −
1
2
(1 + a2)(α + 1/α) − 1
2
R(z)
4a2z(z − a2)(z − a−2)
√
√
(cid:4)
(1 + a2)(α + 1/α) −
R(z)
√
2z(z − a2)(z − a−2)
R(z)
√
(cid:4)
R(z)
(68)
where we used z2 R(cid:10)(z) = 4a2(z2 − 1) in the last step. Comparing (67) and (68) leads
to the following two equations:
(γ1 + γ2z)(z − a2) + a2( ˜γ1z + ˜γ2)(z − a−2) = − 1
2
(z2 − 1)(1 + a2)(α + 1/α),
and
From (70) we find
zγ3(z − a2) + a2 ˜γ3(z − a−2) = 1
2
(z2 − 1).
γ3 = ˜γ3 = 1
2
,
˜γ1 = γ2,
˜γ2 = γ2
and from (69) we find
and
a2γ1 + γ2 = − 1
2
(1 + a2)(α + 1/α).
(69)
(70)
(71)
(72)
(73)
Thus far, we have derived the first two identities in (30).
The value of γ1 can be computed by comparing the asymptotic expansion for the
logarithmic derivative for μ1 from (66) and (62), and comparing the results. Indeed,
123
Biased 2 × 2 periodic Aztec diamond...
from (62) we find
μ(cid:10)
(z)
1
μ1(z)
= − a
2z3/2
⎛
⎝
d−1$
j=0
a(σ j (x, y))
d−1$
k=0
⎞
1/2
1
a(σ k(x, y))
⎠
+ O(1/z2),
as z → ∞, and from (66) we find, using γ1 = ˜γ1 and (16), that
μ(cid:10)
(z)
1
μ1(z)
= daγ1
z3/2
+ O(1/z2),
as z → ∞. Therefore
γ1 = − 1
2
⎛
⎝ 1
d
d−1$
j=0
a(σ j (x, y)) 1
d
d−1$
k=0
1
a(σ k(x, y))
1/2
⎞
⎠
,
(74)
which is the third identity in (30).
Finally, by substituting (65) and (66) using (71), (72), (73) and (74) into (cid:13)(cid:10)(z), and
(cid:15)(cid:16)
using analytic continuation to R, we obtain (29).
5.7 Proof of Lemma 3.11
Proof of Lemma 3.11 The cycle condition (53) implies that (using (29))
(1 − τ )a2
(cid:13)
x2
x1
xγ1 + γ2
√
(x − a2)x
R(x)
d x − τ
(cid:13)
x2
x1
γ1 + xγ2
√
(x − a−2)x
R(x)
d x = 0.
(75)
By a change of variable x (cid:8)→ 1/x we find (using (16))
(cid:13)
x2
x1
a2
xγ1 + γ2
√
(x − a2)x
R(x)
d x = −
(cid:13)
x2
x1
γ1 + xγ2
√
(x − a−2)x
d x,
R(x)
and after substituting this into the first integral, (75) reduces to
(cid:13)
x2
x1
γ1 + xγ2
√
(x − a−2)x
R(x)
d x = 0,
after which the statement easily follows.
(cid:15)(cid:16)
Acknowledgements The authors are grateful to B. Poonen for directing them to [28]. Figure 1 was plotted
using a code that was kindly provided to us by S. Chhita. We thank T. Berggren and M. Bertola for
helpful discussions. A. Borodin was partially supported by the NSF grant DMS-1853981, and the Simons
Investigator program. M. Duits was partially supported by the Swedish Research Council (VR), grant no
2016-05450 and grant no. 2021-06015, and the European Research Council (ERC), Grant Agreement No.
101002013.
Funding Open access funding provided by Royal Institute of Technology.
123
A. Borodin, M. Duits
Data availability Data sharing not applicable to this article as no datasets were generated or analysed during
the current study.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence,
and indicate if changes were made. The images or other third party material in this article are included
in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If
material is not included in the article’s Creative Commons licence and your intended use is not permitted
by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
A Example: torsion point of order six
Let us now assume
a2 =
α
1 + α + α2
.
(76)
Then, as discussed in Sect. 3.1, (a−2, a−2) is a torsion point of order six. Here we
will compute the spectral curves for μ and ν, and derive a degree eight equation for
the boundary of the rough disordered region. We will also show, numerically, how
the steepest descent/ascent path can be chosen when (τ, ξ ) are in the center of the
diamond.
A.1 The flow on the matrices
The linear flow on the elliptic curve is given already in (8). The flow on the matrices
and their decomposition can then be traced giving:
(cid:9)
1 aα2
a
1
zα2
(cid:10)
1 a
a
z 1
(cid:10)
(cid:9)
(cid:10)
(cid:8)→ P
(1)
− (z) =
(cid:9)
1 aα3
a
1
zα3
(cid:10)
(cid:8)→ P
− (z) = 1
(2)
α
(cid:9)
(cid:10)
(cid:8)→ P
(4)
− (z) =
(cid:8)→ P
(5)
− (z) = α
(cid:9)
(cid:10)
1 aα2
a
1
zα2
(77)
1 a
α
aα
z 1
1 a
a
z 1
(cid:9)
(0)
− (z) = α
P
(cid:8)→ P
− (z) = 1
(3)
α
and
(0)
+ (z) =
P
(cid:8)→ P
(3)
+ (z) =
(cid:9)
(cid:9)
(cid:10)
1 az
a
1
α2
α2
1 aα2z
a α2
(1)
+ (z) =
(cid:8)→ P
(cid:10)
(cid:8)→ P
(4)
+ (z) =
(cid:9)
(cid:10)
1 aαz
a
1
α
(cid:9)
1 aαz
a
1
α
(2)
+ (z) =
(cid:8)→ P
(cid:10)
(cid:8)→ P
(5)
+ (z) =
(cid:9)
1 aα2z
a α2
(cid:9)
1 az
a
1
α2
α2
(cid:10)
(cid:10)
. (78)
From here we can compute P
(18).
(d)
± as in (11) and (10) and the spectral curves (19) and
123
Biased 2 × 2 periodic Aztec diamond...
The discriminant R as in (16) can be written as
R(z) = (a−2 − 1)2(a2 + 1)2 + 4a2(z + 1/z − a2 − a−2).
Straightforward computations give det P
(d)
+ (z) = (a2z − 1)6 and
Tr P
(d)
+ (z) = 2 +
(1 − 6a4 + 3a8)z
a6
+ 2(2 − 3a4)z2 + 2a6z3.
The discriminant then becomes
Tr P
(d)
+ (z) − 4 det P
(d)
+ (z)
= a−12z2(1 − 3a4 + 2a6z)2(1 − 6a4 − 3a8 + 4a6(z + 1/z)),
so that
ν(z) = 1 +
(1 − 6a4 + 3a8)z
2a6
+ (2 − 3a4)z2 + a6z3
± z(1 − 3a4 + 2a6z)
2a4
(R(z))1/2.
Similarly, det P
(d)
− (z) = (z − a2)6/z6 and
Tr P
(d)
− (z) = 2 + (2a6z−3 + 4z−2 − 6a4z−2 + ((1 − 6a4 + 3a8)z−1)/a6).
The discriminant then becomes
Tr P
(d)
− (z) − 4 det P
(d)
− (z)
= a−12z−2(1 − 3a4 + 2a6z−1)2(1 − 6a4 − 3a8 + 4a6(z + 1/z)),
so that
μ(z) = 1 +
(1 − 6a4 + 3a8)z−1
2a6
+ (2 − 3a4)z−2 + a6z3
± z−1(1 − 3a4 + 2a6z−1)
2a4
(R(z))1/2.
Note that μ(z) = ν(1/z).
Now that we have μ and ν we can compute the saddle point equation (cid:13)(cid:10)(z)dz = 0.
We start with computing the logarithmic derivatives of μ and ν:
μ(cid:10)(z)
μ(z)
=
−2 + (−1/a2 + 3a2)z + 3a2(R(z))1/2
(z − a2)z(R(z))1/2
(79)
123
and
ν(cid:10)(z)
ν(z)
=
(−1/a2 + 3a2) − 2z + 3a2z(R(z))1/2
(a2z − 1)z(R(z))1/2
.
From the above expressions we can read off the values for γ j :
γ1 = − 1
6a4
+ 1
2
, γ2 = − 1
3a2
,
γ3 = 1
2
.
A. Borodin, M. Duits
(80)
(81)
This, together with (31), allows us to compute the four saddle points as a function of
the parameters a, τ and ξ . The expressions are rather long, and we omit them here.
Instead, we will provide some numerical results in the next subsection.
A.2 Contours of steepest descent/ascent
We will plot the contours of steepest descent/ascent for Re (cid:13), with (cid:13) as in (27), for
the special values
a2 = 1
3
− 1
100
,
τ = 1
2
,
ξ = 0.
This is the midpoint of the Aztec diamond, where we indeed expect a smooth disordered
region to appear. Indeed, with this choice of parameters, we find four saddle points on
the cycle C1. Two of them are on the first sheet:
s1 = −1.97156,
s2 = −0.833032,
and the other two on the second sheet:
s3 = −0.507212,
s4 = −1.20043.
The branch points are at
x1 = −2.01885,
x2 = −0.495331.
Observe that s1 is close to x1 and s3 is close to x2, which we found to be typical
for any choice of parameters. This has the unfortunate consequence that in numerical
illustrations the saddle points s1 and s3 are hard to distinguish from the branch points
x1 and x2, respectively.
The contours of steepest descent/ascent leaving from the saddle points locally
coincide with the level lines of Im (cid:13)(z). The problem is that (cid:13)(z) has logarithmic
terms making Im (cid:13)(z) multi-valued and plotting the level sets Im (cid:13)(z) = Im (cid:13)(s j )
does not give the correct result. For this reason, we compute the vectorfield given by
(Re (cid:13)(cid:10), − Im (cid:13)(cid:10)) and compute the streamlines using the function Streamplot in
Mathematica. The results are given in Fig. 17.
123
Biased 2 × 2 periodic Aztec diamond...
Fig. 17 a Level lines for Im (cid:13). The paths of steepest descent for Re (cid:13) from s1 connect to infinity on the
first sheet. The paths of steepest ascent from s2 end up in a2. b Level lines for Im (cid:13). The paths of steepest
ascent from s2 end up in a−2. To see the paths of steepest descent from s3 we need to zoom in. c Zooming
in on the segment (x2, 0) shows that the path of steepest descent for Re (cid:13) starting from the saddle point s3
end in the origin
From Fig. 17 one can see that the paths of ascent/descent are indeed as illustrated
schematically in Fig. 16g. On the first sheet, the paths of steepest descent leaving from
s1 end up at ∞ and the paths of steepest ascent leaving from s2 end up at a2. On the
second sheet, the paths of steepest descent leaving from s3 end up at 0 and the paths of
steepest ascent leaving from s4 end up at a−2. The statement for s3 is not immediately
obvious from Fig. 17b and this is why we zoom in around s3 as in Fig. 17c.
A.3 Boundary of the rough disordered region
The last result for the case of order six that we will present here, is an explicit expression
for the boundary of the rough disordered region. We follow the procedure indicated in
Sect. 3.4 with a and α related by (76) and values for γ j , j = 1, 2, 3 as in (81). We start
with (31), square both sides and remove the additional factors (z − a2)(z − a−2)/z.
Then we obtain an equation that is polynomial in z and of order four. The values of
(τ, ξ ) where the discriminant vanishes lead to a double saddle point and this will be
the boundary of the liquid region.
123
A. Borodin, M. Duits
The discriminant has degree twelve in τ and ξ , and for general parameters a (and
α related by (76)) the expression is rather long. In order to obtain a shorter expression
it will be convenient to perform the following change of variables
(τ, ξ ) = ((q + v + 1)/2, q/2).
These coordinates change the parallellogram in the left panel into the tilted square in
the right panel of Fig. 12.
The discriminant has two factors. The first factor, of degree four in v and q, reads
(−1 + 9a4 + 9q2 − 9a4q2 − 9a4v2 + 9a8v2)2.
The zero set of this factor is a hyperbola that lies outside the Aztec diamond, and
hence this factor does not contribute to the boundary for the rough region. The second
factor, of degree eight in v and q, is the factor that defines the boundary for the rough
disordered region:
0 = 16 − 336a4 + 1440a8 + 7776a12 − 34992a16 − 104976a20 − 288q2
+6336a4q2 − 45504a8q2 + 124416a12q2
−209952a16q2 + 419904a20q2 + 1296q4 − 32400a4q4
+242352a8q4 − 587088a12q4 + 839808a16q4
−629856a20q4 + 23328a4q6 − 303264a8q6 + 769824a12q6
−909792a16q6 + 419904a20q6 + 104976a8q8
−314928a12q8 + 314928a16q8 − 104976a20q8 − 72v2
+1152a4v2 − 1224a8v2 − 43200a12v2 + 75816a16v2
+419904a20v2 − 157464a24v2 + 1296q2v2
−20088a4q2v2 + 119880a8q2v2 − 527472a12q2v2
+1283040a16q2v2 − 997272a20q2v2 + 472392a24q2v2
−5832q4v2 + 81648a4q4v2 − 367416a8q4v2
+863136a12q4v2 − 833976a16q4v2 + 734832a20q4v2
−472392a24q4v2 + 52488a4q6v2 − 472392a8q6v2
+944784a12q6v2 − 524880a16q6v2 − 157464a20q6v2
+157464a24q6v2 + 81v4 − 1215a4v4 − 3483a8v4
+79461a12v4 − 2349a16v4 − 750141a20v4
+570807a24v4 − 59049a28v4 − 1458q2v4
+21870a4q2v4 − 158922a8q2v4 + 867510a12q2v4 − 1963926a16q2v4
+1418634a20q2v4 − 301806a24q2v4
+118098a28q2v4 + 6561q4v4 − 98415a4q4v4 + 452709a8q4v4
−387099a12q4v4 − 610173a16q4v4
+964467a20q4v4 − 269001a24q4v4 − 59049a28q4v4 + 5832a8v6
123
Biased 2 × 2 periodic Aztec diamond...
−52488a12v6 − 128304a16v6
+734832a20v6 − 717336a24v6 + 157464a28v6
+52488a8q2v6 − 472392a12q2v6 + 944784a16q2v6
−524880a20q2v6 − 157464a24q2v6 + 157464a28q2v6
+104976a16v8 − 314928a20v8 + 314928a24v8 − 104976a28v8.
For α = 1 (and thus a2 = 1/3) this can be reduced to
0 = (3q2 + v2)3(−3 + 12q2 + 4v2).
The first factor is only zero for (q, v) = (0, 0), and what is left is the boundary for the
smooth disordered region. The second factor is an ellipse.
Finally, for α ↓ 0 (and hence a ↓ 0 simultaneously) the curve reduces to
0 = (1 − 9q2)2(4 − 9v2)2,
which gives a rectangular shape. In this case, there is no rough disordered region, but
only a frozen region and a smooth disordered region.
In Fig. 18 we have plotted the boundary of the rough disordered region for several
particular values of a.
B Computation of torsion points
In Sect. 3.1 we gave a few examples of particular choices for the parameters α and
a such that (1/a2, 1/a2) is a torsion point. Here we will indicate how one can find
such examples by recalling the notion of division polynomials. This is a standard
construction for finding the torsion subgroups of the elliptic curve. We will base our
discussion on [28, p. 105].
First, let us introduce new variables
Y = y
2
(a + 1/a)(α + 1/α),
X = x,
and rewrite (5) as
Y 2 = X 3 +
(cid:19)
(a + 1/a)(α + 1/α)2/2 − a2 − 1/a2
(cid:20)
X 2 + X .
123
A. Borodin, M. Duits
Fig. 18 The boundary of the rough disordered regions in the (v, q) plane for the values a = 0, a = 0.4,
a = .55 and a = 1
3 the smooth
3
disordered region has disappeared
3. For a = 0, the rough disordered region has disappeared. For a = 1
3
√
√
In the new variables, we ask for what choices of α and a we have that (1/a2, (a+1/a)(α+1/α)
is a torsion point. With the same notation as in [28, p. 105] we define
2a2
)
(a + 1/a)2(α + 1/α)2,
⎧
⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎩
a2 = −a2 − 1/a2 + 1
4
a4 = 1,
b2 = 4a2,
b4 = 2,
b8 = −1,
123
Biased 2 × 2 periodic Aztec diamond...
and the remaining parameters a1 = a3 = a6 = b6 = 0. Then we define
⎧
⎪⎪⎪⎨
⎪⎪⎪⎩
ψ1 = 1,
ψ2 = 2Y ,
ψ3 = 3X 4 + b2 X 3 + 3b4 X 2 + b8,
ψ4 = ψ2
(cid:3)
2X 6 + b2 X 5 + 5b4 X 4 + 10b8 X 2 + b2b8 X + b4b8
(cid:4)
,
and ψk with k ≥ 5 recursively using:
(cid:18)
ψ2m+1 = ψm+2ψ 3
m
ψmψm+2 − ψm−1ψmψ 2
ψ2ψ2m = ψ 2
− ψm−1ψ 3
m+1
,
m−1
m+1
m ≥ 2,
, m ≥ 3.
The torsion subgroup of order m consists of all zeros of ψm, which are called division
polynomials.
Note that we are not looking for the entire subgroup, but for situations where
(cid:9)
1/a2,
(cid:10)
(a + 1/a)(α + 1/α)
2a2
is a torsion point. After substituting this point into ψm we find a rational function in
a and α. This rational function will have several zeros, but we are only interested in
the zeros that satisfy 0 < α < 1 and 0 < a ≤ 1. For instance, for m = 3 we find the
following equation
−
(1 + a2)2(−1 − α + a2α − α2)(1 − α + a2α + α2)
a8α2
= 0.
This equation has no solutions such that 0 < α < 1 and 0 < a ≤ 1, and thus a third
order torsion point cannot occur.
With the help of a computer code, we found the following equations that give proper
solutions such that we have a torsion point of order m = 4, . . . , 8:
m = 4 : a = 1,
m = 5 : 0 = −a4 + α − a2α + α2 − 2a2α2 − 2a4α2 + α3 + a2α3 − 3a4α3
+ a6α3 + α42a2α4 − 2a4α4 + α5 − a2α5 − a4α,
m = 6 : 0 = (1 + α + α2)a2 − α
m = 7 : 0 = a4 + a4α − a6α + a8α − a10α + 5a4α2 + 2a6α2 − a8α2 − α3
+ 3a2α3 − a4α3 + 5a6α3 − 4a8α3 − 2a10α3 − α4
+ 4a2α4 + 5a4α4 + 12a6α4 − 5a8α4 − α5 − a2α5 + 12a4α5
− 7a8α5 − 3a10α5 − α6 + 2a2α6 + 17a4α6 + 7a8α6
− 6a10α6 + a12α6 − α7 − a2α7 + 12a4α7 − 7a8α7 − 3a10α7 − α8 + 4a2α8
123
+ 5a4α8 + 12a6α8 − 5a8α8 − α9 + 3a2α9 − a4α9 + 5a6α9 − 4a8α9 − 2a10α9
+ 5a4α10 + 2a6α10 − a8α10 + a4α11 − a6α11 + a8α11 − a10α11 + a4α12,
m = 8 : 0 = a − α + a2α + aα2.
A. Borodin, M. Duits
Each term on the right-hand side is a factor of ψm.
References
1. Beffara, V., Chhita, S., Johansson, K.: Airy point process at the liquid-gas boundary. Ann. Probab.
46(5), 2973–3013 (2018)
2. Beffara, V., Chhita, S., Johansson, K.: Local geometry of the rough-smooth interface in the two-periodic
Aztec diamond, arXiv:2004.14068
3. Belokolos, E.D., Bobenko, A.I., Enol’skii, V.Z., Its, A.R., Matveev, V.B.: Algebro-Geometric Approach
to Nonlinear Integrable Equations, Springer Series in Nonlinear Dynamics. Springer, Berlin (1994)
4. Berggren, T.: Domino tilings of the Aztec diamond with doubly periodic weightings, arXiv:1911.01250
5. Berggren, T., Duits, M.: Correlation functions for determinantal processes defined by infinite block
Toeplitz matrices. Adv. Math. 356, 106766, 48 pp (2019)
6. Bertola, M.: Abelianization of Matrix Orthogonal Polynomials, arXiv:2107.12998
7. Chhita, S., Johansson, K.: Domino statistics of the two-periodic Aztec diamond. Adv. Math. 294,
37–149 (2016)
8. Chhita, S., Young, B.: Coupling functions for domino tilings of Aztec diamonds. Adv. Math. 259,
173–251 (2014)
9. Dubrovin, B.A.: Finite-zone linear operators and Abelian varieties. Usp. Mat. Nauk 31(4), 259–260
(1976)
10. Dubrovin, B.A.: Completely integrable Hamiltonian system associated with matrix operators and
Abelian varieties. Funct. Anal. Appl. 11, 28–41 (1977)
11. Duits, M., Kuijlaars, A.B.J.: The two periodic Aztec diamond and matrix valued orthogonal polyno-
mials. J. Eur. Math. Soc 23(4), 1075–1131 (2021)
12. Elkies, N., Kuperberg, G., Larsen, M., Propp, J.: Alternating sign matrices and Domino Tilings (part
I). J. Algebraic Comb 1, 111–132 (1992)
13. Eynard, B., Mehta, M.L.: Matrices coupled in a chain. I. Eigenvalue correlations. J. Phys. A 31,
4449–4456 (1998)
14. Gessel, I., Viennot, G.: Binomial determinants, paths, and hook length formulae. Adv. Math. 58, 300–
321 (1985)
15. Gorin, V.: Lectures on Random Lozenge Tilings (Cambridge Studies in Advanced Mathematics).
Cambridge University Press, Cambridge (2021)
16. Its, A.R.: Canonical systems with finite-zone spectrum and periodic solutions of the nonlinear
Schrödinger equation. Vestn. Leningr. Gos. Univ. 7(2), 39–46 (1976)
17. Johansson, K.: Random matrices and determinantal processes. In: Bovier, A., et al. (eds.) Mathematical
Statistical Physics, Elsevier B.V., Amsterdam, pp. 1–55 (2006)
18. Jonansson, K.: Non-intersecting paths, random tilings and random matrices. Probab. Theory Related
Fields 123, 225–280 (2002)
19. Johansson, K.: The arctic circle boundary and the Airy process. Ann. Probab. 33, 1–30 (2005)
20. Johansson, K., Mason, S.: Dimer-dimer correlations at the rough-smooth boundary, arXiv:2110.14505
21. Kasteleyn, P.W.: Dimer statistics and phase transitions. J. Math. Phys. 4, 287–293 (1963)
22. Kenyon, R., Okounkov, A., Sheffield, S.: Dimers and amoebae. Ann. Math. 163, 1019–1056 (2006)
23. Krichever, I.M.: Algebraic curves and commuting matrix differential operators. Funct. Anal. Appl.
10(2), 144–146 (1976)
24. Krichever, I.M.: Integration of nonlinear equations by the methods of algebraic geometry. Funct. Anal.
Appl. 11(1), 12–26 (1977)
25. Lindström, B.: On the vector representations of induced matroids. Bull. Lond. Math. Soc. 5, 85–90
(1973)
123
Biased 2 × 2 periodic Aztec diamond...
26. Moser, J., Veselov, A.P.: Discrete versions of some classical integrable systems and factorization of
matrix polynomials. Commun. Math. Phys. 139, 217–243 (1991)
27. Okounkov, A., Reshetikhin, N.: Correlation function of Schur process with application to local geom-
etry of a random 3-dimensional Young diagram. J. Am. Math. Soc. 16, 581–603 (2003)
28. Silverman, J.H.: The Arithmetic of Elliptic Curves, Graduate Texts in Mathematics 106, Springer
(2009)
29. Stanley, R.P.: Enumerative Combinatorics, vol. 2. Cambridge University Press, Cambridge (1999)
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.
123
| null |
10.1088_1361-648x_ad1135.pdf
|
Data availability statement
All data that support the findings of this study are included
within the article (and any supplementary files).
|
Data availability statement All data that support the findings of this study are included within the article (and any supplementary files).
|
J. Phys.: Condens. Matter 36 (2024) 115102 (10pp)
Journal of Physics: Condensed Matter
https://doi.org/10.1088/1361-648X/ad1135
Two modes of motions for a single disk
on the vibration stage
Liyang Guan1, Li Tian1, Meiying Hou2,3,∗
and Yilong Han1,∗
1 Department of Physics, Hong Kong University of Science and Technology, Clear Water Bay, Hong
Kong Special Administrative Region of China
2 Key Laboratory of Soft Matter Physics, Beijing National Laboratory for Condensed Matter Physics,
Institute of Physics, Chinese Academy of Sciences, Beijing, People’s Republic of China
3 College of Physics, University of Chinese Academy of Sciences, Beijing, People’s Republic of China
E-mail: [email protected] and [email protected]
Received 25 July 2023, revised 3 November 2023
Accepted for publication 30 November 2023
Published 12 December 2023
Abstract
The motion of a single granular particle is important for understanding their collective motions
on vibration stage, but it remains poorly studied for simple shaped particles, such as a disk.
Here, we systematically measure the motions of a single disk with different vibration amplitudes
A at a fixed vibration frequency f or a fixed acceleration a. The distributions, time-correlations,
and power spectra of displacements per step, mean squared displacements and couplings for
translational and rotational motions are measured. These analyses reveal that the motions
randomly switch between active and inactive modes. Both a and f are important to particle’s
motions and the fraction of active mode. The translational and rotational kinetic energy deviates
from Boltzmann distribution and violates the equipartition theorem in each mode. We find three
types of motion: rolling, lying flat, and fluttering, which give rise to active and inactive modes.
The translational and rotational mean squared displacements, autocorrelations, and power
spectra at different a collapse in active modes, because the disk rolls along its rim with a fixed
inclination angle though our system is under vibration and confinement. The nonzero
cross-correlations between particle’s translational and rotational motions indicate that only
translational motions are insufficient for understanding dense particle systems.
Supplementary material for this article is available online
Keywords: vibration-driven disk, two modes of motion, non-Brownian motions
1. Introduction
Granular particles driven by a vibration or air-blowing stage
[1] have served as a platform for the studies of non-equilibrium
physics [2, 3] and industrial processes, such as ore and
powder separation [1]. They exhibit rich dynamics and pat-
tern formations, such as compaction and segregation [4]
for dense granular particles. For dilute granular gases and
∗
Authors to whom any correspondence should be addressed.
single particle, translational motions have been well stud-
ied, in which displacements per step usually exhibit non-
Gaussian distributions [5–9]. However, the rotational motions
are mainly studied in simulations [10–12], and experimental
measurements are limited [5, 7, 13–15].
Single granular particle, such as the Euler’s disk [16] or
a rolling ring [17] on a fixed table, or a bouncing droplet
on its self-activated surface wave [18, 19], exhibit interesting
motions. For a single particle driven on a vertically vibrating
stage, dimer [20, 21], trimer [22], rod [13, 23–26], and polar
disk [27] have been experimentally measured. These particles
are asymmetric in mass distribution or shape, and some of
1
© 2023 IOP Publishing Ltd
J. Phys.: Condens. Matter 36 (2024) 115102
L Guan et al
them exhibit self-propulsive translational motions on vibra-
tion stage [22, 25–27]. The rotational motion has been meas-
ured for single rod [13] and asymmetric particles [28, 29] in
experiments, and for chiral granular motors in simulations [30,
31]. As one of the simplest shaped isotropic particles, a disk
has recently been studied under a fixed vibration amplitude of
the stage [32]. Here, we study the disk’s motions under differ-
ent vibration amplitudes, frequencies, and accelerations, and
reveal two modes of motions for the first time.
2. Experimental system
The disk with a diameter d = 10 mm, thickness 3.5 mm, and
mass m = 2.0 g is made of the cartridge plastic material
DurusWhite by a 3D printer (Connex350) (figure 1(a)). To
prevent its flipping under vibration, the disk is confined in
a quasi-2D polyethylene terephthalate container whose inner
surfaces are coated by 0.125 mm thick conductive indium-tin-
oxide thin films to prevent electric charges building up from
frictions. The measured static friction coefficient is 0.157 ±
0.001, and the restitution coefficient between the disk and the
container’s surfaces is 0.93. The container is firmly fixed to the
vibration stage (Zhengyi VS-1000VH-51) and levelled hori-
◦
zontally to an accuracy of 0.1
. The container is uniformly
illuminated by LED lights between the container and vibra-
tion stage. A CMOS (complementary metal oxide semicon-
ductor) camera (Lunemera lt225) is placed above the con-
−1 for about
tainer to record the disk’s motions at 150 frames s
5−10 min before the disk hits the boundary of the container.
The translational and rotational motions of the disk are meas-
ured by tracking the position and orientation of the black line
printed on the disk using the cross-platform computer vision
library OpenCV [33]. The disk motions are similar at different
positions in the container, indicating a uniform vibration. The
stage vibrates vertically as A sin (2π f t), where A is the amp-
litude, f is the vibration frequency and t is the time. The vibra-
tion strength is characterised by the maximum acceleration of
the stage:
a = A(2π f )2.
(1)
Two out of the three parameters, A, f and a, are independent.
We vary A and study the disk motions under fixed f or a and
find that A, f and a influence the disk differently. In particu-
lar, two modes of motion are revealed by tuning ( f, a) in the
appropriate range. The two modes of motions are introduced
in section 3.1. The overall motions and its two constituent
modes under different ( f, a) are analysed, including the distri-
bution of displacements per step (section 3.2), energy distribu-
tion (section 3.3), mean squared displacements (section 3.4),
correlations (section 3.5) and power spectra (section 3.6). The
conclusions are summarised and discussed in section 4.
3. Results
The three types of 3D motions are sketched in figure 1. Based
on the raw data of particle’s trajectories (e.g. figures 2(a)
Figure 1. (a) A disk (h = 3.5 mm) is confined in a circular
translucent container with diameter 280 mm and inner wall
separation H = 4.5 mm mounted on the vibration stage. The black
line is printed inside the disk for tracking the rotational motion in
image processing and does not affect the mass distribution. (b) The
disk exhibits three types of vibrations: rolling, fluttering, and lying
flat.
and (b)) and displacements per step (e.g. figures 2(c)–
(e) and 3), we show the static (figures 4–8) and dynamic
(figures 9–14) properties derived from the displacements per
step. They consistently show distinct features for the two
modes of xy motions under different a and f.
3.1. Trajectories
The typical
trajectory is shown in figures 2(a) and (b).
∆x(t) and ∆y(t) are the translational displacements during
1/150 s recognised from the videos with a frame rate of
−1. ∆θ is the rotational displacement during 1/150
150 frame s
section The translational displacements per step ∆x(t) and
∆y(t) in figures 2(c) and (d) show that the motion switches
between active and inactive modes. The disk’s centre of mass
diffuses randomly like Brownian motions during the active
mode, and mainly drifts along one direction during the inact-
ive mode (figure 2(a)). The rotational displacements per step
2
J. Phys.: Condens. Matter 36 (2024) 115102
L Guan et al
Figure 3. Fluctuations of translational and angular displacements
per 1/150 s at f = 60 Hz and (a) a = 3 g, (b) 4 g, (c) 4.5 g and (d)
5 g. g is the free fall acceleration.
types of motions: rolling along the rim, lying flat without wob-
bling and spinning, and fluttering with wobbling, as shown in
figure 1(b) and movie 1. A lying flat disk collides with either
the top or bottom wall. In fluttering, the two ends alternatively
collide with the top and bottom walls, similar to a dimer on
a vibration stage [20]; this state is intermediate between lying
flat and rolling. The frequent collisions in lying flat and flutter-
ing states dissipate the rotational energy. Thus the disk rotates
little, and the disk’s centre moves slowly and steadily without
much direction change, i.e. the inactive mode (figure 2(a)). By
contrast, when the disk rolls along its rim, the static friction
without collision dissipates much less energy, and thus the disk
rolling persistently, resulting in frequent direction changes in
the trajectory (figure 2(a)). When the disk is rolling, the con-
tact point moves along the edge of the disk. The instantaneous
rotational speed of the disk ω is slower than the observed rota-
tional speed of the contact point Ω, which is the same as Euler’s
disk [16].
3.2. Probability distribution function (PDF) of displacements
per step
The PDFs of translational displacement ∆x, P(∆x), and rota-
tional displacement ∆θ, P(∆θ), at different f and a during
∆t = 1/150 s are shown in figure 4. ∆y has similar behaviours
as ∆x, which confirms the isotropy of the vibration stage.
P(∆x) is symmetric around x = 0 and each side can be fitted by
the compressed exponential function f(x) ∝ e
, as shown
in figures 4(a) and (c). Two out of the three fitting parameters
are free because the area under a probability curve must be 1.
The fitted β is insensitive to a at fixed f (figure 4(a) inset) or
to f at fixed a (figure 4(c) inset). The compressed exponential
distribution (i.e. β > 1) is common in granular gases of spheres
[12, 34] and rods [5], but has rarely been observed in the
motion of single particle. P(∆θ) under the fixed f = 60 Hz and
different a or fixed a = 5.0 g and different f all exhibit a peak
−Bxβ
Figure 2. (a) The 200 s trajectory during 150 s < t < 350 s in (c)–(e)
with a time step ∆t = 0.03 s. Active and inactive modes are labelled
by red and blue. (b) The trajectory in (a) coloured by ∆θ during ∆t.
The time evolution of displacements during ∆t (c) ∆x, (d) ∆y and
(e) ∆θ. Red and blue line segments in (e) represent the identified
active and inactive modes, respectively. The typical displacements
of the disk at (A, f, a) = (0.31 mm, 60 Hz, 4.5 g) with a time step
∆t = 1/150 s.
∆θ(t) in figure 2(e) similarly switch between active and inact-
ive modes, and its active mode only occurs when both ∆x(t)
and ∆y(t) are active. The mode switching can be accurately
identified by measuring M = | d2|∆θ|
| on ∆θ(t) with a low-
dt2
pass filter, as shown by the blue and red lines in figure 2(e).
We find that other methods such as the two-threshold Schmitt
trigger, wavelet transform, and the PELT(Pruned Exact Linear
Time) algorithm are less effective than M in the change
point detection of our case (distinguishing the two modes).
The two modes can also be observed in figure 3 under
various a.
By comparing the raw movies about the motions in the hori-
zontal xy plane and on the side view, we find that active and
inactive modes and their transition stage correspond to three
3
J. Phys.: Condens. Matter 36 (2024) 115102
L Guan et al
−B∆xβ
(curves) with the β shown
Figure 4. PDFs of (a) and (c) translational and (b) and (d) rotational
displacements per ∆t = 1/150 s at the fixed (a) and (b) f = 60 Hz
and (c), (d) a = 5.0 g. (a) and (c) PDFs fitted by the compressed
exponential functions P(∆x) = Ae
in the insets. (b) and (d) Each PDF is fitted by a Gaussian function
peaking at ∆θ = 0, and two symmetric Gaussian functions peaking
at ±∆θ0 (curves), respectively. As an example, the dotted curves in
(b) represent the three Gaussian distributions peaking at ∆θ = 0,
±0.084 for a = 4.0 g, which correspond to the inactive rotation,
clockwise and counterclockwise active rotations respectively. The
fitted ∆θ0 is shown in the left inset. The two modes are identified
according to the fluctuation of ∆θ (figure 2(e)). The fraction of
active mode is shown in the right inset. The inverse uncertainty is
used as the weight of each data point in all the fittings, and each date
point is averaged over five trials of experiments for this figure and
figures 4–14.
at ∆θ = 0 and two symmetric subpeaks at ±∆θ0 (figures 4(b)
and (d)).
Curves shift continuously in figures 4(a) and (b) with sim-
ilar ∆θ0 as shown in the left inset of figure 4(b), but change
abruptly at 70 Hz < f < 75 Hz in figures 4(c) and (d) and the
left inset of figure 4(d). The left insets of figures 4(b) and (d)
show that the fitted ∆θ0 is insensitive to a or f, but changes
abruptly at 70 Hz < f < 75 Hz. These can be interpreted by
the right insets of figures 4(b) and (d): the fraction of active
mode changes continuously with a, but abruptly vanishes at
70 Hz < f < 75 Hz by changing f. The PDFs of displacements
per step for the active and inactive modes are plotted separately
in figure 5; These show no significant difference under various
a. P(∆x) are close to Gaussian (solid lines) for active mode
(figure 5(a)) and exponential for inactive mode (figure 5(b)).
P(∆θ) is bimodal for active mode (figure 5(c)) and concen-
trated near 0 for inactive mode (figure 5(d)).
The right insets of figures 4(b) and (d) show that the frac-
tion of active mode increases linearly with a under a fixed f
and abruptly decreases at 70 Hz < f < 75 Hz under a fixed a.
The active mode is nearly absent at f ⩾ 75 Hz, reflecting that
the disk is not activated when the vertical oscillation has a high
Figure 5. (a) and (b) PDFs of translational displacements per
1/150 s for (a) active and (b) inactive modes. (c) and (d) PDFs of
rotational displacements per 1/150 s for (c) active and (d) inactive
modes. f = 60 Hz.
◦
p
f and low a. The disk in the active mode rolls along its rim on
a vibration stage, while its motion is coupled with the vertical
oscillation. In figures 4(b) and (d), the sharp peaks of the PDFs
at ∆θ = 0 correspond to the inactive mode, and the subpeaks at
±∆θ0 correspond to the active mode, as shown in figures 5(c)
and (d). ∆θ0 is almost a constant for curves with different a
and f in figures 4(b) and (d). This finding can be elucidated as
follows. For a rolling disk, the angular speed Ω along the z dir-
ection and ω along the instantaneous rotation axis of the disk,
which is parallel to the disk’s top surface, are connected by
Ω sin α = ω = 2
2g sin α/d, where g is the free-fall accelera-
tion relative to the vibration stage and α is the inclination angle
shown in figure 1(b). α is limited by the container to a max-
. Thus, ∆θ is not sensitive to vibration parameters
imum of 5.8
( f and a). The peak height at ±∆θ0, which is proportional to
the fraction of active mode, increases with a (figure 4(b)) and is
insensitive to f (figure 4(d)). Each section of active mode has a
persistent rotation lasting about 1 s, which is much longer than
the vibration period (see Movie 1). A simulation about a disk
on a vibration stage showed that the persistent rotation of the
disk is accompanied with the slipping at the contact point [35].
Figures 4(c) and (d) under the fixed a = 5.0 g and different f
also show two modes of motion similar to figures 4(a) and (b).
Under low f, the disk motion is very active and insensitive
to f. When f ⩾ 75 Hz under a = 5.0 g, both translational and
rotational motions suddenly weaken. Such two modes have
not observed in [32] under relatively strong vibrations with a
fixed A.
3.3. Energy distribution
The a, f and A all affect the translational kinetic energy Et =
2 mv2 and rotational kinetic energy Er = 1
1
2 Iω2. We measure
4
J. Phys.: Condens. Matter 36 (2024) 115102
L Guan et al
′
−B
Figure 6. (a) PDFs of the translational energies at f = 60 Hz and
−1/2
different a, fitted by p(Et) ∝ E
t
−Bxβ
corresponding rotational kinetic energies fitted by f(x) ∝ e
(curves) with β shown in the inset. E = 0 dominated by inactive
motions is excluded in the fittings. (c) The total kinetic energy
distributions. (d) The fraction of translational and rotational
energies at different a.
Et (curves). (b) PDFs of the
e
′
−B
Figure 7. (a) PDFs of the translational energies at a = 5 g and
−1/2
different f, fitted by p(Et) ∝ E
t
corresponding rotational kinetic fitted by f(x) ∝ e
β shown in the inset. E = 0 dominated by inactive motions is
excluded in the fittings. (c) The distributions of the total kinetic
energy. (d) The fraction of translational and rotational energies at
different f.
Et (curves). (b) PDFs of the
−Bxβ
e
(curves) with
′
◦
′
r = 1
2 I
Ω = ∆θ/∆t in z-direction (see the first panel of figure 1(b))
′Ω2,
and calculate rotational energy kinetic energy E
is in the z-direction. The max-
where the moment of inertia I
imum tilting angle α is very small (≈ 5.8
), thus the estim-
′
r is very close (0.1% error) to the total
ated rotational energy E
rotational kinetic energy along the instantaneous axis Er. The
time step ∆t = 1/150 s is shorter than the vibration periods.
Hence the translational speed v = ∆r/∆t can be estimated.
The angular speed around the z-direction Ω = ∆θ/∆t can be
more accurately measured because the disk persistently rotates
along one direction for seconds, a much longer period than ∆t.
As the kinetic energy is much greater than the gravitational
potential energy change of the disk. Therefore the top and bot-
tom walls have similar effects on the disk. It is the Chi-square
distribution which well fits the distributions of Et at various
a and f (figures 6(a) and 7(a)). We derive that the transla-
tional energy distributions follow p(Et) ∝ E
from
the compressed exponential distribution of ∆x (figures 4(a)
and 5(a)) with the fitted β ′ = 1, which agree well with the fit-
ting results of figures 6(a) and 7(a). The tail of corresponding
distributions of Er (figures 6(b) and 7(b)) can be well fitted
by compressed (β > 1) exponential distributions. The results
indicate that translational and rotational motions have extra
low- and high-energy motions compared with the Boltzmann
distribution (β = 1) for thermal equilibrium. The ratio between
the translational and rotational energies (figures 6(d) and 7(d))
deviates from the equipartition theorem of Et : Er = 2 : 1 for
thermal equilibrium. Under fixed a, P(Et,r) and their fitted β
−1/2
t
−B
e
β
t
E
′
′
Figure 8. Et and Er in the active and inactive modes. White dash
lines mark Et : Er = 2 : 1 predicted by the equipartition theorem.
change abruptly at 70 Hz < f < 75 Hz in figures 7(a) and (b),
reflecting the vanishing active mode. Figures 6(d) and 7(d)
show that the translational kinetic energy dominates in the
inactive mode, and the rotational energy dominates in the
active mode. a and f affect the fraction of active motions
(figures 4(b) and (d) right insets), thereby affecting Et,r and
their ratios in figures 6(d) and 7(d). As a or f changes, Er
changes more dramatically than Et, because the rotations are
much stronger than translational motions in the active mode.
The energy distributions in different degrees of freedom for
active, inactive and both modes are separately presented in
figure 8. Er : Et deviates from 1 : 2 predicted by the equipar-
tition theorem for both active and inactive modes. Both the
5
J. Phys.: Condens. Matter 36 (2024) 115102
L Guan et al
Figure 9. (a) Translational and (b) rotational MSDs ∝ tk under
f = 60 Hz and different a. (c) Translational and (d) rotational MSDs
∝ tk under a = 5.0 g and different f.
Figure 10. (a) and (b) Translational and (c) and (d) rotational MSDs
of (a) and (c) active and (b), (d) inactive modes of motions.
f = 60 Hz. The long-time behaviour is not measurable for some
curves in (b) and (d), because inactive motions do not last long.
translational and rotational energies increase with a dramatic-
ally for the active mode but not for the inactive mode.
3.4. Mean squared displacement
The translational diffusion is characterised by the translational
mean squared displacement
MSDr
≡ ⟨∆r2 (t)⟩ = ⟨(r (t + t0) − r (t0))2⟩,
(2)
where r(t) is the centre position of the particle at time t. ⟨
⟩
averages over all initial times t0 in each trajectory and five tri-
∼ tk with
als of experiment under each (a, f ). We find MSDr
different k in different time regimes, i.e. non-Fickian diffu-
sions. Single particles in a uniform space under thermal equi-
librium usually exhibit normal diffusive motion, i.e. Fickian
diffusion, whose MSD is proportional to time. Under non-
equilibrium, a particle can exhibit non-Fickian diffusion even
at long time scales [36]. The anomalous non-Fickian diffusion
widely exists in complex physical and biological systems [37],
such as flows through disordered media [38], levy flights [39],
chaotic flows [40], and human transports [41]. At t < 0.1 s, the
translational motions under different a and f are subdiffus-
ive, i.e. k < 1 (figures 9(a) and (c)), because collisions tend to
flip the direction of the translational motion. At longer times,
they become diffusive, i.e. k = 1 (figures 9(a) and (c)), except
the superdiffusion, i.e. k > 1, at a ⩽ 4.0 g (figure 9(a)) and
f ⩾ 75 Hz (figure 9(c)).
The rotational diffusion is characterised by the angular
mean squared displacement
MSDθ ≡ ⟨∆θ2 (t)⟩ = ⟨(θ (t + t0) − θ (t0))2⟩,
(3)
where θ(t) is the angular position of the disk at time t.
Figures 9(b) and (d) show ballistic (k = 2) rotations at t < 1 s
and diffusive (k = 1) rotations at t > 1 s, except the subdif-
fusion at 3.0 g (figure 9(b)). The ballistic rotation at short
time is in accordance with the persistent rotations along one
direction.
MSDs in active and inactive modes of motions are shown
separately in figure 10. MSDs almost collapse in figures 10(a)
and (c), indicating that they are insensitive to a in the active
mode of the measured parameter range. This is due to the fact
that the spinning speed of the rolling disk is determined by
the inclination angle α, which is a constant under the fixed
wall separation. By contrast, MSDs are sensitive to a under
inactive mode (figures 10(b) and (d)). A higher a represents
a stronger vibration, thus resulting in higher translational and
rotational MSDs. Under small a, the translational motion is
sub-diffusive at t < 0.1 s and superdiffusive at long times, and
the rotation is sub-diffusive at t < 0.1 s and Brownian at long
times. The long-time MSDs for inactive motions are not meas-
urable at large a, because trajectories are frequently interrup-
ted by active modes. The fraction of active mode increases
with a (figure 4(b) right inset), thus the overall MSDs in
figures 9(a) and (b) are close to the active-mode MSDs shown
in figures 10(a) and (c) at high a and to the inactive-mode
MSDs shown in figures 10(b) and (d) at low a.
3.5. Correlation of displacement per step
The diffusion can be further characterised by the time autocor-
relation functions
C∆x (t) = ⟨∆x (t0) · ∆x (t0 + t)⟩/σ2
C∆θ (t) = ⟨∆θ (t0) · ∆θ (t0 + t)⟩/σ2
∆x,
∆θ.
6
J. Phys.: Condens. Matter 36 (2024) 115102
L Guan et al
Figure 11. Time autocorrelation C(t) of (a) and (b) ∆x, (c) and (d)
∆θ and (e), (f) their cross-correlations. (a), (c) and (e) f = 60 Hz.
(b), (d) and (f) a = 5.0 g. ∆t = 1/150 s.
Figure 12. Time auto-correlation C(t) of (a) and (b) ∆x, (c) and (d)
∆θ and (e), (f) their cross-correlations in (a), (c) and (e) active mode
and (b), (d) and (f) inactive mode. f = 60 Hz.
σ is the standard deviation. C∆x(t) (figures 11(a) and (b)) rap-
idly decays to a negative value within 0.05 s, indicating that the
displacement per step tends to change direction via collisions
and lost memory in 0.05 s, in accordance with the subdiffus-
ive motion at t < 0.1 s in figures 9(a) and (c). C∆x(t) oscillates
around 0 with the same vibration frequency as the stage under
various a and f (figures 11(a) and (b)). Thus, the stage vibra-
tion flips the direction of the disk’s velocity in the horizontal
plane via disk-stage collisions. C∆x(t) and C∆θ(t) in active
and inactive modes are separately shown in figures 12(a)–(d).
The same periodicity of C∆x(t) in active and inactive modes in
figures 12(a) and (b) reflects the robust coupling between the
stage vibration and either rolling, fluttering or lying flat motion
of the disk. According to the Nyquist theorem, the measured
C(t) in figures 11(a)–(d) are accurate when f is less than half
of the sampling frequency of the CCD’s frame rate, i.e. 150/2
= 75 fps. As a increases, C∆x(t) decays faster to the periodic
regime (figures 12(a) and (b)) in the active mode; thus, C∆x(t)
with a more active mode at a higher a (figure 4) decays faster
(figure 11(a)). C∆θ(t) decays to zero in about 1 s (figures 11(c)
and (d)), which agrees with the crossover from superdiffus-
ive to diffusive motion at about 1 s in figures 9(b) and (d).
The angular autocorrelation is stronger in the active mode than
inactive mode (figures 12(c) and (d)); thus, the motions with
more active modes (right insets in figures 4(b) and (d)) exhibit
stronger correlations in figures 11(c) and (d). The translational
or rotational autocorrelations under different a almost overlap
in the active mode (figures 12(a) and (c)) but not in the inact-
ive mode (figures 12(b) and (d)), in accordance with MSDs
(figure 10). C∆θ(t) in the active modes (figure 12(c)) decays in
1 s, reflecting the persistent rotation time. C∆θ(t) in the inact-
ive modes exhibits weak and negative correlations, reflecting
the rapid orientation change of the disk.
The coupling between the translational and the rotational
motions is quantified by the cross-correlation function:
C∆x∆θ (t) = ⟨|∆x (t0)| · |∆θ (t0 + t)|⟩/ (σ∆xσ∆θ)
(4)
It is non-zero at a ⩽ 4.0g at f = 60 Hz (figure 11(e)) and
f > 70 Hz at a = 5.0g (figure 11(f)), showing the coupling
between the translational and rotational motions. C∆x∆θ(t) in
active and inactive modes are separately shown in figures 12(e)
and (f). It is non-zero only at weak vibration, i.e. small a in
both active and inactive modes. The translation-rotation cor-
relations are non-zero for a Brownian ellipsoid [42, 43], a
single granular rod [13, 23] or granular gas of spheres [44].
7
J. Phys.: Condens. Matter 36 (2024) 115102
L Guan et al
Figure 13. Power spectral densities of (a) and (c) ∆x and (b), (d)
∆θ at (a) and (b) f = 60 Hz and various a and at (c) and (d)
a = 5.0 g and various f. ∆t = 1/150 s.
Figure 14. (a) and (b) Translational and (c), (d) rotational power
spectrum densities of (a) and (c) active modes and (b), (d) inactive
modes. ∆t = 1/150 s. f = 60 Hz.
Therefore, the translational-rotational coupling is not sensit-
ive to granular density, but it is sensitive the driving strength,
frequency and particle’s shape.
3.6. Power spectrum
The power spectrum or the power spectral density [45],
S∆x (f) =
−2π ift∆x (t) dt
e
2
(cid:12)
(cid:12)
(cid:12)
(cid:12)
ttot
ˆ
(cid:12)
(cid:12)
(cid:12)
(cid:12)
0
/ (2π ttot)
(5)
provides another angle from which to characterise a time
series ∆x(t) in the frequency domain. ∆x2 ∝ Et and ∆θ2 ∝ Er.
Hence S∆x and S∆θ in figure 13 reflect the kinetic energy
−α, the
distributions in the frequency domain. When S(f) ∼ f
exponent α characterises the colour of the noise. ∆x exhib-
its white noise (α = 0) at low frequencies and blue noise
(α = −1) at high frequencies (figures 13(a) and (c)), whereas
∆θ exhibits white noise at low frequencies and Brownian
noise (α = 2) at high frequencies (figures 13(b) and (d)). These
features are robust for different trials of experiments and under
different a and f. The observed exponent (or the colour of the
noise) has a low-frequency cutoff (about the correlation time)
and a high-frequency cutoff (about stage vibration frequency)
because displacements are less interesting white noises bey-
ond this regime. For example, long-time (longer than correla-
tion time) displacements smear out the short-time complicated
motions and thus become a simple random walk as confirmed
by the linear long-time MSDs in figure 9.
All S∆x in figures 13(a) and (c) exhibit a spike at the
stage vibration frequency when f ⩽ 75 Hz. For f = 85 Hz, the
spike is at 65 Hz, i.e. the folded frequency 150 − 85 = 65 Hz
according to the Nyquist theorem. The rotational spectra in
figures 13(b) and (d) similarly exhibit spikes at the same fre-
quencies but with a lower amplitude, suggesting that the rota-
tions are less coupled with the stage vibration. The frequencies
of the spikes in the power spectra are in accordance with the
oscillation frequencies in C(t) shown in figure 11.
The power spectra are further plotted separately for the
active and inactive modes in figure 14. Under active modes,
∆x exhibits white noise (α = 0) at f < 10 Hz and blue noise
(α = −1) at f > 10 Hz (figure 14(a)), whereas ∆θ exhib-
f < 1 Hz and Brownian noise (α = 2)
its white noise at
at f > 1 Hz (figure 14(c)). These are robust under various
a and f. By contrast, under inactive modes, ∆x exhibits
violet noise (α = −2, figure 14(b)), and ∆θ exhibits blue
noise (α = −1, figure 14(d)) at high frequencies. Similar to
the curve collapse in PDFs (figure 5), MSDs (figure 10)
and autocorrelations (figure 12), the power spectra collapse
under the active mode, indicating that it is insensitive to
vibration.
4. Summary and discussion
The disk’s motion on the vibration stage is dominated by fric-
tion and collisions. Thus, it is difficult to model and predict
in theory or simulation. We observe the active and inactive
modes of motions in both translational and rotational degrees
of freedom. These motions in the xy plane can be interpreted
by its three types of motions in 3D: the disk’s lying flat in the
inactive mode, rolling in the active mode and fluttering dur-
ing their transition. Fluttering is the intermediate state between
lying flat and rolling because it wobbles like rolling and does
not spin as lying flat. When lying flat, the disk frequently
collides with the walls, which dissipates the kinetic energy
8
J. Phys.: Condens. Matter 36 (2024) 115102
L Guan et al
ection of translational motion, resulting in diffusive motions.
The rotational displacements per step exhibit a sharp peak at
∆θ = 0 for inactive mode, and two broad peaks at ±∆θ0 for
active mode, corresponding to clockwise and counterclock-
wise active rotations, respectively. In the active rolling motion,
the rotation typically persists along one direction for about
1 s as shown by the decay time of C∆θ(t), resulting in the
ballistic rotation in MSDs. When C∆θ(t) decays to zero at
t > 1 s, i.e. the rolling direction is randomised, the rotational
MSDs become diffusive. The cross-correlations between the
translational and rotational displacements per step become
non-zero under weak vibrations in both active and inactive
modes. Besides the coupling between some translational and
rotational motions, they both slightly couple with the vertical
vibration of the stage as shown by the spikes in the power
spectra of displacements per step. Moreover, the translational
displacements per step exhibit white noise at low frequencies
under both modes, and blue noise for active mode and violet
noise for inactive mode at high frequencies. The rotational
displacements per step exhibit white noise at low frequencies
under both modes, and Brownian noise for active mode and
blue noise for inactive mode at high frequencies. The trans-
lational autocorrelation functions oscillate at the same fre-
quency as the vibration stage. The vibration frequency of the
stage is clearly resolved from the spike in the power spectra
of translational or rotational displacements per step, reflect-
ing the coupling between the disk’s motions in the xy plane
and vertical vibration of the stage. The per-step displacements’
distributions, power spectra and mean-square displacements
at different a are different for inactive mode but collapse for
active mode, because the two-wall confinement gives a fixed
inclination angle of the disk and consequently a roughly fixed
rolling speed at different a and f in the active mode. Both the
translational and rotational kinetic energies deviate from the
Boltzmann distribution at thermal equilibrium. Moreover, the
ratio of translational to rotational kinetic energy in both active
and inactive mode is lower than 2:1, which was predicted by
the equipartition theorem for thermal equilibrium, reflecting
that energy dissipation via collisions are mainly in the transla-
tional degree of freedom.
Our results show that the granular motions’ sensitivity var-
ies for fixing a, f or A. The nonzero cross-correlation between
the translational and rotational motions indicate that meas-
urements of the translational motions of dense particles may
not be enough to fully understand the particles
motions. We
expect that other plate-shaped particles have more complicated
fluttering and rolling motions. By contrast, rod-shape particles
have fluttering but no rolling motions [20, 47]. Spheres should
have spinning but no fluttering motions. The results cast new
light on the motion of individual particles and the collective
motion of driven granular particles.
′
Data availability statement
All data that support the findings of this study are included
within the article (and any supplementary files).
Figure 15. The vibration parameter regimes explored in this
manuscript (circles and squares) and in [32] (triangles). (a, f ) of
each data point in the top panel is converted to A by equation (1) as
shown in the bottom panel. The observed coexistence of active and
inactive modes, and the abrupt changes at 75 Hz in this study are
absent in the large-A regime explored in [32].
and results in inactive motions. In fluttering, the collisions
usually occur alternatively at the two ends of the disk; thus,
they do not strongly change the orientation of the disk. When
the disk rolls along its rim, its energy dissipation via static
friction is much less. These three types of motions have not
been reported in [32]. The disk’s behaviour under each ( f, a)
can be decomposed into the contributions from the active and
inactive modes. The fraction of active mode increases lin-
early with a at f = 60 Hz and abruptly drops to almost zero in
70 Hz < f < 75 Hz by increasing f at a = 5.0 g. Two modes of
motion have rarely been reported in granular materials except
in a spring-block system on a conveyor belt with slip-stick and
continuous-slip states [46].
Particle’s trajectories are similar to Brownian and ballistic
motions for active and inactive modes, respectively, because
the disk frequently changes direction in active mode but barely
in inactive mode. The coexistence of active and inactive modes
of motions in the xy plane (figure 2(e)) are observed in low-
a and low-A regimes, but was not observed in [32] which
focused on the large-A regime (figure 15). The active and inact-
ive modes in different vibration parameter regimes are sum-
marised in the phase diagram of figure 15. The distributions of
the translational displacements per step are close to Gaussian
for active mode and exponential for inactive mode. The expo-
nential distribution indicates excess amount of small displace-
ments compared with thermal systems because each collision
tends to suppress the translational speed and flip its direction,
as shown by the negative correlation C∆x(t) at the time scale
of one collision or one period of stage vibration. Such dir-
ection flipping also causes the translational subdiffusion in
the short time scale. The translational motions at long times
are diffusive for active mode and surperdiffusive for inactive
mode. The rolling in active mode persistently changes the dir-
9
J. Phys.: Condens. Matter 36 (2024) 115102
L Guan et al
Acknowledgments
This study was supported by the Research Grants Council
of Hong Kong (Grant C6016-20G) (Y H) and the National
Natural Science Foundation of China (Grants 11474326 and
U1738120) (M H).
ORCID iDs
Liyang Guan https://orcid.org/0000-0002-1116-4102
Yilong Han https://orcid.org/0000-0002-1439-0121
References
[1] Aranson I S and Tsimring L S 2006 Rev. Mod. Phys.
78 641
[2] Gollub J P and Langer J S 1999 Rev. Mod. Phys. 71 S396
[3] Casas-Vázquez J and Jou D 2003 Rep. Prog. Phys. 66 1937
[4] Kiesgen de Richter S, Hanotin C, Marchal P, Leclerc S,
Demeurie F and Louvet N 2015 Eur. Phys. J. E 38 74
[5] Harth K, Kornek U, Trittel T, Strachauer U, Höme S, Will K
and Stannarius R 2013 Phys. Rev. Lett. 110 144102
[6] Tatsumi S, Murayama Y, Hayakawa H and Sano M 2009 J.
Fluid Mech. 641 521–39
[7] Scholz C and Pöschel T 2017 Phys. Rev. Lett. 118 198003
[8] Chen W and To K 2009 Phys. Rev. E 80 061305
[9] Kawarada A and Hayakawa H 2004 J. Phys. Soc. Japan
73 2037–40
[10] Huthmann M, Aspelmeier T and Zippelius A 1999 Phys. Rev.
E 60 654
[11] Villemot F and Talbot J 2012 Granular Matter 14 91–97
[12] Barrat A and Trizac E 2002 Phys. Rev. E 66 051303
[13] Trittel T, Harth K and Stannarius R 2017 Phys. Rev. E
95 062904
[14] Grasselli Y, Bossis G and Morini R 2015 Eur. Phys. J. E 38 8
[15] Schmick M and Markus M 2008 Phys. Rev. E 78 010302
[16] Moffatt H 2000 Nature 404 833–4
[17] Jalali M A, Sarebangholi M S and Alam M-R 2015 Phys. Rev.
E 92 032913
[18] Bush J W 2015 Annu. Rev. Fluid Mech. 47 269–92
[19] Couder Y and Fort E 2006 Phys. Rev. Lett. 97 154101
[20] Dorbolo S, Volfson D, Tsimring L and Kudrolli A 2005 Phys.
[21] Wang J, Liu C and Ma D 2014 Proc. R. Soc. A 470 20140439
[22] Dorbolo S, Ludewig F and Vandewalle N 2009 New J. Phys.
11 033016
[23] Wright H, Swift M R and King P 2006 Phys. Rev. E 74 061309
[24] Yamada D, Hondou T and Sano M 2003 Phys. Rev. E
67 040301
[25] Kumar N, Ramaswamy S and Sood A 2011 Phys. Rev. Lett.
106 118001
[26] Kudrolli A, Lumay G, Volfson D and Tsimring L S 2008 Phys.
Rev. Lett. 100 058001
[27] Deseigne J, Léonard S, Dauchot O and Chaté H 2012 Soft
Matter 8 5629–39
[28] Tsai J-C, Ye F, Rodriguez J, Gollub J P and Lubensky T 2005
Phys. Rev. Lett. 94 214301
[29] Altshuler E, Pastor J M, Garcimartín A, Zuriguel I and
Maza D 2013 PLoS One 8 e67838
[30] Talbot J, Burdeau A and Viot P 2011 J. Stat. Mech. 2011 03009
[31] Talbot J, Wildman R D and Viot P 2011 Phys. Rev. Lett.
107 138001
[32] Guan L, Tian L, Hou M and Han Y 2021 Sci. Rep. 11 16561
[33] Bradski G 2000 Dr. Dobb’s J. Softw. Tools (available at:
https://github.com/opencv/opencv/wiki/CiteOpenCV)
[34] McNamara S and Luding S 1998 Phys. Rev. E 58 813
[35] Takano H, Takatori S and Ichino T 2020 Nonlinear Dyn.
100 2205–14
[36] He W, Song H, Su Y, Geng L, Ackerson B J, Peng H and
Tong P 2016 Nat. Commun. 7 11701
[37] Schneider W R and Wyss W 1989 J. Math. Phys. 30 134–44
[38] Bouchaud J-P and Georges A 1990 Phys. Rep. 195 127–293
[39] Chechkin A, Metzler R, Klafter J and Gonchar V Y 2008
Anomalous Transport: Foundations and Applications
(Wiley) ch 5, pp 156–89
[40] Solomon T, Weeks E R and Swinney H L 1993 Phys. Rev. Lett.
71 3975
[41] Brockmann D, Hufnagel L and Geisel T 2006 Nature
439 462–5
[42] Han Y, Alsayed A M, Nobili M, Zhang J, Lubensky T C and
Yodh A G 2006 Science 314 626–30
[43] Peng Y, Lai L, Tai Y-S, Zhang K, Xu X and Cheng X 2016
Phys. Rev. Lett. 116 068303
[44] Brilliantov N V, Pöschel T, Kranz W T and Zippelius A 2007
Phys. Rev. Lett. 98 128001
[45] Thomson D 1982 Proc. IEEE 70 1055–96
[46] Sándor B, Járai-Szabó F, Tél T and Néda Z 2013 Phys. Rev. E
87 042920
Rev. Lett. 95 044101
[47] Abate A and Durian D J 2005 Phys. Rev. E 72 031305
10
| null |
10.1103_physrevresearch.4.043059.pdf
| null | null |
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
Class of distorted Landau levels and Hall phases in a two-dimensional electron gas
subject to an inhomogeneous magnetic field
Dominik Sidler
,1,2,* Vasil Rokaj
,1,3,† Michael Ruggenthaler,1,2,‡ and Angel Rubio 1,2,4,5,§
1Max Planck Institute for the Structure and Dynamics of Matter and Center for Free-Electron Laser Science,
Luruper Chaussee 149, 22761 Hamburg, Germany
2The Hamburg Center for Ultrafast Imaging, Luruper Chaussee 149, 22761 Hamburg, Germany
3ITAMP, Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts 02138, USA
4Center for Computational Quantum Physics, Flatiron Institute, 162 5th Avenue, New York, New York 10010, USA
5Nano-Bio Spectroscopy Group, University of the Basque Country (UPV/EHU), 20018 San Sebastián, Spain
(Received 10 March 2022; accepted 23 September 2022; published 26 October 2022)
An analytic closed form solution is derived for the bound states of a two-dimensional electron gas subject
to a static, inhomogeneous (1/r in plane decaying) magnetic field, including the Zeeman interaction. The
solution provides access to many-body properties of a two-dimensional, noninteracting, electron gas in the
thermodynamic limit. Radially distorted Landau levels can be identified as well as magnetic field induced density
and current oscillations close to the magnetic impurity. These radially localized oscillations depend strongly
on the coupling of the spin to the magnetic field, which gives rise to nontrivial spin currents. Moreover, the
Zeeman interaction introduces a unique flat band, i.e., infinitely degenerate energy level in the ground state,
assuming a spin gs-factor of two. Surprisingly, the charge and current densities can be computed analytically
for this fully filled flat band in the thermodynamic limit. Numerical calculations show that the total magnetic
response of the electron gas remains diamagnetic (similar to Landau levels) independent of the Fermi energy.
However, the contribution of certain, infinitely degenerate energy levels may become paramagnetic. Furthermore,
numerical computations of the Hall conductivity reveal asymptotic properties of the electron gas, which are
driven by the anisotropy of the vector potential instead of the magnetic field, i.e., become independent of spin.
Eventually, the distorted Landau levels give rise to negative and positive Hall conductivity phases, with sharp
transitions at specific Fermi energies. Overall, our work merges “impurity” with Landau-level physics, which
provides novel physical insights, not only locally, but also in the asymptotic limit. This paves the way for a large
number of future theoretical as well as experimental investigations, e.g., to include electronic correlations and to
investigate two-dimensional systems such as graphene or transition metal dichalcogenides under the influence
of inhomogeneous magnetic fields.
DOI: 10.1103/PhysRevResearch.4.043059
I. INTRODUCTION
Lev Landau’s analytic solution for the noninteracting elec-
trons subject to a constant magnetic field, known as Landau
levels, has served as a paradigmatic model system in con-
densed matter physics for almost a century [1,2]. Its basic
concepts are the foundation of numerous groundbreaking dis-
coveries. To mention a few, the integer [3] and fractional
[4,5] quantum Hall effect are fundamentally related to the
*[email protected]
†[email protected]
‡[email protected]
§[email protected]
Published by the American Physical Society under the terms of the
Creative Commons Attribution 4.0 International license. Further
distribution of this work must maintain attribution to the author(s)
and the published article’s title, journal citation, and DOI. Open
access publication funded by the Max Planck Society.
emergence of quantized Landau levels for a two-dimensional
electron gas in a homogeneous magnetic field. Further, in the
presence of a periodic potential, the Landau levels develop
mini-gaps and for the energy spectrum a self-similar fractal
pattern emerges, known as the Hofstadter butterfly [6], which
has become experimentally accessible via magnetotransprot
measurements in Moiré materials [7–10]. The study of Landau
levels and topological edge states is still very actively pursued
and is currently even considered in quantum optics and cavity
quantum electrodynamics (QED), where ultra-strong coupling
of the Landau levels to the quantum vacuum fluctuations
and the control of conduction properties have been achieved
experimentally in a cavity [11–15], with several theoreti-
cal studies and proposals accompanying these developments
[16–20].
In parallel to the fundamental investigation of quantum
systems exposed to magentic fields, the study of impurity
models has a long lasting tradition in solid state physics.
Some of the fundamental theoretical concepts go back to
work of Friedel for charge impurity induced oscillations [21],
whereas Anderson localization [22] or Kondo effect [23] may
2643-1564/2022/4(4)/043059(18)
043059-1
Published by the American Physical Society
SIDLER, ROKAJ, RUGGENTHALER, AND RUBIO
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
emerge due to lattice induced magnetic impurities. Quantum
impurity models are basic to nanoscience as representations
of quantum dots and molecular conductors [24,25] or they
have for example been used to understand the adsorption
of atoms onto surfaces [25–27]. For quantitative predictions
of atomic or molecular impurities, powerful computational
methods are nowadays available (e.g., continuous-time Monte
Carlo [25,28]).
In the following, we will introduce a fundamental theoreti-
cal model, which connects the world of magnetic impurities
with the Landau setting in a nonperturbative way. In more
detail, we will derive a simple closed form solution for an
electron subject to a radial symmetric 1/r-decaying magnetic
field including the spin-dependent Zeeman interaction. Our
solution of the Pauli equation will serve as a fundamental
ingredient to study spin-resolved local and asymptotic prop-
erties of a noninteracting 2D electron gas subject to a radially
symmetric defect, which is induced by the externally applied
magnetic field.
The manuscript is structured as follows. In a first step, we
derive the analytic boundstate solution for our inhomogeneous
field setup. In a second step, single-electron properties are
discussed with their implications for the consecutive many-
body solution. Based on those considerations, local (magnetic
field driven) properties (charge, current and magnetization
densities) of the electron gas are investigated analytically as
well as numerically. Eventually, asymptotic (vector potential
driven) Hall conductivities can be infered for different type of
electric field perturbations, based on locally converged numer-
ical data. Finally, a brief outlook of the various implications
of our exact solution is provided for different future research
directions.
II. ANALYTIC SINGLE ELECTRON SOLUTION
As a starting point for our investigation of a noninteracting
2D electron gas subject to a perpendicular, radially symmetric,
1/r in plane decaying, static magnetic field B(r), we rely
on the minimal coupling Hamiltonian operator in Coulomb
gauge including the Zeeman interaction,
ˆH =
N(cid:2)
j=1
1
2m
[ˆp j
− qA(r j )]2 + gsμB
2
σ j · B(r j ).
(1)
The electron mass is indicated by m with negative unit charge
q = −e. We denote the usual canonical position operator of
particle j as r j and the corresponding momentum operator
as p j and the anisotropic external vector potential is denoted
by A(r). The Bohr magneton is indicated by μB = (e ¯h)/(2m)
and for the spin g factor, we assume the nonrelativistic value
gs = 2 throughout this work. The Pauli vector for electron j
is labeled by σ j.
In a next step, we define the external anisotropic vector
potential within cylindrical coordinates,
A(r) := Aφeφ
(2)
such that it assumes a constant value (Aφ = const) throughout
space, with eφ = 1
r (−yex + xey) indicating the unit vector
along φ direction. The corresponding SI units are [Tm] and
it obeys the Coulomb gauge condition ∇ · A = 0. The corre-
sponding inhomogeneous magnetic field is given by
B(r) = ∇ ∧ A = 1
r
∂ (rAφ )
∂r
ez = Aφ
r
ez.
(3)
the following derivation we assume Aφ < 0,
Throughout
which corresponds to an inhomogeneous magnetic field di-
rected in negative z direction. The permeability of the free
space is given by μ0. Notice that from solving the Maxwell
equations in free space our inhomogeneous magnetic field
corresponds to a radial external current density of the form
Jext = ∇∧B
1
r2 eφ. We will comment on other options later
μ0
in Sec. IV B.
= Aφ
μ0
Having made these preliminary definitions, we can rewrite
the electronic Hamiltonian operator in a more convenient form
that eventually provides access to its simple closed form solu-
tion. The corresponding Hamiltonian of a noninteracting 2D
electron gas, coupled to the classical A(r) and B(r) fields, is
given by
ˆH =
(cid:3)
N(cid:2)
j=1
− ¯h2
2m
∇2
j
+ Aφq ¯h
m
(cid:5)(cid:6)
(cid:4)
i
∂
r j∂φ j
−
σz, j
2r j
+ N
q2A2
φ
2m
,
φ
(4)
in the (r, φ) plane. Fortunately, the contribution of the diamag-
netic term EA2 := q2A2
2m remains constant for all N electrons
in radial coordinates, which reduces the complexity of our
problem considerably. We would like to mention however,
that for a quantized field the diamagnetic A2-term does not
contribute just a constant energy per particle, but modifies
drastically the spectrum and excitations of the electron-photon
system [29]. In a next step, we introduce
> 0,
α := Aφq ¯h
m
which allows a more compact notation for the following
derivation. The resulting stationary Pauli equation for a single
(!) electron can be written as
(cid:3)
− ¯h2
2m
(cid:6)
(cid:6) = E (cid:6),
∂ 2
r2∂φ2
∂
r∂φ
ασz
2r
∂
r∂r
∂ 2
∂r2
+ iα
(5)
+
+
+
(cid:5)
(cid:4)
(6)
where the constant EA2 term is neglected for the moment. No-
tice the close resemblance of Eq. (6) to the two-dimensional
hydrogen atom. For this reason, similar solution strategies ap-
ply for our partial differential equation, as we will demonstrate
subsequently.
The angular and spin problem can trivially be solved by
separation of variables as (cid:6)(r, φ, s) = R(r)(cid:7)(φ)χ (s), with
spin function χ and (cid:7) = eilφ with l ∈ Z, s = ± 1
2 , since
[ ˆH , ∂
∂φ j
] = 0. This leaves us with the radial problem
(cid:6)
(cid:4)
(cid:5)
(cid:3)
ˆHl,sR :=
− ¯h2
2m
∂ 2
∂r2
+
∂
r∂r
− l 2
r2
− α l + s
r
R = E R.
(7)
Before continuing the solution of our radial problem, we need
to distinguish two formally different cases. The interaction
with the B-field gives rise to a Coulomb potential-like 1/r
term, which is attractive if l + s > 0 or repulsive if l + s < 0
043059-2
CLASS OF DISTORTED LANDAU LEVELS AND HALL PHASES …
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
for a fixed α > 0. Notice that the third case l + s = 0 cannot
occur for spin-half particles, due to the Zeeman interaction.
Bound states for l + s > 0. Let us now focus on the at-
tractive eigenvalue problem given in Eq. (7) with l + s >
0. Notice that from the positivity of the Laplacian (kinetic
energy) operator ˆT = ˆHl (α = 0) = 1
2 ( ˆHl+s>0 + ˆHl+s<0), we
find the following relation (cid:4) ˆHl+s>0(cid:5) (cid:2) (cid:4) ˆT (cid:5) (cid:2) (cid:4) ˆHl+s<0(cid:5). To
solve the attractive eigenvalue problem we apply the method
of Frobenius and match orders of a series expansion. There-
fore, we define
(cid:7)
ρ :=
8m|E |
¯h2
(cid:7)
r,
m
2 ¯h2|E |
λl,s := α(l + s)
(8)
(9)
> 0,
for which our radial problem assumes a convenient form
[30,31],
(cid:3)
∂ 2
∂ρ2
+
∂
ρ∂ρ
− l 2
ρ2
+
λl,s
ρ
− 1
4
(cid:6)
R(ρ) = 0.
(10)
To reach a simple closed form solution, we introduce the
Ansatz R(ρ) = e−ρ/2 f (ρ) in agreement with the literature
[30]
(cid:3)
(cid:6)
∂ 2
∂ρ2
−
+
∂
∂ρ
∂
ρ∂ρ
− l 2
ρ2
(cid:8)
+
λl,s − 1
2
(cid:9)
1
ρ
Notice that likewise, hydrogen related, quantization rules arise
for two-dimensional magnetic quantum dots [32,33]. How-
ever, in the following, we can construct the respective explicit
closed form eigenfunctions, which will provide fundamental
physical insights not only analytically but also numerically.
Remark. The exact solution of the nonattractive eigen-
value problem (l + s (cid:2) 0) will remain unknown, since the
Frobenius method does not terminate anymore under these
circumstances.
Eigenfunctions. After having identified the energy eigen-
values for n (cid:3) l, l + s > 0, we can next find the correspond-
ing eigenfunctions by expressing f (ρ) = ρl L(ρ) [31]. This
turns Eq. (11) into
ρ d 2L
dρ2
+ wL = 0, w, ν ∈ N0,
+ (ν + 1 − ρ)
dL
dρ
(16)
which can be solved by the associated Laguerre polynomials
Lν
w of degree w and parameter ν [34]. The associated Laguerre
polynomials are given by Rodrigues’ formula [34],
ρ−νeρ
w!
(e−ρρw+ν ).
Lν
w(ρ) =
d w
dxw
(17)
It is straightforward to show that our radial Eq. (10) trans-
forms into Eq. (16) for the discovered energy eigenvalue En,l,s
with
f (ρ) = 0. (11)
Rn,l,s = e− ρ
2 ρl L2l
n−l (ρ).
(18)
− 1
The Ansatz is motivated, since for large ρ our system ap-
proaches [ ∂ 2
4 ]v(ρ), which has the normalizable solution
∂ρ2
v(ρ) = e−ρ/2. If we apply the series representation f (ρ) =
(cid:10)∞
i=0 ciρi and match the different orders in ρ, we find after an
index shift i (cid:7)→ i + 1 with c−1 = 0:
∞(cid:2)
i=−1
ci+1i(i + 1)ρi−1 − ciiρi−1 + ci+1(i + 1)ρi−1
Consequently, the orthonormal eigenfunctions of our full
problem can be written as
(cid:6)n,l,s = 1(cid:13)
Nn,l
eilφe− ρ
2 ρl L2l
n−l (ρ)χ (s),
(19)
which is identical to the 2D Hydrogen atom solution, except
for a different energy scaling in radial coordinates,
ρ(r) = 2qAφ
¯h
2(l + s)
2n + 1
(20)
r,
ciρi−1 − l 2ci+1ρi−1 + λl,sciρi−1
− 1
2
= 0.
This gives rise to the indicial equation:
ci+1[(i + 1)2 − l 2] = ci
(cid:11)
i + 1
2
(12)
(13)
(cid:12)
.
− λl,s
It implies the “series switches on” for ci+1 when (i + 1)2 = l 2,
i.e., i + 1 = l, and it can terminate only if i + 1
= λ. Oth-
2
erwise one would converge to a non-normalizable solution
ρi
since ci+1 → ci
i for large i and f →
i! . Now, introduc-
ing quantum number n := i = λ − 1
2 leads to a simple closed
form solution for the energy eigenvalues
(cid:10)∞
i=0
En,l,s = −
(cid:3)
q2A2
φ
2m
2(l + s)
2n + 1
(cid:6)
2
, n (cid:3) l, l + s > 0.
(14)
Finally, reintroducing the initially neglected diamagnetic
energy shift EA2 leads to the total one-electron energy within
an inhomogeneous, 1/r-decaying magnetic field:
(cid:3)
(cid:4)
(cid:5)
(cid:6)
E tot
n,l,s
=
q2A2
φ
2m
1 −
2
2(l + s)
2n + 1
, n (cid:3) l, l + s > 0. (15)
which was introduced in Eq. (8), The corresponding normal-
ization is explicitly calculated as [34]
(cid:14)
(cid:15)
n−l )2ρdρdφ
L2l
e−ρρ2l
Nn,l =
(cid:14) ∞
2π
0
0
= 2π (n + l )!
(n − l )!
(2n + 1).
(21)
The orthogonality of the eigenfunctions for different eigenval-
ues is easy to demonstrate. The orthogonality of degenerate
eigenstates is a little more involved, yet will be shown subse-
quently. Notice also that the spin quantum number s enters
the energy dependent scaling factor ρ, which introduces a
radial spin dependence of the wave function. Different nor-
malization constants arising for r-dependent eigenfunctions
are given in Appendix A.
Boundary conditions and uniqueness of eigenstates. Fi-
nally we comment on the uniqueness of the eigenfunctions
and its boundary conditions. Since the radial equation is a
second-order differential equation it allows for, in general,
two linearly independent solutions. A unique solution is then
usually either fixed by choosing appropriate boundary condi-
tions or by normalizability. Since the different forms of the
043059-3
SIDLER, ROKAJ, RUGGENTHALER, AND RUBIO
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
radial equations are of Sturm-Liouville type, there are very
general results available that clarify which conditions are ap-
propriate for a self-adjoint Hamiltonian [35,36]. At ρ → ∞,
no boundary condition can be chosen and normalizability
singles out the unique asymptotic form of the two linearly
independent solutions ν±(ρ) = e±ρ/2. At the lower endpoint
ρ = 0, normalizability singles out the unique form if l (cid:9)= 0
in analogy to the three-dimensional hydrogen case [37]. This
becomes apparent from Eq. (16), for which a Sturm-Liouville
classification of the different endpoints based on the value
of l exists [36,38]. Again, in analogy to the usual hydrogen
case, for l = 0 different boundary conditions at ρ = 0 can be
chosen, and we have selected the usual Friedrich’s boundary
conditions [36,38].
Analytic solutions for gs = 0. Notice that our analytic
solution also applies for gs = 0, which is equivalent to no
Stern-Gerlach term (neglected Zeeman interaction), i.e., the
normal Schrödinger equation is solved instead of the Pauli
equation. For this reduced Hamiltonian, we find that the
expressions for the eigenvalues and eigenstates given in
Eqs. (15) and (19) still apply, when setting s = 0 and n (cid:3) l >
0. However, despite this minor modifications, emerging from
the interaction of the spin-1/2 particles with the magnetic
field, local properties can strongly deviate between gs = 2 and
gs = 0, as will be shown in Sec. IV.
III. SINGLE ELECTRON PROPERTIES AND DISTORTED
LANDAU LEVELS
In a next step, we explore single electron properties based
on our analytic solutions in a 1/r-decaying magnetic field.
From the quantized energy given in Eqs. (15), one can imme-
diately derive fundamental spectral properties [see Fig. 1(a)].
(1) Bounded energy domain. For the allowed quantum
numbers, it is straightforward to see that the constant diamag-
netic energy shift EA compensates exactly for the attractive
potential term entering Eq. (7). Hence, we find
0 (cid:2) E tot
n,l,s
(cid:2) EA2 .
(22)
This implies that the diamagnetic energy EA2 = limn→∞ E tot
n,l,s
determines an Aφ-dependent upper bound in our gauge choice,
beyond which the unknown (!) solutions of the nonattractive
eigenvalue problem commence.
n,l,1/2
n,n,1/2
= E tot
(2) Spin-half degeneracy. Each energy level is spin degen-
erate, i.e., E tot
n,l+1,−1/2, except for the ground-state
= 0, which solely consists of spin up electrons,
energy E tot
assuming a magnetic field direction along the negative z axis.
(3) Dense energy spectrum. It is straightforward to show the
dense nature of the energy spectrum given in Eq. (15). Indeed
every element in the interval given by Eq. (22) is a limit point,
i.e., we have eigenvalues arbitrarily close.
(3) Infinite energy degeneracy. Interestingly, in addition to
the dense nature of the energy spectrum, one can also show
that each energy level is infinitely degenerate (as Landau
levels are), since no restrictions apply to the radial space for
our solution, i.e., ρ ∈ [0, ∞). The degeneracy becomes imme-
= 0. However, it
diately evident for the lowest energy E tot
is a general property of each energy eigenvalue as one sees
!= const, whose representative solution for the
by setting E tot
n,l,s
n,n,1/2
spin up case (s = 1/2) is obtained from:
2l + 1
2n + 1
= odd
odd
= const,
with
l = (2l0 + 1)k − 1
,
2
n = (2n0 + 1)k − 1
,
2
(23)
(24)
(25)
and k ∈ {2D + 1, D ∈ N0} (similar solution applies for spin
down s = −1/2 with l (cid:7)→ l + 1). The introduced subscript
0 indicates the smallest allowed quantum numbers for a
fixed energy eigenvalue. The relations to generate degenerate
eigenvalues given in Eqs. (24) and (25) have important con-
sequences for the eigenfunctions defined in Eq. (19). Indeed,
they ensure that every degenerate eigenvalue (with identical
spin) possesses a unique angular quantum number l. This
automatically imposes orthogonality on the corresponding
eigenstates via angular or spin selection rules. The infinite
degeneracy of the energy spectrum is a very important prop-
erty which connects directly our solution to the well-known
Landau levels. Infinite degeneracy shows up also for the
Landau levels, as the energy spectrum is independent of the
momentum quantum number [1,2]. This fact is directly con-
nected to the quantization of the Hall conductance [3]. In
our inhomogeneous case, however, the infinite degeneracy is
much more intricate, than for the Landau levels, as it depends
on the interplay between two quantum numbers, n and l.
(4) Exponential localization. The radial exponentially sup-
pressed localization of the eigenstates given in Eq. (19)
has an interesting interpretation. First, by definition they
correspond to a boundstate solution of our single electron
Hamiltonian operator. Second, they can be considered as an
“intermediate” regime ∼ exp(−r) between the delocalized
free electron gas solution ∼1 and the more localized Gaussian
form ∼ exp(−r2) of the Landau solution in homogeneous
magnetic fields. Consequently, we anticipate that the influence
of electron-electron interaction should overall be less severe
for our inhomogeneous magnetic setup in comparison with the
more localized Landau case and the subsequently developed
noninteracting many-body solution should indeed represent
a physically reasonable model. For our setting, the radial
localization of the eigenstates is visualized by means of radial
uncertainties (cid:13)r in Fig. 1(a). It reveals that the uncertainty in-
creases considerably, particularly for larger energies and radii,
and can spread over dozens of nm, i.e., can be considered
strongly delocalized with respect to typical molecular scales.
Having access to simple closed form solutions of the eigen-
fuctions in Eq. (19) allows the determination of rigorous
dipole selection rules for angular- and spin-quantum number.
By evaluating (cid:4)n, l, s|qr|n(cid:11), l (cid:11), s(cid:11)(cid:5), one finds
(cid:13)l = ±1, (cid:13)s = 0.
(26)
for dipole allowed transitions. Interestingly, numerical cal-
culations of the radial integrals (see Appendix A) show that
approximately even more stringent dipole selection rules ap-
ply:
(cid:13)n ∈ {0, 1}, (cid:13)l = ±1, (cid:13)s = 0,
(27)
043059-4
CLASS OF DISTORTED LANDAU LEVELS AND HALL PHASES …
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
. Small horizontal lines correspond to the radial uncertainties (cid:13)r =
FIG. 1. Radially resolved (cid:4)r(cid:5)n,l,s energy eigenvalues En,l,s of the electrons for a Aφ/r-decaying magnetic field with Aφ = −0.186 μTm.
The black doted horizontal lines indicate the boundaries of allowed energy eigenvalues for our solution. Notice, that higher-lying energies
exist for l + s (cid:2) 0, but the solutions remain unknown. In (a), B-field aligned spin configuration are marked with
and antialigned electrons
(cid:4)r2(cid:5) − (cid:4)r(cid:5)2. Overall, it is clearly shown that the presence of
with
the inhomogeneous magnetic field increases the radial electron localization close to the origin in combination with a coarse graining of the
discrete energy levels. Further away from the origin the denseness of the energy spectrum becomes apparent. The lowest lying (E = 0),
infinitely degenerate, electronic states can only emerge if the Zeeman interaction between the magnetic field and the spin is considered. In (b),
the dipole allowed transitions are indicated by green lines for (cid:13)n = 0 and purple lines for (cid:13)n = 1. For illustrative purpose, we restrict our
states). The first few distorted Landau levels (purple) are labeled by
visualization to the spin-
ν (cid:3) 0 at their respective highest energies. Notice the strong curvature for ν (cid:3) 1. Similarly, distorted Landau levels could also be identified for
spin-down electrons starting at ν = 1.
states (equivalent results hold for the spin-
(cid:13)
which leads in for the allowed quantum numbers n (cid:3) l >
0, l + s > 0 to effectively only two relevant transition chan-
nels for single electron states n(cid:11), l (cid:11), s(cid:11), as visualized in
Fig. 1(b) for s = 1/2. Indeed, the dipole allowed transition
pattern visualized in Fig. 1(b) suggests the definition of dis-
torted Landau levels ν in our setting:
ν := n − l,
(cid:16)
ν ∈ N0
ν ∈ N
if s = 1/2
if s = −1/2
,
(28)
which are illustrated by the purple lines for s = 1/2. The
corresponding interlevel transitions (green) obey (cid:13)ν = ±1,
transitions (purple) obey
whereas the allowed intralevel
(cid:13)ν = 0. Notice that the lowest lying Level ν = 0 remains
flat, as it is the case for Landau levels, provided that the
Zeeman interaction is considered with gs = 2. For ν > 0,
the levels become strongly distorted with respect to radius
(cid:4)r(cid:5). For relatively low energies (and large radii), they remain
relatively flat, but having a negative curvature dE /d(cid:4)r(cid:5) (cid:2) 0,
which becomes increasingly negative for intermediate ener-
gies, eventually culminating in a flat positive curvature for
E (cid:2) EA2 . The emergence of such strongly distorted Landau
levels has important consequences for the Hall conductivity
as will be seen in Sec. IV B 2. Notice that lifted degeneracies
of the Landau levels have previously been observed in special
043059-5
SIDLER, ROKAJ, RUGGENTHALER, AND RUBIO
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
Landau settings with locally constant (but inhomogeneous)
magnetic fields [39] or under the influence of additional elec-
tric fields [40].
In the last step, the divergence free magnetization current
density was expressed as the curl of the magnetization density
ms. It can be written in a particularly simple form for the
noninteracting electrons of our system,
IV. 2D ELECTRON GAS IN INHOMOGENEOUS
MAGNETIC FIELD
After having discussed the single electron solution for our
inhomogeneous magnetic field setting, we continue by inves-
tigating fundamental many-body properties for noninteracting
electrons. In other words, we leave the “atomistic” single
electron perspective and focus on Fermi energy EF depen-
dent “solid state” characteristics instead. Surprisingly, we will
find that the many-body problem cannot only be evaluated
numerically, but there is even a simple closed form solution
accessible in the thermodynamic limit (N → ∞) at EF =
0+. This allows unique physical insights complementary to
numerical calculations. Overall, we will focus on spatially
sensitive properties (e.g., local densities), which are strongly
affected by the 1/r-decaying B field at the origin, as well
as asymptotic observables (e.g., Hall conductivity) that are
dominated by the influence of the constant vector potential Aφ
instead. Specifically the spatially-dependent properties [see
also Figs 1(a) and 1(b)] will highlight a strong departure
from the usual condensed-matter perspective and show how
extended systems are connected to the more local atomic and
molecular physics.
A. Analytic solution for the distorted Landau level at ν = 0
We continue with the derivation of a simple closed form
solution for the charge, current and magnetization densities of
the fully filled lowest level ν = 0 (EF = 0+). For this purpose,
we introduce the charge density,
as well as the physical charge current density in the Coulomb
gauge [41],
j(r) = jorb(r) + js(r),
(30)
which is decomposed into orbital jorb = jpara + jdia current
contributions, arising from the paramagnetic jpara and diamag-
netic jdia terms, and the spin-dependent magnetization current
density js(r) due to the Stern-Gerlach term. Notice that we
chose to explicitly account for the negative electronic charges
in our density definitions, e.g., the more negative the charge
density becomes, the more electrons accumulate locally. Sim-
ilarly, the different charge current density observables can
explicitly defined as
ms(r) = ¯hq
m
N(cid:2)
i=1
siδ(r − ri )ez.
(34)
Notice that the origin of the magnetization current ˆj
is purely
quantum mechanical, since it
is spin-dependent, whereas
orbital currents can also emerge in a classical setting. Further-
more, the magnetization current can only play a significant
role for inhomogeneous spin magnetizations mz(r), which are
not present in the ubiquitous Landau setting.
s
In a next step, we evaluate the density expressions given
in Eqs. (29)–(34) for the fully occupied, infinitely degenerate,
lowest level at En,n,1/2 = 0. Surprisingly, the resulting infinite
series converges to the following thermodynamic limit solu-
tion (N → ∞) for the charge density in radial coordinates:
n0(r, φ) = q
∞(cid:2)
n=0
ψ ∗
n,n, 1
2
ψ
n,n, 1
2
= q2Aφ
¯hπ
e− 2qAφ
¯h
r sinh
r
(cid:15)
2qAφ
¯h r
(cid:17)
,
(35)
where we applied the eigenstates explicitly given in Eq. (A4)
of Appendix A and used the series expansion of sinh(x)/x =
(cid:10)∞
n=0 x2n/(2n + 1)!. Notice, we can also find the correspond-
ing simple closed form solution for ˜n0(k) in reciprocal k space
(see Appendix C), which diverges (!) for k → 0. Now, we
utilize Eq. (35) to derive simple closed form solutions for the
many-body current densities defined in Eqs. (31)–(33) in a
similar fashion,
∞(cid:2)
r
¯h
n=0
e− 2qAφ
r
(cid:8)
=
q3A2
φ
π ¯hm
(cid:4)
×
cosh
2qAφ
¯h
(cid:9)
−
r
(cid:15)
2qAφ
¯h r
sinh
2qAφ
¯h r
(cid:17)
(cid:5)
eφ,
(36)
n0(r, φ)eφ,
n0(r, φ)ez,
0 (r, φ) = − qAφ
jdia
m
0(r, φ) = ¯h
ms
2m
0(r, φ) = − d
js
dr
z(r) = − jpara
ms
0
(37)
(38)
(39)
(r, φ) − jdia
0 (r, φ).
ˆn(r) = q
N(cid:2)
i=1
δ(r − ri )
(29)
jpara
0
(r, φ) = ¯hq
2π mr
nψ ∗
n,n, 1
2
ψ
n,n, 1
2
N(cid:2)
jpara(r) := ¯hq
2mi
jdia(r) := − q2
mc
i=1
(δ(r − ri )
−→
∇ i −
←−
∇ iδ(r − ri )),
(31)
N(cid:2)
i=1
δ(r − ri )A(r),
(32)
js(r) := ¯hq
2m
N(cid:2)
i=1
−→
∇ × σiδ(r − ri ) = ∇ × ms(r).
(33)
The last relation between orbital and magnetization current
densities derived in Eq. (39) arises, when comparing the
explicit results of jpara
0. This is a truly aston-
ishing result, since this means that the total current density of
the fully occupied lowest band vanishes exactly on the entire
domain (!) of the infinitely extended 2D electron sheet (see
Figs. 2 and 4)
0 with js
and jdia
0
0 (r, φ) := jpara
jtot
0
+ jdia
0
+ js
0
= 0.
(40)
043059-6
CLASS OF DISTORTED LANDAU LEVELS AND HALL PHASES …
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
relativistic quantum fluctuations, which would introduce
small corrections to gs = 2 [42]. The observed subtle can-
cellation effect between orbital and magnetization currents
can only emerge if the Zeeman interaction is included in our
many-electron problem, whereas, for example, for gs = 0 the
lowest flat level μ = 0 must not (!) exist, i.e., one would
observe a similar zero total current density at EF = O+, but
originating from the zero occupancy at finite radii r instead
(absence of charges). For this reason, the total charge densities
strongly deviate between the gs = 0 and the gs = 2 solution,
where only the later one shows a pronounced aggregation at
the origin. This fundamental difference is rather surprising,
give the close resemblance of the respective single electron
solutions derived in Sec. II. Consequently, a quantum effect
(Zeeman interaction) fundamentally alters the (local) proper-
ties of our system.
Notice that on a first sight our lowest flat level ν = 0
closely resembles the Landau solution, for a homogeneous
magnetic field (see Appendix D) applied to a noninteracting
electron gas, which also predicts flat Landau levels with zero
orbital and zero magnetization currents. However, as already
stated, in our case, we find a highly inhomogeneous charge
density distribution instead and the zero total magnetization
is only reached thanks to opposing magnetization and orbital
currents.
B. General many-body solution in the noninteracting
limit for EF (cid:3) 0
In a next step, we investigate the many-electron prob-
lem for EF > 0 numerically, in the vicinity of the magnetic
field impurity at r = 0. Again, we assume fully filled bands
throughout the calculations. Fortunately, the numerical re-
sults reveal that we reach locally converged noninteracting
many-body solutions around the origin, with only a limited
amount of eigenstates. This convergence in real space is rather
surprising, since we deal with an infinite amount of electrons,
where infinite many energy level are infinitely degenerate. A
problem that in principle cannot be represented on a computer.
However, thanks to the exponential localization of the states,
simulations show that we can indeed reach numerically con-
verged real-space, many-body solutions in the vicinity of the
magnetic impurity.
Before continuing our computational analysis, we com-
ment on the delicate choice of a reasonable parameter range
for our investigations, which will hopefully become experi-
mentally accessible in the near future. For this purpose, we
try to minimize the magnetic flux through the 2D electron
gas sample by choosing a small circular vector potential Aφ.
However, on the one hand, this comes at the cost of reducing
the allowed energy domain given in Eq. (22) and thermal noise
may become an issue. On the other hand, reducing Aφ will also
reduce the real space density of states for each energy level,
as we can immediately infer from the scaling of the radial
expectation value of a single electron
(cid:4)ˆr(cid:5)nn1/2 = ¯h
qAφ
at EF = 0. Notice that the derivation of Eq. (41) is straight-
forward, since the associate Laguerre polynomial contribute
(n + 1),
(41)
FIG. 2. From top to bottom: radially resolved charge, current
and magnetization densities for Aφ = −0.186 μTm. Notice that we
explicitly account for the negative electronic charges, i.e., the more
negative the charge density becomes, the more electrons accumulate
locally. Thin continuous lines correspond to the analytic solution of
the fully filled, infinitely degenerate lowest band at EF = 0+. Solid
lines, made of discrete triangles, correspond to the numerical solution
for the fully filled flat band at EF = E2,1,1/2 = 1.96 meV. The later
case nicely exemplifies the B-field induced Friedel oscillations for
the charge density around the origin. They are accompanied by
para- jpara and diamagnetic jdia, as well as magnetization current
js oscillations, which result in a total current jtot oscillating around
the origin. Interestingly, the corresponding total magnetization of
this band indicates a mostly paramagnetic response to the applied
magnetic field, whereas the lowest band does not respond at all, i.e.,
= 0, thanks to the exact cancellation of the orbital and
jtot
0
spin contributions for gs = 2.
= mtot
0
This automatically indicates a zero total magnetic response
mtot = 0, where we used the general definition for the
magnetization density: j := ∇ × m and ∇ · m = 0 with nor-
malizability condition. One would typically only expect such
a vanishing magnetization density (found for EF = 0+) in
free space, but not in the presence of the induced, strongly
inhomogeneous charge distribution, as given by Eq. (35). No-
tice that tiny deviations of the exact cancellation arise from
043059-7
SIDLER, ROKAJ, RUGGENTHALER, AND RUBIO
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
FIG. 3. On the left: Radial density distributions reveal the quantization of the energy continuum around the origin, due to the applied
inhomogeneous magnetic field with Aφ = −0.186 μTm. On the right: The energy and density structure of the emergent levels is visualized
based on the number of electrons Nmax contained within each level up to a radial distance of rmax = 50 nm. The horizontal yellow lines
indicate the standard Landau-levels for a setup with identical magnetic flux, measured through the circular surface area limited by rmax.
Dashed horizontal lines identify the most (red) and second most (black) prominent levels, which are given by ν = 1 and ν = 2. Notice that
the observed DOS discretization seems to be a local effect, which vanishes for larger radii rmax (cid:15) 50 nm, where the discrete energy spectrum
becomes denser (i.e., almost continuous).
= 1 to the involved eigenstates for n = l. Like-
trivially L2l
n−l
wise scaling results hold for the density of states in higher
lying bands, as we can infer from our numerical calculations.
Consequently, with small Aφ a 2D electron gas with very
low charge density may be required to investigate lower ly-
ing energy bands. Therefore, one has to ensure that Wigner
crystallization, i.e., a phase where the Coulomb interaction
between the electrons dominates, does not hamper the results
[43]. Having made all these preliminary consideration, we
suggest Aφ = −0.186 μTm as a reasonable choice, which will
be used throughout our work. It ensures that the localized
states in the lowest energy band at EF = 0+ could in prin-
ciple be populated solely, for a 2D material (e.g., transition
metaldichalcogenide monolayers [43]) with an extremely low
electron density of n2D = 1011 cm2 and an effective mass
m∗ ≈ 1, when considering a radial area defined by rmax = 50
nm. Such dilute electron gas have been realized experimen-
tally, for which Wigner crystallization could be avoided at
temperatures above TW ≈ 11 ± 1 K (measured in absence of
magnetic fields) [43]. Clearly, for higher electron densities,
which implies EF > 0 in our setting, the Wigner crystal-
lization issue becomes less severe and lower temperatures
could be reached to suppress thermal noise. Nevertheless, our
selected Aφ value ensures that the energetic regime of the
derived bound state solution is wider than the thermal noise,
i.e., on the order of a EA2 ≈ 3 meV ∼ TA (cid:2) 35 K, which
should in principle allow measurements down to EF = 0+.
Eventually, our considerations to minimize the magnetic field
strength suggests the preparation of a (state of the art) dilute
2D electron gas within a temperature regime TW (cid:2) T (cid:2) TA.
Notice that the homogeneous Bhom-field equivalent, which
generates the same magnetic flux through a circular surface
with 100 nm diameter as provided by our inhomogeneous
setting, is given by Bhom ≈ 7.5 T. A value that can routinely
be achieved for homogeneous Landau settings [43]. However,
the experimental realization of a constant 1/r-decaying field
shape, will require considerable experimental effort. Potential
setups may facilitate magnetic lensing with metamaterials
or shaping the fields with (pumped) cavities, which offer a
versatile approach to tailor electromagnetic-fields down to the
nanoscale [44].
1. Magnetic field driven charge and current density oscillations
After having determined a reasonable field strength, we
continue by investigating the radially resolved charge density
with respect to the energy around the origin (see Fig. 2). This
analysis reveals a remarkable feature of our system. It depicts
that a discrete, flat band-like, density structure emerges close
to the origin. At a first sight, this appears to contradict our
earlier definition of the distorted Landau levels. However,
the here observed energetic quantization of the density of
state (DOS) has two fundamental limitations. First, it remains
restricted to the vicinity of the origin, whereas for larger radii
the charge density becomes more and more continuous with
respect to E . Second, the usual angular transition dipole selec-
tion rules (cid:13)l = ±1 effectively prevent significant interband
transition between neighboring DOS levels (e.g., see linear re-
sponse Hall conductivity in Sec. IV B 2). Hence, the intriguing
energy band structure in Fig. 3(b) will determine the physics
of our setup most likely only for very specific observables and
perturbations, but not generally as it was the case for Landau
levels.
Nevertheless, the observed discrete DOS pattern has some
interesting properties that we would like to mention and com-
pare with the ubiquitous Landau levels. For example, the usual
043059-8
CLASS OF DISTORTED LANDAU LEVELS AND HALL PHASES …
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
FIG. 4. Angularly resolved density distribution for Aφ = −0.186 μTm. The color-coding of each row is fixed to ensure horizontal
comparability, whereas the displayed color-bars extend over the entire value range, which accounts for the substantial inhomogeneity at
the origin. Left column shows the analytic solution of the lowest band, orbital jorb and magnetization js currents cancel exactly and lead to a
zero magnetic response mtot = 0 to the applied B field. The right column corresponds to the total densities of the system with occupied bands
up to EF = E2,1,1/2. It demonstrates that the magnetic induced oscillatory behavior persists for the entire system and is not only restricted to
specific bands (e.g., at EF = E2,1,1/2 as displayed in the middle column). However, we find that the joint magnetic response of all filled levels
always remains diamagnetic (e.g., mtot (cid:3) 0 ∀r at the bottom of the right column), whereas selected degenerated energy levels may respond
paramagnetically (see bottom figure in the middle column).
equidistant energy spacing is broken for our discrete DOS
plateaus [see Fig. 3(b)]. Furthermore, in our case the DOS
does not only depend nontrivially on the applied vector poten-
tial strength Aφ, but also on the energy E and the considered
integrated surface around the origin (rmax). For this reason,
we do not have access to the scaling of the Fermi-energy
with respect to the externally applied field throughout this
work. As mentioned earlier, the DOS discretization pattern
vanishes for large rmax due to the dense nature of the spectrum.
This indicates that the physical origin of the discretization
043059-9
SIDLER, ROKAJ, RUGGENTHALER, AND RUBIO
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
is most likely related to the decaying magnetic field and
not a direct consequence of the constant vector potential Aφ,
which determines the asymptotic properties of our setting. In
more detail, we find that the most dominant density plateaus
around the origin can be attributed to the quantum numbers
2 , which introduces a n−1 decaying pattern
n, l = n − 1, s = 1
in energy spacing, i.e.,
E tot
n,n−1, 1
2
=
4q2A2
φ
m
·
n
(1 + 2n)2
, n (cid:3) 1
(42)
marked by the red dashed lines in Fig. 3(b). Notice, this pat-
tern exactly corresponds to the allowed intralevel transitions
of our distorted Landau level at ν = 1, which contains the
radially most localized electrons with E > 0. Less dominant
patterns are found with decreasing order for ν > 1 [see black
dashed lines for ν = 2 in Fig. 3(b)]. Another interesting aspect
of the inhomogeneous field arises if one compares the density
of states for our plateaus with the Landau solution assuming
an equal magnetic flux through the surface defined by rmax
[yellow lines in Fig. 3(b)]. One immediately notices that the
Landau levels have an increased DOS and the energetic spac-
ing is considerably larger. Another major difference is that the
number of electrons per Landau level scales quadratically with
rmax, whereas in our case, the scaling is nontrivial except for
the lowest band for which Eq. (41) suggests a linear (!) scaling
in nmax ∝ (cid:4)ˆr(cid:5)nmaxnmax1/2 = rmax.
After having discussed this local DOS discretization, we
focus next on the charge and current-density observables,
around the origin, which are computationally accessible for
a finite number of one-electron states.
In Fig. 2, charge, current, and magnetization density pro-
files are displayed with respect to the radial distance r from
the origin. Thin lines correspond to the previously derived
analytic results for the fully filled band (energy levels) at EF =
0, whereas bold triangles indicate a numerical solution for a
prototypical, fully filled, infinitely degenerated energy level
at EF = E2,1,1/2. One immediately notices that the B-field
inhomogeneity introduces a magnetic defect in the electron
gas, resulting in an accumulation of negative charge around
the origin. When filling higher lying bands (EF > 0), radial
charge density fluctuations occur, which contribute addition-
ally to the negative charge accumulation, introduced by the
lowest-lying flat band (top panel of Fig. 2) given in Eq. (35).
The emergent charge density fluctuations in Fig. 2 resemble
Friedel oscillations, which typically emerge in the vicinity
of charge impurities. While, the oscillatory behavior around
the origin can only be determined numerically for our setup
with EF > 0 assuming fully filled bands, an analytic statement
can be made at the origin, i.e., for r = 0. In more detail, we
notice from Eq. (19) that only states (cid:6)n,0,1/2 of zero angular
momentum (l = 0) are nonvanishing at the origin. Notice that
this implies the corresponding state (cid:6)ν,0,1/2 possesses the
highest total energy within every distorted Landau level ν =
n, s = 1/2. Consequently, the density at the origin counts the
number of fully filled distorted Landau levels. Furthermore,
the charge density at the origin is effectively determined by
only one electron for Fermi energies 0+ (cid:2) EF < E tot
=
8/9EA2 , which already covers a significant part of the bound
state spectrum. The charge density at the origin for a filling up
1,0,1/2
to E ν
F
= E tot
ν,0,1/2 is given by
(cid:17)
(cid:15)
0, E ν
F
n
:= q
(cid:14)
ν(cid:2)
2π
(cid:6)∗
n,0,1/2(0, φ)(cid:6)n,0,1/2(0, φ)dφ (43)
0
n=0
4q3A2
φ
¯h2
=
ν(cid:2)
n=0
1
(2n + 1)3
,
(44)
based on Eq. (A1) with the respective normalization in
Eq. (A2) given in Appendix A. Two particularly interesting
(limiting) cases are
4q3A2
φ
¯h2
,
(cid:17)
(cid:15)
n
=
0, E 0
F
(cid:15)
0, E 0
F
ζ (3)n
(cid:17)
≈ 1.0518 n
(cid:17)
(cid:15)
0, E ∞
F
n
= 7
8
(45)
(46)
(cid:15)
0, E 0
F
(cid:17)
,
where ζ indicates the Riemann Zeta function. Consequently,
even if all bound states are occupied, i.e., infinite electrons can
contribute to the density at the origin, it is only modified by
about 5% compared with the single electronic state occupation
from n = 0, l = 0, and s = 1/2.
In addition to the charge density properties of our system,
we also observe nonvanishing, circular charge currents (in
φ direction) for EF > 0, which show oscillatory behavior in
radial direction (middle panel in Fig. 2) with correspond-
ing magnetization fluctuations in the bottom panel of Fig. 2
( j(r) = ∇ × m, with ∇ · m = 0). Interestingly, the radially
resolved total current density jtot seems to oscillate around
zero, i.e., they can change direction. This is a true quantum
effect, which could not emerge for identical classical charges,
subject to the inhomogeneous magnetic field along z. In more
detail, one can show that every single electron current expec-
tation value is positive (cid:4) jtot(cid:5)n,l,s > 0 for our setting, i.e., the
diamagnetic term dominates [see Eq. (B4) in Appendix B].
This automatically implies that the total magnetic response of
our system will be diamagnetic, which agrees with the Landau
case [1]. However, in the vicinity of the magnetic impurity
things can change fundamentally. While overall the charge
and current density fluctuations remain qualitatively similar,
i.e., independent of considering all bands up to the Fermi level
(right column in Fig. 4) or only the highest occupied band
(middle column in Fig. 4), things change, when investigating
the magnetization density. In that case, one observes that
the magnetic response can become paramagnetic for certain
bands (see bottom row in Fig. 4), while the total magnetic
response always remains diamagnetic. A crucial ingredient for
this effect is the proper consideration of the Zeeman interac-
tion as well as spin-dependent magnetization currents, in order
to achieve the paramagnetic response of certain (degenerate)
energy bands.
To summarize our many-body results up to this point, we
would like to mention that overall the fundamental driving
mechanism investigated so far is mostly local, and is mainly
be related to the strong magnetic field inhomogeneity, which
decays as 1/r. In other words, the constant circular vector
potential Aφ plays only a minor role for the observed local
density aggregation and fluctuations or for the emergence
of the DOS plateaus. However, things change for different
observable, as we will see next.
043059-10
CLASS OF DISTORTED LANDAU LEVELS AND HALL PHASES …
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
FIG. 5. Sketched Hall conductivity and induced currents for two different perturbation with static electric fields: (a) Angularly averaged
Hall conductivity tensor σxy(rx ) at radius rx for a static, homogeneous electric field perturbation Ey = E ey with an inhomogeneous magnetic
field Bz(r) directed along the negative z-axis. Longitudinal electronic current densities induced by Ey are indicated by jy(rx, φ). Transversal
(rx, φ), which are related to the Ey-perturbation by the local transversal Hall conductivity
electronic Hall current densities are given by jHall
σxy(rx, φ). The background color indicates the sign and magnitude of the total current densities jtot (r), induced by the inhomogeneous magnetic
field, as shown in Fig. 4. (b) Similar setup, where the electric field perturbation is given by Ey(rx ) = δ(r − rx )E ey instead, which induces
different Hall currents.
x
2. Vector potential driven sign flip in Hall conductance
In a next step, we investigate the radially resolved rx Hall
conductivity tensor σxy for a static, homogeneous electric field
perturbation in y direction [see Fig. 5(a)]. At zero temperature,
the linear response Hall conductivity assumes the following
simple form [45,46]:
σxy(rx, EF ) = ie2
¯h
(cid:2)
Ea<EF <Eb
1
(Ea − Eb)2
× [(cid:4)a|δ(rx − r)ˆvx|b(cid:5)(cid:4)b|ˆvy|a(cid:5)
− (cid:4)a|ˆvy|b(cid:5)(cid:4)b|δ(rx − r)ˆvx|a(cid:5)],
(47)
(cid:6)
(cid:3)
The velocity operator ˆv = ( ˆp − qA)/m can be written as
(cid:6)(cid:5)
(cid:3)
cos φ∂r − sin φ
sin φ∂r + cos φ
(cid:3)
− qA
− sin φ
cos φ
= 1
m
− i ¯h
∂φ
∂φ
ˆvx
ˆvy
(cid:4)
(cid:6)
r
r
,
(48)
where the Cartesian components x, y are expressed in radial
coordinates.
The radially resolved Hall conductivity in Eq. (47) has
two major advantages compared with the standard integrated
quantity. First, it reveals the rich local conductivity variations,
which we anticipate due to the observed charge and current
density oscillations around the origin. Second, thanks to the
exponential localization of the single electron eigenstates, we
can indeed determine locally converged Hall conductivities at
the magnetic impurity for infinite system sizes, which can be
utilized to infer asymptotic properties of our system. Notice
that the involved angular and spin integrals are solved for the
transition velocity elements, which give rise to the same angu-
lar and spin selection rules as previously seen in Eq. (26) for
the transition dipole moments. This reduces the computational
demand of the summation over occupied |a(cid:5) and unoccupied
states |b(cid:5) considerably, since the only nonvanishing contribu-
tions arise from la = lb ± 1 (see Appendix A for more details
on the numerics).
The locally converged Hall conductivity σxy(rx, EF ) is dis-
played in Fig. 6(a) with respect to rx (angularly integrated
in φx) up to rmax = 50 nm for the static, homogeneous elec-
tric field perturbation along y. One immediately notices that
the magnetic inhomogeneity leads to a depletion of the con-
ductivity close to the origin (rx < 10 nm), i.e., a whitening
of the color pattern, which is caused by the relatively low
electron density of the excited states in combination with
large (cid:13)E = Eb − Ea for the allowed transitions. However,
our locally converged Hall conductivity suggests a remarkable
asymptotic feature for our setting: the emergence of quantum-
Hall phases where the sign of the conductance fluctuates. In
more detail, we find that the local Hall conductivity is negative
(blue) for EF < Eλ, at λ = 1/4 with Eλ given in Eq. (50),
and positive for higher lying Fermi energies (red). This sign
change appears even more pronounced for the integrated con-
(cid:18)
σxy(rx, EF )rxdrx shown in
ductivity pattern σ rmax
Fig. 6(b). It is important to mention that this phenomenon
of the sign change in the Hall conductance also shows up
experimentally for the Hofstadter butterfly in Moiré materials
[7,10], for which the homogeneity of the system is broken by
the lattice periodicity, whereas in our case we rely on a B-field
inhomogeneity.
(EF ) :=
rmax
0
xy
To reach our numerical results, it turns out that the accurate
consideration of a relatively large number of single electron
states is vital to reach converged results and in particular no
reduction of the allowed (cid:13)n transitions can be applied to
speedup the calculations (caused by the 1/(cid:13)E 2-dependency
of σxy). The convergences becomes particularly tricky for
larger Fermi energies, due to previously discussed increase
043059-11
SIDLER, ROKAJ, RUGGENTHALER, AND RUBIO
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
FIG. 6. (a) Radially resolved Hall conductivity σxy for a homogeneous electric field perturbation along the y direction. To reach a locally
converged solution within rmax = 50 nm, a large number of single electron eigenstates with (cid:4)r(cid:5) (cid:15) rmax is required, which indicates that
the locally observed Hall conductivity switch is driven by the constant anisotropic vector potential instead of the localized inhomogeneous
magnetic field. (b) Consequently, the integrated Hall conductivity transition at EF = EA2 (1 − 1/4) is expected to persist in the asymptotic limit
rmax → ∞, which automatically implies that spin contributions are of minor importance for this observable, except for EF = 0+.
in the delocalization of states, i.e., the sharp conductivity
drop (below zero) observed around EF = 3 meV is likely to
be a numerical artifact. The relevance of a large amount of
states with (cid:4)r(cid:5) (cid:15) rmax indicates that our observation is mainly
driven by the constant Aφ vector potential and not by the
1/r-decaying B field, significant solely in the vicinity of the
origin. This automatically suggests that the Zeeman interac-
tion should not play a significant role, which can indeed be
verified numerically [see almost equivalent results for gs = 0
displayed in Figs. 7(a) and 7(b)]. Clearly, from our locally
converged solution we can only infer asymptotic properties,
and considerable future research effort will be required to
further validate our results theoretically as well as experimen-
tally. However, the fundamental origin of the two different
Hall conductivity phases is likely to be a consequence of the
distorted Landau level structure identified in Figs. 1(a) and
1(b), which possess a clearly positive curvature in the positive
Hall conductivity phase and vice versa for the negative phase.
Moreover, the rich spatial as well as energetic structure
(e.g., density variation, distorted Landau levels) provides
numerous opportunities to discover novel physical effects
emerging for different types of perturbations (e.g., spatially or
time resolved). Here, we exemplify our claim for a specific, lo-
cally resolved Hall conductivity measurement, which reveals
particularly interesting properties. We assume our system is
perturbed with a static constant electric fields directed in y
direction, which acts solely on the radial shells located at ry
[see Fig. 5(b)]. The resulting, φx-integrated Hall conductivity
FIG. 7. (a) Radially resolved Hall conductivity σxy for a homogeneous electric field perturbation along the y direction for gs = 0, i.e.,
without Zeeman interaction. To reach a locally converged solution within rmax = 50 nm, a large number of single electron eigenstates with
(cid:4)r(cid:5) (cid:15) rmax is required, which indicates that the locally observed Hall conductivity switch is driven by the constant anisotropic vector potential
instead of the localized inhomogeneous magnetic field. Thus, the locally resolved result is almost identical to the spin-dependent gs = 2
calculation, confirming that the asymptotic properties are already present at rmax = 50 nm. (b) Consequently, the integrated Hall conductivity
transition at EF = EA2 (1 − 1/4) is again considered to persist in the asymptotic limit rmax → ∞.
043059-12
CLASS OF DISTORTED LANDAU LEVELS AND HALL PHASES …
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
FIG. 8. (a) Radially resolved Hall conductivity σxy measured at rx for a cylindrical electric field perturbation at rx along y. Multiple sharp
conductivity transitions (dotted horizontal lines) are observed for this radially localized (!) perturbation, which follow a Rydberg series (labeled
by λ) that alternates with rather smooth transition of opposite sign. (b) Integrated Hall conductivity transition pattern (summation over many
local measurements at different radii), which clearly allows to distinguish smooth from sharp transitions with respect to the Fermi energy EF .
in x-direction is measured at the position of the perturbation,
i.e., at rx = ry:
σxy(rx, rx, EF ) = ie2
¯h
1
(Ea − Eb)2
(cid:2)
Ea<EF <Eb
× [(cid:4)a|δ(rx − r)ˆvx|b(cid:5)(cid:4)b|δ(rx − r)ˆvy|a(cid:5)
− (cid:4)a|δ(rx − r)ˆvy|b(cid:5)(cid:4)b|δ(rx − r)ˆvx|a(cid:5)]. (49)
The resulting Hall conductivity is displayed with respect to
the Fermi level EF in Figs. 8(a) and 8(b). In contrast to
the previous homogeneous perturbation, we find a fractional
quantum Hall conductivity pattern with sharp (!) transitions
that alternate with smooth sign changes in-between. Interest-
ingly, the sharp Hall conductivity transitions follow exactly
the Rydberg series of the Hydrogen energy levels (horizon-
tally dotted lines), i.e., they are observed at
Eλ := EA2 (1 − λ),
λ = 1
n2
λ
, nλ ∈ N.
(50)
(51)
Hence, some Hydrogen properties are recovered at least for
this specific perturbation, which one probably might have
expected, due to the similarity of the corresponding partial
differential equations. Notice that similarly to the previous
computations, the convergence of the Hall conductivity be-
comes increasingly complex at high Fermi Energies, i.e., for
λ (cid:2) 1/16, due to the strong delocalization of the wave func-
tions.
V. CONCLUSION AND OUTLOOK
To our knowledge, this work establishes the first simple
and explicit, analytical solution for an extended 2D electron
gas subject to a static inhomogeneous magnetic field including
the Zeeman interaction. The resulting exact eigenstates pro-
vide access to the many-body properties of a noninteracting
electron gas, which can be calculated numerically and even
analytically (for EF = 0+) in the thermodynamic limit. Based
on those results, distorted Landau levels could be identified,
which eventually give rise to spin-dependent, localized den-
sity / current oscillations as well as distinct switching between
different asymptotic Hall conductivity phases, driven by the
anisotropic vector potential instead of the decaying magnetic
field. Overall our findings highlight that our exact solution
gives rise to a variety of fundamental new physical effects,
which strongly deviate from the Landau solution, locally as
well as in the asymptotic limit.
However, certainly the experimental verification of our
theoretical results will require considerable future research
effort. Nevertheless, we believe this will be a highly rewarding
endeavor. Despite our fundamental observations made so far,
we are far from having explored the full potential of our
solution yet. Indeed, we believe that our exact solution opens
the door to enter novel physical regimes providing numerous
theoretical and experimental opportunities, which are await-
ing to be explored. For example, novel effects are anticipated
to emerge at a zero Hall conductivity phase transition. More-
over, at the moment, we still lack an asymptotic description
of the density of states, which would allow to determine the
dependency of the Fermi energy EF (Aφ ) with respect to the
applied vector potential strength for a fixed electron density
of the underlying 2D material. Having access to these asymp-
totic properties, could enable the exploration of the magnetic
susceptibility (e.g., De-Haas van Alphen like effects [1]) or
the emergence of different Shubnikov-De Haas [47] like con-
ductivity oscillations for varying magnetic field strengths.
Generally speaking, the application of Kubo’s linear response
theory can be extended to further (static, localized, time- and
even spin-dependent) perturbations, beyond the measurement
of the Hall conductivity. Those theoretical investigations are
ideal to propose and design novel experimental setups. From
a theoretical perspective the formal connection to established
impurity models, such as Anderson localization [22] or the
Kondo effect [23], is still absent. In contrast to our setting,
they are formulated in reciprocal space assuming periodic
systems. Introducing periodicity (e.g., due to the presence
of a lattice) may not be trivial in our setting, due to the
043059-13
SIDLER, ROKAJ, RUGGENTHALER, AND RUBIO
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
nonperturbative nature of the magnetic field induced impurity.
Potentially, one could either try to introduce periodicity per-
turbatively, or one could reach for a Hofstadter-butterfly type
of effect [6], assuming periodic magnetic-field perturbations
instead. Apart from raising these open theoretical questions,
our analytic solution may also help the numerical descrip-
tion of light-(quantum)matter interactions in inhomogeneous
magnetic fields. For example, the discovered analytic eigen-
functions could be a reasonable basis-set choice for numerical
simulations of differently decaying, radial symmetric B fields,
which may be easier accessible experimentally than our
1/r solution. Furthermore, corrections from electron-electron
(Coulomb) or even current-current (transverse) interactions
should be straightforward to include numerically on differ-
ent levels of approximations (e.g., Jellium setting). On the
long run, it would also be exciting to investigate our inho-
mogeneous setting in the context of (doped) two-dimensional
heterostructures, similar to Landau levels physics in 2D Moiré
materials [48], which are governed by the interplay of topo-
logical, correlation as well as band structure effects. Overall,
we believe that the discovered analytic solution will serve as
a paradigmatic model for a large number of future theoretical
as well as experimental investigations.
ACKNOWLEDGMENTS
We thank Simone Latini for inspiring discussions. This
work was made possible through the support of the RouTe
Project (13N14839), financed by the Federal Ministry of
Education and Research (Bundesministerium für Bildung
und Forschung (BMBF)) and supported by the European
Research Council (ERC-2015-AdG694097), the Cluster of
Excellence “CUI: Advanced Imaging of Matter” of the
Deutsche Forschungsgemeinschaft (DFG), EXC 2056, project
ID 390715994 and the Grupos Consolidados (IT1249-19).
V.R. acknowledges support from the NSF through a grant
for ITAMP at Harvard University. The Flatiron Institute is a
division of the Simons Foundation.
D.S. initiated the project, discovered the simple closed
form solution and performed corresponding analytic as well as
numerical calculations. M.R. contributed to the mathematical
accuracy and rigorosity. V.R. added expertise and calcula-
tions to connect with the homogeneous Landau setting. All
authors developed the physical interpretation and wrote the
manuscript.
Numerical data available upon request.
APPENDIX A: SCALED EIGENFUNCTIONS
To calculate radial expectation values, it can be convenient
to express the eigenfunctions in Eq. (19) in terms of r:
(cid:5)
l
(cid:4)
(cid:6)n,l,s =
1(cid:13)
eilφe− qAφ
˜Nn,l,s
(cid:4)
¯h
2(l+s)
2n+1 r
2(l + s)
2n + 1
2qAφ
¯h
(cid:5)
× L2l
n−l
2qAφ
¯h
2(l + s)
2n + 1
r
χ (s).
(A1)
Notice the simple expression for the wave-functions at
EF = 0,
ψ
n,n, 1
2
(ρ, φ) =
(cid:19)
√
1
2π (2n + 1)!
einφe− ρ
2 ρnχ
(cid:4)
ψ
n,n, 1
2
(r, φ) =
2q2A2
φ
¯h2π (2n + 1)!
einφe− qAφ
¯h r
(cid:8)
2qAφ
¯h
r
(cid:5)
1
2
(cid:9)
n
, (A3)
(cid:5)
(cid:4)
χ
1
2
(A4)
with ρ = 2qAφ
¯h r, which allows to perform the infinite summa-
tion over this infinitely degenerated many-body state. Notice
that the condition n = l leads to trivial associated Laguerre
polynomials L2l
0
= 1.
The electronic dipole transition selection rules in Eqs. (26)
and (27), of the single electron eigenfunctions given in
Eq. (19) for ρ (cid:7)→ r, arise from
(cid:4)n, l, s|r|n(cid:11), l (cid:11), s(cid:11)(cid:5)
= (cid:4)Rn,l,s|r|Rn(cid:11),l (cid:11),s(cid:11) (cid:5)
× |(cid:4)(cid:7)l | cos(φ)er − sin(φ)eφ|(cid:7)l (cid:11)(cid:5)|(cid:4)χs|χs(cid:11) (cid:5)
(cid:14) ∞
=
1(cid:13)
ρ
n,l,s (r)+ρ
2
e−
n(cid:11) ,l(cid:11) ,s(cid:11) (r)
0
˜Nn,l,s ˜Nn(cid:11),l (cid:11),s(cid:11)
× (ρn,l,s(r))l (ρn(cid:11),l (cid:11),s(cid:11) (r))l (cid:11) · L2l
× L2l (cid:11)
n(cid:11)−l (cid:11) (ρn(cid:11),l (cid:11),s(cid:11) (r))r2drδl±1,l (cid:11) δs,s(cid:11)
n−l (ρn,l,s(r))
∝ 1
Aφ
,
(A5)
(A6)
2
n(cid:11) ,l(cid:11) ,s(cid:11) (r)
where the Aφ proportionality can be shown by suitable change
of variable x = ρn,l,s (r)+ρ
, which eventually removes the
Aφ from the exponential and Laguerre functions. Thus Aφ
appears only as a prefactor of the integral. Accurate and ef-
ficient numerical evaluation of Eq. (A5) can be performed in
x space by means of generalized Gauss-Laguerre quadrature.
The radial parts of the subsequent velocity matrix elements
can be integrated in a similar fashion for the Hall conductivity
in Eq. (47),
(cid:4)n, l, s|ˆvx|n(cid:11), l (cid:11), s(cid:11)(cid:5) =
(cid:3)
−i ¯h
m
(cid:4)
(cid:4)(cid:7)| cos φ|(cid:7)(cid:11)(cid:5)(cid:4)R|∂r|R(cid:11)(cid:5)
(cid:5)
− (cid:4)(cid:7)| sin φ∂φ|(cid:7)(cid:11)(cid:5)(cid:4)R| 1
r
|R(cid:11)(cid:5)
(cid:6)
(cid:4)(cid:7)| sin φ|(cid:7)(cid:11)(cid:5)(cid:4)R|R(cid:11)(cid:5)
+ qAφ
m
(cid:4)
(cid:3)
−i ¯h
m
(cid:4)(cid:7)| sin φ|(cid:7)(cid:11)(cid:5)(cid:4)R|∂r|R(cid:11)(cid:5)
(cid:5)
+ (cid:4)(cid:7)| cos φ∂φ|(cid:7)(cid:11)(cid:5)(cid:4)R| 1
r
|R(cid:11)(cid:5)
(cid:6)
(cid:4)(cid:7)| sin φ|(cid:7)(cid:11)(cid:5)(cid:4)R|R(cid:11)(cid:5)
(cid:4)χs|χs(cid:11) (cid:5),
(A7)
(cid:4)χs|χs(cid:11)(cid:5).
(A8)
r
(cid:4)n, l, s|ˆvy|n(cid:11), l (cid:11), s(cid:11)(cid:5) =
The corresponding normalization constant changes to
˜Nn,l,s = 2π
(cid:4)
2qAφ
¯h
2(l + s)
2n + 1
(cid:5)−2 (n + l )!
(n − l )!
(2n + 1). (A2)
− qAφ
m
043059-14
CLASS OF DISTORTED LANDAU LEVELS AND HALL PHASES …
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
The selection rules (cid:13)l = ±1 apply to all angular transition
matrix elements (with different coefficients) and for the spin
we find (cid:13)s = 0. However, no exact selection rule applies to
the radial transition matrix elements, in particular (cid:4)R|R(cid:11)(cid:5) (cid:9)= 0.
APPENDIX B: SINGLE ELECTRON CURRENTS
AND MAGNETIZATION
Interestingly, for our system, simple closed form solutions
can be calculated for the expected single electron currents:
Jpara
n,l,s
((cid:4)n, l, s|
−→
∇ |n, l, s(cid:5) − (cid:4)n, l, s|
←−
∇ |n, l, s(cid:5))
= ¯hq
2mi
= ¯hq
mi
= q2Aφ
m
(cid:4)n, l, s| 1
r
∂
∂φ
|n, l, s(cid:5)eφ
4l (l + s)
(2n + 1)2
eφ →
(cid:20)
0
q2Aφ
m
2n
2n+1 eφ
if E tot
n,l,s
if E tot
n,n, 1
2
≈ EA2
= 0
,
Jdia
n,l,s
Js
n,l,s
= − q2
m
(cid:14)
2π
=
(cid:4)n, l, s|A|n, l, s(cid:5) = − q2Aφ
m
(cid:14) ∞
eφ,
(∇ × ms)rdrdφ
0
0
(B1)
(B2)
|n, l, s(cid:5)eφ
|n, l, s(cid:5)eφ
∂
∂r
s(cid:4)n, l, s|
= − 2 ¯hq
m
s(cid:4)n, l, s| 1
r
s(l + s)
(2n + 1)2
P.I= ¯hq
m
= 4q2Aφ
m
(cid:20)
0
q2Aφ
m
eφ →
1
2n+1 eφ
if E tot
n,l,s
if E tot
n,n, 1
2
≈ EA2
= 0
,
(B3)
with
n,l,s := Jpara
Jtot
n,l,s
+ Jdia
n,l,s
+ Js
n,l,s
(cid:3) 0 →
(cid:16)
Jdia
n,l,s
0
if E tot
n,l,s
if E tot
n,n, 1
2
≈ EA2
= 0
.
(B4)
in state n, l, s. Single electron currents are visualized in Fig. 9
with respect to their radial expectation values and energy. To
obtain Eq. (B1), it was used that the radial part R(r) of our
wave function is real as well as the gradient operator in er,
which cancels the paramagnetic current in r-direction. With a
similar argument only the eφ component of the applied curl
operator survives for the magnetization current in Eq. (B3).
For the solution of the integrals in Eq. (B1), the following
relation was used [34]
FIG. 9. Expected single electron currents with respect to their
energy and radial position expectation values. They nicely illustrate
that every electron leads has a positive total current J tot (cid:3) 0
(cid:4)n, l, s| 1
r
∂
∂φ
|n, l, s(cid:5) = 2qAφ
¯hNn,l
(cid:14)
2(l + s)
2n + 1
(cid:14) ∞
2π
(il )
with the scaling given in Eq. (20) and (cid:17)(n + l + 1) = (n + l )!
for integers.
×
0
0
2qAφ
¯hNn,l
qAφ
¯h
2l (l + s)
2n + 1
4l (l + s)
(2n + 1)2
= i
= i
e−ρρ2l
(cid:17)
2
(cid:15)
L2l
n−l
dρdφ
2π (cid:17)(n + l + 1)
(n − l )!
(B5)
043059-15
APPENDIX C: LOWEST FLAT BAND DENSITY IN
RECIPROCAL SPACE
Interestingly, a simple closed form solution ˜n0(k), given in
Eq. (35) can be derived for the 2D radial Fourier transform
by utilizing the radial symmetry of n0(ρ) that reduces the
SIDLER, ROKAJ, RUGGENTHALER, AND RUBIO
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
problem to a 2D Hankel transformations instead, i.e.,
˜n0(k) := 2π
(cid:14) ∞
0
(cid:14) ∞
J0(kr)n0(r)rdr
= 2π
0
= 2q2Aφ
¯h
J0(kr)
(cid:4)
−
(cid:21)
1
k
q2Aφ
(cid:15)
1 − e− 4qAφ
π ¯hr
¯h
(cid:17)
r
rdr
(cid:5)
,
1
(cid:15)
(cid:17)
2
4qAφ
¯h
k2 +
(C1)
where J denotes the Bessel function with J0(0) = 1 and the
following Hankel relations were used for n (cid:7)→ ˜n by setting
t = 0 [49],
1
r
(cid:7)→ 1
k
,
1
r
J0(tr)e−sr (cid:7)→
(cid:13)
π
2
(k + t )2 + s2
(cid:4)(cid:19)
K
(cid:5)
.
4kt
(k + t )2 + s2
(cid:4)
(cid:5)
and the eigenfunctions of the operator above are Hermite
functions of the variable y + ¯hkx/eB
mωc
π ¯h
mωc
¯h
(cid:4)
y + ¯hkx
eB
y + ¯hkx
eB
= 1√
2 ¯h (y+ ¯hkx
×Hn
e− ωc
(D6)
n!2n
eB )2
(cid:4)(cid:7)
ψn
(cid:5)(cid:5)
1/4
(cid:8)
(cid:9)
with eigenvalues ¯hωc(n + 1/2)
(cid:4)
(cid:4)
y + ¯hkx
n + 1
eB
2
= ¯hωc
ˆHyψn
(cid:5)
(cid:5)
(cid:4)
y + ¯hkx
eB
(cid:5)
,
ψn
(D7)
with n ∈ N. Thus, applying now ˆH φkx on the shifted Hermite
functions ψn(y + ¯hkx/eB), we obtain
(cid:15)
n + 1
2
ψn =
ˆH φkx
¯hωc
(D8)
ψn.
φkx
(cid:17)(cid:12)
(cid:11)
From the expression above, we deduce that the full set of
eigenfuctions for an electron in a classical homogeneous mag-
netic field is
(C2)
(cid:6)kx,n(r) = φkx (x)ψn
(D9)
(cid:4)
(cid:5)
,
y + ¯hkx
eB
The complete elliptic integral of the first kind is labeled by K
with K (0) = π
2 .
with eigenenergies
= ¯hωc
En,kx
(cid:17)
(cid:15)
n + 1
2
with kx, kz ∈ R, n ∈ N.
(D10)
APPENDIX D: LANDAU LEVELS & ZERO CURRENTS
Next we would like to compute the current of each Landau
Free electrons in a 2D material in the presence of a classi-
cal homogeneous magnetic field along the z direction Bext =
Bez of strength B are described by the minimally coupled
Schrödinger Hamiltonian
ˆH = 1
2me
(i ¯h∇ + eAext(r))2,
(D1)
where in the Landau gauge the external vector potential which
gives rise to the magnetic field is Aext(r) = −exBy. The Lan-
dau gauge is very convenient because it preserves translational
invariance in the x direction. This implies that the Hamiltonian
of Eq. (D1) commutes with the translation operator for the x
direction and consequently the eigenfunctions of ˆH in x will
be plane waves
φkx (x) = eikxx with kx ∈ R.
(D2)
Applying ˆH on the plane waves above, we have
(cid:6)
(cid:5)
2
(cid:3)
− ¯h2
2m
∂ 2
∂y2
+ mω2
2
c
(cid:4)
y + ¯hkx
eB
ˆH φkx
=
φkx
,
(D3)
where we introduced also the cyclotron frequency ωc
ωc = eB
m
.
(D4)
In Eq. (D3), the part depending on the variable y remains to
be treated. The part of ˆH depending on y is a shifted harmonic
oscillator
ˆHy = − ¯h2
2m
∂ 2
y
+ mω2
2
c
(cid:5)
2
(cid:4)
y + ¯hkx
eB
level. The current operator is
ˆJ = e
m
(−i ¯h∇ − eAext(r)) = e
m
(−i ¯h∇ + eByex ).
(D11)
(cid:22)
Then for the current operator on each Landau level, we have
m
e
(cid:23)
(cid:23)(cid:6)kx,n
(cid:6)kx,n
(cid:23)
(cid:23)ˆJ
(cid:24)
= −i ¯hey
(cid:14) ∞
(cid:5)
(cid:4)
y + ¯hkx
eB
(cid:4)
y + ¯hkx
eB
×
dyψn
−∞
(cid:14) ∞
+ ex
× ψn
dyψn
−∞
(cid:4)
y + ¯hkx
eB
(cid:5)
.
(cid:4)
(cid:5)
y + ¯hkx
eB
∂yψn
(cid:5)
( ¯hkx + eBy)
(D12)
To compute the integrals above, we introduce the coordinate
s = y + ¯hkx/eB and we have
(cid:23)
(cid:23)(cid:6)kx,n
dsψn(s)∂sψn(s)
= −i ¯hey
(cid:6)kx,n
(cid:14) ∞
(cid:23)
(cid:23)ˆJ
(cid:22)
(cid:24)
m
e
−∞
(cid:14) ∞
+ exeB
−∞
dsψn(s)sψn(s).
(D13)
Further, we use the recursion relations of the Hermite func-
tions
(cid:7)
∂sψn(s) =
(cid:7)
sψn(s) =
n
2
n
2
ψn−1(s) −
(cid:7)
ψn−1(s) +
(cid:7)
n + 1
2
n + 1
2
ψn+1(s),
(D14)
ψn+1(s),
(D15)
(D5)
and the orthogonality relations of the Hermite functions
(cid:4)ψn|ψm(cid:5) = δnm and we find that expectation value of the cur-
043059-16
CLASS OF DISTORTED LANDAU LEVELS AND HALL PHASES …
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
rent operator for every Landau level is zero
(cid:22)
(cid:24)
(cid:23)
(cid:23)ˆJ
(cid:23)
(cid:23)(cid:6)kx,n
(cid:6)kx,n
= 0.
(D16)
Finally, we would like to note that that in the presence of
external constant electric field, the Landau levels get shifted
by the electric field and form edge states which lead to the
famous integer quantum Hall effect [3,50].
[1] L. D. Landau, Diamagnetismus der metalle, Z. Phys. 64, 629
(1930).
[2] L. D. Landau and E. M. Lifshitz, Quantum Mechanics, Nonrel-
ativistic Theory, 3rd ed. (Pergamon Press, Oxford, 1997).
[3] K. v. Klitzing, G. Dorda, and M. Pepper, New Method for
High-Accuracy Determination of the Fine-Structure Constant
Based on Quantized Hall Resistance, Phys. Rev. Lett. 45, 494
(1980).
[4] D. C. Tsui, H. L. Stormer, and A. C. Gossard, Two-Dimensional
Magnetotransport in the Extreme Quantum Limit, Phys. Rev.
Lett. 48, 1559 (1982).
[5] R. B. Laughlin, Anomalous Quantum Hall Effect: An Incom-
pressible Quantum Fluid with Fractionally Charged Excitations,
Phys. Rev. Lett. 50, 1395 (1983).
[6] D. R. Hofstadter, Energy levels and wave functions of bloch
electrons in rational and irrational magnetic fields, Phys. Rev. B
14, 2239 (1976).
[7] C. R. Dean, L. Wang, P. Maher, C. Forsythe, F. Ghahari, Y. Gao,
J. Katoch, M. Ishigami, P. Moon, M. Koshino, T. Taniguchi, K.
Watanabe, K. L. Shepard, J. Hone, and P. Kim, Hofstadter’s but-
terfly and the fractal quantum Hall effect in Moiré superlattices,
Nature 497, 598 (2013).
[8] J. Barrier, P. Kumaravadivel, R. Krishna Kumar, L. A.
Ponomarenko, N. Xin, M. Holwill, C. Mullan, M. Kim, R. V.
Gorbachev, M. D. Thompson, J. R. Prance, K. Taniguchi, K.
Watanabe, I. V. Grigorieva, K. S. Novoselov, A. Mischenko,
V. I. Fal’ko, A. K. Geim, and A. I. Berdyugin, Long-range
ballistic transport of Brown-Zak fermions in graphene super-
lattices, Nat. Commun. 11, 5756 (2020).
[9] C. Forsythe, X. Zhou, K. Watanabe, Z. Han, T. Taniguchi, K.
Watanabe, M. Koshino, T. Taniguchi, A. Pasupathy, P. Moon,
M. Koshino, P. Kim, and C. R. Dean, Band structure engineer-
ing of 2D materials using patterned dielectric superlattices, Nat.
Nanotechnol. 13, 566 (2018).
[10] L. Wang, Y. Gao, B. Wen, Z. Han, T. Taniguchi, K. Watanabe,
M. Koshino, J. Hone, and C. R. Dean, Evidence for a fractional
fractal quantum Hall effect in graphene superlattices, Science
350, 1231 (2015).
[11] G. Scalari, C. Maissen, D. Turˇcinková, D. Hagenmüller, S.
De Liberato, C. Ciuti, C. Reichl, D. Schuh, W. Wegscheider,
M. Beck, and J. Faist, Ultrastrong coupling of the cyclotron
transition of a 2D electron gas to a THz metamaterial, Science
335, 1323 (2012).
[12] X. Li, M. Bamba, Q. Zhang, S. Fallahi, G. C. Gardner, W.
Gao, M. Lou, K. Yoshioka, M. J. Manfra, and J. Kono, Vac-
uum Bloch-Siegert shift in Landau polaritons with ultra-high
cooperativity, Nat. Photonics 12, 324 (2018).
[13] F. Appugliese, J. Enkner, G. L. Paravicini-Bagliani, M. Beck,
C. Reichl, W. Wegscheider, G. Scalari, C. Ciuti, and J. Faist,
Breakdown of the topological protection by cavity vacuum
fields in the integer quantum hall effect, Science 375, 1030
(2022).
[14] G. L. Paravicini-Bagliani, F. Appugliese, E. Richter, S. Fallahi,
F. Valmorra, J. Keller, M. Beck, N. Bartolo, C. Rössler, T.
Ihn, K. Ensslin, C. Ciuti, G. Scalari, and J. Faist, Magneto-
transport controlled by Landau polariton states, Nat. Phys. 15,
186 (2019).
[15] A. Rubio, A new hall for quantum protection, Science 375, 976
(2022).
[16] D. Hagenmüller, S. De Liberato, and C. Ciuti, Ultrastrong cou-
pling between a cavity resonator and the cyclotron transition of
a two-dimensional electron gas in the case of an integer filling
factor, Phys. Rev. B 81, 235303 (2010).
[17] C. Ciuti, Cavity-mediated electron hopping in disordered quan-
tum hall systems, Phys. Rev. B 104, 155307 (2021).
[18] V. Rokaj, M. Penz, M. A. Sentef, M. Ruggenthaler, and A.
Rubio, Quantum Electrodynamical Bloch Theory with Homo-
geneous Magnetic Fields, Phys. Rev. Lett. 123, 047202 (2019).
[19] V. Rokaj, M. Penz, M. A. Sentef, M. Ruggenthaler, and A.
Rubio, Polaritonic Hofstadter butterfly and cavity-control of the
quantized Hall conductance, Phys. Rev. B 105, 205424 (2022).
[20] D. De Bernardis, Z.-P. Cian, I. Carusotto, M. Hafezi, and P.
Rabl, Light-Matter Interactions in Synthetic Magnetic Fields:
Landau-Photon Polaritons, Phys. Rev. Lett. 126, 103603
(2021).
[21] J. Friedel, Electronic structure of primary solid solutions in
metals, Adv. Phys. 3, 446 (1954).
[22] P. W. Anderson, Absence of diffusion in certain random lattices,
Phys. Rev. 109, 1492 (1958).
[23] J. Kondo, Resistance minimum in dilute magnetic alloys, Prog.
Theor. Phys. 32, 37 (1964).
[24] R. Hanson, L. P. Kouwenhoven, J. R. Petta, Seigo Tarucha, and
Lieven MK Vandersypen, Spins in few-electron quantum dots,
Rev. Mod. Phys. 79, 1217 (2007).
[25] E. Gull, A. J. Millis, A. I. Lichtenstein, A. N. Rubtsov, M.
Troyer, and P. Werner, Continuous-time monte carlo meth-
ods for quantum impurity models, Rev. Mod. Phys. 83, 349
(2011).
[26] R. Brako and D. M. Newns, Slowly varying time-dependent
local perturbations in metals: a new approach, J. Phys. C: Solid
State Phys. 14, 3065 (1981).
[27] D. C. Langreth and P. Nordlander, Derivation of a master equa-
tion for charge-transfer processes in atom-surface collisions,
Phys. Rev. B 43, 2541 (1991).
[28] P. Werner, A. Comanac, L. De’Medici, Matthias Troyer, and
A. J. Millis, Continuous-Time Solver for Quantum Impurity
Models, Phys. Rev. Lett. 97, 076405 (2006).
[29] V. Rokaj, M. Ruggenthaler, F. G. Eich, and A. Rubio, Free
electron gas in cavity quantum electrodynamics, Phys. Rev.
Res. 4, 013012 (2022).
[30] P. Boyle, Lecture notes on physical mathematics, 2010.
[31] H. Schlichtkrull, Lecture notes on differential equations, 2013.
[32] C. A. Downing and M. E. Portnoi, Magnetic quantum dots and
rings in two dimensions, Phys. Rev. B 94, 045430 (2016).
043059-17
SIDLER, ROKAJ, RUGGENTHALER, AND RUBIO
PHYSICAL REVIEW RESEARCH 4, 043059 (2022)
[33] C. A. Downing and M. E. Portnoi, Massless dirac fermions in
two dimensions: Confinement in nonuniform magnetic fields,
Phys. Rev. B 94, 165407 (2016).
[42] R. S. Van Dyck Jr, P. B. Schwinberg, and H. G. Dehmelt, New
High-Precision Comparison of Electron and Positron g Factors,
Phys. Rev. Lett. 59, 26 (1987).
[34] G. B. Arfken and H. J. Weber, Mathematical methods for physi-
cists, 1999.
[35] A. Zettl, Sturm-Liouville Theory (American Mathematical Soc.,
Providence, Rhodes Island, 2010), No. 121.
[36] P. B. Bailey, W. N. Everitt, and A. Zettl, Algorithm 810: The
sleign2 sturm-liouville code, ACM Trans. Math. Softw. 27, 143
(2001).
[37] G. Teschl, Mathematical Methods in Quantum Mechanics with
Applications to Schrödinger Operators (American Mathemati-
cal Soc., Providence, Rhodes Island, 2014), No. 157.
[38] W. N. Everitt, A catalogue of sturm-liouville differential equa-
tions, In Sturm-Liouville Theory (Springer, Basel, Switzerland,
2005), pp. 271–331.
[43] T. Smole´nski, P. E. Dolgirev, C. Kuhlenkamp, A. Popert,
Y. Shimazaki, P. Back, X. Lu, M. Kroner, K. Watanabe, T.
Taniguchi et al., Signatures of wigner crystal of electrons in a
monolayer semiconductor, Nature 595, 53 (2021).
[44] H. Hübener, U. D. Giovannini, C. Schäfer, J. Andberger, M.
Ruggenthaler, J. Faist, and A. Rubio, Engineering quantum
materials with chiral optical cavities, Nat. Mater. 20, 438
(2021).
[45] D. J. Thouless, M. Kohmoto, M. P. Nightingale, and M. den
Nijs, Quantized Hall Conductance in a Two-Dimensional Peri-
odic Potential, Phys. Rev. Lett. 49, 405 (1982).
[46] M. Kohmoto, Topological invariant and the quantization of the
hall conductance, Ann. Phys. 160, 343 (1985).
[39] C. S. Kim and O. Olendski, Landau levels and persistent cur-
rents in nonuniform magnetic fields, Phys. Rev. B 53, 12917
(1996).
[47] L. Schubnikow and W. J. de Haas, Magnetische widerstandsver-
grösserung in einkristallen von wismut bei tiefen temperaturen,
Leiden Commun. 207, 3 (1930).
[40] A. Edery and Y. Audin, New degeneracies and modification of
landau levels in the presence of a parallel linear electric field,
J. Phys. Commun. 3, 025013 (2019).
[41] C. Schäfer, F. Buchholz, M. Penz, M. Ruggenthaler, and A.
Rubio, Making ab initio QED functional(s): Nonperturbative
and photon-free effective frameworks for strong lightmatter
coupling, PNAS 118, e2110464118 (2021).
[48] E. Y. Andrei, D. K. Efetov, P. Jarillo-Herrero, A. H. MacDonald,
K. F. Mak, T. Senthil, E. Tutuc, A. Yazdani, and A. F. Young,
The marvels of moiré materials, Nat. Rev. Mater. 6, 201 (2021).
[49] A. Papoulis, Systems and Transforms with Applications in Optics
(McGraw-Hill Series in System Science, 1968).
[50] R. B. Laughlin, Quantized Hall conductivity in two dimensions,
Phys. Rev. B 23, 5632 (1981).
043059-18
| null |
10.1038/s41523-022-00429-7
| null |
DATA AVAILABILITY The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
|
www.nature.com/npjbcancer
OPEN
ARTICLE
A prospective trial of treatment de-escalation following
neoadjuvant paclitaxel/trastuzumab/pertuzumab in HER2-
positive breast cancer
Adrienne G. Waks1,2,3,13, Neelam V. Desai
Laura M. Spring3,7, Meredith Faggen3,8, Michael Constantine3,6, Otto Metzger1,2,3, Jillian Alberti1, Julia Deane1,10,
Shoshana M. Rosenberg1,2,3,11, Elizabeth Frank1,2, Sara M. Tolaney
Tari A. King2,3,9, Elizabeth A. Mittendorf2,3,9 and Eric P. Winer
3,4,13, Tianyu Li5, Philip D. Poorvu1,2,3, Ann H. Partridge 1,2,3, Natalie Sinclair3,6,
1,2,3,12 ✉
1,2,3, Ian E. Krop 1,3,12, Nadine M. Tung 3,4, Nabihah Tayob3,5,
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
De-escalating adjuvant therapy following pathologic complete response (pCR) to an abbreviated neoadjuvant regimen in human
epidermal growth factor receptor 2-positive (HER2+) breast cancer is the focus of international research efforts. However, the
feasibility of this approach and its appeal to patients and providers had not been formally investigated. We aimed to assess
adherence to de-escalated adjuvant antibody doublet therapy (trastuzumab and pertuzumab [HP], without chemotherapy) among
patients with pCR following neoadjuvant paclitaxel/HP (THP). In this single-arm prospective trial, patients with treatment-naïve
stage II-III HER2+ breast cancer received neoadjuvant weekly paclitaxel ×12 and HP every 3 weeks ×4. The primary endpoint was
receipt of adjuvant non-HER2-directed cytotoxic chemotherapy. Ninety-eight patients received ≥1 dose of THP on study. Patients
had median age of 50 years, 86% had stage II tumors, and 34% were hormone receptor-negative. Five patients had incomplete
clinical response following THP and received doxorubicin and cyclophosphamide before surgery; they were classified as non-pCR
and censored from further analyses. The overall pCR rate was 56.7%. Among patients with pCR, the adherence rate to de-escalated
antibody-only therapy (HP) was 98.2% (95% CI 90.3–100.0%), and the primary feasibility endpoint was reached. The majority of
patients felt positive or neutral about their adjuvant treatment plans. With brief follow-up (median 19.1 months), there were no
breast cancer recurrences. De-escalation of adjuvant chemotherapy among patients who experience pCR in early-stage HER2+
breast cancer is a practicable approach for both patients and physicians. Planned and ongoing prospective trials will determine the
long-term efficacy of this approach.
Trial registration clinicaltrials.gov, NCT03716180, https://clinicaltrials.gov/ct2/show/NCT03716180.
npj Breast Cancer
(2022) 8:63 ; https://doi.org/10.1038/s41523-022-00429-7
INTRODUCTION
Modern treatment regimens for human epidermal growth factor
receptor 2-positive (HER2+) breast cancer produce favorable long-
term outcomes in the vast majority of patients with non-
metastatic disease. The APHINITY trial demonstrated 3-year
invasive disease-free survival (DFS) of 92% among node-positive
early-stage HER2+ breast cancer patients treated with trastuzu-
mab (H) and pertuzumab (P) plus adjuvant chemotherapy1.
regimens for
However, current standard-of-care neo/adjuvant
stage II-III HER2+ breast cancer involve 2–3 chemotherapy agents
plus HER2-directed therapy2, and these regimens are associated
with both serious and burdensome short- and long-term
toxicities3.
It is of great interest to determine if a subset of
patients with anatomic stage II-III HER2+ breast cancer can be
adequately treated with curative intent using less toxic therapy.
Pathologic complete response (pCR) at surgery following
neoadjuvant therapy is a strong favorable prognostic biomarker
in all subtypes of breast cancer, including HER2+ breast cancer
treated with standard modern regimens incorporating HER2-
targeted therapy4–6. pCR is associated with an excellent long-term
outcome and may identify patients who are prime candidates for
de-escalated adjuvant treatment. Preliminary data indicate that
pCR correlates with excellent long-term outcomes in HER2+
regimen is
breast
cancer
non-standard7,8.
or
chemotherapy-sparing
The
CompassHER2-pCR trial
is ongoing and will
(NCT04266249)
determine recurrence-free survival among patients with HER2+
breast cancer who receive an abbreviated neoadjuvant regimen
and experience pCR, then omit additional standard cytotoxic
chemotherapy.
even when the neoadjuvant
otherwise
1Medical Oncology, Dana-Farber Cancer Institute, Boston, MA, USA. 2Breast Oncology Program, Dana-Farber/Brigham and Women’s Cancer Center, Boston, MA, USA. 3Harvard
Medical School, Boston, MA, USA. 4Medical Oncology, Beth Israel Deaconess Medical Center, Boston, MA, USA. 5Department of Data Sciences, Dana-Farber Cancer
Institute, Boston, MA, USA. 6Hematology/Oncology, Dana-Farber/Brigham and Women’s Cancer Center at Milford, Milford, MA, USA. 7Hematology/Oncology, Massachusetts
General Hospital, Boston, MA, USA. 8Hematology/Oncology, Dana-Farber/Brigham and Women’s Cancer Center at South Shore Hospital, South Weymouth, MA, USA. 9Division of
Breast Surgery, Department of Surgery, Brigham and Women’s Hospital, Boston, MA, USA. 10Present address: Clinical Affairs, TransMedics, Inc, Andover, MA, USA. 11Present
address: Department of Population Health Sciences, Weill Cornell Medicine, New York, NY, USA. 12Present address: Yale Cancer Center, New Haven, CT, USA. 13These authors
contributed equally: Adrienne G. Waks, Neelam V. Desai.
email: [email protected]
✉
Published in partnership with the Breast Cancer Research Foundation
Table 1. Patient and tumor characteristics.
No. of patients (%) (N = 98)
49.5 (24–78)
Characteristic
Age, years
Median (range)
Sex
Female
Male
Race
White
Black
Asian
Other
Ethnicity
Hispanic or Latino
Non-Hispanic
Unknown
ECOG PS at baseline
0
1
Unknown
Stage at initial diagnosis
II
III
T status
Tx
T1
T2
T3
T4
N status
N0
N1
N2
N3
;
,
:
)
(
0
9
8
7
6
5
4
3
2
1
2
A.G. Waks et al.
Patients’ and providers’ acceptance of a pCR-based de-
escalated treatment approach has not been formally investigated.
One recent survey found that 43% of breast cancer patients were
not interested in clinical trials investigating chemotherapy de-
escalation, with fear of cancer recurrence and fear of regret being
the most commonly cited reasons for concern9. Understanding
concerns and preferences around this paradigm will be important
the new
for optimizing communication with patients about
potential strategy and encouraging its uptake among appropriate
patients.
The goal of this trial
(DAPHNe: De-escalation to Adjuvant
antibodies Post-pCR to Neoadjuvant THP) was to assess the
feasibility of de-escalating therapy from a multi-agent to a single-
agent chemotherapy backbone plus HP in select patients with
anatomic stage II-III HER2+ breast cancer, based on pCR as a
prognostic biomarker. All patients were planned to receive
neoadjuvant paclitaxel-HP (THP), and patients who experienced
pCR were recommended to receive adjuvant HP only, without
further adjuvant cytotoxic chemotherapy. The primary objective
was to assess adherence to the protocol-specified de-escalated
adjuvant regimen (HP only) among patients with pCR. Post-
operative patient questionnaires were administered to all patients
and physician rationales were reviewed in the medical record to
explore patient and provider attitudes in adjuvant
therapy
decision-making.
RESULTS
Patient characteristics
Table 1 summarizes patient and tumor characteristics for 98
patients who began treatment on trial. The large majority of
patients had clinical anatomic stage II disease (85.7%), and
approximately one-third of patients had hormone receptor-
negative (HR-) tumors (33.7%). Supplementary Table 1 shows all
neoadjuvant treatments received: 84.7% of patients completed all
12 doses of neoadjuvant paclitaxel, and 99%/98% of patients
completed at least 4 doses of neoadjuvant H/P, respectively. One
patient withdrew early for
included in
subsequent analyses. Five patients (5.1%) had obvious residual
disease at the completion of THP and received preoperative
doxorubicin and cyclophosphamide (AC); all other patients
underwent surgery following THP (Fig. 1).
toxicity and is not
Neoadjuvant therapy responses and adjuvant therapy
received
The overall pCR rate was 56.7%, with residual cancer burden (RCB)
I,
II, and III responses in 9.3%, 26.8%, and 2.1% of patients,
respectively. The pCR rate was 42.2% for hormone receptor-
positive (HR+) patients, and 84.8% for HR- patients (Fig. 2). Table 2
shows all adjuvant therapies received by RCB category. Among
patients who experienced pCR following neoadjuvant THP (N =
55), the rate of adherence to de-escalated antibody-only therapy
(HP) was 98.2% (95% confidence interval [CI] 90.3–100.0%). Thus,
the trial met its primary feasibility endpoint (p value from binomial
test: <0.001). Among the remaining 37 patients with non-pCR
responses to neoadjuvant THP, 16 patients received adjuvant
chemotherapy (AC) [N = 14]; cyclophosphamide alone [N = 2], and
21 patients did not receive adjuvant chemotherapy (19 of whom
received adjuvant T-DM1). Overall, 29/37 patients who did not
have a pCR (78%) received at least one dose of adjuvant T-DM1.
84.4% of patients with HR+ disease (54/64 patients) initiated
adjuvant hormonal therapy.
With 19.1 months of median follow-up, there were no breast
cancer recurrences, new primary breast cancers, or deaths. One
patient was diagnosed with metastatic small cell carcinoma of
likely pancreatic primary.
Hormone receptor status
ER+/PR+
ER+/PR−
ER−/PR+
ER−/PR−
HER2 status
Positive
Size of breast tumor by physical exam (cm)
Median (range)
Breast surgery
Lumpectomy
Mastectomy
ECOG PS Eastern Cooperative Oncology Group Performance Status, ER
estrogen receptor, HER2 human epidermal growth factor receptor 2, PR
progesterone receptor.
Patient and provider attitudes toward chemotherapy and de-
escalation
Post-operative questionnaires were administered to 100% of
patients to query patients’ experiences with neoadjuvant che-
motherapy, attitudes toward additional adjuvant chemotherapy,
and perceived alignment with their treating physician about the
97 (99%)
1 (1%)
82 (83.7%)
5 (5.1%)
7 (7.1%)
4 (4.1%)
5 (5.1%)
89 (90.8%)
4 (4.1%)
93 (94.9%)
4 (4.1%)
1 (1%)
84 (85.7%)
14 (14.3%)
1 (1%)
17 (17.3%)
72 (73.5%)
8 (8.2%)
0 (0%)
65 (66.3%)
30 (30.6%)
2 (2%)
1 (1%)
45 (45.9%)
18 (18.4%)
2 (2%)
33 (33.7%)
98 (100%)
3 (0–6)
54 (55.1%)
44 (44.9%)
npj Breast Cancer (2022)
63
Published in partnership with the Breast Cancer Research Foundation
A.G. Waks et al.
3
Fig. 1 Trial flow diagram. pCR pathologic complete response, THP paclitaxel/trastuzumab/pertuzumab.
numerically most
likely to report a better
preoperative chemotherapy experience (Fig. 3a).
than expected
The large majority of patients felt positive or neutral about their
adjuvant treatment plans, regardless of whether they planned to
omit or receive additional chemotherapy such as AC. Among
patients who did not plan to receive adjuvant chemotherapy,
though most felt positive or neutral about that decision (score
1–3), a small minority (3.7% who had experienced pCR, and 9.5%
they
who had not experienced pCR)
“should” receive more chemotherapy (score 4–5)—despite not
planning to receive more. Among patients who planned to
receive adjuvant chemotherapy after not experiencing pCR, 100%
felt positive or neutral about that decision (score 3–5; Fig. 3b).
61.5% of patients overall felt aligned with their treating physician
about adjuvant chemotherapy decisions while 20.9% of patients
felt non-aligned (with 17.6% missing data for this two-question
analysis; Fig. 3c).
reported feeling that
Patient and physician rationale for administering or omitting
adjuvant chemotherapy were also explored through questionnaires
and medical record review, with opportunity for prespecified or free-
text responses. For patients who did not achieve pCR and did not
receive adjuvant chemotherapy such as AC (N = 21), the most
common reason cited for omitting adjuvant chemotherapy was plan
for adjuvant T-DM1 (cited by 14 patients and 17 physicians), and the
second most common reason was a good response to neoadjuvant
chemotherapy (cited by 8 patients and 7 physicians; Supplementary
Table 3). Themes that emerged from free-text responses were
grouped by omission or receipt of adjuvant chemotherapy such as AC
after either pCR or lack of pCR, respectively. Among patients with pCR,
themes related to omission of adjuvant chemotherapy included (1)
following physician advice, (2) emphasizing the importance of pCR
found at surgery, and (3) worry about chemotherapy toxicity. Among
themes related to receipt of adjuvant
patients without pCR,
chemotherapy included (1) high disease risk, and (2) following the
most evidence-based treatment approach regardless of side effects.
Supplementary Table 4 contains all patient-written responses.
Fig. 2 Pathologic response results. Non-pCR indicates patients
who received additional neoadjuvant chemotherapy following
paclitaxel/trastuzumab/pertuzumab. HR hormone receptor, pCR
pathologic complete response, RCB residual cancer burden.
need for additional adjuvant chemotherapy. Response data are
shown according to the following patient categories: no pCR and
did not receive adjuvant chemotherapy; yes pCR and did not
receive adjuvant chemotherapy; no pCR and did receive adjuvant
chemotherapy (Fig. 3, associated data in Supplementary Table 2).
Non-de-escalator patient data (yes pCR and did receive adjuvant
chemotherapy) are included only in the supplement as only one
patient was in this category. There was a 10–20% non-response
rate for all questions, with approximately equivalent non-response
rates across patient categories. In all patient categories, ≥50% of
patients felt that preoperative chemotherapy went better than
expected (score 4–5), and patients who experienced pCR were
Published in partnership with the Breast Cancer Research Foundation
npj Breast Cancer (2022)
63
4
Table 2. All non-hormonal adjuvant systemic therapies received.
A.G. Waks et al.
pCR status
Adjuvant cytotoxic chemotherapy received
Adjuvant antibody therapy received
pCR aka RCB 0
(N = 55)
Regimen
AC ×4 cycles
None
No. patients (%)
Regimen
No. patients (%)
1 (1.8%)
(95% CI 0.05–9.7%)
H (trastuzumab)
P (pertuzumab)
54 (98.2%)
(95% CI 90.3–100%)
RCB I
(N = 9)
RCB II
(N = 26)
RCB III
(N = 2)
AC ×4 cycles
1 (11.1%)
None
8 (88.9%)
AC ×4 cyclesa
12 (46.2%)
Cyclophosphamide x4 cycles
2 (7.7%)
None
12 (46.2%)
AC x4 cycles
None
1 (50%)
1 (50%)
T-DM1
H
P
T-DM1
H
P
T-DM1
H
P
T-DM1
H
P
T-DM1
H
P
T-DM1
H
P
T-DM1
H
P
T-DM1
H
P
T-DM1
1 (100%)
1 (100%)
0
54 (100%)
50 (92.6%)
0
0
0
1 (100%)
5 (62.5%)
4 (50%)
7 (87.5%)
6 (50%)
6 (50%)
7 (58.3%)
1 (50%)
0
1 (50%)
5 (41.7%)
2 (16.7%)
11 (91.7%)
0
0
1 (100%)
0
0
1 (100%)
Patients who received neoadjuvant AC are not included in this table.
AC doxorubicin + cyclophosphamide, CI confidence interval, pCR pathologic complete response, RCB residual cancer burden.
aIn one patient 4 cycles of AC were planned, but stopped early (after 2 cycles) for toxicity.
DISCUSSION
This trial demonstrated the feasibility of de-escalating from multi-
agent to single-agent cytotoxic chemotherapy in combination
with dual anti-HER2 antibody therapy in patients with pCR after
neoadjuvant THP. In this cohort, where the majority of patients
had clinical anatomic stage II disease, just over half (56.7%) of
patients experienced pCR. With brief follow-up in this small cohort,
no breast cancer recurrences were seen. If ongoing larger trials
(e.g. CompassHER2-pCR) demonstrate favorable long-term efficacy
associated with this treatment approach, then the majority of
patients with anatomic stage II-III HER2+ breast cancer may be
able to avoid the substantial toxicities associated with standard
combined chemotherapy regimens.
The overall pCR rate of 56.7% seen in this trial is comparable to
pCR rates previously reported in other cohorts of stage II-III HER2+
breast cancer treated with various chemo-plus-HP regimens. In the
NeoSphere trial, 4 cycles of docetaxel/HP produced a pCR rate
(ypT0/isN0) of 39.3% (N = 107)10; in the KRISTINE trial, 6 cycles of
docetaxel/carboplatin/HP (TCHP) or T-DM1/P produced pCR rates
(ypT0/isN0) of 55.7% (N = 221) and 44.4% (N = 223), respectively11;
and in the TRYPHAENA trial, 6 cycles of 5-fluorouracil/epirubicin/
cyclophosphamide-docetaxel/HP (FEC-THP) or TCHP produced pCR
rates (ypT0N0) of 45.3% (N = 75) and 51.9% (N = 76)12. As in all
other cohorts of HER2+ breast cancer treated with neoadjuvant
therapy, pCR was significantly more likely for those with HR- tumors
compared to HR+ tumors. Though patients with HR+/HER2+
tumors are less likely to experience pCR, pCR carries less prognostic
importance in this subset compared to HR−/HER2+ tumors,
presumably due to the long-term benefits of adjuvant endocrine
therapy4.
The DAPHNe trial represents a formal assessment of feasibility
for a pCR-based de-escalation approach to therapy in HER2+
breast cancer. HER2+ breast cancer is well suited to systemic
therapy de-escalation due to the development of relatively low-
toxicity, high-efficacy targeted therapies beginning with the U.S.
Food and Drug Administration approval of adjuvant trastuzumab
in 2006. The use of pCR as a patient-level surrogate for de-
escalation candidacy13 is supported by the excellent outcomes for
patients with HER2+ breast cancer and pCR regardless of
neoadjuvant
In the KRISTINE trial, patients who
experienced pCR after neoadjuvant T-DM1 plus P had 96.7%
3-year
invasive DFS (despite only 9.1% receiving adjuvant
chemotherapy), and the I-SPY2 trial reported a 93–97% 3-year
for patients with pCR following varied
event-free survival
regimens for stage II–III HER2+ breast cancer,
neoadjuvant
regimens7,8. Therefore, prospectively
including investigational
evaluating the efficacy of pCR-based de-escalation in HER2+
breast cancer is essential. The ongoing CompassHER2-pCR trial will
enroll 1250 patients with stage II-IIIA HER2+ breast cancer and
regimen.
npj Breast Cancer (2022)
63
Published in partnership with the Breast Cancer Research Foundation
A.G. Waks et al.
5
Fig. 3 Patient responses to questionnaire regarding neoadjuvant and adjuvant chemotherapy. a Patient reflections on neoadjuvant
chemotherapy. Specifically, this panel shows responses to the question, “How would you describe your experience with the chemotherapy
you received before surgery”? b Patient perspectives on adjuvant chemotherapy. Specifically, this panel shows responses to the question,
“How strongly do you feel that you should or should not receive more chemotherapy after your surgery?” Patients who selected score 1–2 (“I
feel I should not receive more chemo”) or score 3 (“I feel neutral”) and did not have adjuvant chemotherapy planned were classified as feeling
positive/neutral about their planned adjuvant regimen. Patients who selected score 4–5 (“I feel I should receive more chemo”) or score 3 (“I
feel neutral”) and had adjuvant chemotherapy planned were classified as feeling positive/neutral about their planned adjuvant regimen.
c Patient-physician alignment in planning for adjuvant chemotherapy, as rated by patients. “Aligned” was defined as: patient gave a response
of 1 or 2 on question describing patient’s feeling about adjuvant chemotherapy and question describing treating physician’s feeling about
adjuvant chemotherapy; or patient gave a response of 3 on both questions; or patient gave a response of 4 or 5 on both questions. “Not
aligned” was defined as everything else. pCR pathologic complete response.
determine recurrence-free survival with a treatment approach
to the DAPHNe trial. A similarly structured
nearly identical
European trial (DECRESCENDO) is planned for 1065 patients with
ER−/HER2+ stage I–II breast cancer (tumor size 15–50 mm)13.
Patients with stage III disease likely will not be well-represented in
these trials (with stage IIIB/C entirely excluded), as we observed in
the DAPHNe trial: only 14 stage III patients participated, though all
non-inflammatory stage III tumors were eligible.
For patients without pCR on DAPHNe, several
themes in
adjuvant therapy administration are notable. While all adjuvant
therapy was administered off-trial and therefore up to clinician
discretion, the protocol specifically recommended adjuvant T-DM1
in all patients with residual disease, and additional chemotherapy
in patients with RCB III residual disease at surgery or otherwise
high risk. At least one dose of adjuvant T-DM1 was administered in
78% of patients with residual disease. Adjuvant chemotherapy
was omitted in most patients with RCB I and approximately half of
patients with RCB II residual disease at surgery. This reflects the
fact that long-term disease outcomes are strongly associated with
RCB categorization, with increasing (less favorable) RCB score
predicting worse relapse-free survival14. Though ongoing and
planned trials will inform adjuvant therapy decisions for patients
with pCR, it is unlikely that prospective trials will be performed to
determine the optimal adjuvant regimen for patients with good
but non-pCR response to THP. Accordingly, these decisions will
continue to be made on an individualized basis, as was the case in
the DAPHNe cohort. For patients with significant residual disease
at surgery, the use of adjuvant anthracycline-based chemotherapy
(e.g. AC) will remain an important consideration. If used, AC should
be administered in a dose-dense fashion (every 2 weeks) as this
schedule was associated with improved 10-year breast cancer
outcomes in a large meta-analysis15.
Patients’ and treating physicians’ reports offer insights into the
reasoning and confidence level underlying adjuvant
therapy
decisions. Most patients reported feeling positive or neutral about
their adjuvant regimen, regardless of whether further chemother-
there were modestly
apy was planned or not. However,
numerically higher
feelings toward
adjuvant therapy plan and slightly higher rates of perceived
patient–physician alignment among patients who were planned
for adjuvant chemotherapy, potentially suggestive of a higher
level of ambivalence among patients who did not plan adjuvant
rates of positive/neutral
the potential
chemotherapy. This underscores the importance of thorough
communication about the risks and benefits of de-escalation as
well as acknowledgment of
for psychological
discomfort. Conversely, the fact that planned use of T-DM1 was
the top reason cited for de-escalation among patients without
pCR highlights patients’ and physicians’ relative comfort with the
substitution of a more targeted, less toxic agent for a standard
combination chemotherapy regimen—and likely reflects the fact
that de-escalation of toxic therapy is easier to consider when
something alternative is offered in its place.
Our trial data have several
limitations. Most patients were
enrolled at a single tertiary academic cancer center (DFCI) where
providers already had familiarity with adjuvant de-escalation trials
in HER2+ breast cancer based on participation in prior protocols,
which may have impacted their comfort level with this approach
and experience presenting it to prospective participants. Off-
setting this, approximately one in three enrolled patients were
from other centers including community satellite practices. While
even large trials of a similar de-escalation approach (Com-
passHER2-pCR and DECRESCENDO) will be potentially subject to
the same enrollment biases related to provider experience/
comfort, we expect that given larger sample sizes and broad
recruitment base, those efficacy results will be generalizable for
community uptake. The patient questionnaires used to assess
adjuvant therapy decision-making were developed by the study
team and not previously validated. Finally, we did not gather data
on the number or characteristics of patients who declined to
participate in the trial, though the rapidity of accrual (>7 patients/
month) highlights broad patient interest.
in establishing the long-term efficacy of
The DAPHNe trial formally assessed patients’ acceptance of de-
escalated adjuvant therapy in clinical anatomic stage II-III HER2+
breast cancer. Given the landscape of ongoing trials, we anticipate
that this may be a major emerging treatment paradigm in non-
metastatic HER2+ breast cancer. While larger cohorts will be
instrumental
this
treatment strategy, this trial was unique in its focus on patient
attitudes toward chemotherapy, patient-physician alignment with
to adjuvant chemotherapy, and patients’ sources of
respect
reassurance and reservation about
therapy de-
escalation within this specific patient population. We must
continue to evaluate patients’ and physicians’ perspectives on
facilitate
de-escalation in order
to optimize communication,
adjuvant
Published in partnership with the Breast Cancer Research Foundation
npj Breast Cancer (2022)
63
6
informed decision-making, and ultimately encourage uptake of
this evolving treatment approach that seeks to minimize toxicity
without compromising benefit in the appropriate contexts.
A.G. Waks et al.
METHODS
Patient population
Eligible patients had clinical anatomic stage II-III HER2+ invasive breast
cancer. HER2 positivity was defined according to 2018 American Society of
Clinical Oncology/College of American Pathologists guidelines16. Patients
could have any menopausal or hormone receptor status, and were
required to have performance status ≤1 and adequate organ function at
baseline. Patients with baseline cardiac ejection fraction <50% or
significant peripheral neuropathy (grade ≥ 2 by common terminology
criteria for adverse events v4.0) were excluded. All patients provided
written informed consent and the study was carried out in accordance
with the Declaration of Helsinki.
Treatment protocol
This was a single-arm prospective trial
that enrolled patients from
November 2018 to January 2020 at Dana-Farber/Harvard Cancer Center
(DF/HCC; composed of Dana-Farber Cancer Institute [DFCI], Massachusetts
General Hospital, and Beth Israel Deaconess Medical Center) and affiliated
community satellite practices. All patients were assigned to receive
preoperative paclitaxel (T; 80 mg/m2 weekly for 12 weeks), trastuzumab (H;
loading dose 8 mg/kg, subsequent doses 6 mg/kg, every 3 weeks for 4
cycles), and pertuzumab (P;
loading dose 840 mg, subsequent doses
420 mg, every 3 weeks for 4 cycles) prior to breast surgery. Up to two
additional cycles of HP were allowed in cases of surgical delay. Patients
with obvious residual disease at completion of THP were allowed to
receive additional neoadjuvant therapy at investigator discretion; 4 cycles
of AC was the recommended regimen. Pathologic response to neoadju-
vant therapy was quantified at surgery according to RCB score;17 pCR was
defined as RCB 0 (ypT0/isN0). Patients with pCR were suggested to
complete one year of adjuvant HP, without additional cytotoxic
chemotherapy. In patients without pCR, adjuvant systemic therapy was
per investigator discretion, with 14 cycles of trastuzumab emtansine (T-
DM1) recommended for all patients (per protocol amendment following
the KATHERINE trial data18) and 4 cycles of AC
presentation of
recommended in patients with significant
residual disease. Post-
operative hormonal therapy was administered per investigator discretion.
All patients were followed for disease outcomes post-operatively. All trial
procedures were approved by the DF/HCC institutional review board. The
full protocol is included in Supplementary Material.
Assessment of adjuvant therapy decision-making
After completion of final breast surgery, patients belonged to one of four
adjuvant therapy designations based on their pCR status and receipt of
adjuvant cytotoxic chemotherapy: (1) non-de-escalator: patients with pCR
who received adjuvant cytotoxic chemotherapy; (2) patients without pCR
who did not receive adjuvant cytotoxic chemotherapy; (3) patients without
pCR who received adjuvant cytotoxic chemotherapy; and (4) patients with
pCR who did not receive adjuvant cytotoxic chemotherapy. T-DM1 was not
considered cytotoxic chemotherapy for purposes of this categorization. A
4-item paper-based questionnaire, developed by the study team,
regarding preferences and rationale for receipt/non-receipt of adjuvant
cytotoxic chemotherapy was administered post-operatively and prior to
initiation of adjuvant systemic therapy to all patients. Prior to ques-
tionnaire administration, the final plan for adjuvant cytotoxic chemother-
apy administration (yes/no and regimen) was signed off on by the treating
physician. Treating physician rationale for administration/non-administra-
tion of adjuvant cytotoxic chemotherapy was recorded by two indepen-
dent physician reviewers based on review of progress notes in the medical
record. Discordant opinions were jointly discussed by the two reviewers
and consensus was reached. Questionnaires and standardized medical
record review forms are included in Supplementary Material.
Statistical methods
The primary objective was to assess adherence to protocol-specified
antibody doublet therapy (HP only) in the adjuvant setting among patients
with pCR following neoadjuvant THP. The primary endpoint was receipt of
adjuvant cytotoxic chemotherapy, assessed 3 months post-operatively.
Among patients with pCR to THP, de-escalation would be deemed
infeasible if the true rate of adherence to HP only was ≤80%. With a sample
size of 100 patients, the study was designed to have > 90% power to reject
the null if the true rate of adherence was ≥ 95% (binomial exact test; one-
sided alpha = 0.05). Patients who progressed during neoadjuvant THP,
withdrew consent to participate, received neoadjuvant therapy in addition
to THP, or did not have pCR were not included in the primary analysis
(prespecified). Secondary endpoints included event-free survival and
overall survival. Patients who received additional non-THP neoadjuvant
therapy were counted as non-pCR. Questionnaire and medical record
review results for analysis of adjuvant therapy decision-making were
summarized descriptively and patients who received additional neoadju-
vant therapy following THP were not included in this analysis. SAS v9.4 was
used for data analysis and R v4.0.2 was used to make figures.
Reporting summary
Further information on research design is available in the Nature Research
Reporting Summary linked to this article.
DATA AVAILABILITY
The datasets generated during and/or analyzed during the current study are available
from the corresponding author on reasonable request.
Received: 13 August 2021; Accepted: 14 March 2022;
REFERENCES
1. von Minckwitz, G. et al. Adjuvant pertuzumab and trastuzumab in early HER2-
positive breast cancer. N. Engl. J. Med. 377, 122–131 (2017).
2. Gradishar, W. J. et al. NCCN Clinical Practice Guidelines in Oncology - Breast
Cancer, Version 2.2017 (National Comprehensive Cancer Network, 2017).
3. Slamon, D. et al. Adjuvant trastuzumab in HER2-positive breast cancer. N. Engl. J.
Med. 365, 1273–1283 (2011).
4. Cortazar, P. et al. Pathological complete response and long-term clinical benefit
in breast cancer: the CTNeoBC pooled analysis. Lancet 384, 164–172 (2014).
5. Gianni, L. et al. 5-year analysis of neoadjuvant pertuzumab and trastuzumab in
patients with locally advanced, inflammatory, or early-stage HER2-positive breast
cancer (NeoSphere): a multicentre, open-label, phase 2 randomised trial. Lancet
Oncol. 17, 791–800 (2016).
6. von Minckwitz, G. et al. Definition and impact of pathologic complete response
on prognosis after neoadjuvant chemotherapy in various intrinsic breast cancer
subtypes. J. Clin. Oncol. 30, 1796–1804 (2012).
7. Hurvitz, S. A. et al. Neoadjuvant trastuzumab emtansine and pertuzumab in
human epidermal growth factor receptor 2-positive breast cancer: three-year
outcomes from the phase III KRISTINE study. J. Clin. Oncol. 37, 2206–2216 (2019).
8. Yee, D. et al. Association of event-free and distant recurrence-free survival with
individual-level pathologic complete response in neoadjuvant treatment of
stages 2 and 3 breast cancer: three-year follow-up analysis for the I-SPY2 adap-
tively randomized clinical trial. JAMA Oncol. 6, 1355–1362 (2020).
9. Rocque, G. B. et al. Patient perspectives on chemotherapy de-escalation in breast
cancer. Cancer Med. 10, 3288–3298 (2021).
10. Gianni, L. et al. Efficacy and safety of neoadjuvant pertuzumab and trastuzumab
inflammatory, or early HER2-positive breast
in women with locally advanced,
cancer (NeoSphere): a randomised multicentre, open-label, phase 2 trial. Lancet
Oncol. 13, 25–32 (2012).
11. Hurvitz, S. A. et al. Neoadjuvant trastuzumab, pertuzumab, and chemotherapy
versus trastuzumab emtansine plus pertuzumab in patients with HER2-positive
breast cancer (KRISTINE): a randomised, open-label, multicentre, phase 3 trial.
Lancet Oncol. 19, 115–126 (2018).
12. Schneeweiss, A. et al. Pertuzumab plus trastuzumab in combination with stan-
dard neoadjuvant anthracycline-containing and anthracycline-free chemother-
apy regimens in patients with HER2-positive early breast cancer: a randomized
phase II cardiac safety study (TRYPHAENA). Ann. Oncol. 24, 2278–2284 (2013).
13. Piccart, M. J. et al. Road map to safe and well-designed de-escalation trials of
systemic adjuvant therapy for solid tumors. J. Clin. Oncol. 38, 4120–4129 (2020).
14. Symmans, W. F. et al. Long-term prognostic risk after neoadjuvant chemotherapy
associated with residual cancer burden and breast cancer subtype. J. Clin. Oncol.
35, 1049–1060 (2017).
15. Early Breast Cancer Trialists’ Collaborative Group (EBCTCG). Increasing the dose
intensity of chemotherapy by more frequent administration or sequential
npj Breast Cancer (2022)
63
Published in partnership with the Breast Cancer Research Foundation
scheduling: a patient-level meta-analysis of 37 298 women with early breast
cancer in 26 randomised trials. Lancet 393, 1440–1452 (2019).
16. Wolff, A. C. et al. Human epidermal growth factor receptor 2 testing in breast
cancer: American Society of Clinical Oncology/College of American Pathologists
Clinical Practice Guideline Focused Update. J. Clin. Oncol. 36, 2105–2122 (2018).
17. Symmans, W. F. et al. Measurement of residual breast cancer burden to predict
survival after neoadjuvant chemotherapy. J. Clin. Oncol. 25, 4414–4422 (2007).
18. von Minckwitz, G. et al. Trastuzumab emtansine for residual
positive breast cancer. N. Engl. J. Med. 380, 617–628 (2019).
invasive HER2-
ACKNOWLEDGEMENTS
These data were presented in part at the San Antonio Breast Cancer Symposium
(December 2020). Abstract # PD3-05. Primary funding was provided by the Breast
Cancer Research Foundation (to E.P.W.). Additional funding was provided by the
Breast Cancer Research Foundation, the American Society of Clinical Oncology
Conquer Cancer Foundation, and the Terri Brodeur Breast Cancer Foundation (to A.G.
W.); Susan G. Komen (to E.A.M.).
AUTHOR CONTRIBUTIONS
A.G.W.: co-first author. Concept development; trial oversight; manuscript writing and
editing. N.V.D.: co-first author. Concept development; trial oversight; manuscript
writing and editing. T.L.: concept development, statistical analyses, manuscript
writing and editing. P.D.P.: patient accrual, manuscript writing and editing. A.H.P.:
patient accrual, manuscript writing and editing. N.S.: patient accrual, trial oversight,
manuscript writing and editing. L.M.S.: patient accrual, trial oversight, manuscript
writing and editing. M.F.: patient accrual, trial oversight, manuscript writing and
editing. M.C.:trial oversight, manuscript writing and editing. O.M.: patient accrual,
manuscript writing and editing. J.A.: data management, manuscript writing and
editing. J.D.: data management, manuscript writing and editing. SMR: concept
development, manuscript writing and editing. E.F.: manuscript writing and editing. S.
M.T.: concept development, manuscript writing and editing. I.E.K.: manuscript writing
and editing. N.M.T. manuscript writing and editing. N.T.: Statistical analyses,
manuscript writing and editing. T.A.K.: manuscript writing and editing. E.A.M.:
trial oversight,
manuscript writing and editing. E.P.W.: concept development,
manuscript writing and editing.
COMPETING INTERESTS
A.G.W.: institutional research support from Genentech, MacroGenics, and Merck. A.H.
P.: travel support from Novartis. LMS declares consulting fees from Novartis. O.M.:
receives institutional research funding from Abbvie, Genentech/Pfizer, and Roche;
honoraria from Roche. S.M.T.: receives institutional research funding from AstraZe-
neca, Lilly, Merck, Nektar, Novartis, Pfizer, Genentech/Roche, Immunomedics, Exelixis,
Bristol-Myers Squibb, Eisai, Nanostring, Sanofi, Cyclacel, Odonate, and Seattle
A.G. Waks et al.
7
Genetics; has served as an advisor/consultant to AstraZeneca, Lilly, Merck, Nektar,
Novartis, Pfizer, Genentech/Roche,
Immunomedics, Bristol-Myers Squibb, Eisai,
Nanostring, Puma, Sanofi, Celldex, Paxman, Silverback Therapeutics, G1 Therapeutics,
Gilead, AbbVie, Anthenex, OncoPep, Outcomes4Me, Kyowa Kirin Pharmaceuticals,
Daiichi-Sankyo, Ellipsis, and Samsung Bioepsis Inc. T.A.K.: speakers honoraria Exact
Sciences (formerly Genomic Health); faculty, PrecisCa cancer information services and
compensated service for a Global Advisory Board of Besins Healthcare. E.A.M.:
institutional research from Genentech/Roche via a SU2C grant; research funding from
to
Exact Sciences and Glaxo SmithKline; has served as an advisor/consultant
AstraZeneca, Bristol-Myers Squibb, Exact Sciences, Genentech/Roche, Lilly, Merck and
Sellas. E.P.W.: institutional research funding from Genentech/Roche; consultant for
Athenex, Carrick Therapeutics, G1 Therapeutics, Genentech/Roche, Genomic Health,
Gilead, GSK, Jounce, Lilly, Novartis, Seattle Genetics, Syros, and Zymeworks; scientific
advisory board member at Leap Therapeutics; and serves as President-Elect of the
American Society of Clinical Oncology (ASCO). All remaining authors have declared
no conflicts of interest.
ADDITIONAL INFORMATION
Supplementary information The online version contains supplementary material
available at https://doi.org/10.1038/s41523-022-00429-7.
Correspondence and requests for materials should be addressed to Eric P. Winer.
Reprints and permission information is available at http://www.nature.com/
reprints
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims
in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
material in this article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not included in the
article’s Creative Commons license and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly
from the copyright holder. To view a copy of this license, visit http://creativecommons.
org/licenses/by/4.0/.
© The Author(s) 2022
Published in partnership with the Breast Cancer Research Foundation
npj Breast Cancer (2022)
63
|
Could not heal snippet
|
10.1186_s12915-020-00859-4.pdf
|
Availability of data and materials
All data generated or analysed during this study are included in this
published article, its supplementary information files and publicly available
repositories.
|
Availability of data and materials All data generated or analysed during this study are included in this published article, its supplementary information files and publicly available
|
Burns et al. BMC Biology (2020) 18:133
https://doi.org/10.1186/s12915-020-00859-4
R E S E A R C H A R T I C L E
Open Access
Retargeting azithromycin analogues to
have dual-modality antimalarial activity
Amy L. Burns1, Brad E. Sleebs2,3, Ghizal Siddiqui4, Amanda E. De Paoli4, Dovile Anderson4, Benjamin Liffner1,
Richard Harvey1, James G. Beeson5,6,7, Darren J. Creek4, Christopher D. Goodman8, Geoffrey I. McFadden8 and
Danny W. Wilson1,5*
Abstract
Background: Resistance to front-line antimalarials (artemisinin combination therapies) is spreading, and
development of new drug treatment strategies to rapidly kill Plasmodium spp. malaria parasites is urgently needed.
Azithromycin is a clinically used macrolide antibiotic proposed as a partner drug for combination therapy in malaria,
which has also been tested as monotherapy. However, its slow-killing ‘delayed-death’ activity against the parasite’s
apicoplast organelle and suboptimal activity as monotherapy limit its application as a potential malaria treatment.
Here, we explore a panel of azithromycin analogues and demonstrate that chemical modifications can be used to
greatly improve the speed and potency of antimalarial action.
Results: Investigation of 84 azithromycin analogues revealed nanomolar quick-killing potency directed against the
very earliest stage of parasite development within red blood cells. Indeed, the best analogue exhibited 1600-fold
higher potency than azithromycin with less than 48 hrs treatment in vitro. Analogues were effective against
zoonotic Plasmodium knowlesi malaria parasites and against both multi-drug and artemisinin-resistant Plasmodium
falciparum lines. Metabolomic profiles of azithromycin analogue-treated parasites suggested activity in the parasite
food vacuole and mitochondria were disrupted. Moreover, unlike the food vacuole-targeting drug chloroquine,
azithromycin and analogues were active across blood-stage development, including merozoite invasion, suggesting
that these macrolides have a multi-factorial mechanism of quick-killing activity. The positioning of functional groups
added to azithromycin and its quick-killing analogues altered their activity against bacterial-like ribosomes but had
minimal change on ‘quick-killing’ activity. Apicoplast minus parasites remained susceptible to both azithromycin
and its analogues, further demonstrating that quick-killing is independent of apicoplast-targeting, delayed-death
activity.
Conclusion: We show that azithromycin and analogues can rapidly kill malaria parasite asexual blood stages via a
fast action mechanism. Development of azithromycin and analogues as antimalarials offers the possibility of
targeting parasites through both a quick-killing and delayed-death mechanism of action in a single, multifactorial
chemotype.
Keywords: Plasmodium, Malaria, Antimalarial, Macrolide
* Correspondence: [email protected]
1Research Centre for Infectious Diseases, School of Biological Sciences, The
University of Adelaide, Adelaide 5005, Australia
5Burnet Institute, Melbourne, Victoria 3004, Australia
Full list of author information is available at the end of the article
© The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if
changes were made. The images or other third party material in this article are included in the article's Creative Commons
licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons
licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain
permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the
data made available in this article, unless otherwise stated in a credit line to the data.
Burns et al. BMC Biology (2020) 18:133
Page 2 of 23
Background
Malaria is a mosquito-borne disease caused by proto-
the genus Plasmodium. In 2017,
zoan parasites of
there were ~ 219 million cases of malaria that resulted
in ~ 435,000 deaths [1, 2], with most deaths as the re-
sult of Plasmodium falciparum infection in children
under 5 years of age within sub-Saharan Africa. Current
control strategies include use of insecticide treated bed-
nets and indoor residual spraying, which target mosquito
transmission, chemoprophylaxis in high-risk groups, and
artemisinin combination therapies (ACTs) to both cure
patients and limit their transmission. Widespread use of
these control measures has resulted in significant de-
creases in malaria mortality over the past two decades [1,
2]. However, there is growing concern that artemisinin-
resistant P.
falciparum parasites may spread from the
Greater Mekong sub-region and Eastern India, where they
have previously been identified, and will lead to the loss of
our most effective drug treatments [3–6]. Furthermore,
there is also substantial resistance to some of the current
partner drugs used in ACTs, most notably piperaquine
and mefloquine [7]. Therefore, new antimalarials with
novel mechanisms of action that rapidly clear blood-stage
parasites are urgently needed [8, 9].
Clinically used macrolide antibiotics, in particular azi-
thromycin, have been proposed as partner drugs for
ACTs [10, 11]. Macrolide antibiotics have been shown
to target the malaria parasite’s remnant plastid (apico-
plast), which has a bacterium-like ribosomal complex es-
sential for protein translation and organelle biogenesis
[12–14]. The apicoplast is essential for synthesis of iso-
pentenyl pyrophosphate (IPP) precursors required for
protein prenylation, ubiquinone biosynthesis and doli-
chols required for N-glycosylation and production of
GPI anchors (reviewed in [15] and [16]). Indeed, IPP
synthesis is the sole essential function of the apicoplast
in blood stages, but apicoplast biogenesis and house-
keeping activity is essential for IPP production, making
the apicoplast ribosome an attractive antimalarial target
[13, 14, 17]. P. falciparum parasites treated with clinic-
ally relevant (nanomolar) concentrations of macrolide
antibiotics exhibit a ‘delayed-death’ phenotype in which
parasite growth is arrested during the second replication
cycle after treatment (~ 4 days post-treatment) [13, 14].
Azithromycin exhibits three favourable properties as
an antimalarial: a half-life > 50 hrs making it suitable for
infrequent dosing [18], good in vivo safety profile [19]
and high potency against P. falciparum in vitro [20, 21].
Azithromycin also shows efficacy as a prophylactic [22]
(reviewed in [23]), improved clinical outcomes in com-
bination with pyrimethamine during intermittent pre-
ventative treatment
for malaria in pregnancy (IPTp)
trials [24] and led to a significant decrease in P. falcip-
arum infections following mass drug administrations of
azithromycin monotherapy for trachoma infection [25].
Evidence also suggests that azithromycin inhibits the de-
velopment of mosquito transmissible parasites and liver
stages in rodent models [22, 26, 27]. However, when azi-
thromycin was trialled for treatment of clinical malaria,
it exhibited sub-optimal activity as a monotherapy and
was generally less effective than the similarly acting anti-
biotic clindamycin when used in combination with other
antimalarials [28]. Crucially, the delayed-death activity of
azithromycin has limited its use as a treatment
for
clinical disease. Currently, azithromycin is not used as a
these
first-line
considerations.
for malaria because of
treatment
We previously demonstrated that azithromycin can
also cause rapid parasite death when tested at higher
concentrations (IC50 ~ 10 μM) [27, 29]. Most strikingly,
azithromycin can rapidly inhibit P. falciparum merozoite
invasion of RBCs at
these higher concentrations. In
addition, azithromycin kills parasites within one intracel-
lular blood-stage lifecycle (from immediately post-
merozoite invasion to final schizont maturation at 48
hrs, in-cycle) at a similar IC50 as the drug’s invasion in-
hibitory activity. Testing of a small panel of azithromycin
analogues showed that these ‘quick-killing’ IC50s could
be enhanced through chemical modification. Import-
antly, parasites selected for resistance to azithromycin’s
delayed-death activity (120 hr post-invasion) remained
susceptible to both invasion-inhibition and intracellular
parasite quick-killing activities (invasion, in-cycle and 72
hr inhibition), indicating that azithromycin has a second-
ary, apicoplast-independent, mechanism of action [27,
29]. Therefore, chemical modification of azithromycin
presents a unique opportunity to develop a dual-acting
antimalarial with two independent mechanisms of action
that combines both quick-killing (for rapid clearance of
clinical infection) and delayed-death activities, providing
an element of resistance proofing and improving longer-
term protection from recrudescence or reinfection.
In this study, we screened 84 azithromycin analogues
and defined their efficacy against different stages of the
blood-stage lifecycle. A high proportion of analogues ex-
hibited improved quick-killing activity over azithromycin
against both P. falciparum and P. knowlesi, a model for
P. vivax and human pathogen of developing importance
in Southeast Asia [30], and were equally effective against
lacking an apicoplast. The
parasites containing or
analogues acted rapidly at inhibitory concentrations with
only short treatment times required to kill parasites
throughout
the
blood-stage
low cost of
established safety profile,
manufacture, and previous evaluation in ACTs, the re-
development of azithromycin-like compounds into an
antimalarial with dual mechanisms of action provides a
novel strategy to develop new antimalarials.
development. Given
life,
long-half
Burns et al. BMC Biology (2020) 18:133
Page 3 of 23
Results
Azithromycin analogues show improvement in quick-
killing activity against P. falciparum
We characterised the activity of 84 azithromycin ana-
logues across the malaria parasites asexual blood-stage
development in fine detail, including their activity against
early ring stages. The IC50 values for 72 hr growth-
inhibition assays (drug treatment assays represented in
Fig. 1; 1 cycle assay Fig. 1c) and their toxicity against
mammalian cells for analogues presented in this study
have been published previously [31–35]. Here, we tested
in-cycle assays,
for quick-killing activity using 44 hr
wherein 10 μM of drug was added to early ring-stage D10-
PfPHG parasites within a few hours of invasion and para-
site development quantified at late schizont stage with no
exposure of invading merozoites to the drug. This initial
screen identified 65 of 84 analogues that inhibited growth
by > 30% (Fig. 1b, Additional file 1: Tables S1a-c). The in-
cycle IC50 values for these 65 analogues were determined
(Additional file 1: Tables S1a-c) with all but two analogues
showing improved potency over azithromycin (azithromy-
cin IC50 with 44 hr in-cycle treatment, 11.3 μM) with the
most potent compound exhibiting a 1615-fold lower IC50
than azithromycin (GSK-66 IC50 0.007 μM). Notably, 39
analogues showed > 10-fold improvement over azithromy-
cin (IC50 < 1 μM), with 16 exhibiting a > 55-fold improve-
ment (IC50 < 0.2 μM). Summary inhibitory assay data and
structure for 19 of the most potent analogues featuring
different added functional groups is available in Table 1
and Fig. 2. Published cytotoxicity data against mammalian
cells is available for 13 of the most potent analogues [31,
33–35] with the IC50 against the HepG2 cell line ranging
between 3 and 83 μM and the selectivity index (SI; IC50
against HepG2/44 hr D10-PfPHG IC50 from this study)
ranging between 15 to 415 fold. Eleven of these analogues
had a SI > 50, indicating low mammalian cell toxicity.
The analogues with the low nanomolar 44 hr in-cycle
activity often featured quinoline or chloroquinoline modi-
fications (Table 1, Fig. 2, Additional file 1: Tables S1a-c).
However, there were exceptions including a number of
phenyl-substituted analogues (GSK-5, GSK-6, GSK-9,
GSK-11, GSK-14, GSK-16, GSK-17, GSK-19)
and
naphthalene-substituted analogues (GSK-3, GSK-4, GSK-
15, GSK-18), which all displayed IC50 values < 1 μM.
There was no structural difference between the most po-
tent analogues and the analogues with activity > 1 μM that
could explain the observed activity discrepancy. Consist-
ently, chloroquinoline analogues (GSK-1, GSK-2, GSK-56
and GSK-66) were more potent than their respective
unsubstituted quinoline counterparts (GSK-7, GSK-10,
GSK-58 and GSK-71). Analogues GSK-6 and GSK-9 with
thiourea aryl substitution displayed comparable potency
(IC50 0.2 and 0.44 μM) to naphthalene analogues GSK-3
and GSK-4 (IC50 0.18 and 0.19 μM). However, a large
Fig. 1. Schematic of drug treatment regimens outlining the times of treatment and stage/time of parasitaemia measurement for assays used in
this study. a Merozoite invasion of RBCs: Merozoites were drug treated prior to addition of RBCs. RBC invasion was measured at early ring stages
(< 1 hr rings). b In-cycle: highly synchronous, early ring-stage parasites (0–4 hrs post-invasion) were treated with drug, with the resulting growth
inhibition analysed at schizont stage (44 hrs post-invasion for P. falciparum and 26 hrs for P. knowlesi). c One cycle (0–72 hrs): highly synchronous,
early ring-stage parasites (0–4 hrs post-invasion) were drug-treated and the resulting growth inhibition was measured after ~ 72 hrs of growth,
post one cycle of re-invasion, at schizont stages. d 2 cycle (delayed death); highly synchronous, early ring-stage parasites (0–4 hrs post-invasion)
were drug-treated and allowed to grow for 92 hrs before washing drug with fresh media (post second invasion cycle). Growth inhibition was
assessed approximately 30 hrs later, at schizont stages (0–120 hrs post-invasion for P. falciparum and 0–92 hrs for P. knowlesi)
Burns et al. BMC Biology (2020) 18:133
Page 4 of 23
Table 1 In vitro efficacy of antimalarials and azithromycin analogues against Plasmodium spp. parasites
cmpnd R3 class
R4 class
R5 class
In-cycle (44 hr)
growth DD2 IC50
(μMb, ±SEM)
Invasion inhibition
D10-PfPHG IC50
(μMc, ±SEM)
AZR
Me
Me
CQ
QN
DHA
1
56
59
66
69
70
72
8
10
58
71
73
3
4
15
5
6
9
17
H
H
Chloroquinoline Me
Chloroquinoline H
Chloroquinoline H
Me
Me
Me
Me
Me
Me
Quinoline
Quinoline
Me
Me
Me
Naphthalene
Naphthalene
Naphthalene
Substituted
phenyl thiourea
Substituted
phenyl thiourea
Substituted
phenyl thiourea
Substituted
phenyl thiourea
Me
Me
Me
Me
Me
Me
Quinoline
Me
Me
Me
Me
Me
Me
Me
Me
Me
In-cycle (44 hr)
growth D10-PfPHG
IC50 (μMa, ±SEM)
11.31 (0.49)
0.052 (0.006)
0.39 (0.07)
0.0008 (0.0001)
0.019 (0.004)
0.011 (0.002)
0.073 (0.02)
15.6 (2.1)
0.31 (0.31)
ND
ND
0.082 (0.02)
0.093 (0.02)
0.049 (0.005)
0.043 (0.002)
ND
ND
Chloroquinoline 0.007 (0.001)
Chloroquinoline 0.031 (0.004)
Chloroquinoline 0.05 (0.006)
Chloroquinoline 0.27 (0.01)
0.065 (0.004)
H
H
H
Quinoline
Quinoline
H
H
H
H
H
H
H
0.41 (0.02)
0.48 (0.04)
0.048 (0.004)
0.053 (0.005)
0.31 (0.02)
0.183 (0.02)
0.19 (0.01)
0.67 (0.07)
0.2 (0.01)
0.52 (0.1)
0.748 (0.1)
0.056 (0.01)
0.16 (0.02)
0.48 (0.2)
0.32 (0.07)
ND
0.4 (0.1)
0.4 (0.05)
0.28 (0.05)
0.27 (0.07)
0.44 (0.07)
0.24 (0.04)
0.7 (0.05)
0.54 (0.06)
10 (1.4)
ND
ND
ND
ND
3.2 (0.39)
ND
ND
ND
ND
1.7 (0.02)
4.4 (1.2)
ND
ND
ND
ND
1.8 (0.5)
2.0 (0.2)
3.6 (0.4)
1.61 (0.02)
ND
ND
ND
In-cycle (24 hr)
growth PkYH1 IC50
(μMd, ±SEM)
16 (1.8)
0.017 (0.005)
ND
0.0024 (0.001)
0.2 (0.005)
0.031 (0.008)
ND
0.012 (0.002)
ND
ND
0.15 (0.06)
0.15 (0.01)
0.1 (0.005)
0.071 (0.013)
0.041 (0.005)
0.248 (0.07)
0.095 (0.02)
ND
0.32 (0.12)
0.082 (0.02)
0.16 (0.03)
0.016 (0.005)
0.36 (0.01)
a Drug treatment of intracellular growth, from rings to late schizonts, with no rupture cycle for D10-PfPHG (P. falciparum, 0–44 hrs). Data represents the mean of 3
or more experiments
b Drug treatment of intracellular growth, from rings to late schizonts, with no rupture cycle for DD2 (P. falciparum, 0–44 hrs). Data represents the mean of 3 or
more experiments
c Drug treatment of D10-PfPHG merozoites prior to addition of RBCs. Parasitemia was measured by flow cytometry ~ 30 min post invasion. Data represents the
mean of 2 (for compounds 4 and 5) or 3 experiments
d Drug treatment of intracellular growth, from rings to late schizonts, with no rupture cycle for P. knowlesi YH1 (P. knowlesi, 0–24 h). Data represents the mean of 2
or more experiments
number of analogues supporting thiourea and urea aryl
substitutions were significantly less active, with no clear
distinction between the activity and substitution pattern
on the
thiourea- or urea-substituted
analogues.
ring of
aryl
Analogues with aliphatic substitution on the urea or
thiourea (GSK-31, GSK-35, GSK-38, GSK-45, GSK-47,
GSK-51) generally had reduced activity compared to ana-
logues with pendant aryl moieties (Table 1, Fig. 2, Add-
itional file 1: Tables S1a-c), suggesting the aryl substituent
was important for modulating potency. Consistent with
this observation, analogues that did not terminate with an
aromatic substituent and were only decorated with small
aliphatic functionality (analogues GSK-34, GSK-44, GSK-
46, GSK-50, GSK-52, GSK-53, GSK-54, GSK-55, GSK-62,
GSK-64, GSK-83, GSK-84) were either weakly active
(> 3 μM) or inactive. These data suggested that the
type of functionality and the length of the carbon-chain
linking the aromatic group to the macrolactone was not
important for activity. However, analogues GSK-56, GSK-
57, GSK-58, GSK-66, GSK-67 and GSK-71, with short 3-
carbon linkers between the macrolactone and the
quinoline group, were amongst the most potent. Overall,
there was no consistent trend between the type of func-
tionality and the length of the carbon-chain linking the
aromatic group to the macrolactone.
The position of the pendant quinoline or aromatic
system attached to the macrolactone—either N6-, O-
Burns et al. BMC Biology (2020) 18:133
Page 5 of 23
N R4
X
X
n
R3
N
HO
OH
O
O
OH
OR1
OR2
R5O
O
Desosaminyl
O
OH
O
Cladinosyl
R
R
N
quinoline
X
X
n
naphthalene
H
N
H
N
n
X
substituted phenyl urea
R
Modifications
Azithromycin
R1 = desosaminyl, R2 = cladinosyl, R3 and R4 = CH3, R5 = H
Analogues in this study
R1 = desosaminyl, R2 = cladinosyl or H, R3, R 4,
R5 = H, Me or modifications
Fig. 2. Structure of azithromycin and analogues. Outline of the structure of the parent molecule azithromycin, structural side-chains and sites of
attachment of functional groups (R1–5) for compounds shown in Table 1. Structure of functional groups added is listed in Table 1
desosaminyl or N-desosaminyl—did not affect the in-
cycle 44 hr activity of analogues (Table 1, Fig. 2, Add-
itional file 1: Tables S1a-c). For example, analogues with
the same quinoline functionality, GSK-1, GSK-56 and
GSK-66 attached to either N6-, N-desosaminyl or O-
desosaminyl positions, displayed similar IC50 values be-
tween 7 and 19 nM. This trend was observed amongst
other analogues for which there were matched pairs.
The cladinosyl group did not affect 44 hr in-cycle activ-
ity, for example respective analogues with the cladinosyl
group, GSK-1, GSK-10, GSK-56 and GSK-66, possessed
similar activity compared to analogues without the cladi-
nosyl group, GSK-67, GSK-7 and GSK-57. This observa-
tion is consistent with our previous findings on the
azalide structure activity relationship [29].
Azithromycin analogues show improved activity against
merozoite RBC invasion
We previously showed that azithromycin and analogues
inhibit merozoite invasion, with merozoites found to con-
tact and briefly deform the RBC membrane, and then de-
tach when examined in the presence of azithromycin [29].
We investigated whether the 39 analogues that had an in-
cycle (44 hr) IC50 < 1 μM could inhibit merozoite invasion
at a concentration of 1 μM and identified eight analogues
that inhibited invasion by > 20% at 1 μM (Fig. 1a, Table 1,
Fig. 2, Additional file 1: Tables S1a-c). The invasion inhibi-
tory IC50 for seven of these analogues with sufficient avail-
able sample were determined; there was a 2- to 6-fold
reduction in the invasion inhibitory IC50 over azithromy-
cin (range GSK-8 4.4 μM to GSK-5 1.6 μM) (Table 1,
Additional file 2: Figure S1). Importantly, azithromycin
analogues with improved in-cycle activity also had im-
proved potency against merozoite invasion, confirming
previous observations that both invasion and in-cycle
quick-killing activities can be improved with a single
chemical modification [29]. We next tested whether azi-
thromycin analogue invasion inhibitory activity was di-
rected against
treating purified
the merozoite by
merozoites with 10 μM of GSK-72 (invasion inhibitory
IC50 1.7 μM), followed by washing drug off the merozoites,
and then mixing merozoites with RBCs (Additional file 3
Figure S2). GSK-72-treated merozoites were stopped from
invading RBCs after washing off the drug, suggesting that
the invasion inhibitory activity of azithromycin analogues
is irreversible and directed towards the merozoite.
Quick-killing activity is independent of apicoplast
targeting
We previously showed that quick-killing activity is main-
tained against delayed-death-resistant parasites [29], sug-
gesting that quick-killing occurs through a mechanism of
Burns et al. BMC Biology (2020) 18:133
Page 6 of 23
action independent of the apicoplast. However, the fact
that the apicoplast and apicoplast-ribosome were still
present in these drug-treated parasites left open the possi-
bility that quick-killing activity could still be linked to the
apicoplast [36]. To confirm quick-killing is completely in-
dependent of
the apicoplast, we generated apicoplast
minus (PfPHGapicoplast-null) parasites through prolonged
treatment with azithromycin and then rescued with media
supplementation with the isoprenoid precursor, isopente-
nyl pyrophosphate (IPP) [17, 36]. PfPHGapicoplast-null para-
sites showed a complete loss of sensitivity to azithromycin
in 120 hr delayed-death assays, confirming that the apico-
plast had been removed (Additional file 4: Figure S3a) [17,
36]. In contrast, there was no difference in growth inhib-
ition for the PfPHGapicoplast-null and PfPHGwildtype parasites
when treated with azithromycin (Additional
file 4:
Figure S3b) and 15 lead analogues at the in-cycle D10-
PfPHGwildtype IC90 concentration for 44 hrs (Additional file
4: Figure S3c; Additional file 1: Table S1a-b). These data
confirm that quick-killing activity is independent of the
apicoplast, indicating that there is a secondary mechanism
of action for azithromycin and analogues.
Azithromycin is a rapid and irreversible inhibitor across
blood-stage parasite growth
After confirming that azithromycin and analogues have
both invasion (Table 1, Additional file 2: Figure S1) and
intracellular (Fig. 3a) blood-stage quick-killing activity that
is independent of apicoplast-targeting delayed death (Add-
itional file 4: Figure S3a-c), we next determined drug ac-
tivity across early rings (0–12 hrs post invasion), early
trophozoites (12–24 hrs post invasion), late trophozoites
(24–36 hrs post invasion) and schizonts (36–44 hrs post
invasion). Azithromycin demonstrated a similar
IC50
across each pulsed treatment stage (0–12 hr IC50 14 μM,
12–24 hr IC50 16 μM, 24–36 hr IC50 15 μM) with these
values similar to the IC50 values obtained for 44 hr (IC50
11.3 μM) and invasion inhibition (IC50 10 μM) treatments
(Fig. 3b, c). We confirmed that azithromycin’s quick-
killing activity works rapidly by assessing the morpho-
logical effects of pulsed treatment with a 2× IC90 drug
concentration. Ring-stage treatments (0–12 hrs) showed
pronounced vacuolation of the cytoplasm, a typical sign of
parasite stress. Trophozoite stages (12–24 hrs and 24–36
hrs) appeared either pyknotic or severely vacuolated with
indicative of rapid cell death
only a 12-hr treatment,
(Fig. 3b, e). Although azithromycin treated schizont stages
(36–44 hrs post-invasion) did not show potent growth in-
hibitory activity when assessed by flow cytometry, light
microscopy smears showed late-stage parasites with severe
vacuolation and minimal merozoite maturation, indicating
this population was indeed killed by azithromycin treat-
ment (Fig. 3e). These data, together with our earlier data,
provide direct evidence that azithromycin acts broadly
across invasion and throughout the entire blood-stage life-
cycle, including early ring stages.
Azithromycin and analogues rapidly kill early ring-stage
parasites
Our finding that azithromycin could kill ring-stage para-
sites (0–12 hrs post invasion) with similar efficacy to 44
hrs of drug treatment is of major interest since the ma-
jority of clinically used antimalarials, with the notable
exception of the artemisinins [37, 38], have relatively
poor activity against newly invaded ring stages [39–42].
early
To provide further insights into how quickly azithro-
mycin and analogues act against early ring stages, we ex-
amined activity of 6- and 12-hr treatments of early ring
stages (0–6 hrs and 0–12 hrs post-invasion treatments)
for azithromycin and a panel of diverse analogues that
had activity at nanomolar concentrations in parallel. Azi-
thromycin and the analogues tested showed < 2-fold re-
duction in potency with a 6-hr
ring-stage
treatment compared to a 12-hr ring-stage or full 1 cycle
(44 hr) treatment, highlighting the drug efficacy against
early ring stages (Fig. 3d, e, Fig. 4a, b, Additional file 5:
Table S2). Consistent with previous publications, dihy-
droartemisinin (DHA)
resulted in severe
growth retardation with early ring-stage treatment [37–
39]. DHA is considered to be one of the few clinically
used antimalarials with reasonable efficacy against early
ring-stage parasites [37–39], making the ability of azi-
thromycin and analogues to also cause rapid death of
these stages a promising finding. In contrast, chloro-
quine had comparatively poor activity for early ring-
stage treatments, which is as expected since chloroquine
is known to lack potency against ring-stage parasites.
treatment
Microscopy analysis was performed for parasites
treated with a 2× IC90 (0–44 hrs) of azithromycin and
analogues to examine the phenotypic changes associated
with early ring-stage drug treatment (Fig. 4b). Early (0–
6 hrs) ring stages treated with azithromycin GSK-66 and
GSK-3 exhibited vacuolation, with evidence of pyknotic
cells developing with extended treatment for GSK-71
and GSK-3 (0–12 hrs). Notably, GSK-5 resulted in a
large number of pyknotic parasites within only 6 hrs of
drug treatment, highlighting the speed with which these
compounds can act. DHA treatment of early (0–6 hrs)
ring stages did not lead to a clear change in parasite
morphology. However, after extended ring-stage treat-
ment (0–12 hrs) pyknotic cells became prominent. No
aberrant growth phenotype was observed with chloro-
quine with treatment of early ring stages (0–6 hrs), with
evidence of vacuolation only occurring after extended
ring-stage treatment (0–12 hrs). Short-term pulse treat-
ments confirmed that azithromycin and analogues rap-
idly kill early ring-stage parasites, the growth inhibitory
of
effects
and modification
reversible,
not
are
Burns et al. BMC Biology (2020) 18:133
Page 7 of 23
A
)
l
o
r
t
n
o
C
%
(
h
t
w
o
r
G
120
100
80
60
40
20
0
C
)
l
o
r
t
n
o
C
%
(
h
t
w
o
r
G
120
100
80
60
40
20
B
0-44 hrs
0-72 hrs
0-120 hrs
0-6 hrs 0-12 hrs 12-24 hrs 24-36 hrs 36-48 hrs
0.01 0.1 1
10 100
Azithromycin (µM)
0-12 hrs
12-24 hrs
24-36 hrs
36-44 hrs
0
0.1
1
100
Azithromycin (µM)
10
D
)
l
o
r
t
n
o
C
%
(
h
t
w
o
r
G
120
100
80
60
40
20
0
0-6 hrs
0-12 hrs
0-44 hrs
0.01 0.1 1
10 100
Azithromycin (µM)
Fig. 3. Azithromycin has broad activity against blood-stage parasites. a Early ring-stage P. falciparum parasites (0–4 hrs post-invasion) were treated
with doubling dilutions of azithromycin and inhibition of growth measured for in-cycle (44 hr, IC50, 11 μM), 1-cycle (72 hr, IC50, 14 μM) and 2-cycle
(delayed death, 120 hr, IC50, 0.07 μM) assays (44 hr vs 72 hr, P=NS; 120 hr vs 44 hr P < 0.0001; 120 hr vs 72 hr P < 0.0001). b Schematic of drug
washout treatment scheme to assess azithromycin’s quick-killing stage of activity. Early ring-stage parasites (0–4 hrs post-invasion) were aliquoted
to a 96-well plate and doubling dilutions of azithromycin added between 0-12 hrs, 12–24 hrs, 24–36 hrs and 36–44 hrs post invasion prior to drug
removal by washing with fresh media. c Growth inhibition of azithromycin across 0–12 hrs, 12–24 hrs, 24–36 hrs and 36–44 hrs post invasion prior
to drug removal by washing with fresh media. There was no significance between treatment times for 0–12 hrs, 12–24 hrs, 24–36 hrs, but there
was for 0–12 hrs vs 36–44 hrs (P = 0.005), 12–24 hrs vs 36–44 hrs (P = 0.01) and 24–36 hrs vs 36–44 hrs (P = 0.01). d Growth inhibition of
azithromycin with very early ring-stage treatment across 0–6 hrs and 0–12 hrs post-invasion compared to a full in-cycle (0–44 hr) treatment.
Treatments showed significant difference (P < 0.0001) with the exception of 0–12 hrs vs 0–44 hrs (P = 0.19). For all growth curves, parasitemia was
measured at 44 hrs post invasion at schizont stage via flow cytometry. Data represents the means of 3 or more experiments expressed as a
percentage of non-inhibitory control and error bars represent ± SEM. Dose response IC50s compared using extra sum of squares F-test. Repeat
measure data is available in Additional file 15 Supporting Value Data. e Representative Giemsa-stained thin blood smears showing the growth
phenotypes seen for non-inhibitory media controls (top panels) and in the presence of 2× IC90 concentration of azithromycin (bottom panels)
across different stages of intraerythrocytic blood-stage development (0–6 hrs, 0–12 hrs, 12–24 hrs, 24–36 hrs and 36–44 hrs)
Burns et al. BMC Biology (2020) 18:133
Page 8 of 23
Fig. 4. (See legend on next page.)
Burns et al. BMC Biology (2020) 18:133
Page 9 of 23
(See figure on previous page.)
Fig. 4. Growth inhibition profiles of azithromycin analogues and control drugs with short-term and in-cycle drug treatments. a Early ring-stage P.
falciparum parasites (0–4 hrs post-invasion) were treated with doubling dilutions of azithromycin analogues/control drugs for 0–6 hrs and 0–12 hrs
prior to washing the drug out of cultures allowing growth to continue until parasites were 44 hrs old. A 0–44 hr continuous drug control
treatment was also included. a Growth inhibition profile of GSK-3 (naphthalene), GSK-5 (substituted phenyl), GSK-66 (chloroquinoline), GSK-71
(quinoline), dihydroartemisinin (DHA) and chloroquine with very early ring-stage treatment across 0–6 hrs and 0–12 hrs post-invasion compared
to a full in-cycle treatment. There was no significant difference in drug efficacy between the treatment times of GSK-5 or GSK-71 (P > 0.01). GSK-
66 showed a significant difference between 0-6 hr vs 0–12 hr treatments (P < 0.0079) and 0–6 hr vs 0–44 hr (P = 0.001), but there was no
significant difference in drug efficacy between 0-12 hr vs 0–44 hr treatments (P = 0.96). GSK-3 and DHA showed no significant difference in
efficacy between treatment times (P > 0.01), with the exception of 0–6 hr vs 0–44 hr (P = 0.005 and P = 0.01, respectively). In contrast, chloroquine
demonstrated a significant difference in drug efficacy between all treatment times (0–6 hr vs 0–12 hr P < 0.0001; 0–6 hr vs 0–44 hr P < 0.0001; 0–
12 hr vs 0–44 hr P < 0.0001). Parasitemia was measured via flow cytometry 44 hrs post-invasion. Data represents the means of 3 or more
experiments expressed as a percentage of non-inhibitory control and error bars represent ± SEM. Dose response IC50s compared using extra sum
of squares F-test. Repeat measure data is available in Additional file 15 Supporting Value Data. b Representative Giemsa-stained thin blood smears
showing the growth phenotypes seen for non-inhibitory media controls, and treatment with 2× IC90 of azithromycin analogues GSK-3 (0.74 μM),
GSK-5 (0.62 μM), GSK-66 (0.034 μM), GSK-71 (0.18 μM) and control drugs DHA (0.003 μM) and chloroquine (0.222 μM) (bottom panels) 0–6 hrs post
treatment and 0–12 hrs post treatment
azithromycin can produce analogues with broad and po-
tent efficacy across blood-stage parasite growth.
the
against
Quick-killing azithromycin analogues maintain activity
against drug-resistant P. falciparum and P. knowlesi
We next investigated whether analogues retained po-
chloroquine/mefloquine/pyrimeth-
tency
falciparum DD2 line [43], and an
amine-resistant P.
artemisinin-resistant P.
falciparum Cambodian isolate
[44–46] (Table 1). Relative to the chloroquine sensitive
D10-PfPHG line, DD2 parasites exhibited a 0.24- to 8.4-
fold loss of sensitivity to azithromycin and analogues. Of
note, analogues
featuring a chloroquinoline moiety
(GSK-1, GSK-56, GSK-66, GSK-72) were 4.77-fold less
chloroquine-resistant DD2, whereas
potent
and
quinoline-,
analogues
phenyl-substituted moieties were on average 1.35-fold
less
(Table 1, Add-
itional file 6: Table S3).
sensitive (n = 11 compounds)
naphthalene-
featuring
against
We next
tested the efficacy of azithromycin ana-
logues against the P.
falciparum artemisinin-resistant
clinical isolate Cam3.II, which has a mutation within
the Kelch13 (PF3D7_1343700) propeller gene (R539T,
Cam3.IIDHA resistant(R539T)) associated with increased
early ring-stage (0–3 hrs) survival in vitro with DHA treat-
ment [44–46]. Early ring-stage Cam3.IIDHA resistant(R539T)
resistant and a reverted sensitive line (Cam3.IIsensitive)
were pulsed for 4 hrs before the drug was washed off, with
growth determined 66 hrs later via flow cytometry [46,
47]. Since comparison of IC50 has limited relevance in
ring-stage survival assays, we compared instead the per-
centage (%) parasite growth of Cam3.IIDHA resistant(R539T)
parasites at the drug concentration that inhibited 95% of
growth for the Cam3.IIsensitive line. As expected, ~ 41%
Cam3.IIDHA resistant(R539T) parasites survived DHA treat-
ment
killed 95% of
concentration that
Cam3.IIsensitive parasites (Fig. 5, Table 2). In contrast,
the
at
growth of both the Cam3.IIDHA resistant(R539T) and the
Cam3.IIsensitive lines were equally inhibited at the con-
centration that killed 95% of DHA-sensitive parasites
for azithromycin and analogues GSK-56, GSK-71,
GSK-3 and GSK-5.
for
Of note, the IC50 of 4 hr ring-stage treatments ob-
served for the Cam3.IIsensitive line was similar to that of
6 hr ring-stage treatment seen in D10-PfPHG line upon
treatment of azithromycin (Cam3.IIsensitive IC50 31 μM,
D10-PfPHG IC50 30 μM) and GSK-5 (Cam3.IIsensitive
IC50 0.20 μM, D10-PfPHG IC50 0.3 μM). Furthermore,
activity against early ring-stage Cam3.IIsensitive parasites
was also similar to the in-cycle (44 hr) treatment activity
against D10-PfPHG parasites
azithromycin
(Cam3.IIsensitive IC50 31 μM, D10-PfPHG IC50 14 μM),
GSK-5 (Cam3.IIsensitive IC50 0.20 μM, D10-PfPHG IC50
0.26 μM), GSK-56 (Cam3.IIsensitive IC50 0.006 μM, D10-
PfPHG IC50 0.010 μM), GSK-58 (Cam3.IIsensitive
IC50
0.075 μM, D10-PfPHG IC50 0.048 μM) and GSK-4
0.04 μM, D10-PfPHG IC50
(Cam3.IIsensitive
0.19 μM). Despite the much more stringent drug wash-
out procedure employed for the Cam3.IIsensitive ring-
stage survival assays, activity against early ring stages
was equivalent to that seen for 6 hr treatment of D10-
PfPHG ring stages and similar to in-cycle treatments of
D10-PfPHG. These results support that azithromycin
and analogues have rapid activity against early ring-stage
parasites of different P. falciparum lines.
IC50
We next tested the activity of azithromycin and ana-
logues against the zoonotic malaria parasite P. knowlesi,
which is a significant human pathogen in regions of
Southeast Asia [48] and an in vitro culturable model for
P. vivax [49]. We found that azithromycin maintains po-
tency against P. knowlesi in both in-cycle (28 hr for P.
knowlesi, Pk) and delayed-death (92 hr) assays compared
to P. falciparum (Pf) (Pk in-cycle IC50 13 μM, delayed-
death IC50 0.08 μM, Pf in-cycle IC50 11.3 μM, delayed-
Burns et al. BMC Biology (2020) 18:133
Page 10 of 23
Fig. 5. Activity of azithromycin analogues against artemisinin-resistant parasites. Lead azithromycin analogues were tested against artemisinin-
resistant Cam3.IIDHA resistant(R539T) parasites containing the K13 propeller mutation and reverted, artemisinin-sensitive, Cam3.IIsensitive parasites in
ring-stage survival assays (4 hr drug pulse of very early rings 0–3 hrs post invasion) prior to washing off drug and assessment of parasitaemia (66
hrs later by flow cytometry). Dihydroartemisinin (DHA), azithromycin, GSK-3 (naphthalene), GSK-5 (substituted phenyl), GSK-56 (chloroquinoline)
and GSK-71 (quinoline). Parasitemia was measured via flow cytometry ~ 72 hrs post-invasion. Data represents the mean of 2 or more experiments
expressed as a percentage of non-inhibitory control and error bars represent ± SEM. Repeat measure data is available in Additional file 15
Supporting Value Data
Table 2 Ring-stage survival assay percent survival values from drug treated artemisinin-resistant and artemisinin-sensitive parasites
Modification
Compound
DHA
Cam3.IIsensitive 72 hr
growth IC50 (μM, ±SEM)
0.007 (0.002)
Cam3.IIresistant 72 hr
growth IC50 (μM, ±SEM)
0.011 (0.001)
Azithromycin 30 (5.5)
0.035 (0.004)
0.2 (0.02)
0.004 (0.001)
Naphthalene
2-Chlorophenyl
7-
Chloroquinoline
Quinolone
3
5
56
71
30 (0.005)
0.04 (0.004)
0.28 (0.005)
0.009 (0.001)
0.07 (0.007)
0.15 (0.03)
Concentration of drug = 5%
growth of Cam3.IIsensitive (μM)
Growth
Cam3.IIDHA resistant(R539T) (%)
0.05
100
0.4
0.5
0.055
0.6
41
1
8
6
7
6
The μM concentration of drug (DHA, azithromycin, GSK-3, GSK-5, GSK-56 and GSK-71) that resulted in a 5% survival value for artemisinin-sensitive Cam3.IIsensitive
parasites was then used to treat artemisinin-resistant Cam3.IIDHA resistant(R539T) parasites, and the resulting % parasite survival value for the resistant parasites is
displayed in the table. The IC50 value of the drugs against Cam3.IIsensitive and Cam3.IIresistant strains is also shown to indicate their overall potency against
artemisinin-sensitive and artemisinin-resistant parasites. Parasites were incubated for one cycle (72 hrs) after pulsed drug treatment and washing prior to
measurement of parasitaemia by flow cytometry. Data represent the mean of 2 experiments
Burns et al. BMC Biology (2020) 18:133
Page 11 of 23
this divergent parasite species
death IC50 0.07 μM) (Table 1) as previously shown [50].
We next tested a panel of azithromycin analogues that
had potent quick-killing activity against P.
falciparum
for their efficacy against P. knowlesi and identified that
the majority of analogues had similar quick-killing po-
tency against
(Add-
itional file 7: Table S4). Of interest, the analogue GSK-9
exhibited a significant 33.1-fold improvement in activity
against P. knowlesi when compared to activity against P.
falciparum, suggesting that some species-specific differ-
ences in drug activity can occur. Together, these data
efficacy
that azithromycin analogues have
support
against diverse human malaria parasites and across DHA
and multi-drug-resistant parasites.
bacteria
pneumoniae
Analogues modified at the macrolactone-ring maintain
dual mechanisms of action
We next sought to define whether the more potent quick-
killing azithromycin analogues maintained apicoplast-
targeting delayed-death activity. As quick-killing IC50s for a
number of analogues (GSK-1, GSK-4, GSK-5, GSK-29,
GSK-57, GSK-66, GSK-71, GSK-78) approached that of the
delayed-death IC50 values of azithromycin (120 hr IC50
0.07 μM), the measurement of apicoplast targeting delayed-
death activity (i.e. activity after 120 hrs of treatment, Fig. 1d)
would likely be compromised by quick-killing potency.
Therefore, we assessed the activity of azithromycin and a
panel of quick-killing analogues against the azithromycin-
Streptococcus
(Add-
sensitive
itional file 8: Table S5) on the basis that this Gram-positive
bacteria’s ribosome could serve as a proxy for the malaria
parasite bacterium-like apicoplast ribosome [12, 51]. Con-
sistent with previously published results, limited inhibition
of bacterial growth was observed for analogues with an N-
substitution on the desosamine sugar moiety [34, 35, 52].
Indeed, N-substituted analogues of azithromycin have been
deliberately designed to reduce off-target drug activity
against bacteria for use in alternative drug applications [34,
35, 52]. In contrast, all analogues with N6-substitutions on
the macrolactone backbone (GSK-1, GSK-4, GSK-5, GSK-
6, GSK-9, GSK-11, GSK-12, GSK-16, GSK-17, GSK-21,
GSK-25) had activity against S. pneumoniae similar to azi-
thromycin. Thus, selecting the site of azithromycin modifi-
cation can allow improved quick-killing activity while
maintaining apicoplast targeting delayed-death activity, or
delayed-death activity can be removed along with off-target
antibacterial effects to produce a quick-killing specific
antimalarial.
Analysis of the quick-killing mechanism of action
suggests a multi-factorial mechanism of action
In an attempt to identify the molecular target of quick-
killing activity, we selected for in vitro drug resistance by
subjecting an azithromycin delayed-death-resistant D10
line (D10-AZRr) with a stepwise increase [12] of the
quick-killing azithromycin analogue GSK-59 featuring a
chloroquinoline-substituted desosamine moiety
that
lacks delayed-death activity. After three attempts, we
failed to select for resistant parasites > 3 months after
drug removal, suggesting that the mechanism of quick-
killing cannot be readily selected for in vitro.
We next undertook an untargeted metabolomics
screen to identify changes in the metabolomic signature
of azithromycin and the quick-killing analogues GSK-5
(substituted phenyl), GSK-66 (chloroquinoline) and
GSK-71 (quinoline) and to compare changes during
treatment with these analogues to known antimalarials,
such as chloroquine and DHA (Fig. 6, Additional file 9:
Figure S4, Additional file 10: Table S6, Additional file
11: Table S7, Additional file 12: Table S8). Following a
2 hr treatment of trophozoite-stage parasites at a 5×
IC50 (44 hr) concentration, supervised multivariate ana-
lysis (partial least squares-discriminate analysis) and heat
map showed that the most prominent metabolomic sig-
nature shared between azithromycin and analogues was
a series of short peptides that were increased for all of
azithromycin, GSK-71, GSK-5 and the food vacuole-
targeting control drug chloroquine
(Fig. 6, Add-
itional file 10: Table S6a&b). Since increases in these
peptides have previously been demonstrated for chloro-
quine- and piperaquine-treated trophozoites [53],
it is
possible that this signature indicates a mechanism of ac-
tion similar to the 4-aminoquinolines, which are thought
to act by inhibiting crystallisation of haemoglobin-
derived haem to form haemozoin within the parasite’s
food vacuole. However, it was also noted in the study by
Creek et al. that the sequences for the majority of these
peptides are not derived from degraded haemoglobin, in-
dicating that the metabolomic signature shared between
chloroquine, azithromycin, GSK-71 and GSK-5 are likely
due to disruption of proteolytic processes other than
haemoglobin digestion. In addition, GSK-66 which has
the most chloroquine-like functional group in terms of
structure and was the most potent analogue tested in
this study, showed little in the way of changed metabo-
lites and gave a profile most similar to untreated control.
Since chloroquine is known to disrupt the haemoglobin
digestion pathway by inhibition of haemozoin formation
[54–56], we next measured the levels of haemoglobin,
haem and haemozoin in the parasites following treat-
ment with analogues GSK-66 (chloroquinoline) and
GSK-71 (quinoline) [57] (Fig. 7). Trophozoite-stage par-
asites were treated with CQ, GSK-71, and GSK-66 at
10× IC50 for 5 hrs. There was an increase in measurable
haemoglobin and a reduction in haemozoin formation
for parasites treated with chloroquine, as expected for
this known inhibitor of haemoglobin digestion and hae-
mozoin formation. A similar build-up in haemoglobin
Burns et al. BMC Biology (2020) 18:133
Page 12 of 23
Fig. 6. Hierarchical clustering of the different sample groups, treated with chloroquine (CQ) (blue), DHA (green), azithromycin (Az) (light blue),
GSK-5 (purple), GSK-71 (yellow), GSK-66 (grey) and ethanol control (red). Vertical clustering displays similarities between sample groups, while
horizontal clusters reveal the relative abundances of the 50 most significantly different metabolites from experiment 1. The significantly
differentially regulated metabolites are further classified into three different groups, the CQ-like peptides (blue line), TCA cycle (red line) and
haemoglobin-derived peptides (orange lines). All compounds were tested with three technical replicates. White indicates no change, while red
and blue indicates increased and decreased abundances respectively. Ward’s minimum variance method algorithm was used to generate the
hierarchical cluster analysis
was seen for GSK-71; however, there was no decrease in
haemozoin, supporting that this drug may have activity
in the food vacuole, but this did not involve measurable
inhibition of haemozoin formation. Again, GSK-66
treatment had no effect on haemoglobin or haemozoin
levels, supporting the non-targeted metabolomics data
which suggests that this drug has limited effects on para-
the concentration and duration
site metabolism at
Burns et al. BMC Biology (2020) 18:133
Page 13 of 23
Fig. 7. Haemoglobin fractionation of GSK-71, chloroquine and GSK-66-treated Plasmodium falciparum (3D7) parasites compared to an ethanol
control. Scatter dot plots representing the relative levels of a haemoglobin, b free haem and c haemozoin in trophozoite-stage parasites
following a 5 hr incubation with 10× IC50 (44 hr) concentration of GSK-71 (1100 nM), chloroquine (520 nM) and GSK-66 (70 nM) expressed as the
fold change when compared to an EtOH control. Data is represented as the mean of > 3 paired replicates from three independent experiments
with the error bars expressed as SEM. Significant differences were assessed using Student’s t test. Repeat measure data is available in
Additional file 15 Supporting Value Data. d A panel of representative Giemsa-stained parasites treated with 10× IC50 (44 hr) concentration of GSK-
71, chloroquine, GSK-66 and the ethanol negative control after 5 hrs.
tested. These data support that azithromycin and ana-
logues have activity in the food vacuole of drug-treated
trophozoites, but also indicate additional activity outside
of haemoglobin digestion.
A second shared metabolomic signature was observed
for azithromycin and the phenyl-substituted analogue
GSK-5, with a major reduction in key metabolites (in-
cluding succinate, fumarate, malate) of the mitochon-
drial tricarboxylic acid (TCA) cycle (Additional file 11:
Table S7a&b, Additional file 13: Figure S5). The reduc-
tion in TCA metabolites was evident across repeat ex-
periments for azithromycin, but was less prominent for
GSK-5 in the second experiment (Additional
file 11:
Table S7a&b). Although several steps in the Plasmodium
TCA cycle are considered dispensable in blood-stage
parasites, the fumarate hydratase conversion of fumarate
to malate followed by the malate quinone oxidoreduc-
to
(MQO) mediated conversion of malate
tase
oxaloacetate are thought to have important roles in the
parasite’s purine salvage pathway [58, 59]. Reduced bio-
availability of fumarate and malate, two key metabolites
required for efficient purine salvage, would negatively
impact on purine production and parasite growth over
time and offers a novel drug development strategy. In-
deed, a recent paper has identified blood-stage inhibitors
of MQO in the Pathogen Box [60] suggesting that this
pathway is a viable drug target against asexual-stage par-
asites. These data implicate a second membrane-bound
organelle as a potential target during trophozoite stages
of the parasite lifecycle, underlining the potential for
multifactorial mechanisms of action.
Azithromycin and GSK-5 also caused a reduction in
haemoglobin-derived peptides across both experiments
to levels lower than seen for chloroquine and DHA, two
food vacuole targeting drugs (Additional file 12: Table
S8a&b). Thus, treatment with azithromycin and GSK-5
Burns et al. BMC Biology (2020) 18:133
Page 14 of 23
caused an increase in specific non-haemoglobin-derived
peptides similar to that seen for chloroquine, a consist-
ent decrease in haemoglobin-derived peptides (most
prominently for GSK-5 in this data set) and a decrease
in TCA cycle metabolites. In contrast, GSK-71 was most
notably associated with an increase in non-haemoglobin
chloroquine-like peptides, while GSK-66 and DHA had
minimal impact on the metabolic profile under the con-
ditions analysed here. This highlights the potential abil-
ity of azithromycin analogues with different structures to
interrupt normal metabolic functions across the cell and
in different organelles, even when used at the same fold-
IC50 and against the same lifecycle stages.
Given the metabolomics evidence suggesting that
azithromycin and analogues may target the food vacuole,
we investigated whether the rapid ring-stage killing
activity of the chloroquinoline analogue GSK-66 (Fig. 4a,
b, Table 1) may be a result of azithromycin pre-
sensitising ring stages to the chloroquinoline moiety. We
treated early ring-stage D10-PfPHG parasites (0–6 hrs)
with azithromycin at an IC10 concentration and added a
dilution series of chloroquine. Addition of azithromycin
did not potentiate chloroquine’s activity against early
ring stages, with the IC50 of azithromycin+chloroquine
remaining well above the activity of GSK-66 (Add-
itional file 14: Figure S6). In addition, a range of func-
tional groups were found to potentiate azithromycin’s
quick-killing activity. These combined data suggest that
to
azithromycin does not
chloroquinoline-like moieties nor act through disruption
of haem polymerisation per se as chloroquine is believed
to, but rather may act more broadly within the parasite’s
food vacuole as well as potentially other cellular and
organellar targets such as the parasite’s mitochondrion.
pre-sensitise
parasites
Discussion
The spread of parasites resistant to artemisinin combin-
ation therapies (ACTs) in Southeast Asia, India and
other regions highlights the need for novel antimalarial
drug treatment strategies to ensure timely and effective
treatment of clinical disease [3–6, 8]. Despite limited use
against clinical cases of malaria, macrolide antibiotics re-
main of interest as potential partner drugs in antimalar-
ial combinations due to their activity against malaria
parasites and well-established safety profile in children
and pregnant women [10, 11, 24, 61]. Recently, we iden-
tified that high concentrations of clinically used macro-
lides inhibit merozoite invasion in vitro and showed that
this mechanism of action was independent of apicoplast-
targeting delayed death [29]. Here, we demonstrate the
potential for the antibiotic azithromycin to be repur-
posed as an antimalarial with two potent mechanisms of
action with the identification of azithromycin analogues
that have potent activity throughout intra-erythrocytic
parasite development and against merozoite invasion.
We established that this activity is through a mechanism
independent of
the known activity of azithromycin
against the parasite apicoplast, revealing potential new
pathways for development of novel antimalarials.
We investigated the activity of a panel of the analogues
and identified 65 with improved in-cycle activity (44 hr
early rings to schizont treatment) compared to azithro-
mycin. Of these, 39 analogues with diverse functional
groups
IC50
0.02 μM), naphthalene (GSK-3, IC50 0.183 μM), quin-
IC50 0.048 μM) and chloroquinoline
oline (GSK-58,
(GSK-66, IC50 0.007 μM) had nanomolar IC50s, provid-
ing between an 11- to 1615-fold improvement over
azithromycin.
including substituted phenyl
(GSK-5,
Azithromycin and analogues exhibited equipotent
intracellular blood-stage
quick-killing activity across
parasite growth. This included rapid activity against early
ring-stage development (both 0–6 and 0–12 hrs post in-
vasion) at a similar potency to 0–44 hr (one cycle) treat-
ments. Therefore, azithromycin and analogues have a
similar efficacy profile to the artemisinins [37, 38], being
effective against early ring stages and across the blood-
stage lifecycle, but with additional potential to be active
against liver and transmission-stage parasites [22, 26,
27]. We found that the azithromycin analogues with the
best activity in 44 hr assays (GSK-3, GSK-5, GSK-56 and
GSK-72) also exhibited the greatest improvement in in-
vasion inhibitory activity over azithromycin, highlighting
that both quick-killing activities can be improved over
azithromycin. However, the ability to push potency of
merozoite invasion inhibition into clinically relevant
concentrations below 1 μM may be limited. Importantly,
assays where merozoites were treated directly prior to
compound removal and addition of RBCs to begin inva-
sion show that the invasion inhibitory activity of azithro-
mycin and analogues is directed against the merozoite
and not against the RBC. A number of invasion inhibi-
tory antimalarial strategies are being pursued globally
(reviewed in [62]), and there remains the possibility that
further improvements in azithromycin analogue invasion
inhibitory
additional
development.
achievable with
IC50
are
and
quinoline
chloroquinoline
It is interesting to note that improved quick-killing ac-
tivity is ubiquitous across analogues with phenyl, naph-
functional
thalene,
groups. It has previously been hypothesised that the high
potency of several analogues featuring quinoline and
chloroquinoline moieties was due to these analogues act-
ing like hybrid azithromycin (apicoplast ribosome target-
ing) and chloroquine (food vacuole target) activity [33,
34] molecules.
Interestingly, azithromycin analogues
with the four functional groups display properties dis-
similar to chloroquine, these being (i) improved invasion
Burns et al. BMC Biology (2020) 18:133
Page 15 of 23
for
lines
against
activity
chloroquine-resistant
analogues
inhibitory activity compared to azithromycin, whereas
chloroquine does not inhibit invasion [39, 63], and (ii)
and
similar
featuring
chloroquine-sensitive
substituted phenyl, naphthalene and quinoline moieties.
Activity against chloroquine-resistant DD2 for analogues
with chloroquinoline functional groups was variable with
two analogues showing improved potency against the
chloroquine-resistant line over the chloroquine-sensitive
line, while three compounds were less potent against the
resistant line; and (iii) potent inhibition of very early ring
stages (0–6 hrs post invasion), which are largely insensi-
tive to chloroquine. However, additional evidence from
this study does support the idea that azithromycin and
analogues quick-killing activity may, in part, be acting
against the parasite’s food vacuole.
trends were observed with the
Although our ability to perform comprehensive and
detailed SAR comparison was limited by compound
availability impacting on matched-pair analysis, some
general
analogues
available. Analogues with chloroquinoline and quinoline
substituents were generally the most potent in one-cycle
44 hr assays. Naphthalene had modest potency and is a
close bioisostere of quinoline. In general, analogues with
a short carbon linking the amino quinoline to the N6-
position of the macrocycle or the O- or N-position of
the desosamine group were the most active. Appending
functional moieties to the N6-position of the macrolac-
tone, or to the desosamine sugar, both conferred signifi-
cantly improved in-cycle activity, with a slight tendency
for improved quick-killing activity when the functional
group was either attached to the N- or the O- of the des-
osamine sugar as opposed to the N6-position of the
macrolactone (i.e. chloroquinoline GSK-66desos
(IC50
0.007 μM) and GSK-1macro (IC50 0.019 μM); naphthalene
GSK-78desos
(IC50
0.59 μM)). Thus, the position of the functional group on
the macrocyclic did not greatly impact activity, suggest-
ing the macrocycle may be acting as a vehicle for trans-
portation of the active functionality.
(IC50 0.51 μM) and GSK-12macro
Within the parasite, it is possible that analogues are
metabolised and then release the pendant quinoline or
aromatic system as the active component of compound.
This is possible either by an oxidative mechanism hydro-
lysing amine-linked substituents, or by proteolytic or
hydrolytic degradation of the amide and urea functional-
ity linking the pendant quinoline or aromatic group to
the macrolactone. In this study, we could not conclu-
sively address whether metabolism was occurring, but
this will be an important facet to address in a future
mechanistic study of these azalide analogues. The possibil-
ity of the macrolactone acting as a delivery vehicle with
subsequent metabolic release of the active payload in the
parasite raises the prospect for the azithromycin scaffold
to be tethered to and act as a delivery vehicle for other an-
timalarials that act at a similar asexual killing rate to
chloroquine, akin to antimalarial candidates undergoing
clinical trials such as KAF156 or MMV048 [64]. Such a
strategy to improve dual target efficacy of azithromycin
analogues, and delay the onset of resistance, is an attract-
ive option. Furthermore, while it has been demonstrated
that these analogues have efficacy in in vivo rodent models
[31, 33, 35], the effective contribution of quick-killing has
not been assessed. In addition, whether these analogues
would be stable to first pass metabolism in the liver is an-
other important aspect to consider in future development
of the azalide analogue class.
Although the azithromycin analogues identified as
having improved quick-killing activity in this study fea-
ture a range of added functional groups, compounds
with quinoline and chloroquinoline moieties feature
prominently amongst the most potent quick-killing ana-
logues. Hybrid molecules featuring quinolines fused to a
second chemotype with antimalarial properties such as
endoperoxides [65] or reversed chloroquine drugs that
are linked to a reversal agent, a molecule known to in-
hibit or circumvent the activity of the chloroquine resist-
ance transporter PfCRT [66, 67], have been developed
and shown to have efficacy in rodent malaria models
(reviewed in [68]). The current lead reversed chloro-
quine compound, DM1157 [69], has shown low nano-
molar potency against chloroquine-resistant parasites,
demonstrated efficacy against P. chabaudi rodent mal-
aria parasites and has recently undergone Phase I trials
in humans (NCT03490162, [70]). Despite the potential
of DM1157, hybrid molecules have faced hurdles in de-
velopment including examples of endoperoxide hybrids
unable to overcome existing resistance mechanisms [71]
and the high MW of the compounds impacting on desir-
able drug-like properties. In this regard, it is interesting
to note that the ketolide antibiotics solithromycin and
telithromycin, semi-synthetic derivatives of erythromycin
which both feature a large functional group added to the
macrolactone ring, have been progressed for clinical use.
This highlights that modified macrolides can be devel-
oped that maintain favourable drug-like properties des-
pite their high MW.
Metabolomic analysis of azithromycin and analogue-
treated parasites suggests one potential site of drug
activity in trophozoite stages is the parasite’s food vacu-
ole, with a similar build-up of largely non-haemoglobin
peptides observed for azithromycin, analogues GSK-5
and GSK-71 as seen for chloroquine. However, a num-
ber of differences to chloroquine were also observed in-
cluding the chloroquinoline-modified analogue GSK-66
causing minimal change in parasite metabolism, azithro-
mycin and GSK-5 having activity against mitochondrial
metabolism and GSK-5 also causing a reduction in
Burns et al. BMC Biology (2020) 18:133
Page 16 of 23
haemoglobin-derived peptides. Previous studies have
shown that trophozoite-stage treatment with the mito-
chondrial targeting drug atovaquone, alone and in com-
bination with proguanil, leads to a build-up of the TCA
metabolite fumarate [53, 72]. It was postulated that this
could be a result of the TCA enzyme malate-quinone
oxidoreductase complex also having a role in the mito-
chondrial electron transport chain (the target of atova-
quone) that may be affected by atovaquone, leading to
off-target disruption of the TCA cycle. In contrast, azi-
thromycin and GSK-5 treatment caused a reduction in
fumarate and other TCA metabolites, a signature differ-
ent to that of atovaquone. Interestingly, treatment with
the membrane-bound glucose transporter inhibitor 3361
led to a reduction in TCA and haemoglobin-derived
peptides after 6 hrs of drug treatment [72], similar to
that seen for azithromycin and GSK-5 here. The mul-
tiple changes in parasite metabolic networks seen when
inhibiting glucose uptake supports data generated in this
study that suggests azithromycin and analogues quick-
through multifactorial
occur
killing
mechanisms.
activity may
While there are limitations in this analysis, including
only one lifecycle stage and drug concentration (5× the
44 hr IC50) tested for each analogue, these data clearly
demonstrate that azithromycin and analogues likely have
multi-factorial mechanisms of action even against a sin-
gle lifecycle stage. Given the apparent site of activity for
azithromycin and analogues includes the membrane-
bound food vacuole and mitochondrion,
it is possible
that additional membrane-bound organelles in other life-
cycle stages (i.e. the rhoptry in merozoites) could also be
the target of these drugs. Additional experimental valid-
ation for the site of activity across a range of analogues
and lifecycle stages will need to be undertaken in order
to detail the potential promiscuity of these drugs in
stopping parasite growth.
Previous studies have suggested that azithromycin ana-
logues may act through a chloroquine-like mechanism
[33–35] (reviewed in [73]), and evidence presented in this
study from metabolomic experiments and haemoglobin
fractionation assays supports that one of the sites of activ-
ity for azithromycin and analogues is the parasite’s food
vacuole. If a chloroquine-like targeting of the food vacuole
is an important component of azithromycin and analogue
quick-killing activity, these modified analogues have two
major advantages over chloroquine and quinine for clin-
ical
and
quinoline-substituted analogues maintained reasonable
activity against chloroquine-resistant DD2 parasites. The
maintenance of potency against chloroquine-resistant par-
asites could be explained by the different properties of the
drug limiting the ability of the mutated chloroquine-
the drug from the
to expel
resistant
Firstly, phenyl-, naphthalene-
transporter
treatment.
developing vacuole [74, 75]. Secondly, azithromycin and
analogues have rapid activity against early ring-stage para-
sites. Rapid activity against ring stages is in stark contrast
to the poor activity of chloroquine and quinine against
these early parasites and it is certainly possible that azi-
thromycin and analogues could access the site of the ini-
tial stages of haemoglobin digestion, similar to artemisinin
[37, 38, 76], via superior lipophilic properties [33, 34].
[43]
D10-PfPHG
chloroquine-
and
Azithromycin and analogues display several other
properties of
interest. The majority of quick-killing
analogues tested against chloroquine/pyrimethamine-
artemisinin-resistant
DD2
resistant
and
Cam3.IIDHA resistant(R539T)
[44, 45] retained potency
and artemisinin-
compared to the
sensitive
artemisinin-sensitive
Cam3.IIsensitive lines. While there were examples of
chloroquinoline containing analogues being less potent
against DD2 parasites, these data broadly indicate that a
wide range of azithromycin analogue modifications can
significantly improve quick-killing activity in a way that
overcomes a number of established resistance mecha-
nisms. Azithromycin and analogue invasion blocking ac-
tivity is shared across distantly related Apicomplexan
parasites such as Toxoplasma gondii [29, 77], P. berghei
[29] and the zoonotic human malaria parasite P. knowlesi.
Since neither T. gondii nor Plasmodium spp. merozoites
contain a food vacuole, the target of chloroquine, it seems
likely that azithromycin and analogues have additional
mechanisms of action, with properties such as modulation
of intraerythrocytic calcium (Ca2+), interference of kinase
signalling pathways, cationic trapping and sequestration
within acidic environments, as well as decreasing mobility
of phospholipid bilayers demonstrated for azithromycin in
other eukaryotic cell systems, all potential alternative
MOAs contributing to quick-killing [78–82].
Finally, the influence of the site of modification to azi-
functional
thromycin and the addition of different
groups was investigated in the context of delayed-death
activity. Previous studies have demonstrated that the
desosamine sugar is critical for binding to bacterial ribo-
somes, and we anticipated that modifications to this re-
gion would stop apicoplast-targeting delayed-death
activity [12, 51, 52]. However, the potent quick-killing
activity of azithromycin analogues (GSK-4, GSK-5, GSK-
12, GSK-16, GSK-57, GSK-71, etc.) precluded assess-
ment of delayed-death activity using traditional 120 hr
parasite assays. Therefore, we assessed whether a fo-
cused set of azithromycin analogues maintained their ac-
tivity against prokaryotic ribosomes by determining the
minimum inhibitory concentration (MIC) activity of the
gram-positive bacteria, S. pneumoniae. Comparison of P.
falciparum quick-killing IC50 and S. pneumoniae MIC
confirmed that attaching the functional group to the
desosamine sugar (GSK-57, GSK-66, GSK-71 and GSK-
Burns et al. BMC Biology (2020) 18:133
Page 17 of 23
78) abrogated activity against bacterial ribosomes as ex-
pected. In contrast, analogues with the functional group
attached to the N6-positon of the macrolactone (GSK-1,
GSK-4, GSK-5, GSK-6, GSK-9, GSK-11, GSK-12, GSK-
16, GSK-17, GSK-21, GSK-25) maintained activity
against S. pneumoniae, suggesting that delayed-death ac-
tivity via targeting the bacterium-like ribosome of the
apicoplast is maintained in analogues featuring modifica-
tion to the N6-positon of the macrolactone (GSK-1,
GSK-4, GSK-5, GSK-6, GSK-9, GSK-11, GSK-12, GSK-
16, GSK-17, GSK-21, GSK-25). Thus, analogues could
be modified to act through either single (i.e. quick-
killing) or dual
(i.e. quick-killing and delayed-death)
mechanisms of action depending on the properties de-
sired (i.e. quick parasite clearance and/or long-term
prophylaxis) and whether removal of non-selective anti-
biotic activity is preferred over apicoplast-targeting
delayed-death prophylaxis.
azithromycin analogues
intracellular blood-stage development,
Conclusion
We have shown that azithromycin and analogues have a
quick-killing mechanism of action that kills parasites
throughout
in-
cluding inhibition of merozoite invasion of RBCs. Add-
exhibit promising
itionally,
potency against very early ring-stage parasites, which is a
rare feature amongst existing antimalarials. Importantly,
quick-killing can be improved without losing activity
against protein synthesis by the apicoplast ribosome (de-
layed death). Conversely, the option to engineer azithro-
mycin to remove activity against a bacterium-like
ribosome and thereby avoid selection for ‘bystander’
bacterial resistance is available. Further development of
azithromycin analogues offers the prospect of designing
compounds with either quick-killing (quick-parasite
clearance) mode of action or both quick-killing and
slow-killing prophylactic activity. This design strategy
should also retard resistance acquisition by hitting two
targets. Fine-tuning the quick-killing activity of azithro-
mycin analogues significantly broadens its clinical appli-
cations and offers resistance proofing through two
independent mechanisms of action. Therefore, the iden-
tification of potent azithromycin analogues with rapid
killing phenotypes and dual mechanisms of action (de-
layed-death and quick-killing activity) provide a new av-
enue for anti-malarial drug development.
Sigma), azithromycin (100 mM, AK-Scientific) and GSK
analogues (10 mM, GSK-1–84) were made up in ethanol
as vehicle. Chloroquine diphosphate salt (10 mM, Sigma-
Aldrich) was dissolved in 10% acetic acid in H2O. Dihy-
droartemisinin (10 mM, DHA, Sigma-Aldrich) were dis-
solved in dimethyl sulfoxide (DMSO). Drugs were added
such that the vehicle was diluted > 100-fold for merozo-
ite invasion assays and > 1000-fold for intracellular
growth assays to minimise non-specific inhibition.
Culture and synchronisation of Plasmodium spp. parasites
falcip-
Green fluorescent protein (GFP) expressing P.
arum D10-PfPHG parasites [84], DD2 [43], artemisinin-
(Cam3.IIDHA resistant(R539T)) and artemisinin-
resistant
sensitive (Cam3.IIsensitive) Cambodian isolates [45] and P.
knowlesi PkYH1 [85] were cultured in human O+ eryth-
rocytes (RBCs) (Australian Red Cross Blood Service).
Parasites were cultured in RPMI-HEPES culture medium
(pH 7.4, 50 μg/mL hypoxanthine, 25 mM NaHCO3,
20 μg/mL gentamicin, 0.5% Albumax II (Thermo Fisher
Scientific)) and maintained in an atmosphere of 1% O2,
4% CO2 and 95% N2 according to established protocols
[86]. Tight synchronisation of D10-PfPHG parasites was
achieved using sodium heparin [63, 87]. P. falciparum
DD2, the Cambodian isolates and P. knowlesi (PkYHI),
were synchronised with continuous passage over a gradi-
ent of 70% Percoll (Sigma-Aldrich) for purification of
late-stage schizonts and 5% w/v sorbitol (Sigma-Aldrich)
treatments for ring stages.
Drug inhibition assays
A diagram outlining the different Plasmodium spp. drug
inhibition assays used in this study is available in Fig. 1
and has been described previously [29, 63]. Stage specifi-
city assessment of azithromycin or analogues during
blood-stage P. falciparum development was undertaken
through the addition of the drug at the specified time
points (0–6 hrs, 0–12 hrs, 12–24 hrs, 24–36 hrs or 36–
44 hrs post merozoite invasion) and the subsequent re-
moval through three consecutive washes with 200 μl of
medium (centrifuged at 300×g for 2 min) before resus-
pending in a final volume of 200 μl. Parasite growth was
quantified at late schizont stages (44–48 hrs post inva-
sion) by flow cytometry of parasites stained with 10 μg/
mL ethidium bromide (EtBr) for 1 hr prior to washing
with PBS.
Methods
Antimalarial drugs
Azithromycin analogues (GSK-1–84) were a gift from
GlaxoSmithKline and were synthesised as described pre-
viously [31–35, 83]. Additional file 1: Tables S1a-c pro-
vides further details of chemical structure and analogue
(3075 mM
origin. Stock concentrations of quinine
Invasion inhibition assays
Purification of viable merozoites and merozoite invasion
inhibition assays has been described previously [29, 63,
87]. Briefly, 300 mL of D10-PfPHG schizont culture, 3%
haematocrit and 4–5% parasitaemia tightly synchronised
to a 6 hr window of invasion with heparin were magnet
purified (Mitenyi Biotech) away from RBCs at 40–46 hrs
Burns et al. BMC Biology (2020) 18:133
Page 18 of 23
post-invasion. Purified schizonts were eluted in up to
30 mL of media, 10 μM of E64 (Sigma-Aldrich) was
added and the parasites were left to mature for 5 hrs.
Schizonts were filtered through a 1.2-μm syringe filter
(Minisart, Sartorius) in incomplete media with NaHCO3
to release merozoites and 22.5 μl of filtrate was added to
2.5 μl of drug prior to addition of RBC (0.5% final haem-
atocrit). Plates were agitated at 400 rpm for 10 min at
37 °C to promote invasion.
For drug washout, 90 μL of purified merozoites was
added to 10 μL of either incomplete media (no serum)
or incomplete media plus drug before transfer to a 0.22-
μm Ultrafree-MC centrifugal filter (Thermo Fisher). Fil-
ter columns were centrifuged at 750 rcf for 1 min and
washed with incomplete media twice. Free merozoites
were resuspended off the filter in 45 μL of incomplete
media and transferred to 96-well U-bottom plates con-
taining 5 μL of RBCs at 1% haematocrit (final haemato-
crit of 0.1%). Plates were agitated at 400 rpm for 10 min
at 37 °C and cultures were incubated at 37 °C for 30 min.
Cells were treated with 5 μg/mL EtBr for 10 min prior to
being washed in 1 x PBS and ring-stage parasitemia
measured by flow cytometry.
Ring-stage survival assays (RSA0-3h)
For ring-stage survival assays [44–46], tightly synchro-
nised artemisinin-resistant Cam3.IIDHA resistant(R539T) and
artemisinin-sensitive Cam3.IIsensitive late schizont stage
parasites were concentrated over a gradient of 70% Per-
coll (Sigma-Aldrich), washed once in complete medium
and incubated for 3 hrs with fresh RBCs to allow inva-
sion. Cultures were sorbitol treated to eliminate the
remaining schizonts. The 0–3 hr post-invasion rings
were adjusted to 1% parasitemia and 1% haematocrit be-
fore exposure to a serial dilution of DHA, azithromycin
and azithromycin analogue concentrations for 4 hrs.
Plates were washed five times with 200 μl of medium be-
fore parasites were transferred into a new 96-well plate
to ensure the complete removal of drug [47]. Parasites
were grown for a further 66 hrs, before parasitemia was
assessed by flow cytometry.
Apicoplast-null inhibition assays
Apicoplast-null (D10-PfPHGapicoplast-null) [17, 36] parasites
were generated through supplementation of culture media
with 200 μM isopentenyl pyrophosphate (IPP) and apico-
plast removal through treatment with 0.35 μM (5× IC50)
of azithromycin for 6 days, with parasites cultured con-
tinuously thereafter with IPP. Removal of the apicoplast
was confirmed by growing D10-PfPHGwildtype and D10-
PfPHGapicoplast-null (+IPP) parasites with reducing concen-
trations of azithromycin for ~ 120 hrs which identified a
~ 64 fold-change in the IC50 values between the parasite
populations (D10-PfPHGapicoplast-null IC50, 4.5 μM; D10-
PfPHGwildtype IC50, 0.07 μM) confirming apicoplast re-
moval. To test
for azithromycin analogue activity
against the apicoplast, tightly synchronised ring-stage
D10-PfPHGapicoplast-null (+IPP) and D10-PfPHGwildtype
parasites were treated with the in-cycle 90% inhibitory
concentration (IC90) of drugs obtained for D10-
PfPHGwildtype for ~ 44 hrs (in-cycle) and the resulting
growth inhibition determined by flow cytometry.
Flow cytometry and microscopy analysis of inhibition
Parasitaemia was measured on an LSR Fortessa (Becton
Dickinson) with a 96-well plate reader. Mature (> 36 hr
post-invasion) P. falciparum D10-PfPHG parasites were
counted using Fl-1-high (GFP; excitation wavelength,
488 nm) and Fl-2-high (EtBr; excitation wavelength, 488
nm). D10-PfPHG ring-stage parasites (< 6 hrs post inva-
sion) were counted using a Fl-1-high (GFP) and Fl-2-low
(EtBr) gate [63]. Mature parasites of the remaining lines
were gated with a forward scatter (FSC) and FL-2-high
(EtBr) gate [63]. Typically, 20,000–40,000 RBCs were
counted in each well. Samples were analysed using
FlowJo software (TreeStar Inc) with growth of drug
treatments normalised against media control wells to
calculate the percentage survival. Thin smears for mi-
croscopy were fixed with fresh methanol and stained in
10% Giemsa (Merck) for 10 min. IC50s and IC90s were
determined for each drug using GraphPad Prism
(GraphPad Software) according to the recommended
protocol for nonlinear regression (constrained to top =
100 and bottom = 0) of a log-(inhibitor)-versus-response
curve.
Selection of azithromycin-resistant P. falciparum lines
In vitro selection of quick-killing-resistant lines was car-
ried out using a P. falciparum (D10-PfPHG) line featur-
ing a G91D mutation in the apicoplast ribosomal gene,
rpl4, resulting in a ~ 57-fold loss of sensitivity to azithro-
mycin’s delayed-death activity (2 cycles, Fig. 1d) (D10-
AZRr). To select for quick-killing resistance [12], D10-
AZRr parasites were first exposed to 3× IC50 of GSK-59
(chloroquinoline moiety, delayed-death inactive drug)
for 3 days, followed by a 5× IC50 concentration for 4
days then 3× IC50 for an additional 2 days prior to re-
moval of the drug. After treatment, parasites were fed
once every 2 days, and once a week, 30–40% of culture
was replaced with fresh RBCs. Parasites were examined
every 2 to 3 days by Giemsa-stained thin blood films for
between 3 (90 days) and 5 months (150 days) with no re-
crudescent parasites observed.
Antibacterial screen
Antibacterial activity of azithromycin and analogues
against Streptococcus pneumoniae was determined using
96-well minimum inhibitory concentration (MIC) assays
Burns et al. BMC Biology (2020) 18:133
Page 19 of 23
[88]. Two-fold serial dilutions were added to macrolide-
sensitive D39 S. pneumoniae in 100 μL Mueller Hinton
Broth supplemented with 5% lysed horse blood. Bacterial
growth was assessed after 24 hrs incubation with drug by
estimating the MIC where bacterial growth, as indicated
by a media colour change, could be identified (MIC
expressed as μM).
Sample extraction for metabolomics analysis
For metabolomics experiments, two 150-mL flasks at 6%
haematocrit containing tightly synchronised ~ 30–34 hr
D10-PfPHG trophozoites were harvested via magnet
purification (Miltenyi Biotech).
Infected RBC density
was quantitated by flow cytometry [89], and 2 mL of 3 ×
107 parasites were added to and incubated in 24-well mi-
crotiter plates for 1 hr at 37 °C to stabilise the culture.
Drugs (5× IC50) were added and incubated for a further
2 hrs prior to removal of the supernatant, 2× washes
with 800 μL ice-cold 1× PBS with cells pelleted via cen-
trifugation at 400×g for 5 min at 4 °C. The cell pellets
were resuspended in 150 μL of ice-cold extraction buffer
(MeOH) containing 1 μM internal standards; CHAPS
and PIPES, and incubated on ice for 1 hr with shaking at
200 rpm. Insoluble material was pelleted with centrifuga-
tion at 14,800×g for 10 min at 4 °C and 120 μL of super-
natant was collected and stored at − 80 °C until analysis.
spectrometry
LC-MS analysis
(LC-MS)
Liquid chromatography-mass
data was acquired on a Q-Exactive Orbitrap mass
spectrometer (Thermo Scientific) coupled with high-
performance liquid chromatography system (HPLC,
Dionex Ultimate® 3000 RS, Thermo Scientific) as previ-
ously described [53]. Briefly, chromatographic separation
was performed on a ZIC-pHILIC column equipped with
a guard (5 μm, 4.6 × 150 mm, SeQuant®, Merck). The
mobile phase (A) was 20 mM ammonium carbonate
(Sigma Aldrich), and (B) acetonitrile (Burdick and Jack-
son) and needle wash solution was 50% isopropanol.
The column flow rate was maintained at 0.3 ml/min with
temperature at 25 °C and the gradient programme was
as follows: 80% B decreasing to 50% B over 15 min, then
to 5% B at 18 min until 21 min, increasing to 80% B at
24 min until 32 min. Total run time was 32 min with an
injection volume of 10 μL. A mass spectrometer was op-
erated in full scan mode with positive and negative po-
larity switching at 35k resolution at 200 m/z, with
detection range of 85 to 1275 m/z, AGC target was 1e6
ions with a maximum injection time of 50 ms. Electro-
spray ionisation source (HESI) was set to 4.0 kV voltage
for positive and negative mode, and sheath gas was set
to 50, aux gas to 20 and sweep gas to 2 arbitrary units,
capillary temperature to 300 °C and probe heater
temperature to 120 °C. The samples were analysed as a
single batch to avoid batch-to-batch variation and ran-
domised to account for LCMS system drift over time.
Repeated analysis of pooled quality control samples was
the batch to confirm signal
performed throughout
reproducibility.
Data processing using IDEOM
The acquired LCMS data was processed in untargeted
fashion using open source software, IDEOM [90] (http://
Initially, Proteo-
mzmatch.sourceforge.net/ideom.php).
Wizard was used to convert raw LC-MS files to mzXML
format and XCMS was used to pick peaks. Mzmatch.R
was used to convert to peakML files, align samples and
filter peaks using minimum detectable intensity of 100,
000, relative standard deviation (RSD) of < 0.5 (reprodu-
cibility), and peak shape (codadw) of > 0.8. Mzmatch was
also used to retrieve missing peaks and annotate related
peaks. Default IDEOM parameters were used to elimin-
ate unwanted noise and artefact peaks. Loss or gain of a
proton was corrected in negative and positive ESI
modes, respectively, followed by putative identification
of metabolites by accurate mass within 3 ppm mass error
searching against common metabolite databases includ-
ing the Kyoto Encyclopedia of Genes and Genomes
(KEGG), MetaCyc and LIPIDMAPS. To reduce the
number of false positive identifications, retention time
error was calculated for each putatively identified metab-
olite using IDEOM’s build-in retention time model
which uses actual retention time data of authentic
standards (~ 350 standards). Metabolites identified by
comparison to authentic standards (including TCA cycle
metabolites) are level 1 identifications according to the
other
Metabolomics
putatively identified metabolites (including all peptides)
are assigned as level 2. Statistical analysis on filtered data
was performed using the Metaboanalyst web interface
[91].
Standards
Initiative,
and all
Haemoglobin fractionation
The haemaglobin fractionation assay was adapted from
[57]. Aliquots of 6.5 mL of 30–32 hr post invasion para-
site cultures were adjusted to 8% parasitaemia and 2%
haematocrit and then incubated with chloroquine, GSK-
66, GSK-71 or ethanol (vehicle control) for 5 hrs. Treat-
ments were performed in triplicate. Following incuba-
tion, the media was aspirated off and the culture was
incubated with 2.3 mL of 0.1% saponin in 1× PBS with
protease inhibitors (complete mini protease inhibitor
cocktail (Roche)) for 10 min at 4 °C in order to lyse the
iRBCs. The parasites were washed three times with PBS
and stored at − 80 °C.
For the haemoglobin fractionation,
lysed parasites
were resuspended in 50 μL of Milli-Q water and soni-
cated for 5 min in a water bath sonicator. Following
Burns et al. BMC Biology (2020) 18:133
Page 20 of 23
sonication, 50 μL of 0.2 M HEPES (pH 7.5) was added
and the samples were centrifuged at 4000 rpm for 20
min. The supernatant containing the haemoglobin frac-
tion was carefully transferred to new tubes and 50 μL of
4% of SDS was added before the samples were incubated
at 95 °C for 5 min. Following heating, 50 μL of 0.3 M
NaCl and 50 μL of 25% (v/v) pyridine (Sigma) in 0.2 M
HEPES was added, the sample containing the haemoglo-
bin fraction were vortexed and transferred to a 96-well
plate.
The remaining pellets were treated with 50 μL of
MilliQ water and 50 μL of 4% SDS and resuspended be-
fore being sonicated for 5 min and incubated at 95 °C for
5 min in order to solubilise the free haem. Following in-
cubation, 50 μL of 0.2 M HEPES, 0.3 M NaCl and 25%
pyridine were added to the samples. The samples were
then subsequently centrifuged at 4000 rpm for 20 min.
The supernatant was transferred to the 96-well plate,
corresponding to the free haem fraction.
The remaining pellet containing the haemozoin frac-
tion was solubilised by resuspending with 50 μL of
MilliQ water and 50 μL of 0.3 M NaOH. The samples
were sonicated for 15 min before 50 μL of 0.2 M HEPES,
0.3 M HCl and 25% pyridine was added. The samples
were then transferred to the 96-well plate, corresponding
to the haemozoin fraction. The total amount of haem in
each fraction was quantified using a haem standard
curve prepared from a 100 μg/mL standard solution of
haematin in 0.3 M NaOH. Serial dilution of the standard
curve was carried out in a 96-well plate in triplicate, and
50 μL of 0.2 M HEPES, 4% SDS, 0.3 M NaCl, 0.3 M HCl
and 25% pyridine was added. The absorbance of the
standard curve and each fraction was measured at a 405-
nm wavelength using a Perkin Elmer Ensight Plate
Reader. The samples were normalised via a paired ana-
lysis to the ethanol control and graphed as their fold
change vs ethanol ± SEM. All fractions had > 2 replicates
from 2 independent experiments.
Supplementary information
Supplementary information accompanies this paper at https://doi.org/10.
1186/s12915-020-00859-4.
Additional file 1 : Table S1. Activities of azithromycin analogues.
Additional file 2 : Figure S1. Azithromycin analogues show
improvement in invasion inhibitory activity. (A) Screening a panel of
azithromycin analogues identified 7 with up to 6-fold lower invasion in-
hibitory IC50 activity in contrast to the parental azithromycin. IC50 Azithro-
mycin 10 μM; GSK-4, 2.0 μM (Azithromycin vs GSK-4 P < 0.0001); GSK-5,
1.61 μM (Azithromycin vs GSK-5 P < 0.0001); GSK-56, 3.2 μM (Azithromycin
vs GSK-56 P < 0.0001); GSK-8, 4.4 μM (Azithromycin vs GSK-8 P = 0.2); GSK-
3, 1.8 μM (Azithromycin vs GSK-3 P < 0.0001); GSK-15, 3.6 μM (Azithromy-
cin vs GSK-15 P < 0.001); GSK-72, 1.7 μM (Azithromycin vs GSK-72 P <
0.0001). Newly invaded ring-stage parasitemia was measured at 1 hr post
invasion via flow cytometry. Data represents the mean of 2 (GSK 5) or
more experiments expressed as percentage of non-inhibitory control.
Error bars represent ± SEM. Dose response IC50s compared using extra
sum of squares F-test. (B) The food-vacuole targeting antimalarial drugs
chloroquine and quinine showed minimal invasion inhibitory activity at
10 μM while merozoite invasion was blocked by the invasion inhibitory
control heparin (25 μg/mL). Data represents the mean of 3 experiments
expressed as percentage of non-inhibitory control. Error bars represent ±
SEM. Repeat measure data is available in Additional file 15 Supporting
Value Data.
Additional file 3 : Figure S2. Azithromycin analogues inhibit merozoite
invasion irreversibly. Whether azithromycin analogues inhibited invasion
through a direct effect on the merozoite, rather than an effect on the
RBC, was assessed by directly treating and then washing the drug off
purified merozoites. Analogue GSK-72 was chosen as a compound with
improved invasion inhibitory activity over azithromycin with merozoites
treated at 10 μM. The actin inhibitor cytochalasin D (cytoD, 500 μM) was
included as an irreversible washout control. The RON2 binding peptide
R1 (100 μg/mL) was included as a reversible control. Ring-stage parasit-
aemia of newly invaded parasites was determined ~ 30 min post invasion
by flow cytometry, with results presented as % parasitaemia relative to a
media control. Results represent the mean of 2 experiments and the error
bars represent the ± SEM. Repeat measure data is available in Add-
itional file 15 Supporting Value Data.
Additional file 4 : Figure S3. Growth inhibition profiles of azithromycin
and analogues in parasites lacking the apicoplast. Early ring-stage P. fal-
ciparum parasites (0–4 hrs post-invasion) were treated with doubling dilu-
tions of azithromycin and inhibition of growth measured for (A) 2 cycle
(delayed death, 120 hr) assays (D10-PfPHGapicoplast-null IC50, 4.5 μM; D10-
PfPHGwildtype IC50, 0.07 μM. P = < 0.0001) or (B) 44 hr (in-cycle) (D10-
PfPHGapicoplast-null IC50, 16 μM; D10-PfPHGwildtype IC50, 11.3 μM. P = 0.24) as-
says. Parasitemia was measured at 120 hrs or 44 hrs post invasion, re-
spectively, at schizont stage via flow cytometry. Data represents the
mean of 3 (or more) experiments expressed as percentage of non-
inhibitory control and error bars represent ± SEM. (C) There was no differ-
ence in 44 hr IC50s between D10-PfPHGapicoplast-null and D10-PfPHGwildtype
parasites when treated with the azithromycin analogues GSK 1 (D10-
PfPHGapicoplast-null IC50, 0.028 μM; D10-PfPHGwildtype IC50, 0.023 μM. P = 0.36)
and GSK 66 (D10-PfPHGapicoplast-null IC50, 0.009 μM; D10-PfPHGwildtype IC50,
0.007 μM. P = 0.08). Data represents the mean of 2 (D10-PfPHGapicoplast-null)
or 3 (D10-PfPHGwildtype) experiments expressed as percentage of non-
inhibitory control and error bars represent ± SEM. Dose response IC50s
compared using extra sum of squares F-test. Repeat measure data is
available in Additional file 15 Supporting Value Data.
Additional file 5 : Table S2. Azithromycin analogue activity across
different age ranges of D10-PfPHG blood stage development.
Additional file 6 : Table S3. Azithromycin analogue inhibition for
chloroquine sensitive and resistant lines.
Additional file 7 : Table S4. Azithromycin analogue activity against P.
falciparum D10-PfPHG and P. knowlesi YH1 parasites.
Additional file 8 : Table S5. Azithromycin analogue activity against the
bacterial pathogen Streptococcus pneumoniae compared to P. falciparum
D10-PfPHG.
Additional file 9 : Figure S4. Sparse partial least square-discriminant ana-
lysis (SPLS-DA) of Plasmodium falciparum (D10-PfPHG)-infected red blood
cells following treatment with DHA (green), chloroquine (blue), azithro-
mycin (light blue), GSK-5 (purple), GSK-71 (yellow), GSK-66 (grey), and
ethanol control (red) from experiment 1. sPLS-DA showing scores plot for
components one and two, the plots were generated using the top 10
metabolites for each component. Points represent individual sample rep-
licates while the 95% confidence interval is represented by the shaded re-
gion. (File format .pdf).
Additional file 10 : Table S6. Changes in metabolites upon azithromycin
and analogue treatment shared with chloroquine treated parasites.
Additional file 11 : Table S7. Changes in metabolites upon azithromycin
and analogue treatment associated with the parasite TCA cycle.
Additional file 12 : Table S8. Changes in metabolites upon azithromycin
and analogue treatment mapping to haemoglobin after drug treatment.
Burns et al. BMC Biology (2020) 18:133
Page 21 of 23
Additional file 13 : Figure S5. Model for TCA metabolism following
treatment of Plasmodium falciparum (D10-PfPHG)-infected red blood cells.
Relative abundance of the TCA metabolites from infected red blood cells
treated with DHA (blue), chloroquine (red), azithromycin (green), GSK-5
(purple), GSK-71 (orange), GSK-66 (black), compared with the Ethanol con-
trol from experiment 1. Data are represented as mean fold change from
triplicate treatments multiplied by corresponding RSD values. Abbrevia-
tions: OAA, oxaloacetate; PEP, phosphoenolpyruvate.
Additional file 14 : Figure S6. Azithromycin does not pre-sensitise early-
ring stages to chloroquine. Early ring-stage P. falciparum parasites (0–4 hrs
post-invasion) were treated with doubling dilutions of chloroquine (IC50;
0–6 hrs, 0.73 μM), chloroquine + IC10 of azithromycin (IC50; 0–6 hrs,
1.1 μM) or highly potent analogue GSK-66 which features a chloroquino-
line moiety (IC50; 0–6 hrs, 0.004 μM) for 0–6 hrs, prior to removal of drugs
by washing. Comparison of the resulting in-cycle growth shows a small
change between growth of chloroquine vs chloroquine + azithromycin
treated parasites (P = 0.0041). This compares to a large difference be-
tween growth inhibitory IC50 of GSK-66 and chloroquine (P < 0.0001) and
chloroquine + azithromycin (P < 0.0001), indicating that azithromycin
does not potentiate ring stage activity of chloroquine. Parasitemia was
measured at 44 hrs post invasion at schizont stage via flow cytometry.
Data represents the mean of 3 (or more) experiments expressed as per-
centage of non-inhibitory control and error bars represent ± SEM. Dose
response IC50s compared using extra sum of squares F-test. Repeat meas-
ure data is available in Additional file 15 Supporting Value Data. (File for-
mat .pdf).
Additional file 15 : Supporting data values. Excel Spreadsheet
containing repeat measure data for Figs. 3, 4, 5 and 7, and
Additional files 2, 3, 4 and 15.
Abbreviations
ACT: Artemisinin combination therapies; RBC: Red blood cell; Ca2+: Calcium;
DMSO: Dimethyl sulfoxide; EtBr: Ethidium bromide; FIC: Fractional Inhibitory
Concentration; FSC: Forward scatter; GFP: Green fluorescent protein;
GSK: GlaxoSmithKline; HPLC: High-performance liquid chromatography;
IC: Inhibitory concentration; IPP: Isoprenoid pyrophosphate; IPTp: Intermittent
preventative treatment for malaria in pregnancy; KEGG: Kyoto Encyclopedia
of Genes and Genomes; LC-MS: Liquid chromatography-mass spectrometry;
MAPK: Mitogen-activated protein kinase; MIC: Minimum inhibitory
concentration; pf: Plasmodium falciparum; pk: Plasmodium knowlesi;
RSD: Relative standard deviation; RSA: Ring-stage survival assay;
SEM: Standard error of the mean; WHO: World Health Organization
Acknowledgements
Dr. Francisco Javier Gamo and Dr. Noemi Bahamontes Rosa (GlaxoSmithKline,
Tres Cantos facility, Spain) for the provision of modified azalides. Dr. Jeremy
Burrows, Medicines for Malaria Venture, for helpful discussion and advice.
David Fidock and Leann Tilley for providing the laboratory adapted
Cam3.IIDHA resistant(R539T) and Cam3.IIsensitive lines. We thank Juan Miguel Balbin
for help in generating the diagrams. Human erythrocytes were kindly
provided by the Red Cross Blood Bank (Adelaide, Australia). Metabolomics
analysis was performed at the Monash Proteomics and Metabolomics
Facility.
Authors’ contributions
DW, BS, CG, JB, DC, GS, and GM contributed to the conceptualization. AB, GS,
AD, DA, BL, RH, and DW contributed to the experiments and validation. GS,
BS, and DC contributed to the specialised analysis. AB, BS, GS, AD, DA, BL,
RH, JB, DC, CD, GM, and DW contributed to the writing, reviewing and
editing of the manuscript. All authors read and approved the final
manuscript.
Funding
This work was made possible through the National Health and Medical
Research Council of Australia (Project Grant 1143974 to D.W.W., G.I.M, B.E.S.
and C.D.G; Development Grant 1113712 to B.E.S.; Senior Research Fellowship
1077636 to JGB; Career Development (II) Fellowship 1148700 to DJC) and the
Victorian State Government Operational Infrastructure Support and Australian
Government NHMRC IRIISS. D.W.W. is a University of Adelaide Beacon Fellow
and Hospital Research Foundation Fellow. B.E.S. is a Corin Centenary Fellow.
Availability of data and materials
All data generated or analysed during this study are included in this
published article, its supplementary information files and publicly available
repositories.
The metabolomics spectrometry data and search results [92] supporting the
conclusions of this article are available at the NIH Common Fund’s National
Metabolomics Data Repository (NMDR) website, the Metabolomics
Workbench, https://www.metabolomicsworkbench.org where it has been
assigned Project ID (ST001315). The data can be accessed directly via it’s
Project DOI: (https://doi.org/10.21228/M8CX0M). This work is supported by
NIH grant U2C-DK119886.
Supporting data values for other experiments are included in
Additional file 15 Supporting Data Values. Other datasets used and/or
analysed during the current study are available from the corresponding
author on request.
Ethics approval and consent to participate
Human RBCs were provided by the Australian Red Cross Blood Bank with
ethics approval for use of the cells obtained from the University of Adelaide
Human Ethics Committee.
Consent for publication
Not Applicable.
Competing interests
The funders had no role in study design, data collection and analysis,
decision to publish, or preparation of the manuscript. The authors have
declared that no conflict of interest exists.
Author details
1Research Centre for Infectious Diseases, School of Biological Sciences, The
University of Adelaide, Adelaide 5005, Australia. 2Walter and Eliza Hall
Institute of Medical Research, Melbourne, Victoria 3050, Australia.
3Department of Medical Biology, University of Melbourne, Melbourne,
Victoria 3050, Australia. 4Monash Institute of Pharmaceutical Sciences,
Monash University, Melbourne, Victoria 3052, Australia. 5Burnet Institute,
Melbourne, Victoria 3004, Australia. 6Department of Medicine, University of
Melbourne, Melbourne, Australia. 7Central Clinical School and Department of
Microbiology, Monash University, Melbourne, Australia. 8School of
Biosciences, University of Melbourne, Melbourne, Victoria 3010, Australia.
Received: 9 March 2020 Accepted: 28 August 2020
References
1. Murray CJ, Ortblad KF, Guinovart C, Lim SS, Wolock TM, Roberts DA, et al.
Global, regional, and national incidence and mortality for HIV, tuberculosis,
and malaria during 1990-2013: a systematic analysis for the Global Burden
of Disease Study 2013. Lancet. 2014;384(9947):1005–70.
2. World Malaira Report. Geneva: World Health Organization (WHO). 2018.
3.
Ariey F, Witkowski B, Amaratunga C, Beghain J, Langlois AC, Khim N, et al. A
molecular marker of artemisinin-resistant Plasmodium falciparum malaria.
Nature. 2014;505(7481):50–5.
5.
6.
4. Miotto O, Sekihara M, Tachibana S-I, Yamauchi M, Pearson RD, Amato R,
et al. Emergence of artemisinin-resistant Plasmodium falciparum with
kelch13 C580Y mutations on the island of New Guinea. bioRxiv. 2019:
621813.
Das S, Saha B, Hati AK, Roy S. Evidence of artemisinin-resistant Plasmodium
falciparum malaria in Eastern India. N.Engl. J. Med. 2018;379(20):1962–4.
Ashley E, Dhorda M, Fairhurst R, Amaratunga C, Lim P, Suon S, et al. Spread
of Artemisinin resistance in Plasmodium falciparum malaria. N Engl J Med.
2014;371(5):411–23.
Nsanzabana C. Resistance to artemisinin combination therapies (ACTs): do
not forget the partner drug! Int J Infect Dis. 2019; 4(1): 26.
Blasco B, Leroy D, Fidock DA. Antimalarial drug resistance: linking
Plasmodium falciparum parasite biology to the clinic. Nat Med. 2017;23(8):
917–28.
Burrows JN, Duparc S, Gutteridge WE, Hooft van Huijsduijnen R, Kaszubska
W, Macintyre F, et al. New developments in anti-malarial target candidate
and product profiles. Malar J. 2017;16(1):26.
7.
8.
9.
Burns et al. BMC Biology (2020) 18:133
Page 22 of 23
10. Noedl H, Krudsood S, Chalermratana K, Silachamroon U, Leowattana W,
32. Hutinec A, Rupcic R, Ziher D, Smith KS, Milhous W, Ellis W, et al. An
Tangpukdee N, et al. Azithromycin combination therapy with artesunate or
quinine for the treatment of uncomplicated Plasmodium falciparum malaria
in adults: a randomized, phase 2 clinical trial in Thailand. Clin Infect Dis.
2006;43(10):1264–71.
Sykes A, Hendriksen I, Mtove G, Mandea V, Mrema H, Rutta B, et al.
Azithromycin plus artesunate versus artemether-lumefantrine for treatment
of uncomplicated malaria in Tanzanian children: a randomized, controlled
trial. Clin Infect Dis. 2009;49(8):1195–201.
Sidhu AB, Sun Q, Nkrumah LJ, Dunne MW, Sacchettini JC, Fidock DA. In
vitro efficacy, resistance selection, and structural modeling studies implicate
the malarial parasite apicoplast as the target of azithromycin. J Biol Chem.
2007;282(4):2494–504.
11.
12.
13. Dahl EL, Rosenthal PJ. Multiple antibiotics exert delayed effects against the
Plasmodium falciparum apicoplast. Antimicrob Agents Chemother. 2007;
51(10):3485–90.
14. Goodman CD, Su V, McFadden GI. The effects of anti-bacterials on the
malaria parasite Plasmodium falciparum. Mol Biochem Parasitol. 2007;152(2):
181–91.
Imlay L, Odom AR. Isoprenoid metabolism in apicomplexan parasites. Curr
Clin Microbiol Reo. 2014;1(3–4):37–50.
15.
16. Guggisberg AM, Amthor RE, Odom AR. Isoprenoid biosynthesis in
Plasmodium falciparum. Eukaryot Cell. 2014;13(11):1348–59.
17. Yeh E, DeRisi JL. Chemical rescue of malaria parasites lacking an apicoplast
defines organelle function in blood-stage Plasmodium falciparum. PLoS Biol.
2011;9(8):e1001138.
Lode H, Borner K, Koeppe P, Schaberg T. Azithromycin--review of key
chemical, pharmacokinetic and microbiological features. J Antimicrob
Agents Chemother. 1996;37(Suppl C):1–8.
18.
19. Hopkins S. Clinical toleration and safety of azithromycin. Am J Med. 1991;
20.
21.
22.
91(3a):40s–5s.
Sagara I, Oduro AR, Mulenga M, Dieng Y, Ogutu B, Tiono AB, et al. Efficacy
and safety of a combination of azithromycin and chloroquine for the
treatment of uncomplicated Plasmodium falciparum malaria in two multi-
country randomised clinical trials in African adults. Malar J. 2014;13:458.
Kimani J, Phiri K, Kamiza S, Duparc S, Ayoub A, Rojo R, et al. Efficacy and
safety of azithromycin-chloroquine versus sulfadoxine-pyrimethamine for
intermittent preventive treatment of Plasmodium falciparum malaria
infection in pregnant women in Africa: an open-label, Randomized Trial.
PloS One. 2016;11(6):e0157045.
Friesen J, Silvie O, Putrianti ED, Hafalla JC, Matuschewski K, Borrmann S.
Natural immunization against malaria: causal prophylaxis with antibiotics. Sci
Transl Med. 2010;2(40):40ra9.
23. Gaillard T, Madamet M, Pradines B. Tetracyclines in malaria. Malar J. 2015;14:
24.
25.
26.
445.
Luntamo M, Kulmala T, Mbewe B, Cheung YB, Maleta K, Ashorn P. Effect of
repeated treatment of pregnant women with sulfadoxine-pyrimethamine
and azithromycin on preterm delivery in Malawi: a randomized controlled
trial. Am J Trop Med Hyg. 2010;83(6):1212–20.
Schachterle SE, Mtove G, Levens JP, Clemens E, Shi L, Raj A, et al. Short-term
malaria reduction by single-dose azithromycin during mass drug
administration for trachoma, Tanzania. J Emerg Infect Dis. 2014;20(6):941–9.
Shimizu S, Osada Y, Kanazawa T, Tanaka Y, Arai M. Suppressive effect of
azithromycin on Plasmodium berghei mosquito stage development and
apicoplast replication. Malar J. 2010;9:73.
automated, polymer-assisted strategy for the preparation of urea and
thiourea derivatives of 15-membered azalides as potential antimalarial
chemotherapeutics. Bioorg Med Chem. 2011;19(5):1692–701.
33. Peric M, Fajdetic A, Rupcic R, Alihodzic S, Ziher D, Bukvic Krajacic M, et al.
Antimalarial activity of 9a-N substituted 15-membered azalides with
improved in vitro and in vivo activity over azithromycin. J Med Chem. 2012;
55(3):1389–401.
34. Pesic D, Starcevic K, Toplak A, Herreros E, Vidal J, Almela MJ, et al. Design,
synthesis, and in vitro activity of novel 2′-O-substituted 15-membered
azalides. J Med Chem. 2012;55(7):3216–27.
Starcevic K, Pesic D, Toplak A, Landek G, Alihodzic S, Herreros E, et al. Novel
hybrid molecules based on 15-membered azalide as potential antimalarial
agents. Eur J Med Chem. 2012;49:365–78.
35.
36. Uddin T, McFadden GI, Goodman CD. Validation of putative apicoplast-
targeting drugs using a chemical supplementation assay in cultured human
malaria parasites. Antimicrob Agents Chemother. 2018;62(1): e01161-17.
37. Dogovski C, Xie SC, Burgio G, Bridgford J, Mok S, McCaw JM, et al. Targeting
38.
the cell stress response of Plasmodium falciparum to overcome artemisinin
resistance. PLoS Biol. 2015;13(4):e1002132.
Klonis N, Xie SC, McCaw JM, Crespo-Ortiz MP, Zaloumis SG, Simpson JA,
et al. Altered temporal response of malaria parasites determines differential
sensitivity to artemisinin. Proc Natl Acad Sci U S A. 2013;110(13):5157–62.
39. Zhang Y, Asante KS, Jung A. Stage-dependent inhibition of chloroquine on
Plasmodium falciparum in vitro. J Parasitol. 1986;72(6):830–6.
41.
40. Cowman AF, Morry MJ, Biggs BA, Cross GA, Foote SJ. Amino acid changes
linked to pyrimethamine resistance in the dihydrofolate reductase-
thymidylate synthase gene of Plasmodium falciparum. Proc Natl Acad Sci U
S A. 1988;85(23):9109–13.
Fry M, Pudney M. Site of action of the antimalarial hydroxynaphthoquinone,
2-[trans-4-(4′-chlorophenyl) cyclohexyl]-3-hydroxy-1,4-naphthoquinone
(566C80). Biochem Pharmacol. 1992;43(7):1545–53.
Egan TJ, Ross DC, Adams PA. Quinoline anti-malarial drugs inhibit spontaneous
formation of beta-haematin (malaria pigment). FEBS Lett. 1994;352(1):54–7.
43. Wellems TE, Panton LJ, Gluzman IY. do Rosario VE, Gwadz RW, Walker-Jonah
42.
A, et al. Chloroquine resistance not linked to mdr-like genes in a
Plasmodium falciparum cross. Nature. 1990;345(6272):253–5.
45.
44. Witkowski B, Amaratunga C, Khim N, Sreng S, Chim P, Kim S, et al. Novel
phenotypic assays for the detection of artemisinin-resistant Plasmodium
falciparum malaria in Cambodia: in-vitro and ex-vivo drug-response studies.
Lancet Infect Dis. 2013;13(12):1043–9.
Straimer J, Gnadig NF, Witkowski B, Amaratunga C, Duru V, Ramadani AP,
et al. K13-propeller mutations confer artemisinin resistance in Plasmodium
falciparum clinical isolates. Science. 2015;347(6220):428–31.
Straimer J, Gnadig NF, Stokes BH, Ehrenberger M, Crane AA, Fidock DA.
Plasmodium falciparum K13 Mutations Differentially Impact Ozonide
Susceptibility and Parasite Fitness In Vitro. mBio. 2017;8(2): e00172-17.
47. Yang T, Xie SC, Cao P, Giannangelo C, McCaw J, Creek DJ, et al. Comparison of
46.
the exposure time dependence of the activities of synthetic ozonide
antimalarials and dihydroartemisinin against K13 wild-type and mutant
Plasmodium falciparum strains. Antimicrob Agents Chemother. 2016;60(8):
4501–10.
Singh B, Kim Sung L, Matusop A, Radhakrishnan A, Shamsul SS, Cox-Singh J,
et al. A large focus of naturally acquired Plasmodium knowlesi infections in
human beings. Lancet. 2004;363(9414):1017–24.
48.
27. Goodman CD, Useglio M, Peiru S, Labadie GR, McFadden GI, Rodriguez E,
49. Moon RW, Sharaf H, Hastings CH, Ho YS, Nair MB, Rchiad Z, et al.
et al. Chemobiosynthesis of new antimalarial macrolides. Antimicrob Agents
Chemother. 2013;57(2):907–13.
van Eijk AM, Terlouw DJ. Azithromycin for treating uncomplicated malaria.
Cochrane database Syst Rev. 2011;2011(2):Cd006688.
28.
29. Wilson DW, Goodman CD, Sleebs BE, Weiss GE, de Jong NW, Angrisano F,
et al. Macrolides rapidly inhibit red blood cell invasion by the human
malaria parasite, Plasmodium falciparum BMC Biol 2015;13:52.
30. Cooper DJ, Rajahram GS, William T, Jelip J, Mohammad R, Benedict J, et al.
Plasmodium knowlesi malaria in Sabah, Malaysia, 2015-2017: ongoing
increase in incidence despite near-elimination of the human-only
Plasmodium species. Clin Infect Dis. 2020;70(3):361–7.
31. Bukvic Krajacic M, Peric M, Smith KS, Schonfeld ZI, Ziher D, Fajdetic A, et al.
Synthesis, structure-activity relationship, and antimalarial activity of ureas
and thioureas of 15-membered azalides. J Med Chem. 2011;54(10):3595–
605.
50.
Normocyte-binding protein required for human erythrocyte invasion by the
zoonotic malaria parasite Plasmodium knowlesi. Proc Natl Acad Sci U S A.
2016;113(26):7231–6.
van Schalkwyk DA, Moon RW, Blasco B, Sutherland CJ. Comparison of the
susceptibility of Plasmodium knowlesi and Plasmodium falciparum to
antimalarial agents. J Antimicrob Chemother. 2017;72(11):3051–8.
51. Poehlsgaard J, Douthwaite S. The bacterial ribosome as a target for
52.
antibiotics. Nat Rev Microbiol. 2005;3(11):870–81.
Schlunzen F, Zarivach R, Harms J, Bashan A, Tocilj A, Albrecht R, et al.
Structural basis for the interaction of antibiotics with the peptidyl
transferase centre in eubacteria. Nature. 2001;413(6858):814–21.
53. Creek DJ, Chua HH, Cobbold SA, Nijagal B, MacRae JI, Dickerman BK, et al.
Metabolomics-based screening of the malaria box reveals both novel and
established mechanisms of action. Antimicrob Agents Chemother. 2016;
60(11):6650–63.
Burns et al. BMC Biology (2020) 18:133
Page 23 of 23
77.
76. Abu Bakar N, Klonis N, Hanssen E, Chan C, Tilley L. Digestive-vacuole genesis
and endocytic processes in the early intraerythrocytic stages of Plasmodium
falciparum. J Cell Sci. 2010;123(Pt 3):441–50.
Lee Y, Choi JY, Fu H, Harvey C, Ravindran S, Roush WR, et al. Chemistry and
biology of macrolide antiparasitic agents. J Med Chem. 2011;54(8):2792–804.
Kanoh S, Rubin BK. Mechanisms of action and clinical application of
macrolides as immunomodulatory medications. Clin Microbol Rev. 2010;
23(3):590–615.
78.
79. Wasserman M, Alarcon C, Mendoza PM. Effects of Ca++ depletion on the
80.
81.
asexual cell cycle of Plasmodium falciparum. Am J Trop Med Hyg. 1982;
31(4):711–7.
Sicard A, Semblat JP, Doerig C, Hamelin R, Moniatte M, Dorin-Semblat D,
et al. Activation of a PAK-MEK signalling pathway in malaria parasite-
infected erythrocytes. Cell Microbiol. 2011;13(6):836–45.
Stepanic V, Kostrun S, Malnar I, Hlevnjak M, Butkovic K, Caleta I, et al.
Modeling cellular pharmacokinetics of 14- and 15-membered macrolides
with physicochemical properties. J Med Chem. 2011;54(3):719–33.
82. Berquand A, Mingeot-Leclercq MP, Dufrene YF. Real-time imaging of drug-
membrane interactions by atomic force microscopy. Biochim Biophys Acta.
2004;1664(2):198–205.
83. Bukvic Krajacic M, Novak P, Dumic M, Cindric M, Paljetak HC, Kujundzic N.
Novel ureas and thioureas of 15-membered azalides with antibacterial
activity against key respiratory pathogens. Eur J Med Chem. 2009;44(9):
3459–70.
84. Wilson DW, Crabb BS, Beeson JG. Development of fluorescent Plasmodium
85.
86.
falciparum for in vitro growth inhibition assays. Malar J. 2010;9:152.
Lim C, Hansen E, DeSimone TM, Moreno Y, Junker K, Bei A, et al. Expansion
of host cellular niche can drive adaptation of a zoonotic malaria parasite to
humans. Nat Commun. 2013;4:1638.
Trager W, Jensen JB. Human malaria parasites in continuous culture.
Science. 1976;193(4254):673–5.
87. Boyle MJ, Wilson DW, Richards JS, Riglar DT, Tetteh KK, Conway DJ, et al.
Isolation of viable Plasmodium falciparum merozoites to define erythrocyte
invasion events and advance vaccine and drug development. Proc Natl
Acad Sci U S A. 2010;107(32):14378–83.
88. Wiegand I, Hilpert K, Hancock RE. Agar and broth dilution methods to
determine the minimal inhibitory concentration (MIC) of antimicrobial
substances. Nat Protoc. 2008;3(2):163–75.
Tham WH, Wilson DW, Lopaticki S, Schmidt CQ, Tetteh-Quarcoo PB, Barlow
PN, et al. Complement receptor 1 is the host erythrocyte receptor for
Plasmodium falciparum PfRh4 invasion ligand. Proc Natl Acad Sci U S A.
2010;107(40):17327–32.
89.
90. Creek DJ, Jankevics A, Burgess KE, Breitling R, Barrett MP. IDEOM: an excel
interface for analysis of LC-MS-based metabolomics data. Bioinformatics.
2012;28(7):1048–9.
91. Chong J, Soufan O, Li C, Caraus I, Li S, Bourque G, et al. MetaboAnalyst 4.0:
towards more transparent and integrative metabolomics analysis. Nucleic
Acids Res. 2018;46(W1):W486–w94.
92. Burns AL, Sleebs, BE., Siddiqui, G De Paoli, AE, Anderson, D, Liffner, B,
Harvey, R, Beeson, JG, Creek, DJ, Goodman, CD, McFadden, GI, Wilson, DW.
Project ID (ST001315) Retargeting azithromycin-like compounds as
antimalarials with dual modality. Metabolmics Workbench. 2020; doi:
https://doi.org/10.21228/M8CX0M.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
54. Olafson KN, Ketchum MA, Rimer JD, Vekilov PG. Mechanisms of hematin
crystallization and inhibition by the antimalarial drug chloroquine. Proc Natl
Acad Sci U S A. 2015;112(16):4946–51.
55. Dorn A, Vippagunta SR, Matile H, Jaquet C, Vennerstrom JL, Ridley RG. An
assessment of drug-haematin binding as a mechanism for inhibition of
haematin polymerisation by quinoline antimalarials. Biochem Pharmacol.
1998;55(6):727–36.
56. Olafson KN, Nguyen TQ, Rimer JD, Vekilov PG. Antimalarials inhibit hematin
crystallization by unique drug-surface site interactions. Proc Natl Acad Sci U
S A. 2017;114(29):7531–6.
57. Combrinck JM, Fong KY, Gibhard L, Smith PJ, Wright DW, Egan TJ.
Optimization of a multi-well colorimetric assay to determine haem species in
Plasmodium falciparum in the presence of anti-malarials. Malar J. 2015;14:253.
58. Bulusu V, Jayaraman V, Balaram H. Metabolic fate of fumarate, a side
59.
product of the purine salvage pathway in the intraerythrocytic stages of
Plasmodium falciparum. J Biol Chem. 2011;286(11):9236–45.
Ke H, Lewis IA, Morrisey JM, McLean KJ, Ganesan SM, Painter HJ, et al.
Genetic investigation of tricarboxylic acid metabolism during the
Plasmodium falciparum life cycle. Cell Rep. 2015;11(1):164–74.
60. Wang X, Miyazaki Y, Inaoka DK, Hartuti ED, Watanabe YI, Shiba T, et al.
Identification of Plasmodium falciparum mitochondrial malate: quinone
oxidoreductase inhibitors from the pathogen box. Genes. 2019;10(6): 471.
61. Rosenthal PJ. Azithromycin for malaria? Am J Trop Med Hyg. 2016;95(1):2–4.
62. Burns AL, Dans MG, Balbin JM, de Koning-Ward T, Gilson PR, Beeson JG,
et al. Targeting malaria parasite invasion of red blood cells as an
antimalarial strategy. FEMS Microbiol Rev. 2019;43(3):223–38.
63. Wilson DW, Langer C, Goodman CD, McFadden GI, Beeson JG. Defining the
timing of action of antimalarial drugs against Plasmodium falciparum.
Antimicrob Agents Chemother. 2013;57(3):1455–67.
64. Ashton TD, Devine SM, Mohrle JJ, Laleu B, Burrows JN, Charman SA, et al.
The development process for discovery and clinical advancement of
modern antimalarials. J Med Chem. 2019; 62(23): 10526-10562.
65. Coslédan F, Fraisse L, Pellet A, Guillou F, Mordmüller B, Kremsner PG, et al.
Selection of a trioxaquine as an antimalarial drug candidate. Proc Natl Acad
Sci U S A. 2008;105(45):17579–84.
66. Martin SK, Oduola AM, Milhous WK. Reversal of chloroquine resistance in
Plasmodium falciparum by verapamil. Science. 1987;235(4791):899–901.
67. Burgess SJ, Selzer A, Kelly JX, Smilkstein MJ, Riscoe MK, Peyton DH. A
chloroquine-like molecule designed to reverse resistance in Plasmodium
falciparum. J Med Chem. 2006;49(18):5623–5.
68. Oliveira R, Miranda D, Magalhães J, Capela R, Perry MJ, O'Neill PM, et al.
From hybrid compounds to targeted drug delivery in antimalarial therapy.
Bioorganic Med Chem. 2015;23(16):5120–30.
69. Burgess SJ, Kelly JX, Shomloo S, Wittlin S, Brun R, Liebmann K, et al.
Synthesis, structure-activity relationship, and mode-of-action studies of
antimalarial reversed chloroquine compounds. J Med Chem. 2010;53(17):
6477–89.
70. ClinicalTrials.gov. Safety and Pharmacokinetics Study of DM1157 to Treat
Malaria. (NCT03490162) NIH, U.S. National Library of Medicine: National
Institute of Allergy and Infectious Diseases (NIAID); 2018 (updated June 17,
2019). Available from: https://clinicaltrials.gov/ct2/show/NCT03490162.
Accessed July 2020.
71. Paloque L, Witkowski B, Lelièvre J, Ouji M, Ben Haddou T, Ariey F, et al.
Endoperoxide-based compounds: cross-resistance with artemisinins and
selection of a Plasmodium falciparum lineage with a K13 non-synonymous
polymorphism. J Antimicrob Chemother. 2018;73(2):395–403.
72. Cobbold SA, Chua HH, Nijagal B, Creek DJ, Ralph SA, McConville MJ.
Metabolic dysregulation induced in Plasmodium falciparum by
dihydroartemisinin and other front-line antimalarial drugs. J Infect Dis. 2016;
213(2):276–86.
73. Paljetak HC, Tomaskovic L, Matijasic M, Bukvic M, Fajdetic A, Verbanac D,
et al. Macrolide hybrid compounds: drug discovery opportunities in anti-
infective and anti-inflammatory area. Curr Top Med Chem. 2017;17(8):919–
40.
Juge N, Moriyama S, Miyaji T, Kawakami M, Iwai H, Fukui T, et al.
Plasmodium falciparum chloroquine resistance transporter is a H+-coupled
polyspecific nutrient and drug exporter. Proc Natl Acad Sci U S A. 2015;
112(11):3356–61.
74.
75. Martin RE, Marchetti RV, Cowan AI, Howitt SM, Broer S, Kirk K. Chloroquine
transport via the malaria parasite's chloroquine resistance transporter.
Science. 2009;325(5948):1680–2.
| null |
10.1088_1361-6587_acff7f.pdf
|
The data that support the findings of this study are openly
available at
the following URL/DOI. https://github.com/
YukiJajima/improved-spork/blob/main/NLD_data_CNN.
|
Data availability statement The data that support the findings of this study are openly available at the following URL/DOI. https://github.com/ YukiJajima/improved-spork/blob/main/NLD_data_CNN .
|
Plasma Phys. Control. Fusion 65 (2023) 125003 (8pp)
Plasma Physics and Controlled Fusion
https://doi.org/10.1088/1361-6587/acff7f
Estimation of 2D profile dynamics of
electrostatic potential fluctuations using
multi-scale deep learning
Yuki Jajima1, Makoto Sasaki1,∗, Ryohtaroh T Ishikawa2, Motoki Nakata2,3,
Tatsuya Kobayashi2, Yuichi Kawachi4 and Hiroyuki Arakawa5
1 College of Industrial Technology, Nihon University, Narashino 275-8575, Japan
2 National Institute for Fusion Science, Toki 509-5292, Japan
3 PRESTO, Japan Science and Technology Agency, Kawaguchi 332-0012, Japan
4 Department of Electronics, Kyoto Institute of Technology, Sakyo 606-8585, Japan
5 Faculty of Medical Sciences, Kyushu University, Fukuoka 812-8582, Japan
E-mail: [email protected]
Received 10 June 2023, revised 17 September 2023
Accepted for publication 3 October 2023
Published 26 October 2023
Abstract
Dynamics in magnetically confined plasmas are dominated by turbulence driven by spatial
inhomogeneities in density and temperature. Simultaneous measurement of velocity field and
density fluctuations is necessary to observe the particle transport, but the measurement of the
velocity field fluctuations is often challenging. Here, we propose a method to estimation
velocity field fluctuations from density fluctuations by using plasma turbulence simulations and
a deep technique learning. In order to take multi-scale characteristics into account, the several
number of spatial filters are used in the convolutional neural network. The velocity field
fluctuations are successfully predicted, and the particle transport estimated from the predicted
velocity field fluctuations is within 93.1% accuracy. The deep learning could be used for the
prediction of physical variables which are difficult to be measured.
Keywords: plasma turbulence, particle transport, deep learning
(Some figures may appear in colour only in the online journal)
1. Introduction
Dynamics in magnetically inhomogeneous plasmas are dom-
inated by turbulence, which is driven by spatial gradient of
density and temperature [1–3]. It is important to observe turbu-
lence transport for understanding the nature of fusion plasmas.
Simultaneous measurement of scalar and vector fields, such
as density, temperature, and velocity field vectors is neces-
sary to obtain turbulence transport [4]. However, the meas-
urement of the velocity field fluctuations is often challen-
ging due to the need for complex equipment [5–9] or various
assumptions [5, 10].
∗
Author to whom any correspondence should be addressed.
Recently, deep learning technique has been actively
developed in the area of image analysis [11]. It has been suc-
cessfully applied to the prediction of spatio-temporal dynam-
ics of turbulence on the Sun, which is difficult to measure [12].
In order to take multi-scale characteristics into account, the
several number of spatial filters are used in the convolutional
neural network, which is called multi-scale deep learning [13].
In this study, we apply this method to turbulence in mag-
netically confined plasmas. We propose a method to predict
the velocity field fluctuations from the density fluctuations
by the combination of plasma turbulence simulation and the
deep learning. We construct the network to estimate velo-
city field fluctuations, by learning the relation between dens-
ity and electrostatic potential fluctuations. As the first step, in
order to simplify the problem, we chose the turbulence in a
1361-6587/23/125003+8$33.00 Printed in the UK
1
© 2023 IOP Publishing Ltd
Plasma Phys. Control. Fusion 65 (2023) 125003
Y Jajima et al
linear device, where the small number of degrees of freedom
exist and the electron temperature fluctuation is often negli-
gibly small [14–16]. It is demonstrated that the quantitative
estimation of turbulence driven transport is possible using the
network.
The paper is organized as follows. In section 2, the tur-
bulence data used for the deep learning is introduced. In
section 3, the performance of the prediction is discussed.
Summary is given in section 4.
2. Prediction method of turbulence
2.1. Training data: turbulence simulation
Figure 1. Snap-shots of density(left) and electrostatics potential
fluctuations(right).
The turbulence simulation data is used for the deep learning to
predict the velocity field fluctuations. The simulation is based
on a reduced fluid model in cylindrical plasmas. The code is
called Numerical Linear Device [14, 17], which calculates the
nonlinear turbulence state such as resistive drift waves [14],
D’Angelo modes [17], and Kelvin-Helmholtz instabilities
[18–20]. The cylindrical coordinate (r, θ, z) is adopted as fol-
lows; the background density gradient is in r-direction, and the
magnetic field is in the z-direction. The nonlinear steady state
of the resistive drift waves is calculated with the following
set of parameters; the magnetic field B = 0.1 [T], the plasma
radius a = 10 [cm], the device length λ = 4 [m], the electron-
ion and ion-neutral collision frequencies normalized by the
ion cyclotron frequency νe = 510, νin = 0.035, and the set of
the viscosities for the density, the parallel flow and the vorti-
−4 respectively [17]. Here, the
city µN = µV = 10
−2, where ωci is the ion
viscosities are normalized by ωci
gyrofrequency and ρs is the gyroradius measured by the sound
velocity. The two-dimensional density and electrostatic poten-
tial patterns perpendicular to the magnetic field obtained by
the simulation are shown in figure 1. Density and electrostatics
potential are normalized by N = ln (ne/n0),ϕ = eϕ /Te, where
ne is the electron density, n0 is that at plasma center, and Te is
the electron temperature. For the density, azimuthally homo-
geneous background profile is formed. The fluctuations exists
in both density and electrostatic potential, which propagate in
the electron diamagnetic direction. Here, it is noted that for
the simplicity the electron temperature is treated as a fixed
parameter, Te = 3 [eV] , and the electron temperature fluctu-
ation is not considered. Actually, the electron temperature fluc-
tuation has been confirmed to be small in the corresponding
experiment [15, 16].
−2, µU = 10
−1ρs
The relation between the density and the electrostatic
potential pattern in nonlinear regime is learned by the neural
networks, which is mentioned in the next subsection. Here, it
is noted that in the linear regime, their relation becomes close
to the Boltzmann relation [4]. The time evolution of the fluctu-
ation pattern in the azimuthal direction is illustrated in figure 2.
In the nonlinearly saturated state, the data during 2000 < t <
3400 (the 11 periods of the dominant mode) is used for training
the network, and the period during 3450 < t < 3650 (approx-
imately two cycles of the dominant mode) is used for the
2
Figure 2. Time evolutions of azimuthal patterns of the density and
the electrostatic potential. The data during 2000 < t < 3400 is used
for the training, and the data during 3450 < t < 3650 is used for the
prediction.
prediction. Here, the modulation of the azimuthal profile of
the fluctuation corresponds to the formation of the streamer
[14, 17].
2.2. Multi-scale convolutional neural network
A multi-scale convolutional neural network model [13], which
was originally developed for estimating hardly-observable tur-
bulent fields on the solar surface, is applied to this study as
shown in figure 3. The network is constructed to estimate the
spatial distribution of electrostatic potential fluctuations (out-
put) from density fluctuations (input).
The multi-scale convolutional neural network includes fil-
ters with various sizes, in order to detect spatially localized
patterns and global structures simultaneously. Five different
sizes of filters are used: 3 × 3,7 × 7,15 × 15,31 × 31, and
51 × 51,where the first and the second numbers correspond to
the number of pixels in r and θ directions, respectively. The
size of the input image is 86 × 64 pixels. The multiple filter
size is effective for detecting eddies with various sizes. These
Plasma Phys. Control. Fusion 65 (2023) 125003
Y Jajima et al
Figure 3. Structure of multi-scale convolutional neural network [13]. Here, Conv, BN,SE denote convolution layer, batch normalization
layer and squeeze-and-excitation, respectively.
Figure 4. Snap-shots of electrostatic potential fluctuations, Left: Prediction result, Right: Simulation data(answer).
filters are used to perform convolution for the spatial and tem-
poral axes. For the temporal convolution, 3 time steps are used
to consider time variations.
Next, the structure of the network is explained. In the first
convolution layer shown in the left block in figure 3, a three-
dimensional convolution along the spatial and temporal axes
is performed, resulting in a four-dimensional array containing
the spatial axis, the temporal axis, and the number of filters.
In the final block, shown in the right block figure 3, convolu-
tion is performed only on the spatial axis. We include 1 × 1
bottle-neck layer before each convolution layer [11]. The loss
function is the mean squared error, and the learning is per-
formed to minimize it. Adaptive Moment Estimation (Adam)
is used for the optimization of this network [21]. The learn-
ing rate is selected as η = 0.001. The number of frames for
the training, the validation, and the test are 280, 40, and 40,
respectively. Here each data set was normalized to have a mean
of 0 and a standard deviation of 1 for each physical quantity.
The number of epochs is 20, and the batch size is 10. Here,
the number of test data is checked to be sufficient for the pre-
diction. Under the above conditions, the network learns the
relationship between the input and output images of the train-
ing data. Here, it is noted that the estimation of the potential
is performed by using the density data at the different tem-
poral phase from that used for training. The background dens-
ity profile is not extracted for the training. Thus, the neural
network predicts the background profile, the fluctuation amp-
litudes, and their phase relations.
3. Prediction of turbulence and particle transport
3.1. Spatio-temporal structure estimation of electrostatic
potential dynamics
The prediction of the time evolution of the electrostatic poten-
tial fluctuations is performed by the trained network with the
time trace of the 2D density distribution. The result is shown
in figure 4. The left panel is the spatial distribution predicted
by the network, and the right panel is that obtained by the sim-
ulation. Both structures are in good agreement in the radial
and azimuthal directions. The correlation coefficient includ-
ing time evolution is 0.98. The prediction is successful with
high accuracy.
Next,
the azimuthal spectra are evaluated to quantify
the accuracy of
the estimation. The azimuthal Fourier
mode decomposition of the electrostatic potential fluctuation,
ϕ (r, θ, t), is given as
ϕ (r, θ, t) =
X
m
ϕ m (r, t) eimθ,
(1)
where m is the number of the azimuthal mode number. Figure 5
shows the azimuthal mode spectrum at r/a = 0.58, where
3
Plasma Phys. Control. Fusion 65 (2023) 125003
Y Jajima et al
Figure 5. Azimuthal mode spectrum of electrostatic potential fluctuations where the blue and red lines correspond to the spectrum of the
predicted result, and that of the simulation data.
Figure 6. (a) Coherence Cm between estimated results and simulation data. (b) Phase angle ψ m between estimated results and simulation
data.
a is the plasma radius. The dominant modes are m = 3,5.
The azimuthal mode spectrum of the simulated and predicted
p (t), respectively. The
data are introduced as ϕ m
coherence and the phase angle between them are calculated as
follows.
a (t) and ϕ m
2
3
Cm = Re
(cid:10)
6
6
4
rD
ϕ m
p(t)
∗
ϕ m
E D
a (t)
(cid:11)
7
7
5 ,
E
|ϕ m
p (t)|2
|ϕ m
a (t)|2
ψ m =
−1
tan
1
π
(cid:2)(cid:10)
(cid:2)(cid:10)
Re
Im
∗
∗
ϕ m
ϕ m
p(t)
p(t)
ϕ m
ϕ m
a (t)
a (t)
!
(cid:11)(cid:3)
(cid:11)(cid:3)
,
(2)
(3)
Figure 7. Radial profiles of background densities in cases of
different collisional diffusion µN = 0.01, 0.012, 0.013, 0.02.
∗
where ⟨ ⟩ denotes the time average and
denotes the complex
conjugate. The coherence and the phase angle for each mode
number is illustrated in figure 6. The coherence for the dom-
inant modes is C3 = 0.998 and C5 = 0.996 for m = 3 and 5,
respectively. The phase angles have values of ψ 3 =−0.0008
and ψ 5 =−0.0157, which indicate good temporal and spatial
agreement between simulated and predicted data. The char-
acteristics of the prediction accuracy with the change of the
learning rate and the filter size are described in the appendix.
Here, we consider simulations with different collisional
diffusion coefficients µN = 0.01, 0.012, 0.013, 0.02, where
the above result corresponds to µN = 0.01. Depending on
the change of the parameter, the background density pro-
file and the fluctuation pattern changes as shown in figures 7
and 8. First, we show the prediction result for the tur-
bulence with different parameters by using the network
trained for a single data set. The network is trained by the
4
Plasma Phys. Control. Fusion 65 (2023) 125003
Y Jajima et al
Figure 8. Snapshots of electrostatic potential in cases of µN = 0.01, 0.012, 0.013, 0.02.
Figure 9. Correlation coefficient of electrostatic potentials by the prediction and that by simulation. Cases with training data of the single
and multiple data set are compared.
simulation data with µN = 0.01. The obtained network is used
for the prediction of the electrostatic potential patterns for
µN = 0.012, 0.013, 0.02, which are not used for the training.
Here, the numbers of the images for the training, validation
and test are 280, 40, and 40, respectively, and the epoch num-
ber is 40. Even in this case, the prediction is successful with the
correlation coefficient between the prediction and the answer
is around 0.9, as shown in the blue line of figure 9. Next,
the prediction is performed by the network trained by using
the multiple data set µN = 0.01, 0.012, 0.013, 0.02. The set
of the training data is created by mixing the turbulence data
obtained from the multiple simulations with different µN. The
numbers of the images for the training, validation and test are
280 × 4 = 1120, 40, and 40, respectively, and the epoch num-
ber is 80. The prediction quality improves compared with the
case by the single training data, which is shown by the red line
5
Plasma Phys. Control. Fusion 65 (2023) 125003
Y Jajima et al
to the experiments, it is necessary to learn the relation between
the diagnostic signal and the potential fluctuation, where the
diagnostic signal can be demonstrated by using the turbulence
simulation.
4. Summary
In this study, the electrostatic potential fluctuations are pre-
dicted from density fluctuations, using the convolutional
neural network. Simulated data of resistive drift wave turbu-
lence in linear plasmas is used as the training data. To account
for the multi-scale characteristics of turbulence, the multi-
scale convolutional neural network including multiple filters
is employed to construct a network that calculates the electro-
static potential fluctuations as output with the input of density
fluctuations. The time evolution of the electrostatic potential
fluctuation is predicted by using the density fluctuation from
time to time. The prediction is successful, with the correla-
tion coefficient of 0.98. It is also shown that the quantitative
evaluation of particle transport is possible from the estimated
electrostatic potential fluctuation. This study is the first step
for the prediction of the unobservable quantities, and thus the
problem is simplified as much as possible. Applications of the
proposed method to more complicated cases, such as turbu-
lence in fusion devices or synthetic diagnostics, are the future
work.
Data availability statement
Figure 10. Turbulence particle transport driven by each azimuthal
mode, where the blue and red lines are the estimated turbulent
particle transport, and that of the simulation data.
of figure 9. Thus, the proposed method is robust, because it can
estimate the electrostatic potential with high accuracy under a
wide range of parameter conditions which are not used for the
training.
3.2. Estimation of turbulence-driven particle transport
The particle transport is calculated from the predicted electro-
static potential fluctuations. The particle transport Γ is calcu-
lated as follows
Γ = − 1
r
X
m
Im [mnmϕ m
∗]
(4)
The data that support the findings of this study are openly
available at
the following URL/DOI. https://github.com/
YukiJajima/improved-spork/blob/main/NLD_data_CNN.
where nm is the mth mode of the density fluctuation [4]
figure 10 shows the particle transport driven by each mode at
r/a = 0.58. For the dominant mode, the prediction and the
simulation agree very well. The sum of the transport from
each mode is found to be estimated in 93.1% accuracy. For
the modes with the high wave numbers, the prediction is not
so good. Since the energy of high wave number modes are
very small, which corresponds to 0.1% of the total energy,
the total transport is predicted with high accuracy. Using the
estimated electrostatic potential fluctuations, quantitative eval-
uation of particle transport is possible. In this study, we use
the simulation data in which the electron temperature fluctu-
ation is neglected. When one considers plasmas under the elec-
tron heating, the electron temperature gradient is formed, and,
as a result, the electron temperature fluctuation becomes sig-
nificant. In this case, only the density may not be sufficient
for the prediction of the electrostatic potential fluctuation. The
extension of the network structure would be necessary for this
case, which is a future work. Furthermore, the method could be
applied to the experimental data. In order to apply this method
Acknowledgments
This work was partly supported by JSPS Kakenhi Grant
Nos. JP21K03509, JP21K03513,
the collaboration pro-
grams of NIFS (NIFS22KIPH015, NIFS22KIST019), Kyushu
University RIAM (2022S2-CD-1), and the Nihon University
Fund for Supporting Young Scientists.
Appendix. Characteristics of learning performance
The learning performance with regards to the learning rates
η is investigated. To obtain the statistical performance of the
learning of the network parameters on initial values, ten pre-
dictions with different initial value are performed, where the
initial values are given as random values. The average over
the ensembles is shown as the thick line. The loss function
becomes minimum at the learning rate η is 0.001. The depend-
ence on the filter size is illustrated in figure 11. For the low
6
Plasma Phys. Control. Fusion 65 (2023) 125003
Y Jajima et al
Figure 11. Filter dependence: The red, green, blue, light blue, and purple lines are the values of the evaluation function in the number of
modes when the filter size is 3 × 3,7 × 7,15 × 15,31 × 31, and 51 × 51, respectively, and the black line is the value of the evaluation
function when training with all the above filters.
Figure 12. Learning rate dependence: the blue and red lines correspond to the convergence value of the loss function, and that of the
validation loss function.
mode number (m ⩽ 5), the prediction is successful independ-
ent on the filter size. For the mode with m ⩾ 5, the perform-
ance depends on the filter size and number. The learning
with the multi-filters is the best for the prediction of all the
modes.
Learning performance of multi-scale convolutional neural
networks changes, depending on the learning rate and the
size/number of filters. We investigate how the performance
changes in the cases with a single filter size and with the mul-
tiple filter sizes.
the
We
compare
cases with each different
size
(3 × 3,7 × 7,15 × 15,31 × 31, and 51 × 51), and with all the
above filters. To evaluate the learning performance under each
condition, we define the error ε as follows.
(cid:12)
(cid:12)
(cid:12)
(cid:12)
p
a
ϕ m
ϕ m
2
(cid:12)
(cid:12)
(cid:12)
(cid:12)
ε =
− 1,
(5)
a and ϕ m
p denote the simulated and predicted data
where ϕ m
of the azimuthal mode spectrum at r/a = 0.58 respectively.
Figure 12 shows the error at each mode number. It can be seen
that the performance is higher when multiple filter sizes are
used.
ORCID iDs
Makoto Sasaki https://orcid.org/0000-0001-6835-1569
Ryohtaroh T Ishikawa https://orcid.org/0000-0002-4669-
5376
7
Plasma Phys. Control. Fusion 65 (2023) 125003
Y Jajima et al
Motoki Nakata https://orcid.org/0000-0003-2693-4859
Yuichi Kawachi https://orcid.org/0000-0002-5222-6082
Hiroyuki Arakawa https://orcid.org/0000-0001-9793-
099X
References
[1] Wagner F 2007 Fritz Phys. Control. Fusion 49 B1
[2] Diamond P H, Itoh S-I, Itoh K and Hahm T S 2005 Plasma
Phys. Control. Fusion 47 R35
[3] Terry P W 2000 Rev. Mod. Phys. 72 109
[4] Horton W 1999 Rev. Mod. Phys. 71 735
[5] Bretz N 1997 Rev. Sci. Instrum. 68 2927–64
[6] McKee G R, Fonck R J, Gupta D K, Schlossberg D J,
Shafer M W, Holland C and Tynan G 2004 Rev. Sci.
Instrum. 75 3490–2
[7] Schmitz L, Wang G, Hillesheim J C, Rhodes T L,
Peebles W A, White A E, Zeng L, Carter T A
and Solomon W 2008 Rev. Sci. Instrum.
79 10F113
[8] Ido T, Shimizu A, Nishiura M, Hamada Y, Kato S,
Nishizawa A and Nakano H 2007 Plasma Fusion Res.
2 S1100
[9] Hillesheim J C, Peebles W A, Rhodes T L, Schmitz L,
Carter T A, Gourdain P-A and Wang G 2009 Rev. Sci.
Instrum. 80 083507
[10] Nagashima Y et al 2011 Rev. Sci. Instrum. 82 033503
[11] He K, Zhang X, Ren S and Sun J 2016 Proc. CVPR pp 770–8
[12] Ramos A, Requerey I S and Vitas N 2017 Astron. Astrophys.
604 A11
[13] Ishikawa R T, Nakata M, Katsukawa Y, Masada Y and
Riethmüller T L 2022 Astron. Astrophys. 658 A142
[14] Kasuya N, Yagi M, Itoh K and Itoh S-I 2008 Phys. Plasmas
15 052302
[15] Oldenburger S, Uriu K, Kobayashi T, Inagaki S, Sasaki M,
Nagashima Y, Yamada T, Fujisawa A, ITOH S-I and
ITOH K 2012 Plasma Fusion Res. 7 2401146
[16] Kawashima K et al 2011 Plasma Fusion Res. 6 2406118
[17] Sasaki M et al 2017 Phys. Plasmas 24 112103
[18] Sasaki M, Camenen Y, Escarguel A, Inagaki S, Kasuya N,
Itoh K and Kobayashi T 2019 Phys. Plasmas 26 042305
[19] Sasaki M, Kawachi Y, Dendy R O, Arakawa H, Kasuya N,
Kin F, Yamasaki K and Inagaki S 2019 Plasma Phys.
Control. Fusion 61 112001
[20] Sasaki M, Kobayashi T, Dendy R O, Kawachi Y, Arakawa H
and Inagaki S 2020 Plasma Phys. Control. Fusion
63 025004
[21] Kingma D P and Jimmy B 2014 arXiv:1412.6980
8
| null |
10.1007_s11538-020-00756-5.pdf
| null | null |
Bulletin of Mathematical Biology (2020) 82:74
https://doi.org/10.1007/s11538-020-00756-5
O R I G I N A L P A P E R
Population Dynamics with Threshold Effects Give Rise
to a Diverse Family of Allee Effects
Nabil T. Fadai1
· Matthew J. Simpson2
Received: 16 April 2020 / Accepted: 27 May 2020 / Published online: 12 June 2020
© The Author(s) 2020
Abstract
The Allee effect describes populations that deviate from logistic growth models and
arises in applications including ecology and cell biology. A common justification for
incorporating Allee effects into population models is that the population in question
has altered growth mechanisms at some critical density, often referred to as a threshold
effect. Despite the ubiquitous nature of threshold effects arising in various biological
applications, the explicit link between local threshold effects and global Allee effects
has not been considered. In this work, we examine a continuum population model
that incorporates threshold effects in the local growth mechanisms. We show that this
model gives rise to a diverse family of Allee effects, and we provide a comprehensive
analysis of which choices of local growth mechanisms give rise to specific Allee
effects. Calibrating this model to a recent set of experimental data describing the
growth of a population of cancer cells provides an interpretation of the threshold
population density and growth mechanisms associated with the population.
Keywords Logistic growth · Per-capita growth rate · Population dynamics ·
Population models
1 Introduction
Mathematical models of population dynamics often include an Allee effect to account
for dynamics that deviate from logistic growth (Stephens et al. 1999; Allee and Bowen
1932; Courchamp et al. 1999; Taylor and Hastings 2005; Courchamp et al. 2008). The
Electronic supplementary material The online version of this article (https://doi.org/10.1007/s11538-
020-00756-5) contains supplementary material, which is available to authorized users.
B Nabil T. Fadai
[email protected]
1
2
School of Mathematical Sciences, University of Nottingham, Nottingham NG7 2RD, UK
School of Mathematical Sciences, Queensland University of Technology, Brisbane, QLD 4001,
Australia
123
74
Page 2 of 22
N. T. Fadai, M. J. Simpson
logistic growth model (Table 1, Fig. 1) describes the growth rate, dC(t)/dt, as a
quadratic function of density, C(t), at time t ≥ 0. The logistic growth model has two
equilibria: C ∗ = 0 and C ∗ = K , where an equilibrium is any value C ∗ such that
dC(t)/dt = 0 when C(t) ≡ C ∗. Since densities near C(t) ≡ K will approach K ,
while densities near C(t) ≡ 0 diverge away from zero (Fig. 1), we say that C ∗ = K
is a stable equilibrium, while C ∗ = 0 is an unstable equilibrium. This means that
the logistic growth model implicitly assumes that all densities, no matter how small,
eventually thrive.
Mathematical models that include an Allee effect relax the assumption that all
population densities will thrive and survive, which is inherent in logistic growth mod-
els (Murray 2003; Edelstein-Keshet 2005; Stephens et al. 1999; Taylor and Hastings
2005; Courchamp et al. 2008). Consequently, populations described using Allee effect
models exhibit more complicated and nuanced dynamics, including reduced growth
at low densities (Neufeld et al. 2017; Johnson et al. 2006; Gerlee 2013) and extinc-
tion below a critical density threshold (Courchamp et al. 1999; Allee and Bowen
1932; Taylor and Hastings 2005; Courchamp et al. 2008). The phrase Allee effect can
have many different interpretations in different parts of the literature. For instance,
the Weak Allee effect (Table 1, Fig. 1) is used to describe density growth rates that
deviate from logistic growth, but do not include additional equilibria (Murray 2003;
Edelstein-Keshet 2005; Taylor and Hastings 2005; Stephens et al. 1999). A common
mathematical description of the Weak Allee effect is shown in Table 1, where the
factor 1 + C(t)/A represents the deviation from the classical logistic growth model.
Despite the similarity between logistic growth and the Weak Allee effect, it is not
possible to write down an explicit solution for Weak Allee effect in terms of C(t), like
we can for logistic growth. Despite this, we are still able to examine the equilibria of
the Weak Allee effect to understand its salient features. Since A > 0, the Weak Allee
effect does not incorporate any additional equilibria other than C ∗ = 0 and C ∗ = K .
Noting that the main feature of an Allee effect is a deviation from logistic growth, the
cubic representation of the growth rate shown in Table 1 is employed predominantly
for simplicity rather than explicit biological significance (Taylor and Hastings 2005;
Stefan et al. 2012; Stephens et al. 1999). Therefore, in this work, we refer to the Weak
Allee effect as any population density growth rate that deviates from logistic growth
without incorporating additional equilibria.
Another common type of Allee effect is the Strong Allee effect (Table 1, Fig. 1), in
which an additional unstable intermediate equilibrium, C ∗ = B, with 0 < B < K ,
Table 1 Typical mathematical descriptions of logistic growth, the Weak Allee effect, and the Strong Allee
effect
Typical mathematical description
(cid:2)
(cid:3)
dC(t)
dt
dC(t)
dt
dC(t)
dt
= rC(t)
= rC(t)
= rC(t)
(cid:2)
(cid:2)
1 − C(t)
K
1 − C(t)
K
1 − C(t)
K
(cid:3) (cid:2)
(cid:3) (cid:2)
(cid:3)
(cid:3)
1 + C(t)
A
C(t)
B
− 1
Notes
r > 0, K > 0
r > 0, A > 0, K > 0
r > 0, 0 < B < K
Effect
Logistic growth
Weak Allee
Strong Allee
123
Population Dynamics with Threshold Effects Give Rise to a…
Page 3 of 22
74
i
c
i
t
s
g
o
L
C
d
t
d
0
0
0
0
0
0
K
K
C
C
B
C
K
C
d
t
d
C
d
t
d
e
e
l
l
A
k
a
e
W
e
e
l
l
A
g
n
o
r
t
S
)
t
(
C
)
t
(
C
)
t
(
C
K
0
K
0
K
B
0
t
t
t
Fig. 1 Comparison of typical logistic growth, Weak Allee, and Strong Allee models. The mathematical
descriptions of the three models are shown in Table 1
is incorporated (Murray 2003; Edelstein-Keshet 2005; Taylor and Hastings 2005;
Stephens et al. 1999; Courchamp et al. 1999). In a similar fashion to the Weak Allee
effect, the cubic form of the Strong Allee effect (Table 1) is chosen predominantly
for simplicity (Taylor and Hastings 2005; Stefan et al. 2012; Stephens et al. 1999).
Therefore, we will refer to any growth rate with two stable equilibria, C ∗ = 0 and
C ∗ = K , and an additional intermediate unstable equilibrium as the Strong Allee
effect. Throughout this work, we refer to growth rates that deviate from logistic growth
as an Allee effect, whereas specific Allee effects (e.g. the Weak Allee effect and the
Strong Allee effect) are referred to using more specific terminologies.
123
74
Page 4 of 22
N. T. Fadai, M. J. Simpson
While Allee effects were originally used to describe population dynamics arising
in ecology (Taylor and Hastings 2005; Tu et al. 2019; Courchamp et al. 1999; Johnson
et al. 2006; Simberloff et al. 2013; Seebens et al. 2017; Drake 2004; Courchamp et al.
2008), there has been increasing interest in examining the potential for Allee effects in
population dynamics relating to cell biology (Neufeld et al. 2017; Böttger et al. 2015;
Gerlee 2013; Sarapata and de Pillis 2014; Jenner et al. 2018, 2019; Bobadilla et al.
2019; Johnston et al. 2017; Jin et al. 2017; Johnson et al. 2019; de Pillis et al. 2005;
de Pillis and Radunskaya 2003). In both cell biology and ecological applications,
the Allee effect provides a suitable modelling framework to describe the dynamics
of well-mixed populations that exhibit non-logistic features. However, because stan-
dard models incorporating Allee effects are continuum models that describe global,
population-level features of the population dynamics, the interpretation of Allee effects
at the individual scale remains less clear (Johnston et al. 2017; Böttger et al. 2015).
Understanding how local, stochastic growth mechanisms give rise to global Allee
effects in a population is important, since these individual-level mechanisms can ulti-
mately determine whether a population will survive or be driven to extinction (Johnston
et al. 2017; Scott et al. 2014; Colon et al. 2015; Böttger et al. 2015). Certain individual-
level biological features are ubiquitous among populations displaying Allee effects,
providing a unifying feature in both cell biology and ecological applications. One of
these phenomena is a threshold effect (Frankham 1995; Rossignol et al. 1999; Metzger
and Décamps 1997), which we also refer to as a binary switch. We define a binary
switch as a local feature of a population that behaves differently when a particular bio-
logical mechanism is present or absent. Some examples of binary switches include: the
go-or-grow hypothesis in cell biology (Hatzikirou et al. 2012; Vittadello et al. 2020),
phenotypic plasticity (Friedl and Alexander 2011; Böttger et al. 2015), tree mast-
ing (Koenig and Knops 2005), external harvesting pressure (Courchamp et al. 1999;
Kuparinen et al. 2014), density-dependent clustering (Martínez-García et al. 2015),
and resource depletion (Hopf and Hopf 1985). For all of these examples, Allee effects
have been proposed to potentially explain more complicated and nuanced population
dynamics than are possible in a logistic growth framework. However, the link between
the details of such a local binary switch and the resulting population-level Allee effect
is unclear. Given that local binary switches are thought to be widely important in biol-
ogy and ecology, we ask two questions: (i) how does the incorporation of a local binary
switch in proliferation and death rates affect the global dynamics of a population? and
(ii) how does this local binary switch relate to different forms of Allee effects?
In this work, we show that incorporating local-level binary switches in a contin-
uum, population-level mathematical modelling framework gives rise to a surprisingly
diverse family of Allee effects. Some switches in proliferation and death rates give
rise to established Allee effects, whereas other binary switches lead to more gener-
alised Allee effects that have not been previously reported. We show that incorporating
local-level binary switches in proliferation and death rates leads to a diverse family
of Allee effects with only a few model parameters. This model, which we refer to as
the Binary Switch Model, captures key biological features, but continues to exhibit
the same qualitative features as various Allee effects. We conclude by applying the
Binary Switch Model to a recent cell biology data set. Interpreting this data with our
123
Population Dynamics with Threshold Effects Give Rise to a…
Page 5 of 22
74
a
b
M = 2
n
o
i
t
a
r
e
f
i
l
o
r
P
s
e
t
a
R
R
r
0
1
2
3
4
5
6
c
h
t
a
e
D
s
e
t
a
R
Rβ
rα
0
1
2
3
4
5
6
n
n
Fig. 2 Schematic for the Binary Switch Model. Individuals in a population a can sense nearby individuals,
providing a simple measure of local density. Individuals who sense higher than a threshold density, M,
are shown in blue, while more isolated individuals are shown in red. This threshold density determines the
constant rates at which individuals proliferate and die. b, c The binary switch shown here occurs when
individuals can sense more than M = 2 neighbours
modelling framework suggests that the observed growth is non-logistic and that the
phenomena are best explained by a binary switch at low density.
2 The Binary Switch Model
We consider an individual-based model framework that incorporates individual-level
growth mechanisms varying with local population density to describe the temporal
evolution of the global population density. One framework incorporating these afore-
mentioned features is the stochastic agent-based model framework that we proposed
in Fadai et al. (2020), in which individuals of the same size move, die, and proliferate
on a two-dimensional hexagonal lattice. This discrete model incorporates exclusion
(crowding) effects, allowing the population density to saturate at a finite capacity, as
well as proliferation and death rates that vary with the local population density. While
local population density can be measured in many different ways, Fadai et al. (2020)
take the simplest approach and use the number of nearest neighbours as a measure of
local density (Fig. 2).
As the individual dynamics of the stochastic agent-based model are difficult to
analyse mathematically, we examine the continuum limit per-capita growth rate as
a means of representing the average dynamics of the spatially uniform population,
noting that there is good agreement between these two modelling approaches (Fig. 3).
Full details of the discrete–continuum comparison are summarised in the Supplemen-
tary Information. Since the average population dynamics obtained from the discrete
stochastic individual-based model agree well with its continuum description (Fig. 3),
we will only consider the features of the continuum description of the model, whose
per-capita growth rate is given by
123
74
Page 6 of 22
N. T. Fadai, M. J. Simpson
1
C(t)
dC(t)
dt
= (1 − C(t))
(cid:5)
5
n
γn
5(cid:4)
n=0
(cid:6)
C(t)n (1 − C(t))5−n − γ6C(t)6,
where
(cid:7)
γn =
,
pn − 6dn
6−n
d6,
n = 0, . . . , 5,
n = 6.
(1)
(2)
Here, C(t) is the population density at time t, while pn and dn are the proliferation and
death rates that vary with the number of nearest neighbours, n (Fadai et al. 2020). The
parameter grouping γn can be interpreted as the net growth mechanism for a particular
local population density. Noting that C(t) ≡ 1 represents the maximum packing
density, we have C(t) ∈ [0, 1]. Equation (1) has a thirteen-dimensional parameter
space: namely, Θ = ( p0, . . . , p5, d0, . . . , d6).
1
n
p
0.5
0.2
n
d
0.1
a
2
1
n
p
0.2
n
d
0.1
b
1
n
p
0.5
0.2
n
d
0.1
c
)
t
(
C
)
t
(
C
)
t
(
C
1
0.8
0.6
0.4
0.2
0
0
1
0.8
0.6
0.4
0.2
0
0
1
0.8
0.6
0.4
0.2
0
0
n
n
n
n
n
n
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
C
d
t
d
C
d
t
d
C
d
t
d
0.4
0.3
0.2
0.1
0
-0.1
-0.2
0
0.4
0.3
0.2
0.1
0
-0.1
-0.2
0
0.4
0.3
0.2
0.1
0
-0.1
-0.2
0
2
4
6
8
10
t
2
4
6
8
10
t
2
4
6
8
10
t
0.2
0.4
C
0.6
0.8
1
0.2
0.4
C
0.6
0.8
1
0.2
0.4
C
0.6
0.8
1
Fig. 3 a When no binary switch is present, (1) reduces to logistic growth. b, c When a binary switch is
incorporated in proliferation and death rates (M = 2), the continuum limit is no longer logistic. In all of
these parameter regimes, the average density data determined from discrete model simulations, shown in
red dashed curves in the middle column (Supplementary Information), agrees well with the continuum limit
predictions (4), shown in black solid curves. Density growth rates in the right-most column show that (a)
is logistic, while (b, c) are not
123
Population Dynamics with Threshold Effects Give Rise to a…
Page 7 of 22
74
We incorporate a binary switch into (1) by choosing
(cid:7)
r ,
R,
pn =
n = 0, . . . , M,
n = M + 1, . . . , 6,
dn =
(cid:7)
r α,
Rβ,
n = 0, . . . , M,
n = M + 1, . . . , 6.
(3)
This choice of parameters means that we have the proliferation rate pn = r when the
local density is at or below the critical density M, or pn = R when the local density is
above M. We refer to M ∈ {0, 1, 2, 3, 4, 5} as the threshold density. For simplicity, we
assume that the death rates are a particular fraction of the proliferation rates: α ∈ [0, 1]
and β ∈ [0, 1]. It is useful to note that (1)–(3) relaxes to the classical logistic growth
model, for any choice of M ∈ {0, 1, 2, 3, 4, 5} by setting r = R and α = β (Fig. 3a).
By substituting (3) into (1), we obtain the Binary Switch Model,
1
C(t)
dC(t)
dt
= r
(cid:5)
M(cid:4)
j=0
5
j
(cid:6)
C(t) j (1 − C(t))6− j
(cid:9)
(cid:8)
1 − 6α
6 − j
− RβC(t)6
+ 1(M ≤ 4) · R
(cid:6)
C(t) j (1 − C(t))6− j
(cid:8)
1 − 6β
6 − j
(cid:9)
,
5(cid:4)
j=M+1
(cid:5)
5
j
(cid:7)
1(M ≤ 4) =
1, M ≤ 4,
0, M = 5,
(4)
(5)
where
is an indicator function. The Binary Switch Model shows, for the first time, how a
local binary switch in individual-level proliferation and death rates leads to a particular
global density growth rate. A summary of parameters and their particular biological
interpretation is shown in Table 2. In particular, we note that the Binary Switch Model
reduces the thirteen-dimensional parameter space in (1) to a five-dimensional param-
eter space: Θ = (r , R, α, β, M). This reduced parameter space means that the Binary
Switch Model can be used with less risk of over-fitting than (1) (Warne et al. 2019).
We will discuss further merits of this reduced parameter space when calibrating the
Binary Switch Model to experimental data in Sect. 3.
In Fig. 3, we show how the Binary Switch Model gives rise to non-logistic growth
mechanisms. When no binary switch is present (Fig. 3a), the growth mechanisms
Table 2 Summary of parameters used in the Binary Switch Model
Parameter
Biological interpretation
r ∈ [0, ∞)
R ∈ [0, ∞)
α ∈ [0, 1]
β ∈ [0, 1]
M ∈ {0, 1, 2, 3, 4, 5}
Low-density proliferation rate
High-density proliferation rate
Ratio of low-density death rate to low-density proliferation rate
Ratio of high-density death rate to high-density proliferation rate
Threshold density
123
74
Page 8 of 22
N. T. Fadai, M. J. Simpson
are independent of local density and assume a single proliferation and death rate,
resulting in logistic growth. However, when a binary switch is incorporated into the
proliferation and death rates (Fig. 3b, c), the population dynamics described by (4)
deviates from the classical logistic growth model. Consequently, we now wish to
examine the various kinds of Allee effects the Binary Switch Model can give rise to.
The main qualitative differences between logistic growth and various Allee effects are
based on the number of equilibria and their stability; therefore, we now examine the
roots of (4) for various parameter values. In all parameter regimes considered in the
work, the zero equilibrium, C ∗ = 0, will always be present. Additional equilibria, if
present, will be denoted as C ∗ = Ci ∈ (0, 1], where i = 1, 2, ... and are ordered such
that Ci < Ci+1 for all i. Since the right-hand side of (4) is a sixth-degree polynomial,
a maximum of six equilibria can be present in (0, 1], but explicit expressions for the
solutions of the polynomial cannot be determined in general. We will show that in
the Binary Switch Model, a maximum of three equilibria can be present in (0, 1].
Setting r = 0 and R > 0 (Case 1) or R = 0 and r > 0 (Case 2), we will show
that fewer equilibria are present in (0, 1]. In Case 3, corresponding to r > 0 and
R > 0, certain combinations of parameter values produce equilibria with additional
qualitative features, such as double-root and triple-root equilibria. For these special
equilibria, we will designate particular symbols to Ci , which appear as required.
2.1 Case 1: r = 0 and R > 0
This case corresponds to situations where individuals below the threshold density M
do not proliferate or die. We will now show that in Case 1, either no equilibria are
present in (0, 1], or we have one equilibrium C1 ∈ (0, 1], depending on the choice of
β and M. In this regime, (4) simplifies to
1
RC(t)
dC(t)
dt
= S(C(t); β, M)
:= −βC(t)6
+ 1(M ≤ 4) ·
(cid:5)
5
j
5(cid:4)
j=M+1
(cid:6)
C(t) j (1 − C(t))6− j
(cid:8)
(cid:9)
.
1 − 6β
6 − j
(6)
Since β appears as a linear coefficient in (6), it is easier to solve S(C1, β, M) = 0 for
β than for C1. The resulting relationship between C1 and β depends on the integer
value of M ∈ {0, 1, 2, 3, 4, 5}; however, a general solution in terms of arbitrary M is
difficult to obtain. Instead, we define the family of functions, f M (C1), for a particular
value of M, such that
β = f M (C1) ⇐⇒ S(C1, f M (C1), M) = 0.
(7)
Using f M (C1), we determine the unique value of β that solves S(C1, β, M) = 0 for
a given value of C1 ∈ (0, 1], shown in Table 3. Plotting β = f M (C1) for all M ∈
{0, 1, 2, 3, 4, 5} and C1 ∈ (0, 1] indicates that f M (C1) is one-to-one on C1 ∈ (0, 1].
123
Population Dynamics with Threshold Effects Give Rise to a…
Page 9 of 22
74
Table 3 Relationships between the nonzero equilibrium of the Binary Switch Model, C1, to β and M for
Case 1 when r = 0 (6)
M
0
1
2
3
4
5
β = f M (C1)
Range of β : C1 ∈ (0, 1]
+15C1−5
+15C1−6
−20C2
1
−20C2
1
−30C1+10
−40C1+15
−25C1+10
−45C1+20
+15C3
1
+15C3
1
+35C2
1
+45C2
1
C5
−6C4
1
1
C5
−6C4
1
1
4C4
−19C3
1
1
5C4
−24C3
1
1
−6C3
+21C2
1
1
−10C3
+36C2
1
1
4C2
−9C1+5
1
10C2
−24C1+15
1
−C1+1
−5C1+6
0
β ∈ [0, 5/6)
β ∈ [0, 2/3)
β ∈ [0, 1/2)
β ∈ [0, 1/3)
β ∈ [0, 1/6)
∅
−1
(β) also has one solution, provided that
Therefore, the inverse function C1 = f
M
β ∈ [0, (5 − M)/6). This range of β is obtained by mapping the C1 interval (0, 1] via
the functions f M (C1). The functions f M (C1) in Table 3 provide a link between β and
C1: if C1 is known, β = f M (C1) provides the parameter value to input in the model
to obtain such an equilibrium. Conversely, if β is known, Table 3 indicates whether
or not C1 ∈ (0, 1]. Finally, we note that when β ≥ (5 − M)/6, or when M = 5, only
the zero equilibrium, C ∗ = 0, is present.
To determine the stability of the equilibria, we consider the cases when β ∈
[0, (5 − M)/6) and when β ≥ (5 − M)/6 separately. When β ∈ [0, (5 − M)/6), two
distinct equilibria are present: C ∗ = 0 and C ∗ = C1 ∈ (0, 1]. Based on the sign of
∂S(C; f M (C ∗), M)/∂C at these equilibria, C ∗ = 0 is always unstable and C ∗ = C1
is always stable. These features are consistent with the Weak Allee effect, whereby
the density growth rate deviates from logistic growth without incorporating additional
equilibria. When β ≥ (5 − M)/6, or when M = 5, C ∗ = 0 is the only equilibrium and
it is always stable, corresponding to the qualitative features of an extinction density
growth rate, where limt→∞ C(t) = 0 for all C(0). Both qualitative features in this
parameter regime are shown in the bifurcation diagram in Fig. 4. We conclude that in
Case 1, either zero or one equilibria is present in the interval (0, 1], corresponding to
extinction and Weak Allee parameter regimes, respectively.
2.2 Case 2: r > 0 and R = 0
This case corresponds to when individuals above M do not proliferate or die. When
R = 0, we have
1
rC(t)
dC(t)
dt
= T (C(t); α, M)
:= (1 − C(t))
(cid:5)
M(cid:4)
j=0
5
j
(cid:6)
C(t) j (1 − C(t))5− j
(cid:8)
(cid:9)
,
(8)
1 − 6α
6 − j
123
74
Page 10 of 22
N. T. Fadai, M. J. Simpson
0
5-M
6
[
1
]
β
Weak Allee Effect
dC
dt
dC
dt
Extinction
C
C
Stable Equilibrium
Unstable Equilibrium
Fig. 4 Bifurcation diagram of the Binary Switch Model, shown in (6), for Case 1 when r = 0. Varying β
produces different qualitative features in terms of equilibria and their stability. The resulting density growth
rates, dC/dt, are shown as a function of C, where a stable equilibrium is represented with a black circle
and an unstable equilibrium with a white circle
which is independent of β. In a similar fashion to Case 1, we consider the equilibria
for various choices of α and M, noting that C ∗ = 0 and C ∗ = 1 are always equilibria
in this case. However, we will show that in Case 2, we have the possibility of a third
equilibrium in (0, 1). When this additional equilibria is present, then C2 = 1 and
C1 ∈ (0, 1); otherwise, C1 = 1. To determine if C ∗ = 1 is the first or second nonzero
equilibrium, we define
α = gM (C1) ⇐⇒ T (C1, gM (C1), M) = 0,
(9)
−1
M
and determine the value of α that solves T (C1, α, M) = 0 for a given value of
C1 ∈ (0, 1), shown in Table 4. Like Case 1, the family of functions α = gM (C1)
provide an explicit relationship between α and C1. Since α = gM (C1) is one-to-one
on C1 ∈ (0, 1), the inverse function C1 = g
(α) also has one solution, C1 ∈ (0, 1),
provided α ∈ ((6 − M)/6, 1). This value of C1 ∈ (0, 1) provides a third equilibrium
of (8); conversely, when α ≤ (6 − M)/6, or when M = 0, the only two equilibria are
C ∗ = 0 and C1 = 1.
In the case where C1 ∈ (0, 1), examining the sign of ∂S(C; f M (C ∗), M)/∂C shows
that C ∗ = 0 and C ∗ = 1 are unstable, whereas C ∗ = C1 is stable. This combination
of equilibria has the opposite stability properties of the Strong Allee effect (Table 1),
and so we refer to density growth rates with these stability properties as the Reverse
Allee effect. In the case where α ≤ (6 − M)/6, or when M = 0, stability analysis
shows that C1 = 1 is stable and C ∗ = 0 is unstable, which is consistent with the
qualitative features of the Weak Allee effect. Finally, when α = 1, we return to having
only two equilibria, C ∗ = 0 and C ∗ = 1, but the stability is the opposite of the usual
Weak Allee effect. Therefore, when α = 1, limt→∞ C(t) = 0 for C(0) < 1. All these
qualitative features in this parameter regime are shown in the bifurcation diagram in
Fig. 5. We conclude that in Case 2, either one or two equilibria are present in (0, 1],
with the Extinction regime occurring when α = 1. For α < 1, a new kind of Allee
123
Population Dynamics with Threshold Effects Give Rise to a…
Page 11 of 22
74
Table 4 Relation between nonzero equilibrium, 0 < C1 < 1, to α and M for Case 2 when R = 0 (8)
M
0
1
2
3
4
5
α = gM (C1)
Range of α : C1 ∈ (0, 1)
1
4C1+1
5C1+1
6C2
+3C1+1
1
10C2
+4C1+1
1
+3C2
4C3
+2C1+1
1
1
+6C2
10C3
+3C1+1
1
1
+C2
+C3
C4
1
1
1
+3C2
+4C3
5C4
1
1
1
1
+C2
+C3
1
1
+C4
1
C5
1
+C1+1
+2C1+1
+C1+1
∅
α ∈ (5/6, 1)
α ∈ (2/3, 1)
α ∈ (1/2, 1)
α ∈ (1/3, 1)
α ∈ (1/6, 1)
0
6-M
6
Weak Allee Effect
dC
dt
dC
dt
C
Reverse Allee Effect
1
dC
dt
Extinction
C
C
Stable Equilibrium
Unstable Equilibrium
Fig. 5 Bifurcation diagram of the Binary Switch Model, shown in (8), for Case 2 when R = 0. Varying α
produces different qualitative features in terms of equilibria and their stability. The resulting density growth
rates, dC/dt, are shown as a function of C, where a stable equilibrium is represented with a black circle
and an unstable equilibrium with a white circle
effect, which we call the Reverse Allee effect, occurs if two equilibria are present in
(0, 1]; otherwise, we retrieve the Weak Allee effect.
2.3 Case 3: r > 0 and R > 0
In the most general case, the proliferation and death rates of individuals change at the
threshold density M, but remain nonzero on either side of the threshold density. As a
result, (4) can be written as
123
74
Page 12 of 22
N. T. Fadai, M. J. Simpson
1
rC(t)
dC(t)
dt
= − R
r
βC(t)6 +
(cid:5)
M(cid:4)
j=0
5
j
(cid:6)
C(t) j (1 − C(t))6− j
(cid:9)
(cid:8)
1 − 6α
6 − j
(cid:8)
+ 1(M ≤ 4) · R
r
5(cid:4)
j=M+1
(cid:5)
5
j
(cid:6)
C(t) j (1 − C(t))6− j
(cid:9)
.
1 − 6β
6 − j
(10)
Without loss of generality, we assume that r = 1, since other nonzero values or r
can be rescaled to unity by changing the timescale in (4), which does not affect its
equilibria. Consequently, with some rearranging, we have
1
C(t)
dC(t)
dt
= V(C(t); R, α, β, M)
:= 1 − C(t) − α(1 − C(t)6) − RβC(t)6
5(cid:4)
(cid:5)
+ 1(M ≤ 4) ·
j=M+1
(cid:6)
C(t) j (1 − C(t))6− j
5
j
(cid:8)
R − 1 + 6(α − β R)
6 − j
(cid:9)
.
(11)
We will show that in Case 3, there can be between zero and three equilibria in (0, 1],
noting that C ∗ = 1 is an equilibrium of (11) if and only if β = 0. When we have
three distinct equilibria in (0, 1], we obtain a new type of Allee effect, referred to
here as the Hyper-Allee effect (Fadai et al. 2020), in which the zero equilibrium is
unstable, and an intermediate unstable equilibrium is contained between two positive,
stable equilibria. However, in order for the parameter space to continuously transition
from the Weak Allee effect, as in Cases 1 and 2, to the Hyper-Allee effect, there must
exist a critical set of model parameters at which a double-root equilibrium occurs.
Therefore, to determine what regions of (R, α, β, M) parameter space exhibit Hyper-
Allee effects instead of the Weak Allee effect, we focus on determining the boundary
of these effects in terms of model parameters and equilibria. This boundary, defined
as the Tangential Manifold, will be the focus of our analysis in this section.
In addition to determining the boundary between Weak Allee and Hyper-Allee
parameter spaces, we will also show that even more Allee effects are present when
α = 1. In particular, we show that in Case 3, the Extinction parameter regime continues
to exist, along with the Strong Allee effect, when α = 1. We also determine an explicit
relationship between R, β, and M for when the Extinction regime becomes the Strong
Allee effect, which is linked to the Tangential Manifold. We now focus our attention
on determining additional equilibria Ci ∈ (0, 1].
Numerical observations indicate that certain combinations of (R, α, β, M) can pro-
duce up to three distinct values of Ci ∈ (0, 1] satisfying V = 0. Furthermore, in
parameter regimes where three distinct equilibria are present in (0, 1], stability anal-
ysis about these equilibria reveals that C ∗ = 0 and C ∗ = C2 are unstable equilibria,
whereas C ∗ = C1 and C ∗ = C3 are stable equilibria. These qualitative features are
consistent with the aforementioned Hyper-Allee effect, which is a higher-order effect
that is very different to the usual Weak Allee and Strong Allee effects (Fig. 6).
123
Population Dynamics with Threshold Effects Give Rise to a…
Page 13 of 22
74
dC
dt
Extinction
dC
dt
Junction Point
C
C
dC
dt
dC
dt
Negative Tangential Manifold
C
C
Triple Point
1
0.8
0.6
0.4
0.2
0
0
dC
dt
2
4
R
6
8
10
Weak Allee Effect
C
Strong Allee Effect
Hyper-Allee Effect
Positive Tangential Manifold
dC
dt
dC
dt
dC
dt
C
C
C
Stable Equilibrium
Unstable Equilibrium
Semi-stable Equilibrium
Fig. 6 Bifurcation diagram of the Binary Switch Model for Case 3, shown in (11), with β = 0.06, r =
1, R > 0, and M = 4. Pairs of (α, R) parameters produce different qualitative features, in terms of equilibria
and their stability. The resulting density growth rates, dC/dt, are shown as a function of C, where a stable
equilibrium is represented with a black circle, an unstable equilibrium with a white circle, and a semi-stable
equilibrium with a half-filled circle
For solutions to continuously transition from one equilibrium in (0, 1], like the
Weak Allee effect in Cases 1 and 2, to three equilibria in (0, 1], such as the Hyper-
Allee effect, we must have certain values of (R, α, β, M) that produce a double root
for Ci . We denote this special case of a double root equilibrium as ˆC, which can occur
in either the C1 or C2 equilibrium position. In addition to satisfying V = 0, the double
root equilibrium, C ∗ = ˆC, must also satisfy
V( ˆC; R, α, β, M) =
∂
∂C
V(C; R, α, β, M)
(cid:10)
(cid:10)
(cid:10)
(cid:10)
= 0.
C= ˆC
(12)
The set of parameters satisfying (12) is referred to as the Tangential Manifold, where
the double root equilibrium, ˆC, is a semi-stable equilibrium of (11) (Strogatz 2018).
A semi-stable equilibrium C ∗ = ˆC has the properties that populations slightly larger
than C(t) ≡ ˆC remain close to ˆC, but populations slightly smaller than C(t) ≡ ˆC
diverge away from ˆC, or vice-versa. Since we have two equations with four unknowns,
we parametrise the Tangential Manifold as (R, α) = (FM ( ˆC, β), G M ( ˆC, β)), for par-
123
74
Page 14 of 22
N. T. Fadai, M. J. Simpson
Table 5 Relation between the semi-stable equilibrium, ˆC, to α, β, R, and M for Case 3. Parameter values
satisfying R = FM ( ˆC, β) and α = G M ( ˆC, β) are members of the Tangential Manifold. If ˆC < C, then
ˆC is a member of the Positive Tangential Manifold; if C < ˆC < ˜C, then ˆC is a member of the Negative
Tangential Manifold. The Triple Point, C, is defined implicitly via β = HM (C), while the Junction Point,
˜C, is defined implicitly via β = JM ( ˜C)
M
0
1
2
3
4
5
M
0
1
2
3
4
5
R = FM ( ˆC, β)
0
( ˆC−1)6
ˆC( ˆC5−6 ˆC4+15 ˆC3−20 ˆC2−10 ˆC−30β+20)
( ˆC−1)5(6 ˆC2+8 ˆC+1)
ˆC2(6 ˆC5−22 ˆC4+21 ˆC3+15 ˆC2+10 ˆC+60β−30)
( ˆC−1)4(6 ˆC4+16 ˆC3+21 ˆC2+6 ˆC+1)
ˆC3(6 ˆC5−8 ˆC4−7 ˆC3−6 ˆC2−5 ˆC−60β+20)
( ˆC−1)3( ˆC6+4 ˆC5+10 ˆC4+20 ˆC3+10 ˆC2+4 ˆC+1)
ˆC4( ˆC5+ ˆC4+ ˆC3+ ˆC2+ ˆC+30β−5)
0
α = G M ( ˆC, β)
1
β( ˆC5−6 ˆC4+15 ˆC3−20 ˆC2+15 ˆC−30)−20( ˆC−1)
ˆC5−6 ˆC4+15 ˆC3−20 ˆC2−10 ˆC−30β+20
β(6 ˆC5−22 ˆC4+21 ˆC3+15 ˆC2−40 ˆC+60)+30( ˆC−1)
6 ˆC5−22 ˆC4+21 ˆC3+15 ˆC2+10 ˆC+60β−30
β(6 ˆC5−8 ˆC4−7 ˆC3−6 ˆC2+45 ˆC−60)−20( ˆC−1)
6 ˆC5−8 ˆC4−7 ˆC3−6 ˆC2−5 ˆC−60β+20
β( ˆC5+ ˆC4+ ˆC3+ ˆC2−24 ˆC+30)+5( ˆC−1)
ˆC5+ ˆC4+ ˆC3+ ˆC2+ ˆC+30β−5
1/6
β = HM (C)
∅
2(1−C)
3
(1−C)(1+2C)
3C+2
(1−C)(1+2C+2C2)
3C2+4C+3
(1−C2)(2C2+C+2)
3(C3+2C2+3C+4)
∅
β = JM ( ˜C)
∅
( ˜C−1)( ˜C3−5 ˜C2+10 ˜C−10)
˜C4−6 ˜C3+15 ˜C2−20 ˜C+15
( ˜C−1)(6 ˜C3−16 ˜C2+5 ˜C+20)
6 ˜C4−22 ˜C3+21 ˜C2+15 ˜C−40
( ˜C−1)(6 ˜C3−2 ˜C2−9 ˜C−15)
6 ˜C4−8 ˜C3−7 ˜C2−6 ˜C+45
( ˜C−1)( ˜C3+2 ˜C2+3 ˜C+4)
˜C4+ ˜C3+ ˜C2+ ˜C−24
∅
ticular values of ˆC and β (Fig. 6). The functions FM ( ˆC, β) and G M ( ˆC, β) describing
the Tangential Manifold are shown in Table 5.
While the Tangential Manifold can be determined explicitly by solving (12), we
observe that two forms of a semi-stable equilibrium can occur (Fig. 6). If the double
root ˆC is below some critical value, C, then this semi-stable equilibrium occurs between
C ∗ = 0, which is unstable, and some larger equilibrium C ∗ = C2, which is stable. If
ˆC > C, then this semi-stable equilibrium is larger than both C ∗ = 0 and C ∗ = C1,
which remain unstable and stable, respectively. We refer to the branch of the Tangential
Manifold where ˆC < C as the Positive Tangential Manifold, based on the sign of the
density growth rate between ˆC and C2 (Fig. 6). In a similar fashion, we refer to the
branch of the Tangential Manifold where ˆC > C as the Negative Tangential Manifold.
When ˆC = C, the double root becomes a stable triple root and C satisfies
∂ 2
∂C 2
V (C; FM (C, β), G M (C, β), β, M)
(cid:10)
(cid:10)
(cid:10)
(cid:10)
C=C
= 0,
(13)
where R = FM (C, β) and α = G M (C, β) are chosen to ensure we remain on the
Tangential Manifold. Equation (13) provides an additional constraint on the Tangential
123
Population Dynamics with Threshold Effects Give Rise to a…
Page 15 of 22
74
Table 6 Summary of qualitative features seen in the Binary Switch Model. The stability of each equilibrium,
listed in increasing order of magnitude, can be stable (S), unstable (U ), or semi-stable (SS)
Effect name
Extinction
Logistic growth
Weak Allee/Triple Point
Junction Point
Strong Allee
Reverse Allee
Positive Tangential Manifold
Negative Tangential Manifold
Hyper-Allee
Equilibria
Stability
Notes
{0}
{0, C1}
{0, C1}
{0, C1}
{0, C1, C2}
{0, C1, C2}
{0, C1, C2}
{0, C1, C2}
{0, C1, C2, C3}
{S}
{U , S}
{U , S}
{S, SS}
{S, U , S}
{U , S, U }
{U , SS, S}
{U , S, SS}
{U , S, U , S}
r = R, α = β
Triple: C1 = C
C1 = ˜C
C2 = 1
C1 = ˆC
C2 = ˆC
Manifold, implying that we can relate C to a unique value of β. We denote β = HM (C)
if (13) is satisfied, with C denoting the Triple Point of (11) (Table 5).
Additionally, from Fig. 6, we note that when α = 1, the equilibria and their resulting
stability change, compared to α < 1. When α = 1, the Negative Tangential Manifold is
valid for a unique pair of (β, R) parameters, for a particular equilibrium value, C ∗ = ˜C.
We define this critical equilibrium value as the Junction Point, which satisfies
G M ( ˜C, β) = 1.
(14)
We denote β = JM ( ˜C) if (14) is satisfied (Table 5); furthermore, we determine the
corresponding value of R at the Junction Point by evaluating R = FM ( ˜C, JM ( ˜C)).
When α = 1 and R < R, the only equilibrium value of (11) is C ∗ = 0, which is stable.
This implies that all population densities go extinct in this parameter regime. When
α = 1 and R > R, (11) has three solutions: C ∗ = 0, which is stable, an intermediate-
valued unstable equilibrium C ∗ = C1, and a larger-valued stable equilibrium C ∗ = C2
(Fig. 6). Thus, the stability features of this density growth rate are the same as the Strong
Allee effect. When R = R, the Junction Point, C ∗ = ˜C, is semi-stable, while C ∗ = 0
remains stable. A summary of this diverse family of Allee effects, in terms of the
number and stability of the equilibria, is shown in Table 6.
From Table 5, we note some key features of the Tangential Manifold. Firstly, when
β = 0, we note that the Triple Point is C = 1 for 1 ≤ M ≤ 4. Since the Negative
Tangential Manifold must have ˆC > C, we conclude that the Negative Tangential
Manifold does not exist when β = 0, which is also observed in Fig. 7. When β =
(5 − M)/6 and 1 ≤ M ≤ 4, the Triple Point and the Junction Point are both C = ˜C =
0, implying that no points are contained in the Tangential Manifold. Consequently,
parameter pairs (α, R) that result in qualitative features other than the Extinction
regime or the Weak Allee effect can only occur when α < 1 and β ∈ [0, (5 − M)/6),
as shown in Fig. 7. Finally, we note that when M = 0 or M = 5, the Tangential
Manifold does not exist, since the solution of (12) requires R = 0. Therefore, the
123
74
Page 16 of 22
N. T. Fadai, M. J. Simpson
Fig. 7 Bifurcation diagram of the Binary Switch Model for Case 3, shown in (11), with M = 4, r = 1, R >
0, and varying β. The qualitative forms of various effects are shown in the legend, described in further
detail in Fig. 6. The parameter space exhibiting Hyper-Allee features vanishes as β → 1/6
123
Population Dynamics with Threshold Effects Give Rise to a…
Page 17 of 22
74
1
y
t
i
s
n
e
D
d
e
a
c
s
e
R
l
0.8
0.6
0.4
0.2
0
0
a
100
10-1
y
t
i
s
n
e
D
d
e
a
c
s
e
R
l
50
100
Time (h)
150
200
10-2
b
Time (h)
Experimental Data
Binary Switch (M=0)
Binary Switch (M=1)
Binary Switch (M=2)
Binary Switch (M=3)
Binary Switch (M=4)
Binary Switch (M=5)
Fig. 8 Population density of U87 glioblastoma cells compared to the calibrated Binary Switch Model. U87
glioblastoma cells, with initial densities of c1(0) = 0.02, c2(0) = 0.06, and c3(0) = 0.2, are observed
over the span of 120 h (black circles) (Neufeld et al. 2017). The Binary Switch Model (solid curves) is fit
to minimise the combined least-square error (15), Σχ 2, of three experimental data sets shown in Neufeld
et al. (2017). The estimates of the optimal model parameter set, for each value of M, are shown in Table 7.
b A semi-log plot makes it easier to visually compare the quality of match between the data and the model
qualitative features of (11) in the entire (α, R) parameter space are those seen in the
Weak Allee effect when α < 1 and the Extinction regime when α = 1.
To summarise, we determine that in Case 3 when M ∈ {1, 2, 3, 4}, and β ∈ [0, (5−
M)/6), a diverse family of Allee effects can be found. Among these Allee effects are:
the Weak Allee effect, the Extinction regime, the Strong Allee effect, and a Hyper-Allee
effect parameter regime with three distinct equilibria in (0, 1]. Additional Allee effects
can be observed at the boundaries of the aforementioned Allee effects, including the
Tangential Manifold and Junction Point with semi-stable equilibria, as well as the
Triple Point with a single stable equilibria in (0, 1]. In all of these cases, there are
between zero and three equilibria in the interval (0, 1].
3 Interpreting Experimental Data Using the Binary Switch Model
To demonstrate how the Binary Switch Model can be used to provide biological insight,
we consider population-level data sets describing the growth of populations of cancer
cells. Neufeld et al. (2017) perform three experiments with U87 glioblastoma cells.
Uniform monolayers of cells are grown from three different initial densities, with
the data shown in Fig. 8. Here, we see that all three experiments lead to increasing
population densities with time. The two experiments with the smallest initial densities
lead to increasing, concave up C(t) profiles. The experiment with the largest initial
density leads to an increasing C(t) profile that changes concavity at approximately
t = 100 h.
123
74
Page 18 of 22
N. T. Fadai, M. J. Simpson
Table 7 Estimates of the Binary Switch Model parameters that minimise the combined least-squares error
(15) between model predictions and experimental data from Neufeld et al. (2017). The optimal parameter
set with M = 1, highlighted in bold, provides the smallest combined least-squares error for all values of M
M
r
R
0
1
2
3
4
5
0.0113
0.0168
0.0262
0.0345
0.0180
0.0576
0.0206
0.0642
0.0218
0.134
0.0237
0.0110
α
0.174
0.0608
2.84 × 10
3.66 × 10
3.43 × 10
3.73 × 10
−5
−9
−9
−10
β
C1(0)
C2(0)
C3(0)
Σχ 2
−6
2.82 × 10
0.0692
0.139
0.0892
0.0623
2.34 × 10
−4
0.0250
0.0192
0.0160
0.0126
0.0112
0.0661
0.0652
0.0619
0.0534
0.0489
0.00933
0.0420
0.184
0.188
0.191
0.193
0.191
0.183
0.0179
0.0154
0.0169
0.0268
0.0366
0.0571
The density of U87 glioblastoma cells has already been rescaled by its maximum
packing density in Neufeld et al. (2017), so we assume that C = 1 corresponds
to the maximum rescaled density. Our aim is to choose Θ = (α, β, r , R, M), with
C1(0), C2(0), and C3(0) as initial conditions, such that the model parameters provide
the best match to all three experimental conditions simultaneously. It is important
to calibrate the model to match all three data sets simultaneously, because if (4)
is consistent with the experimental data, there should be a single choice of model
parameters that matches the observed population dynamics, regardless of initial density
(Jin et al. 2016b).
To match all experimental data sets simultaneously, we consider the combined
least-squares error between model predictions and all data:
Σχ 2(Θ) :=
(cid:11)
C(t j ; Θ) − c j
(cid:12)
2 .
(cid:4)
j
(15)
Here, we treat the initial densities, C1(0), C2(0), C3(0) as parameters to be deter-
mined; therefore, we consider the extended parameter vector, Θ = (M, r , R,
α, β, C1(0), C2(0), C3(0)). In (15), c j represents all three experimental data sets
obtained at times t j , concatenated into a single vector. While the Binary Switch
Model uses the initial conditions C1(0), C2(0), and C3(0), we denote the experi-
mental measurements at t = 0 h as c1(0), c2(0), and c3(0), respectively (Fig. 8).
Using fminsearch in MATLAB (MathWorks 2020), we estimate Θ ∗ such that
Σχ 2 is minimised. Since M is discrete, while (r , R, α, β, C1(0), C2(0), C3(0)) are
continuous, we estimate Θ ∗ for each value of M ∈ {0, 1, 2, 3, 4, 5} and then choose
the value of M that minimises Σχ 2. A MATLAB implementation of this least-squares
procedure is discussed in the Supplementary Information.
In Fig. 8, we show the best match that the Binary Switch Model can provide to all
three data sets from Neufeld et al. (2017) for each value of M. The optimal parameter
set Θ ∗ and minimal Σχ 2 for each value of M are reported in Table 7. We conclude
that setting a threshold of M = 1 provides the best match to these data sets. While
larger values of M clearly deviate from the experimental data sets at low population
densities (Fig. 8b), setting M = 0 or M = 2 also leads to a reasonable visual match
123
Population Dynamics with Threshold Effects Give Rise to a…
Page 19 of 22
74
for all three experimental data sets (Fig. 8). Furthermore, it is of interest to note that
the optimal model parameters associated with small values of M correspond to non-
logistic growth features, since logistic growth can only be obtained when r = R
and α = β (Table 7). The match between the experimental data and the model at
M = 1 has several consequences: (i) this exercise confirms that the data reported by
Neufeld et al. (2017) does not follow standard logistic growth; (ii) the high-quality
match between the Binary Switch Model and the data for M = 1 is consistent with
population dynamics similar to a Weak Allee effect, and (iii) interpreting this data
using the Binary Switch Model indicates that the best way to explain the population
dynamics with a relatively small threshold population density.
3.1 Applications to Ecology
Threshold effects are thought to be a common feature in biological population dynam-
ics, both in cell biology and in ecology. In the previous section, we demonstrated that
a population of U87 glioblastoma cells did not follow logistic growth and was bet-
ter described using the Weak Allee effect. In a similar fashion, various populations
in ecology with known threshold effects are better described using Allee effects. A
common threshold effect arising in ecology is a threshold population density (Cour-
champ et al. 2008), whereby a particular species will go extinct below this critical
density. Species that have been noted to go extinct below a threshold density include
the quokka (Sinclair and Pech 1996), the woodland caribou (Wittmer et al. 2005), the
red-backed vole (Morris 2002), and the gypsy moth (Tcheslavskaia et al. 2002; Lieb-
hold and Bascompte 2003). In many of these populations, the threshold density has
been measured (Courchamp et al. 2008), thereby providing an appropriate estimate of
the equilibrium density C ∗ employed in the Binary Switch Model. Consequently, the
Binary Switch Model aligns with threshold effects known to arise in ecology, while
also providing insight into the underlying individual-level mechanisms that give rise
to Allee effects. Furthermore, by making use of measured threshold population den-
sities, we are thereby able to obtain an estimate of the threshold parameter M, further
reducing the parameter search space needed to calibrate the Binary Switch Model to
match experimental data.
4 Conclusions
In this work, we examine the link between threshold effects in population growth
mechanisms and Allee effects. An abrupt change in growth mechanisms, which we
refer to as a binary switch, is thought to be a common feature of biological popula-
tion dynamics. Despite the ubiquitous nature of local binary switches in population
dynamics, an explicit connection to Allee effects has not been considered. To explore
this connection in greater detail, we examine a population density growth model, in
which the proliferation and death rates vary with the local density of the population.
By incorporating a local binary switch in these proliferation and death rates, we greatly
123
74
Page 20 of 22
N. T. Fadai, M. J. Simpson
reduce the size of the parameter space while explicitly incorporating a biologically
realistic threshold effect in the proliferation and death rates.
To provide insight into the qualitative features of population dynamics arising in the
Binary Switch Model, we examine the presence and stability of the resulting equilibria.
We show that when the binary switch occurs at some intermediate population density
and the high-density death rate is not too large, a diverse family of Allee effects is
supported by the model. Among these Allee effects are: (i) logistic growth, when no
binary switch is present; (ii) the Weak Allee effect, which modifies the simpler logistic
growth model without changing its equilibria or their stability; (iii) an Extinction
regime, where all population densities will eventually go extinct; (iv) the Strong Allee
effect, where population below a critical density will go extinct rather than grow, and
(v) the Hyper-Allee effect, which has two distinct positive stable population densities.
Furthermore, we show that there are additional forms of Allee effects at the boundaries
in the parameter space that separate these five main classes of Allee effects.
Along with exhibiting a wide range of Allee effects, the Binary Switch Model has a
restricted parameter regime, making the interpretation of the local binary switch clearer
while requiring fewer parameters to identify when calibrating to experimental data. To
demonstrate these advantages, we calibrate the Binary Switch Model to experimental
data sets arising in cell biology. Not only can the Binary Switch Model provide a
good match to all experimental data across three different initial densities, we also
find that the parameters used to match the data provide a more explicit interpretation
of the underlying local growth mechanisms arising in the population. Specifically, we
confirm that the experimental data are inconsistent with the standard logistic model
and that the phenomena is best explained by a binary switch at low density. We
conclude that the Binary Switch Model is useful to theorists and experimentalists
alike in providing insight into binary switches at the individual scale that produce
Allee effects at the population scale.
While one of the merits of the Binary Switch Model is to show how a single local
binary switch gives rise to a variety of Allee effects, further extensions of the mod-
elling framework can be made. For instance, additional switches can be incorporated
into the modelling framework, representing populations whose proliferation and death
rates change at more than one density. We anticipate that this kind of extension would
lead to additional forms of Allee effects in the resulting population dynamics. Another
potential modification would be to generalise the notion how we measure local density.
In this work, we take the simplest possible approach use the number of nearest neigh-
bours on a hexagonal lattice to represent the local density. Several generalisations,
such as working with next nearest neighbours or working with a weighted average of
nearest neighbours, could be incorporated into our modelling framework (Fadai et al.
2020; Jin et al. 2016a). Again, we expect that such extensions would lead to an even
richer family of population dynamics models. We leave these extensions for future
considerations.
Acknowledgements This work is supported by the Australian Research Council (DP170100474). The
authors thank the anonymous referee for their helpful comments.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
123
Population Dynamics with Threshold Effects Give Rise to a…
Page 21 of 22
74
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence,
and indicate if changes were made. The images or other third party material in this article are included
in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If
material is not included in the article’s Creative Commons licence and your intended use is not permitted
by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
References
Allee WC, Bowen ES (1932) Studies in animal aggregations: mass protection against colloidal silver among
goldfishes. J Exp Zool 61(2):185–207
Bobadilla AVP, Carraro T, Byrne HM, Maini PK, Alarcón T (2019) Age structure can account for delayed
logistic proliferation of scratch assays. Bull Math Biol 81(7):2706–2724
Böttger K, Hatzikirou H, Voss-Böhme A, Cavalcanti-Adam EA, Herrero MA, Deutsch A (2015) An emerg-
ing Allee effect is critical for tumor initiation and persistence. PLoS Comput Biol 11(9):e1004366
Colon C, Claessen D, Ghil M (2015) Bifurcation analysis of an agent-based model for predator–prey
interactions. Ecol Model 317:93–106
Courchamp F, Clutton-Brock T, Grenfell B (1999) Inverse density dependence and the Allee effect. Trends
Ecol Evol 14(10):405–410
Courchamp F, Berec L, Gascoigne J (2008) Allee effects in ecology and conservation. Oxford University
Press, Oxford
de Pillis LG, Radunskaya A (2003) A mathematical model of immune response to tumor invasion. In: Bathe
KJ (ed) Computational fluid and solid mechanics 2003. Elsevier, Amsterdam, pp 1661–1668
de Pillis LG, Radunskaya AE, Wiseman CL (2005) A validated mathematical model of cell-mediated
immune response to tumor growth. Cancer Res 65(17):7950–7958
Drake JM (2004) Allee effects and the risk of biological invasion. Risk Anal Int J 24(4):795–802
Edelstein-Keshet L (2005) Mathematical models in biology. SIAM, Philadelphia
Fadai NT, Johnston ST, Simpson MJ (2020) Unpacking the Allee effect: determining individual-level
mechanisms that drive global population dynamics. bioRxiv https://www.biorxiv.org/content/10.1101/
774000v3
Frankham R (1995) Inbreeding and extinction: a threshold effect. Conserv Biol 9(4):792–799
Friedl P, Alexander S (2011) Cancer invasion and the microenvironment: plasticity and reciprocity. Cell
147(5):992–1009
Gerlee P (2013) The model muddle: in search of tumor growth laws. Cancer Res 73(8):2407–2411
Hatzikirou H, Basanta D, Simon M, Schaller K, Deutsch A (2012) ‘Go or grow’: the key to the emergence
of invasion in tumour progression? Math Med Biol J IMA 29(1):49–65
Hopf F, Hopf F (1985) The role of the Allee effect in species packing. Theor Popul Biol 27(1):27–50
Jenner AL, Yun C-O, Kim PS, Coster AC (2018) Mathematical modelling of the interaction between
cancer cells and an oncolytic virus: insights into the effects of treatment protocols. Bull Math Biol
80(6):1615–1629
Jenner AL, Kim PS, Frascoli F (2019) Oncolytic virotherapy for tumours following a Gompertz growth
law. J Theor Biol 480:129–140
Jin W, Penington CJ, McCue SW, Simpson MJ (2016a) Stochastic simulation tools and continuum models
for describing two-dimensional collective cell spreading with universal growth functions. Phys Biol
13(5):056003
Jin W, Shah ET, Penington CJ, McCue SW, Chopin LK, Simpson MJ (2016b) Reproducibility of scratch
assays is affected by the initial degree of confluence: experiments, modelling and model selection. J
Theor Biol 390:136–145
Jin W, Shah ET, Penington CJ, McCue SW, Maini PK, Simpson MJ (2017) Logistic proliferation of cells
in scratch assays is delayed. Bull Math Biol 79(5):1028–1050
Johnson DM, Liebhold AM, Tobin PC, Bjørnstad ON (2006) Allee effects and pulsed invasion by the gypsy
moth. Nature 444(7117):361–363
Johnson KE, Howard G, Mo W, Strasser MK, Lima EABF, Huang S, Brock A (2019) Cancer cell population
growth kinetics at low densities deviate from the exponential growth model and suggest an Allee effect.
PLoS Biol 17:e3000399
123
74
Page 22 of 22
N. T. Fadai, M. J. Simpson
Johnston ST, Baker RE, McElwain DLS, Simpson MJ (2017) Co-operation, competition and crowding:
a discrete framework linking Allee kinetics, nonlinear diffusion, shocks and sharp-fronted travelling
waves. Sci Rep 7:42134
Koenig WD, Knops JM (2005) The mystery of masting in trees: some trees reproduce synchronously over
large areas, with widespread ecological effects, but how and why? Am Sci 93(4):340–347
Kuparinen A, Keith DM, Hutchings JA (2014) Allee effect and the uncertainty of population recovery.
Conserv Biol 28(3):790–798
Liebhold A, Bascompte J (2003) The allee effect, stochastic dynamics and the eradication of alien species.
Ecol Lett 6(2):133–140
Martínez-García R, Murgui C, Hernández-García E, López C (2015) Pattern formation in populations with
density-dependent movement and two interaction scales. PLoS ONE 10(7):e0132261
MathWorks (2020) fminsearch: MathWorks documentation. https://au.mathworks.com/help/matlab/ref/
fminsearch.html. Accessed May 2020
Metzger J-P, Décamps H (1997) The structural connectivity threshold: an hypothesis in conservation biology
at the landscape scale. Acta Oecol 18(1):1–12
Morris DW (2002) Measuring the allee effect: positive density dependence in small mammals. Ecology
83(1):14–20
Murray JD (2003) Mathematical biology I: an introduction. Springer, New York
Neufeld Z, von Witt W, Lakatos D, Wang J, Hegedus B, Czirok A (2017) The role of Allee effect in
modelling post resection recurrence of glioblastoma. PLoS Comput Biol 13:e1005818
Rossignol R, Malgat M, Mazat J-P, Letellier T (1999) Threshold effect and tissue specificity implication
for mitochondrial cytopathies. J Biol Chem 274(47):33426–33432
Sarapata EA, de Pillis LG (2014) A comparison and catalog of intrinsic tumor growth models. Bull Math
Biol 76(8):2010–2024
Scott SM, Bodine EN, Yust A (2014) An agent-based model of Santa Cruz island foxes (urocyon littoralis
santacruzae) which exhibits an Allee effect. Lett Biomath 1(1):97–109
Seebens H, Blackburn TM, Dyer EE, Genovesi P, Hulme PE, Jeschke JM, Pagad S, Pyšek P, Winter M,
Arianoutsou M et al (2017) No saturation in the accumulation of alien species worldwide. Nat Commun
8:14435
Simberloff D, Martin J-L, Genovesi P, Maris V, Wardle DA, Aronson J, Courchamp F, Galil B, García-
Berthou E, Pascal M et al (2013) Impacts of biological invasions: what’s what and the way forward.
Trends Ecol Evol 28(1):58–66
Sinclair ARE, Pech RP (1996) Density dependence, stochasticity, compensation and predator regulation.
Oikos 75:164–173
Stefan A, Geritz H, Kisdi É (2012) Mathematical ecology: why mechanistic models? J Math Biol 65(6–
7):1411
Stephens PA, Sutherland WJ, Freckleton RP (1999) What is the Allee effect? Oikos 87:185–190
Strogatz SH (2018) Nonlinear dynamics and chaos: with applications to physics, biology, chemistry, and
engineering. CRC Press, Boca Raton
Taylor CM, Hastings A (2005) Allee effects in biological invasions. Ecol Lett 8(8):895–908
Tcheslavskaia K, Brewster CC, Sharov AA (2002) Mating success of gypsy moth (Lepidoptera: Lymantri-
idae) females in Southern Wisconsin. Great Lakes Entomol 35(1):1
Tu C, Suweis S, D’Odorico P (2019) Impact of globalization on the resilience and sustainability of natural
resources. Nat Sustain 2:283–289
Vittadello ST, McCue SW, Gunasingh G, Haass NK, Simpson MJ (2020) Examining go-or-grow using
fluorescent cell-cycle indicators and cell-cycle-inhibiting drugs. Biophys J 118(6):1243–1247
Warne DJ, Baker RE, Simpson MJ (2019) Using experimental data and information criteria to guide model
selection for reaction-diffusion problems in mathematical biology. Bull Math Biol 81(6):1760–1804
Wittmer HU, Sinclair ARE, McLellan BN (2005) The role of predation in the decline and extirpation of
woodland caribou. Oecologia 144(2):257–267
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.
123
| null |
10.1038_s41598-022-05932-2.pdf
|
Data availability
All ISI system code is deposited at https:// github. com/ haide rlab/ ISI, and source data and analysis code to rep-
licate the main results will be publicly available at DOI (https:// doi. org/ 10. 6084/ m9. figsh are. 16200 711) upon
|
Data availability All ISI system code is deposited at https:// github. com/ haide rlab/ ISI , and source data and analysis code to replicate the main results will be publicly available at DOI ( https:// doi. org/ 10. 6084/ m9. figsh are. 16200 711 ) upon publication.
|
OPEN
Optimizing intact skull intrinsic
signal imaging for subsequent
targeted electrophysiology
across mouse visual cortex
Armel Nsiangani1,3, Joseph Del Rosario1, Alan C. Yeh1, Donghoon Shin2, Shea Wells2,
Tidhar Lev‑Ari1, Brice Williams1 & Bilal Haider1*
Understanding brain function requires repeatable measurements of neural activity across multiple
scales and multiple brain areas. In mice, large scale cortical neural activity evokes hemodynamic
changes readily observable with intrinsic signal imaging (ISI). Pairing ISI with visual stimulation
allows identification of primary visual cortex (V1) and higher visual areas (HVAs), typically through
cranial windows that thin or remove the skull. These procedures can diminish long‑term mechanical
and physiological stability required for delicate electrophysiological measurements made weeks
to months after imaging (e.g., in subjects undergoing behavioral training). Here, we optimized
and directly validated an intact skull ISI system in mice. We first assessed how imaging quality and
duration affect reliability of retinotopic maps in V1 and HVAs. We then verified ISI map retinotopy in
V1 and HVAs with targeted, multi‑site electrophysiology several weeks after imaging. Reliable ISI
maps of V1 and multiple HVAs emerged with ~ 60 trials of imaging (65 ± 6 min), and these showed
strong correlation to local field potential (LFP) retinotopy in superficial cortical layers (r2 = 0.74–0.82).
This system is thus well‑suited for targeted, multi‑area electrophysiology weeks to months after
imaging. We provide detailed instructions and code for other researchers to implement this system.
The mouse has become an important tool for investigation of the mammalian visual cortex. Anatomical studies
in mice reveal strong interconnections of primary visual cortex (V1) with multiple higher visual areas (HVAs)1–3;
this hierarchical cortical organization parallels that of the primate visual system4. Moreover, V1 and HVAs in
mice show retinotopic organization, such that visual space maps topographically to cortical space. Measuring
cortical responses to retinotopic visual stimulation with intrinsic signal imaging (ISI) of hemodynamics allows
functional localization of V1 and HVAs in the intact mouse brain5,6. However, cortical hemodynamic responses
show small amplitude changes (< 1% relative to ongoing fluctuations7), so investigators typically thin or remove
the skull (and often the dura) of adult mice for maximum signal quality in V1 and HVAs2,5,7; further, ISI is typi-
cally performed during anesthesia, where controlled conditions permit large numbers of stimulus repetitions
and averaging that overcomes small signal amplitudes and background fluctuations5,8,9.
These well-established ISI methods pose some limitations for subsequent electrophysiological recordings.
First, skull removal or thinning can lead to inflammation, bone remodelling, scar formation, and neural plas-
ticity in as little as a week5,10–12, constraining the timeframe of subsequent electrophysiological measurements.
Optimizing a transcranial ISI system would preserve skull integrity and ensure optimum mechanical stability
and physiological conditions during delicate electrophysiological recordings, particularly multi-site silicon probe
or patch clamp recordings13,14. Second, most ISI protocols utilize anesthesia and sedation, where hundreds of
visual stimulus repetitions and hours of imaging are needed for high-resolution retinotopic maps5,8,9,15. Prolonged
anesthesia, even at low concentrations, can induce lasting effects on visual task performance in rodents16,17,
potentially impairing subjects undergoing behavioral training that lasts weeks to months18,19. Quantifying the
minimum amount of data necessary for intact skull ISI would help mitigate any unneeded consequences of pro-
longed or repeated anesthesia. To our knowledge, no study has optimized ISI for intact skull conditions in adult
mice, quantified the minimum sample size (imaging duration) necessary for reliable and repeatable maps of V1
and HVAs, then verified these retinotopic maps directly with electrophysiological measurements.
1Biomedical Engineering, Georgia Institute of Technology & Emory University, Atlanta, USA. 2Electrical and
Computer Engineering, Georgia Institute of Technology, Atlanta, USA. 3Biology & Computer Science, Georgia
State University, Atlanta, USA. *email: [email protected]
Scientific Reports | (2022) 12:2063
| https://doi.org/10.1038/s41598-022-05932-2
1
Vol.:(0123456789)www.nature.com/scientificreports Here, we optimized and validated performance of an intact skull ISI system for mice. We instilled several
quality control checkpoints to ensure robust detection of transcranial ISI signals. We quantified the duration of
imaging necessary for reliable retinotopic maps of V1 and HVAs, and then verified these ISI maps with targeted
awake and anesthetized electrophysiological recordings, often several weeks after the initial imaging. Our findings
reveal a high degree of correspondence between ISI map retinotopy and direct electrophysiological measurements
in superficial cortical layers. Our system specifications, protocol, and codebase are all made publicly available,
providing a useful tool for investigators wishing to pair minimally invasive transcranial ISI with subsequent
targeted electrophysiology in the mouse visual system.
Results
Strategy for optimal acquisition of intact skull intrinsic signal imaging. Our ISI protocol is
inspired by prior work2,5, but provides two major improvements tailored for ISI integration with subsequent
multi-site electrophysiology. First, we have optimized and validated imaging signal quality using a modified
intact skull cranial window preparation (Fig. 1). Unlike skull thinning or removal, which can cause inflamma-
tion, scarring, or bone regrowth in as little as 7 days5,20, our method allows for less invasive pre-imaging prepa-
ration, preserving the skull integrity for high-quality electrophysiological recordings with multiple sequential
craniotomies. The optimal combinations of headplates, glass coverslips, and adhesives were determined after a
series of pilot experiments. We tested different coverslip thicknesses (0.09–0.12 mm, 0.13–0.17 mm) and diam-
eters (e.g., 3 or 5 mm). We selected coverslips that provided the best balance of spatial coverage, qualitative
optical clarity (sharpness of vasculature immediately after glue polymerization), and minimal amount of glue
(and waiting time) needed for stable polymerization. The 5 mm diameter coverslip with 0.09–012 mm thickness
and Vetbond provided the optimal combination of clarity, cortical spatial coverage, and ease of implantation.
Thicker coverslips required too much glue and/or time for polymerization, and Vetbond provided the greatest
clarity of vasculature as compared to an alternate adhesive (CA Glue). Further, Vetbond plus a coverslip pro-
vided long-lasting clarity and smoothly graded retinotopic maps as compared to glue alone with no coverslip
(Fig. S2C). This is likely due to the coverslip providing a barrier from mechanical degradation and air exposure
that increases opacity over time. Once these window parameters were optimized, we tested various headplates
with different sized chambers to assess mechanical stability and ease of access during electrophysiology experi-
ments. We found that the 5 mm coverslip bonded inside a headplate with an 11 mm circular opening (Fig. S2)
provided a highly repeatable and mechanically stable platform for ISI, subsequent cranial window removal, and
multiple sessions of mechanically stable high-quality electrophysiology recordings.
Our cranial windows provided good quality imaging several weeks after implantation. Cranial windows and
ISI maps are shown for all individual mice (Fig. S3; n = 10 mice imaged from 1 to 6 weeks after implantation). We
observed strong correlations between first day and subsequent day maps (average 15.5 days later), both relative
to a fixed reference map within mouse (Fig. S3A; Pearson r = 0.82 ± 0.07 across all imaging days; n = 4 mice with
long term imaging). The general steps of the imaging protocol and references to all MATLAB code required for
image acquisition, processing, and troubleshooting are described in Table 1.
We also added online processing to our system to facilitate troubleshooting during the experiment. Before
experimenters commit to multiple repeated blocks of red light imaging, green light images were acquired and rap-
idly analyzed (with temporal compression) to provide feedback of signal quality to the user (Fig. 1C). Although
green light imaging captures spatially coarse signals due to changes in blood volume, vasculature dilation, and
capillary blood recruitment in addition to cortical activity21, the higher SNR of green versus red light signals
provides a robust estimate of overall hemodynamic signals in few trials. If coarse hemodynamic signals appear
poor, users can immediately proceed to common troubleshooting measures (detailed in Table 2).
To further increase signal quality control, we added an algorithm to detect and discard individual trials with
poor signals that degrade average retinotopic map quality (Fig. 1D). This step is consistent with previous studies
showing that noisy frames degrade imaging results22. The algorithm first extracted the Fourier phase and ampli-
tude of the signal at the stimulus drift frequency for each trial of red imaging (Fig. 2F). The normalized variance
of each trial phase map was then computed. Phase maps with low variance were found to consist mostly of noisy
trials without stimulus-driven pixel changes in expected ROIs. We found that a threshold for discarding frames
with normalized variance < 0.6 provided the best results across all experiments. To increase user control of this
process, the code displays averaged phase maps from each block of trials at different variance thresholds, and
the user is free to adjust this. On average ~ 10% of trials per session were discarded due to noisy image data, and
these tended to cluster at the beginning of imaging sessions, when anesthesia level transitioned from induction
to maintenance. This suggests a main source of noise arose from anesthetic depth and effects on neurovascular
coupling7. These noise-reduction procedures likely improved detection of HVAs in fewer trials, since HVAs are
smaller than V1 and more sensitive to small changes in signal quality.
We also quantified the effect of smoothing parameters on map reliability in V1 and HVAs. We determined
the optimal amount of 2-D Gaussian spatial filtering that improves the clarity and sharpness of retinotopic maps
and visual area borders (Fig. 1E), without distorting ground-truth retinotopic coordinates. We swept through 6
spatial filtering parameters and computed the correlation between smoothed areal maps and a (fixed) reference
map. Areal reliability of V1 and HVAs was significantly impacted with parameters lower or greater than the
optimal Gaussian kernel width (found to be σ = 3–5 pixels, equivalent to 18–30 µm, similar to prior studies4,7,8);
similarly, the smoothness and steepness of borders between V1 and HVAs was also compromised when using
non-optimal smoothing parameters.
Finally, we implemented an algorithm for improved and automatic alignment of image frames across sessions
and days from the same subjects (Fig. S1D). This enables investigators to perform multiple short imaging sessions
across multiple days and align all image frames to produce an averaged map across sessions. This reduces the
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
2
Vol:.(1234567890)www.nature.com/scientificreports/ Figure 1. Experimental setup and optimization steps for transcranial intrinsic signal imaging. (A, B) Position
of mouse and monitors displaying visual stimuli. Eyes were vertically and horizontally centered at each monitor
(~ 19 cm away from each), and these formed a right angle. A primary computer controlled the main system
components: light intensity (Red light: λ = > 610 nm; Green light: λ = ~ 525 nm); a complementary metal–
oxide–semiconductor (CMOS) camera coupled to a tandem lens macroscope; and a photodiode recording
the timing of visual display events. A secondary computer displays visual stimuli and communicates via UDP
with the primary computer. The visual stimulus is a contrast-reversing (6 Hz) checkerboard pattern drifting
across the screen (0.055 Hz). Stimuli drifted from left to right (and right to left) to map retinotopy in azimuth
and drifted from bottom to top (and top to bottom) to map elevation. (C) Experiments start with a coarse test
for hemodynamic signals under green light (see Table 1, “Methods” section), then move to acquisition with
red light. If green light imaging failed to generate high quality signals (top, before), the illumination and focal
plane was adjusted (bottom). Color scale shows normalized signal intensity and does not correspond to visual
space. Scale bar = 1 mm. (D) After signal optimization, red light imaging commences. Algorithms exclude noisy
frames based on a minimum signal to noise (SNR) threshold for periodic responses at the stimulus frequency
(see “Methods” section), resulting in higher quality maps (bottom). Resulting absolute phase maps are shown
(− 10° to 120° in azimuth). (E) Optimal spatial filtering of high SNR frames defines clear areal boundaries in
visual field sign maps (see Fig. 2C and Fig. S4C–F). Visual field sign (VFS) maps shown, scale from − 1 (sign
negative areas) to 1 (sign positive areas). (F) Retinotopic maps, VFS maps and area contours (top) are aligned
to vasculature images (bottom, acquired with green light) for registration of areas with visible vasculature
landmarks. Investigators assess alignment, coverage range, size, and location of areas relative to expected6. See
also Table 2. (C–F) All in same mouse (Mouse 1; Fig. S3).
need for prolonged anesthetized experiments, minimizing physiological stress associated with recovery from
anesthesia23, which could be particularly beneficial in subjects undergoing concurrent behavioral training in
challenging tasks18,24,25.
These modified procedures enable construction of high-quality transcranial retinotopic maps, comparable in
quality to previous studies where the cranium is thinned or removed (including dura removal in some cases)2,5,7.
Typical visual field coverage spanned − 30° to 30° in altitude, and − 10° to 120° in azimuth (Fig. 2A). A frequency
domain analysis of reflectance from V1 in representative retinotopic maps showed a clear peak at the drift
frequency of the visual stimulus (Fig. 2F), and a sharp change in reflected light intensity at the expected spatial
location (Fig. 2G; slightly displaced due to expected hemodynamic signal lag), and no such changes in adjacent
non-visual cortical regions. Our protocol and code also provide semi-automated alignment of the retinotopic and
visual field sign maps to the vasculature observed through the cranial window (Fig. 2D,E; Fig. S2D). This step is
important because it allows investigators to align the ISI retinotopic maps and VFS maps to visible vasculature
landmarks that guide selection of sites for craniotomies.
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
3
Vol.:(0123456789)www.nature.com/scientificreports/ Step #
Time (min)
Software used
Description
primary_script.m secondary_script.m The camera is focused to the cortex and acquires images
Anesthesia, sedation, camera placement
Experiment stages
Evoked signal detection (Fig. 1C)
Prep for retinotopic map acquisition
Azimuth map acquisition
Elevation map acquisition
5
10
0.5
0.5
1
2
3
4
5
6
Matrox Intellicam
primary_script.m; secondary_script.m
30.5–61
primary_script.m; secondary_script.m
30.5–61
of vasculature. A test experiment (green filter and
illumination) with a single block of visual stimulus pres-
entation is used to detect global hemodynamic signal
Adjust focus to intracortical plane (~ 0.1–0.5 mm below
cranial surface)
Switch to red filter and illumination
Load horizontal.param in primary_script.m and run
script for visual stimulation in azimuth
Load vertical.param in primary_script.m and run script
for visual stimulation in elevation
Load Post-imaging folder and run run_first.m to con-
struct retinotopic and VFS maps. Processing includes:
– Retrieval of slow fluctuations in intensity across all
imaging frames (frequency domain)
– Azimuth and elevation phase maps for each trial
– SNR thresholding
– Average azimuth and elevation retinotopic maps per
session
– Combination of maps from multiple imaging sessions
– Overlay of retinotopic, VFS, vasculature
Load Post-imaging folder and run Align.m to align
craniotomy images to retinotopic contours
Retinotopic map analysis and display (Fig. 1F; Sup.
Fig. S3)
7
10–20
run_first.m
Alignment of craniotomies to retinotopic contours
(Fig. 4B)
8
5
Align.m
Table 1. ISI protocol.
Case
Intervention
Comment
Failure to detect signal (Table 1, steps 1–2)
a. Adjust camera focus deeper
b. Adjust camera position
c. Decrease/increase light intensity
d. Adjust anesthesia level
a. Focusing the camera at inappropriate level (e.g. at the dural surface) reduces
signal amplitude (see “Methods”, “Imaging procedures” sections)
b. Cranial window must be positioned at the center of the camera, reflecting the
brightest light
c. Hemodynamic signals evoke small reflectance intensity changes. Too little light
may prevent signal detection
d. High levels of anesthesia impair visual signals. It is important to pair sedation
with minimal maintenance anesthesia during imaging
a. Clear all or Restart MATLAB
a. Clear workspace or restart MATLAB to flush memory
Software or Image acquisition freezes/crashes
b. Unplug/plug data acquisition (DAQ) device
b. DAQ device may power off if software freezes, preventing additional recordings
c. Reboot primary computer
c. Extreme case, only when the experimenter is unable to acquire images
Table 2. Troubleshooting.
Data length requirements for reliable and repeatable retinotopic maps. We next determined
the optimal amount of data needed to generate reliable and repeatable retinotopic maps in our conditions. Prior
work with thin skull or excised skull cranial windows suggests that ~ 100 sweeps of visual stimuli are required
for adequate retinotopic and VFS maps; however, to the best of our knowledge, these suggested sample sizes lack
clear quantitative justification5,8. Further, it was also possible that transcranial ISI would require significantly
more data to reliably resolve signals from V1 and HVAs. Therefore, we first determined the amount of data
necessary to construct well-defined retinotopic maps. Our general strategy was to compute retinotopic maps
using different subsampled amounts of trials and to compare these to an overall ‘reference’ retinotopic map
constructed from all eligible trials across all imaging sessions within subject (> 190 trials over several days; see
“Methods” section). We measured the centers (centroid) and extent (boundaries) of each identified cortical
area defined by the VFS reference maps. We then plotted the average Euclidian distance (error) between V1
centroids derived from the various subsampled maps and the reference map. We found that the average error
fell below 100 μm after ~ 60 visual stimulus trials were included for map generation (Fig. 3A,B). We defined an
error of < 100 μm acceptable since this is the average size of craniotomies made for silicon probe recordings. We
found no significant difference between V1 centroids in the reference map versus subsampled maps constructed
from > 60 visual stimulus trials (p > 0.05, Mann–Whitney U with Bonferroni correction). This minimum trial
number was established from one mouse (Fig. 3) then tested and confirmed with 4 other mice (Supplementary
Fig. S4B; Average error of 80 ± 16 μm after 69 ± 5 trials, mean ± SD).
We next determined the amount of data needed to estimate the boundaries of V1. To do this, we used the
reference VFS map to define the extent of V1 (Fig. 3C; reference map constructed from > 190 single trial azimuth
and elevation maps; see Methods). We then used the same subsampling strategy to determine the minimum
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
4
Vol:.(1234567890)www.nature.com/scientificreports/ Figure 2. Retinotopic maps of azimuth, elevation, and visual field sign with intact skull ISI. (A) Retinotopic
maps in azimuth from 2 mice (Mouse 1 and 2; see Fig. S3) with 5-mm intact skull cranial windows. Colorbar
shows spatial location of visual stimulus driving maximal Fourier response at each image pixel (see “Methods”
section). Scale bar is 1 mm. (B) As in (A), for elevation maps in same mice. (C) Visual field sign (VFS) map for
mouse 1 showing cortical area boundaries, computed from azimuth and elevation maps. Scale is − 1 to 1 (see
“Methods” section for calculation). Identified areas V1 (primary visual cortex), Area P (posterior aspect of visual
cortex), LM (lateromedial), AL (anterolateral), RL (rostrolateral), AM (anteriomedial), PM (posteromedial),
MMA (Medio-medial-anterior), and MMP (Medio-medial-posterior). Identified areas correspond to prior
reports with excised skull cranial windows6. (D) Overlay of VFS (white) and azimuth (red) retinotopic map
contours (10° increments) on image of vasculature of Mouse 1. Custom processing in finalized software package
allows users to define visual areas from VFS map, and automatically align maps to vasculature. (E) As (D), for
elevation map contours. (F) Power spectrum of raw reflectance from multiple pixels (black circle in A) across
2880 imaging frames and 16 stimulus trials (grey traces). A peak is present in the average response (black) at the
frequency of visual stimulus (0.055 Hz, red dashed line). Blue trace shows power spectrum of pixels in adjacent
non-visual cortical area from same trials. (G) Average intensity versus stimulus position averaged across 200
stimulus cycles. Maximum intensity change (decreased reflectance) occurs near preferred stimulus location
expected from azimuth map (black circle in A). Maximum slightly displaced due to hemodynamic delay for
stimulus drifting from negative to positive azimuth (Fig. 1A). Blue trace shows cycle average for pixels outside of
visual areas.
number of trials necessary for resolution of V1 from the background signal. All pixels within V1 in the reference
VFS maps were considered “signal”. We then defined a “noise” region outside of V1 that also did not contain any
HVA (Supplementary Fig. S2D). The separability of the signal and noise regions were compared in the various
subsampled versus reference VFS maps to define the minimum amount of data needed to define the extent of
V1. The analysis revealed that a clear separation between V1 signal versus noise distributions starts when > 43
visual stimulus trials were used to construct a VFS map (Fig. S4A). This finding was consistent with the visual
inspection of VFS maps (Fig. 3C).
We next used this same procedure to identify the minimum number of trials needed to resolve multiple HVAs.
ROC analysis was performed to determine whether pixel intensity (signal) in a visual cortical region of interest
(ROI) is distinguishable from intensity in an adjacent non-visual cortical region (noise). Visual cortical ROIs
were determined using the reference map constructed from all trials, and an adjacent non-visual ROI defined
the noise distribution (Fig. S2D). A classification boundary was determined and then applied to ROIs in maps
constructed from subsampled data with increasing numbers of trials. The curve plotting the ratio of true positives
to false positives determines the accuracy of the classification boundary, with an area under the receiver operat-
ing characteristic curve (AUROC) of 0.5 (diagonal line in Fig. 3D) equal to chance level classification. We then
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
5
Vol.:(0123456789)www.nature.com/scientificreports/ Figure 3. Quantifying resolvability of V1 and HVAs as a function of trials. (A) Error of estimated azimuth
receptive field (RF) locations in V1 as a function of number of trials. Retinotopic contour maps computed from
varying trial numbers (abscissa), and error estimated as Euclidean distance of RF locations in trial-limited
maps versus reference map (mean distance ± SD; centroids of contours binned at 10°, Fig. 2D). Reference map
computed from 190 trials (5 recording sessions), trial-limited maps computed by randomly subsampling from
these 190 trials with (grey) or without (black) replacement. Error in estimated RF locations falls < 0.1 mm
(dashed line) within 60 trials. No significant difference in centroids of azimuth contours for reference maps
and resampled maps with > 60 trials (see Results). (B) Same mouse as (A), for elevation RFs. Reference maps
from 170 trials (5 recording sessions). No significant difference in centroids of elevation contours (e.g., Fig. 2E)
between reference maps and resampled maps with > 60 trials (p = 0.325, Mann–Whitney U with Bonferroni
correction). (C) Trial limited VFS maps (left) versus reference VFS map (right). Same sessions as (A,B).
Reference map shows contours for V1 and only 3 HVAs for clarity: LM (lateromedial area), RL (rostrolateral),
and PM (posteromedial). (D) Receiver operating characteristic (ROC) curves computed for centroid detection
of V1 and 3 sign positive HVAs (LM, RL, PM) for same trial limited VFS maps in (C). True positives evaluated
as average pixel values in areal contours defined from reference VFS maps versus noise areas outside of visual
cortex (Fig. S2). Areal boundaries for V1, RL, and PM pass detection threshold (75% accuracy, + symbol)
after > 50 visual stimulus trials. LM areal boundaries pass threshold after > 70 trials. Sign negative areas AL and
AM were also readily identified with > 70 trials (e.g., Fig. 2C). (A–D) All from Mouse 1 (Fig. S3).
evaluated classification performance in the visual vs non-visual ROIs from subsampled VFS maps constructed
with increasing numbers of trials.
We generated receiver operating characteristic (ROC) curves for primary visual cortex (V1), and for latero-
medial area (LM), rostrolateral area (RL), and posteromedial area (PM) to assess our ability to detect these HVAs
from noise as the number of trials increases. For this analysis, the size and extent of each area was determined
from a reference VFS map. Then, using similar resampling methods as described previously, multiple subsampled
VFS maps were created by aggregating different numbers of visual stimulus trials (Fig. 3C). The area under the
ROC curves (AUROC) comparing the pixel intensity inside and outside these visual areas shows that the full
extents of visual areas PM and RL were detectable at 75% accuracy level after ~ 54 trials (Fig. 3D), while area LM
necessitated ~ 75 trials. Similar trial duration criteria were found even when we did not constrain the pixel area
by the VFS reference map, but instead analyzed the pixel SNR at centroids of sign positive (or negative) areas
detected in maps generated with increasing numbers of trials. Although we only quantified detectability of areas
LM, RL, PM, the reference maps also show that areas P, AL, AM, and several others were clearly resolvable after
75 trials (see Fig. 2C). These other areas are not analyzed in detail here since they were not extensively targeted
for electrophysiology, discussed next.
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
6
Vol:.(1234567890)www.nature.com/scientificreports/ Validating intrinsic signal imaging with targeted electrophysiology. We next used ground-truth
extracellular electrophysiological recordings to validate retinotopy estimated in the ISI maps of V1 and HVAs.
The cranial window was removed by carefully drilling away the Metabond surrounding the coverslip, then
detaching it and the supporting Vetbond. Once the window was removed for electrophysiology, we were able to
visualize vascular landmarks and use these to target multiple craniotomies to primary and higher visual areas for
multiple days in a row (n = 10 mice, 4–14 recording days), with high single unit yield and low noise signals. Cra-
niotomies were performed in the primary visual cortex (V1), and higher visual areas: AL, LM, AM, PM, and RL
(Fig. 4A). We aligned and overlaid the azimuth and VFS maps on the vasculature image to target specific retino-
topic regions of the visual areas (Fig. 4B). For instance, LM and RL were expected to respond to a wide extent of
azimuthal visual space (~ 0° to 100°). However, PM was expected to be responsive to stimuli in a more restricted
azimuthal portion of monocular visual space (45° to 80°), consistent with previous findings8.
Extracellular recordings were performed to determine the preferred stimulus position in both azimuth and
elevation for local field potential (LFP) responses24,25. Laminar LFP responses were separated into superficial and
deep cortical layers (see “Methods” section). In V1, we found a high correlation between retinotopic coordinates
estimated from ISI and electrophysiology in superficial cortical layers (Fig. 4C; r2 = 0.82, p = 3.8e−31; average
error: 8.2° ± 7.6°, n = 22 recordings in 6 mice). During these same recordings, deep layer LFPs also showed
significant but weaker correlation to ISI coordinates (Fig. 4D; r2 = 0.51, p = 5.2e−14) and with greater average
error (13.4° ± 13.1°). Nevertheless, the error between ISI versus LFP retinotopic coordinates was not statistically
significant in either superficial (p = 0.39) or deep layers (p = 0.065, Wilcoxon signed rank tests). Due to experi-
mental considerations and time constraints during electrophysiological recordings, we prioritized verification
of azimuth retinotopy in V1 and HVAs (discussed next). However, in a subset of V1 recordings (n = 3 mice, 13
recordings), we also measured elevation retinotopy. We again found low error between ISI maps and V1 LFP
elevation retinotopy in superficial layers (3.5 ± 2.6°), compared to deep layers (6.9 ± 2.8°; mean ± SD). Accounting
for error in both azimuth and elevation, the total estimated error in Euclidean space was 8.9 ± 8.0° (mean ± SD)
in superficial layers and 15.1 ± 13.4° in deep layers of V1 (n = 35 recordings).
The correlation between ISI and LFP retinotopy in HVAs was also significant in all recordings, and greater in
superficial layers (Fig. 4E; r2 = 0.74, p = 2.6e−13) versus deep layers (Fig. 4F; r2 = 0.54, p = 2.9e−8). In HVAs, only area
LM exhibited a statistically significant difference between ISI and LFP retinotopy in both superficial (p = 0.023)
and deep cortical layers (p = 0.0027; Wilcoxon signed rank tests). No significant differences were found between
retinotopy estimated from ISI versus LFP in superficial or deep layers in areas AL (superficial, p = 0.625; deep,
p = 0.375), PM (superficial, p = 0.195; deep, p = 0.0781), and RL (superficial, p = 0.583; deep, p = 0.1721).
Finally, in a subset of experiments, we measured functional properties of single neurons in HVAs and com-
pared these to benchmark literature. We isolated single neurons in V1 and 5 of the HVAs targeted by ISI (LM,
AL, RL, AM, PM) with silicon probe recordings in awake mice (n = 734 neurons, 3 mice, 29 recordings). Spatial
and temporal frequency (SF and TF) tuning in regular spiking (RS) putative excitatory neurons in HVAs tar-
geted by ISI showed broad consistency with prior reports, although small samples sizes largely precluded robust
findings of statistical significance across all areas. We found a significant main effect of visual area on SF tuning
(p = 0.049), with LM preferring higher SFs than RL (Fig. 4I; Kruskal–Wallis tests with Bonferroni correction for
all comparisons). There was no significant main effect of visual area on TF tuning (p = 0.17). However, when
examining preferred stimulus speed (the ratio of preferred TF to preferred SF), we found a significant main effect
of visual area on speed tuning (p = 0.0148), with RL preferring higher speed stimuli than V1 and PM, broadly
consistent with prior findings in V1, RL, and PM26. Again, these findings should be interpreted cautiously given
low sample sizes relative to prior studies that sampled thousands of neurons26. Other differences between our
and prior results include measuring neural activity across all layers, measuring spikes with silicon probes, not
imposing single neuron inclusion criteria27, and measuring in awake stationary mice (rather than prior studies
measuring Ca2+ responses from highly responsive neurons only in L2/3 of anesthetized26 or running28 mice).
Overall, our findings suggest RS neurons prefer different stimulus speeds in V1, RL, and PM, a topic for future
electrophysiological studies. These results, alongside verifications of expected retinotopy in HVAs (Fig. 4E,F),
provide further evidence of the viability of our system for ISI targeting of HVAs with subsequent multi-site
electrophysiology in awake mice.
Discussion
Here, we validated the performance of an intrinsic signal optical imaging (ISI) system optimized for intact skull
imaging in the adult mouse visual system. We characterized the quality, resolution, and trial dependence of
retinotopic maps in multiple visual cortical areas, and then validated these with targeted electrophysiological
measurements of retinotopy and functional properties. Our intact skull imaging in adult mice matches well-
established benchmarks for thinned skull or excised skull cranial window preps2,5,7,29, but provides a specific
advantage for investigators wishing to identify retinotopic maps and then perform subsequent visually targeted
multi-site electrophysiology, perhaps weeks or months later (e.g. in mice undergoing training in behavioral
tasks). Maintaining skull integrity over weeks to months ensures optimum mechanical stability and physiological
conditions during sensitive electrophysiological recordings, particularly high-density multi-site silicon probe
or patch clamp recordings13,14. We also improved quality control and alignment algorithms so that multiple
imaging sessions within and across days can be readily combined; this enables multiple short-duration imaging
sessions to be aggregated to resolve small or low-signal HVAs, rather than necessitating a single long-duration
imaging session. We provide all necessary details to replicate these procedures and have made all code and
methods available for those wishing to implement this minimally invasive ISI imaging that is readily combined
with subsequent targeted, multi-site electrophysiology.
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
7
Vol.:(0123456789)www.nature.com/scientificreports/ Figure 4. Validation of ISI map retinotopy in V1 and HVAs with electrophysiology. (A) Example overlay of VFS and azimuth
retinotopy. Note HVAs show distinct retinotopic coverage (e.g., LM versus PM). (B) Overlay of VFS (black) and azimuth retinotopic
map contours (blue, 10° increments) aligned with vasculature. White circles: sites and average size of craniotomies after cranial
window removal and alignment to vasculature. Location of craniotomies is used to determine expected azimuth RF location within V1
or HVAs. Data in (A,B) from Mouse 1 (Fig. S3). (C) Correlation between expected ISI azimuth coordinates (abscissa) versus observed
RF location from local field potential (LFP) responses (ordinate) in superficial layers of V1 (n = 6 mice, 22 recording sessions). Error:
8.2° ± 7.6° (mean ± SD). Overall r2 = 0.82, p = 3.8e−31; Black stimulus r2 = 0.84, p = 5.9e−17; White stimulus r2 = 0.80, p = 2.4e−15. no
significant difference between expected and observed (p = 0.39) and white and black not significantly different (p = 0.46, Wilcoxon
signed rank test). Blue circles indicate recordings from mouse and sites in (B). Data from mice 1,2,3,5,6,7 in Supplemental Fig. S3.
(D) As (C), for deep V1 layers. Error: 13.4° ± 13.1° (mean ± SD). Overall r2 = 0.51, p = 5.2e−14; Black stimulus r2 = 0.61, p = 2.1e−9; White
stimulus r2 = 0.43, p = 3.0e−6. No significant difference between expected and observed (p = 0.07). (E,F) Like (C,D) for higher visual
areas AL, LM, PM, and RL (n = 4 mice; 15 recording sessions). Superficial layers of HVAs show greater correlation to ISI coordinates
(r2 = 0.74, p = 2.6e−13) than deep layers (r2 = 0.54, p = 2.9e−8). Data from mice 1, 2, 8, 9 in Supplemental Fig. S3 are shown here. (G–I)
Regular spiking (RS) neuron temporal frequency, spatial frequency, and speed tuning in V1 (n = 82 neurons) and HVAs (LM: n = 181;
AL: n = 46; RL: n = 108; AM: n = 67; PM: n = 125). Data from mice 1, 2, 9 (Fig. S3). No significant effect of area for TF tuning (p = 0.17).
Significant main effect of area on SF tuning (p = 0.049), with LM preferring higher SFs than RL. Significant main effect of area on speed
tuning (p = 0.0148), with RL preferring higher speed stimuli than V1 and PM. Kruskal–Wallis tests with Bonferroni correction for all
comparisons. (J–L) Same sessions as (G–I), for fast spiking (FS) neurons in V1 (n = 27 neurons) and HVAs (LM: n = 19; AL: n = 4; RL:
n = 28; AM: n = 13; PM: n = 34). No significant effect of area on FS tuning properties (TF: p = 0.615; SF: p = 0.057; Speed: p = 0.624).
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
8
Vol:.(1234567890)www.nature.com/scientificreports/ Development of this system also allowed us to provide quantification for the relationship between imaging
duration and the resolvability of retinotopic maps in V1 and HVAs. Unlike previous research that has thoroughly
inspected the variations in location and size of V1 and HVAs using excised skull cranial windows6, our study
sought to determine the minimum number of trials (and thus minimum duration of anesthesia) needed to
accurately resolve V1 and multiple HVAs through an intact skull cranial window. We found that displaying the
visual stimulus for 50 to 60 trials in forward and reverse directions (~ 65 to 75 min) is sufficient to generate high-
quality retinotopic maps that define the extent of V1 and 2 commonly investigated HVAs (PM, RL). ROC analysis
revealed that ~ 60 visual stimulus trials identify retinotopy and delimit borders for V1 and these HVAs with > 75%
accuracy, with ~ 90 trials needed to readily identify V1, LM, RL, AL, AM, PM (Fig. 2C). Somewhat surprisingly,
resolving the full extent of area LM required the most trials. This could be because the full extent of LM can only
be defined once V1 and AL (sign negative areas) and RL and Area P (sign positive) are also resolvable.
Comparison of our maps to prior studies reveals additional factors to consider for the resolvability of HVAs.
First, imaging through thinned or excised skull for longer periods of time will yield better identification of areas
beyond the main group of lateral (LM, RL, AL) and medial (AM, PM) HVAs, an important consideration for
targeted investigation of such areas30; Second, VFS maps constructed from widefield GCaMP6 fluorescence—a
direct neuronal signal—provide higher resolvability and faster identification of HVAs than ISI maps2,6, providing
advantages over hemodynamic imaging but limiting experiments to transgenic mice expressing calcium indica-
tors. Other benchmark studies of hemodynamic ISI mapping that use both intact skull and transcranial imaging
show high resolution maps8,9, but the exact skull preparation and imaging durations generating the exemplar
maps are not specified and thus difficult to directly compare with ours; one of these prior studies acquired maps
with very long periods of imaging (up to 6 h) in acute tracheostomized subjects. Although many prior studies
have used intact skull transcranial preps for acute ISI in juvenile29 or adult mice15,31,32, our study specifically
developed a chronic transcranial window for adult mice that (1) generates high quality ISI maps of V1 and HVAs
while minimizing the extent of anesthetized imaging (2) allows visualization, monitoring, and maintenance of
cranial and cortical health for weeks to months, and (3) facilitates visually targeted multi-site electrophysiology
from V1 and HVAs within subjects, discussed next.
Validation of this system with electrophysiology allowed us to quantify the relationship between retinotopy
inferred by ISI maps versus retinotopy measured from neural activity across cortical layers. We confirmed that
ISI retinotopy showed significant correlation with LFP retinotopy in V1 and multiple HVAs; furthermore, ISI
retinotopy corresponded most closely with neural activity acquired from the superficial layers of cortex, with an
error (± 8.2°) comparable to the width of the visual stimulus presented during electrophysiology experiments (9°).
This error is much smaller than the average receptive field size of V1 excitatory neurons (between 15°–30°)33,34.
ISI also showed significant correlation with retinotopy in deeper layers, but with greater error (± 13.4°), con-
sistent with prior observations that deep layer V1 neurons show larger RFs and greater retinotopic scatter33,34.
These findings carry some limitations. First, we did not systematically verify that each electrode penetration was
completely perpendicular to the cortical surface, which could contribute to greater variability in our deeper layer
measurements; second, electrophysiology was performed in both anesthetized and awake mice, which could
contribute to greater variability. Nevertheless, to the best of our knowledge, our study provides the first error
estimates for retinotopy inferred from ISI maps versus laminar-specific neural activity across multiple mouse
visual cortical areas. In all cases, measures of retinotopy inferred from transcranial ISI showed significant cor-
relation with direct electrophysiological measures in V1 and HVAs. These findings also provide considerations
for future studies of laminar-specific neural activity underlying ISI signals. Finally, functional visual properties
of single neurons in V1 and HVAs in our ISI targeted recordings showed some consistency with benchmark
literature26, providing a second independent electrophysiological metric of ISI map fidelity.
Our system and protocol were optimized for both novice and experienced users, yet some limitations remain.
First, the quality of signals and retinotopic maps depends critically upon the clarity and stability of the window;
this requires some skill and experience for success but is no more difficult than many other in vivo mouse pro-
cedures requiring careful execution (e.g., headplate implantation, cannulation, stereotaxic injections). Second,
the system and protocol has only been optimized with a single brand of CMOS camera and frame grabber,
although the code could be readily adaptable to other hardware, including sCMOS cameras. Third, the system
and protocol still require human intervention (e.g., adjustment of camera focus, light intensity, or anesthesia
level), but this is described here step-by-step. Fourth, our quantitative assessment of differences between ISI vs
neural signals mainly considered azimuthal retinotopy, but future studies could consider other aspects of visual
selectivity. Finally, our methods for transcranial ISI and subsequent electrophysiology seem readily testable in
other sensory systems35,36.
In summary, this system provides a way for investigators of the mouse visual system to pair well-established
hemodynamic mapping of visual cortical brain activity9,37, with subsequent long-term, stable, retinotopically
targeted neural recordings across multiple cortical visual areas. We have characterized system performance for
minimally invasive intact skull cranial windows and measured the minimum amount of anesthetized imaging
data required to infer reliable retinotopic maps of V1 and several HVAs. Minimizing the need for bone thinning,
bone or dura removal, or recovery from long bouts of anesthesia is particularly advantageous to investigators
using mice for experiments requiring manipulations lasting weeks to months (e.g., behavioral tasks, plastic-
ity studies, recovery of visual function, studies of aging). Lastly, the minimally invasive requirements for high
resolution ISI mapping shown here may be more amenable for sensitive or costly transgenic strains, therefore
expanding capabilities for experiments that require precise targeting of cortical visual areas to study mouse
models of neurological disorders.
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
9
Vol.:(0123456789)www.nature.com/scientificreports/ Methods
All experiments were approved by the Georgia Institute of Technology Institutional Animal Care and Use Com-
mittee (IACUC) and conform to guidelines established by the National Institutes of Health. All methods were
performed in accordance with the relevant guidelines and regulations. The study design did not use treatments
necessitating blinding, or comparison of experimental versus control groups. All descriptions of experimental
procedures, sample sizes, data analysis, resampling methods, statistical comparisons, and outcome measures are
consistent with the ARRIVE guidelines 2.0.
ISI hardware and software. All relevant software packages and toolboxes for intrinsic signal optical imag-
ing are included in the ISI package available in the lab’s public code repository (https:// github. com/ haide rlab/
ISI). The initial list of components and software was obtained from a previously published protocol5. It was then
adjusted to meet logistical requirements and conditions in our lab (e.g., updated software for acquisition and
processing, updated DAQ and camera interface, updated acquisition of frame timestamps, expansion of visual
display from 1 to 2 monitors, etc.). Briefly, our ISI system was composed of a primary and secondary computer
(control center and stimulus display), a light supply, a photodiode for temporal alignment, and a tandem-lens
macroscope for image acquisition (Fig. 1A). MATLAB R2018b was used to develop, optimize, and run all soft-
ware, and process and analyze all data.
The current system codebase has been improved in several ways relative to prior open-source ISI systems and
extended to facilitate integration with electrophysiology experiments. First, we have adapted the original source
code5 (MATLAB v. 2008) to function with modern versions of MATLAB that use different DAQ interfacing (vali-
dated here from MATLAB v. 2018 onward). Second, we have expanded the visual stimulus display to include two
monitors placed further away from the mouse, enabling placement of equipment necessary for neural recordings
in the same set-ups (micromanipulators, lick detectors, recording accessories) while still permitting stimulation
and mapping of large regions of visual space. Third, we have removed system dependency on the Matrox Imaging
Library with custom C++ code. Fourth, we deployed a method to trigger frame captures independently from visual
display draw, to timestamp these frames using the system clock, and then to align acquired frames to the start
and stop of visual stimuli and monitor frame flips using a simultaneously acquired photodiode signal. Fifth, we
have created the ability for users to quickly overlay and register visible light images of the skull and craniotomies
with user-defined fiduciary landmarks to overlay craniotomies on retinotopic maps. Lastly, we have provided a
simple GUI for system control for novice users, with an option for greater control by expert users.
To image the cortex, a wide-field camera, controlled by a frame grabber (Matrox Radient eV; Matrox), with
lenses (Lens 1: Nikon—AI-S FX Nikkor 50 mm f/1.2 manual focus lens; Lens 2: Nikon—Ai 85 mm f/2 manual
focus lens) in tandem configuration was positioned above a transcranial window. Frames were captured while
visual stimuli (see “Visual stimuli” below for details) were presented on two screens covering the binocular and
monocular visual space (Fig. 1A). Custom-built software was used to interface all hardware components, provide
feedback during hemodynamic imaging, conduct in-depth post-recording analysis, and align ground-truth
electrophysiology craniotomies to retinotopic maps.
The general steps and code are outlined in Table 1.
Common Troubleshooting steps are outlined in Table 2.
Surgical procedures. All procedures were approved by the Georgia Institute of Technology Institutional
Animal Care and Use Committee (IACUC). We present data from the same n = 10 implanted mice throughout
the study (Fig. S3), including 5 C57BL/6J (Mice 1, 2, 6, 7, 8), 3 B6PVCre x Ai32 (Mice 3, 9, 10), and 1 Sst-IRES-
cre x Ai32 mouse (Mouse 5). We did not observe obvious differences in signal quality in mice expressing a
fluorescent protein (Fig. S3). Additional imaging data from 1 CNTNAP2−/− KO mouse24 (Mouse 4) is shown
for repeatability analysis, with no contribution to electrophysiological analysis. Mouse 10 also contributed only
imaging data (Fig. S4B), so n = 8 mice were used for both imaging and electrophysiology (Mice 1–3, 5–9). All
mice were male and 4–12 weeks old at time of implant. Detailed procedures for head-plate implantation have
been described elsewhere25. Briefly, mice were anesthetized with isoflurane (3% induction, 1–2% maintenance),
body temperature was kept at ~ 37 °C using a heating pad, and the eyes were protected with veterinary oph-
thalmic ointment (Puralube). The skin was removed, and the fascia and periosteum overlying the skull were
carefully resected and removed with a cotton bud and/or scalpel blade (no. 11) under physiological saline solu-
tion, avoiding scratches or bleeding of the cranial surface. No removal or thinning of the skull was performed
following this step. Once dural and cortical vasculature was cleanly and clearly visible under saline, the skull was
allowed to air dry thoroughly. A custom-built stainless steel head post with a recording chamber (11 mm inner
diameter) was lightly affixed to the skull using veterinary adhesive (Vetbond) (Supplementary Fig. S2). Follow-
ing headplate fixation, a glass coverslip (5 mm diameter, #1 thickness ~ 0.15 mm, Warner CS-5R) was centred
over the representation of V1 and HVAs (centre of window at ~ 2.4 mm lateral to midline and ~ 2.4 mm anterior
to lambda) and bonded to the skull using Vetbond (Supplementary Fig. S2). The layer of Vetbond between the
glass window and skull was allowed to fully dry (45–75 min), leaving a fully transparent transcranial view of
cortical surface vasculature. The edges of the cranial window were then sealed with dental polymer (Metabond),
and the headplate was fully bonded to the skull. Mice were individually housed and monitored for full recovery
for at least 3 days before imaging.
Imaging procedures. Mice were anesthetized with isoflurane (3% induction), given a sedative via intra-
peritoneal injection (Chlorprothixene, 10−5 mg/kg), and placed on a heating pad to maintain body temperature
(~ 37 °C). 3-mm contact lenses (Ocuscience) were inserted to prevent dehydration of the eyes and maintain
ocular clarity during imaging sessions. During imaging anesthesia was lowered to 0.5–0.75%. The cortex was
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
10
Vol:.(1234567890)www.nature.com/scientificreports/ illuminated using fiber optic guides and a high intensity tungsten halogen lamp (Illumination technologies
3900E, 9596A lamp) that emits high intensity light across a broad wavelength spectrum (Fig. S2B). We placed a
bandpass green filter (450–600 nm; Illumination Tech P/N 9542) between the light source and fiber optic guides
to better isolate signals from vasculature, or used a longpass red filter (> 610 nm; Illumination Tech P/N 9541)
to better isolate changes in blood oxygenation (see discussion below). A filter wheel (Thorlabs LCFW5) with
bandpass emission filters (Edmund Optics) was installed between the macroscope lenses to capture reflected
photons from green (525 ± 25 nm) or red (700 ± 10 nm) imaging, as in prior studies5,29,38. A CMOS camera
(Falcon2; Teledyne DALSA) acquired images of the cortex at a frame rate of 10 Hz. We choose to use a CMOS
camera to maintain system consistency with prior work, but the integration of sCMOS cameras is an avenue for
future system optimization.
The general procedure for finding the optimal focal plane consists of the following steps. First, the mouse is
positioned and secured on the recording rig, and the camera is placed above the cranial window, and brought to
maximal height. Green light and filters are enabled, and the height of the camera is decreased just until blood ves-
sels are clearly visible. Then, using a calibrated scale attached to the camera, we lower the focal plane by another
0.5 mm so that vasculature appears slightly blurry. We then proceed to perform a test experiment (with green
light). If the test fails to show adequate signal, the experimenter can perform a series of interventions, including
re-adjustment of the camera focus. The best rule of thumb is to re-focus on the vasculature then decrease the
focal plane by greater than 0.5 mm, and proceed with another test acquisition. This step can be iterated up to
1 mm below the vasculature surface, but no deeper. If the experimenter is not able to find a clearly optimal focal
plane, imaging just below the level of vasculature is usually adequate for subsequent red light imaging.
The imaging wavelengths of our system are based on prior established studies of intrinsic signal imaging in
mice5,7,8,26,39. Wavelengths 510–590 nm (green) primarily isolate changes in blood volume that result from a com-
bination of vasculature dilation, capillary blood recruitment and cortical activity21. Wavelengths > 600 nm (red)
permit better isolation of the oxymetric component of the hemodynamic signal, due to differences in sensitivity
of light absorption for deoxygenated versus oxygenated hemoglobin. Imaging the brain at these wavelengths
produces ISI maps that are more spatially correlated to the underlying neuronal response than maps produced
from shorter wavelengths that primarily reflect widespread blood volume changes21,40. Although some studies
suggest optimal signal to noise ratio for oxymetric signals at 610–630 nm41, at ~ 700 nm there is a substantial
increase in the relative absorption differences between deoxygenated versus oxygenated hemoglobin42, which
further accentuates detection of active metabolic changes due to neural activity. These considerations motivate
the wavelength choices in our system, as well as in recent studies that use identical wavelengths for cortical ISI
in mice5,29,38.
Visual stimulus for ISI. Mice were positioned in front of two computer monitors that were at right angles
from one another (Fig. 1A). Mice were facing the center of one monitor covering the binocular visual field, and
90° from the center of the second monitor covering the monocular visual field. Stimuli were presented on lin-
earized LCD monitors (60 or 80 Hz refresh rate; Dell U2419H with maximum brightness of 250 cd * m−2; mean
[black, grey, white] screen luminance during recordings of [0, 112, and 238] cd * m−2). The stimulus was a 20°
wide with black and white full contrast reversing checkerboard (6 Hz), drifting at 0.055 Hz across the visual field
on a black background. The horizontal drifting stimulus was corrected for spherical visual coordinates26.Our
decision to use a continuous reversing checkerboard drifting bar to drive the cortex was to maintain consistency
with stimuli used to generate ISI maps in prior studies5,9,26,43.
Screens were positioned ~ 19 cm away from the eyes, and the mouse was vertically positioned at the midpoint
of the screens (~ 17 cm above air table) (Fig. 1B). The vertical and the horizontal planes through the eyes were
used to define the origin of azimuth and elevation visual coordinates (0° directly in front of the mouse). Visual
stimulus presentation consisted of a checkerboard drifting in any of the four cardinal directions (Nasal-Temporal,
Temporal-Nasal, Superior-Inferior, Inferior-Superior) for 18 s. Each block consisted of 10 unidirectional sweeps
lasting a total of 180 s. Imaging sessions were comprised of multiple blocks. Pairs of stimulus sweeps (in oppo-
site directions) generally defined a single trial for absolute phase map construction (see below, “Hemodynamic
correction”).
Images acquisition, processing, and quality control. Signal quality check with green light imag-
ing. At the beginning of each experiment, a short test acquisition (Duration: 3–5 min) was performed in-focus
with the surface vasculature with green light (λ 450–600 nm) and a 525 nm bandpass filter (see “Imaging proce-
dures” section). This allowed experimenters to rapidly assess coarse visually evoked hemodynamic signals and
perform any adjustments before further acquisition (Tables 1 and 2). Common adjustments included adjusting
lighting position, intensity, and depth of focus. Adjustments and rapid test acquisitions were performed repeat-
edly until appropriate signals were detected (Table 1; Fig. 1C). Following this verification step, the camera was
focused below the cortical surface (~ 100–500 μm below the brain surface), and long imaging sessions (Dura-
tion: 1–2.5 h) were performed with red light (λ > 610 nm) and 700 nm bandpass filter to isolate changes in deoxy-
hemoglobin (HbR) concentrations across the visual cortex. Mice were subjected to one imaging session per day.
Each imaging session was comprised of multiple blocks of visual stimulus presentation in multiple directions
(see “Visual stimulus for ISI” section). Images were collected at 10 Hz (180 frames per trial).
Hemodynamic delay correction. The slow temporal sampling of ISI signals (~ 10 Hz) is adequate to capture
hemodynamic responses because they evolve slowly compared to the time course of neuronal activity44,45. Pre-
vious research has shown that the detection of adequate hemodynamic responses during anesthesia necessi-
tates visual stimuli with periods > 10 s9. Unlike electrical activity, blood perfusion-related responses are usually
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
11
Vol.:(0123456789)www.nature.com/scientificreports/ delayed by at least 1–6 s from the onset of a stimulus7,40. This hemodynamic shift can be corrected by recording
responses in one cardinal direction (forward motion) and its reverse (backward motion), then subtracting them
to create an absolute response map9. Thus, intrinsic signals are recorded with repeated continuous visual stimuli
sweeping across the screens in directional pairs: Nasal-Temporal and Temporal-Nasal for mapping azimuth, or
Superior-Inferior and Inferior-Superior for mapping elevation. Here, retinotopic maps (azimuth and elevation)
were produced from a minimum of 5 blocks/session (50 trials) in both the forward (Azimuth: Nasal–Temporal;
Elevation: Inferior–Superior) and backward (Azimuth: Temporal–Nasal; Elevation: Superior–Inferior) direc-
tions. A maximum of 10 blocks (100 trials) per imaging session were acquired. One directional pair constitutes
the definition of a trial for analysis of absolute phase maps of azimuth and elevation (Fig. 3A,B), and then
matched directional pairs (one pair for azimuth maps, one for elevation maps) constitute a trial for a VFS map
(since VFS maps are necessarily computed from the angular difference between the azimuth and elevation abso-
lute phase maps).
Images were processed following previously published methods2,8,46, with additional adjustments. First, images
from a block were aligned, cropped, and resized to the vasculature image acquired at the start of each session
using green light (Supplemental Fig. S1A–C). Then, the baseline signal (average of frames acquired during the
first 5 s when no visual stimulus was presented) was subtracted from each frame to retrieve the change in light
absorbance. This was done using the modified Beer-Lambert law46:
�A = εl�C = log 10
Ia
I0 (cid:30)
(cid:31)
where the change in light absorbance is (cid:31)A , molar absorptivity is ε , path length is l , molar concentration change
compared to baseline measurement is (cid:31)C , post stimuli light intensity is Ia , and baseline light intensity is I0.
Fourier analysis of response phase maps. A Discrete Fourier Transform (DFT) was used to extract each pixel
response at the frequency of the visual stimulus to create periodic phase maps2. Phase maps depicting the aver-
age change during each trial within a block were constructed (i.e., 10 single trial phase maps per block). A qual-
ity criterion was then applied to only select single trial phase maps that exceeded a normalized variance of 0.6
(calculated as variance of pixel intensity from the mean in absolute phase maps). Single trial phase maps that
exceeded the normalized variance threshold were combined to create the average block phase map. To overcome
the hemodynamic delay following the presentation of a stimulus, absolute phase maps were constructed by
subtracting the block averaged phase map of the backward motion (e.g., Temporal–Nasal) from the block aver-
aged phase map for forward motion (e.g., Nasal–Temporal). These absolute phase maps were then translated to
the spatial location of the visual stimulus to create azimuth or elevation retinotopic maps (Fig. 2A,B). Multiple
block averaged absolute phase maps were combined to constitute the imaging session phase map. At each pixel,
we computed the sine angle difference between azimuth and elevation maps to create a visual field sign (VFS)
map (Fig. 2C). VFS maps use features of retinotopic gradients (e.g. regions where retinotopic preferences reverse
polarity) to define the size and boundaries of the primary (V1) and higher visual areas (HVAs). Here, the direc-
tion of each pixel’s retinotopic gradient is represented as a value ranging from sign negative (− 1) to sign positive
(1). Additionally, the software aligns and overlays the VFS map on the vasculature and retinotopic contours
(Fig. 2D,E).
Alignment and overlay of retinotopic maps to vasculature. Contour maps in azimuth and elevation
were automatically aligned to the reference vasculature image, obtained with green light during post-imaging
analysis (Supplemental Fig. S1A–C). First, images were cropped to only preserve the region of interest from the
retinotopic and vasculature images. Then, the images were resized before being overlaid (Fig. 2D,E). After the
cranial window was removed for electrophysiological recordings, a reference image showing the site(s) of cra-
niotomies, vasculature, and fiduciary landmarks in the chamber was used to semi-automatically align the reti-
notopic maps using custom functions (“Align.m”). These are rigid transformations (X–Y translations) with no
rotation or warping to account for different focal planes. The investigator selects common fiduciary features in
both retinotopic map + vasculature image as well as craniotomies + vasculature image to initiate the alignment.
Neural recordings and analysis of retinotopy. The detailed steps for laminar local field potential
recordings have previously been published25,47. In short, small craniotomies (~ 100–500 μm) were made over
V1 or HVAs using the ISI maps as reference. Recordings were made with multi-site silicon probes (Neuronexus)
consisting of a single 32-channel shank spanning all layers of the cortex. Electrodes were advanced ~ 1000 μm
below the cortical surface. The signals were acquired at 30 kHz (Blackrock Microsystems) and filtered at 0.3–
300 Hz to acquire the LFP signal. To measure the preferred retinotopic locations for neural responses, 100%
contrast vertical white or black bars (width: 9°, duration: 0.1 s, inter-stimulus interval: 0.3 s) were presented
in random locations spanning the binocular and monocular visual field (~ − 60° to + 150° in azimuth) on grey
linearized LCD monitors (see specs in “Visual Stimuli for ISI"). LFP recordings were performed in anesthetized
(n = 5) and awake (n = 5) mice as detailed previously24. We observed no major differences in retinotopic corre-
spondence of ISI versus LFP across anesthetized (V1, RL) and awake (V1, RL, LM, PM, AL) recordings, so these
were combined (Fig. 4).
Analysis of retinotopic maps in V1 and HVA.
In order to create contour maps, azimuth and eleva-
tion retinotopic maps were rounded to the nearest 10°, with each contour section representing the cortical area
that responds to that visual stimulus spatial location. Data from all imaging sessions (i.e., all ISI sessions across
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
12
Vol:.(1234567890)www.nature.com/scientificreports/ multiple days) were then used to create randomly drawn subsamples of specific sizes from the total dataset. Resa-
mpling of data was done with or without replacement, with no obvious differences between methods (Fig. 3A).
These subsampled maps of increasing data length were then compared to a reference map computed using all
data collected from all ISI imaging sessions within a mouse. The similarity of subsampled and reference maps
was evaluated by determining the centroid of each contour section for the subsampled and reference maps, then
calculating the average Euclidean distance between the two (similar to prior approaches6,8). This enabled us to
study the error in estimating cortical retinotopy as a function of number of trials. Subsampled and reference VFS
maps were also created as above (Fig. S3), allowing us to plot the pixel intensity distribution within boundaries
of identified cortical areas (Fig. 3C; Fig. S4A). We calculated receiver operating characteristic (ROC) curves of
VFS pixel intensity using contour areas defined by the reference maps (“signal”) versus adjacent non-visual cor-
tical (“noise”) regions outside of areas localized in the reference map (see Fig. 3D and Fig. S2D) to identify the
number of samples needed to resolve V1 and HVAs.
Analysis of retinotopic signals in ISI versus electrophysiology. We used retinotopic ISI maps to
perform targeted craniotomies and neural recordings of specific retinotopic subregions within V1 and HVAs.
The sites of craniotomies and azimuth retinotopic maps were aligned to identify expected ISI retinotopic coor-
dinates and compare these to ground-truth electrophysiological measurements at these same sites (Fig. 4B).
LFP responses were separated into cortical layers based on the earliest visual response latency, which typically
corresponds to the input layer1,24. In V1, the channel with the lowest latency represents the middle of L4, which
is about 200 µm in thickness. L4 was defined as the average of all channels within ± 100 microns of this site.
The average of all channels above the upper boundary of L4 was termed superficial layers, whereas the aver-
age of all channels below the lower boundary of L4 was termed deep layers. Retinotopic preferences of the
maximum LFP response (averaged within the superficial and deep layers) were then compared to the preferred
locations predicted from the retinotopic ISI maps. Pearson correlations (r) were computed (MATLAB ‘corr’
function) between the ISI and LFP preferred locations to obtain r2 values. Additionally, the differences between
ISI (expected) and LFP (observed) retinotopic coordinates provided an error estimate that could be compared
across all experiments. We inspected if there was any dependence of expected versus observed coordinates upon
the luminance polarity of the visual stimulus (black versus white, 100% contrast). Since there were no obvious
differences, these are presented together (Fig. 4).
Single neuron selectivity analysis in HVAs. We performed awake extracellular recordings as described
above (“Neural recordings and analysis of retinotopy” section) across visual areas (V1, LM, AL, RL, AM, and
PM). We isolated single neuron action potentials and separated these into regular spiking (RS) putative excita-
tory and fast spiking (FS) putative inhibitory neurons as in our prior studies24,25,48. Drifting gratings (σ = 10°,
100% contrast) were positioned within 10° degrees of the receptive field as determined by the LFP response
online. Gratings varied in orientation and drift direction (0–360° at 45° intervals), spatial frequency (0.02, 0.04,
0.1, 0.16 cpd), and temporal frequency (0.5, 1, 2, 6 Hz). These parameters were chosen to match benchmark
studies of neural selectivity in mouse higher visual areas26,28, and only recordings with receptive fields in the
monocular visual fields were analyzed, again consistent with previous reports26. Single neuron responses were
analyzed by first determining preferred direction (direction with the highest evoked firing rate) and then meas-
uring the preferred spatial frequency and temporal frequency at the preferred direction (Fig. 4). Speed tuning
was calculated as the preferred temporal frequency divided by the preferred spatial frequency.
Data availability
All ISI system code is deposited at https:// github. com/ haide rlab/ ISI, and source data and analysis code to rep-
licate the main results will be publicly available at DOI (https:// doi. org/ 10. 6084/ m9. figsh are. 16200 711) upon
publication.
Received: 21 August 2021; Accepted: 18 January 2022
References
1. Niell, C. M. & Scanziani, M. How cortical circuits implement cortical computations: Mouse visual cortex as a model. Annu. Rev.
Neurosci. https:// doi. org/ 10. 1146/ annur ev- neuro- 102320- 085825 (2021).
2. Zhuang, J. et al. An extended retinotopic map of mouse cortex. Elife 6, e18372. https:// doi. org/ 10. 7554/ eLife. 18372 (2017).
3. Wang, Q. & Burkhalter, A. Area map of mouse visual cortex. J. Comp. Neurol. 502, 339–357. https:// doi. org/ 10. 1002/ cne. 21286
(2007).
4. Siegle, J. H. et al. Survey of spiking in the mouse visual system reveals functional hierarchy. Nature https:// doi. org/ 10. 1038/ s41586-
020- 03171-x (2021).
5. Juavinett, A. L., Nauhaus, I., Garrett, M. E., Zhuang, J. & Callaway, E. M. Automated identification of mouse visual areas with
intrinsic signal imaging. Nat. Protoc. 12, 32–43. https:// doi. org/ 10. 1038/ nprot. 2016. 158 (2017).
6. Waters, J. et al. Biological variation in the sizes, shapes and locations of visual cortical areas in the mouse. PLoS ONE 14, e0213924.
https:// doi. org/ 10. 1371/ journ al. pone. 02139 24 (2019).
7. Pisauro, M. A., Dhruv, N. T., Carandini, M. & Benucci, A. Fast hemodynamic responses in the visual cortex of the awake mouse.
J. Neurosci. 33, 18343–18351. https:// doi. org/ 10. 1523/ jneur osci. 2130- 13. 2013 (2013).
8. Garrett, M. E., Nauhaus, I., Marshel, J. H. & Callaway, E. M. Topography and areal organization of mouse visual cortex. J. Neurosci.
34, 12587–12600. https:// doi. org/ 10. 1523/ jneur osci. 1124- 14. 2014 (2014).
9. Kalatsky, V. A. & Stryker, M. P. New paradigm for optical imaging: Temporally encoded maps of intrinsic signal. Neuron 38,
529–545. https:// doi. org/ 10. 1016/ s0896- 6273(03) 00286-1 (2003).
10. Silasi, G., Xiao, D., Vanni, M. P., Chen, A. C. & Murphy, T. H. Intact skull chronic windows for mesoscopic wide-field imaging in
awake mice. J. Neurosci. Methods 267, 141–149. https:// doi. org/ 10. 1016/j. jneum eth. 2016. 04. 012 (2016).
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
13
Vol.:(0123456789)www.nature.com/scientificreports/ 11. Cramer, S. W. et al. Through the looking glass: A review of cranial window technology for optical access to the brain. J. Neurosci.
Methods 354, 109100. https:// doi. org/ 10. 1016/j. jneum eth. 2021. 109100 (2021).
12. Xu, H. T., Pan, F., Yang, G. & Gan, W. B. Choice of cranial window type for in vivo imaging affects dendritic spine turnover in the
cortex. Nat. Neurosci. 10, 549–551. https:// doi. org/ 10. 1038/ nn1883 (2007).
13. Kalmbach, A. S. & Waters, J. Brain surface temperature under a craniotomy. J. Neurophysiol. 108, 3138–3146. https:// doi. org/ 10.
1152/ jn. 00557. 2012 (2012).
14. Jouhanneau, J. S. & Poulet, J. F. A. Multiple two-photon targeted whole-cell patch-clamp recordings from monosynaptically con-
nected neurons in vivo. Front. Synaptic Neurosci. 11, 15. https:// doi. org/ 10. 3389/ fnsyn. 2019. 00015 (2019).
15. Kaneko, M., Stellwagen, D., Malenka, R. C. & Stryker, M. P. Tumor necrosis factor-alpha mediates one component of competitive,
experience-dependent plasticity in developing visual cortex. Neuron 58, 673–680. https:// doi. org/ 10. 1016/j. neuron. 2008. 04. 023
(2008).
16. Liu, J., Wang, P., Zhang, X., Zhang, W. & Gu, G. Effects of different concentration and duration time of isoflurane on acute and
long-term neurocognitive function of young adult C57BL/6 mouse. Int. J. Clin. Exp. Pathol. 7, 5828–5836 (2014).
17. Valentim, A. M., Alves, H. C., Olsson, I. A. & Antunes, L. M. The effects of depth of isoflurane anesthesia on the performance of
mice in a simple spatial learning task. J. Am. Assoc. Lab. Anim. Sci. 47, 16–19 (2008).
18. International Brain L. et al. Standardized and reproducible measurement of decision-making in mice. Elife 10, e63711. https:// doi.
org/ 10. 7554/ eLife. 63711 (2021).
19. Ruediger, S. & Scanziani, M. Learning speed and detection sensitivity controlled by distinct cortico-fugal neurons in visual cortex.
Elife 9, e59247. https:// doi. org/ 10. 7554/ eLife. 59247 (2020).
20. Zhao, Y. J. et al. Skull optical clearing window for in vivo imaging of the mouse cortex at synaptic resolution. Light Sci. Appl. 7,
17153. https:// doi. org/ 10. 1038/ lsa. 2017. 153 (2018).
21. Frostig, R. D., Lieke, E. E., Ts’o, D. Y. & Grinvald, A. Cortical functional architecture and local coupling between neuronal activity
and the microcirculation revealed by in vivo high-resolution optical imaging of intrinsic signals. Proc. Natl. Acad. Sci. U. S. A. 87,
6082–6086. https:// doi. org/ 10. 1073/ pnas. 87. 16. 6082 (1990).
22. Glasser, M. F. et al. Using temporal ICA to selectively remove global noise while preserving global signal in functional MRI data.
Neuroimage 181, 692–717. https:// doi. org/ 10. 1016/j. neuro image. 2018. 04. 076 (2018).
23. Jacus, M. O., Rahija, R. J., Davis, A. D., Throm, S. L. & Stewart, C. F. Observational evaluations of mice during cerebral microdialysis
for pediatric brain tumor research. J. Am. Assoc. Lab. Anim. Sci. 54, 304–310 (2015).
24. Del Rosario, J. et al. Diminished cortical excitation and elevated inhibition during perceptual impairments in a mouse model of
autism. Cereb. Cortex 31, 3462–3474. https:// doi. org/ 10. 1093/ cercor/ bhab0 25 (2021).
25. Speed, A., Del Rosario, J., Mikail, N. & Haider, B. Spatial attention enhances network, cellular and subthreshold responses in mouse
visual cortex. Nat. Commun. 11, 505. https:// doi. org/ 10. 1038/ s41467- 020- 14355-4 (2020).
26. Marshel, J. H., Garrett, M. E., Nauhaus, I. & Callaway, E. M. Functional specialization of seven mouse visual cortical areas. Neuron
72, 1040–1054. https:// doi. org/ 10. 1016/j. neuron. 2011. 12. 004 (2011).
27. Mesa, N., Waters, J. & de Vries, S. E. J. The effect of inclusion criteria on the functional properties reported in mouse visual cortex.
eNeuro 8, ENEURO.0188-20.2021. https:// doi. org/ 10. 1523/ ENEURO. 0188- 20. 2021 (2021).
28. Andermann, M. L., Kerlin, A. M., Roumis, D. K., Glickfeld, L. L. & Reid, R. C. Functional specialization of mouse higher visual
cortical areas. Neuron 72, 1025–1039. https:// doi. org/ 10. 1016/j. neuron. 2011. 11. 013 (2011).
29. Smith, I. T., Townsend, L. B., Huh, R., Zhu, H. & Smith, S. L. Stream-dependent development of higher visual cortical areas. Nat.
Neurosci. 20, 200–208. https:// doi. org/ 10. 1038/ nn. 4469 (2017).
30. Beltramo, R. & Scanziani, M. A collicular visual cortex: Neocortical space for an ancient midbrain visual structure. Science 363,
64–69. https:// doi. org/ 10. 1126/ scien ce. aau70 52 (2019).
31. Ranson, A., Cheetham, C. E., Fox, K. & Sengpiel, F. Homeostatic plasticity mechanisms are required for juvenile, but not adult,
ocular dominance plasticity. Proc. Natl. Acad. Sci. U. S. A. 109, 1311–1316. https:// doi. org/ 10. 1073/ pnas. 11122 04109 (2012).
32. Hofer, S. B., Mrsic-Flogel, T. D., Bonhoeffer, T. & Hubener, M. Prior experience enhances plasticity in adult visual cortex. Nat.
Neurosci. 9, 127–132. https:// doi. org/ 10. 1038/ nn1610 (2006).
33. Van den Bergh, G., Zhang, B., Arckens, L. & Chino, Y. M. Receptive-field properties of V1 and V2 neurons in mice and macaque
monkeys. J. Comp. Neurol. 518, 2051–2070. https:// doi. org/ 10. 1002/ cne. 22321 (2010).
34. Niell, C. M. & Stryker, M. P. Highly selective receptive fields in mouse visual cortex. J. Neurosci. 28, 7520–7536. https:// doi. org/
10. 1523/ jneur osci. 0623- 08. 2008 (2008).
35. Knutsen, P. M., Mateo, C. & Kleinfeld, D. Precision mapping of the vibrissa representation within murine primary somatosensory
cortex. Philos. Trans. R. Soc. Lond. B Biol. Sci. 371, 20150351. https:// doi. org/ 10. 1098/ rstb. 2015. 0351 (2016).
36. Tsukano, H. et al. Quantitative map of multiple auditory cortical regions with a stereotaxic fine-scale atlas of the mouse brain. Sci.
Rep. 6, 22315. https:// doi. org/ 10. 1038/ srep2 2315 (2016).
37. Sirotin, Y. B. & Das, A. Anticipatory haemodynamic signals in sensory cortex not predicted by local neuronal activity. Nature 457,
475–479. https:// doi. org/ 10. 1038/ natur e07664 (2009).
38. Rasmussen, R., Matsumoto, A., Dahlstrup Sietam, M. & Yonehara, K. A segregated cortical stream for retinal direction selectivity.
Nat. Commun. 11, 831. https:// doi. org/ 10. 1038/ s41467- 020- 14643-z (2020).
39. Siegle, J. H. et al. A survey of spiking activity reveals a functional hierarchy of mouse corticothalamic visual areas. bioRxiv, 805010.
https:// doi. org/ 10. 1101/ 805010 (2019).
40. Vanzetta, I. & Grinvald, A. Increased cortical oxidative metabolism due to sensory stimulation: Implications for functional brain
imaging. Science 286, 1555–1558. https:// doi. org/ 10. 1126/ scien ce. 286. 5444. 1555 (1999).
41. Toga, A. W. & Mazziotta, J. C. Brain Mapping: The Methods 2nd edn. (Academic Press, 2002).
42. Dervieux, E., Bodinier, Q., Uhring, W. & Theron, M. Measuring hemoglobin spectra: Searching for carbamino-hemoglobin. J.
Biomed. Opt. 25, 105001. https:// doi. org/ 10. 1117/1. JBO. 25. 10. 105001 (2020).
43. Polimeni, J. R., Granquist-Fraser, D., Wood, R. J. & Schwartz, E. L. Physical limits to spatial resolution of optical recording: Clari-
fying the spatial structure of cortical hypercolumns. Proc. Natl. Acad. Sci. U. S. A. 102, 4158–4163. https:// doi. org/ 10. 1073/ pnas.
05002 91102 (2005).
44. Lu, H. D., Chen, G., Cai, J. & Roe, A. W. Intrinsic signal optical imaging of visual brain activity: Tracking of fast cortical dynamics.
Neuroimage 148, 160–168. https:// doi. org/ 10. 1016/j. neuro image. 2017. 01. 006 (2017).
45. Pouratian, N., Sheth, S. A., Martin, N. A. & Toga, A. W. Shedding light on brain mapping: Advances in human optical imaging.
Trends Neurosci. 26, 277–282. https:// doi. org/ 10. 1016/ s0166- 2236(03) 00070-5 (2003).
46. Turley, J. A., Zalewska, K., Nilsson, M., Walker, F. R. & Johnson, S. J. An analysis of signal processing algorithm performance for
cortical intrinsic optical signal imaging and strategies for algorithm selection. Sci. Rep. 7, 7198. https:// doi. org/ 10. 1038/ s41598-
017- 06864-y (2017).
47. Williams, B., Speed, A. & Haider, B. A novel device for real-time measurement and manipulation of licking behavior in head-fixed
mice. J. Neurophysiol. 120, 2975–2987. https:// doi. org/ 10. 1152/ jn. 00500. 2018 (2018).
48. Speed, A., Del Rosario, J., Burgess, C. P. & Haider, B. Cortical state fluctuations across layers of V1 during visual spatial perception.
Cell Rep. 26, 2868–2874.e2863. https:// doi. org/ 10. 1016/j. celrep. 2019. 02. 045 (2019).
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
14
Vol:.(1234567890)www.nature.com/scientificreports/ Acknowledgements
We thank Anderson Speed for technical support, members of the Haider lab and Jordan Hamm for feedback,
and Ruben Uribe and Chris Howard (Physimetrics, Inc.) for assistance in machine development. This work was
supported by the Whitehall Foundation, the Alfred P. Sloan Foundation, National Institutes of Health Neurologi-
cal Disorders and Stroke (NS107968), National Institutes of Health BRAIN Initiative (NS109978), and a grant
from the Simons Foundation (SFARI 600343).
Author contributions
A.N., A.C.Y. and B.W. assembled original hardware components and developed software; A.N. and S.W. opti-
mized hardware and software; A.N. wrote analysis code and analyzed all experiments; A.N. and D.S. optimized
signal processing; A.N., J.D.R. and T.L.A. carried out ISI imaging; J.D.R. and T.L.A. performed silicon probe
experiments; A.N. and B.H. wrote the manuscript with feedback from all authors.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary Information The online version contains supplementary material available at https:// doi. org/
10. 1038/ s41598- 022- 05932-2.
Correspondence and requests for materials should be addressed to B.H.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
© The Author(s) 2022
Scientific Reports | (2022) 12:2063 |
https://doi.org/10.1038/s41598-022-05932-2
15
Vol.:(0123456789)www.nature.com/scientificreports/
| null |
10.1038_s41598-022-07802-3.pdf
| null | null |
OPEN
Predicting individual perceptual
scent impression from imbalanced
dataset using mass spectrum
of odorant molecules
Tanoy Debnath1,3* & Takamichi Nakamoto1,2,3*
Predicting odor impression is considered an important step towards measuring the quality of scent
in the food, perfume, and cosmetic industries. In odor impression identification and classification,
the main target is to predict scent impression while identifying non-target odor impressions are
less significant. However, the effectiveness of predictive models depends on the quality of data
distribution. Since it is difficult to collect large scale sensory data to create an evenly distributed
positive (target odor) and negative (non-target odor) samples, a method is necessary to predict
the individual characteristics of scent according to the number of positive samples. Moreover, it is
required to predict large number of individual odor impressions from such kind of imbalanced dataset.
In this study, we used mass spectrum of flavor molecules and their corresponding odor impressions
which have a very disproportioned ratio of positive and negative samples. Thus, we used One-class
Classification Support Vector Machine (OCSVM) and Cost-Sensitive MLP (CSMLP) to precisely classify
target scent impression. Our experimental results show satisfactory performance in terms of AUC ROC
to detect the olfactory impressions of 89 odor descriptors from the mass spectra of flavor molecules.
Odor impression prediction is an active area of research that is important for evaluating product quality1.
Human sense of smell is a complex process than visual and auditory perception2 as one odorant molecule can
be described with more than two odor descriptors which is influenced by a person’s cultural background3 and
experience. Odor is generally a complex mixture of many mono-molecular molecules that attaches and activates
the olfactory receptors (ORs) of an Olfactory Sensory Neuron (OSN)4,5. Odors then stimulate our nasal olfactory
neurons and the olfactory bulbs, thus converting these olfactory signals into odor impressions, such as ‘fruity’,
‘citrus’, ‘spicy’ etc.
Machine learning techniques have made significant progress in predicting odor impressions using molecular
structure parameters6, activation information of the olfactory bulb7. Sanchez-Lengeling et. al8 used graph neural
network to predict odor descriptors using a molecular graph structure. Another study had shown a report for
predicting natural language descriptions of mono-molecular odorants using odor wheel9. Recently, Deepak et al.
predicted smell impressions of ‘sweet’ and ‘musky’ using molecular structure parameters10. Odor, however, is
usually a complex mixture, so it is better to establish a general method where we can use the same chemical fea-
tures as the inputs of the machine learning model regardless of mixture or single molecule. The mass spectrum,
an analytical technique that ionizes chemicals and sorts the ions based on their mass-to-charge ratio (m/z), can
be used as input to the neural network model as it can be collected for both mono-monomolecular chemicals
or chemical mixtures.
We used small mass spectrum dataset of mono-molecular chemicals, including continuous11 sensory data
from the Dravnieks12 to predict the odor impressions. Then, relatively large mass spectrum dataset was used
with binary form of odor descriptors from Sigma-Aldrich catalog13, which appears mutually exclusive, to predict
the odor character of chemical using the natural language processing technique14. However, these studies9,14–16
focused on predicting odorant impressions by clustering similar smell impressions. Therefore, individual percep-
tion of odors is necessary for odor molecules to describe them.
Information and Communications Engineering, Tokyo
1Department of
Institute of Technology, Tokyo,
Japan. 2Laboratory for Future Interdisciplinary Research in Science and Technology, Tokyo Institute of
Technology, Tokyo, Japan. 3These authors contributed equally: Tanoy Debnath and Takamichi Nakamoto *email:
[email protected]; [email protected]
Scientific Reports | (2022) 12:3778
| https://doi.org/10.1038/s41598-022-07802-3
1
Vol.:(0123456789)www.nature.com/scientificreports Moreover, the problem of imbalanced ratio of negative to positive samples, which frequently appears in odor-
impression prediction, deteriorates the prediction accuracy. We tried to solve this problem using oversampling
technique16,17 where a part of the data was artificially generated. Although it improved the prediction accuracy to
some extent, the improvement was limited because we could use only restricted artificial data. Creating artificial
samples can duplicate samples from the minority class and this increases the likelihood of overfitting especially
for high oversampling rates when class skew was severe19. The task of odor prediction is often imbalanced due
to the presence of over and under-represented odor descriptors. So, our goal is to establish how to make a clas-
sification using the original small number of positive samples. Here the positive sample means the target odor
with specified odor descriptors (e.g., fruity, pine, etc.) which we would like to predict from non-target (negative)
samples. Non-target sample does not have specified odor descriptor. In this study, we will use the positive/nega-
tive and target/non-target words interchangeably.
Several odor impressions (e.g., fruity, sweet) appear very often and make it easier for human participants to
describe with these common words. Thus, the number of positive appearances of uncommon odor impressions
(‘hazelnut’, ‘peach’) is small. If the number of target-odors is much smaller than non-target odor samples, machine
learning model only learns non-target odor samples well which affects the overall predictive performance. One
of the methods to handle these large numbers of negative samples with small target odors is to use one class clas-
sification where we do not need equal proportions of positive and negative samples18,19. One class classification,
a well-known method that has been applied to many research themes such as outlier detection etc. However,
it has never been employed in odor prediction task. We chose OCSVM because only small number of positive
samples was available here. Another possibility is to give weighted cost the minority samples during the training
process with the help of neural network models trained with weighted loss functions.
Therefore, in this study we would like to establish a method for predicting individual odor perception from
the mass spectrum of odorant molecules using highly imbalanced odor descriptors datasets without creating
artificial samples. In odor prediction task, the main goal is to predict positive samples (target odor like ‘fruity’,
‘pine’) regardless of its occurrence frequency. However, the same algorithm might not be useful when the ratio
of negative samples to positive ones varied. In this work, we used two separate algorithms (one class classifica-
tion Support Vector Machine20,21and Cost Sensitive MLP22) to predict 89 odor impressions where each odor
impression is highly imbalanced and has different occurrence frequency. We experimentally divided these odor
descriptors into three categories, ‘large’, ‘middle’ & ‘small’, depending on the number of positive samples. We
investigated the experimental results to select the correct algorithm as a function of the number of positive (target
odor) and negative (non-target odor) samples.
The main contribution of this paper is the use of one-class classification Support Vector Machine (OCSVM)
and cost sensitive multilayer perceptron (CSMLP) to evaluate the prediction performance from a large number
of negative samples depending on odor descriptor occurrence in the dataset. Moreover, we established a rule for
selecting the correct algorithm based on the ratio of negative to positive samples. We report that it can achieve
better sensitivity or in other words, obtain better performance in predicting target odor with small category odor
descriptors. To the best of our knowledge, our proposed method is the first to establish odor prediction system
depending on the odor descriptor occurrences.
Materials and methods
Flavor database descriptions. Leffingwell23 Flavor Database (n = 2345) was used for this study where
flavor molecules were labeled by one or more odor descriptors. The database contains chemical names with CAS
number and their corresponding odor descriptions which are in free-form text. We obtained the mass spectrum
of these flavor molecules from the Chemistry Webbook provided by National Institute of Standards and Tech-
nology (NIST)24 using the corresponding CAS number. Although original flavor database has more data, the
verbal data without mass spectrum is eliminated. As one molecule is described with multiple odor descriptors,
it creates a multi-label prediction problem. We try to solve it by decomposing the problem in several binary
classification models. we got 89 odor descriptors including one odorless descriptor. We listed the name of odor
descriptors and their frequencies of appearances among 2345 samples in Fig. 1.
Data processing. Original mass spectrum of NIST has more than 300 dimensions. Intensity of 50–262 m/z
(mass-to-charge ratio) was used because the intensity of m/z below 50 is mainly derived from odorless molecules
like oxygen and intensity of m/z higher than 262 originates from molecules with low volatility. Thus, the data
matrix is expressed by rows of 2345 samples and columns of 212 intensities. This data matrix of mass spectra was
normalized in the range of 0 and 1 after dividing by the maximum value in the same mass spectrum. Principal
component analysis (PCA) was used to reduce the dimensionality from these 212 intensities of mass spectrum.
All the 2345 odorant molecules are listed with their CAS numbers in Supplementary file 1.
Predictive model. One class classifier and binary SVM classifier. We used OCSVM to predict the target
odor from mass spectrum of the imbalanced set of flavor molecules. OCSVM learns the task of making a deci-
sion boundary to classify new data as similar or different from the training set. The classifier tries to detect a sin-
gle class and rejects the others. At first, the dataset was divided into majority (samples without target attribute)
and minority class (samples with target attribute). Next, we created (train, test) tuples of majority samples for
five-fold cross-validation. We split the majority class as we only have these majority samples during the train-
ing and minority class was added with the test split during the validation time. GridSearchCV (using Python
libraries) was used to optimize the hyperparameters of OCSVM. We present the results for the model with the
mean AUROC on each testing fold. We run the algorithm 89 times as we considered this analysis as a binary
classification task of 89 odor descriptors.
Scientific Reports | (2022) 12:3778 |
https://doi.org/10.1038/s41598-022-07802-3
2
Vol:.(1234567890)www.nature.com/scientificreports/ Figure 1. 89 Odor descriptors with corresponding positive samples among 2345 odorant molecules.
We compare the performance of OCSVM with the binary SVM classifier using 5-fold stratified cross-vali-
dation where the only exception is that both positive and negative samples were used to train the binary SVM
classifier.
Traditional & cost‑sensitive multilayer perceptron. Two types of multilayer perceptrons were used for this study.
One is traditional multilayer perceptron (MLP) and other is cost- sensitive multilayer perceptron (CSMLP).
Traditional MLP trained by backpropagation of error algorithm considered misclassification costs (false nega-
tive and false positive) are the same, so a false negative is worse or more expensive than a false positive25,26. In
the cost sensitive MLP, we assigned higher weight for the minority class and at the same time reduced the weight
for the majority class. We determined the class-weights for the majority and minority classes in such a way that
the model pays more attention to the observations from minority class. We scaled the weights of both classes so
that the sum of the weights of all examples keeps the same, in other words, we assigned the class weights which
is inversely proportional to their respective frequencies27.
5-Fold stratified cross-validation was used for evaluating both methods. The model was trained with 16 hid-
den neurons (empirically) in the 2nd layer, Relu as the activation function and sigmoid at the final layer. Binary
cross entropy was used as a loss function and model was trained with Adam28 (keras-optimizer) with learning
rate 0.001. Drop out was used as a regularizer for preventing overfitting. The batch size was 64 and there was
total 109 epochs. The difference between the two models was whether we determined the weight of class for was
equal or not. The modified binary cross-entropy loss function that was used for CSMLP in Eq. (1).
weighted Binary crossentropy = −
1
N
N
(cid:31)
i=1
where wj = total samples / (n-classes * n-samplesj).
[w0(cid:30)yi ∗ log(p(yi)) + w1((1 − yi(cid:29)∗log(cid:30)1 − p(cid:30)yi(cid:29)(cid:29)))]
(1)
Here, wj is the weight for each class (j = 1: positive sample; j = 0: negative sample); total samples are the total
number of samples or rows in the dataset; n-classes are the total number classes (in our case 2 class) in the target;
n-samplesj is the total number of samples of the respective class.
Results
Principal component analysis. PCA was used to check the distribution of positive and negative samples
for 89 odor descriptors as well as reducing the dimension to 25 optimally from its 212 intensities. We used 25
PCs for each odor descriptor in this study which capture more than 60% of the total variation. Increasing the
number of PCs to more than 25 had no effect on the overall performance of the model. Figure 2 (top row)
depicted scatter plot of the first two principal components (PCs) for fruity, banana & spearmint.
There was no clear separation between fruity and non-fruity samples as shown in 2D PCA plot Fig. 2A which
indicates that samples are overlapping with one another and non-linear in structure also. Thus, the problem for
discriminating its boundary gets complicated not only because of the data distribution but also for the skewness
towards negative samples. For these types of samples where the data distribution overlaps, we can consider it as
a large category odor descriptor as these odor descriptors appear most of the time among 2345 samples. There
are totally 3 such kind of odor descriptors (fruity, sweet & green) in this study. The ratio of non-fruity to fruity,
non-sweet to sweet and non-green to green are 1.64, 2.60 & 3.12 respectively in this experiment where the posi-
tive samples were around half of the negative ones. Such odor impressions can be predicted using MLP model
by properly splitting the positive and negative samples during training period and using the suitable evaluation
Scientific Reports | (2022) 12:3778 |
https://doi.org/10.1038/s41598-022-07802-3
3
Vol.:(0123456789)www.nature.com/scientificreports/ Figure 2. Top row, Principal component analysis in 2D space for each odor category. (A) Fruity (B) Banana &
(C) Spearmint odor samples for large, middle & small category respectively. All are different figure although the
location of each data point is the same. Red circle is different. Bottom row, (D), (E), (F) depicted the predicted
vs ground truth odor detection for fruity, banana & spearmint using One class Support Vector Machine where
green dots indicate the non-target odor samples for each example. Please see the supplementary file 2 for each
odor descriptor detection using OCSVM for every category.
metric for describing the result of odor prediction of imbalanced dataset. We chose area under ROC curve as
that curve keeps AUC high by scoring most of negatives very lower.
There are 34 more odor descriptors where the ratio of negative to positive samples was between 5 ~ 44, thus
the problem to predict the target odor becomes difficult for increasing the number of negative samples compared
to the large category samples. Figure 2B depicts the banana and non-banana odor samples in 2D space where
target odor (banana) overlaps with non-banana samples. We can consider these kind of odor impressions as a
middle-class odor descriptor.
However, the prediction gets more complicated when the ratio of negative to positive samples were too
high (ratio of non-spearmint to spearmint is 166.5), for example spearmint (14 positive samples) was depicted
as Fig. 2C. Such problems are difficult to solve due to the high skewness towards negative class (2331 negative
samples). One hypothesis is to solve this problem by considering the minority samples as an outlier because most
target odors are outside the dense boundary of negative samples. In this analysis, we found 51 odor descriptors
where the number of positive samples was below 50 and, in all cases, these target odors were almost outside the
decision boundary of non-odor samples. The ratio of negative to positive samples is between 46 ~ 334 for these
51 odor descriptors. We have renamed it as a small category positive sample.
One class SVM & binary SVM classifier. The most important part of One class SVM was optimizing the
hyperparameters kernel, gamma (σ), nu(ν). The number of support vectors decreases with σ increasing and the
decision boundary becomes unclear. The parameter ν also affects the shape of decision boundary; as ν increases,
the number of support vectors increases, and so does the number of incorrectly classified training samples
increase. It is usually set to a small value to ensure a small misclassification rate on the training phase. During
the optimization, we selected the range ν from 0.01 to 0.3, set the σ as scale or auto and used radial basis function
kernel (RBF) as a kernel for optimizing these hyperparameters using scikit-learn One class SVM29. Surprisingly,
the hyper-parameters optimized (ν = 0.01 and σ = auto) for one class SVM of 89 odor descriptors were the same.
One hypothesis for such results is the training dataset of negative (non- target odors) samples which is almost
the same for all classifications for selecting the boundaries of these samples because their original distribution
was the same. We will fit a distribution or decision boundary for the negative samples and then use the trained
model to label the validation set to see if the given sample is positive or negative.
The data distribution of large category odor descriptors (fruity, green, sweet) was complex and the ratio of
negative to positive samples was not so high. The results of the fivefold cross validation (Mean AUC ROC) of these
three descriptors are shown in Fig. 3A (purple). The Mean AUC ROC for these three odor descriptors is very low
(below 0.60). For example, the decision boundary made by the non-fruity samples during the training completely
overlaps with the validation dataset that included both the non-fruity and fruity samples shown in Fig. 2D. These
Scientific Reports | (2022) 12:3778 |
https://doi.org/10.1038/s41598-022-07802-3
4
Vol:.(1234567890)www.nature.com/scientificreports/ Figure 3. 5-fold cross validation results to show AUC for each odor category; (A) large (B) Middle & (C) Small
category odor samples. Blue, Red, Green & purple line depicted the result for Traditional MLP, cost sensitive
MLP, Binary SVM and One class SVM respectively. We present the results for the model with the mean AUROC
on each testing fold for five-fold cross-validation method.
results indicate that it is not feasible to use one class classification when the ratio of negative to positive samples
is small (between 1.5 ~ 4 approximately).
Odor classification becomes more difficult when the number of target class is too small, meaning the ratio of
negative to positive samples is very high. Each of these 51 odors descriptors have less than 50 positive samples
out of 2345. Thus, it is a problem to discriminate such positive samples (specific odor descriptor such as spear-
mint) from negative samples (non-spearmint) as shown in Fig. 2C. For example, 11 spearmint odors were clearly
predicted, and other ground truth samples could not be accurately identified as shown in Fig. 2F. There were
some errors with red circle that only indicated false detection of odors but note that such false detection occurs
at the border of the boundary of negative (non-spearmint) class. Figure 3C (purple) is the result (Mean AUC
ROC) of a classification of 51 odor descriptors. Traditional binary classification is not a suitable method to use in
such cases because of the class skewness, thus the model only learns the negative class and does not generalize
well for the unseen positive class. A good AUC ROC score (more than 0.90) has been obtained for all of these odor
descriptors using one class support vector machine except alcoholic, coffee (below 0.80). So, it is feasible to use
OCSVM for identifying odor when its attribute appears infrequently.
When the number of positive classes is more than 50 in other words when the ratio of negative to positive
samples increases compared to the small category samples, the results were not good as the previous small
category samples. The decision boundary of positive and negative samples was difficult to discriminate due to
overlapping between positive and negative samples. We found 35 odor descriptors with positive samples between
51 and 350. For example, banana samples (68 positive samples) were not accurately predicted because they were
overlapping with non-banana (2277) samples as illustrated in Fig. 2E. Banana samples that were located within
the decision boundary of the non-banana samples were very difficult to distinguish from the non-banana sample
and this was true for the other 34 middle category odor descriptors. We got comparatively low AUC ROC score
(below 0.80) for apple, banana, burnt, earthy, ethereal, fermented, floral, fresh, garlic, herbaceous, meaty, nutty,
onion, pineapple, pungent, roast, spicy, sulfurous, tropical, winey as shown in Fig. 3B (purple).
We used SVM to compare with OCSVM. We found a better performance (based on AUROC shown in green
line of Fig. 3A) of SVM than OCSVM for the large category. Out of 35 odor descriptors in middle category,
only 11(banana, burnt, ethereal, floral, meaty, nutty, onion, pungent, roast, sulfurous, winey) have better per-
formance using SVM shown in Fig. 3B-green (see supplementary file 6-SVM column). However, none of the
odor descriptors show good performance (Fig. 3C) for small category dataset. This is due to the small number
of target samples. So, typical binary classification is not a good choice when the occurrence of the target (posi-
tive) samples is very small (below 50 in our case). We reported the optimal hyperparameters (C, and gamma)
for binary SVM in supplementary file 3.
So, it can be said that we can use OCSVM when we have a very small number of target odors. In this type of
small category samples where the positive samples are outside the decision boundary of the negative samples,
it will be better to use OCSVM to accurately predict the target odor. On the other hand, it is not appropriate to
use OCSVM for middle and large category data.
Traditional and cost-sensitive multilayer perceptron. Since one class classification failed to detect
the target odor samples when the number of positive samples was higher than middle and small category odor
samples, we used cost-sensitive multilayer perceptron & compared them to traditional MLP.
Scientific Reports | (2022) 12:3778 |
https://doi.org/10.1038/s41598-022-07802-3
5
Vol.:(0123456789)www.nature.com/scientificreports/ Large Category Odor Descriptors Traditional MLP
Name of
OD
Fruity
Green
Sweet
TP
30
0
11
TN FP
FN ROC AUC Recall
Target smell during testing
119
182
154
24
0
5
62
53
65
0.655
0.585
0.681
0.326
0.000
0.145
92
53
76
Large Category Odor Descriptors Cost Sensitive MLP
Name of OD
TP
TN FP
FN ROC AUC Recall
Target smell during testing Weight for class 0 Weight for class 1
Fruity
Green
Sweet
76
28
46
61
119
103
82
63
56
16
25
30
0.658
0.591
0.676
0.826
0.528
0.605
92
53
76
0.8
0.66
0.69
1.32
2.06
1.8
Table 1. Traditional MLP (top) & cost sensitive MLP (bottom) result for predicting target odor (large
category) for the testing set. TP = True Positive; TN = True Negative; FP = False Positive; FN = False Negative.
We got a satisfactory Mean AUC ROC (in both cases around 0.80 which is better than OCSVM) as shown in
Fig. 3A (Blue & Red line for Traditional and Cost-sensitive MLP respectively) for these three descriptors using
a 5-fold stratified cross-validation of these two methods. Mean AUC ROC score has been slightly increased for
these high category odor descriptors using Cost-sensitive MLP. Compared to the results of traditional (blue) &
cost-sensitive (red) MLP shown in Fig. 3C, OCSVM (red) has shown better performance in predicting target
odors from the large number of negative samples, in other words, when the number of positive samples was very
small (< 50). OCSVM has reported the best prediction performance (> 0.90) for almost all odor descriptors. We
got a relatively good AUC ROC score (around 0.80) using CSMLP compared to OCSVM for each middle category
odor descriptors except berry, earthy and musty, as shown in Fig. 3B. However, these three odor descriptors had
lower prediction performance using traditional MLP, which was improved slightly using cost sensitive MLP.
However, this experiment will be clearer if we do the experiment using train-validation-test split method.
Considering the low numerosity of data, we experimented this analysis by dividing the dataset into training,
validation, and test sets. The model was trained and validated with 1899 & 211 samples respectively. Although
initially we set the epoch 109, we stopped model training at the best validation error instead of a fixed number
of epochs. The model was then tested on the 235 samples that included both target and non-target samples.
Similar sets of training/validation/testing samples were used in both MLP and CSMLP cases. We have shown
the results of large category odor descriptors in Table 1. We have provided supplementary file 4 for the middle
and small category datasets.
Table 1 shows the comparison between traditional (top) and cost sensitive MLP (bottom) to predict more
true positive samples from the test sets. For example, there were total 92 fruity samples in testing set. Traditional
MLP model identified only 30 true positive (fruity) samples, thus recall and the area under ROC curve was not
so high. Cost sensitive MLP detected 76 true positive samples, in other words, class weights increased recall
because the model found more true positives samples, thus decreasing the false negatives.
Statistical analysis for OCSVM on small category dataset. We have further analyzed on the small
category of odor descriptors because our claim is: ‘OCSVM is more precise for target odors with low occurrence
of odor descriptors (51 odor descriptors used in our case)’. We have analyzed in reverse process. (1) First, we cre-
ated (train, test) from positive samples for the five-fold cross validation; (2) While four-fold of positive samples
were used for training, the remaining fold and negative samples were used together for testing each time. Thus,
we got 51 pairs (first method where we used negative samples for OCSVM training and another one where we
used positive samples for training) [see supplementary file 5].
A paired samples t-test (significance level was set at an alpha of P < 0.05) was conducted to compare the dif-
ference of training with positive samples and training with negative samples. There was a significant difference
in the scores for training with positive samples (M = 0.607, Var = 0.0058) and training with negative samples
(M = 0.970, Var = 0.0026) conditions; t (50) = − 23.22, P = 9.65 × 10−29 . These results suggest a statistically sig-
nificant difference between OCSVM performance and odor descriptor numerosity (explained with training
samples) without considering the data distribution.
Discussion. Finally, based on the above experiment we have made a decision for choosing the appropriate
algorithm to predict the odor impression as a function of the number of positive samples as depicted in Fig.4. We
found a large number of small category odor descriptors (no. of positive samples between 7 and 50) where the
ratio of negative to positive was between 46 ~ 334 using the flavor database. In odor prediction tasks, the small
number of target odors affects the overall performance of the classification. As the number of target odors is very
small, one class classification is a better approach for such kind of situation. We can treat these small number
of positive samples as an outlier, so it will be easier to predict the odor impression with these limited number
of positive samples. On the other hand, 35 odor descriptors have been found in this study where the number
of positive samples were between 52 to 346. Ratio of negative to positive samples were between 5 ~ 44. Experi-
mental results showed that CSMLP gave better result to detect the true positive samples for most of the odor
descriptors in this category compared to the one class SVM. However, when the ratio of negative to positive sam-
Scientific Reports | (2022) 12:3778 |
https://doi.org/10.1038/s41598-022-07802-3
6
Vol:.(1234567890)www.nature.com/scientificreports/ Figure 4. Horizontal axis shows the range of positive samples & vertical axis is the no. of odor descriptors of
each category. Ratio of negative to positive samples are also shown here for each category (inside the box).
ples was between 1.5~4, in other words, number of positive samples was half of the negative samples, CSMLP
showed better performance than OCSVM. Most recent work8 used Leffingwell database to predict odor percep-
tion using Graph neural network. We reported comparisons in supplementary file 6 for each odor descriptor that
are common both in our study and in the paper8 that used molecular graph structures as inputs for the Graph
neural network. Although this is completely different from our case because we used the mass spectrum as an
input, we can compare OCSVM & CSMLP with GNN8, traditional MLP and normal SVM in terms of predicted
performance of each odor descriptor. We present the results of the model with the mean AUROC in each test
fold for five-fold cross-validation method. We noticed that 51 (43 & 8 from small & middle category respec-
tively) odor descriptors (marked in red color in supplementary File 6) showed better AUROC performance in
our approach. Most of the odor descriptors are from small category where number of positive samples is very
small. Based on the supplementary file 6, we did a paired samples t-test (significance level was set at an alpha
of P < 0.05) on small category odor descriptor’s result (47 odor descriptors are common in GNN result and our
result (OCSVM). There was a significant difference in results between GNN (M = 0.892, Var = 0.004) & OCSVM
(M = 0.968, Var = 0.003); t (46) = − 5.97429, P = 1.58 × 10−7 . Although GNN is slightly better than other methods
in case of large and middle categories, OCSVM is much better than GNN in small category.
Conclusion
Machine learning approaches have been used to predict the smell impression; however, previous studies did not
use odor descriptors themselves but used odor descriptor groups to describe the scent. Since there are very few
target samples for most of the odor descriptors, selecting an appropriate computational method that can address
this limitation for the odor prediction task can be considered as the solution with most reliability.
In this work, we propose OCSVM as a one-class classification method, and Cost-sensitive MLP to classify
and predict target odors from a large number of negative (non-odor) samples using mass spectrum of odorant
molecules. The main goal of one-class classifier is to separate positive samples from others, we use it to find target
odors (positive set) that has the similar objective. In this specific problem, the target is to detect odors while
identifying non-target samples are of less or no significant.
We have divided the odor descriptors into three categories according to their presence in the odorant mol-
ecules so that the over-represented and under-represented odor descriptors are clearly understood. Since such
under-represented (peach, pine, etc.) odor impressions are large (in our case 51), we should keep good accuracy
for under-represented odor descriptors.. GNN model8 can perform well in large (only 3%) and medium (39%)
category descriptors, but we have still 57% small category descriptors that we need to predict. Experimental
results suggest that OCSVM is suitable for use when the number of such odor descriptors is very small (less
than 50 in our experiment). We also found a statistically significant difference between OCSVM performance
and odor descriptor numerosity (explained with training samples) without considering the data distribution
and result between previous study8 & OCSVM. So, our proposed method could be a way to predict odors for
small category odor descriptors.
Moreover, traditional, and Cost-sensitive MLP was used for comparing the results with one class classifica-
tion. Empirical results showed that cost sensitive MLP had better predictive performance than traditional one.
Scientific Reports | (2022) 12:3778 |
https://doi.org/10.1038/s41598-022-07802-3
7
Vol.:(0123456789)www.nature.com/scientificreports/ It can be used for all the cases (large, middle & small). However, when the number of odor descriptors are very
small, OCSVM is a better predictive model. So, the numerosity of a given odor descriptor is sufficient to select
the method to be used (CSMLP or OCSVM).
The major limitation in our study is that our current study is limited to only Leffingwell database only. It
can be extended to another available flavor dataset in the future. In conclusion, our study provided an idea
that OCSVM and Cost-sensitive MLP could be useful for predicting scent impression, with a limited number
of target samples and without generating the artificial observations for balancing the number of positive and
negative samples.
Received: 29 June 2021; Accepted: 22 February 2022
References
1. Zarzo, M. & Stanton, D. T. Understanding the underlying dimensions in perfumers’ odor perception space as a basis for developing
meaningful odor maps. Atten. Percept. Psychophys. 71, 225–247. https:// doi. org/ 10. 3758/ APP. 71.2. 225 (2009).
2. Haddad, R. et al. A metric for odorant comparison. Nat. Methods 5, 425–429. https:// doi. org/ 10. 1038/ nmeth. 1197 (2008).
3. Chrea, C. et al. Culture and odor categorization: agreement between cultures depends upon the odors. Food Qual. Prefer. 15,
669–679. https:// doi. org/ 10. 1016/j. foodq ual. 2003. 10. 005 (2004).
4. Buck, L. & Axel, R. A novel multigene family may encode odorant receptors: a molecular basis for odor recognition. Cell 65,
175–187. https:// doi. org/ 10. 1016/ 0092- 8674(91) 90418-x (1991).
5. Nakamoto, T. Essentials of Machine Olfaction and Taste Vol. 1 (Wiley, New York, 2016).
6. Keller, A. et al. Predicting human olfactory perception from chemical features of odor molecules. Science 355, 820–826. https://
doi. org/ 10. 1126/ scien ce. aal20 14 (2017).
7. Shang, L., Liu, C., Tomiura, Y. & Hayashi, K. Odorant clustering based on molecular parameter-feature extraction and imaging
analysis of olfactory bulb odor maps. Sens. Actuators B Chem. 255, 508–518. https:// doi. org/ 10. 1016/j. snb. 2017. 08. 024 (2018).
8. Sanchez-Lengeling, B. et al. Machine Learning for Scent: Learning Generalizable Perceptual Representations of Small Molecules.
Preprint at https:// arxiv. org/ abs/ 1910. 10685 (2019).
9. Gutiérrez, E. D., Dhurandhar, A., Keller, A., Meyer, P. & Cecchi, G. A. Predicting natural language descriptions of mono-molecular
odorants. Nat. Commun. https:// doi. org/ 10. 1038/ s41467- 018- 07439-9 (2018).
10. Chacko, R. et al. Data based predictive models for odor perception. Sci. Rep. 10, 17136. https:// doi. org/ 10. 1038/ s41598- 020- 73978-1
(2020).
11. Nozaki, Y. & Nakamoto, T. Odor impression prediction from mass spectra. PLoS ONE https:// doi. org/ 10. 1371/ journ al. pone. 01570
30 (2016).
12. Dravnieks, A. Atlas of odor character profiles (ASTM, 1985).
13. SIgma-Aldrich, “Flavors and Fragrances.” Available: http:// www. sigma aldri ch. com/ indus tries/ flavo rs- and- fragr ances. html.
14. Nozaki, Y. & Nakamoto, T. Predictive modeling for odor character of a chemical using machine learning combined with natural
language processing. PLoS ONE https:// doi. org/ 10. 1371/ journ al. pone. 01984 75 (2018).
15. Debnath, T., Prasetyawan, D., & Nakamoto, T. Prediction of odor descriptor group of essential oils from mass spectra using machine
learning. IEEE, 2019 IEEE Int. Symp. Olfaction Electron. Nose 1–3 (2019) https:// doi. org/ 10. 1109/ ISOEN. 2019. 88232 26
16. Debnath, T. & Nakamoto, T. Predicting human odor perception represented by continuous values from mass spectra of essential
oils resembling chemical mixtures. PLoS ONE https:// doi. org/ 10. 1371/ journ al. pone. 02346 88 (2020).
17. Chawla, N. V., Bowyer, K. W., Hall, L. O. & Kegelmeyer, W. P. SMOTE: synthetic minority over-sampling technique. J. Artif. Intell.
Res. 16, 321–357. https:// doi. org/ 10. 1613/ jair. 953 (2002).
18. Japkowicz, N. Learning from Imbalanced Data sets: A Comparison of Various Strategies. In Proceedings of the AAAI’2000 Work-
shop on Learning from Imbalanced Data Sets, Austin, TX. (2000b)
19. Fernández Hilario, A. et al. Learning from Imbalanced data sets 1st edn. (Springer, Cham, 2018).
20. Khan, S. S. & Madden, M. G. A survey of recent trends in one class classification. In Artificial intelligence and cognitive science.
AICS 2009. Lecture notes in computer science Vol. 6206 (eds Coyle, L. & Freyne, J.) (Springer, Berlin, Heidelberg, 2010). https://
doi. org/ 10. 1007/ 978-3- 642- 17080-5_ 21.
21. Schölkopf, B., Platt, J. C., Shawe-Taylor, J., Smola, A. J. & Williamson, R. C. Estimating the support of a high-dimensional distribu-
tion. Neural Comput. 13, 1443–1471. https:// doi. org/ 10. 1162/ 08997 66017 50264 965 (2001).
22. Castro, C. L. & Braga, A. P. Novel cost-sensitive approach to improve the MLP performance on imbalanced data. IEEE Trans. MLPs
Learn. Syst. 24, 888–899. https:// doi. org/ 10. 1109/ TNNLS. 2013. 22461 88 (2013).
23. Leffingwell J. C. Leffingwell & associates, (2005).
24. CAS Number Search. http:// webbo ok. nist. gov/ chemi stry/ cas- ser. html
25. Ling, C. X. & Sheng, V. S. Cost-sensitive learning. In Encyclopedia of machine learning (eds Sammut, C. & Webb, G. I.) (Springer,
Boston, MA, 2011). https:// doi. org/ 10. 1007/ 978-0- 387- 30164-8_ 181.
26. Charles, E. The foundations of cost-sensitive learning, IJCAI’01: Proceedings of the 17th international joint conference on Artificial
intelligence - Vol 2, 973–978. https:// doi. org/ 10. 5555/ 16421 94. 16422 24 (2001).
27. Classification on imbalanced data. https:// www. tenso rflow. org/ tutor ials/ struc tured_ data/ imbal anced_ data# class_ weigh ts
28. Optimizers. https:// keras. io/ api/ optim izers/.
29. Sklearn.svm.OneClassSVM—https:// scikit- learn. org/ stable/ modul es/ gener ated/ sklea rn. svm. OneCl assSVM. html
Author contributions
T.N., T.D.: Conceptualization, data curationT.D.: Investigation, Methodology, Software, VisualizationT.D.: Writ-
ing– original draftT.N.: Funding acquisition, Project administration, SupervisionT.N.: Writing– review&editing.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary Information The online version contains supplementary material available at https:// doi. org/
10. 1038/ s41598- 022- 07802-3.
Correspondence and requests for materials should be addressed to T.D. or T.N.
Scientific Reports | (2022) 12:3778 |
https://doi.org/10.1038/s41598-022-07802-3
8
Vol:.(1234567890)www.nature.com/scientificreports/ Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
© The Author(s) 2022
Scientific Reports | (2022) 12:3778 |
https://doi.org/10.1038/s41598-022-07802-3
9
Vol.:(0123456789)www.nature.com/scientificreports/
| null |
10.1090_bproc_99.pdf
| null | null |
PROCEEDINGS OF THE
AMERICAN MATHEMATICAL SOCIETY, SERIES B
Volume 9, Pages 159–173 (April 12, 2022)
https://doi.org/10.1090/bproc/99
SEGRE-DEGENERATE POINTS FORM A SEMIANALYTIC SET
JI ˇR´I LEBL
(Communicated by Harold P. Boas)
Abstract. We prove that the set of Segre-degenerate points of a real-analytic
subvariety X in Cn is a closed semianalytic set.
It is a subvariety if X is
coherent. More precisely, the set of points where the germ of the Segre variety
is of dimension k or greater is a closed semianalytic set in general, and for a
coherent X, it is a real-analytic subvariety of X. For a hypersurface X in Cn,
the set of Segre-degenerate points, X[n], is a semianalytic set of dimension at
most 2n − 4. If X is coherent, then X[n] is a complex subvariety of (complex)
dimension n − 2. Example hypersurfaces are given showing that X[n] need not
be a subvariety and that it also need not be complex; X[n] can, for instance,
be a real line.
1. Introduction
Segre varieties are a widely used tool for dealing with real-analytic submanifolds
in complex manifolds. Recently, there have been many applications of Segre vari-
ety techniques to singular real-analytic subvarieties, and while the techniques are
powerful, they have to be applied carefully. It is tempting to cite an argument or
result for submanifolds to prove the same result for subvarieties, but there are two
things that can go wrong. First, the Segre variety can be degenerate (of wrong
dimension), and second, the variety itself may be not coherent, and the Segre va-
riety cannot be defined by the same function(s) at all points. One cannot define
Segre varieties with respect to the complexification at one point and expect this
complexification to give a well-defined Segre variety at all nearby points (germs
have complexifications, but their representatives may not). One incorrect but very
tempting statement is that the set of Segre-degenerate points of a real hypersurface
in Cn is necessarily a complex-analytic subvariety. The result follows for coherent
hypersurfaces, but not in general. The set of Segre-degenerate points of a hyper-
surface is not only not a complex-analytic subvariety in general, it need not even
be a real-analytic subvariety, it is merely a semianalytic set. We give an example
where it is not a subvariety, and one where it is of odd real dimension.
The idea of using Segre varieties is old, although the techniques for using them
in CR geometry were brought into prominence first by Webster [14] and Diederich–
Fornæss [7]. For a good introduction to their use for submanifolds, see the book by
Baouendi–Ebenfelt–Rothschild [2]. They started to be used for singular subvarieties
recently, see for example Burns–Gong [5], Diederich–Mazzilli [8], the author [11],
Received by the editors April 2, 2021, and, in revised form, August 2, 2021, and August 12,
2021.
2020 Mathematics Subject Classification. Primary 32C07; Secondary 32B20, 14P15.
Key words and phrases. Segre-degenerate, Segre variety, semianalytic.
The author was supported in part by Simons Foundation collaboration grant 710294.
c(cid:2)2022 by the author(s) under Creative Commons Attribution 3.0 License (CC BY 3.0)
159
160
JI ˇR´I LEBL
Adamus–Randriambololona–Shafikov [1], Fern´andez-P´erez [9], Pinchuk–Shafikov–
Sukhov [13], and many others. However, the reader should be aware that sometimes
in the literature on singular subvarieties a Segre variety is defined with respect to a
single defining function and it is not made clear that the Segre variety is then not
well-defined if the point moves.
A good reference for real-analytic geometry is Guaraldo–Macr`ı–Tancredi [10],
and a good reference for complex analytic subvarieties is Whitney [16].
A real-analytic subvariety of an open U ⊂ Cn is a relatively closed subset X ⊂ U
defined locally by the vanishing of real-analytic functions. If p ∈ X, then the ideal
Ip(X) of real-analytic germs at p vanishing on X is generated by the components
of a mapping f (z, ¯z). Let ΣpX, the germ of the Segre variety at p, be the germ at
p of a complex-analytic subvariety given by the vanishing of z (cid:4)→ f (z, ¯p) (ΣpX is
independent of the generator f ). Normally ΣpX is of the same complex codimension
as is the real codimension of X. So if X is a real hypersurface, then ΣpX is usually
a germ of a complex hypersurface. For a hypersurface, we say X is Segre-degenerate
at p if ΣpX is not a complex hypersurface, that is, if ΣpX = (Cn, p). See §3 for a
more precise definition.
One of the main differences of real and complex varieties is that real varieties need
not be coherent. A real-analytic subvariety is coherent if the sheaf of germs of real-
analytic functions vanishing on X is a coherent sheaf. Equivalently, X is coherent if
it has a complexification, that is, a single variety that defines the complexification of
all germs of X, or in yet other words, if for every p, representatives of the generators
of Ip(X) generate the ideals Iq(X) for all nearby q. For the hypersurface case, we
prove the following result.
Theorem 1.1. Let U ⊂ Cn be open and X ⊂ U a real-analytic subvariety of
codimension 1 (a hypersurface). Let X[n] ⊂ X be the set of Segre-degenerate points.
Then:
(i) X[n] is a semianalytic set of dimension at most 2n − 4, which is locally
contained in a complex-analytic subvariety of (complex) dimension at most
n − 2.
(ii) If X is coherent, then X[n] is a complex-analytic subvariety of (complex)
dimension at most n − 2.
The dimension of the complex subvariety may be smaller than n−2. Example 6.1
gives a coherent hypersurface in C3 where X[n] is an isolated point. For noncoherent
X, examples exist for which X[n] is not a complex variety, or that are not even a
In particular, the dimension of X[n] need not be even.
real-analytic subvariety.
Example 6.6 is a hypersurface in C3 such that (real) dimension of X[n] is 1.
In
Example 6.5, X[n] is only semianalytic and not a real-analytic subvariety.
p X is then ΣpX. However, for a noncoherent X, the germ of ΣU
The Segre variety can be defined with respect to a specific defining function, or
a neighborhood U of a point p. For a small enough U , take the representatives of
q X for all q ∈ X. The germ of
the generators of Ip(X), and use those to define ΣU
ΣU
q X at q need
not be the same as the germ ΣqX, no matter how small U is and how close q is
to p, since the representatives of the generators of Ip(X) may not generate Iq(X).
There may even be regular points q arbitrarily close to p where ΣU
q X is singular
(reducible) at q. See Example 6.4. If q is a regular point where X is generic (e.g.
a hypersurface), the germ ΣqX is always regular. The point is that the germs
SEGRE-DEGENERATE POINTS FORM A SEMIANALYTIC SET
161
ΣqX cannot be defined coherently by a single set of equations for a noncoherent
subvariety.
The results above are a special case of results for higher codimension. In general,
the set of “Segre-degenerate points” would be points where the Segre variety is not
of the expected dimension. The main result of this paper is that for general X, we
can stratify X into semianalytic sets by the dimension of the Segre variety.
Theorem 1.2. Let U ⊂ Cn be open and X ⊂ U a real-analytic subvariety of
dimension d < 2n (i.e. X (cid:6)= U ). Let X[k] ⊂ X be the set of points where the Segre
variety is of dimension k or higher. Then:
(i) For every k = 0, 1, . . . , n, X[k] is a closed semianalytic subset of X, and
X[n] is locally (as germs at every point) contained in a complex-analytic
subvariety of dimension at most n − d − 1.
(ii) If X is coherent, then for every k = 0, 1, . . . , n, X[k] is a closed real-
analytic subvariety of X, and X[n] is a complex-analytic subvariety of di-
mension at most n − d − 1.
The sets X[k] are nested: X[k+1] ⊂ X[k].
If X is of pure dimension d ≥ n,
we find that X[d−n] = X. Then X[n] ⊂ · · · ⊂ X[k] ⊂ · · · ⊂ X[d−n] = X.
If,
furthermore, there exists a regular point of X where X is a generic submanifold,
then X[d−n+1] (the reasonable definition of “Segre-degenerate points” in this case)
is a semianalytic subset of X of dimension less than d, since where X is a generic
submanifold the dimension of the Segre variety is necessarily d − n. We avoid
defining the term Segre-degenerate for general X as the Segre varieties can be
degenerate in various ways; it is better to just talk about the sets X[k] or the sets
X[k] \ X[k+1]. In any case, since the sets X[k] are semianalytic, every reasonable
definition of “Segre-degenerate” based on dimension leads to a semianalytic set.
Notice that for k < n, the set X[k] is not necessarily complex even if it is a proper
subset of a coherent X, see Example 6.2.
The structure of this paper is as follows. First, we cover some preliminary
results on subvarieties and semianalytic sets in §2. We introduce Segre varieties in
the singular case in §3. In §4, we prove the simpler results for the coherent case,
and we cover the noncoherent case in §5. In §6 we present some of the examples
showing that the results are optimal and particularly illustrating the degeneracy of
the noncoherent case.
2. Preliminaries
We remark that the content of this section is not new but totally classical, and
the degeneracies shown in the examples have been known for a long time, already
by Cartan, Whitney, Bruhat, and others. See e.g. [6, 15].
Definition 2.1. Let U ⊂ Rk (respectively U ⊂ Ck) be open. The set X ⊂ U is a
real-analytic subvariety (resp. a complex-analytic subvariety) of U if for each point
p ∈ U , there exists a neighborhood V ⊂ U of p and a set of real-analytic (resp.
holomorphic) functions F(V ) such that
(1)
X ∩ V = {p ∈ V : f (p) = 0 for all f ∈ F(V )}.
162
JI ˇR´I LEBL
Write Xreg ⊂ X for the set of points which are regular, that is,
(2)
Xreg
def= {p ∈ X : ∃ neighborhood V of p, such that
V ∩ X is a real-analytic (resp. complex) submanifold}.
def= X \ Xreg. The dimension
The set of singular points is the complement: Xsing
of X at p ∈ Xreg, written as dimp X, is the real (resp. complex) dimension of the
real-analytic (resp. complex) manifold at p. The dimension of X, written as dim X,
is the maximum dimension at any regular point. The dimension of X at p ∈ Xsing
is the minimum dimension of X ∩ V over all neighborhoods V of p. Define
(3)
X ∗ def= {p ∈ Xreg : dimp X = dim X}.
A variety or germ is irreducible if it cannot be written as a union of two proper
subvarieties. Let Iq(X) denote the ideal of germs (f, q) of functions that vanish on
the germ (X, q).
An analytic space is, like an abstract manifold, a topological space with an
atlas of charts with real-analytic (resp. holomorphic) transition maps, but the local
models are subvarieties rather than open sets of Rn or Cn. See e.g. [10, 16].
Subvarieties are closed subsets of U . If a topology on X is required, we take the
subspace topology. Unlike in the complex case, a real-analytic subvariety can be a
C k-manifold while being singular as a subvariety. For example, x2 − y2k+1 = 0 in
R2. Also, in the real case, the set of singular points need not be a subvariety and
X ∗ need not equal Xreg.
Definition 2.2 (See e.g. [3, 12]). For a set V (an open set in Rn, or a subvariety),
let S
be the smallest family of sets (the intersection of all such families)
that is closed under finite unions, finite intersections, and complements of sets of
the form
(cid:5)
x ∈ V : f (x) ≥ 0
(4)
where f ∈ C ω(V ) (f real-analytic in V , or a restriction of a real-analytic function
if V is a subvariety).
(cid:2)
C ω(V )
(cid:4)
(cid:3)
,
A set X ⊂ U is semianalytic (in U ) if for each p ∈ U , there is a neighborhood V
(cid:3)
. Here U is an open set in Rn, a subvariety, or
(cid:2)
C ω(V )
of p such that X ∩ V ∈ S
an analytic space.
Note that {x : f (x) ≤ 0} = {x : −f (x) ≥ 0}. Equality is obtained by intersecting
{x : f (x) ≥ 0} and {x : −f (x) ≥ 0}. Complement obtains sets of the form
{x : f (x) > 0} and {x : f (x) (cid:6)= 0}. Thus we have all equalities and inequalities.
Subvarieties are semianalytic, but the family of semianalytic sets is much richer.
If X is a complex-analytic subvariety, then Xsing is a complex-analytic subvariety,
while if X is only real-analytic, then Xsing is only a semianalytic subset.
Example 2.3. The Whitney umbrella, sx2 = y2 in R3 using coordinates (x, y, s),
is a set X where Xsing is the set given by x = 0, y = 0, and s ≥ 0.
It is a common misconception related to the subject of this paper to think that
the set of singular points of a real subvariety X can be defined by the vanishing of
the derivatives of functions that vanish on X. For a subvariety X defined near p,
it is possible that dψ vanishes on some regular points of X arbitrarily near p for
SEGRE-DEGENERATE POINTS FORM A SEMIANALYTIC SET
163
every function ψ defined near p such that ψ = 0 on X. Before proving this fact, let
us prove a simple lemma.
Lemma 2.4. Suppose X = {x ∈ Rk : P (x) = 0} for an irreducible homogeneous
polynomial P (irreducible in the ring of polynomials) and X is a hypersurface (di-
mension k − 1).
If (f, 0) is a germ of a real-analytic function that vanishes on X, then (f, 0) is a
multiple of the germ (P, 0). In other words, I0(X) is generated by the germ (P, 0).
Proof. The proof is standard, it is a version of one of the claims from the proof
of Chow’s theorem. Clearly, X is a real cone, that is, if x ∈ X then λx ∈ X for
all λ ∈ R. Write a representative f (x) =
(cid:3)=0 f(cid:3)(x) in terms of homogeneous
parts. Suppose x ∈ X, so f (x) = 0. As λx ∈ X, then f (λx) ≡ 0. But then
(cid:6)∞
(cid:3)=0 λ(cid:3)f(cid:3)(x) is identically zero, meaning f(cid:3)(x) = 0 for all (cid:4). Since
X is a hypersurface, the polynomial P generates the ideal of all polynomials van-
ishing on X and thus P divides all the polynomials f(cid:3) (See e.g. Theorem 4.5.1
(cid:2)
in [4]). Thence, the germ (P, 0) divides the germ (f, 0).
(cid:3)=0 f(cid:3)(λx) =
(cid:6)∞
(cid:6)∞
Example 2.5. Let us give an example of a pure 2-dimensional real-analytic subvari-
ety X ⊂ R3 with an isolated singularity at the origin, such that for any real-analytic
defining function ψ of X near the origin, the set where both dψ and ψ vanish is a
1-dimensional subset of X. Therefore, the set where the derivative vanishes for the
defining function is of larger dimension than the singular set, and dψ vanishes at
some regular points. This subvariety will be a useful example later (Example 6.4),
and it is a useful example of a noncoherent subvariety where coherence breaks not
because of a smaller dimensional component.
Let X be the subvariety of R3 in the coordinates (x, y, s) ∈ R3:
(5)
(x2 + y2)6 − s8x3(s − x) = 0.
We claim that X is as above. Despite the singularity being just the origin, for any
real-analytic ψ defined near the origin that vanishes on X, we get dψ(0, 0, s) = 0,
so the derivative vanishes on
(cid:4)
(cid:5)
(x, y, s) ∈ R3 : x = 0, y = 0
= {0} × {0} × R ⊂ X.
As this example will be useful for Segre varieties, we prove the claim in detail.
The subvariety in R2 defined by (x2 + y2)6 − x3(1 − x) = 0 is irreducible. Indeed,
it is a connected compact submanifold. To see that it is connected and compact,
solve for y. The tricky part is showing that the subvariety is nonsingular near the
origin, which can be seen by writing (x2 + y2)6 = x3(1 − x) and taking the third
root to get
√
(x2 + y2)2 = x 3
1 − x.
(6)
(7)
Near the origin, we can solve for x using the implicit function theorem.
Homogenize (x2 + y2)6 − x3(1 − x) with s to get the set X in R3 given by (5).
The set X is a cone with an isolated singularity; it is a cone over a manifold. By
Lemma 2.4, if ψ vanishes on X, then
(8)
ψ =
(cid:2)
(x2 + y2)6 − s8x3(s − x)
(cid:3)
ϕ.
In other words, on X, dψ must vanish where the derivative of (x2 +y2)6−s8x3(s−x)
vanishes.
164
JI ˇR´I LEBL
Consider subvarieties of Cn ∼
3. Segre varieties
= R2n. Let U ⊂ Cn be open and X ⊂ U a real-
analytic subvariety. Write U conj = {z : ¯z ∈ U } for the complex conjugate. Let
ι(z) = (z, ¯z) be the embedding of Cn into the “diagonal” in Cn × Cn. Denote by
X U the smallest complex-analytic subvariety of U × U conj such that ι(X) ⊂ X U .
By smallest we mean the intersection of all such subvarieties. It is standard that
there exists a small enough U (see below) such that X U ∩ ι(Cn) = ι(X). Let
σ : Cn × Cn → Cn × Cn denote the involution σ(z, w) = ( ¯w, ¯z). Note that the
“diagonal” ι(Cn) is the fixed set of σ.
Proposition 3.1. Let U ⊂ Cn be open and X ⊂ U be a real-analytic subvariety.
Then σ(X U ) = X U .
Proof. The set σ(X U ) is a complex-analytic subvariety as it is defined by vanishing
of anti-holomorphic functions, and hence by holomorphic functions. As X is fixed
by σ, we have X ⊂ σ(X U ) ∩ X U , and the result follows as X U is the smallest
(cid:2)
subvariety containing X.
The ideal Ip(X) can be generated by the real and imaginary parts of the gen-
erators of the ideal of germs of holomorphic functions defined at (p, ¯p) in the com-
plexification that vanish on the germ of ι(X) at (p, ¯p). Call the ideal of these
holomorphic functions Ip(X).
Given a germ of a real-analytic subvariety (X, p), denote by Xp the smallest
germ of a complex-analytic subvariety of
that contains the image of
(X, p) by ι. The germ Xp is called the complexification of (X, p). It is not hard to see
that the irreducible components of (X, p) correspond to the irreducible components
of Xp; if (X, p) is irreducible, so is Xp. In the theory of real-analytic subvarieties,
X U would not be called a complexification of X unless
= Xp for all
p ∈ X, and that cannot always be achieved.
(cid:2)
Cn×Cn, (p, ¯p)
(cid:2)
X U , (p, ¯p)
(cid:3)
(cid:3)
As we will need a specific neighborhood often, we make Definition 3.2.
Definition 3.2. Let X ⊂ U be a real-analytic subvariety of dimension d of an
open set U ⊂ Cn. We say U is good for X at p ∈ X if the following conditions are
satisfied:
(i) U is connected.
(ii) The real dimension of (X, p) is d and the complex dimension of Xp and X U
is also d.
(iii) There exists a real-analytic function ψ : U → Rk whose complexification
converges in U ×U conj, whose zero set is X, and whose germ (ψ, p) generates
Ip(X).
(cid:2)
X U , (p, ¯p)
(iv) X U ∩ ι(Cn) = ι(X).
(v)
(vi) The irreducible components of X U correspond in a one-to-one fashion to
= Xp.
(cid:3)
the irreducible components of the germ Xp.
If U (cid:4) ⊂ U is good for X ∩ U (cid:4) at p we say simply that U (cid:4) is good for X at p.
Proposition 3.3. Suppose U ⊂ Cn is open, X ⊂ U is a real-analytic subvariety,
and p ∈ X. Then there exists a neighborhood U (cid:4) ⊂ U of p such that U (cid:4) is good for
X at p.
Furthermore, for any neighborhood W of p, there exists a neighborhood W (cid:4) ⊂ W
of p that is good for X at p.
SEGRE-DEGENERATE POINTS FORM A SEMIANALYTIC SET
165
Proof. The idea is standard (see e.g.
[10]), but let us sketch a proof. The main
difficulty is mostly notational. Take the germ Ip(X) of complexified functions
that vanish on the germ of ι(X) at (p, ¯p). Note that Ip(X) is closed under the
conjugation taking ψ to ψ ◦ σ, that is, ψ(z, ζ) to ¯ψ(ζ, z). It is generated by a finite
set of functions f1, . . . , fk, which are all defined in some polydisc Δ×Δconj centered
at p. The real and imaginary parts of these functions also generate an ideal, and
this ideal must be equal to Ip(X). We can also assume that Δ is small enough that
all the components of the subvariety V defined by f1, . . . , fk go through (p, ¯p) (in
other words V is the smallest subvariety of Δ containing the germ of ι(X) at (p, ¯p)).
Similarly, make Δ small enough that the real and imaginary parts of f1, . . . , fk
restricted to the diagonal give the subvariety X ∩ Δ all of whose components go
through p. We can take Δ to also be small enough that all components of Xp have
distinct representatives in Δ. The set Δ is our U (cid:4).
(cid:2)
Definition 3.4. Suppose U ⊂ Cn is open and X ⊂ U is a real-analytic subvariety.
The Segre variety of X at p ∈ U relative to U is the set
(cid:5)
(cid:4)
(9)
If U (cid:4) ⊂ U , we write ΣU (cid:2)
p X def=
ΣU
p X for ΣU (cid:2)
When U (cid:4) is good for X at p ∈ X, define the germ
p (X ∩ U (cid:4)).
z ∈ U : (z, ¯p) ∈ X U
.
(10)
Define
(11)
(12)
ΣpX def=
(cid:2)
ΣU (cid:2)
p X, p
(cid:3)
.
(cid:4)
(cid:4)
def=
def=
X[k]
XU[k]
(cid:5)
z ∈ U : dim ΣzX ≥ k
z X ≥ k
z ∈ U : dimz ΣU
,
(cid:5)
.
The germ ΣpX is well-defined by the proposition. First, there exists a good
neighborhood of p, and any smaller good neighborhood of p would give us the same
germ of the complexification at p.
If X is an irreducible hypersurface, X is Segre-degenerate at p ∈ X if ΣpX =
(Cn, p), that is, if p ∈ X[n]. A point p is Segre-degenerate relative to U if dimp ΣU
p X
= n, that is, if p ∈ XU[n]. A key point of this paper is that these two notions can
be different. We will see that XU[n] is always a complex subvariety and contains
X[n], and the two are not necessarily equal even for a small enough U . They may
not be even of the same dimension.
For a general dimension d set, we will simply talk about the sets X[k] and we
will not make a judgement on what is the best definition for the word “Segre-
degenerate.”
A (smooth) submanifold is called generic (see [2]) at p if in some local holomor-
phic coordinates (z, w) ∈ Cn−k × Ck vanishing at p it is defined by
(13)
Im w1 = r1(z, ¯z, Re w),
. . . ,
Im wk = rk(z, ¯z, Re w),
with rj and its derivative vanishing at 0. For instance, a hypersurface is generic.
Proposition 3.5. Suppose X is a real-analytic submanifold of Cn of dimension d
(so codimension 2n − d) and p ∈ X.
(i) dim ΣpX ≤ d
(ii) If X is generic at p, then ΣpX is a germ of a complex submanifold and
2 . In particular, if d < 2n, then X[n] = ∅.
dim ΣpX = d − n.
166
JI ˇR´I LEBL
Proof. We start with the generic case. Using the defining functions above, k =
2n − d, we note that if we plug in ¯w = 0 and ¯z = 0, we get k linearly independent
defining equations for a complex submanifold.
If X is not generic, then we can write down similar equations but solve for real
and imaginary parts some of the variables. Since these could conceivably be the real
2 = 2n−d
and imaginary parts of the same variable, we may only get k
independent
equations, so the dimension of ΣpX could be as high as n − 2n−d
(cid:2)
2 = d
2 .
2
If X is regular at p but not generic, the germ ΣpX could possibly be singular
and the dimension may vary as p moves on the submanifold. See Example 6.3.
Let us collect some basic properties of Segre varieties in the singular case.
p X, p
Proposition 3.6. Let X ⊂ U ⊂ Cn be a real-analytic subvariety of dimension d
and p ∈ X. Then
(cid:3)
(cid:2)
(i) ΣpX ⊂
ΣU
.
(ii) dim ΣpX ≥ d − n.
(iii) X[k] ⊂ XU[k] for all k.
(iv) If U is good for X at p and d ≤ n, then dim ΣU
(v) If d ≤ n, then X[n] = ∅.
(vi) q ∈ ΣqX if and only if q ∈ X, and so X[k] ⊂ X for all k = 0, 1, 2, . . . , n.
(vii) If U is good for X at p, then q ∈ ΣU
q X if and only if q ∈ X, and so
XU[k] ⊂ X for all k = 0, 1, 2, . . . , n.
p X < n.
Proof. If U (cid:4) ⊂ U then ΣU (cid:2)
p , as any analytic function defined on U that
p
vanishes on X is an analytic function on U (cid:4) that vanishes on X ∩ U (cid:4). Parts (i) and
(iii) follow.
⊂ ΣU
For part (ii), the complexification Xp has dimension d. Let U (cid:4) be good for X at
p. The germ ΣpX is the germ at p of the intersection of X U (cid:2)
and Cn × {¯p}. The
codimension of X U (cid:2)
at (p, ¯p) in Cn × Cn is 2n − d, and the codimension of Cn × {¯p}
is n. Hence, their intersection is of codimension at most 3n − d or dimension
2n − (3n − d) = d − n.
To see (iv) first note that if d < n, then it is impossible for ΣU
q X to be of
dimension n as it is a subvariety of X U , which is of dimension d < n. If d = n,
then without loss of generality suppose that (X, p) is irreducible. As U is good
for X at p, then X U is also irreducible. By dimension, as X U is of dimension n
and dim ΣpX = n, then U × {¯p} would be an irreducible component of X U . By
symmetry (applying σ), {p} × U conj is also an irreducible component of X U . This
is a contradiction as X U is irreducible.
Then (v) follows from (iv) by considering a small enough good neighborhood of
every q ∈ X.
For (vi), if q ∈ X, then q ∈ ΣqX, since ψ(q, ¯q) = 0 for any germ of a function at
q that vanishes on X. If q /∈ X, then clearly ΣqX = ∅. So X[k] ⊂ X.
For (vii), again if q ∈ X, then it must be that q ∈ ΣU
U , we have X U = ι(X), and so q ∈ ΣU
q X means that q ∈ X.
q X. Similarly, for a good
(cid:2)
The point of this paper is that even for arbitrarily small neighborhoods U of p
(even good for X at p) and a q ∈ U that is arbitrarily close to p, it is possible that
(14)
(ΣU
p X, q) (cid:6)= ΣqX.
SEGRE-DEGENERATE POINTS FORM A SEMIANALYTIC SET
167
That is, unless X is coherent. Let us focus on X[n] for a moment. It is possible
that for all neighborhoods U of a point p ∈ X,
X[n] (cid:6)= XU[n].
(15)
The set XU[n] is rather well-behaved.
Proposition 3.7. Let X ⊂ U ⊂ Cn be a real-analytic subvariety of dimension
d < 2n, and suppose U is good for X at some p ∈ X. Then XU[n] is a complex-
analytic subvariety of dimension at most n − d − 1. In particular, X[n] is contained
in a complex-analytic subvariety of dimension at most n − d − 1.
Proof. Without loss of generality, suppose that (X, p) is irreducible. The variety
X U is fixed by the involution σ. In other words, (z, ¯w) ∈ X U if and only if (w, ¯z) ∈
X U . So if q ∈ XU[n], then (z, ¯q) ∈ X U for all z ∈ U , and therefore (q, ¯z) ∈ X U
for all z ∈ U .
z X is a complex-
analytic subvariety, generically of dimension d−n, then XU[n] is a complex-analytic
subvariety of dimension at most d − n.
z X for all z ∈ U . As ΣU
In particular, q ∈ ΣU
The only way that XU[n] could be of dimension d − n is if all the varieties ΣU
z X
contained a fixed complex-analytic subvariety V of dimension d − n. This means
that V × Cn ⊂ X U and Cn × V conj ⊂ X U (by applying σ). By dimension, these are
components of X U . Since we assumed that (X, p) is irreducible, so is Xp and so is
X U if U is good for X at p, and we obtain a contradiction. Hence, XU[n] must be
(cid:2)
of dimension at most d − n − 1.
4. Coherent varieties
A real-analytic subvariety is coherent if the sheaf of germs of real-analytic func-
tions vanishing on X is a coherent sheaf. The fundamental fact about coherent
subvarieties is that they possess a global complexification. That is, if X is co-
herent, then there exists a complex-analytic subvariety X of some neighborhood
(cid:2)
of X in Cn × Cn such that X ∩ ι(Cn) = X and
is equal to Xp, the
X , (p, ¯p)
complexification of the germ (X, p) at every p ∈ X. See [10].
(cid:3)
Lemma 4.1. Let X ⊂ U ⊂ Cn be a real-analytic subvariety. If X ⊂ U is coherent
q X, q) for all q ∈ X. In particular,
and U is good for X at p ∈ X, then ΣqX = (ΣU
XU[n] = X[n].
Proof. Since X is coherent, we have a global complexification X and hence X U =
X ∩ U . In particular, this is true for any good neighborhood U (cid:4) ⊂ U of any point
q ∈ X, so ΣqX = (ΣU (cid:2)
q X, q) = (ΣU
q X, q).
As XU[n] ⊂ X and it is the set where (ΣU
q X, q) is of dimension n, we find that it
is equal to the set where ΣqX is of dimension n. In other words, XU[n] = X[n]. (cid:2)
We can now prove the theorem for coherent subvarieties. Theorem 4.2 implies the
coherent part of Theorem 1.2 for k = n, and hence the coherent part of Theorem 1.1.
Theorem 4.2. Let U ⊂ Cn be open and X ⊂ U be a coherent real-analytic subvari-
ety of dimension d < 2n. Then X[n] is a complex-analytic subvariety of dimension
at most d − n − 1.
Proof. It is sufficient to work in a good neighborhood of some point, without loss
of generality, assume that U is good for some p ∈ X. Apply the lemma and
(cid:2)
Proposition 3.7.
168
JI ˇR´I LEBL
For general k, we have Theorem 4.3, which finishes the coherent case of The-
orem 1.2 for k < n. That is, for every k, the X[k] sets are subvarieties of X for
coherent X. These subvarieties no longer need to be complex-analytic.
Theorem 4.3. Let U ⊂ Cn be open and X ⊂ U be a coherent real-analytic subva-
riety. Then for every k = 0, 1, . . . , n, X[k] is a real-analytic subvariety of X.
A generic submanifold has the Segre variety of the least possible dimension. Let
X be an irreducible coherent subvariety of dimension d. If Xreg is generic at some
point, then ΣqX is of (the minimum possible) dimension d − n somewhere. The
Segre-degenerate set is the set where ΣqX is higher than d − n, that is, it is the
set X[d−n+1]. According to this theorem, this Segre-degenerate set X[d−n+1] is a
real-analytic subvariety of X.
Proof. It is a local result and so without loss of generality, assume that U is good
for X at some p ∈ X. Let (z, ξ) be the complexified variables of Cn × Cn. Consider
the projection π(z, ξ) = ξ defined on X U . The Segre variety ΣU
z X is (identified
(cid:3)
with) the fiber π−1
. The dimension of the germ ΣzX is the dimension at
(cid:2)
(z, ¯z) of π−1
π(z, ¯z)
(cid:2)
π(z, ¯z)
(cid:3)
as X is coherent. For any integer k, the set
(cid:8)
(cid:7)
(z, ξ) ∈ X U : dim(z,ξ) π−1
(cid:2)
π(z, ξ)
(cid:3)
≥ k
(16)
Vk =
is a complex-analytic subvariety of X U (see e.g. Theorem 9F in chapter 7 of Whit-
(cid:2)
ney [16]). Then Vk ∩ {ξ = ¯z} = ι(X[k]) is a real-analytic subvariety of ι(X).
5. The set of Segre-degenerate points is semianalytic
It is rather simple to prove that X[k] is always closed (in classical, not Zariski,
topology).
Proposition 5.1. Let U ⊂ Cn be open and X ⊂ U be a real-analytic subvariety.
Then q (cid:4)→ dim ΣqX is an upper semicontinuous function on X. In particular, for
every k, X[k] is closed.
Proof. Let p ∈ X be some point and let U be good for X at p and follow the
construction in the proof of Theorem 4.3, that is let X U and π be as before. The
(cid:3)
(cid:2)
z X is the fiber π−1
. For all z ∈ X, ΣzX is a subset
Segre variety ΣU
π(z, ¯z)
z X ≥
(possibly proper as X is not coherent) of the germ (ΣU
p X, p) = ΣpX. As the sets Vk ∩ ι(X) are
dim ΣzX. As U is good for X at p, (ΣU
closed, dim ΣpX = (ΣU
q X, q) for all
(cid:2)
sufficiently nearby q, and these are in turn bounded below by dim ΣqX.
p X, p) is bounded below by dimensions of (ΣU
z X, z), and so dimz ΣU
We need some results about semianalytic subsets. We are going to use nor-
malization on X U and so we need to prove that semianalytic sets are preserved
under finite holomorphic mappings. The key point in that proof is Theorem 5.2 on
projection of semialgebraic sets extended to handle certain semianalytic sets.
Theorem 5.2 ((cid:6)Lojasiewicz–Tarski–Seidenberg (see [3, 12])). Let A be a ring of
real-valued functions on a set U , and let π : U × Rm → U be the projection.
(cid:3)
, then π(X) ∈ S(A).
(cid:2)
A[t1, . . . , tm]
If X ∈ S
Complex-analytic subvarieties are preserved under finite (or just proper) holo-
morphic maps. Real semialgebraic sets are preserved under all real polynomial
maps. On the other hand real-analytic subvarieties or semianalytic sets are not
SEGRE-DEGENERATE POINTS FORM A SEMIANALYTIC SET
169
preserved by finite or proper real-analytic maps. But, as long as the map is holo-
morphic and finite, semianalytic sets are preserved. Here is an intuitive useful
argument of why this is expected: Map forward the complexification of a real-
(cid:3)
analytic subvariety by the complexification of the map (z, ¯z) (cid:4)→
, which
is still finite, so it maps the complexification to a complex-analytic subvariety. So
the image of a real-analytic subvariety of dimension d via a finite holomorphic map
is contained in a real-analytic subvariety of dimension d. To get equality we need
to go to semianalytic sets: Think of z (cid:4)→ z2 as the map and the real line as the
real-analytic subvariety. The holomorphicity is required as the complexification of
a finite real-analytic map need not be finite (simple example: z (cid:4)→ z ¯z + i(z + ¯z)).
(cid:2)
f (z), ¯f (¯z)
Lemma 5.3. Let V, W be complex analytic spaces, S ⊂ V a semianalytic set,
and f : V → W a finite holomorphic map. Then f (S) is semianalytic of the same
dimension as S.
Proof. Without loss of generality, assume that f (V ) = W . Furthermore, since the
map is finite, and finite unions of semianalytic sets are semianalytic, assume that
V, W are actual complex-analytic subvarieties by working locally in some chart, and
in general we can just assume we are working in an arbitrarily small neighborhood
of the origin 0 ∈ V , and that f (0) = 0. Suppose V is a subvariety of some
neighborhood U ⊂ Cn, and W is a subvariety of some open set U (cid:4) ⊂ Cm. By
adding components to f equal to the defining functions of V (and thus possibly
increasing m) we can assume without loss of generality that f : U → Cm is a finite
map on U and not just V .
Consider the graph Γf of f in U × Cm. As f is finite, the projection of Γf to Cm
is finite. Hence, the variety Γf can be defined by functions that are polynomials in
the first n variables (in fact polynomials in the first n variables and a few of the last
m variables depending on the codimension of W = f (V ) in Cm). Let z = x + iy
denote the first n variables, and ξ denote the last m variables. The variety Γf as a
real subvariety is defined by functions that are polynomials in x and y.
(cid:2)
C ω(U )
Also assume that U is small enough so that S is defined by real-analytic functions
(cid:3)
. The set S corresponds to a semianalytic set (cid:9)S ⊂ Γf .
in U , that is, S ∈ S
The set (cid:9)S is defined by functions defined in some U × U (cid:4), suppose ϕ is one of
these functions. The subvariety Γf is defined by polynomials in x and y, so we
find Weierstrass polynomials in every one of x and y with coefficients real-analytic
functions in ξ that are in the real-analytic ideal for Γf at (0, 0). Since adding
anything in the ideal does not change ϕ where it matters (on Γf ), we can divide
by these polynomials and find a remainder ψ, which is a polynomial in x and y
(cid:3)
such that ψ = ϕ on Γf . In other words, (cid:9)S ∈ S
. By the (cid:6)Lojasiewicz–
Tarski–Seidenberg theorem, the projection of (cid:9)S to U (cid:4) is semianalytic.
The fact that the dimension is preserved follows from f being finite.
(cid:2)
C ω(U (cid:4))[x, y]
(cid:2)
Remark 5.4. The conclusion of the lemma is not true if f is not holomorphic and
finite. If f is proper but not holomorphic, the best we can conclude is that f (S)
is subanalytic as long as we also assume that S is precompact. Our task would be
easier if we only desired to prove that X[k] is subanalytic.
The proof that X[k] is semianalytic for non-coherent subvarieties is similar to
Theorem 4.3, but we work on the normalization of the complex variety X U .
170
JI ˇR´I LEBL
Theorem 5.5. Let U ⊂ Cn be open and X ⊂ U be a real-analytic subvariety. Then
for every k = 0, 1, . . . , n, X[k] is a closed semianalytic subset of X.
Proof. Again, it is a local result, so without loss of generality, assume that U is
good for X at some p ∈ X and suppose that X is irreducible at p and that X is
of dimension d. Consider h : Y → X U , the normalization of X U . There are two
reasons why X U is not the complexification at some point q. For points z arbitrarily
near q, either the set X is of lower dimension at z or there are multiple irreducible
components of the germ
(cid:2)
X U , (z, ¯z)
(cid:3)
.
Let X ∗ denote the relative closure in U of the set of points of dimension d. The
set X \ X ∗ is semianalytic, and so locally near any q ∈ X it is possible to write
X = X ∗ ∪ X (cid:4) for X (cid:4) a real-analytic subvariety of lower dimension (possibly empty)
defined in a neighborhood of q. Suppose for induction that X (cid:4)
[k] is semianalytic.
Then X (cid:4)
\ X ∗ = X[k] \ X ∗ is also semianalytic (in a neighborhood of q). In other
[k]
words, it remains to prove that X[k] ∩ X ∗ is semianalytic.
(cid:3)
, and note that this is a closed semianalytic subset of Y
of dimension d, although it can have points of various dimensions. Therefore, take
X2 = X ∗
1 to be the closure (in Y) of the nonsingular points of X1 of dimension d.
It is clear that h(X2) = ι(X ∗).
Let X1 = h−1
(cid:2)
ι(X ∗)
Let (z, ξ) be the complexified variables of Cn × Cn, where X U lives. Consider
z X is the fiber
(cid:3)
(cid:2)
, but the germ at (z, ¯z) may contain other components, so we pull back
the projection π(z, ξ) = ξ defined on X U . The Segre variety ΣU
π−1
to Y.
Let η be the variable on Y and we pull back via h as (π ◦ h)−1
(cid:3)
. The
space Y is normal and so the germ (Y, η) is irreducible for all η. Near some η ∈ X2,
the set X2 is a totally-real subset of Y of dimension d. Hence (Y, η), which is
irreducible and of dimension d, contains (X2, η) and is then the smallest complex
subvariety containing (X2, η). The germ of the complexification of X at h(η) has as
its components the images of (Y, η(cid:4)) via h for all η(cid:4) ∈ h−1(h(η)) ∩ X2. These images
must be contained in the complexification and as h(X2) = ι(X ∗), their union is the
entire complexification of X at h(η). We thus need to consider the sets
η ∈ Y : dimη(π ◦ h)−1
(cid:2)
π ◦ h(η)
(cid:2)
π ◦ h(η)
π(z, ¯z)
Wk =
≥ k
(17)
(cid:8)
(cid:7)
(cid:3)
,
which are again complex analytic. We are interested in the sets X2 ∩ Wk, which
are semianalytic, and we have just proved above that h(X2 ∩ Wk) = ι(X[k]). As h
(cid:2)
is finite and X2 ∩ Wk is semianalytic, we are finished.
6. Examples of Segre variety degeneracies
Example 6.1. The set of Segre-degenerate points of a coherent hypersurface in
Cn can be a complex subvariety of dimension strictly less than n − 2. Let X ⊂ C3
in coordinates (z, w, ξ) ∈ C3 be given by
(18)
z ¯z + w ¯w − ξ ¯ξ = 0.
The set of regular points is everything except the origin, so only the origin can be
Segre-degenerate, and for this subvariety, it is, as the above equation generates the
ideal I0(X) by Lemma 2.4. So X[3] = {0}, which is of dimension n − 3 = 0.
Example 6.2. For a higher codimensional subvariety, the set X[k] for k < n is
generally not complex. Clearly if k ≤ d − n, then X[k] = X and X is not necessarily
SEGRE-DEGENERATE POINTS FORM A SEMIANALYTIC SET
171
complex. But even for higher k less than n, the set need not be complex. Let
X ⊂ C3 in coordinates (z, w, ξ) ∈ C3 be given by
(19)
z ¯z − w ¯w = 0,
Im ξ = 0.
The subvariety X is 4-dimensional and coherent. It is easy to see that X[1] = X,
X[2] = {z = 0, w = 0, Im ξ = 0}, and X[3] = ∅. The set X[2] is not complex.
Example 6.3. A submanifold may be Segre-degenerate, if it is CR singular. Let
(z, w) be the coordinates in C2 and consider the manifold X given by
(20)
w = z ¯z.
As this is a complex equation, to find the generators of the ideal, we must take
the real and imaginary parts, or equivalently, also consider the conjugate of the
equation, ¯w = z ¯z. For points where z (cid:6)= 0, the Segre variety is just the trivial
point, so zero dimensional. But at the point (0, 0) the Segre variety is the complex
line {w = 0}. In other words, X[0] = X, X[1] = {(0, 0)}, and X[2] = ∅.
Similarly, the Segre variety of a submanifold can be singular if the manifold is
CR singular. Let (z, w, ξ) be coordinates in C3 and consider X given by
w = z2 + ¯z2 + ξ2 + ¯ξ2.
(21)
The Segre subvariety at the origin Σ0X is the pair of complex lines given by {w =
0, (z + iξ)(z − iξ) = 0}.
Example 6.4. Consider Example 2.5, that is (x2 + y2)6 − s8x3(s − x) = 0, and
extend it to C2 using z = x + iy and w = s + it. In other words, we use X × R if
X is the variety of the previous example. That is, let X in (z, w) ∈ C2 be given by
f (z, w, ¯z, ¯w) = (z ¯z)6 − (Re w)8(Re z)3(Re w − Re z) = 0.
(22)
Similarly to Example 2.5, this f generates the ideal at I0(X), its derivatives vanish
when z = 0, but X is regular outside of {z = 0, Re w = 0}. So there are regular
(hypersurface, thus generic) points of X where the complexified f defines a singular
subvariety. That is, regular points of X where the corresponding X U is singular for
any neighborhood U of 0. For such a point q, for any U , ΣqX is regular, but ΣU
q X
is singular at q. In particular,
(23)
ΣqX (cid:2) (ΣU
q , q).
So ΣqX is just one component of the germ (ΣU
q , q).
Example 6.5. The set of Segre-degenerate points of a hypersurface need not be a
subvariety for noncoherent X. Let X ⊂ C3 in coordinates (z, w, ξ) ∈ C3 be given
by
z ¯z − (ξ + ¯ξ)w ¯w = 0.
(24)
The set is reminiscent of the Whitney umbrella. The set U = C3 is a good neighbor-
hood for X at 0. The set of Segre-degenerate points with respect to U (actually any
neighborhood U of the origin) is XU[3] = {w = z = 0}, that is, a one-dimensional
complex line. However, where Re ξ < 0, the variety X is locally just the line
{w = z = 0}. Therefore, the variety is a real manifold of dimension 2 (complex
manifold of dimension 1). At such points ΣpX is one-dimensional and such points
are not in X[3] (not Segre-degenerate). Hence,
(25)
X[3] = {(z, w, ξ) ∈ X : w = z = 0, Re ξ ≥ 0}
172
JI ˇR´I LEBL
and this set is not a subvariety, it is a semianalytic set.
Example 6.6. Let us construct the promised noncoherent hypersurface in C3
where the set X[n] of Segre-degenerate points is not complex, in fact, it is a real
line.
Let X ⊂ C3 in coordinates (z, w, ξ) ∈ C3 be given by
(26)
ψ = w2 ¯w2(Re ξ) + 4(Re z)(Re ξ)2w ¯w + 4(Re z)3z ¯z = 0.
The function is irreducible as a polynomial and homogeneous and thus (X, 0) is
irreducible as a germ of a real-analytic subvariety.
The set where dψ = 0 is given by Re z = 0, w = 0, and this set lies in X.
Therefore, {dψ = 0} ⊂ X is 3-real dimensional. However, the singular set Xsing is
2-dimensional given by Re z = 0, w = 0, and Re ξ = 0. Let us prove this fact. For
simplicity let z = x + iy and ξ = s + it and assume s (cid:6)= 0. Solve for w ¯w as
(cid:12)
(27)
w ¯w = x
s4 − sx(x2 + y2)
.
(cid:10)
−2s ± 2
s
(cid:11)
When the sign is negative and s (cid:6)= 0, we can solve for x by the implicit function
theorem and the subvariety has a regular point there. If the sign is positive and
s (cid:6)= 0, then we claim that there is no solution except x = 0, s = 0, w = 0. We
s4 − sx(x2 + y2) < 2s,
must check a few possibilities. If x > 0 and s > 0, then 2
s
and as w ¯w must be positive there are no such real solutions. Similarly for every
other sign combination. That means that the only solution when s (cid:6)= 0 is when the
sign is positive. So X is regular when Re ξ = s (cid:6)= 0. Similarly, it is not difficult to
show that X is singular at points where Re z = 0, w = 0, Re ξ = 0: For example,
at such points, were they regular, the Re z = 0 hyperplane and the Re z = − Re ξ
hyperplane would both have to be tangent as their intersections with X are singular
(both reducible). That is impossible for a regular point.
(cid:11)
Since ψ generates the ideal at the origin, it is easy to see that XU[n] = {z = 0, w =
0} near the origin for any good neighborhood U of the origin. As X[n] ⊂ XU[n] and
X[n] ⊂ Xsing, we can see that X[n] ⊂ {z = 0, w = 0, Re ξ = 0}. Since the defining
function does not depend on Im ξ, all the points of the set {z = 0, w = 0, Re ξ = 0}
are in X[n] or none of them are. The origin is definitely Segre-degenerate as ψ is
the generator of the ideal there, and thus X[n] = {z = 0, w = 0, Re ξ = 0}. So the
set X[n] where X is Segre-degenerate is of real dimension 1.
In other words:
(i) dim Xsing = 2.
(ii) {df = 0} ∩ X is 3 real-dimensional for every real-analytic germ f vanishing
on X (and not identically zero).
(iii) The set of Segre-degenerate points X[n] is a real one-dimensional line.
(iv) The set of Segre-degenerate points relative to U , XU[n], is a complex one-
dimensional line at the origin for every good neighborhood U of the origin,
and XU[n] ∩ Xreg (cid:6)= ∅.
Acknowledgments
The author would like to acknowledge Fabrizio Broglia for very insightful com-
ments and pointing out some missing hypotheses. The author would also like to
thank the anonymous referee and also Harold Boas for careful reading of the man-
uscript and for suggesting quite a few improvements to the exposition.
SEGRE-DEGENERATE POINTS FORM A SEMIANALYTIC SET
173
References
[1] Janusz Adamus, Serge Randriambololona, and Rasul Shafikov, Tameness of complex dimen-
sion in a real analytic set, Canad. J. Math. 65 (2013), no. 4, 721–739, DOI 10.4153/CJM-
2012-019-4. MR3071076
[2] M. Salah Baouendi, Peter Ebenfelt, and Linda Preiss Rothschild, Real submanifolds in com-
plex space and their mappings, Princeton Mathematical Series, vol. 47, Princeton University
Press, Princeton, NJ, 1999, DOI 10.1515/9781400883967. MR1668103
[3] Edward Bierstone and Pierre D. Milman, Semianalytic and subanalytic sets, Inst. Hautes
´Etudes Sci. Publ. Math. 67 (1988), 5–42. MR972342
[4] Jacek Bochnak, Michel Coste, and Marie-Fran¸coise Roy, Real algebraic geometry, Ergebnisse
der Mathematik und ihrer Grenzgebiete (3) [Results in Mathematics and Related Areas (3)],
vol. 36, Springer-Verlag, Berlin, 1998. Translated from the 1987 French original; Revised by
the authors, DOI 10.1007/978-3-662-03718-8. MR1659509
[5] Daniel Burns and Xianghong Gong, Singular Levi-flat real analytic hypersurfaces, Amer. J.
Math. 121 (1999), no. 1, 23–53. MR1704996
[6] Henri Cartan, Vari´et´es analytiques r´eelles et vari´et´es analytiques complexes (French), Bull.
Soc. Math. France 85 (1957), 77–99. MR94830
[7] Klas Diederich and John E. Fornaess, Pseudoconvex domains with real-analytic boundary,
Ann. of Math. (2) 107 (1978), no. 2, 371–384, DOI 10.2307/1971120. MR477153
[8] Klas Diederich and Emmanuel Mazzilli, Real and complex analytic sets. The relevance of
Segre varieties, Ann. Sc. Norm. Super. Pisa Cl. Sci. (5) 7 (2008), no. 3, 447–454. MR2466436
[9] Arturo Fern´andez-P´erez, Levi-flat hypersurfaces tangent to projective foliations, J. Geom.
Anal. 24 (2014), no. 4, 1959–1970, DOI 10.1007/s12220-013-9404-y. MR3261727
[10] Francesco Guaraldo, Patrizia Macr`ı, and Alessandro Tancredi, Topics on real analytic spaces,
Advanced Lectures in Mathematics, Friedr. Vieweg & Sohn, Braunschweig, 1986, DOI
10.1007/978-3-322-84243-5. MR1013362
[11] Jiˇr´ı Lebl, Singular set of a Levi-flat hypersurface is Levi-flat, Math. Ann. 355 (2013), no. 3,
1177–1199, DOI 10.1007/s00208-012-0821-1. MR3020158
[12] Stanislas (cid:2)Lojasiewicz, Ensembles semi-analytiques, Inst. Hautes ´Etudes Sci., Bures-sur-
Yvette, 1965. Available at http://perso.univ-rennes1.fr/michel.coste/Lojasiewicz.pdf.
[13] S. I. Pinchuk, R. G. Shafikov, and A. B. Sukhov, Dicritical singularities and laminar currents
on Levi-flat hypersurfaces (Russian, with Russian summary), Izv. Ross. Akad. Nauk Ser. Mat.
81 (2017), no. 5, 150–164, DOI 10.4213/im8582; English transl., Izv. Math. 81 (2017), no. 5,
1030–1043. MR3706863
[14] S. M. Webster, On the mapping problem for algebraic real hypersurfaces, Invent. Math. 43
(1977), no. 1, 53–68, DOI 10.1007/BF01390203. MR463482
[15] H. Whitney and F. Bruhat, Quelques propri´et´es fondamentales des ensembles analytiques-
r´eels (French), Comment. Math. Helv. 33 (1959), 132–160, DOI 10.1007/BF02565913.
MR102094
[16] Hassler Whitney, Complex analytic varieties, Addison-Wesley Publishing Co., Reading,
Mass.-London-Don Mills, Ont., 1972. MR0387634
Department of Mathematics, Oklahoma State University, Stillwater, Oklahoma
74078
Email address: [email protected]
| null |
10.1371_journal.pone.0255335.pdf
|
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
|
All relevant data are within the manuscript and its Supporting Information files. Data curation: Melanie Ricke-Hoch, Elisabeth Stelling, Martina Kasten, Thomas Gausepohl,
|
RESEARCH ARTICLE
Impaired immune response mediated by
prostaglandin E2 promotes severe COVID-19
disease
1‡*, Elisabeth Stelling1‡, Lisa Lasswitz2, Antonia P. Gunesch2,3,4,
2,5,6, Thomas Pietschmann2,3, Virginie Montiel7, Jean-Luc Balligand7,
Melanie Ricke-HochID
Martina Kasten1, Francisco J. Zapatero-Belincho´ n2,5, Graham Brogden2,
Gisa GeroldID
Federica Facciotti8, Emilio Hirsch9, Thomas GausepohlID
F. Rimmelzwaan10, Anne Ho¨ fer11,12, Mark P. Ku¨ hnel11,12, Danny Jonigk11,12,
Julian Eigendorf13, Uwe Tegtbur13, Lena Mink13, Michaela Scherr14, Thomas Illig15,
Axel Schambach16,17, Tobias J. Pfeffer1, Andres Hilfiker18, Axel Haverich18,
Denise Hilfiker-Kleiner1,19
1, Husni ElbaheshID
10, Guus
1 Department of Cardiology and Angiology, Hannover Medical School, Hanover, Germany, 2 Institute of
Experimental Virology, TWINCORE, Center for Experimental and Clinical Infection Research Hannover,
Hanover, Germany, 3 German Center for Infection Research, Hanover-Braunschweig Site, Braunschweig,
Germany, 4 Department of Gastroenterology, Hepatology and Endocrinology, Hannover Medical School,
Hanover, Germany, 5 Department of Clinical Microbiology, Virology & Wallenberg Centre for Molecular
Medicine (WCMM), Umeå University, Umeå, Sweden, 6 Department of Biochemistry, University of
Veterinary Medicine Hannover, Hanover Germany, 7 Pole of Pharmacology and Therapeutics, Institut de
Recherche Expe´ rimentale et Clinique, and Cliniques Universitaires Saint-Luc, Universite´ catholique de
Louvain (UCLouvain), Brussels, Belgium, 8 Department of Experimental Oncology, European Institute of
Oncology IRCCS, Milan, Italy, 9 Department of Molecular Biotechnology and Health Sciences, Molecular
Biotechnology Center, University of Torino, Torino, Italy, 10 Research Center for Emerging Infections and
Zoonoses (RIZ), University of Veterinary Medicine in Hannover (TiHo), Hannover, Germany, 11 Biomedical
Research in Endstage and Obstructive Lung Disease (BREATH), German Center for Lung Research,
Hanover, Germany, 12 Institute for Pathology, Hannover Medical School, Hanover, Germany, 13 Institute of
Sports Medicine, Hannover Medical School, Hanover, Germany, 14 Department of Hematology, Hemostasis,
Oncology and Stem Cell Transplantation, Hannover Medical School, Hanover, Germany, 15 Hannover
Unified Biobank (HUB), Hannover Medical School, Hanover, Germany, 16 Institute of Experimental
Hematology, Hannover Medical School, Hanover, Germany, 17 Division of Hematology and Oncology,
Boston Children’s Hospital, Harvard Medical School, Boston, MA, United States of America, 18 Department
of Cardiac, Thoracic, Transplantation and Vascular Surgery, Hannover Medical School, Hanover, Germany,
19 Department of Cardiovascular Complications of Oncologic Therapies, Medical Faculty of the Philipps
University Marburg, Marburg, Germany
‡ SIGR and MK contributed equally to this work as first co-authors.
* [email protected]
Abstract
The SARS-CoV-2 coronavirus has led to a pandemic with millions of people affected. The
present study finds that risk-factors for severe COVID-19 disease courses, i.e. male sex,
older age and sedentary life style are associated with higher prostaglandin E2 (PGE2)
serum levels in blood samples from unaffected subjects. In COVID-19 patients, PGE2 blood
levels are markedly elevated and correlate positively with disease severity. SARS-CoV-2
induces PGE2 generation and secretion in infected lung epithelial cells by upregulating
cyclo-oxygenase (COX)-2 and reducing the PG-degrading enzyme 15-hydroxyprostaglan-
din-dehydrogenase. Also living human precision cut lung slices (PCLS) infected with SARS-
CoV-2 display upregulated COX-2. Regular exercise in aged individuals lowers PGE2
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Ricke-Hoch M, Stelling E, Lasswitz L,
Gunesch AP, Kasten M, Zapatero-Belincho´n FJ, et
al. (2021) Impaired immune response mediated by
prostaglandin E2 promotes severe COVID-19
disease. PLoS ONE 16(8): e0255335. https://doi.
org/10.1371/journal.pone.0255335
Editor: Paulo Lee Ho, Instituto Butantan, BRAZIL
Received: March 11, 2021
Accepted: July 14, 2021
Published: August 4, 2021
Copyright: © 2021 Ricke-Hoch et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: This work was supported by: The
German Research Foundation (DFG, HI 842/3-2 to
D.H.-K.), by the DFG Clinical Research Group (DFG
KFO311, HI 842/10-1, HI 842/10-2 to D.H.-K.; RI
2531/2-1, RI 2531/2-2 to M.R.-H.), by REBIRTH I/
II to D.H.-K., by the Foundation Leducq (Project ID
19CVD02) to D.H.-K. and E.H., DFG as part of the
German Strategy for Excellence (EXC 2155
“RESIST”, Project ID 39087428 to D.J.), The
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
1 / 24
PLOS ONE DEFEAT PANDEMIcs (AP6-9, to D.J. and M.P.K.),
by the (DFG – Projektnummer 158989968 - SFB
900 project C7 and DFG project GE 2145/3-2 to G.
G.), the ‘Niedersa¨chsischen Vorab’ program
(project 76251-99-3/19 to G.G.) through the
Ministry of Lower Saxony (MWK) and the
Volkswagen Foundation (Volkswagen Stiftung), by
the Federal Ministry of education and research
(project COVID-Protect, Project: 01KI20143C to G.
G.), the Knut and Alice Wallenberg Foundation and
the Federal Ministry of Education and Research
together with the the Ministry for Science and
Culture (MWG) through the ‘Professorinnen
Programm III’ to G.G., by the European Research
Council Consolidator Grant (XHale; 771883 to D.
J.), by REBIRTH I/II and REBIRTH Center for
Regenerative Translational Medicine (MWK, project
ZN3440) to A.S., by Cariplo Foundation (Project
#2018-0498 to E.H.), by MWG project 14-76103-
184 CORONA-1/20 to T.I. and by the European
Virus Archive GLOBAL (EVA-GLOBAL) project
funded by the European Union’s Horizon 2020
research and innovation program under grant
agreement No 871029 (to Christian Drosten). This
work was partly supported by the Alexander von
Humboldt Foundation in the framework of the
Alexander von Humboldt Professorship endowed
by the German Federal Ministry of Education and
Research and by funding from the Ministry for
Science and Culture (MWK), Lower Saxony,
Germany (14 - 76103-184 CORONA-15/20 to G.F.
R.). A.P.G. was supported by the Deutsches
Zentrum fu¨r Infektionsforschung (DZIF; German
Center for Infection Research; Grant No. TTU
05.816 00 to T.P.). Work by J.L.B. was supported
by grants from Fonds National de la Recherche
Scientifique (FNRS) and WEBIO. J.L.B is an
established investigator of the WELBIO institute.
The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Impaired immune response promotes severe COVID-19 disease
serum levels, which leads to increased Paired-Box-Protein-Pax-5 (PAX5) expression, a
master regulator of B-cell survival, proliferation and differentiation also towards long lived
memory B-cells, in human pre-B-cell lines. Moreover, PGE2 levels in serum of COVID-19
patients lowers the expression of PAX5 in human pre-B-cell lines. The PGE2 inhibitor Taxi-
folin reduces SARS-CoV-2-induced PGE2 production. In conclusion, SARS-CoV-2, male
sex, old age, and sedentary life style increase PGE2 levels, which may reduce the early
anti-viral defense as well as the development of immunity promoting severe disease
courses and multiple infections. Regular exercise and Taxifolin treatment may reduce these
risks and prevent severe disease courses.
Introduction
The 2019 strain of coronavirus (severe acute respiratory syndrome coronavirus-2 SARS-CoV-
2) caused a pandemic with COVID-19 disease affecting millions of people worldwide. Patients
with serious disease courses frequently present with severe acute respiratory syndrome that
can progress to pneumonia and acute respiratory distress syndrome and shock [1–3]. Systemic
inflammation, acute cardiac injury, heart failure, and hypercoagulability are critical complica-
tions in COVID-19 disease [1, 4–9]. Identified cell types infected with SARS-CoV-2 include
pulmonary epithelial cells, renal cells, cardiomyocytes, endothelial cells and pericytes [10–12].
An increased risk for infection and severe disease courses have been found in association
with older age, male sex, cardiovascular comorbidities and air pollution [7, 13–15]. Immuno-
thrombosis integrates innate immunity, activation of platelets, and clotting factors to fight
invading pathogens and concurrently promotes inflammation-related tissue damage; in the
context of COVID-19 disease, this may explain the systemic hypercoagulability frequently
present in COVID-19 patients [8]. Further alterations in the immune system with partially
opposing mechanisms have been reported in acute and chronic COVID-19 disease. On one
hand, COVID-19 infection appears associated with an upregulation and activation of neutro-
phils while at the same time lymphocytes are diminished [16]. Reduced lymphocyte popula-
tions seem to correlate with more severe organ injury and higher mortality in hospitalized
COVID-19 patients [16]. In this regard, T-cell exhaustion [3, 17], reduced circulating and resi-
dent B-cell population and loss of germinal centers associated with viral persistence and severe
disease courses correlate with high mortality in the acute phase [3, 18, 19]. On the other hand,
a growing body of clinical data suggests that a cytokine storm is associated with COVID-19
severity and is also a crucial cause of death from COVID-19 [20–22]. Among potential mecha-
nisms, SARS-CoV-2 induced formation of autoantibodies, tissue and organ injury as well as
secondary infection with bacteria and fungi [23, 24].
Prostaglandin (PG) E2, a metabolite of arachidonic acid, is a well-known modulator of viral
infection [25]. As such, PGE2 suppresses the adaptive and innate immune systems and pro-
motes infection, e.g., by influenza A virus (IAV) [26, 27]. Moreover, increased circulating
PGE2 levels have been associated with reduced immunity in response to IAV vaccination [26,
27]. Interestingly, IAV infection also promotes the production of PGE2 [28]. Cyclooxygenase-
2 (COX-2) is a rate-limiting enzyme for the generation of PGE2 and Hydroxyprostaglandin
Dehydrogenase 15-(NAD) (HPGD) is an enzyme responsible for the degradation of PGE2
[29]. These findings, supported further by a recent literature review [30] naturally suggested a
connection between arachidonic acid metabolism and PGE2 in COVID-19 disease.
We hypothesized that PGE2 modulates the immune response in individuals at risk for
severe COVID-19 disease. To test this, we first measured serum PGE2 levels in COVID-19
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
2 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
patients with different levels of disease severity, as well as in subjects with putative risk factors
(age, sex, physical fitness) for a severe disease course. To analyze the direct effects of SARS-
CoV-2 on PGE2 production, we infected human lung epithelial cells and human precision-
cut-lung-slices (PCLS) with SARS-CoV-2. Additionally, we further dissected the mechanisms
of PGE2 modulation of immune defense, e.g. through B-cell maturation and the formation of
memory cells, and correlated disease severity with lung B-cell content in patient samples. We
further tested strategies to reduce PGE2 production or the effect on the above parameters as
preventive or therapeutic modalities against severe COVID-19.
Materials and methods
Unless otherwise stated, chemicals and reagents were all purchased from Sigma-Aldrich.
Study design
COVID-19 study.
In this study of 89 patients diagnosed with COVID-19, 41 presented
with mild/moderate symptoms and 48 were hospitalized with severe disease. Blood samples
were also obtained from male (n = 18) and female subjects (n = 28) (age 18–50 years) from a
healthy population established by Hannover Unified Biobank (HUB).
At the time of blood sampling, for 29 patients it was known whether they obtained corti-
coids or not. Among those n = 14 obtained no corticoids and n = 15 COVID-19 patients with
mild and severe disease course received corticoids (Dexamethasone n = 11 or Medrol n = 4).
Information on the use of NSAIDs or leukotriene modifiers were not available. None of the
healthy controls were under corticoids or nonsteroidal anti-inflammatory drugs (NSAIDs)
treatment.
The local ethics committees at Hannover Medical School, Comite´ d’Ethique Hospitalo-
Facultaire of UCLouvain, and the Ethical Committee of IEO has been obtained (IEO1271)
approved this study. All patients and healthy control subjects provided written informed con-
sent. The study conforms to the principles outlined in the Declaration of Helsinki.
Physical assessment and exercise program in healthy elderly individuals (rebirth 60plus
cohort, DRKS00013885). All subjects in the Rebirth 60plus cohort (DRKS00013885) were
initially tested for maximum power output on a cycle ergometer with graded exercise test
(GXT). Based on their activities, physical fitness and pathologies, each subject was given an
aerobic exercise training program. Once a month, the subjects were contacted by phone to
assess training progress and adjust the exercise program, if necessary. All subjects of the
Rebirth 60plus study were informed about benefits and risks regarding all study procedures.
Height and weight were measured using a scale (seca gmbh & co. kg, Hamburg, Germany).
Body fat was measured with a medical Body Composition Analyzer mBCA (seca gmbh & co.
kg, Hamburg, Germany). The physical activity was tracked using a GPS watch Forerunner 30
(Garmin Deutschland GmbH, Munich, Germany) and a daily diary where all physical activi-
ties were additionally documented. All study procedures were approved by the local ethics
committee of Hannover Medical School (Vote #7617) and all subjects provided informed writ-
ten consent prior to the commencement of the study procedures.
Blood sampling and blood tests
Blood samples were collected in S-Monovette1 tubes containing ethylenediaminetetraacetic
acid (EDTA, for plasma) or clot activator (for serum) at the time of hospital admission or at
study inclusion (baseline, BL) and at the follow-up (FU) visits after 12 months for the Rebirth
60Plus male and female subjects (age >60 years). Blood samples were also obtained from
young male and female subjects (age 18–50 years) from a healthy population established by
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
3 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Hannover Unified Biobank (HUB). Plasma or serum was separated by centrifugation at 1500
rpm for 10 min and aliquots were stored at -80˚C. Laboratory workup was performed as part
of routine analysis by hospital laboratories for leukocytes, neutrophils, lymphocytes, platelets,
CRP and LDH. PGE2 serum and plasma levels were measured using the prostaglandin E2
ELISA kit (abcam ab133021) according to the manufacturer’s protocol.
Infection of Calu-3 cells with SARS-CoV-2 and Taxifolin treatment
Calu-3 cells (kindly provided by Prof. Po¨hlmann, German Primate Center, Go¨ttingen; ATCC
Cat# HTB-55; RRID:CVCL_0609) were maintained in Dulbecco’s’ modified Eagle medium
and Vero cells (ATCC-CCL-81; Lot 58484194) in Advanced MEM at 37˚C and 5% CO2. Both
media were supplemented with 10% fetal bovine serum, 2 mM glutamine, 0.1 mM non-essen-
tial amino acids and 1% Penicillin/Streptomycin. Calu-3 cells (4.5x105 cells/well) were seeded
in collagen-coated 24-well plates. For infection, the SARS-CoV-2 (strain SARS-CoV-2/Mu¨n-
chen-1.2/2020/984,p3) [31] kindly provided by Christian Drosten (Charite´, Berlin) through
the European Virus Archive–Global (EVAg) was used. The isolate was propagated and titrated
in Vero cells. Calu-3 cells were pretreated with 100 μM Taxifolin or DMSO (0.15%) for 24 h.
Infection with SARS-CoV-2 isolate was performed at a multiplicity of infection (MOI) of
2.0x10-5 for 4 h at 37˚C in the presence of the compounds. Heat-inactivated virus (15 min,
70˚C) served a negative control. After infection, cells were washed twice with PBS before the
medium containing the respective compound was added. At 48 h post infection, culture super-
natant was collected and heat-inactivated (15 min, 70˚C) prior to the detection of PGE2. RNA
was isolated from cell lysates using a NucleoSpin RNA kit (Macherey-Nagel) according to the
manufacturer’s instructions to analyze virus genome copy numbers, COX-2, HPGD, PTGES2,
PTGES3, TNFa and IFNg expression.
Virus titration in Vero E6 cells for infection of lung slices with SARS-CoV-
2
Vero E6 (ATCC CRL-1586) and Vero cells (ATCC CCL-81) were maintained in Eagle’s Mini-
mum Essential Medium (EMEM) (Lonza) supplemented with 25 mM of HEPES (Gibco),
1 × GlutaMAX (Gibco), 100 U/ml penicillin and 100 μg/ml streptomycin. SARS-CoV2 isolate
(strain SARS-CoV-2/Mu¨nchen-1.2/2020/984,p3) [31] was kindly provided by Christian Dros-
ten. SARS-CoV-2 seed stocks were generated by inoculating Vero E6 (ATCC CRL-1586) at a
MOI of 0.001, collecting and aliqouting the culture supernatant at 72 h post infection (hpi),
then storing at -80˚C in aliquots. SARS-CoV-2 working stocks were generated by an additional
passage on Vero cells (ATCC CCL-81) at a MOI of 0.001. Plaque and median tissue culture
infectious dose (TCID50) assays were performed to titrate the cultured virus after both passages
using Vero cells. This stock was used for the ex vivo infections of human tissues.
Infections of precision-cut human lung slices (PCLS) with SARS-CoV-2
PCLS were maintained in DMEM/F12 medium (Gibco, Thermo Fisher Scientific) supple-
mented with 2 mM of HEPES (Gibco), 1 × GlutaMAX (Gibco), 100 U/ml penicillin and
100 μg/ml streptomycin; this media was also used for virus dilutions and post-infection incu-
bation. On the day of infection, PCLS were rinsed with PBS (without Mg2+ and Ca2+) then
inoculated with 1 × 105 PFU SARS-CoV-2 in 250 μl of media per well in 48-well plates and
incubated at 37˚C. After 2 h, the inoculum was removed and the PCLS were then cultured in
250 μl of DMEM/F12 medium. At 72 and 120 hpi, supernatants were collected and PCLS were
fixed with fixation buffer (4% PFA, 0.1% glutaraldehyde and 200 mM HEPES in ddH2O) for 1
h at room temperature followed by 24 h at 4˚C.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
4 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
QRT-PCR for NSP7 to confirm SARS-CoV-2 infection
SARS-CoV-2 infections in human Calu-3 cells and human lung slices and tissue were verified
by NSP7 mRNA expression using qRT-PCR (forward primer: GGG CTC AAT GTG TCC
AGT TAC, reverse primer: TTG CCC TGT CCA GCA TT).
Human lung biopsies from acute COVID-19 patients
Patients with acute COVID-19 (AC, n = 6) have been diagnosed with COVID-19 and were
positively tested via PCR as described [4]. All AC patients used in this study showed typical
acute respiratory distress syndrome (ARDS) histopathology typical for COVID-19 disease. In
addition, NSP7 expression was used to detect SARS-CoV-2 virus in biopsies with the limita-
tion that due to heterogeneous distribution of the virus or already cleared acute infection, PCR
is not always positive in every area of the lungs and therefore NSP7 might be not detected.
Multiplex immunohistochemistry of human lung biopsies
The FFPE sections for each group (Control (Ctrl) n = 3, acute COVID-19 (AC) n = 6, trans-
plant rejected (TR) n = 4) were representatively stained with the manual Opal 7-Color IHC Kit
(Akoya Biosciences, Marlborough, MA) as previously described [32]. The primary antibodies
CD4 (Cytomed SP35, 1:50), CD8 (Dako M0755, 1:600), CD68 (Dako PGM1, 1:750) and CD20
(Dako M0755, 1:1000) were combined in sequence with the opal fluorophore CD4-Opal520,
CD8-Opal570, CD20-Opal540 and CD68-Opal650. The sections were scanned with the Vectra
3 System (Akoya Biosciences, Marlborough, MA). The Regions of Interest (ROIs) were
selected representative for small, medium and large vessels for the entire tissue section. The
number of analyzed stamps was 43 for Ctrl, 74 for AC and 56 for TR. For the detection of
CD20+ B cells, the analysis was performed with the inForm Advanced Image Analysis Software
Version 2.3.0 (Akoya Biosciences, Marlborough, MA) and ImageJ 1.53c (Wayne Rasband,
National Institutes of Health, USA). Statistical analysis was performed using the generalized
linear model with Gaussian distribution and weights adjusted according to the number of
ROIs per patient.
Stimulation of human pre-B-cell lines
Human pre-B-cell lines 697 (ACC42 DSMZ collection) and SUP-B15 (ACC389 DSMZ collec-
tion) were cultivated in RPMI (Gibco) supplemented with 10% FBS. 5x105 cells per ml were
pre-incubated with either the EP1/EP2 receptor antagonist AH6809 (10 μM, Tocris) or the
EP4 receptor antagonist GW627368 (10 μM, Tocris) for 2 h. PGE2 (10 μM, Sigma-Aldrich)
was added and cells were harvested after 48 h in TRIzol, or stained with trypan blue (Bio-Rad
laboratories) and counted for measuring live to dead ratio and cell numbers using the TC20
automated cell counter (Bio-Rad laboratories). Control cells were incubated with dissolvents
(DMSO or ethanol (ETHO), 1 μL/ml media). Alternatively, 5x105 per ml 697 and SUP-B15
cells were incubated with 10% human serum from older individuals (>60 y) prior to the com-
mencement of the exercise program at baseline (BL) and after 12M (12M FU) for 48 h and har-
vested in TRIzol. SUP-B15 cells were incubated with 10% human serum from COVID-19
patients and from healthy controls. Cells were harvested after 48 h in TRIzol.
PGE2 and prostaglandin D2 (PGD2) detection in supernatants of Calu-3
PGE2 and PGD2 levels in the supernatants of the cell lines Calu-3 (normalized to total RNA
content) were measured using the prostaglandin E2 ELISA kit (abcam ab133021) or the
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
5 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
prostaglandin D2 ELISA kit (Cayman Chemicals, No. 512031) respectively, according to the
manufacturer’s protocols.
Isolation of RNA and qRT-PCR
Total RNA was isolated with TRIzol (Thermo Fisher Scientific) and cDNA synthesis was per-
formed as described previously [33]. Real-time PCR with the SYBR green dye method (Bril-
liant SYBR Green Mastermix-Kit, Thermo Fisher Scientific) was performed with the AriaMx
Real-Time PCR System (Agilent Technologies) as described [33]. Expression of mRNA levels
was normalized using the 2-ΔΔCT method relative to 18S, beta-2-microglobulin (B2M) and
glyceraldehyde-3-phosphate dehydrogenase (GAPDH). A list of qRT-PCR primers used in
this study is provided in the supplements file S1 Table.
RNA isolation from formalin fixed and paraffin embedded tissue
RNA isolation from formalin-fixed and paraffin embedded tissue was performed using the
Maxwell1 RSC RNA FFPE Purification Kit (Promega Corporation, Madison, WI). RNA con-
tent was measured by using the Qubit RNA IQ Assay (Thermo Fisher Scientific, Waltham,
MA).
Statistical analyses
Statistical analysis was performed using GraphPad Prism version 5.0a, 7.0 and 8.1.2 for Mac
OS X (GraphPad Software, San Diego, CA, USA).
Normal distribution was tested using the D’Agostino normality test or Shapiro-Wilk nor-
mality test if the sample was too small for D’Agostino normality test. Continuous data were
expressed as mean ± SD or median and interquartile range (IQR), according to the normality
of distribution. Comparison between two groups was performed using one sample t-test or
unpaired two-tailed t-test for Gaussian distributed data and the Mann-Whitney-U test where
at least one column was not normally distributed. When comparing more than two groups,
ANOVA and Bonferroni’s post hoc test or Dunnett’s post hoc test were used according to the
normality of distribution. Categorical variables are presented as frequencies (percentages) and
compared using Fisher’s exact test. A two-tailed P value of <0.05 was considered statistically
significant. Correlation for BMI, BW, body fat content and age was analyzed via ozone correla-
tion analysis by using Pearson correlation coefficients for Gaussian distributions or for non-
parametric Spearman correlation coefficients for non-normal distribution.
Results
PGE2 levels in healthy individuals in relation to sex and age
In healthy control individuals aged <50, circulating PGE2 levels were higher (P>0.01) in men
than in women (Fig 1A). Sex-related differences in circulating PGE2 levels were not observed
in older (<60 years) healthy individuals (Fig 1B). Circulating PGE2 levels were markedly
higher in older (>60 years) healthy males and females than in respective sex-matched younger
(<50 years) individuals (Fig 1C and 1D). Both males and females showed a significant positive
correlation of circulating PGE2 levels with age (Fig 1E and 1F), while no correlation with BMI,
body weight (BW) or body fat content was observed (S2 Table, S1 Fig). Controlled physical
exercise for 12 months reduced PGE2 in elderly male and female individuals compared with
their baseline (BL) levels (Fig 1G and 1H and S2 Table). With these indications, we next set to
explore whether PGE2 levels changes in COVID-19 and whether differences in PGE2 levels
could explain severe disease courses after SARS-CoV-2 infection.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
6 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 1. (A) The dot plots summarize circulating serum PGE2 levels (pg/ml) of males (n = 18) and females (n = 28)
below the age of 50 years. (B) Dot plots summarize circulating serum PGE2 levels (pg/ml) of males (n = 40) and
females (n = 46) over the age of 60 years. (C) The dot plots summarize circulating serum PGE2 levels (pg/ml) of males
(n = 18) <50y and males (n = 40) >60y. (D) Dot plots summarize circulating serum PGE2 levels (pg/ml) of females
(n = 28) <50y and females (n = 46) >60y. Ozone correlation analysis of serum PGE2 levels with age in (E) males
(n = 66, Spearman r: 0.2564, P-value: 0.0377) and (F) females (n = 76, Spearman r: 0.638, P-value: <0.0001).
Circulating serum PGE2 levels at baseline (BL) and after 12-months follow-up (FU) following controlled physical
training from (G) males (n = 31) and (H) females (n = 37). (A, B, D, G, H) Data are presented as median±IQR,
��P<0.01, ���P<0.001, ����P<0.0001, Mann-Whitney-U test. (C) Data are presented as mean±SD, ��P<0.01,
unpaired two-tailed t-test. (E, F) Ozone correlation, Spearman correlation coefficients, two-tailed P value. Underlying
data can be found in S1 Data.
https://doi.org/10.1371/journal.pone.0255335.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
7 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Circulating levels of PGE2 in COVID-19 patients and age-matched healthy
controls
We analyzed PGE2 levels in individuals with mild/moderate (n = 41) and severe (n = 48)
COVID-19 disease from hospitals in Hanover (Germany), Milan (Italy) and Brussels (Bel-
gium) and in age-matched healthy controls (n = 31) (Table 1, S3 Table). Clinical data and labo-
ratory characteristics of the COVID-19 patients revealed that the more severely affected
patients were significantly older with a higher proportion of males than females compared
with the mildly/moderately affected group (Table 1). BMI and diabetes rate are increased in
the entire COVID-19 cohort with no significant difference between the mild/moderate and
the severe groups (Table 1). In addition, C reactive protein (CRP) was elevated, while the total
leukocyte- and neutrophil counts were within the normal range, although some patients dis-
played markedly increased levels (Table 1). The mean lymphocyte counts (T- and B-cells) were
reduced in the majority of COVID-19 patients and were specifically low in patients with severe
disease courses (Table 1). Platelets were in the normal range in all COVID-19 patient groups
and lactate dehydrogenase (LDH) was increased and highest in the severely affected patients
(Table 1). Mortality was 15% for the entire cohort with no patient deaths in the mild/moderate
group and 27% of patients dying in the severe disease group who were all of male sex
(Table 1). Circulating PGE2 levels were increased in COVID-19 patients at the time of hospital
admission compared with healthy controls, and PGE2 levels were significantly higher in the
severely affected patients compared with mildly/moderately affected patients (Fig 2A–2F,
Table 1). A direct relationship of PGE2 levels to death events was not observed (Fig 2D–2F).
COVID-19 patients who need hospitalization were defined as severe COVID-19 patients.
Body mass index (BMI), C-reactive protein (CRP), lactate dehydrogenase (LDH), leukocytes
Table 1. Summary of clinical data of the COVID-19 patients.
Parameters
COVID-19 patients total
Mild to moderate COVID-19 disease
Severe COVID-19 disease
(N = 89)
59 (46–68)
30% (27/89)
79.5 (67.75–96.5)
(n = 42)
171.9±9.4
(n = 43)
27.2 (23.7–30.2)
(n = 43)
28% (11/50)
8036±5831
(n = 50)
4999±2697
(n = 35)
1114±564
(n = 36)
107±83
(n = 50)
Age (years, median ± IQR)
Sex female (%)
Body weight (kg, median ± IQR)
Body height (cm, mean ± SD)
BMI (median ± IQR)
Diabetes (%)
Total leucocytes, counts/μl (mean ± SD)
Standard value: 3900–10200 counts/μl
Neutrophils, counts/μl (mean ± SD)
Standard value: 1500–7700 counts/μl
Lymphocytes, counts/μl (mean ± SD)
Standard value: 1100–4500 counts/μl
CRP mg/L (mean ± SD)
Standard value: <5 mg/L
LDH at hospitalization UI/L (median ± IQR)
Standard value: <248 UI/L
Platelets at hospitalization 103/μl (mean ± SD)
Standard value: 160–370 103/μl
Mortality (%)
https://doi.org/10.1371/journal.pone.0255335.t001
(N = 41)
51 (40–67)
44% (18/41)
75 (65.5–86.5)
(n = 25)
170±9.28
(n = 25)
(N = 48)
62 (51–68.75)�
19% (9/48)�
90 (76–100)�
(n = 17)
174.6±9.1
(n = 18)
26.56 (22.96–28.9)
28 (25.5–31.95)
(n = 26)
19% (5/26)
6803±3213
(n = 26)
4664±2614
(n = 21)
1266±608
(n = 21)
69.4±55.95
(n = 26)
(n = 17)
25% (6/24)
9373±7592
(n = 24)
5502±2839
(n = 14)
902±428
(n = 15)
147.1±89���
(n = 24)
459 (348–659)���
(n = 23)
221±83
(n = 24)
27% (13/48)���
363.5 (263.8–518.8)
299 (229–375)
(n = 48)
229±81
(n = 50)
15% (13/89)
(n = 25)
235±80
(n = 26)
0% (0/41)
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
8 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 2. Circulating PGE2 levels are elevated in COVID-19 patients. The dot plots summarize circulating serum PGE2
levels (pg/ml) of (A) COVID-19 patients (n = 29) and healthy controls (n = 31), and separately (B) for males (COVID-
19 male patients n = 19; healthy male controls n = 14) and (C) for females (COVID-19 female patients n = 10; healthy
female controls n = 17). The dot plots summarize relative circulating plasma PGE2 levels (in %) (D) of patients with
severe (n = 36) and mild (n = 24) disease, and separately (E) for males (severe affected males n = 29; mild affected
males n = 14) and (F) for females (severe affected females n = 7; mild affected females n = 10); the median of patients
with mild disease was set at 100%. Dots representing patients who died with COVID-19 disease are highlighted in red.
(A, B, D-F) Data are presented as median±IQR, �P<0.05, ���P<0.001, ����P<0.0001, Mann-Whitney-U test. (C) Data
are presented as mean±SD, �P<0.05, unpaired two-tailed t-test. Underlying data can be found in S1 Data.
https://doi.org/10.1371/journal.pone.0255335.g002
normal count, neutrophils normal count, and lymphocytes below normal counts, were ana-
lyzed at the time of hospital admission in routine clinical lab tests. Standard values of blood
parameters were indicated in the parameter column. Values outside the normal range were
indicated in bold font. Comparison between the groups of mild and severe COVID-19 was
performed using Student’s t-test for Gaussian distributed data (presented as mean ± SD) and
the Mann-Whitney-U test where at least one column was not normally distributed (presented
as median and interquartile range (IQR)). Categorical variables are presented as frequencies
(percentages) and were compared using Fisher’s exact test. �P<0.05, ��P<0.01, ���P<0.001
severe COVID-19 vs mild to moderate COVID-19 disease. Underlying data can be found in
S1 Data.
Expression of COX-2 and HPGD and secretion of PGE2 in human lung
epithelial cells and precision-cut lung slices infected with SARS-CoV-2
Next, we investigated whether SARS-CoV-2 would enhance PGE2 production in infected host
cells. Human lung epithelial cells (Calu-3 cells) were infected with SARS-CoV-2 (strain SARS-
CoV-2/Mu¨nchen-1.2/2020/984,p3) [31] and infection was confirmed with qRT-PCR for the
SARS-CoV-2 gene encoding nonstructural protein (NSP)7 [34] (Fig 3A). Heat-inactivation of
SARS-CoV-2 infected supernatants of Calu-3 cells was not associated with degradation of
PGE2 (S2A Fig). Infected cells displayed increased secretion of PGE2, which was specifically
prevented by incubation with the PGE2 inhibitor Taxifolin [35, 36] (Fig 3B). The synthesis of
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
9 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
other prostaglandins like PGD2 was not altered by Taxifolin in infected Calu-3 cells (S3A Fig).
Moreover, Taxifolin treatment was not associated with changes in the proliferation capacity of
Calu-3 cells (S3B–S3D Fig).
SARS-CoV-2 infection increased the expression of COX-2 and reduced the expression of
the PGE2 degrading enzyme HPGD but did not alter the expression of the PGE synthase
(PTGES) in Calu-3 cells (Fig 3C–3E). In contrast, the expression of PGE synthase 2 (PTGES2)
and PGE synthase 3 (PTGES3) were significantly reduced by SARS-CoV-2 in Calu-3 lung cells
(Fig 3F and 3G). In line with these results, the production of PGD2 was also increased in
infected Calu-3 cells (S3E Fig). Additionally, SARS-CoV-2 infection markedly induced the
expression of TNFα (644-fold; P<0.05, S2B Fig), which is known to induce COX-2 expression
and with this the PGE2 production in human fibroblasts [37]. The expression of IFNγ could
not be detected in control or in SARS-CoV-2 infected in human Calu-3 lung cells. Also, the ex
vivo infection of living human PCLS with SARS-CoV-2 (viral infection analyzed by NSP7
qRT-PCR, Fig 3H) led to an upregulation of COX-2 expression compared with non-infected
control slices, while HPGD mRNA levels were unchanged and PGE synthase (PTGES) expres-
sion tended to be increased (Fig 3I–3K).
Effect of PGE2 on the expression of pre-B-cell differentiation and survival
factor PAX5 in human pre-B-cells
PGE2 is known to attenuate the proliferation, differentiation and survival of B-cells [38, 39].
Here, we observed that the addition of PGE2 (10 μM, i.e. 3525 pg/ml), in the range measured
in COVID-19 patients’ sera (1300 to >20.000 pg/ml), to two human B-cell precursor lines, 697
and SUP-B15, significantly reduced PAX5 mRNA expression (Fig 4A and 4B). The effect of
PGE2 on PAX5 in 697 and SUP-B15 cells could be blocked by co-treatment with the PGE2
receptor 4 (EP4; PTGER4) antagonist, GW627368 but not with the EP2 receptor antagonist,
AH6809 (Fig 4A). The expression of PTGER4 in 697 and SUP-B15. Cells was confirmed by
qRT-PCR (S4A and S4B Fig). Additionally, PGE2 (10 μM) stimulation was associated with a
reduced 697 cell number (51%) compared to control (100%, p<0.01) treated 697 cells. The
ratio of live to dead pre-B-cells was not altered through PGE2 stimulation indicating that the
decrease in pre-B-cell number is not mediated by enhanced cell death (S4C Fig). However,
PGE2 stimulation was associated with a reduced expression of the proliferationmarkers Ki67,
TOP2A and TPX2 (S4D–S4F Fig) indicating that it reduces the proliferation capacity of pre-B
cells.
Effect of PGE2 on the expression of inflammatory cytokines TNFα and
IFNγ in human pre-B-cells
During SARS-CoV-2 infection upregulation of PANoptosis inducing cytokines, i.e. TNFα and
IFNγ have been reported in immune cells [40]. Here, PGE2 stimulation reduced the expression
of TNFα in both pre-B cell lines 697 and SUP-B15 (S5A and S5B Fig). The expression of IFNγ
was not changed in 697 cells and in SUP-B15 cells, PGE2 reduced its expression (S5C and S5D
Fig).
Effect of serum from elderly individuals before/after physical exercise on
PAX5 expression in human pre-B-cells
PAX5 expression was higher in 697 and SUP-B15 pre-B-cells incubated with serum from
elderly individuals collected after 12 months of controlled physical exercise compared with
their serum before exercise (Fig 4C and 4D and S2 Table). In addition, the EP4 antagonist,
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
10 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 3. SARS-CoV-2 infection modulates PGE2 secretion and COX-2 and HPGD expression. (A) Representative gel
image of NSP7 mRNA expression of Calu-3 cells infected with SARS-CoV-2 and control cells. (B) The bar graph
summarizes PGE2 content in supernatants of Calu-3 cells infected with SARS-CoV-2 and treated with Taxifolin
(n = 4) compared with untreated mock (n = 6), DMSO control (n = 8) and heat-inactivated (h.i.) SARS-CoV-2 (n = 6)
normalized to total RNA. The bar graphs summarize mRNA expressions of (C) COX-2, (D) HPGD, (E) PTGES, (F)
PTGES2 and (G) PTGES3 of SARS-CoV-2 infected Calu-3 cells (n = 3). (H) Representative gel image of NSP7 and
B2M mRNA expression of SARS-CoV-2 infected lung slices (120 hpi) and control slices. The bar graphs summarize
mRNA expressions of (I) COX-2, (J) HPGD and (K) PTGES of SARS-CoV-2 infected lung slices (120 hpi; n = 3 for
ctrl, n = 4 for SARS-CoV-2 infection). Data are presented as mean±SD, (B) unpaired two-tailed t-test, �P<0.05 vs.
mock, ��P<0.01 vs. mock, ##P<0.01 vs. SARS-CoV-2 + DMSO. (C-G) One sample t-test, �P<0.05, ��P<0.01 vs. ctrl,
(I-K) unpaired two-tailed t-test, �P<0.05 vs. ctrl. Underlying data can be found in S1 Data and uncropped gel images
in S6 Fig.
https://doi.org/10.1371/journal.pone.0255335.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
11 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 4. PGE2 stimulation of B-cells modulates the immune response. (A) The bar graph summarizes PAX5 mRNA
expression of 697 pre-B-cells treated with AH6809 (10 μM) or GW627368 (10 μM) and PGE2 (10 μM) for 48 h (n = 18
for ctrl and PGE2 treated cells, n = 3 for AH6809 treated cells and n = 6 for GW627368 treated cells). (B) The bar graph
summarizes PAX5 mRNA expression of human pre-B-cell line SUP-B15 with PGE2 (10 μM) for 48 h (n = 9). (C) The
bar graph summarizes PAX5 mRNA expression of 697 pre-B-cells treated with human serum collected at BL and after
12-months FU of controlled physical training (n = 11). (D) The bar graph summarizes PAX5 mRNA expression of
SUP-B15 pre-B-cells treated with human serum collected at BL and after 12-months FU of controlled physical training
(n = 4). The bar graph summarizes PAX5 mRNA expression of (E) 697 (n = 11) and (F) SUP-B15 (n = 2) pre-B-cells
treated with serum from elderly individuals with high PGE2 levels with and without GW627368 (10 μM). Control pre-
B cells were treated with the solvent DMSO. (A) unpaired two-tailed t-test, ��P<0.01 vs. ctrl, #P<0.05 vs. PGE2, (B-F)
One sample t-test, �P<0.05, ��P<0.01 vs. ctrl or BL, the mean of ctrl or BL was set at 100%. Underlying data can be
found in S1 Data.
https://doi.org/10.1371/journal.pone.0255335.g004
GW627368 increased PAX5 in 697 and SUP-B15 pre-B-cells exposed to serum collected before
physical exercise, indicating that the suppressive effect is mediated by PGE2-EP4 (Fig 4E and
4F).
Effect of serum from COVID-19 patients on PAX5 expression in human
pre-B-cells
Serum from COVID-19 patients with elevated PGE2 levels reduced the expression of PAX5 in
SUP-B15 cells compared with serum from healthy controls. Again, this effect was blocked
upon co-treatment with the PGE2 receptor 4 (EP4) antagonist, GW627368 (Fig 5A).
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
12 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 5. Modulation of the immune response in COVID-19 patients. (A) The bar graph summarizes PAX5 mRNA
expression of SUP-B15 pre-B-cells treated with serum from healthy controls (serum pooled from 9 controls) and from
COVID-19 patients (serum pooled from 9 COVID-19 patients) incubated with and without GW627368 (10 μM).
Control cells were treated with the solvent DMSO (n = 6 wells with control serum and n = 3 wells with serum of
COVID-19 patients with and without GW627368). (B) Representative gel image of NSP7 and B2M mRNA expression
in control lung tissue (ctrl), in lung tissue of patients with severe acute COVID-19 disease (AC) and in lung tissue
obtained after transplant rejection (TR). (C) Immunohistochemical staining for CD68+, CD4+, CD8+ and CD20+
immune cells (scale bar: 100 μm), (D) Dot plot summarizing the immunohistological quantification of CD20 positive
B-cells per area (mm), dot plots summarize mRNA expression of (E) CD20 and (F) of CD138 in control lung tissue
(ctrl), in the lung tissue of patients with severe acute COVID-19 disease (AC) and in lung tissue obtained after
transplant rejection (TR). (A) One sample t-test, ��P<0.01 vs. ctrl, # P<0.05 vs. serum from COVID-19 patients. (D)
Statistical analysis was performed using the generalized linear model with Gaussian distribution and weights adjusted
according to the number of ROIs per patient, ���P<0.001 vs. ctrl, ###P<0.001 vs AC. (E, F) unpaired two-tailed t-test,
��P<0.01 vs. ctrl, �P<0.05 vs. ctrl. Underlying data can be found in S1 Data and uncropped gel images in S7 Fig.
https://doi.org/10.1371/journal.pone.0255335.g005
Analyses of B-cells in lungs from patients who died of severe acute COVID-
19 disease compared with healthy controls and transplant rejection
biopsies
In lung biopsies from patients who died of severe acute COVID-19 disease (AC group, con-
firmed by qRT-PCR for NSP7, Fig 5B), the signals for CD20 pre-B-cells (qRT-PCR and immu-
nohistochemical quantification) and plasma cells (qRT-PCR for CD138) were barely
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
13 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
detectable and lower than in control lung tissue (ctrl) and markedly lower than in lung tissue
obtained after transplant rejection (TR, Fig 5C–5F). Lung tissue immunostaining showed
increased numbers of CD68+ macrophages and CD4+ T-cells in AC and TR compared with
ctrl lung biopsies (Fig 5C).
Discussion
The key finding of this study is that PGE2 is elevated in patients with COVID-19 disease, with
the highest blood levels observed in those severely affected. Furthermore, SARS-CoV-2 itself
upregulates PGE2 in infected host cells and risk factors such as male sex, age and sedentary life
style are also associated with higher PGE2 serum levels. Finally, PGE2 impairs the B-cell medi-
ated immune response at least in part by reducing PAX5 while the PGE2 inhibitor Taxifolin
attenuates SARS-CoV-2 induced PGE2 production. Moreover, regular exercise also reduces
PGE2 levels in elderly subjects, which is associated with increased PAX5 production in B-cells
exposed to these sera. Thus, PGE2 may emerge as a modulating factor for disease severity and
development of immunity and could therefore be a therapeutic target in COVID-19 preven-
tion and treatment.
Since it is known that PGE2 can exert immunosuppressive effects during viral infection
[25–27], its elevation might critically reduce the initial defense against SARS-CoV-2 and may
thereby lead to more severe disease courses. Interestingly, our data show that the SARS-CoV-2
virus, not only hijacks the host cell gene expression machinery in order to replicate, but also
forces infected host cells to produce PGE2 by upregulating the PGE-generating enzyme COX-
2, and at least in part by reducing the expression of the PGE2-degrading enzyme HPGD (Fig
6). In line with the upregulation of COX-2 but without a specific upregulation of PGE2
synthases by SARS-CoV-2 in infected human lung cells, the production of another prostaglan-
din, PGD2, was also increased. However, to study the regulation and role of PGD2 in COVID-
19 disease was beyond the scope of the present study and needs further investigation. In addi-
tion, we provide evidence that reported risk factors for more severe COVID-19 disease
courses, i.e. male sex, age and a sedentary life style [13, 41] are associated with higher PGE2
levels as PGE2 serum levels are higher in men than women, higher in elderly (>60 years) indi-
viduals of both sexes than in younger individuals, and PGE2 levels in elderly could be reduced
by regular exercise (Fig 6). These findings might explain why males or elderly individuals are
more affected than females or younger individuals. Sex-related differences in circulating PGE2
levels appeared to be specific for younger individuals since in the healthy cohort older <60
years no such differences were observed. Whether age-related hormonal changes in older
females contributes to the age effect in women needs to elucidated in future studies.
In addition to already known effects of PGE2 on immune cells, we discovered a novel
mechanism by which PGE2 in serum from COVID-19 patients specifically impacts on pre-B-
cells since PGE2 in the sera of COVID-19 patients reduces the expression of PAX5 in human
pre-B-cells via its EP4 receptor. PAX5 is a master regulator of most aspects of the life cycle of
B-cells as it represses the transcription of genes required for the development of other hemato-
poietic lineages and plasma cells and by controlling numerous genes that are required for early
development, antigen-receptor recombination, signaling and adhesion [42–44]. Moreover,
while high PAX5 expression is necessary for the above described processes, its reduction is
important for the final differentiation of short-lived plasma cells and their antibody (AB) pro-
duction. Thereby, high PGE2 serum levels on one hand reduces the number of pre-B-cells, but
on the other hand boosts the terminal differentiation of B-cells towards short-lived plasma
cells, two features that on the long run would lead to depleting the B-cell reservoir. This feature
may explain why some patients with initially high SARS-CoV-2-directed AB titers but evolving
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
14 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
15 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 6. Schematic representation of pleiotropic influences of SARS-CoV-2 infection, physical activity and age on
PGE2 levels and the ensuing altered immune response. (A) Modulators of PGE2 synthesis and degradation are
SARS-CoV-2 infection, but also physical inactivity, sex and older age, which are all risk factors for more severe
COVID-19 disease courses [7, 13–15]. Additionally, SARS-CoV-2 infection induces TNFα expression that is known to
mediate increased COX-2 expression [37]. These modulators upregulate the expression of the PGE2-generating
enzyme COX-2 and at least in part reduce the expression of the PGE2-degrading enzyme HPGD, which results in
increased generation and secretion of PGE2. PGE2 targets the innate immune system (monocytes/macrophages),
where it lowers its efficacy to remove pathogens in part by reducing the release of cytokines [46, 59]. Additionally,
PGE2 impairs the response of the adaptive immune system against pathogens by lowering proliferation and survival of
T-cells and inducing T-cell [3, 52, 60, 61]. Furthermore, PGE2 is impairing the B-cell response to pathogens in part by
directly suppressing the B-cell specific transcription factor PAX5 [62]. Increased PGE2 secretion can be prevented by
physical exercise and specific PGE2 inhibitors such as Taxifolin. In addition, Taxifolin reduces viral replication. The
low immune response (phase 1) may enable the entry of secondary infections with bacteria and fungi and reinfections
with SARS-CoV-2 associated with tissue and organ injury, formation of autoantibodies potentially leading to a
cytokine storm and an excessive immune response [20–24]. (B) In pre-B-cells, PAX5 is responsible for suppressing
other hematopoietic differentiation programs and promotes proliferation, survival and differentiation of pre-B-cells
[42–44]. PGE2 reduces PAX5 expression via its EP4 receptor, which not only reduces their survival and proliferation
but boosts the differentiation of B-cells towards plasma cells and may even allow transdifferentiation, features that may
lead to the cytokine storm but also the depletion of the B-cell pool (and germinal centers) [19, 42]. In addition, since
PAX5 is important for the formation memory cells, PGE2 is therefore also lowering the formation of immunity [44,
47]. Blocking the EP4 signaling with the EP4 receptor antagonist GW627368 prevents downregulation of PAX5 in pre-
B-cells and may improve viral defense and formation of immunity against SARS-CoV-2.
https://doi.org/10.1371/journal.pone.0255335.g006
towards a severe disease course display a reduction in germinal centers [19] and reduced B-cell
response thereafter. Our findings in postmortem lung tissue of patients who died of COVID-
19 are in line with this interpretation. Indeed, we detected reduced CD20+ B-cells numbers in
COVID-19 lung tissue in comparison with healthy control tissue or with transplant rejection
lung biopsies. Likewise, other reports show no significant lymphocyte invasion in cardiac tis-
sue despite the presence of SARS-CoV-2 particles [10, 11]. Additional studies suggest higher
risks for severe disease courses in COVID-19 patients with dysfunctional B-cells due to com-
mon variable immune deficiencies (CVIDs) [18], while in turn, patients with larger pools of
naïve B-cells seem to build a more effective immune response to SARS-CoV-2 [45].
The observed low B-cell signals in lung biopsies from patients who died during acute
SARS-CoV-2 infection may also point to loss of these immune cells by PANoptosis (inflamma-
tory cell death). In this regard, Karki et al. reported that during SARS-CoV-2 infection a com-
bination of TNFα and IFNγ could induce PANoptosis [40]. However, we observed the
opposite, i.e. PGE2 reduced the expression of TNFα and IFNγ in pre-B-cells, a feature that has
also been reported for monocytes and macrophages [46].
Moreover, we found that PGE2 reduces the proliferation of human pre-B-cells, an observa-
tion that fits well with the PGE2-mediated reduction of PAX5 and may thereby contribute to
rarification of B-cells in infected tissue.
In addition, since we observed that SARS-CoV-2-infected lung cells upregulate TNFα
expression and since TNFα is known to induce COX-2 expression, we found one possible
mechanism how SARS-CoV-2 may upregulated PGE2 production in infected tissues (Fig 6A)
[37].
As reported above, high PGE2 in COVID-19 serum impairs the B-cell mediated immune
response at least in part by reducing PAX5. PAX5 expression is also necessary for the develop-
ment of memory B-cells after follicular B-cells have encountered antigens [44, 47]. In this
regard, elevated PGE2 would also reduce the ability of an organism to develop longstanding
immunity after COVID-19 infection. Indeed, there are reports on reinfection in individuals
with SARS-CoV-2 [48–50] including a recent case report of a patient with a CD20+ B-cell
acute lymphoblastic leukemia who developed high AB titers against COVID-19 after an initial
recovery. However, the patient experienced a viral reactivation after she lost her COVID-19
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
16 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
AB following the administration of rituximab, cytarabine, and dasatinib for her leukemia, and
experienced severe COVID-19 pneumonia with lymphopenia and high inflammatory markers
[51]. PGE2 not only affects B-cells, but also promotes T-cell exhaustion and viral expansion
through EP2 and EP4, as revealed by recent studies [52] and immunosuppression caused by
T-cell depletion and exhaustion have been suggested as contributing to viral persistence and
mortality in COVID-19 patients [3].
Based on the suspected crucial role of PGE2 for COVID-19 disease courses, we tested the
potential of the PGE2 inhibitor Taxifolin, also known as dihydroquercetin, to limit SARS-CoV-
2-induced PGE2 production in human lung cells (Fig 6A). In agreement with our hypothesis
that PGE2 contributes to severe COVID-19 disease, Taxifolin significantly reduced PGE2 pro-
duction in infected lung cells. Additionally, a recent publication on screening for natural inhibi-
tors for SARS-CoV-2 in silico identified Taxifolin as a direct inhibitor of the SARS-CoV-2 main
protease [53]. Taxifolin is a potent flavonoid with anti-inflammatory activity, which is present as
a natural compound in vegetables and fruits and the Siberian larch, Larix sibirica, [35, 36]. It is
readily available in foodstuffs and could be tested directly in COVID-19 patients. PGE2 synthesis
can be inhibited by NSAIDs, which block COX-1 and -2. However, it is known that NSAIDs are
interfering with the RAAS [54] and in this context, controversial data have been reported sug-
gesting that NSAIDs may favor SARS-CoV-2 entry by upregulating ACE2 [55, 56]. Moreover,
NSAIDs by inhibiting COX-1 and -2 may also reduce the generation of additional prostaglan-
dins, which may have beneficial effects. Therefore, and because the safety of using NSAIDs in
the treatment of COVID-19 patients is discussed critical, we decided to use Taxifolin as an alter-
native treatment strategy. Indeed, we could show that Taxifolin blocked only the SARS-CoV-
2-induced PGE2 synthesis but not the synthesis PGD2 in infected lung cells. Inhibition of the
microsomal prostaglandin E synthase-1 (mPGES-1) by sonlicromanol (Khondrion; a drug cur-
rently in phase 2b studies for mitochondrial disease), may also be beneficial in COVID-19
patients (Fig 6A). Moreover, COVID-19 patients could also benefit from COX-inhibitors such
as aspirin and ibuprofen in the early phase of disease as suggested by a recent review [57]. Treat-
ment of mild and severely affected patients with corticoids, like Dexamethasone or Medrol, has
been associated with better outcome. Here, we observed that corticoids seem to have no effect
on circulating PGE2 levels although number of patients in these subgroup analyses was too low
to be conclusive. Finally, we provide evidence that regular physical activity lowers PGE2 in the
serum of elderly individuals without COVID-19 infection and may thereby support their
immune systems in fighting SARS-CoV-2 infection (Fig 6A).
Thus, known risk factors for severe COVID-19 disease such as age, sex and physical inactiv-
ity are associated with elevated PGE2 levels prior infection and may thereby contribute to a
reduced immune response at the time of SARS-CoV-2 infection. In addition, the SARS-CoV-2
infection may further compromise the immune response by further upregulating PGE2 in
those individuals with pre-existing higher PGE2 levels. Furthermore, it is known that also the
exposure to high levels SARS-CoV-2 virus particles contribute to severe COVID-19 disease
also in individuals with otherwise low risk factors (for example severe disease cases in nurses
and physicians) [58]. As we could demonstrate that SARS-CoV-2-infected host cells produce
high levels of PGE2, a massive infection with SARS-CoV-2 virus may lead to high PGE2 secre-
tion and high circulating PGE2 levels, which subsequently reduced the immune response also
in individual with otherwise low risk for severe disease.
Conclusions
In conclusion, our data suggest that PGE2 production, either induced by SARS-CoV-2 infec-
tion or determined by endogenous and exogenous risk factors critically influences COVID-19
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
17 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
disease severity, (Fig 6A). Mechanistically, we show that PGE2 specifically targets B-cells by
reducing PAX5, a key factor for B-cell proliferation and differentiation (Fig 6A and 6B).
Reducing PGE2 levels preventively and/or during COVID-19 disease may therefore provide a
valuable therapeutic strategy to prevent and fight SARS-CoV-2 infection and to enhance and
prolong immunity.
Limitations of the study
Limitations of our study include the limited numbers of blood samples from COVID-19
patients and that clinical data on COVID-19 patients, i.e. as C-reactive protein (CRP), lactate
dehydrogenase (LDH), leukocytes normal count, neutrophils normal count, and lymphocytes
were not available for all patients.
PGE2 synthesis can be blocked by corticosteroids that inhibit the phospholipases or by
NSAIDs that inhibit the cyclooxygenase. In this study, at the time of blood sampling a part of
the COVID-19 patients with mild or severe disease were treated with corticosteroids or
NSAIDs. Information on the use of NSAIDs or leukotriene modifiers were not available. PGE2
levels in those patients might be underestimated, since both medications may reduce PGE2
biosynthesis.
Most individuals in the healthy elderly collective displayed age-related normal BMI and
numbers in subgroup with increased or reduced BMI were too low to perform conclusive cor-
relation analyses with PGE2 levels.
Serum and plasma samples have to be stored at -80˚C immediately to avoid degradation of
PGE2 and to avoid further prostaglandin synthesis by COX-2. For the present study serum
and plasma was immediately being processed, frozen and stored at -80˚C.
Venipuncture and ex vivo platelet activation may alter plasma prostanoid concentrations, a
feature that cannot be completely excluded.
Supporting information
S1 Fig. PGE2 serum levels showed no correlation with BMI, BW or body fat content.
Ozone correlation analysis of serum PGE2 levels with (A-E) BMI ((A) males: n = 40, Spearman
r: -0.1485, P value: 0.3604; (B) males in normal range BMI 25–30: n = 24, Spearman r: -0.1231,
P value: 0.5667; (C) males with a BMI >30: n = 9, Spearman r: 0.3167, P value: 0.4101 (D)
males with a BMI <25: n = 7, Spearman r: -0.2143, P value: 0.6615 (E) females: n = 45, Pearson
r: 0.03956, P value: 0.7964), (F, G) BW (males: n = 40, Spearman r:-0.08246, P value: 0.6130;
females: n = 45, Pearson r: 0.05614, P value: 0.7142) and (H, I) body fat content (males: n = 37,
Pearson r:-0.03295, P value: 0.8465; females: n = 43, Pearson r: 0.1374, P value: 0.3797) in
(A-D, F, H) males and (E, G, I) females. (A-I) Ozone correlation, Spearman or Pearson corre-
lation coefficients, two-tailed P value. Underlying data can be found in S1 Data.
(TIFF)
S2 Fig. SARS-CoV-2 infection in Calu-3 cells. (A) Heat-inactivation (h.i.) of PGE2 for 30
min at 70˚C compared to untreated PGE2 (ctrl) from the same sample (n = 4). Data are pre-
sented as mean±SD, ctrl was set at 100%, one-sample t-test. (B) The bar graph summarizes
TNFa mRNA expression of SARS-CoV-2 infection in Calu-3 cells in cell culture lysates (n = 3
independent cell culture experiments). Data are presented as mean±SD, mock was set at 100%,
�P<0.05 vs mock, #P<0.05 vs h.i., one-way ANOVA, Dunnett post hoc test. Underlying data
can be found in S1 Data.
(TIFF)
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
18 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
S3 Fig. Taxifolin treatment has no effect on the secretion of PGD2 or the proliferation
capacity of Calu-3 cells. The bar graph summarizes PGD2 content in supernatants of Calu-3
cells infected with SARS-CoV-2 and treated for 48 h with Taxifolin (100 μM; n = 12) compared
with DMSO control (n = 10). The bar graphs summarize the mRNA expression of the prolifer-
ation markers (B) Ki67, (C) TOP2A and (D) TPX2 of Calu-3 cells treated with Taxifolin
(100 μM) for 48 h (n = 7 for ctrl and PGE2 treated cells). (E) The bar graph summarizes PGD2
content in supernatants of Calu-3 cells infected with SARS-CoV-2 (n = 6) compared with
untreated mock (n = 6) and heat-inactivated (h.i.) SARS-CoV-2 (n = 6) normalized to total
RNA. (A-E) Data are presented as mean±SD, (A, B, C) unpaired two-tailed t-test, n.s. (D)
Mann-Whitney-U test, n.s. (E) mock was set at 100%, ��P<0.01 vs mock, #P<0.05 vs h.i., one-
way ANOVA, Bonferroni’s post hoc test. Underlying data can be found in S1 Data.
(TIFF)
S4 Fig. PGE2 stimulation of pre-B-cells modulates the cell number due to alterations of
the proliferation capacity. Representative gel images of PTGER4 and B2M mRNA expression
in pre-B-cell lines (A) 697 and (B) SUP-B15. (C) The bar graph summarizes the percentage of
live cells of control treated and PGE2 (10 μM) treated 697 cells after 48h stimulation. Total cell
number was set at 100%. (D-F) The bar graphs summarize the mRNA expression of the prolif-
eration markers (C) Ki67, (D) TOP2A and (E) TPX2 of pre-B-cells 697 treated with PGE2
(10 μM) for 48 h (n = 5 for ctrl and PGE2 treated cells). (C-E) Data are presented as mean±SD,
(C-E) n. s., ��P<0.01 vs ctrl, unpaired two-tailed t-test. Underlying data can be found in S1
Data and uncropped gel images in S8 Fig.
(TIFF)
S5 Fig. PGE2 stimulation of pre-B-cells is not associated with elevated TNFa or IFNg
expression. The bar graph summarizes TNFa mRNA expression of (A) 697 or (B) SUP-B18
pre-B-cells treated with PGE2 (10 μM) after 48 h (n = 5). Control cells were treated with the
solvent ETHO (n = 5). The bar graph summarizes IFNg mRNA expression of (C) 697 or (D)
SUP-B18 pre-B-cells treated with PGE2 (10 μM) after 48 h (n = 5). Control cells were treated
with the solvent ETHO (n = 4). (A-D) Data are presented as mean±SD, (A, C, D) n. s.,
��P<0.01 vs ctrl, unpaired two-tailed t-test and (B) ��P<0.01, Mann-Whitney-U test. Underly-
ing data can be found in S1 Data.
(TIFF)
S6 Fig. The uncropped gel for Fig 3A and 3H.
(TIFF)
S7 Fig. The uncropped gel for Fig 5B.
(TIFF)
S8 Fig. The uncropped gel for S4A and S4B Fig.
(TIFF)
S1 Data. Numerical raw data. All numerical raw data are combined in a single excel file,
“S1_Data.xlsx,” this file consists of several spreadsheets and each contains the data of 1 figure
or table.
(XLSX)
S1 Table. List of human qRT-PCR primers.
(DOCX)
S2 Table. Summary of clinical data from male and female probands baseline (BL) and
after 12 M Follow-Up (FU) controlled exercise (E). Body mass index (BMI) was determined
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
19 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
with BMI = bodyweight (BW) / squared height. Body weight, body height, BMI, body fat and
activity were analyzed at BL and after 12M FU controlled exercise. Comparison between the
groups BL vs 12M FU was performed using Student’s t-test for Gaussian distributed data (pre-
sented as mean ± SD) and the Mann-Whitney-U test where at least one column was not nor-
mally distributed (presented as median and interquartile range (IQR)). ���P<0.001,
����P<0.00001 BL vs 12M FU. Underlying data can be found in S1 Data.
(DOCX)
S3 Table. Summary of clinical data of healthy controls from Fig 1A.
(DOCX)
Acknowledgments
We thank Sergej Erschow, Silvia Gutzke, Brigit Brandt, Angelica Julieth Diaz Basabe, Delphine
De Mulder, Thomas Gerlach and Giulietta Saletti for excellent technical assistance, and Dr.
Helge Stark for the bioinformatic analysis input. The SARS-CoV-2 virus isolate was kindly
provided by Christian Drosten, Charite´, Berlin.
Author Contributions
Conceptualization: Melanie Ricke-Hoch, Elisabeth Stelling, Thomas Pietschmann, Emilio
Hirsch, Danny Jonigk, Uwe Tegtbur, Axel Schambach, Axel Haverich, Denise Hilfiker-
Kleiner.
Data curation: Melanie Ricke-Hoch, Elisabeth Stelling, Martina Kasten, Thomas Gausepohl,
Anne Ho¨fer, Danny Jonigk, Julian Eigendorf, Lena Mink, Michaela Scherr, Tobias J. Pfeffer,
Denise Hilfiker-Kleiner.
Formal analysis: Melanie Ricke-Hoch, Elisabeth Stelling, Graham Brogden, Gisa Gerold,
Anne Ho¨fer, Danny Jonigk, Denise Hilfiker-Kleiner.
Funding acquisition: Melanie Ricke-Hoch, Gisa Gerold, Thomas Pietschmann, Jean-Luc Bal-
ligand, Emilio Hirsch, Guus F. Rimmelzwaan, Mark P. Ku¨hnel, Danny Jonigk, Thomas
Illig, Axel Schambach, Denise Hilfiker-Kleiner.
Investigation: Melanie Ricke-Hoch, Elisabeth Stelling, Lisa Lasswitz, Antonia P. Gunesch,
Francisco J. Zapatero-Belincho´n, Graham Brogden, Gisa Gerold, Federica Facciotti, Husni
Elbahesh, Guus F. Rimmelzwaan, Anne Ho¨fer, Danny Jonigk, Julian Eigendorf, Uwe Tegt-
bur, Lena Mink, Tobias J. Pfeffer, Axel Haverich, Denise Hilfiker-Kleiner.
Methodology: Melanie Ricke-Hoch, Elisabeth Stelling, Lisa Lasswitz, Antonia P. Gunesch,
Francisco J. Zapatero-Belincho´n, Graham Brogden, Gisa Gerold, Thomas Pietschmann,
Husni Elbahesh, Anne Ho¨fer, Danny Jonigk, Thomas Illig, Andres Hilfiker, Axel Haverich,
Denise Hilfiker-Kleiner.
Project administration: Melanie Ricke-Hoch, Gisa Gerold, Danny Jonigk, Julian Eigendorf,
Uwe Tegtbur, Denise Hilfiker-Kleiner.
Resources: Lisa Lasswitz, Antonia P. Gunesch, Francisco J. Zapatero-Belincho´n, Graham
Brogden, Gisa Gerold, Virginie Montiel, Jean-Luc Balligand, Federica Facciotti, Emilio
Hirsch, Anne Ho¨fer, Mark P. Ku¨hnel, Danny Jonigk, Uwe Tegtbur, Lena Mink, Thomas
Illig, Tobias J. Pfeffer, Andres Hilfiker, Axel Haverich.
Supervision: Melanie Ricke-Hoch, Gisa Gerold, Thomas Pietschmann, Mark P. Ku¨hnel,
Danny Jonigk, Denise Hilfiker-Kleiner.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
20 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Validation: Melanie Ricke-Hoch, Elisabeth Stelling, Gisa Gerold, Anne Ho¨fer, Danny Jonigk,
Denise Hilfiker-Kleiner.
Visualization: Melanie Ricke-Hoch, Elisabeth Stelling, Graham Brogden, Anne Ho¨fer, Danny
Jonigk, Michaela Scherr, Denise Hilfiker-Kleiner.
Writing – original draft: Melanie Ricke-Hoch, Elisabeth Stelling, Lisa Lasswitz, Antonia P.
Gunesch, Husni Elbahesh, Anne Ho¨fer, Mark P. Ku¨hnel, Danny Jonigk, Denise Hilfiker-
Kleiner.
Writing – review & editing: Melanie Ricke-Hoch, Elisabeth Stelling, Francisco J. Zapatero-
Belincho´n, Graham Brogden, Gisa Gerold, Thomas Pietschmann, Virginie Montiel, Jean-
Luc Balligand, Thomas Gausepohl, Husni Elbahesh, Guus F. Rimmelzwaan, Anne Ho¨fer,
Mark P. Ku¨hnel, Danny Jonigk, Julian Eigendorf, Uwe Tegtbur, Lena Mink, Thomas Illig,
Axel Schambach, Tobias J. Pfeffer, Andres Hilfiker, Axel Haverich, Denise Hilfiker-Kleiner.
References
1.
Fried JA, Ramasubbu K, Bhatt R, Topkara VK, Clerkin KJ, Horn E, et al. The Variety of Cardiovascular
Presentations of COVID-19. Circulation. 2020; 141(23):1930–6. https://doi.org/10.1161/
CIRCULATIONAHA.120.047164 PMID: 32243205
2.
Lau MSY, Grenfell B, Thomas M, Bryan M, Nelson K, Lopman B. Characterizing superspreading events
and age-specific infectiousness of SARS-CoV-2 transmission in Georgia, USA. Proc Natl Acad Sci U S
A. 2020; 117(36):22430–5. https://doi.org/10.1073/pnas.2011802117 PMID: 32820074
3. Vardhana SA, Wolchok JD. The many faces of the anti-COVID immune response. The Journal of exper-
imental medicine. 2020; 217(6). https://doi.org/10.1084/jem.20200678 PMID: 32353870
4. Ackermann M, Verleden SE, Kuehnel M, Haverich A, Welte T, Laenger F, et al. Pulmonary Vascular
Endothelialitis, Thrombosis, and Angiogenesis in Covid-19. N Engl J Med. 2020; 383(2):120–8. https://
doi.org/10.1056/NEJMoa2015432 PMID: 32437596
5. Yin S, Huang M, Li D, Tang N. Difference of coagulation features between severe pneumonia induced
by SARS-CoV2 and non-SARS-CoV2. J Thromb Thrombolysis. 2020.
6.
7.
Zheng YY, Ma YT, Zhang JY, Xie X. COVID-19 and the cardiovascular system. Nature reviews Cardiol-
ogy. 2020. https://doi.org/10.1038/s41569-020-0360-5 PMID: 32139904
Zheng Z, Peng F, Xu B, Zhao J, Liu H, Peng J, et al. Risk factors of critical & mortal COVID-19 cases: A
systematic literature review and meta-analysis. J Infect. 2020.
8. Nicolai L, Leunig A, Brambs S, Kaiser R, Weinberger T, Weigand M, et al. Immunothrombotic Dysregu-
lation in COVID-19 Pneumonia Is Associated With Respiratory Failure and Coagulopathy. Circulation.
2020; 142(12):1176–89. https://doi.org/10.1161/CIRCULATIONAHA.120.048488 PMID: 32755393
9. Szekely Y, Lichter Y, Taieb P, Banai A, Hochstadt A, Merdler I, et al. Spectrum of Cardiac Manifesta-
tions in COVID-19: A Systematic Echocardiographic Study. Circulation. 2020; 142(4):342–53. https://
doi.org/10.1161/CIRCULATIONAHA.120.047971 PMID: 32469253
10. Chen L, Li X, Chen M, Feng Y, Xiong C. The ACE2 expression in human heart indicates new potential
mechanism of heart injury among patients infected with SARS-CoV-2. Cardiovasc Res. 2020; 116
(6):1097–100. https://doi.org/10.1093/cvr/cvaa078 PMID: 32227090
11.
Fox SE, Li G, Akmatbekov A, Harbert JL, Lameira FS, Brown JQ, et al. Unexpected Features of Cardiac
Pathology in COVID-19 Infection. Circulation. 2020; 142(11):1123–5. https://doi.org/10.1161/
CIRCULATIONAHA.120.049465 PMID: 32689809
12. Sharma A, Garcia G Jr., Wang Y, Plummer JT, Morizono K, Arumugaswami V, et al. Human iPSC-
Derived Cardiomyocytes Are Susceptible to SARS-CoV-2 Infection. Cell Rep Med. 2020; 1(4):100052.
https://doi.org/10.1016/j.xcrm.2020.100052 PMID: 32835305
13. Wenham C, Smith J, Morgan R, Gender, Group C-W. COVID-19: the gendered impacts of the outbreak.
Lancet. 2020; 395(10227):846–8. https://doi.org/10.1016/S0140-6736(20)30526-2 PMID: 32151325
14. Brandt EB, Beck AF, Mersha TB. Air pollution, racial disparities, and COVID-19 mortality. J Allergy Clin
Immunol. 2020; 146(1):61–3. https://doi.org/10.1016/j.jaci.2020.04.035 PMID: 32389591
15. Martelletti L, Martelletti P. Air Pollution and the Novel Covid-19 Disease: a Putative Disease Risk Factor.
SN Compr Clin Med. 2020:1–5. https://doi.org/10.1007/s42399-020-00274-4 PMID: 32296757
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
21 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
16.
Li D, Chen Y, Liu H, Jia Y, Li F, Wang W, et al. Immune dysfunction leads to mortality and organ injury in
patients with COVID-19 in China: insights from ERS-COVID-19 study. Signal Transduct Target Ther.
2020; 5(1):62. https://doi.org/10.1038/s41392-020-0163-5 PMID: 32371949
17. Wilk AJ, Rustagi A, Zhao NQ, Roque J, Martinez-Colon GJ, McKechnie JL, et al. A single-cell atlas of
the peripheral immune response in patients with severe COVID-19. Nat Med. 2020; 26(7):1070–6.
https://doi.org/10.1038/s41591-020-0944-y PMID: 32514174
18. Quinti I, Lougaris V, Milito C, Cinetto F, Pecoraro A, Mezzaroma I, et al. A possible role for B cells in
COVID-19? Lesson from patients with agammaglobulinemia. J Allergy Clin Immunol. 2020. https://doi.
org/10.1016/j.jaci.2020.04.013 PMID: 32333914
19. Kaneko N, Kuo HH, Boucau J, Farmer JR, Allard-Chamard H, Mahajan VS, et al. Loss of Bcl-6-
Expressing T Follicular Helper Cells and Germinal Centers in COVID-19. Cell. 2020.
20. Catanzaro M, Fagiani F, Racchi M, Corsini E, Govoni S, Lanni C. Immune response in COVID-19:
addressing a pharmacological challenge by targeting pathways triggered by SARS-CoV-2. Signal
Transduct Target Ther. 2020; 5(1):84. https://doi.org/10.1038/s41392-020-0191-1 PMID: 32467561
21. Hu B, Huang S, Yin L. The cytokine storm and COVID-19. J Med Virol. 2020.
22. Mehta P, McAuley DF, Brown M, Sanchez E, Tattersall RS, Manson JJ, et al. COVID-19: consider cyto-
kine storm syndromes and immunosuppression. Lancet. 2020; 395(10229):1033–4. https://doi.org/10.
1016/S0140-6736(20)30628-0 PMID: 32192578
23. Wang SF, Tseng SP, Yen CH, Yang JY, Tsao CH, Shen CW, et al. Antibody-dependent SARS corona-
virus infection is mediated by antibodies against spike proteins. Biochemical and biophysical research
communications. 2014; 451(2):208–14. https://doi.org/10.1016/j.bbrc.2014.07.090 PMID: 25073113
24.
Zhang Q, Bastard P, Liu Z, Le Pen J, Moncada-Velez M, Chen J, et al. Inborn errors of type I IFN immu-
nity in patients with life-threatening COVID-19. Science. 2020; 370(6515).
25. Sander WJ, O’Neill HG, Pohl CH. Prostaglandin E2 As a Modulator of Viral Infections. Front Physiol.
2017; 8:89. https://doi.org/10.3389/fphys.2017.00089 PMID: 28261111
26. Coulombe F, Jaworska J, Verway M, Tzelepis F, Massoud A, Gillard J, et al. Targeted prostaglandin E2
inhibition enhances antiviral immunity through induction of type I interferon and apoptosis in macro-
phages. Immunity. 2014; 40(4):554–68. https://doi.org/10.1016/j.immuni.2014.02.013 PMID: 24726877
27.
28.
Full F, Gack MU. Prostaglandin E2: the villain in the host response to influenza virus. Immunity. 2014;
40(4):453–4. https://doi.org/10.1016/j.immuni.2014.03.008 PMID: 24745327
Liu L, Cao Z, Chen J, Li R, Cao Y, Zhu C, et al. Influenza A virus induces interleukin-27 through cycloox-
ygenase-2 and protein kinase A signaling. J Biol Chem. 2012; 287(15):11899–910. https://doi.org/10.
1074/jbc.M111.308064 PMID: 22343630
29. Schmidleithner L, Thabet Y, Schonfeld E, Kohne M, Sommer D, Abdullah Z, et al. Enzymatic Activity of
HPGD in Treg Cells Suppresses Tconv Cells to Maintain Adipose Tissue Homeostasis and Prevent
Metabolic Dysfunction. Immunity. 2019; 50(5):1232–48 e14. https://doi.org/10.1016/j.immuni.2019.03.
014 PMID: 31027998
30. Hoxha M. What about COVID-19 and arachidonic acid pathway? Eur J Clin Pharmacol. 2020; 76
(11):1501–4. https://doi.org/10.1007/s00228-020-02941-w PMID: 32583353
31. Wolfel R, Corman VM, Guggemos W, Seilmaier M, Zange S, Muller MA, et al. Virological assessment
of hospitalized patients with COVID-2019. Nature. 2020; 581(7809):465–9. https://doi.org/10.1038/
s41586-020-2196-x PMID: 32235945
32. Mascaux C, Angelova M, Vasaturo A, Beane J, Hijazi K, Anthoine G, et al. Immune evasion before
tumour invasion in early lung squamous carcinogenesis. Nature. 2019; 571(7766):570–5. https://doi.
org/10.1038/s41586-019-1330-0 PMID: 31243362
33. Hoch M, Fischer P, Stapel B, Missol-Kolka E, Sekkali B, Scherr M, et al. Erythropoietin preserves the
endothelial differentiation capacity of cardiac progenitor cells and reduces heart failure during antican-
cer therapies. Cell Stem Cell. 2011; 9(2):131–43. https://doi.org/10.1016/j.stem.2011.07.001 PMID:
21816364
34. Khailany RA, Safdar M, Ozaslan M. Genomic characterization of a novel SARS-CoV-2. Gene Rep.
2020; 19:100682. https://doi.org/10.1016/j.genrep.2020.100682 PMID: 32300673
35. Hamalainen M, Nieminen R, Asmawi MZ, Vuorela P, Vapaatalo H, Moilanen E. Effects of flavonoids on
prostaglandin E2 production and on COX-2 and mPGES-1 expressions in activated macrophages.
Planta Med. 2011; 77(13):1504–11. https://doi.org/10.1055/s-0030-1270762 PMID: 21341175
36. Wang Y, Wang Q, Bao X, Ding Y, Shentu J, Cui W, et al. Taxifolin prevents beta-amyloid-induced
impairments of synaptic formation and deficits of memory via the inhibition of cytosolic phospholipase
A2/prostaglandin E2 content. Metab Brain Dis. 2018; 33(4):1069–79. https://doi.org/10.1007/s11011-
018-0207-5 PMID: 29542038
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
22 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
37. Nakao S, Ogtata Y, Shimizu E, Yamazaki M, Furuyama S, Sugiya H. Tumor necrosis factor alpha
(TNF-alpha)-induced prostaglandin E2 release is mediated by the activation of cyclooxygenase-2
(COX-2) transcription via NFkappaB in human gingival fibroblasts. Mol Cell Biochem. 2002; 238(1–
2):11–8. https://doi.org/10.1023/a:1019927616000 PMID: 12349897
38. Soleymani Fard S, Jeddi Tehrani M, Ardekani AM. Prostaglandin E2 induces growth inhibition, apopto-
sis and differentiation in T and B cell-derived acute lymphoblastic leukemia cell lines (CCRF-CEM and
Nalm-6). Prostaglandins Leukot Essent Fatty Acids. 2012; 87(1):17–24. https://doi.org/10.1016/j.plefa.
2012.04.012 PMID: 22749740
39. Chen R, Cao Y, Tian Y, Gu Y, Lu H, Zhang S, et al. PGE2 ameliorated viral myocarditis development
and promoted IL-10-producing regulatory B cell expansion via MAPKs/AKT-AP1 axis or AhR signaling.
Cell Immunol. 2020; 347:104025. https://doi.org/10.1016/j.cellimm.2019.104025 PMID: 31837749
40. Karki R, Sharma BR, Tuladhar S, Williams EP, Zalduondo L, Samir P, et al. Synergism of TNF-alpha
and IFN-gamma Triggers Inflammatory Cell Death, Tissue Damage, and Mortality in SARS-CoV-2
Infection and Cytokine Shock Syndromes. Cell. 2021; 184(1):149–68 e17. https://doi.org/10.1016/j.cell.
2020.11.025 PMID: 33278357
41. Davies NG, Klepac P, Liu Y, Prem K, Jit M, group CC-w, et al. Age-dependent effects in the transmis-
sion and control of COVID-19 epidemics. Nat Med. 2020; 26(8):1205–11. https://doi.org/10.1038/
s41591-020-0962-9 PMID: 32546824
42. Delogu A, Schebesta A, Sun Q, Aschenbrenner K, Perlot T, Busslinger M. Gene repression by Pax5 in
B cells is essential for blood cell homeostasis and is reversed in plasma cells. Immunity. 2006; 24
(3):269–81. https://doi.org/10.1016/j.immuni.2006.01.012 PMID: 16546096
43. Schebesta A, McManus S, Salvagiotto G, Delogu A, Busslinger GA, Busslinger M. Transcription factor
Pax5 activates the chromatin of key genes involved in B cell signaling, adhesion, migration, and immune
function. Immunity. 2007; 27(1):49–63. https://doi.org/10.1016/j.immuni.2007.05.019 PMID: 17658281
44. Nutt SL, Tarlinton DM. Germinal center B and follicular helper T cells: siblings, cousins or just good
friends? Nat Immunol. 2011; 12(6):472–7. https://doi.org/10.1038/ni.2019 PMID: 21739669
45. Schultheiss C, Paschold L, Simnica D, Mohme M, Willscher E, von Wenserski L, et al. Next-Generation
Sequencing of T and B Cell Receptor Repertoires from COVID-19 Patients Showed Signatures Associ-
ated with Severity of Disease. Immunity. 2020; 53(2):442–55 e4. https://doi.org/10.1016/j.immuni.2020.
06.024 PMID: 32668194
46. Agard M, Asakrah S, Morici LA. PGE(2) suppression of innate immunity during mucosal bacterial infec-
tion. Front Cell Infect Microbiol. 2013; 3:45. https://doi.org/10.3389/fcimb.2013.00045 PMID: 23971009
47.
48.
49.
50.
51.
Johnson K, Shapiro-Shelef M, Tunyaplin C, Calame K. Regulatory events in early and late B-cell differ-
entiation. Mol Immunol. 2005; 42(7):749–61. https://doi.org/10.1016/j.molimm.2004.06.039 PMID:
15829263
Iwasaki A. What reinfections mean for COVID-19. Lancet Infect Dis. 2020. https://doi.org/10.1016/
S1473-3099(20)30783-0 PMID: 33058796
Tillett RL, Sevinsky JR, Hartley PD, Kerwin H, Crawford N, Gorzalski A, et al. Genomic evidence for
reinfection with SARS-CoV-2: a case study. Lancet Infect Dis. 2020. https://doi.org/10.1016/S1473-
3099(20)30764-7 PMID: 33058797
Legros V, Denolly S, Vogrig M, Boson B, Siret E, Rigaill J, et al. A longitudinal study of SARS-CoV-2-
infected patients reveals a high correlation between neutralizing antibodies and COVID-19 severity.
Cell Mol Immunol. 2021; 18(2):318–27. https://doi.org/10.1038/s41423-020-00588-2 PMID: 33408342
Lancman G, Mascarenhas J, Bar-Natan M. Severe COVID-19 virus reactivation following treatment for
B cell acute lymphoblastic leukemia. J Hematol Oncol. 2020; 13(1):131. https://doi.org/10.1186/
s13045-020-00968-1 PMID: 33008453
52. Sajiki Y, Konnai S, Okagawa T, Nishimori A, Maekawa N, Goto S, et al. Prostaglandin E2-Induced
Immune Exhaustion and Enhancement of Antiviral Effects by Anti-PD-L1 Antibody Combined with
COX-2 Inhibitor in Bovine Leukemia Virus Infection. J Immunol. 2019; 203(5):1313–24. https://doi.org/
10.4049/jimmunol.1900342 PMID: 31366713
53. Andre´ F, Manuel S, Santhosh N, Markus A. L., Martin S. Inhibitors for Novel Coronavirus Protease Iden-
tified by Virtual Screening of 687 Million Compounds 2020.
54. Meune C, Mourad JJ, Bergmann JF, Spaulding C. Interaction between cyclooxygenase and the renin-
angiotensin-aldosterone system: rationale and clinical relevance. J Renin Angiotensin Aldosterone
Syst. 2003; 4(3):149–54. https://doi.org/10.3317/jraas.2003.023 PMID: 14608518
55. Smart L, Fawkes N, Goggin P, Pennick G, Rainsford KD, Charlesworth B, et al. A narrative review of
the potential pharmacological influence and safety of ibuprofen on coronavirus disease 19 (COVID-19),
ACE2, and the immune system: a dichotomy of expectation and reality. Inflammopharmacology. 2020;
28(5):1141–52. https://doi.org/10.1007/s10787-020-00745-z PMID: 32797326
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
23 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
56. Valenzuela R, Pedrosa MA, Garrido-Gil P, Labandeira CM, Navarro G, Franco R, et al. Interactions
between ibuprofen, ACE2, renin-angiotensin system, and spike protein in the lung. Implications for
COVID-19. Clin Transl Med. 2021; 11(4):e371. https://doi.org/10.1002/ctm2.371 PMID: 33931961
57. Baghaki S, Yalcin CE, Baghaki HS, Aydin SY, Daghan B, Yavuz E. COX2 inhibition in the treatment of
COVID-19: Review of literature to propose repositioning of celecoxib for randomized controlled studies.
Int J Infect Dis. 2020; 101:29–32. https://doi.org/10.1016/j.ijid.2020.09.1466 PMID: 33007455
58.
59.
Fajnzylber J, Regan J, Coxen K, Corry H, Wong C, Rosenthal A, et al. SARS-CoV-2 viral load is associ-
ated with increased disease severity and mortality. Nature communications. 2020; 11(1):5493. https://
doi.org/10.1038/s41467-020-19057-5 PMID: 33127906
Tang T, Scambler TE, Smallie T, Cunliffe HE, Ross EA, Rosner DR, et al. Macrophage responses to
lipopolysaccharide are modulated by a feedback loop involving prostaglandin E2, dual specificity phos-
phatase 1 and tristetraprolin. Sci Rep. 2017; 7(1):4350. https://doi.org/10.1038/s41598-017-04100-1
PMID: 28659609
60. Boniface K, Bak-Jensen KS, Li Y, Blumenschein WM, McGeachy MJ, McClanahan TK, et al. Prosta-
glandin E2 regulates Th17 cell differentiation and function through cyclic AMP and EP2/EP4 receptor
signaling. The Journal of experimental medicine. 2009; 206(3):535–48. https://doi.org/10.1084/jem.
20082293 PMID: 19273625
61.
Lee J, Aoki T, Thumkeo D, Siriwach R, Yao C, Narumiya S. T cell-intrinsic prostaglandin E2-EP2/EP4
signaling is critical in pathogenic TH17 cell-driven inflammation. J Allergy Clin Immunol. 2019; 143
(2):631–43. https://doi.org/10.1016/j.jaci.2018.05.036 PMID: 29935220
62. Simkin NJ, Jelinek DF, Lipsky PE. Inhibition of human B cell responsiveness by prostaglandin E2. J
Immunol. 1987; 138(4):1074–81. PMID: 3027169
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
24 / 24
PLOS ONE
| null |
10.1089_cmb.2022.0149.pdf
|
AVAILABILITY
IntOMICS is open source software written in R, under GPL-2, and available at https://bioconductor.org/
packages/IntOMICS. The implementation relies on existing R package
|
AVAILABILITY IntOMICS is open source software written in R, under GPL-2, and available at https://bioconductor.org/ packages/IntOMICS . The implementation relies on existing R packages.
|
JOURNAL OF COMPUTATIONAL BIOLOGY
Volume 30, Number 5, 2023
Mary Ann Liebert, Inc.
Pp. 569–574
DOI: 10.1089/cmb.2022.0149
Open camera or QR reader and
scan code to access this article
and other resources online.
IntOMICS: A Bayesian Framework for Reconstructing
Regulatory Networks Using Multi-Omics Data
ANNA PACˇ I´NKOVA´ 1,2 and VLAD POPOVICI1
ABSTRACT
Integration of multi-omics data can provide a more complex view of the biological system
consisting of different interconnected molecular components. We present a new compre-
hensive R/Bioconductor-package, IntOMICS, which implements a Bayesian framework for
multi-omics data integration. IntOMICS adopts a Markov Chain Monte Carlo sampling
scheme to systematically analyze gene expression, copy number variation, DNA methyla-
tion, and biological prior knowledge to infer regulatory networks. The unique feature of
IntOMICS is an empirical biological knowledge estimation from the available experimental
data, which complements the missing biological prior knowledge. IntOMICS has the poten-
tial to be a powerful resource for exploratory systems biology.
Keywords: Bayesian networks, integrative analysis, multi-omics, regulatory network.
1. INTRODUCTION
M ulti-omics data collect multiple modalities from the same set of samples and describe different
aspects of cellular functioning. Integrative analysis combining multi-omics data can enhance our
understanding of biological systems consisting of interconnected molecular components, which is crucial for
developing novel personalized therapeutic strategies for complex diseases. Therefore, developing a freely
available and user-friendly computational framework to infer regulatory relationships by integrating multiple
omics data is one of the most relevant problems in systems biology (Hasin et al., 2017; Subramanian et al.,
2020; Kang et al., 2022). Bayesian networks (BNs) are models used to represent probabilistic relationships
between multiple interacting entities (Pearl, 1988; Cooper, 1989; Neapolitan, 1990). Over the past decades,
BNs have become popular in computational biology (Lucas et al., 2004).
We present a new comprehensive R package, IntOMICS—a Bayesian framework based on Markov
Chain Monte Carlo (MCMC) (Madigan et al., 1995) for multi-omics data integration, which combines prior
1RECETOX, Faculty of Science, Masaryk University, Brno, Czech Republic.
2Faculty of Informatics, Masaryk University, Brno, Czech Republic.
# Anna Pa(cid:2)cı´nkova´ and Vlad Popovici, 2023. Published by Mary Ann Liebert, Inc. This Open Access article is
distributed under the terms of the Creative Commons License [CC-BY] (http://creativecommons.org/licenses/by/4.0),
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
569
570
PACˇ I´NKOVA´ AND POPOVICI
knowledge with data-derived evidence for inferring regulatory networks. IntOMICS complements the
missing prior knowledge using empirical biological knowledge estimated from the available experimental
data. For further details about the IntOMICS algorithm, its performance and benchmark analysis, see
Pa(cid:2)cı´nkova´ and Popovici (2022). IntOMICS implementation also includes functions to visualize empirical
biological knowledge and generate diagnostic plots of an MCMC sampling scheme (Madigan et al., 1995).
2. DESIGN AND IMPLEMENTATION
IntOMICS implementation consists of two modules (Fig. 1). The OMICS module includes data pre-
processing and computing some quantities needed to score a BN. IntOMICS apply the BGe score (Geiger
and Heckerman, 1994) developed for continuous data. The BN module includes the MCMC sampling
scheme for structure learning and sampling of BNs. In the last part of the BN module, IntOMICS infers the
resulting network structure, including the edge weights representing the empirical frequency of given edges
over the sample of network structures.
GE
n1
CNV
n2
METH
n3
B
n1
layers
definition
m
m
m
n1
FIG. 1.
IntOMICS workflow. IntOMICS
two independent modules:
consists of
the OMICS
yellow nodes are part of
module and gray nodes are part of the BN
module. Edge weight wi represents the
empirical frequency of given edge over
samples of network structures. B, biolog-
ical knowledge; BN, Bayesian network;
CNV, copy number variation; GE, gene
expression; METH, methylation.
filtering
METH
n3'
m
parent set configurations and BGe score computation
B
empirical
n1+n2+n3'
n1+n2+n3'
BGe score update
Markov Chain Monte Carlo structure learning
edge posterior probabilities calculation
CNV1
w1
CNV3
GE1
w5
METH3
CNV2
w3
w2
GE3
w7
GE2
w4
w6
GE4
INTOMICS: REGULATORY NETWORKS FROM MULTI-OMICS
571
Input to IntOMICS are
(cid:2) data matrices that represent collections of features for a set of samples (gene expression matrix [GE],
copy number variation matrix [CNV], and DNA methylation matrix [METH]) and
(cid:2) biological prior knowledge, which contains the information on known interactions among molecular
features from public database(s).
IntOMICS is designed to infer regulatory networks, even if copy number variation or DNA methylation
data (or both) are not available.
IntOMICS adapts MCMC scheme to multi-omics data—GE, CNV, and METH—by layers definition.
Edges from the GE to the CNV/METH layers are excluded from the set of candidate edges. The resulting
regulatory network structure consists of three types of nodes: GE nodes refer to gene expression levels,
CNV nodes refer to copy number variations, and METH nodes refer to DNA methylation levels. Edge
weight wi represents the empirical frequency of a given edge over samples of network structures.
Although the method is designed to work on any modalities defined in a continuous domain, the current
implementation is tuned for gene expression, copy number variation, and DNA methylation. Adding a new
modality requires the implementation of a new interface for the OMICS module, whereas the computational engine
in the BN module remains the same. In that case, the OMICS module interface needs to be modified to capture all
possible regulators of nodes from the given layer and accordingly define all possible parent set configurations.
2.1. Usage example
We use IntOMICS to investigate Wnt signaling and the role of the FOXM1 gene in epithelial ovarian
cancer (EOC) using 17 samples from the GSE146556 data set (Zhang et al., 2020) consisting of GE, CNV,
and METH data. EOC is characterized by TP53 mutations, DNA copy number aberrations, numerous pro-
moter methylation events, and NOTCH and FOXM1 signaling activation (The Cancer Genome Atlas
Research Network, 2011). FOXM1, one of the crucial oncogene drivers of EOC proliferation, is upregu-
lated in EOC (The Cancer Genome Atlas Research Network, 2011; Zhang et al., 2020).
Chen et al. (2016) identified FOXM1 as a novel target of the Wnt signaling essential for b-catenin
activation. FOXM1 accumulation in the nucleus promotes activation of Wnt signaling pathway by pro-
tecting the b-catenin/TCF4 complex from inhibition by CTNNBIP1. USP5–FOXM1 association abolishes
the CTNNBIP1 inhibition of the b-catenin/TCF4 complex. GSK3 activity enhances FBXW7-mediated
FOXM1 ubiquitination resulting in protein degradation. We select 14 genes from the Kyoto Encyclopedia
of Genes and Genomes (Ogata et al., 1999) Wnt signaling pathway together with FOXM1, USP5, and
FBXW7 genes to infer the regulatory network using IntOMICS.
The first step is to perform data preprocessing and compute quantities needed to score a BN using
omics_module() function:
> OMICS_mod_res <- omics_module(omics = omics, PK = PK, layers_def = layers_def,
TFtargs = TFtarg_mat,
lm_METH = TRUE,
annot = annot, gene_annot = gene_annot,
r_squared_thres = 0.5)
It is possible to use linear regression to filter irrelevant DNA methylation probes through lm_METH =
TRUE. Arguments such as r_squared_thres or p_val_thres can be used to define the minimal R2 or the
p-value threshold to determine a significant result.
The next step is to estimate model parameters and generate a sample of BNs from posterior distribution:
> BN_mod_res <-
OMICS_mod_res, minseglen = 50000)
bn_module(burn_in = 100000,
thin = 500, OMICS_mod_res =
Now we can generate the diagnostic plots of the MCMC simulation and filter the most reliable edges in the
resulting network structure (in this example, we use 0.75 quantile of all edge weights as the edge weight threshold):
> trace_plots(mcmc_res = BN_mod_res, burn_in = 10000, thin = 500, edge_freq_thres = 0.75)
> res_weighted <- edge_weights(mcmc_res = BN_mod_res, burn_in = 10000, thin = 500,
edge_freq_thres = 0.5)
> weighted_net_res <- weighted_net(cpdag_weights = res_weighted, gene_annot =
gene_annot, PK = PK, OMICS_mod_res = OMICS_mod_res, gene_ID = ‘‘gene_symbol,’’
TFtargs = TFtarg_mat, B_prior_mat_weighted = B_prior_mat_weighted(BN_mod_res))
572
PACˇ I´NKOVA´ AND POPOVICI
ggraph_weighted_net() function is used to visualize the resulting network structure with the color scale
for all modalities used in the network structure inference:
> ggraph_weighted_net(net = weighted_net_res)
The resulting regulatory network can be seen in Figure 2. We can see several interactions known from
the biological prior knowledge, including interactions from CTNNB1 (b-catenin) to TCF4 and from TCF4
to CCND1. IntOMICS also identified the interaction between USP5 and FOXM1. On the contrary, the
interaction from CTNNBIP1 to CTNNB1 is not identified. CNV associated with GE is identified in several
genes, including tumor suppressor FBXW7. Some of them were identified as significantly deleted in ovarian
cancer, such as LEF1 or CTNNBIP1 (The Cancer Genome Atlas Research Network, 2011). METH-GE
interactions were identified in AXIN2 and LRP5 (previously reported as hypermethylated in EOC; Dai et al.,
2011). These results suggest IntOMICS identified interactions expected to be observed in EOC samples
with FOXM1 overexpression.
FIG. 2. Example of the IntOMICS output. GE features are denoted by upper case, CNV features are denoted by lower
case, and DNA METH features are denoted by methylation probe names (cgxxxx). PK, prior knowledge.
INTOMICS: REGULATORY NETWORKS FROM MULTI-OMICS
573
3. CONCLUSION
We present IntOMICS as a comprehensive and powerful tool for regulatory network inference using
multi-omics data. IntOMICS combines prior knowledge with data-derived evidence to advance regula-
tory networks inference. IntOMICS is designed to be easily extended by another modality. The current
implementation is tuned for gene expression, copy number variation, and DNA methylation data. However,
the user can infer regulatory network, even if copy number variation or DNA methylation data (or both) are
not available. IntOMICS is a powerful resource for exploratory systems biology and can provide valuable
insights into biological processes’ complex mechanisms that have a vital role in personalized medicine.
ACKNOWLEDGMENTS
The authors thank Research Infrastructure RECETOX RI [LM2018121] financed by the Ministry of
Education, Youth, and Sports, and Operational Programme Research, Development, and Innovation—
project CETOCOEN EXCELLENCE [CZ.02.1.01/0.0/0.0/17_043/0009632] for supportive background.
Access to computing and storage facilities owned by parties and projects contributing to the National Grid
Infrastructure MetaCentrum provided under the programme ‘‘Projects of Large Research, Development,
and Innovations Infrastructures’’ CESNET [LM2015042] is greatly appreciated.
This work was supported from the European Union’s Horizon 2020 research and Innovation program
under grant agreement No. 857560. This publication reflects only the author’s view, and the European
Commission is not responsible for any use that may be made of the information it contains.
AUTHORS’ CONTRIBUTIONS
Conceptualization, software, validation, formal analysis, data curation, writing—original draft, and
visualization by A.P. Conceptualization, methodology, writing—review and editing, supervision, project
administration, and funding acquisition by V.P.
AVAILABILITY
IntOMICS is open source software written in R, under GPL-2, and available at https://bioconductor.org/
packages/IntOMICS. The implementation relies on existing R packages.
AUTHOR DISCLOSURE STATEMENT
The authors declare that they have no conflicting financial interests.
FUNDING INFORMATION
This study was supported by Czech Science Foundation (GACR) through Grant No. 19-08646S and the
European Union’s Horizon 2020 research and innovation programme under Grant Agreement No. 825410.
REFERENCES
Chen, Y., Li, Y., Xue, J., et al. 2016. Wnt-induced deubiquitination FoxM1 ensures nucleus b-catenin transactivation.
EMBO J. 35, 668–684.
Cooper, G.F. 1989. Current research directions in the development of expert systems based on belief networks. Appl.
Stochast. Models Data Analysis. 5, 39–52.
Dai, W., Teodoridis, J.M., Zeller, C., et al. 2011. Systematic CpG Islands Methylation Profiling of Genes in the Wnt Pathway in
Epithelial Ovarian Cancer Identifies Biomarkers of Progression-Free Survival. Clin. Cancer Res. 17, 4052–4062.
574
PACˇ I´NKOVA´ AND POPOVICI
Geiger, D., and Heckerman, D. 1994. Learning gaussian networks, 235–243. Proceedings of the 10th Conference on
Uncertainty in Artificial Intelligence.
Hasin, Y., Seldin, M., and Lusis, A. 2017. Multi-omics approaches to disease. Genome Biol. 18, 83.
Kang, M., Ko, E., and Mersha, T.B. 2022. A roadmap for multi-omics data integration using deep learning. Brief.
Bioinform. 23, bbab454.
Lucas, P.J., van der Gaag, L.C., and Abu-Hanna, A. 2004. Bayesian networks in biomedicine and health-care. Artif.
Intell. Med. 30, 201–214.
Madigan, D., York, J., and Allard, D. 1995. Bayesian graphical models for discrete data. Int. Stat. Rev. Revue Int. De
Stat. 63, 215–232.
Neapolitan, R.E. 1990. Probabilistic Reasoning in Expert Systems: Theory and Algorithms. John Wiley & Sons, Inc.,
New York, NY, USA.
Ogata, H., Goto, S., Sato, K., et al. 1999. KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res. 27,
29–34.
Pa(cid:2)cı´nkova´, A., and Popovici, V. 2022. Using empirical biological knowledge to infer regulatory networks from multi-
omics data. BMC Bioinformatics. 23, 351.
Pearl, J. 1988. Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference. Morgan Kaufmann
Publishers Inc., San Francisco, CA, USA.
Subramanian, I., Verma, S., Kumar, S., et al. 2020. Multi-omics data integration, interpretation, and its application.
Bioinform. Biol. Insights. 14, 1–24.
The Cancer Genome Atlas Research Network. 2011. Integrated genomic analyses of ovarian carcinoma. Nature. 474,
609–615.
Zhang, W., Klinkebiel, D., Barger, C.J., et al. 2020. Global DNA hypomethylation in epithelial ovarian cancer: Passive
demethylation and association with genomic instability. Cancers. 12, 764.
Address correspondence to:
Mgr. Anna Pacˇı´nkova´
RECETOX
Faculty of Science
Masaryk University
Kotlarska 2
Brno 61137
Czech Republic
E-mail: [email protected]
| null |
10.1093_gbe_evad119.pdf
|
Data Availability
NCBI accession numbers for all genes analyzed are provided
in supplementary data file 1, Supplementary Material online.
Alignments and trees used in positive selection analyses are
provided in supplementary data file 2, Supplementary
Material online. Genotypes for great ape species at the
position of the rs2306894 human polymorphism were ob-
tained from: https://www.biologiaevolutiva.org/greatape/
data.html. For analyses of MelLec in human populations,
the following links were used to download or access the rele-
vant datasets:
• Sampling locations: https://www.internationalgenome.
org/data-portal/population.
• HGDP Chr12: https://ngs.sanger.ac.uk/production/hgdp/
hgdp_wgs.20190516/.
• 1KG Chr12: http://ftp.1000genomes.ebi.ac.uk/vol1/ftp/
data_collections/1000G_2504_high_coverage/working/
20201028_3202_phased/.
• Ancestral FASTA files for GRCh38 (homo sapiens ances-
tor GRCh38.tar.gz downloaded March 2023): https://ftp.
ensembl.org/pub/current_fasta/ancestral_alleles/.
home/snp/rs2306894.
FST, Tajima’s D, β(1) statistics, and statistics from linkage
disequilibrium analysis are provided in supplementary data
file 1, Supplementary Material online. AlphaFold-modeled
CLEC12A CTLD structures can be found on figshare at:
dx.doi.org/10.6084/m9.figshare.23535738.
|
Data Availability NCBI accession numbers for all genes analyzed are provided in supplementary data file 1, Supplementary Material online. Alignments and trees used in positive selection analyses are provided in supplementary data file 2, Supplementary Material online. Genotypes for great ape species at the position of the rs2306894 human polymorphism were obtained from: https://www.biologiaevolutiva.org/greatape/ data.html . For analyses of MelLec in human populations, the following links were used to download or access the relevant datasets: • Sampling locations: https://www.internationalgenome . org/data-portal/population . • HGDP Chr12: https://ngs.sanger.ac.uk/production/hgdp/ hgdp_wgs.20190516/ . • 1KG Chr12: http://ftp.1000genomes.ebi.ac.uk/vol1/ftp/ data_collections/1000G_2504_high_coverage/working/ 20201028_3202_phased/ . • Ancestral FASTA files for GRCh38 (homo sapiens ancestor GRCh38.tar.gz downloaded March 2023): https://ftp. ensembl.org/pub/current_fasta/ancestral_alleles/ . • eQTL analysis available from GTEx: https://gtexportal.org/ home/snp/rs2306894 . F ST , Tajima's D, β (1) statistics, and statistics from linkage disequilibrium analysis are provided in supplementary data file 1, Supplementary Material online. AlphaFold-modeled CLEC12A CTLD structures can be found on figshare at: dx.doi.org/10.6084/m9.figshare.23535738.
|
GBE
Rapid Evolution of Glycan Recognition Receptors Reveals
an Axis of Host–Microbe Arms Races beyond Canonical
Protein–Protein Interfaces
Zoë A. Hilbert
Ellen M. Leffler
1,2,*, Paige E. Haffener1, Hannah J. Young1,2, Mara J.W. Schwiesow1,2,
1, and Nels C. Elde1,2,*
1Department of Human Genetics, University of Utah, Salt Lake City, Utah, USA
2Howard Hughes Medical Institute, University of Utah School of Medicine, Salt Lake City, UT, USA
*Corresponding authors: E-mails: [email protected], [email protected].
Accepted: 23 June 2023
Abstract
Detection of microbial pathogens is a primary function of many mammalian immune proteins. This is accomplished through
the recognition of diverse microbial-produced macromolecules including proteins, nucleic acids, and carbohydrates.
Pathogens subvert host defenses by rapidly changing these structures to avoid detection, placing strong selective pressures
on host immune proteins that repeatedly adapt to remain effective. Signatures of rapid evolution have been identified in nu-
merous immunity proteins involved in the detection of pathogenic protein substrates, but whether similar signals can be ob-
served in host proteins engaged in interactions with other types of pathogen-derived molecules has received less attention.
This focus on protein–protein interfaces has largely obscured the study of fungi as contributors to host–pathogen conflicts,
despite their importance as a formidable class of vertebrate pathogens. Here, we provide evidence that mammalian immune
receptors involved in the detection of microbial glycans have been subject to recurrent positive selection. We find that rapidly
evolving sites in these genes cluster in key functional domains involved in carbohydrate recognition. Further, we identify con-
vergent patterns of substitution and evidence for balancing selection in one particular gene, MelLec, which plays a critical role
in controlling invasive fungal disease. Our results also highlight the power of evolutionary analyses to reveal uncharacterized
interfaces of host–pathogen conflict by identifying genes, like CLEC12A, with strong signals of positive selection across mam-
malian lineages. These results suggest that the realm of interfaces shaped by host–microbe conflicts extends beyond the
world of host–viral protein–protein interactions and into the world of microbial glycans and fungi.
Key words: host–pathogen interactions, evolutionary conflict, rapid evolution, balancing selection, pattern recognition
receptor, microbial glycans.
Significance
The impact of host–pathogen conflicts in driving evolutionary innovation in mammalian immune proteins is well docu-
mented; however, the role of nonprotein components of microbial pathogens in contributing to such evolutionary pro-
cesses is not well understood. We identify widespread signals of adaptive evolution in mammalian immune receptors
that engage largely with carbohydrate components that decorate the outer surfaces of diverse microbial pathogens,
from viruses to fungi. Further, we demonstrate how interactions involving nonproteinaceous components of microbes
have driven evolutionary change in mammalian genes across multiple timescales, including evidence for balancing se-
lection in a fungal melanin receptor gene in many human populations. Collectively, these findings extend the realm of
host–microbe evolutionary conflicts beyond traditionally studied protein–protein interfaces and demonstrate the im-
pressively broad impact microbes have on the evolution of their animal hosts.
© The Author(s) 2023. Published by Oxford University Press on behalf of Society for Molecular Biology and Evolution.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse,
distribution, and reproduction in any medium, provided the original work is properly cited.
Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023 1
Hilbert et al. GBE
Introduction
Recognition of microbial pathogens by mammalian immune
proteins is essential for activation of protective immune re-
sponses and organismal survival. Pattern recognition recep-
tors (PRRs) encompass a diverse group of host proteins
which are integral in detecting microbial pathogens as for-
eign invaders through recognition of unique molecular
features (Medzhitov 2007; Kawai and Akira 2008; Vance
et al. 2009; Tan et al. 2015). These pathogen-associated
molecular patterns are similarly as diverse as the receptors
that they engage with and range from proteins, like bac-
terial flagellins, to nucleic acids, to complex carbohy-
drates, or glycans.
Microbial glycans are a defining feature of the cell walls of
bacteria and fungi and decorate the outer membranes and
surfaces of parasites, whereas glycosylation of coat and sur-
face proteins is also well documented in many viruses
(Nyame et al. 2004; Comstock and Kasper 2006; van Kooyk
and Rabinovich 2008; Raman et al. 2016; Gow et al. 2017).
Glycan-recognizing PRRs include, among others, a subset of
the Toll-like receptors (TLRs) as well as many members of
the calcium-binding C-type lectin receptor (CLR) family.
Although the specific glycans recognized by some of these
PRRs are known—such as Dectin1’s affinity for ß-glucans or
TLR4’s for lipopolysaccharide—for many of these receptors,
the exact molecular patterns on microbial surfaces required
for recognition are unclear, as is the extent to which variation
of these patterns among different microbial species might af-
fect recognition (Poltorak et al. 1998; Brown and Gordon
2001; Herre et al. 2004; Park et al. 2009; Werling et al. 2009).
Phylogenetic analysis of immune genes, including PRRs,
has revealed them to be among the most rapidly evolving
genes in mammalian genomes, reflecting the pace of evolu-
tion needed to keep up with constantly shape-shifting patho-
gens (George et al. 2011; Daugherty and Malik 2012; Rausell
and Telenti 2014; Wang and Han 2021). Studies of rapidly
evolving immune genes in mammals have largely focused
on genes involved in interactions with pathogen-produced
protein factors. Comparative analyses of recurrent rapid evo-
lution (or positive selection) on the amino acid level frequently
reveal the consequential interaction interfaces between host
and pathogen proteins. Related experimental studies show
how evolution on both sides of these interactions can have
functional implications for both host and pathogen (Sawyer
et al. 2005; Elde et al. 2009; Mitchell et al. 2012; Barber
and Elde 2014; Tenthorey et al. 2020; Carey et al. 2021).
These studies reveal the extent to which microbes can spur
diversification and evolutionary innovation in the hosts they
infect. However, detection of these host–pathogen “arms
races” has so far been primarily limited to protein–protein in-
terfaces involving viruses and bacteria, even though engage-
ment between hosts and infectious microbes involves a wide
variety of biological macromolecules and species.
Fungi, in particular, represent a major class of human
pathogens which are currently auspiciously absent from stud-
ies of host–pathogen evolutionary conflict. Systemic fungal
infections are associated with severe disease and high mortal-
ity rates in human patients and the emergence of multidrug
resistant strains has increased dramatically in recent years
(Fisher et al. 2022). Beyond human patients, fungal infections
pose a severe threat to the health of food crops, and fungal
pathogens are currently responsible for massive declines in
amphibian and hibernating bat populations world-wide
(Fisher et al. 2020). Despite the importance of these patho-
gens for the health of evolutionarily diverse organisms, our
understanding of the role of host–fungal conflicts in shaping
vertebrate immune defenses has been hampered by the rela-
tive lack of known protein-based fungal virulence factors.
As the first line of defense against recognition by host im-
mune factors, diversification in microbial cell wall components
and organization has been well documented in bacterial and
fungal pathogens (Gow et al. 2017; Imperiali 2019). Further,
molecular mimicry of host glycan structures, such as sialic
acids, and hijacking of glycosylation pathways has been de-
monstrated to be a common mechanism of immune evasion
in numerous pathogenic bacteria and viruses (Comstock and
Kasper 2006; Vigerust and Shepherd 2007; Carlin et al. 2009;
Varki and Gagneux 2012; Raman et al. 2016). Although gly-
can hijacking and mimicry in fungi is less well documented, re-
ports of sialic acids and sialoglycoconjugates in the cell walls
of several fungal species, including the pathogenic species
Candida albicans and Cryptococcus neoformans, suggest
that fungi may also use methods of molecular mimicry to
evade host immune recognition (Rodrigues et al. 1997;
Soares et al. 2000; Masuoka 2004). And in fungi, regulated
secretion of exopolysaccharide “decoys” correlates with de-
creased immune infiltration, suggesting these microbes have
developed numerous strategies to prevent their recognition
by host immune systems (Denham et al. 2018).
Such evasion strategies among microbes suggest the po-
tential for selective pressures to exist on immune receptors
to be able to maintain the ability to recognize microbial gly-
cans and initiate immune responses to control infection. In
this study, we identify signatures of positive selection in a
set of primarily glycan-recognizing PRRs across three dis-
tinct mammalian lineages, suggesting that host–pathogen
interfaces involving nonproteinaceous macromolecules
may represent a new dimension of host–microbe arms
races and can spur evolution in all species involved.
Results
Signatures of Rapid Evolution Are Pervasive Among
Mammalian CLRs and Other Carbohydrate Recognition
PRRs
To assess whether host genes involved in microbial carbo-
hydrate recognition are rapidly evolving in mammals, we
2 Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023
Rapid Evolution of Glycan Recognition Receptors GBE
compiled a list of 26 relevant genes for analysis (fig. 1A
and B and Supplementary Material online). These genes
were selected based on annotated functions in the recogni-
tion of microbial cell walls or other carbohydrate compo-
nents of microbial cells. Genes were also prioritized for
analysis based on documented expression patterns.
Namely, genes expressed by immune cells or on mucosal
surfaces were prioritized given their relevance for interac-
tions with microbes and defense against infection.
More than half of the selected PRR genes contain an an-
notated C-type lectin domain (CTLD), including a number
of CLR family members with a single CTLD (e.g., Dectin1/
CLEC7A, Langerin/CD207/CLEC4K, Mincle/CLEC4E) as
well as the soluble CTLD-containing proteins (MBL2,
SP-A, SP-D) and the multiple CTLD-containing mannose re-
ceptors (MRC1 and MRC2). Beyond the CLRs and other
CTLD-containing proteins, our list also included a putative
chitin receptor (FIBCD1), complement receptor 3 (CD11B/
CD18), and TLRs (TLR2 and TLR4). Among this latter group,
there have been previous reports of signatures of positive
selection in the TLR genes as well as CD11B, which we
were able to replicate in this study, while also extending
analyses of selection in these genes to additional mamma-
lian lineages (Wlasiuk and Nachman 2010; Areal et al.
2011; Liu et al. 2019; Boguslawski et al. 2020; Judd et al.
2021). Finally, we also included in our analyses the CTLDs
of three conserved mammalian selectin genes: E-Selectin,
L-Selectin, and P-Selectin. These CTLD containing proteins
are expressed on a variety of different cell types and act
to coordinate cell adhesion and leukocyte trafficking
through recognition of “self”-produced carbohydrate li-
gands or self-associated molecular patterns (SAMPs)
(Varki 2011; Cummings et al. 2022). Given their important
role in recognition of these SAMPs on leukocytes and other
mammalian cells and no documented role in the recogni-
tion of microbes, we hypothesized that the CTLDs from
these Selectin genes would not be subject to the same evo-
lutionary pressures as other candidate genes involved in dir-
ect interactions with infectious microbes.
For each of these genes, we obtained orthologous se-
quences from publicly available databases for species with-
in three distinct mammalian lineages: simian primates,
mouse-like rodents (Myomorpha), and bats. Primates
were chosen given their relevance to human health, where-
as bats and rodents have been implicated as important re-
servoirs for many microbial species with zoonotic potential,
suggesting that such evolutionary analysis may reveal un-
ique patterns of selection among PRRs across these three
mammalian lineages (Han et al. 2015; Guth et al. 2022).
The orthologous gene sequences within each lineage
were aligned and each gene was assessed for signals of re-
current positive selection using a combination of different
analysis algorithms, including Phylogenetic Analysis by
Maximum Likelihood (PAML) and Branch-Site Unrestricted
in
(BUSTED)
for Episodic Diversification
Test
the
Hypothesis Testing using Phylogenies (HyPhy) suite (Pond
et al. 2005; Yang 2007; Murrell et al. 2015). Both algo-
rithms use the calculation of the ratio of the nonsynon-
ymous to synonymous substitution rates (dN/dS) and
model fitting comparisons in order to make inferences
about signatures of selection across genes and phylogenies.
For genes or codons under purifying selection, nonsynon-
ymous substitutions are selected against, leading to dN/
dS values less than 1. In contrast, positive selection—or ra-
pid evolution—is characterized by the relative enrichment
of nonsynonymous substitution rates, which can be identi-
fied by elevated dN/dS values (>1) in these genes or at spe-
cific codons within genes.
Using the site models implemented in PAML along with
BUSTED, we identified signatures of site-specific positive se-
lection by at least one of the two algorithms (BUSTED P <
0.05 or PAML M7 vs. M8 likelihood-ratio test [LRT] P <
0.05) in nine (35%) of the primate PRRs (fig. 1 and
supplementary data file 1, Supplementary Material online).
This number was strikingly elevated among the rodent and
bat lineages, with 16 (62%) and 21 (81%) genes under
positive selection in these groups, respectively. Mapping
these positively selected genes onto a phylogenetic tree
of the CTLDs from the CLR-type PRRs revealed no clear pat-
tern to the distribution of positive selection across this fam-
ily of receptors (fig. 1B and supplementary fig. S1B,
Supplementary Material online). Instead, rapid evolution
seems pervasive across the entire family of CLRs that
were analyzed.
Through these approaches, we identified a core set of six
PRRs predicted to be under positive selection by one or both
algorithms in all mammalian lineages tested. These core
genes include those, such as TLR4, with long-established
roles in microbial recognition and previously defined li-
gands. However, this core group, surprisingly, also includes
the CLR gene CLEC12A, whose role in interactions with mi-
crobes is still emerging, pointing to the possibility of as yet
undefined, but important, roles for this CLR in microbial
recognition. Beyond the shared signatures of positive selec-
tion across lineages, these core rapidly evolving PRRs also
tended to have a higher number of sites predicted to be un-
der positive selection, with many of the rapidly evolving
amino acid residues falling into functionally relevant re-
gions of these receptors, namely the extracellular
carbohydrate-binding domains.
Outside of this core set of positively selected genes, we
observed lineage-specific patterns of positive selection
among the remaining PRRs. These different patterns of se-
lection across the three mammalian lineages suggest the
possibility that distinct populations of microbial species
may have played a role in shaping the evolution of these
mammalian receptors. Importantly, our analyses of the
CTLDs of mammalian selectins revealed little evidence for
Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023 3
Hilbert et al. GBE
FIG. 1.—Positive selection across mammalian carbohydrate recognition PRRs. (A) Positive selection analyses of 26 glycan PRRs in primates (left column),
rodents (middle), and bats (right column). Colored boxes indicate whether evidence of positive selection was supported by PAML analyses only (medium blue)
or by both PAML and BUSTED analyses (dark blue). Genes with no evidence for positive selection are represented by pale blue boxes. Statistical cutoffs were
P < 0.05 for PAML M7 versus M8 likelihood ratio tests and for BUSTED analysis. (B) Patterns of positive selection mapped onto a phylogenetic tree of the
human CTLD domains. Only genes from the gene set with CTLDs are represented. Colored circles represent evidence of positive selection in the primate
(orange), rodent (purple), and/or bat (blue) lineages. Genes with black circles were not analyzed in this study because of unclear ortholog relationships across
mammals but do have important roles in pathogen detection in mammals. Numbers indicate bootstrap values from phylogenetic tree construction using
IQ-TREE.
4 Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023
AB Rapid Evolution of Glycan Recognition Receptors GBE
positive selection in these genes with high levels of conser-
vation across lineages. This further underscores the role of
microbial pathogen interactions in driving the evolutionary
signatures we observe across this gene set of PRRs.
Rapidly Evolving Codons in Mammalian Langerin
(CD207) Correspond with Amino Acid Positions at Key
Ligand Recognition Interfaces
The set of PRRs under positive selection in all three of the
tested mammalian lineages includes Langerin (CD207), a
CLR expressed primarily by the Langerhans cells of the
skin as well as other professional antigen presenting cells.
Langerin has an established role in the activation of critical
inflammatory responses following direct detection of di-
verse microbial pathogens, including fungi, viruses, and
bacteria (de Witte et al. 2007; de Jong et al. 2010; van
der Vlist et al. 2011; van Dalen et al. 2019). In particular,
Langerin has been shown to be able to recognize and
bind to ß-glucans in Candida species as well as the
skin-associated fungal species Malassezia furfur (de Jong
et al. 2010). Bacterial recognition by Langerin has been ob-
served for multiple species, including Staphylococcus aur-
eus, a major cause of skin infections (Yang et al. 2015;
van Dalen et al. 2019). In the context of both fungal and
S. aureus infection, Langerin has been shown to play a
role in regulating inflammatory Th17 responses (Sparber
et al. 2018; van Dalen et al. 2019). Structural studies of hu-
man Langerin have revealed it to have a canonical CLR fold,
with a Glu-Pro-Asn (EPN) motif in the primary ligand bind-
ing site, suggestive of a ligand preference for mannose and
mannose-type carbohydrates (Tateno et al. 2010; Feinberg
et al. 2011; Hanske et al. 2017). Interestingly, recent work
examining the ligand-binding profiles of Langerin homo-
logs from humans and mice identified distinct differences
in the binding specificities for more complex bacterial-
derived glycans among these homologs, despite conserva-
tion of the EPN motif in the binding site (Hanske et al.
2017). This suggests that sequence variation in the
Langerin CTLD may play an important role in modulating
microbial recognition.
To determine whether the signals of rapid evolution that
we observe in Langerin across mammalian lineages might
functionally correlate with differences in ligand preference,
we first mapped the sites under positive selection in each
lineage to the annotated protein domains (fig. 2A). A large
proportion of positively selected sites in all three lineages
mapped to the extracellular region of the protein, with
many falling into the CTLD itself, including several overlap-
ping amino acid positions which were predicted to be un-
der positive selection in all three mammalian lineages.
In addition to the PAML algorithm, we also used the
HyPhy suite programs mixed effects model of evolution
(MEME) and fast unbiased Bayesian approximation
(FUBAR) to independently assess individual amino acid sites
for elevated dN/dS values across the Langerin coding se-
quence (Murrell et al. 2012, 2013). Although MEME, like
PAML, assesses patterns of episodic selection occurring
on at least one branch of the phylogeny, the FUBAR algo-
rithm can be used to identify sites under pervasive positive
selection across an entire phylogeny. These additional ana-
lyses, thus, provide both confirmatory and complementary
methods to PAML for assessing site-specific rapid evolution.
Agreement between the three algorithms was high across all
positively selected sites in Langerin (fig. 2B). In particular,
amino acid positions 213 and 289, which were identified
by PAML analyses in all three lineages, showed signatures
of positive selection in the MEME and FUBAR analyses in
both primates and bats. Similarly, multiple methods inde-
pendently highlighted position 313 as rapidly evolving in
bats and rodents, in agreement with the PAML analyses of
primate sequences. Rapid evolution of other lineage-specific
sites was also supported by all three analyses (fig. 2B).
The convergence of these signatures of rapid evolution
on the Langerin CTLD and these three residues (213, 289,
and 313) across multiple mammalian lineages hints at pos-
sible functional significance to amino acid changes at these
positions. When mapped onto a crystal structure of the
Langerin CTLD in complex with a mannose ligand and a co-
ordinating calcium ion, we observed that many of the resi-
dues under positive selection clustered around the ligand
binding site (fig. 2C). This supports the hypothesis that vari-
ation at these positions across mammalian Langerin homo-
logs might result in differences in microbial glycan binding
specificities. Furthermore, this suggests the possibility that
the signals of rapid evolution we observe in mammalian
Langerin homologs was driven by the selective pressure
to maintain the ability to recognize specific microbial spe-
cies through distinct microbial glycans on their surfaces
and in their cell walls.
Mapping Patterns of Substitution in an Invasive
Aspergillosis Susceptibility Allele of MelLec (Melanin
Lectin/CLEC1A) across Primates
Unlike many CLRs, which can recognize similar ligands
present on many different species of microbes, MelLec
(also known as CLEC1A), was recently identified as being
a highly specific receptor for 1,8-dihydroxynaphthalene
(DHN)-melanin, a critical component of the cell walls of
a relatively limited group of fungal species (Stappers
et al. 2018). Included in these DHN-melanin-producing
fungi are the human fungal pathogens Aspergillus
fumigatus and the black yeasts, which account for
significant morbidity and mortality in both immune-
suppressed and immunocompetent patients worldwide
(Brown et al. 2012; Seyedmousavi et al. 2014).
Recognition of DHN-melanin in fungal cells via MelLec
Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023 5
Hilbert et al. GBE
FIG. 2.—Diversification of Langerin (CD207) ligand-binding interfaces in all mammalian lineages. (A) Positively selected residues (triangles) predicted by
PAML (Model 8, BEB > 0.9) cluster primarily in the extracellular portion of Langerin (CD207), with many in the CTLD. A number of positively selected sites in the
CTLD are common across primates (orange triangles), rodents (purple triangles), and bats (blue triangles). (B) Agreement between different algorithms for
identifying site-specific positive selection in Langerin of different mammalian groups. Listed residue numbers correspond to the position in the human
Langerin sequence. Single letter residues correspond to the amino acid identity in human (primates, left), house mouse (rodents, middle), or black flying
fox (bats, right) sequences. Bolded residues are those predicted to be under positive selection across all mammals by one or more tests. (C) Positively selected
sites mapped onto a crystal structure of the human Langerin CTLD (gray, PDB:3p5d) in complex with a mannose ligand (yellow) and Ca2+ ion (magenta)
(Feinberg et al. 2011). Positively selected sites in all three lineages (colored in green) along with several sites from rodent (blue) and bat (purple) analyses
are shown with sidechains and surround the ligand binding site.
has been demonstrated to be critical for the activation of
an antifungal immune response and survival of systemic
A. fumigatus infection in in vivo models. Notably, a com-
mon human polymorphism causing a single amino acid
change (Gly26Ala, rs2306894) has been identified in
the cytoplasmic region of the MelLec protein. This
Ala26 allele has been associated with higher probability
of invasive Aspergillosis in transplant patients and has
also been shown to result in decreased production of crit-
ical cytokines in response to fungal stimulation in in vitro
experiments (Stappers et al. 2018). Combined, these data
support a role for MelLec in the immune responses to fun-
gal infection in both mice and humans.
Our PAML analyses revealed signatures of recurrent posi-
tive selection in MelLec in both the primate and rodent
lineages (fig. 1). Although significance by LRT varied for pri-
mate analyses of MelLec depending on whether a species
or gene tree was used in the analysis, manual inspection
of the alignments revealed extensive sequence variation
at PAML-identified sites across the primate MelLec
orthologs (see Methods and supplementary data file 1,
Supplementary Material online). This suggests that interac-
tions between these mammalian groups and pathogenic
fungi may have played a role in shaping amino acid diversi-
fication in this PRR. Furthermore, the rapidly evolving amino
acids within MelLec include several in the CTLD, consistent
6 Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023
ABC Rapid Evolution of Glycan Recognition Receptors GBE
with the potential for sequence variation to confer changes
in ligand-binding affinity or specificity among different
MelLec
1,
homologs
Supplementary Material online).
(supplementary
data
file
While mapping the positively selected sites in primate
MelLec orthologs, we were surprised to find that at the
site of the human polymorphism, Gly26, we observed a con-
served alanine residue in all primates except humans and
black-capped squirrel monkeys (Saimiri boliviensis bolivien-
sis, fig. 3A). This suggests that Gly26 likely represents the de-
rived human allele, while alanine is the ancestral allele
among primates. Whether the alanine at position 26 in other
primate homologs confers the same defects in cytokine pro-
duction observed for the human allele is presently unknown.
Although it is possible that sequence variation elsewhere in
the primate MelLec homologs might compensate for the
alanine at position 26, future experimental studies will be
needed to assess how sequence variation at this and other
sites contribute to function of the MelLec receptor.
We next explored the distribution of these two MelLec al-
leles in human populations. Across human populations in
the 1000 Genomes Project (1KG) dataset, the frequency of
the derived Gly26 allele varies widely, from only 0.11 in
African (AFR) and 0.13 in European (EUR) populations to
0.65 in East Asian (EAS) populations (fig. 3C) (The 1000
Genomes Project Consortium 2015). Given the high fre-
quency of the Gly26 allele in EAS populations, we turned
to two additional resources to more comprehensively assess
the distribution of this allele across Asia (GenomeAsia100K
Consortium et al. 2019; Bergström et al. 2020). Using the
Genome Asia 100K Browser and the Human Genome
Diversity Project (HGDP), we observed that the Gly26 allele
reached even higher frequencies in Oceanic (OC) and
Southeast Asian (SAS) populations that were not repre-
sented in the 1000 Genomes dataset. The Gly26 allele was
fixed in the populations from Papua New Guinea in the
HGDP, though the sample size was small (n = 17) and at
an allele frequency (AF) of 0.77 in PNG in the Genome
Asia 100K dataset (n = 70) (fig. 3C). The HGDP also revealed
a high frequency of the Gly26 allele in multiple American
(AMR) populations (e.g., AF = 1 in Colombian, AF = 0.94 in
Karitiana and AF = 0.95 in Pima), which may reflect the
shared ancestry between native American and Asian popula-
tions. To quantify the allele frequency differences observed
across these populations, we calculated pairwise FST be-
tween EUR populations (with low Gly26 frequencies) and
the OC, SAS, and AMR populations and tested for signifi-
cance relative to other single nucleotide polymorphisms
(SNPs) on chromosome 12 (supplementary data file 1,
Supplementary Material online). FST was high between all
tested populations, falling in the tail of the empirical distribu-
tions, indicating an elevated signal of differentiation consist-
ent with the allele frequency differences observed between
these groups.
The extreme population differentiation of
the
rs2306894 Gly26Ala SNP could reflect that this locus has
been a target of selection in human populations. Both posi-
tive and balancing selection can affect population differen-
tiation and FST values. We first assessed whether rs2306894
or any other SNPs in MelLec showed signatures of local
positive selection. Both searches of published scans for re-
cent positive selection focusing on Asian populations as
well as our own analysis of the Colombian population
from the 1KG database using Relate showed no evidence
for positive selection in MelLec in human populations
(supplementary fig. S2, Supplementary Material online)
(Voight et al. 2006; Liu et al. 2017; Speidel et al. 2019,
2021). Next, we calculated Tajima’s D in 1 kb windows
across all of Chromosome 12 in each population from the
HGDP and 1KG datasets. Notably, we observed elevated
Tajima’s D values for the window containing MelLec and
rs2306894 in the majority of the tested populations, with
a significantly positive value in 31 of 62 populations as-
sessed (empirical P < 0.05), suggestive of balancing selec-
tion acting at this locus (fig. 3C, middle). To further
confirm this, we ran BetaScan, a more sensitive method
for detecting balancing selection, where high β(1) statistics
are indicative of an excess of SNPs at similar frequencies,
a key feature of genomic regions under balancing selection
(Siewert and Voight 2017, 2020). The β(1) statistic was
significantly elevated (empirical P < 0.05) for MelLec in all
of the 1KG populations except for the AFR populations,
further suggesting that this gene has been subject to
balancing selection in many human populations (fig. 3C,
bottom).
is
in perfect
the selective signatures we
It is important to note that while previous functional
studies have focused solely on the Gly26Ala SNP, our ana-
lyses revealed that this SNP
linkage
disequilibrium (LD) with a large number of other SNPs with-
in MelLec (e.g., 42 SNPs in r2 = 1 with rs2306894 in EAS,
spanning 8 kb) making it challenging to distinguish the tar-
identify here
get of
(supplementary data file 1, Supplementary Material online).
The vast majority of these SNPs fall into intronic regions and
are documented eQTLs for MelLec in multiple tissues in the
Genotype-Tissue Expression (GTEx) project (Lonsdale et al.
2013). Two of these SNPS in LD with rs2306894 fall within
regulatory regions which could have direct regulatory ef-
fects on expression of MelLec: rs2306893 in the 5′UTR
and rs2277416 in a splice region. Future studies probing
the effects of these SNPs on MelLec function may further
our understanding of how they individually or collectively
contribute to fungal disease and reveal a more nuanced un-
derstanding of the target of the balancing selection signa-
tures we observe.
Beyond humans, we also noted that the black-capped
squirrel monkey sequence from the NCBI GenBank data-
base carried a valine at position 26, in contrast to the
Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023 7
Hilbert et al. GBE
FIG. 3.—Single nucleotide polymorphisms in primate populations converge on a single site in Melanin Lectin (CLEC1A). (A) Patterns of conservation and
variation at amino acid position 26 of MelLec across primates. Most primate species carry the ancestral alanine allele (orange highlighting), whereas single
nucleotide polymorphisms in both humans (glycine, green highlighting) and squirrel monkeys (valine, pink highlighting) confer missense mutations. (B)
Genotypes of 19 unrelated squirrel monkey gDNA samples from three S. boliviensis subspecies. The sex and the amino acid identity at position 26 for
each individual are indicated, with heterozygous individuals indicated as carrying both Ala and Val amino acids (A/V in Black-capped and Peruvian squirrel
monkeys). (C) (top) Geographic distribution of the glycine 26 allele (green) at SNP rs2306894 in human populations. Allele frequencies are shown for popula-
tions from the 1KG Project and the HGDP. Individuals carrying the Ala26 allele (orange) have been previously shown to have higher risk of invasive fungal
infections in stem-cell transplant patients (Stappers et al. 2018). (middle) Tajima’s D values for populations from the HGDP and 1KG and (bottom) β(1) for
populations from the 1KG project showing evidence of balancing selection at the MelLec locus. For both plots, * empirical P-value < 0.05, ** empirical
P-value < 0.01. Population abbreviations are as follows: AMR, America; AFR, Africa; EUR, Europe; CSA, Central-South Asia; ME, Middle East; SAS, South
Asia; EAS, East Asia; OC, Oceania.
8 Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023
ABC Rapid Evolution of Glycan Recognition Receptors GBE
alanine of all other primates (fig. 3A). To confirm this obser-
vation and investigate the patterns of substitution at this
position among squirrel monkey populations, we amplified
the region surrounding this SNP from multiple genomic
DNA (gDNA) samples from black-capped squirrel monkeys
(S. boliviensis boliviensis) as well as two other closely related
squirrel monkey subspecies: Peruvian squirrel monkeys
(S. boliviensis peruvinsis) and Guianan squirrel monkeys
(S. sciureus sciureus). In total, we genotyped 19 unrelated
individuals from these three subspecies. Interestingly, the
Guianan squirrel monkeys were universally homozygous
for the ancestral Ala26 allele, whereas no individuals homo-
zygous for this allele could be found in the other two sub-
species (fig. 3B and supplementary fig. S3, Supplementary
Material online). Among black-capped and Peruvian squir-
rel monkeys, there was a mix of individuals homozygous for
the derived Val26, as well as heterozygous individuals,
again raising intriguing questions about the potential se-
lective pressures that have shaped allele frequency distribu-
tions in squirrel monkeys as in humans.
To rule out the possibility that the lack of observed se-
quence variation in other primates might be due to sam-
pling bias of the publicly available sequences in GenBank,
we also looked for variation at this locus among hominoid
primates using data from the Great Ape Genome Project
(Prado-Martinez et al. 2013). There was no evidence in
these data for any sequence variation at amino acid pos-
ition 26 in gorillas, bonobos, chimpanzees, or orangutans
(supplementary data file 1, Supplementary Material online).
Combined, these data strongly suggest that mutation of
this locus has occurred independently in humans and squir-
rel monkeys, perhaps due to similar evolutionary pressures
in these species from fungi or other microbial species.
Extensive Positive Selection across CLEC12A in Primates,
Bats, and Rodents Portends an Unidentified Role in
Microbial Recognition and Binding
In addition to genes with well-established roles in immune
responses to microbial pathogens, our analyses also re-
vealed extensive positive selection occurring at sites within
the CLEC12A gene, a more mysterious member of the CLR
family of receptors. Originally identified as a receptor for
uric acid, a marker of cell death, other reports have identi-
fied roles for this receptor in the recognition of hemozoin
produced by Plasmodium spp. during infection as well as
in the regulation of antibacterial autophagy responses
(Neumann et al. 2014; Begun et al. 2015; Raulf et al.
2019). Most recently, CLEC12A, has been shown to directly
bind to a number of gut-resident bacteria and is required
for the phagocytosis of these bacteria and subsequent
modulation of microbiome community composition
(Chiaro et al. 2023). Although the exact moiety that
CLEC12A engages remains undefined, these data strongly
suggest the possibility that CLEC12A is also capable of rec-
ognizing molecular patterns found in the bacterial cell wall,
including bacterial glycans. Given the breadth of the cur-
rently known ligands and roles of CLEC12A and its expres-
sion predominantly in myeloid cells, it is likely that the full
scope and nature of the interfaces between CLEC12A
and pathogenic microbes has not yet been revealed.
Further supporting this idea, our phylogenetic analyses of
CLEC12A revealed strong signals of positive selection on
this gene across all mammalian lineages, suggestive of
strong selection imposed on this gene by interactions
with, perhaps, diverse pathogens (fig. 4). In fact, in both
bats and primates, the gene-wide dN/dS calculated by
PAML was >1 (supplementary data file 1, Supplementary
Material online). CLEC12A was the only gene analyzed in
this study for which this was true and supports the model
that CLEC12A is evolving under remarkably strong positive
selection in mammals.
Although positively selected sites were distributed across
the entire coding sequence of CLEC12A, a large number fall
directly in the CTLD, a pattern which is most pronounced in
primates (orange triangles, fig. 4A). Many of these sites
were independently predicted to be rapidly evolving by
PAML, MEME and FUBAR and tend to cluster in the same
regions in all three mammalian groups, suggesting these
may be regions important for the immune or ligand binding
functions of the protein (supplementary data file 1,
Supplementary Material online). Given the large number
of sites under positive selection in the CTLD, no discernable
patterns emerged from mapping these sites onto
AlphaFold-predicted structures of CLEC12A CTLD homo-
logs from different species that might hint at effects of se-
quence diversification on ligand binding. Of note, however,
was the fact that despite the primary sequence divergence
across mammals, there were no significant differences in
the AlphaFold-predicted structures of primate, rodent and
bat homologs suggesting that more subtle modifications
in structure may underlie any functional differences be-
tween homologs (supplementary fig. S4, Supplementary
Material online).
To identify specific rapidly evolving branches in each
mammalian lineage, we applied models implemented in
PAML that allow calculation of dN/dS for each branch of
a given phylogenetic tree (fig. 4B–D). This temporal view
of the evolution of CLEC12A revealed extensive episodic
positive selection across each of the mammalian phyloge-
nies. Among the simian primates, all three major groups
(Hominids, Old World, and New World Monkeys) contained
branches with elevated dN/dS values, though these values
were slightly higher among both the ancient and recent
branches in the hominid and New World Monkey lineages
(fig. 4B). Similar patterns can be seen in the rodent and bat
phylogenies, where positive selection was also rampant
(fig. 4C and D). Consistent with the elevated gene-wide
Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023 9
Hilbert et al. GBE
FIG. 4.—Extensive positive selection in CLEC12A across mammals reveals a new host–pathogen battleground. (A) Diagram showing sites under positive
selection in CLEC12A in primates (orange triangles), rodents (purple triangles) and bats (blue triangles). Indicated sites were predicted by PAML (Model 8, BEB
> 0.9). Locations of the CTLD and transmembrane domain are indicated on the left. (B)–(D) dN/dS values for CLEC12A were calculated across the species
phylogenies of primates (B), rodents (C), and bats (D) using PAML (free ratios, Model = 1 setting). Lineages with elevated dN/dS values (>1), suggestive of
positive selection along that branch, are indicated with colored lines. Calculated dN/dS values are listed above each branch and for branches lacking either
nonsynonymous or synonymous sites; ratios of the respective substitution numbers (N:S) are indicated.
dN/dS value observed for bat CLEC12A (dN/dS = 1.2,
supplementary data file 1, Supplementary Material online),
especially high substitution rates were abundant across the
bat phylogeny, and in particular among the new world leaf-
nosed bats (Phyllostimidae), a group which includes the
spear-nosed bats, Jamaican fruit bat and the Honduran
yellow-shouldered bat (fig. 4D). Combined, the strength
of the signals of rapid evolution that our analyses revealed
10 Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023
ABCD Rapid Evolution of Glycan Recognition Receptors GBE
in CLEC12A across multiple mammalian lineages, suggest it
functions as an underappreciated but critical component in
the arsenal of immune receptors that engage with micro-
bial pathogens and play a role in immune defenses against
infection. Although it is theoretically possible that the sig-
nals we observe in CLEC12A have been driven by already
identified ligands and interactions, we hypothesize that
interactions between
there are
CLEC12A and other microbial species for which this se-
quence variation will have functional implications.
likely undiscovered
Discussion
Our study revealed widespread signatures of rapid evolu-
tion across glycan-recognition PRRs in three major mamma-
lian lineages: primates, rodents, and bats. Such strong
signatures of positive selection are frequently associated
with host–pathogen arms races, signifying the consequen-
tial impacts on fitness associated with these interactions.
We hypothesize that the evolutionary signatures we ob-
serve among CLRs and related factors represent a new
axis in these arms races where hosts keep pace with the nu-
merous and well-studied evasive strategies microbes use to
prevent detection of their immunogenic glycan-rich sur-
faces. Consistent with this hypothesis, we found that posi-
tive selection among these genes is often enriched in
functionally significant portions of the protein, namely
in the CTLDs which directly interact with glycans. In
Langerin, this pattern was particularly clear, with a cluster
of rapidly evolving residues falling directly surrounding
the ligand binding pocket of the CTLD (fig. 2C). Positively
selected sites in Langerin include amino acid position 313,
which has previously been determined to contribute signifi-
cantly to ligand binding, with mutations at this position re-
sulting in a complete lack of recognition of certain simple
carbohydrate ligands (Tateno et al. 2010). Across all the
mammalian species we analyzed in this study, we observed
eight different amino acids sampled at this position, a find-
ing that strongly points to functional differences in ligand
binding and specificity.
The finding that the highly specific DHN-melanin binding
MelLec receptor is rapidly evolving in both primates and
rodents is particularly exciting. To date, studies of
host–microbe evolutionary arms races have largely involved
only interactions with viruses or bacteria; the role of eukary-
otic pathogens, such as fungi, in shaping the evolution of
mammalian host species has remained unexplored. Rapid
evolution in MelLec across species when paired with the
emerging patterns of substitution at a functionally import-
ant site in both humans and squirrel monkeys strongly sug-
gests that fungi can, in fact, play an important role in
shaping the evolution of mammalian immune systems.
Additionally, many of the other PRRs identified as rapidly
evolving in this study also engage with fungal pathogens,
suggesting that the breadth of host proteins shaped by in-
teractions with pathogenic fungi may be extensive.
Our population genetics analyses of the human MelLec
Gly26Ala SNP further revealed strong population differenti-
ation in the allele frequencies of this SNP along with signals
of balancing selection within this locus in many human popu-
lations. This raises several intriguing hypotheses: first, that dif-
ferent association with fungal species across geographic
regions might partially account for the allele frequency differ-
ences observed across human populations. Other factors,
such as lifestyle and/or dietary differences across human po-
pulations could also play a role in driving the population dif-
ferentiation we observe. Whether and how these different
pressures shaped the distribution of these MelLec alleles in
human populations remains a fascinating challenge to dis-
sect. A second hypothesis that arises from our population
genetic analysis of MelLec suggests that although the
Gly26 allele appears to be protective under some circum-
stances, there may be tradeoffs associated with changes at
this position, reflected in the maintenance of the ancestral
Ala26 allele in human populations and the signals of balan-
cing selection we observe. Indeed, although MelLec is essen-
tial for protection against invasive disease caused by fungal
species like A. fumigatus, its function was shown to be detri-
mental in in vivo models of asthma driven by the same fungal
species suggesting that MelLec activity has a complex impact
on establishing appropriate immune responses to fungi
(Stappers et al. 2018; Tone et al. 2021). Whether and how
mutation of position 26 (or other sites) within the MelLec lo-
cus might contribute to these differing outcomes remains to
be seen but may provide some insight into the signals of bal-
ancing selection we observe in this gene.
Previous analysis of carbohydrate-ligand binding in dif-
ferent mammalian Langerin homologs led to the surprising
finding that although specificity in ligand binding for simple
carbohydrates was similar across different Langerin var-
iants, dramatic differences were observed in the context
of complex carbohydrates and intact bacterial cells
(Hanske et al. 2017). These differences were identified des-
pite high conservation in the solved crystal structures of the
CTLDs from these homologs, suggesting that more subtle
structural or sequence variation underlies variability in lig-
and binding. Our analyses of the CLEC12A gene suggest
this may be a general feature among these rapidly evolving
CLRs. In CLEC12A, we observed extensive diversification of
the primary sequence in all mammalian lineages analyzed,
but very little change in the predicted structures of diverse
variants of this protein (supplementary fig. S4, Supplementary
Material online). This suggests that the CLR fold is highly ro-
bust to sequence variation and underscores the need for fu-
ture studies to parse the functional implications of the
sequence variation we observe.
Our results raise intriguing questions about the interac-
tions that drive rapid evolution in glycan-recognition
Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023 11
Hilbert et al. GBE
receptors and what the tradeoffs may be for interactions
with other microbes. Many of these PRRs are nonspecific,
involved in the recognition of many diverse glycan struc-
tures found in multiple microbial species. This suggests
that diversification of the carbohydrate recognition do-
mains of these PRRs could have a profound impact on the
recognition of numerous microbial species. Although this
may make it challenging to identify the exact molecular
changes or microbial species that have driven rapid evolu-
tion in these glycan PRRs, this system represents a unique
opportunity to study the tradeoffs associated with rapid
evolution, a topic that has been largely ignored in pro-
tein–protein arms races, where the focus has remained on
1:1 interactions between host proteins and highly specific
pathogenic substrates. Recent advances in high-throughput
profiling of host lectin interactions with complex microbial
glycans when applied to these rapidly evolving PRRs will like-
ly help to shed light on these questions of what drove these
signals of evolution and what the consequences might be
for specific microbial recognition (Stowell et al. 2014;
Jégouzo et al. 2020).
Finally, our phylogenetic screen identified extensive posi-
tive selection among rodent and, in particular, bat glycan
PRRs, where a striking 81% of the genes we analyzed
were found to be rapidly evolving. This suggests that for
these carbohydrate-recognition receptors, evolution has
been driven by lineage-specific microbial communities, per-
haps including both pathogenic and commensal species.
Combined, our data reveal a new axis of evolutionary
arms races—involving microbial glycan detection—and
dramatically expand the realm of host–microbe interactions
to include fungal pathogens with consequential influence
on the evolution of eukaryotic biology.
Materials and Methods
Phylogenetic Analyses
Candidate gene ortholog sequences were obtained from
NCBI GenBank either through gene name searches or by
BLAST searches using the Human ortholog sequence as query
(see supplementary data file 1, Supplementary Material on-
line for full list of accession numbers). Additional BLAST
searches were carried out using alternate species as query
to confirm that the same subsets of genes were being iden-
tified through different searches. Orthologous relationships
between genes were further confirmed by phylogenetic
and synteny analysis and species were excluded from evolu-
tionary analysis if clear orthology could not be established.
Phylogenetic tree analysis of some of the more divergent
genes, like CLEC12A, confirmed that orthologs of CLEC12A
from all three mammalian groups cluster together on a single
branch, removed from the other CLR genes (supplementary
fig. S1A, Supplementary Material online).
Sequences were obtained for all available simian primate
species, Myomorpha species (minus Jaculus jaculus, for
which we could not consistently find well-annotated ortho-
logs), and the Chiroptera. Coding sequences were down-
loaded and aligned using the Geneious Translation Align
function with the MUSCLE algorithm option. Alignments
were manually inspected and trimmed to remove gaps,
ambiguous regions of the alignment and stop codons.
Alignments were used to construct gene trees using
IQ-TREE and the GTR + G + I model with 100 nonparametric
bootstraps (Nguyen et al. 2015). Both gene trees and gen-
erally accepted species phylogenies for each of the mam-
malian groups were used for downstream evolutionary
analyses. Alignments and trees used in analysis can be
found in supplementary data file 2, Supplementary Material
online. Data shown in figure 1 are based on analyses done
with species trees, but all of the results of the analyses can
be found in supplementary data file 1, Supplementary
Material online. Unless otherwise noted, all computational
analysis was performed using the University of Utah Center
for High Performance Computing.
Positive selection was assessed using the codeml func-
tion of the PAML software package (v4.9) with the F3 × 4
codon frequency model (Yang 2007). Gene-wide dN/dS va-
lues were calculated using model 0. To test whether a sub-
set of amino acid sites were evolving under positive
selection, we performed LRTs, comparing pairs of NSsites
models including: M1 (neutral evolution) versus M2 (posi-
tive selection) and M7 (neutral, beta distribution dN/dS ≤
1) versus M8 (positive selection, beta distribution allowing
for dN/dS > 1). For genes with statistical support for positive
selection, specific amino acid positions were identified as
being under positive selection based on having a Bayes
Empirical Bayes (BEB) posterior probability of greater than
90% in the M8 model. For the free ratios analysis of
CLEC12A, codeml Model 1, allowing variation of dN/dS
across branches of the phylogeny, was run on the
CLEC12A alignments with an unrooted species tree for
each lineage.
The BUSTED, MEME, and FUBAR programs from the
HyPhy suite (version 2.5.41) were run through the com-
mand line with the same input alignments and trees used
for PAML analyses and default options (Pond et al. 2005;
Murrell et al. 2012, 2013, 2015). Results were visualized
using the HyPhy Vision web server. For several of the
BUSTED analyses, we noticed that the algorithm found
statistically significant support for positive selection in align-
ments that had very high levels of conservation determined
by other methods (e.g., Primate FIBCD1 and Dectin1).
When we examined these results, we found that the signal
was being driven entirely by codons containing multiple nu-
cleotide substitutions, which has been a documented con-
founding variable in branch-site models of rapid evolution
(Venkat et al. 2018). For these anomalous results, we re-ran
12 Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023
Rapid Evolution of Glycan Recognition Receptors GBE
the analyses without these multiply substituted sites and
found that these genes were no longer predicted to be un-
der positive selection by BUSTED (see “BUSTED P-value with
MNMs removed” column in supplementary data file 1,
Supplementary Material online). These re-runs are reflected
in the results displayed in figure 1.
Codon alignments of the Human CTLDs from each of the
CLRs in the gene set were used as input to IQ-TREE for
phylogenetic tree construction (fig. 1B) (Nguyen et al.
2015). The VT + G4 substitution model was selected as
the best fit model by the ModelFinder function, and 100
nonparametric bootstrap
replicates were performed
(Kalyaanamoorthy et al. 2017). Some IQ-TREE analyses
were performed with the IQ-Tree webserver (Trifinopoulos
et al. 2016). CTLDs were identified based on annotated do-
mains from UniProt and genes with multiple CTLDs (e.g.,
MRC1 and MRC2) were excluded. An alternate version of
this tree built from an alignment of nine representative spe-
cies spanning all three mammalian groups assessed is shown
in supplementary figure S1B, Supplementary Material on-
line. Species included were: Homo sapiens, Mucaca mulat-
ta, S. boliviensis, Mus musculus, Microtus ochrogaster,
Nannospalax galili, Myotis myotis, Pteropus alecto, and
Rhinolophus sinicus. Tree topology varied only slightly
across species and in this pan-species tree.
MelLec Human Population Genetics Analyses
To map the geographic distribution of the G26A poly-
morphism (rs2306894) in Human MelLec (CLEC1A), sam-
pling
locations of 1KG on GRCh38 and HGDP
populations were downloaded from the International
Genome Sample Resource (Zheng-Bradley et al. 2017;
Lowy-Gallego et al. 2019; Bergström et al. 2020).
Chromosome 12 VCF files for HGDP and 1KG datasets
were downloaded from their respective FTP sites (see
Data Availability statement below). VCFtools was used to
obtain the allele frequency at G26A for all populations,
and the map was created using the R library ggmap
(Danecek et al. 2011).
Tajima’s D was calculated using VCFtools and β(1) statis-
tics using BetaScan2 (Siewert and Voight 2020). The de-
rived allele was obtained from ancestral FASTA files
downloaded from Ensembl (see Data Availability statement
below). Empirical P-values were calculated in R by compari-
son with all other test statistic values on chr12 and plots
were generated with ggplot2 (R Core Team 2022).
Cowplot was used to combine the map, Tajima’s D, and
β(1) plots.
r2 was calculated between rs2306894 and SNPs within
100 kb in either direction to identify pairs in high linkage
disequilibrium using VCFtools and plink2 (Chang et al.
2015). We also generated a population-specific chromo-
some 12 VCF, using VCFtools, from the 1KG Colombian
population to test for positive selection using Relate
v1.1.8 and the add-on module for selection, which infers
how quickly a mutation spread through the population
based on genome-wide genealogies (Speidel et al. 2019,
2021).
Squirrel Monkey gDNA MelLec Genotyping
Squirrel monkey gDNA was originally isolated from blood
samples kindly provided by the MD Anderson Squirrel
Monkey Resource and Breeding Center in September 2015.
The provided samples came from unrelated individuals and
additional information including Sample IDs, sex and age of
the animals can be found in supplementary figure S3,
Supplementary Material online. One additional gDNA sample
from S. sciureus sciureus was isolated from the AG05311
fibroblast cell line provided by the Coriell Institute. All gDNA
samples have been stored at −20 °C.
Primers MS_B17 and MS_B20 were designed to amp-
lify a ∼500 bp fragment including the entirety of Exon 1
of MelLec (CLEC1A) which contains the polymorphic
site (amino acid 26), along with flanking sequence. The
black-capped squirrel monkey genome saiBol1was used
as a reference for primer design. Polymerase chain reac-
tions were performed using Phusion Flash polymerase
and 50 ng of each gDNA sample from the squirrel mon-
key individuals. PCR products were confirmed on a gel,
purified with Exo-SAP and Sanger sequenced at the
University of Utah Sequencing Core using primer
MS_B19. Genotypes were called based on visualization
of Sanger sequencing traces in Geneious. Primer se-
quences are as follows:
MS_B17 TCCATGAGAGGTGCAAACAG
MS_B20 AGTTGTGGAAAGCGCACAG
MS_B19 ACATGCTGTTTCCCTTCAGC
Structural Modeling and Comparisons of CLEC12A
CTLDs
The structures of the CTLDs of nine mammalian CLEC12A
orthologs were modeled using AlphaFold
(v 2.1.2)
(Jumper et al. 2021). The predicted structure with the high-
est confidence (ranked_0.pdb) for each ortholog was com-
pared with all other species using jFATCAT through the
RCSB PDB Pairwise Structure Alignment tool (Prlić et al.
2010; Burley et al. 2018; Li et al. 2020). Alignments were
performed using both the rigid and flexible alignment algo-
rithms and results were identical between the two. RMSD
values were plotted as a heatmap in R (supplementary fig.
S4, Supplementary Material online). All ranked_0 predicted
structures and CTLD sequences used for modeling can be
found at: dx.doi.org/10.6084/m9.figshare.23535738.
Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023 13
Hilbert et al. GBE
Supplementary Material
• eQTL analysis available from GTEx: https://gtexportal.org/
Supplementary data are available at Genome Biology and
Evolution online (http://www.gbe.oxfordjournals.org/).
Acknowledgments
We thank members of the Elde lab for helpful discussions in
the development of this project. We thank Stephen
Goldstein for suggestions on tree-building and primate
population genetics and Ian Boys for help with AlphaFold
modeling. This work was supported by the National
Institutes of Health (grant number R35 GM147709 to
E.M.L, grant number R35 GM134936 to N.C.E., and grant
number T32GM141848 to H.J.Y.); a Burroughs Wellcome
Fund Investigator in the Pathogenesis of Infectious
Disease Award to N.C.E.; and a postdoctoral fellowship
from the Helen Hay Whitney Foundation to Z.A.H.
Author Contributions
Z.A.H. and N.C.E. designed the study and wrote the manu-
script. Z.A.H. performed evolutionary analyses, structural
modeling, and interpreted results. P.E.H. and E.M.L per-
formed population genetics analyses on MelLec and inter-
preted results. H.J.Y. performed BLAST searches and
sequence alignments for phylogenetic analyses. M.J.W.S.
performed squirrel monkey sample PCRs, sequencing,
and data analysis. All authors reviewed and edited the
manuscript.
Data Availability
NCBI accession numbers for all genes analyzed are provided
in supplementary data file 1, Supplementary Material online.
Alignments and trees used in positive selection analyses are
provided in supplementary data file 2, Supplementary
Material online. Genotypes for great ape species at the
position of the rs2306894 human polymorphism were ob-
tained from: https://www.biologiaevolutiva.org/greatape/
data.html. For analyses of MelLec in human populations,
the following links were used to download or access the rele-
vant datasets:
• Sampling locations: https://www.internationalgenome.
org/data-portal/population.
• HGDP Chr12: https://ngs.sanger.ac.uk/production/hgdp/
hgdp_wgs.20190516/.
• 1KG Chr12: http://ftp.1000genomes.ebi.ac.uk/vol1/ftp/
data_collections/1000G_2504_high_coverage/working/
20201028_3202_phased/.
• Ancestral FASTA files for GRCh38 (homo sapiens ances-
tor GRCh38.tar.gz downloaded March 2023): https://ftp.
ensembl.org/pub/current_fasta/ancestral_alleles/.
home/snp/rs2306894.
FST, Tajima’s D, β(1) statistics, and statistics from linkage
disequilibrium analysis are provided in supplementary data
file 1, Supplementary Material online. AlphaFold-modeled
CLEC12A CTLD structures can be found on figshare at:
dx.doi.org/10.6084/m9.figshare.23535738.
Literature Cited
The 1000 Genomes Project Consortium. 2015. A global reference for
human genetic variation. Nature 526:68–74.
Areal H, Abrantes J, Esteves PJ. 2011. Signatures of positive selection in
Toll-like receptor (TLR) genes in mammals. BMC Evol Biol. 11:368.
Barber MF, Elde NC. 2014. Escape from bacterial iron piracy through
rapid evolution of transferrin. Science 346:1362–1366.
Begun J, et al. 2015. Integrated genomics of Crohn’s disease risk vari-
ant identifies a role for CLEC12A in antibacterial autophagy. Cell
Rep. 11:1905–1918.
Bergström A, et al. 2020. Insights into human genetic variation and
population history from 929 diverse genomes. Science 367:
eaay5012.
Boguslawski KM, et al. 2020. Exploiting species specificity to under-
stand the tropism of a human-specific toxin. Sci Adv. 6:eaax7515.
Brown GD, et al. 2012. Hidden killers: human fungal infections. Sci
Transl Med. 4(165):165rv13.
Brown GD, Gordon S. 2001. A new receptor for β-glucans. Nature
413:36–37.
Burley SK, et al. 2018. RCSB Protein Data Bank: biological macromol-
ecular structures enabling research and education in fundamental
biology, biomedicine, biotechnology and energy. Nucleic Acids
Res. 47:D464–D474.
Carey CM, Apple SE, Hilbert ZA, Kay MS, Elde NC. 2021. Diarrheal
pathogens trigger rapid evolution of the guanylate cyclase-C sig-
naling axis in bats. Cell Host Microbe. 29:1342–1350.e5.
Carlin AF, et al. 2009. Molecular mimicry of host sialylated glycans al-
lows a bacterial pathogen to engage neutrophil siglec-9 and
dampen the innate immune response. Blood 113:3333–3336.
Chang CC, et al. 2015. Second-generation PLINK: rising to the chal-
lenge of larger and richer datasets. Gigascience 4:1–16.
Chiaro TR, et al. 2023. Clec12a tempers inflammation while restricting
expansion of a colitogenic commensal. Biorxiv. 2023.03.16.532997.
doi:10.1101/2023.03.16.532997.
Comstock LE, Kasper DL. 2006. Bacterial glycans: key mediators of di-
verse host immune responses. Cell 126:847–850.
Cummings RD, Chiffoleau E, van Kyook Y, McEver RP. 2022. Chapter
34: C-type lectins. In: Varki A et al., editors. Essentials of glycobiol-
ogy [Internet]. 4th ed. Cold Spring Harbor, NY: Cold Spring Harbor
Laboratory Press. doi:10.1101/glycobiology.4e.34.
Danecek P, et al. 2011. The variant call format and VCFtools.
Bioinformatics 27:2156–2158.
Daugherty MD, Malik HS. 2012. Rules of engagement: molecular in-
sights from host-virus arms races. Annu Rev Genet. 46:677–700.
de Jong MAWP, et al. 2010. C-type lectin langerin is a β-glucan recep-
tor on human langerhans cells that recognizes opportunistic and
pathogenic fungi. Mol Immunol. 47:1216–1225.
Denham ST, et al. 2018. Regulated release of cryptococcal polysac-
charide drives virulence and suppresses immune cell infiltration
into the central nervous system. Infect Immun. 86:e00662-17.
de Witte L, et al. 2007. Langerin is a natural barrier to HIV-1 transmis-
sion by Langerhans cells. Nat Med. 13:367–371.
14 Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023
Rapid Evolution of Glycan Recognition Receptors GBE
Elde NC, Child SJ, Geballe AP, Malik HS. 2009. Protein kinase R reveals
an evolutionary model for defeating viral mimicry. Nature 457:
485–489.
Feinberg H, et al. 2011. Structural basis for langerin recognition of di-
verse pathogen and mammalian glycans through a single binding
site. J Mol Biol. 405:1027–1039.
Fisher MC, et al. 2020. Threats posed by the fungal kingdom to hu-
mans, wildlife, and agriculture. mBio 11: e00449–20.
Fisher MC, et al. 2022. Tackling the emerging threat of antifungal re-
sistance to human health. Nat Rev Microbiol. 20:557–571.
GenomeAsia100K Consortium et al. 2019. The GenomeAsia 100K
Project enables genetic discoveries across Asia. Nature 576:
106–111.
George RD, et al. 2011. Trans genomic capture and sequencing of pri-
mate exomes reveals new targets of positive selection. Genome
Res. 21:1686–1694.
Gow NAR, Latge J-P, Munro CA. 2017. The fungal cell wall: structure,
biosynthesis, and function. Microbiol Spectr. 5:FUNK-0035-2016.
Guth S, et al. 2022. Bats host the most virulent—but not the most dan-
gerous—zoonotic viruses. Proc Natl Acad Sci. 119:e2113628119.
Han BA, Schmidt JP, Bowden SE, Drake JM. 2015. Rodent reservoirs of
future zoonotic diseases. Proc Natl Acad Sci. 112:7039–7044.
Hanske J, et al. 2017. Bacterial polysaccharide specificity of the pattern
recognition receptor langerin is highly species-dependent. J Biol
Chem. 292:862–871.
Herre J, Gordon S, Brown GD. 2004. Dectin-1 and its role in the recog-
nition of β-glucans by macrophages. Mol Immunol. 40:869–876.
Imperiali B. 2019. Bacterial carbohydrate diversity—a brave new
world. Curr Opin Chem Biol. 53:1–8.
Jégouzo SAF, et al. 2020. Mammalian lectin arrays for screening host–
microbe interactions. J Biol Chem. 295:4541–4555.
Judd EN, Gilchrist AR, Meyerson NR, Sawyer SL. 2021. Positive natural
selection in primate genes of the type I interferon response. BMC
Ecol Evol. 21:65.
Jumper J, et al. 2021. Highly accurate protein structure prediction with
AlphaFold. Nature 596:583–589.
Kalyaanamoorthy S, Minh BQ, Wong TKF, von Haeseler A, Jermiin LS.
2017. Modelfinder: fast model selection for accurate phylogenetic
estimates. Nat Methods. 14:587–589.
Kawai T, Akira S. 2008. Toll-like receptor and RIG-1-like receptor sig-
naling. Ann N Y Acad Sci. 1143:1–20.
Li Z, Jaroszewski L, Iyer M, Sedova M, Godzik A. 2020. FATCAT 2.0: to-
wards a better understanding of the structural diversity of proteins.
Nucleic Acids Res. 48:W60–W64.
Liu X, et al. 2017. Characterising private and shared signatures of positive
selection in 37 Asian populations. Eur J Hum Genet. 25:499–508.
Liu G, Huanxin Z, Zhao C, Honghai Z. 2019. Evolutionary history of the
Toll-like receptor gene family across vertebrates. Genome Biol Evol.
12:3615–3634.
Lonsdale J, et al. 2013. The genotype-tissue expression (GTEx) project.
Nat Genet. 45:580–585.
Lowy-Gallego E, et al. 2019. Variant calling on the GRCh38 assembly
with the data from phase three of the 1000 genomes project.
Wellcome Open Res. 4:50.
Masuoka J. 2004. Surface glycans of Candida albicans and other
pathogenic fungi: physiological roles, clinical uses, and experimen-
tal challenges. Clin Microbiol Rev. 17:281–310.
Medzhitov R. 2007. Recognition of microorganisms and activation of
the immune response. Nature 449:819–826.
Mitchell PS, et al. 2012. Evolution-guided identification of antiviral spe-
cificity determinants in the broadly acting interferon-induced in-
nate immunity factor MxA. Cell Host Microbe. 12:598–604.
Murrell B, et al. 2012. Detecting individual sites subject to episodic di-
versifying selection. PLoS Genet. 8:e1002764.
Murrell B, et al. 2013. FUBAR: a fast, unconstrained Bayesian approxi-
mation for inferring selection. Mol Biol Evol. 30:1196–1205.
Murrell B, et al. 2015. Gene-wide identification of episodic selection.
Mol Biol Evol. 32:1365–1371.
Neumann K, et al. 2014. Clec12a is an inhibitory receptor for uric acid
crystals that regulates inflammation in response to cell death.
Immunity 40:389–399.
Nguyen L-T, Schmidt HA, von Haeseler A, Minh BQ. 2015. IQ-TREE: a
fast and effective stochastic algorithm for estimating maximum-
likelihood phylogenies. Mol Biol Evol. 32:268–274.
Nyame AK, Kawar ZS, Cummings RD. 2004. Antigenic glycans in para-
sitic infections: implications for vaccines and diagnostics. Arch
Biochem Biophys. 426:182–200.
Park BS, et al. 2009. The structural basis of lipopolysaccharide recogni-
tion by the TLR4–MD-2 complex. Nature 458:1191–1195.
Poltorak A, et al. 1998. Defective LPS signaling in C3H/HeJ and
C57BL/10ScCr mice: mutations in Tlr4 gene. Science 282:
2085–2088.
Pond SLK, Frost SDW, Muse SV. 2005. Hyphy: hypothesis testing using
phylogenies. Bioinformatics 21:676–679.
Prado-Martinez J, et al. 2013. Great ape genetic diversity and popula-
tion history. Nature 499:471–475.
Prlić A, et al. 2010. Pre-calculated protein structure alignments at the
RCSB PDB website. Bioinformatics 26:2983–2985.
Raman R, Tharakaraman K, Sasisekharan V, Sasisekharan R. 2016.
Glycan–protein interactions in viral pathogenesis. Curr Opin
Struct Biol. 40:153–162.
Raulf M-K, et al. 2019. The C-type lectin receptor CLEC12A recognizes
plasmodial hemozoin and contributes to cerebral malaria develop-
ment. Cell Rep. 28:30–38.e5.
Rausell A, Telenti A. 2014. Genomics of host–pathogen interactions.
Curr Opin Immunol. 30:32–38.
R Core Team. 2022. R: a language and environment for statistical com-
puting. Vienna, Austria: R Foundation for Statistical Computing.
https://www.R-project.org/.
Rodrigues ML, et al. 1997. Identification of N-acetylneuraminic acid
and
its 9-O-acetylated derivative on the cell surface of
Cryptococcus neoformans: influence on fungal phagocytosis.
Infect Immun. 65:4937–4942.
Sawyer SL, Wu LI, Emerman M, Malik HS. 2005. Positive selection of
primate TRIM5α identifies a critical species-specific retroviral re-
striction domain. Proc Natl Acad Sci. 102:2832–2837.
Seyedmousavi S, et al. 2014. Black yeasts and their filamentous rela-
tives: principles of pathogenesis and host defense. Clin Microbiol
Rev. 27:527–542.
Siewert KM, Voight BF. 2017. Detecting long-term balancing selection
using allele frequency correlation. Mol Biol Evol. 34:2996–3005.
Siewert KM, Voight BF. 2020. Betascan2: standardized statistics to de-
tect balancing selection utilizing substitution data. Genome Biol
Evol. 12:3873–3877.
Soares RMA, et al. 2000. Identification of sialic acids on the cell
surface of Candida albicans. Biochim Biophys Acta. 1474:
262–268.
Sparber F, et al. 2018. Langerin+ DCs regulate innate IL-17 production
in the oral mucosa during Candida albicans-mediated infection.
PLoS Pathog. 14:e1007069.
Speidel L, et al. 2021. Inferring population histories for ancient gen-
omes using genome-wide genealogies. Mol Biol Evol. 38:3497.
Speidel L, Forest M, Shi S, Myers SR. 2019. A method for genome-wide
genealogy estimation for thousands of samples. Nat Genet. 51:
1321–1329.
Stappers MHT, et al. 2018. Recognition of DHN-melanin by a C-type
lectin receptor is required for immunity to Aspergillus. Nature
555:382–386.
Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023 15
Hilbert et al. GBE
Stowell SR, et al. 2014. Microbial glycan microarrays define key fea-
tures of host-microbial interactions. Nat Chem Biol. 10:470–476.
Tan X, Sun L, Chen J, Chen ZJ. 2015. Detection of microbial infections
through innate immune sensing of nucleic acids. Annu Rev
Microbiol. 72:447–478.
Tateno H, et al. 2010. Dual specificity of Langerin to sulfated and man-
nosylated glycans via a single C-type carbohydrate recognition do-
main*. J Biol Chem. 285:6390–6400.
Tenthorey JL, Young C, Sodeinde A, Emerman M, Malik HS. 2020.
Mutational resilience of antiviral restriction favors primate
TRIM5α in host-virus evolutionary arms races. Elife 9:e59988.
Tone K, et al. 2021. Mellec exacerbates the pathogenesis of
Aspergillus fumigatus-induced allergic inflammation in mice.
Front Immunol. 12:675702.
Trifinopoulos J, Nguyen L-T, von Haeseler A, Minh BQ. 2016.
W-IQ-TREE: a fast online phylogenetic tool for maximum likelihood
analysis. Nucleic Acids Res. 44:W232–W235.
Vance RE, Isberg RR, Portnoy DA. 2009. Patterns of pathogenesis:
discrimination of pathogenic and nonpathogenic microbes by
the innate immune system. Cell Host Microbe. 6:10–21.
Van Dalen R, et al. 2019. Langerhans cells sense Staphylococcus aur-
eus wall teichoic acid through langerin to induce inflammatory re-
sponses. mBio 10:e00330-19.
van der Vlist M, et al. 2011. Human Langerhans cells capture measles virus
through Langerin and present viral antigens to CD4+ T cells but are
incapable of cross-presentation. Eur J Immunol. 41:2619–2631.
Van Kooyk Y, Rabinovich GA. 2008. Protein-glycan interactions in the
control of innate and adaptive immune responses. Nat Immunol. 9:
593–601.
Varki A. 2011. Letter to the glyco-forum: since there are PAMPs and
DAMPs, there must be SAMPs? Glycan “self-associated molecular
patterns” dampen innate immunity, but pathogens can mimic
them. Glycobiology 21:1121–1124.
Varki A, Gagneux P. 2012. Multifarious roles of sialic acids in immunity.
Ann N Y Acad Sci. 1253:16–36.
Venkat A, Hahn MW, Thornton JW. 2018. Multinucleotide mutations
cause false inferences of lineage-specific positive selection. Nat
Ecol Evol. 2:1280–1288.
Vigerust DJ, Shepherd VL. 2007. Virus glycosylation: role in virulence
and immune interactions. Trends Microbiol. 15:211–218.
Voight BF, Kudaravalli S, Wen X, Pritchard JK. 2006. A map of recent
positive selection in the human genome. PLoS Biol. 4:e72.
Wang W, Han G-Z. 2021. Pervasive positive selection on virus receptors
driven by host-virus conflicts in mammals. J Virol. 95:e01029-21.
Werling D, Jann OC, Offord V, Glass EJ, Coffey TJ. 2009. Variation mat-
ters: TLR structure and species-specific pathogen recognition.
Trends Immunol. 30:124–130.
Wlasiuk G, Nachman MW. 2010. Adaptation and constraint at Toll-like
receptors in primates. Mol Biol Evol. 27:2172–2186.
Yang Z. 2007. PAML 4: phylogenetic analysis by maximum likelihood.
Mol Biol Evol. 24:1586–1591.
Yang K, et al. 2015. Host Langerin (CD207) is a receptor for Yersinia
pestis phagocytosis and promotes dissemination. Immunol Cell
Biol. 93:815–824.
Zheng-Bradley X, et al. 2017. Alignment of 1000 Genomes Project
reads to reference assembly GRCh38. Gigascience 6:1–8.
Associate editor: Dr. George Zhang
16 Genome Biol. Evol. 15(7) https://doi.org/10.1093/gbe/evad119 Advance Access publication 30 June 2023
| null |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.